Second International Symposium
FIELD SCREENING METHODS FOR
HAZARDOUS WASTES AND
TOXIC CHEMICALS
February 12-14, 1991
Symposium Proceedings
-------
SECOND INTERNATIONAL SYMPOSIUM
FIELD SCREENING METHODS FOR
HAZARDOUS WASTES AND
TOXIC CHEMICALS
Februaiy 12-14,1991
CO-SPONSORS
U.S. Environmental Protection Agency
U.S. Department of Energy
U.S. Army Toxic and Hazardous Materials Agency
U.S. Army Chemical Research, Development and Engineering Center
U.S. Air Force
Florida State University
National Environmental Technology Applications Corporation
National Institute for Occupational Safety and Health
-------
DISCLAIMER
Although this Proceedings Document reports the oral and poster presentations and
discussions that occurred during this Symposium funded by the United States
Environmental Protection Agency, the contents represent views independent of
Agency Policy. This Document has not been subjected to the Agency's peer review
process and does not necessarily reflect the Agency views. No official endorse-
ment should be inferred.
-------
SYMPOSIUM ORGANIZATION
Symposium Chairman - Llewellyn Williams, EPA/EMSL-Las Vegas, NV
Vice-Chairman - Eric Koglin, EPA/EMSL-Las Vegas, NV
Executive Secretary - John Koutsandreas, Florida State University
ACKNOWLEDGEMENTS
This symposium has been arranged through a contract with
ICAIR, Life Systems, Inc. The following personnel were involved in
coordinating this symposium:
Program Manager - Ms. Jo Ann Duchene
Presentation Coordinators - Mr. Ron Polhill
Ms. Donna Studniarz
Exhibit Coordinator - Mr. Charles Tanner
Registration Coordinator - Ms. Linda Hashlamoun
-------
FOREWORD
The role of and need for field screening methods for the identification and quantification of contaminants in
environmental media is growing rapidly. This nation and its European neighbors are faced with the tremendous task
of remediating thousands of hazardous waste sites -- the legacy of our much less environmentally aware
predecessors. Field screening methods that generate real-time information on the nature and extent of contamina-
tion improve the cost-effectiveness of remediation. Many of these same methods can, and in some cases are already
being used to improve our capability to measure exposure, at the point of exposure, thereby improving our ability
to assess risks to human health and the environment.
The U.S. EPA is not the only viable user of field screening methods; that fact is reflected in the list of this
Symposium's co-sponsors. Other agencies are discovering applications for these same technologies to address
issues such as worker safety, drug interdiction, and chemical warfare defense. The research activities supported by
these same agencies are advancing innovative technologies that may have application in environmental monitoring
and field screening.
To present a global view of technological developments, this Symposium featured over 120 platform and poster
presentations from the United States and around the world. The papers and discussions that follow represent three
days of intense communication and cooperation among a variety of communities—regulatory, academic, industrial
and users. It is my hope that the products of this Symposium will find many uses and will provide the impetus for
new initiatives in field screening methods.
Llewellyn R. Williams
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada
-------
CONTENTS
OPENING PLENARY SESSION
Opening Remarks — Dr. Llewellyn Williams, U.S. EPA, Environmental Monitoring Systems Laboratory, Las Vegas 1
Keynote Address — Analytical Issues in the U.S. EPA Superfund Program
Larry Reed, U.S. EPA, Director Hazardous Site Evaluation Division, Office of Emergency and Remedial Response ....3
Overview ofDOE's Field Screening Technology Development Activities
C.W. Frank, T.D. Anderson, C.R. Cooley, K.E. Hain, S.C.T. Lien, U.S. Department of Energy; R.L. Snipes,
Martin Marietta Energy Systems; M.D. Erickson, Argonne National Laboratory 5
Department of Defense Field Screening Methods Requirements in the Installation Restoration Program
Dennis J. Wynne, U.S. Army Toxic and Hazardous Materials Agency 15
An Overview of Army Sensor Technology Applicable to Field Screening of Environmental Pollutants
Raymond A. Mackay, U.S. Army Chemical Research, Development and Engineering Center 17
Field Analytical Methods for Superfund
Howard M. Fribush and Joan F. Fisk, U.S. EPA 25
Field Delineation of Soils Contamination on Hazardous Waste Sites Regulated Under New Jersey's Hazardous Waste Program
Frederick W. Cornell, New Jersey Department of Environmental Protection 31
Plenary Session Discussion 40
SESSION 1:
Chemical Sensors
Chairperson: Dr. Ed Poziomek, University of Nevada Environmental Research Center
A FiberOptic Sensor for the Continuous Monitoring of Chlorinated Hydrocarbons
P.P. Milanovich, P.P. Daley, K. Langry, B.W. Colston, S.B. Brown and S.M. Angel, Lawrence
Livermore National Laboratory 43
Chemical Sensors for Hazardous Waste Monitoring
M.B. Tabacco, Q. Zhou, K. Rosenblum, Geo-Centers, Inc.; M.R. Shahriari, Rutgers University 49
Rapid, Subsurface, In Situ Field Screening of Petroleum Hydrocarbon Contamination Using Laser Induced
Fluorescence Over Optical Fibers
S.H. Lieberman, G.A. Theriault, Naval Ocean Systems Center; S.S. Cooper, P.O. Malone and R.S. Olsen, U.S. Army
Waterways Experiment Station, Vicksburg; P.W. Lurk, U.S. Army Toxic and Hazardous Materials Agency 57
Chemical Sensors Panel Discussion 64
Spectroelectrochemical Sensing of Chlorinated Hydrocarbons for Field Screening and In Situ Monitoring Applications
Michael M. Carrabba, Robert B. Edmonds and R. David Rauh, EIC Laboratories, Inc.; John W. Haas, III,
Oak Ridge National Laboratories 67
Surface Acoustic Wave (SAW) Personal Monitor for Toxic Gases
N.L. Jarvis, H. Wohltjen and J.R. Lint, Microsensor Systems, Inc 73
Arrays of Sensors and Microsensors for Field Screening of Unknown Chemical Wastes
W.R. Penrose, J.R. Stetter and W.J. Buttner, Transducer Research, Inc.; Z. Cao, Illinois Institute of Technology 85
SESSION 2:
Ion Mobility Spectrometry
Chairperson: Dr. Steve Harden, U.S. Army Chemical Research, Development and Engineering Center
Real-Time Detection of Aniline in Hexane By Flow Injection Ion Mobility Spectrometry
G.E. Burroughs, National Institute for Occupational Safety and Health; G.A. Eiceman and L. Garcia-Gonzalez,
New Mexico State University 95
Detection of Microorganisms by Ion Mobility Spectrometry
A.P. Snyder, M. Miller and D.B. Shoff, U.S. Army Chemical Research, Development and Engineering Center;
Gary A. Eiceman, New Mexico State University; D. A. Blyth, J. A Parsons, Geo-Centers, Inc 103
Data Analysis Techniques for Ion Mobility Spectrometry
Dennis M. Davis, U.S. Army Chemical Research, Development and Engineering Center 113
-------
Ion Mobility Spectrometry as a Field Screening Technique
Lynn D. Hoffland and Donald B. Shoff, U.S. Army Chemical Research, Development and Engineering Center 137
Hand-Held GC-Ion Mobility Spectrometry for On-Site Analysis of Complex Organic Mixtures in Air or Vapors Over Waste Sites
Suzanne Ehart Bell, Los Alamos National Laboratory; G.A. Eiceman, New Mexico State University 153
Remote and In Situ Sensing of Hazardous Materials by Infrared Laser Absorption, Ion Mobility Spectrometry and Fluorescence
Peter Richter, Technical University of Budapest 167
SESSION 3:
Robotics
Chairperson: Dr. Carolyn Esposito, U.S. EPA Risk Reduction Engineering Laboratory
The Department of Energy's Robotics Technology Development Program for Environmental Restoration and
Waste Management
A.C. Heywood, Science Applications International Corporation; S.A. Meacham, Oak Ridge National Laboratory;
P.J. Eicker, Sandia National Laboratories 173
Field Robots for Waste Characterization and Remediation
William L. Whittaker, David M. Pahnos; Field Robotics Center, Carnegie Mellon Institute 181
Space Technology for Application to Terrestrial Hazardous Materials Analysis and Acquisition
Brian Muirhead, Susan Eberlein, James Bradley and William Kaiser, NASA/Jet Propulsion Laboratory 187
Development of a Remote Tank Inspection (RTI) Robotic System
Chris Fromme, Barbara P. Knape, Bruce Thompson, RedZone Robotics, Inc 197
Automated Subsurface Mapping
Jim Osbom, Field Robotics Center, Carnegie Mellon Institute 205
SESSION 4:
QA and Study Design
Chairperson: Dr. Janine Jessup Arvizu
A Quality Assurance Sampling Plan for Emergency Response (QASPER)
John M. Mateo, Christine M. Andreas, Roy F. Weston; William Coakley, U.S. EPA 217
A Rationale for the Assessment of Errors in Soil Sampling
Jeffrey van Ee, U.S. EPA; Clare L. Gerlach, Lockheed Engineering & Sciences Company 227
A Review of Existing Soil Quality Assurance Materials
K. Zarrabi, A.J. Cross-Smiecinski and T. Starks, University of Nevada 235
SESSION 5:
Air Pathway Monitoring at Superfund Sites
Chairperson: Dr. William McClenny
Evaluation of Emission Sources and Hazardous Waste Sites Using Portable Chromatographs
R.E. Berkley, U.S. EPA 253
High Speed Gas Chromatographyfor Air Monitoring
S.P, Levine, H.Q. Ke and R.F. Mouradian, University of Michigan; R. Berkley, U.S. EPA; J. Marshall, HNU Systems.... 265
Screening Volatile Organics By Direct Sampling Ion Trap and Glow Discharge Mass Spectrometry
Marcus B. Wise, G.B. Hurst, C.V. Thompson, Michelle V. Buchanan and Michael R. Guerin, Oak Ridge National
Laboratory 273
Development and Testing of a Man-Portable Gas Chromatography/Mass Spectrometry System for Air Monitoring
Henk L.C. Meuzelaar, Dale T. Urban and Neil S. Arnold, University of Utah 289
On-Site Multimedia Analyzers: Advanced Sample Processing with On-Line Analysis
S. Liebman, Geo-Centers, Inc.; M.B. Wasserman, U.S. Army Chemical Research, Development and Engineering
Center, E.J. Levy and S. Lurcott, Computer Chemical Systems, Inc 299
Using a FID-Based Organic Vapor Analyzer in Conjunction with GC/MS Summa Canister Analyses to Assess the Impact of
Landfill Gassesfrom a Superfund Site on the Indoor Air Quality of an Adjacent Commercial Property
T.H. Pritchett, U.S. EPA; D. Mickunas and S. Schuetz, IT Corporation 307
-------
SESSION 6:
Field Mobile GC/MS Techniques
Chairperson: Dr. Stephen Billets, U.S. EPA Environmental Monitoring Systems Laboratory, Las Vegas
Field Analytical Support Project (FASP) Use to Provide Data for Characterization of Hazardous Waste Sites for Nomination
to the National Priorities List (NPL): Analysis ofPolycyclic Aromatic Hydrocarbons (PAHs) and Pentachlorophenol (PCP)
Lila AccraTransue, Andrew Hafferty and Tracy Yerian, Ecology and Environment 309
Thermal Desorption Gas Chromatograph-Mass Spectrometry Field Methods for the Detection of Organic Compounds
A. Robbat, Jr. T-Y Liu, B. Abraham and C-J- Liu, Tufts University 319
Rapid Determination of Semivolatile Pollutants by Thermal Extraction/Gas Chromatography/Mass Spectrometry
T. Junk, V. Shirley, C.B. Henry, T.R. Irvin and E.B. Overton, Louisiana State University; J.E. Zumberge,
C. Sutton and R.D. Worden, Ruska Laboratories, Inc 327
The Application of a Mobile Ion Trap Mass Spectrometer System to Environmental Screening and Monitoring
William H. McClennen, Neil, S. Arnold, Henk L.C. Meuzelaar, JoAnn A. Lighty, University of Utah;
Erich Ludwig, GSF Munchen, Institut fur Okologische Chemie 339
Field Measurement of Volatile Organic Compounds by Ion Trap Mass Spectrometry
M.E. Cisper, J.E. Alarid, P.H. Hemberger, E.P. Vanderveer, Los Alamos National Laboratory 351
Transportable GC/lon Trap Mass Spectrometry for Trace Field Analysis of Organic Compounds
Chris P. Leibman and David Dogruel, Eric P. Vanderveer, Los Alamos National Laboratory 367
SESSION 7
Portable Gas Chromatography
Chairperson: Dr. Thomas Spittler, U.S. EPA New England Regional Laboratory
The Use of Field Gas Chromatography to Protect Gmundwater Supplies
Thomas M. Spittler, U.S. EPA 377
Field Screening Procedures for Determining the Presence of Volatile Organic Compounds in Soil
Alan B. Crockett and Mark S. DeHaan, EG&G Idaho, Inc 383
Comparison of Field Headspace Vs. Field Soil Gas Analysis Vs. Standard Method Analysis of Volatile Petroleum
Hydrocarbons in Water and Soil
Randy D. Golding, Marty Favero, Glen Thompson, Tracer Research Corporation 395
Field Screening ofBTEX in Gasoline-Contaminated Groundwater and Soil Samples by a Manual, Static Headspace GC Method
James D. Stuart, Suya Wang and Gary A. Robbins, University of Connecticut; Clayton Wood, HNU Systems, Inc. ...407
Comparison of Aqueous Headspace Air Standard Vs. SUMMA Canister Air Standard for Volatile Organic
Compound Field Screening
H. Wang, Roy F. Weston, Inc.; W.S. Clifford, U.S. EPA 415
Quantitative Soil Gas Sampler Implant for Monitoring Dump Site Subsurface Hazardous Fluids
Kenneth T. Lang, Douglas T. Scarborough, U.S. Army Toxic and Hazardous Materials Agency; Mark Glover,
D.P. Lucero, IIT Research Institute 423
SESSION 8
Field Screening Methods for Worker Safety
Chairperson: Dr. Judd Posner, National Institute for Occupational Safety and Health
Tunable COi Laser-Based Photo-Optical Systems for Surveillance of Indoor Workplace Pollutants
Harley V. Piltingsrud, National Institute for Occupational Safety and Health 433
Immuno-Based Personal Exposure Monitors
Arbor Drinkwine, Stan Spurlin, Midwest Research Institute; Jeanette Van Emon, U.S. EPA;
Viorica Lopez-Avila, Mid-Pacific Environmental Laboratory, Inc 449
A Remote Sensing Infrared Air Monitoring System for Gases and Vapors
S.P. Levine, H.K. Xiao, University of Michigan; W. Herget, Nicolet Analytical; R. Spear, University of California;
T. Pritchett, U.S. EPA 461
Adriamycin Exposure Study Among Hospital Personnel
R.L. Stephenson, Thomson Consumer Electronics Inc.; C.H. Rice, J. Dimos, University of Cincinnati 465
-------
Real-Time Personal Monitoring in the Workplace Using Radio Telemetry
Ronald J. Kovein and Paul Hentz, National Institute for Occupational Safety and Health 473
Improvements in the Monitoring of PPM Level Organic Vapors with Field Portable Instruments
Gerald Moore, GMD Systems, Inc 483
SESSION 9
X-Ray Fluorescence
Chairperson: Dr. John Barich, U.S. EPA Region X
Rapid Assessment of Superfund Sites for Hazardous Materials with X-Ray Fluorescence Spectrometry
W.H. Cole III, R.E. Enwall, G.A. Raab, C.A. Kuharic, Lockheed Engineering and Sciences Co.;
W.H. Engelmann, L.A. Eccles, U.S. EPA 497
A High Resolution Portable XRF Hgh Spectrometer for Field Screening of Hazardous Wastes
J.B. Ashe, Ashe Analytics; P.P. Berry and G.R. Voots, TN Technologies, Inc.; M. Bemick, Roy F. Weston, Inc.;
G. Prince, U.S. EPA 507
Low Concentration Soil Contaminant Characterization Using EDXRF Analysis
A.R. Harding, Spectrace Instruments, Inc 517
Data Quality Assurance/Quality Control for Field X-Ray Fluorescence Spectrometry
Clark D. Carlson, John R. Alexander, The Bionetics Corporation 525
A Study of the Calibration of a Portable Energy Dispersive X-Ray Fluorescence Spectrometer
C.A. Ramsey, D.J. Smith and E.L. Bour, U.S. EPA 535
SESSION 10
Fourier Transform Infrared Spectrometry and Other Spectroscopy Methods
Chairperson: Dr. Donald Gurka, U.S. EPA Environmental Monitoring Systems Laboratory, Las Vegas
Use of Long-Path FTIR Spectrometry in Conjunction with Scintillometry to Measure Gas Fluxes
Douglas I. Moore, Clifford N. Dahrn, James R. Gosz, University of New Mexico; Reginald J. Hill, NOAA 541
Pattern Recognition Methods for FTIR Remote Sensing
Gary W. Small, University of Iowa; Robert T. Kroutil, U.S. Army Chemical Research,
Development and Engineering Center 549
Remote Vapor Sensing Using a Mobile FTIR Sensor
R.T. Kroutil, J.T. Ditillo, R.L. Gross, R.J. Combs, W.R. Loerop, U.S. Army Chemical Research,
Development and Engineering Center; G.W. Small, University of Iowa 559
Use of Wind Data to Compare Point-Sample Ambient Air VOC Concentrations with Those Obtained by Open-Path FT-IR
Ray E. Carter, Jr. and Dennis D. Lane, Glen A. Marotz, University of Kansas;
Mark J. Thomas, Jody L. Hudson, U.S. EPA 571
Remote Detection ofOrganics Using Fourier Transform Infrared Spectroscopy
Jack C. Demirgian and Sandra M. Spurgash, Argonne National Laboratory 583
Intrepretation ofPPM-Meter Data from Long-Path Optical Monitoring Systems as They Would be Used at
Superfund Hazardous Waste Sites
Thomas H. Pritchett, U.S. EPA; Timothy R. Minnich, Robert L. Scotto and Margaret R. Leo,
Blasland, Bouck & Lee 591
CLOSING PLENARY SESSION
Awards Ceremony 593
Closing Remarks 595
POSTERS
Calibration of Fiber Optic Chemical Sensors
W.F. Arendale and Richard Hatcher, University of Alabama; Bruce Nielsen, Hq. AFESC/RDVW 597
Gas-Chromatographic Analysis of Soil-Gas Samples at a Gasoline Spill
R.J. Baker, J.M. Ficher, N.P. Smith, S.A. Koehnlein, A.L. Baehr, U.S. Geological Survey 599
-------
Significant Physical Effects on Surface Acoustic Wave (SAW) Sensors
David L. Bartley, National Institute for Occupational Safety and Health 601
An Evaluation of Field Portable XRF Soil Preparation Methods
Mark Bemick, Donna Idler, Lawrence Kaelin, Dave Miller, Jayanti Patel, Roy F. Weston; George Prince and
Mark Sprenger, U.S. EPA 603
Development of a Field Screening Technique for Dimethyl Mercury in Air
Brian E. Brass, Lawrence P. Kaelin, Roy F. Weston; Thomas H. Pritchett, U.S. EPA 609
Applicability of Thin-Layer Chromatography to Field Screening of Nitrogen-Containing Aromatic Compounds
William C. Brumley, Cynthia M. Brownrigg, U.S. EPA 615
Assessing the Air Emissions from a Contaminated Aquifer at a Superfund Site
S. Burchette and T.H. Prichett, U.S. EPA; S. Schuetz, IT Corporation; K. Harvey, Roy F. Weston, Inc 619
Calculation and Use of Retention Indices for Identification of Volatile Organic Compounds with a Microchip
Gas Chromatograph
K.R. Carney, E.B. Overton and R.L. Wong, Louisiana State University 621
Determination ofPCBs by Enzyme Immunoassy
Mary Anne Chamerlik-Cooper, Robert E. Carlson, ECOCHEM Research, Inc; Robert O. Harrison,
ImmunoSystems, Inc 625
Practical Limits in Field Determination of Fluorescence Using Fiber Optic Sensors
Wayne Chudyk, Kenneth Pohlig, Carol Botteron and Rose Najjar, Tufts University 629
The Colloidal Borescope—A Means of Assessing Local Colloidal Flux and Groundwater Velocity in Porous Media
T.A. Cronk, P.M. Kearl, Oak Ridge National Laboratory 631
Fieldable Enzyme Immunoassay Kits for Drugs and Environmental Chemicals
Peter H. Duquette, Patrick E. Guire, Melvin J. Swanson, Martha, J. Hamilton, Stephen J. Chudzik and
Ralph A. Chappa, Bio-Metric Systems, Inc 633
Xuma Expert System for Support of Investigation and Evaluation of Contaminated Sites
W. Eitel, R. Hahn Landesanstalt f. Umweltschutz B.; W.W. Geiger and R. Weidemann,
Institut f. Datenverarbeitung 645
A Rapid Response SAW-GC Chemical Monitor for Low-Level Vapor Detection
John A. Elton, James F. Houle, Eastman Kodak Company 649
Passive Cryogenic Whole Air Field Sampler
Steven J. Fernandez, Bill G. Motes, Joseph P. Dugan Jr., Susan K. Bird, Gary J. McManus, Westinghouse Idaho
Nuclear Company 653
Effectiveness of Porous Glass Elements for Suction Lysimeters to Monitor Soil Water for Organic Contaminants
Stanley M. Finger, Hamid Hojaji, Morad Boroomand and Pedro B. Macedo, Catholic University of America 657
Comparison of Mobile Laboratory XRF and CLP Split Sample Lead Results from a Superfund Site Remediation in New Jersey
Jon C. Gabry, Ebasco Environmental 671
Screening of Groundwater for Aromatics by Synchronous Fluorescence
R.B. Gammage, J.W. Haas, III and T.M. Allen, Oak Ridge National Laboratory 673
In Situ Detection of Toxic Aromatic Compounds in Groundwater Using Fiberoptic UV Spectroscopy
J.W. Haas III, T.G. Matthews and R.B. Gammage, Oak Ridge National Laboratory 677
Development of Field Screening Methods for TNT and RDX in Soil and Ground Water
Thomas F. Jenkins and Marianne E. Walsh, U.S. Army Cold Regions Research and Engineering Laboratory;
Martin H. Stutzand Kenneth T. Lang, U.S. Army Toxic and Hazardous Materials Agency 683
Quantification of Pesticides on Soils by Thermal Extraction-GC/MS
T. Junk, T.R. Irvin, Louisiana State University; K.C. Donnelly and D. Marek, Texas A&M University 687
A Portable Gas Chromatograph with an Argon lonization Detector for the Field Analysis of Volatile Organics
Lawrence P. Kaelin, Roy F. Weston, Thomas H. Pritchett, U.S. EPA 689
Sea Mist—A Technique for Rapid and Effective Screening of Contaminated Waste Sites
Carl Keller and Bill Lowry, Science and Engineering Associates, Inc 693
xi
-------
Portable Gas Chromatograph Field Monitoring ofPCB Levels in Soil at the El:a Gate Property
Marty R. Keller and Gomes Ganapathi, Bechtel National, Inc 697
Real Time Monitoring of the Flue of a Chemical Demilitarization Incinerator
S.N. Ketkar and S.M. Penn, Extrel Corporation 701
Field Evaluation of the Bruker Mobile Mass Spectrometer Under the U.S. EPA SITE Program
S.M. Klainer, M.E. Silverstein, V.A. Ecker, D.J. Chaloud, Lockheed Engineering and Sciences Company and
S. Billets, U.S. EPA 705
The DITAM Assay A - Fast, Fieldable Method to Detect Hazardous Wastes, Toxic Chemicals, and Drugs
Cynthia Ladouceur, U.S. Army Chemical Research, Development and Engineering Center 709
Rapid Screening of Ground Water Contaminants Using Innovative Field Instrumentation
Amos Linenberg and David Robinson, Sentex Sensing Technology, Inc 711
Improved Detection of Volatile Organic Compounds in a Microchip Gas Chromatograph
Aaron M. Mainga and Edward B. Overton, Louisiana State University 713
On-Line Screening Analyzers for Trace Organics Utilizing a Membrane Extraction Interface
Richard G. Melcherand Paul L. Morabito, The Dow Chemical Company 717
Candidate Protocols for Sampling and Analysis of Chemicals from the Clean Air Act List
R.G. Merrill, J.T. Bursey, D.L. Jones, T.K. Moody, C.R. Blackley, Radian Corporation;
W.B. Kuykendal, U.S. EPA 721
The Investigation of Soil Sampling Devices and Shipping and Holding Time Effects on Soil Volatile Organic Compounds
J.R. Parolini, V.G. King, T.W. Nail and T.E. Lewis, Lockheed Engineering and Sciences Company 725
Developmental Logic for Robotic Sampling Operations
Michael D. Pavelek II, Micren Associates, Chris C. Fromme, RedZone Robotics, Inc 729
Practical Problems Encountered in Remote Sensing of Atmospheric Contaminants
Kirkman R. Phelps and Michael S. DeSha, U.S. Army Chemical Research, Development and Engineering Center....733
A SI/LI Based High Resolution Portable X-Ray Analyzer for Field Screening of Hazardous Waste
Stanislaw Piorek and James R. Pasmore, Outokumpu Electronics, Inc 737
Measurement and Analysis ofAdsistor and Figaro Gas Sensor Used for Underground Storage Tank Leak Detection
Marc A. Portnoff, Richard Grace, Alberto M. Guzman, Jeff Hibner, Carnegie Mellon University 741
Extraction Disks for Spectroscopic Field Screening Applications
Edward J. Poziomek, University of Nevada; DeLyle Eastwood, Russell L. Lidberg,
Gail Gibson, Lockheed Engineering and Sciences Co 747
Field Analytical Support Project (FASP) Development of High-Peiformance Liquid Chromatography (HPCL) Techniques
for On-Site Analysis ofPolycyclic Aromatic Hydrocarbons (PAHs) at PreRemedial Superfund Sites
Andrew Riddell, Andrew Hafferty and Tracy Yerian, Ecology and Environment, Inc 751
A Field Comparison of Monitoring Methods for Waste Anesthetic Gases and Ethylene Oxide
Stanley A. Salisbury, G.E. Burroughs, William J. Daniels, Charles McCammon and Steven A. Lee,
National Institute for Occupational Safety and Health 755
On-Site and On-Line Spectroscopic Monitoring of Toxic Metal Ions Using Fiber Optic Ultraviolet Absorption Spectrometry
Kenneth J. Schlager, Biotronics Technology, Inc.; Bernard J. Beemster, Beemster and Associates 759
Rapid Screening of Soil Samples for Chlorinated Organic Compounds
H. Schlesing, N. Darskus, C. Von Hoist and R. Wallon, Biocontrol Institute for Chemische Und Biologische
Untersuchungen Ingelheim 763
Development of a Microbore Capillaiy Column GC-Focal Plane Mass Spectrograph with an Array Detector
for Field Measurements
M.P. Sinha, California Institute of Technology 765
Application of a Retention Index Approach Using Internal Standards to a Linear Regression Model for Retention Time
Windows in Volatile Organic Analysis
Russell Sloboda, NUS Corporation 775
-------
Detection of Airborne Microorganisms Using a Hand-Held Ion Mobility Spectrometer
A. Peter Snyder, U.S. Army Chemical Research, Development & Engineering Center; David A. Blyth,
John A. Parsons, Geo-Centers, Inc; Gary A. Eiceman, New Mexico State University 783
Field Analysis for Hexavalent Chrome in Soil
Robert L. Stamnes, U.S. EPA; Greg D. DeYong, HACH Company; Clark D. Carlson, Bionetics Corp 785
Transportable Tunable Dye Laser for Field Analysis of Aromatic Hydrocarbons in Groundwater
Randy W. St. Germain and Gregory D. Gillispie, North Dakota State University 789
Real Time Detection of Biological Aerosols
Peter J. Stopa, Michael T. Goode, Alan W. Zulich, David W. Sickenberger, E. William Sarver and
Raymond A. Mackay, U.S. Army Chemical Research, Development and Engineering Center 793
Laser Fluorescence EEM Instrument for In-Situ Groundwater Screening
Todd A. Taylor, Hong Xu and Jonathan E. Kenny, Tufts University 797
Analysis of Total Poly aromatic Hydrocarbon Using Ultraviolet-Fluorescence Spectrometry
T.L. Theis, A.G. Collins, P.J. Monsour, S.G. Pavlostathis and C.D. Theis, Clarkson University 805
On-Site Analysis of Chlorinated Solvents in Groundwater by Purge and Trap GC
Stephen A. Turner, Daniel Twomey, Jr., Thomas L. Francoeur and Brian K. Butler, ABB
Environmental Services, Inc 811
U.S. EPA Evaluation of Two Pentachlorophenol Immunoassay Systems
J.M. Van Emon, U.S. EPA; R.W. Gerlach, R.J. White and M.E. Silverstein, Lockheed Engineering and
Sciences Company 815
Rapid Screening Technique for Polychlorinated Biphenyis (PCBs) Using Room Temperature Phosphorescence
T. Vo-Dinh, G.H. Miller, A. Pal, W. Watts and M. Uziel, Oak Ridge National Laboratory;
D. Eastwood and R. Lidberg, Lockheed Engineering & Management Services Co 819
Rapid Determination of Drugs and Semivolatile Organics by Direct Thermal Desorption Ion Trap Mass Spectrometry
Marcus B. Wise, Ralph H. Ilgner, Michelle V. Buchanan and Michael R. Guerin, Oak Ridge National Laboratory ....823
A New Approach for On-Site Monitoring of Organic Vapors at Low PPB Levels
H. Wohltjen, N.L. Jarvis and J.R. Lint, Microsensor Systems 829
A Rapid Screening Procedure for Determining Tritium in Soil
K.M. Wong and T.M. Carsen, Lawrence Livermore National Laboratory 835
Field Preparation and Stabilization of Volatile Organic Constituents of Water Samples by Off-Line Purge and Trap
Elizabeth Woolfenden, Perkin-Elmer Limited, James Ryan, The Perkin-Elmer Corporation 837
A Field-Portable Supercritical Fluid Extractor for Characterizing Semivolatile Organic Compounds in Waste and Soil Samples
Bob W. Wright, Cherylyn W. Wright and Jonathan S. Fruchter, Battelle, Pacific Northwest Laboratories 841
Detection of Mercuric Ion in Water with a Mercury-Specific Antibody
Dwane E. Wylie, Larry D. Carlson, Randy Carlson, Fred W. Wagner, Sheldon M. Schuster, BioNebraska 845
The Effects of Preservatives on Recovery and Analysis of Volatile Organic Compounds
Kaveh Zarrabi, Steven Ward, Thomas Starks and Charles Fitzsimmons, University of Nevada 849
Participants' List 851
XIII
-------
OPENING REMARKS
Welcome to the Second International Symposium on Field Screening Methods for Hazardous Waste and Toxic Chemicals.
Twenty-eight months ago, the first of these symposia was held here in Las Vegas, here at the Sahara, and the response to
that Symposium clearly indicated that the time was right. There was really a need for a forum to exchange information
about the emerging technologies that can be and have been applied to environmental monitoring in the field.
As you can see from the list of the Symposium sponsors, EPA is certainly not alone in its appreciation for these technolo-
gies and their potential for the future. I believe that we have assembled a powerful program for you this next two and a
half days.
The team responsible for this program was made up of, Mr. John Koutsandreas, the Executive Secretary from Florida
State University, Mr. Eric Koglin, the Matrix Manager here at EPA-Las Vegas for the Advanced Field Monitoring Meth-
ods Program and the coordinators at Life Systems, Inc. But for all of the efforts of the Symposium team, it's really the
interest, the enthusiasm, and the participation, over the next couple of days, of all the attendees, that will really set this
Symposium apart.
We are already planning for the Third International Symposium. We try to stagger them in such a way that enough time
elapses — that the papers aren't the same and the technologies have had an opportunity to advance. We're looking at just
about two years from now.
We are very interested in getting your feedback on what you like and what you don't like about the way the Symposium goes
this year, and any recommendations you can make to help us strengthen the next Symposium will be greatly appreciated.
This year, we have added a Scientific Awards Committee. Some of you have had an opportunity to see the certificates and
a couple of dramatic eagle trophies as you came in.
We're privileged to have a number of leaders in the area of environmental measurement here at this Symposium. We'll
share their views and the views of their organizations about current and future applications of field screening and field
analytical technologies.
Someone once said, Llew, why don't you write a poem? It was a long time ago, but it was never quite forgotten, so if
you'll bear with me:
The Second International has finally arrived.
Your program will suggest to you just how hard we have strived.
To bring to you the latest scoops and field technology
That's based on engineering, chemistry, biology.
Besides the platform papers that I know you'll want to hear.
The poster session entrees will just knock you on your ear.
And for the technophilic crowd, exhibitors galore
Will tell you all about their products, and a wee bit more.
We'll try to slake your appetite for the newest and the best.
And give you opportunities to mingle with the rest,
To share and learn, to see and show our efforts, may they yield
Accelerated products we can take into the field.
The future of our measurements, if any bets I'd hedge.
Resides in these technologies, we're on the leading edge.
So welcome to this overview of all those things to come.
And welcome to Las Vegas, where the Rebs are number one.
Llewellyn R. Williams
Symposium Chairperson
-------
KEYNOTE ADDRESS
ANALYTICAL ISSUES IN THE U.S. EPA SUPERFUND PROGRAM
Larry Reed, U.S. Environmental Protection Agency, Director Hazardous Site Evaluation Division, Office of Emergency and Remedial Response
I am glad to be invited to this Field Screening Symposium because our
Superfund office in EPA is such a primary user and booster of the
technology. We want to get more and more use out of the technology
and the field analytic methods. It's always good to be here and
participate. We've been very strong participants and boosters of the
EMSL-Las Vegas operation in field methods, and we'll continue to do
so for years to come.
I wanted to begin by setting the backdrop of where we are in the
Hazardous Waste Superfund Program, and then discuss the vital role
that field methods plays in that program.
We now have a Superfund Program that encompasses a full pipeline:
from discovery of sites (1,500) to two thousand new potential sites
identified to us each year, to the listing of approximately about one
hundred sites a year on the National Priorities List, through the
remedial action and remedial design process. The whole pipeline is in
complete use, and now more than ever, we're putting higher and
higher emphasis on focusing on the worst sites first throughout the
program. This obviously puts a premium on having the best and the
quickest environmental data available to evaluate and clean up sites
in the program.
The second aspect of where we are now in the Superfund Program that
bears on this Symposium, is we have just, (in December) promulgated
revisions to the Hazard Ranking System (HRS) which will become
effective March 12, 1991. This will expand the types of sites that we
will be looking at and screening. We have added new concerns, a
greater emphasis on ecological concerns, incorporated direct expo-
sure to soils and more emphasis on sediments. We are very proud of
this rule, and we will be gathering a lot more information for screening
sites for future National Priorities List updates.
Also, we have finalized the last of our proposed sites. In the Federal
Register, we proposed ten sites under the old HRS. All sites have
therefore been finalized. Eleven hundred and eighty-nine (1,189)
final sites are now on the National Priorities List. We will be hitting
the ground running, listing new sites as quickly as possible under the
new Hazard Ranking System, for the rest of the Superfund Program.
The focus of our Superfund Program has been on enforcement first,
integrating the use of the fund with the use of our enforcement
authorities. This is focusing more and more on a consistent use of
analytic methods, including both field methods and fixed lab meth-
ods, across the program, on appropriate QA procedures across all the
different types of sites, regardless of whether they are enforcement
lead, state lead or fund lead.
The final background point as far as field methods is concerned is our
adoption and phased incorporation of the principles of Total Quality
Management (TQM) into the Superfund Program. We began with
pilot projects last year, designed to embrace the principles of TQM.
The basic concepts of this program include:
• continual improvement in the process
• identifying our clients (ensuring that you know who they are, and
since there are various levels of clients and different relationships
with those clients)
• working with our clients
• identifying and addressing the worst problems first
• gathering data for informed decision making
These are the kinds of principles we're trying to address in all the
aspects of our program. So more and more, we'll be working with you,
and participating in this kind of audience where, at various times,
either you're our clients or we're your clients. This type of gathering
enforces that interaction among the various communities that deal
with field screening.
I now want to discuss some specific points on field analysis.
Howard Fribush of my staff went out and visited all 10 of our regional
offices to determine what is the state of the use of field screening in
our very decentralized program. We found field screening has a lot of
purposes including determining worker safety requirements, particu-
larly for our removal program and for the site assessment program,
which lists sites.
Field screening obviously provides immediate feedback to the site
assessors, to the samplers and to our clean-up contractors. That, again,
is a strong benefit that we see in encouraging the use of field methods
to continually improve and streamline our Superfund process.
An important application of field screening methods is how they can
be used to shorten the time that it takes to evaluate the risk posed at a
site. This can also be used to generate data to determine the appropri-
ate technologies to be used for clean-up and what levels of clean-up
are appropriate. These applications are evident looking at Regional
history—field screening technologies have been used in the Superfund
Program, basically from its inception. We have seen advances in field
instruments, and this is making on-site analysis at Superfund sites
much more desirable.
As part of Howard's Regional visits, the different aspects of our
Superfund program, were polled. The arms of the Program can be
divided into three functional aspects 1) the Site Assessment Program,
the front end of the Program that generates data needed to evaluate the
site and whether it needs to be included on the National Priorities List,
2) the Remedial Program, where once a site is on National Priorities
List the actual clean-up process is initiated, and 3) the Removal
Program, which can be called out at any time to clean up immediate
health threats at sites. We found a split among those different parts of
the program. About ten percent of the data being gathered for the Site
Assessment Program was from field screening. Similarly for the
Remedial Program, about ten percent of the data gathered was with
field screening methods, field analytic methods. The biggest user
proportionally was our Removal Program, slightly over a third of the
data gathered from the Removal Program is related to field screening
methods. What we'd like to do is, working through symposia such as
this, try to encourage and increase that use to even higher levels as
appropriate.
The role that we play in the Office of Emergency and Remedial
Response, and my division, the Hazardous Site Evaluation Division,
is basically providing guidance for this on-site analysis. As I men-
tioned before, when you're dealing with a decentralized program, you
always have to encourage consistency of methods among sites, but
you also have to deal with the uniqueness of each site. We are bridging
the gap by coming up with guidance to provide the Regional offices
on the use of methods.
-------
As follow-up to the Regional review we are evaluating the advantages
of field analysis in the Superfund Program and building on that to
expand its uses as appropriate in the future. Our future guidance
documents will address evaluating when to use it and then how to use
it. We are also trying to get consistent terminology in our guidance.
Screening technology, portable methods, fieldable methods, mobile
methods, all of these terms have been used. We've been trying in our
field methods catalog to come up with some consistency so even those
unfamiliar with these various technologies, can become familiar with
the basic terminology.
Several major efforts are underway in the Superfund Program. Within
the last year we established our first field methods management
forum. The focus of that field methods management forum was to get
managers involved, not just those that have to go out in the field and
implement the technology, but the managers who would be the ones
to determine what proportion of overall analytic support is necessary
for field methods versus fixed labs. The first meeting of this gpoup was
in June, 1990. We had seven regions, headquarter offices, and the
EMSL-Las Vegas group at this session. The objective of this effort
was to get management involved and to focus on the blockages
preventing us from getting field methods used to a greater extent.
Future topics for meetings include: 1) regional administration of field
screening (where does it go, who is in charge;) 2) collecting the
method and instrument performance information, and 3) trying to get
this data out to the field in the best usable form to those familiar with
the technologies, their usage, their limitations and their strengths.
There's another effort underway — the Field Methods Work Group.
This group contains the worker bees, the people that have to go out and
get the job done. This group has been meeting since 1987. Their initial
focus was looking at things at the very basic level of data quality
objectives—how to define them in order to get them in a more useful
format understood by both the chemists and the field engineers. In
July, 1990, they met and focused on the catalog of field methods and
the need for a new version. The Field Methods Screening Catalog
User's Guide came out three or four years ago, and we realized the
limitations of it. At the time we wanted to get information on some 30
different screening methods. Obviously the next stage is updating
this, adding more methods and more data that will be useful to our
field offices. We expect to release this update of the Field Methods
Catalog some time this year. It will triple the number of methods that
are contained in the original catalog to about 100.
We are obviously going to be looking at both QA and QC of field
methods. The basic question is the need for Regional consistency.
What is appropriate Q A for a field lab? How can we get that guidance
out? What are the appropriate QC requirements for field methods? We
need to get that information out to the field again, by bringing in the
user community.
Another issue obviously encouraging consistency and appropriate
use of the technology is training. We have been working to come up
with a training program on field methods with the regions and EMSL-
Las Vegas. We've even gotten one of our regional offices to hopefully
loan some of their field equipment in a true bureaucratic gesture to
EMSL-Las Vegas to use as a basis for training programs. We hope to
have this training program developed this year. Obviously the level
we'll have to look at then is how much and what level of training do
we need to provide out there? How much training should be done for
the people using the field methods? At what levels should it be
presented, and how much should be mandatory to ensure and promote
consistency?
There are several basic field method issues that I haven't mentioned,
but that I'd like to touch on before closing: how do we capture
performance information on methods and the instruments? The state-
of-the-art is obviously rapidly changing. How do we capture that
information given, among other things, federal regulations about how
much we can provide in working with industry. How do we capture
that performance data and get it to the field for use in the most useable
form? We have 100 methods that we have looked at for the upcoming
catalog update. What type of data do the people want, and what type
of format? How much? Do they want extensive data, shortened data
or very abstract data. What type of data will encourage the use in the
field?
The final point, and one that I know this Symposium will be working
on, is introducing improved methods, particularly to the Superfund
Program. How do we get the new methods out? What are the incentive
systems? How do we call out and identify the best methods so that they
are being selected for use in the field?
In closing, there are a lot of efforts we have underway to encourage
a maximal, appropriate use of field screening methods. This sympo-
sium is a key one. I mentioned the Field Methods Management Forum
and the Field Methods Work Group, two continuing efforts to provide
direction and recommendations for additional guidance for consis-
tency and use of technology in the field. Field screening methods is a
big field. It is a continuing, emerging field that will continue to
command national attention. We in the Superfund Program are great
boosters and great users of it. I speak as both a provider, working with
EMSL-Las Vegas and their services, and a user, working on risk
assessments and the site assessment program. I encourage you in your
pursuits to increase the use of field screening methods.
-------
OVERVIEW OF DOE'S FIELD SCREENING
TECHNOLOGY DEVELOPMENT ACTIVITIES
by
C.W. Frank, T.D. Anderson, C.R. Cooley, KJE. Hain, and S.C.T. Lien
Office of Technology Development
U.S. Department of Energy
Washington, DC 20874
R.L. Snipes
Support Contractor Office
Martin Marietta Energy Systems
Oak Ridge, Tennessee 37831
M.D. Erickson
Research and Development Program Coordination Office
Chemical Technical Division
Argonne National Laboratory
Argonne, Illinois 60439
ABSTRACT
The Department of Energy (DOE) has recently created
the Office of Environmental Restoration and Waste
Management, into which it consolidated those activities.
Within this new organization, the Office of Technology
Development (OTD) is responsible for research,
development, demonstration, testing, and evaluation
(RDDT&E) activities aimed at meeting DOE cleanup
goals, while minimizing cost and risk. Site
characterization using traditional drilling, sampling, and
analytical methods comprises a significant part of the
environmental restoration efforts in terms of both cost
and time to accomplish. It can also be invasive and
create additional pathways for spread of contaminants.
Consequently, DOE is focusing on site characterization
as one of the areas in which significant technological
Work supported by the U.S. Department of Energy,
under contract W-31-109-Eng 38.
advances are possible which will decrease cost, reduce
risk, and shorten schedules for achieving restoration
goals. DOE is investing considerably in R&D and
demonstration activities which will improve the abilities
to screen chemical, radiological, and physical parameters
in the field. This paper presents an overview of the
program objectives and status and reviews some of the
projects which are currently underway in the area.
INTRODUCTION
The Department of Energy (DOE) has recently
consolidated its environmental restoration and waste
management activities into the Office of Environmental
Restoration and Waste Management, formed by
Secretary James Watkins in early 1989. Within that
Office of Technology Development, in part
The_submitted manuscript has been authored
by a contractor of the U. S. Government
under contract No. W-3M09-ENG-38.
Accordingly, the U. S Government retains a
nonexclusive, royalty-free license to publish
or reproduce the published form of this
contribution, or allow others to do so, for
U. S. Government purposes.
-------
new organization, the Office of Technology
Development (OTD) oversees DOE's Technology
Development Program, whose objective is to establish
and maintain a national program for applied research,
development, demonstration, testing, and evaluation
(RDDT&E). These activities will pursue technologies
that will enable DOE to meet its 30-year compliance
and cleanup goals safely, efficiently, and effectively.(1)
The first step in environmental restoration is site and
contaminant characterization. Characterization of the
current distribution of contaminants and the
geohydrological factors that promote and control their
spread will provide the starting point for determining
what must be remediated and for selecting and
designing remediation methods.
STATUS OF OTD ACTIVITIES
A cross section of the technology development activities
which have been or are being conducted are described
below. Space limitations preclude describing all
activities in this area. Some of these activities will be
described in more detail by the principal investigators at
this conference.
DUVAS Fiberscope for in Situ Groundwater
Monitoring. Because of its proven ability to detect
compounds such as benzene and its derivatives, which
are common solvents and components of fuels,
derivative ultraviolet absorption spectrometry (DUVAS)
is being developed as a rapid and reliable method for in
situ detection of aromatic pollutants. To date, a
prototype DUVAS fiberscope has been constructed and
tested for measuring spatial and temporal distribution of
organics in groundwater. An important component of
the fiberscope is a rugged, down-well probe with a
unique "detector-in-head" design that increases the
maximum depth of subsurface detection. Results
comparable to those obtained with a conventional
laboratory spectrometer have been achieved with optical
fiber lengths up to 50 meters. The portable DUVAS
fiberscope will provide faster, more reliable, and less
expensive measurement of subsurface groundwater
contamination. For further information, contact the
Principal Investigators, J.W. Haas in and R.B.
Gammage, Oak Ridge National Laboratory, P.O. Box
2008, Oak Ridge, TN 37831-6113. Phone: (615) 574-
5042 (Haas), (615) 574-6256 (Gammage).
Advances in Surface-Enhanced Raman Spectroscopv for
Applications in Real-Time Subsurface Monitoring.
Because of its excellent selectivity, surface-enhanced
Raman scattering (SERS) has attracted considerable
attention as a potentially powerful analytical tool for
detecting and screening trace-level contaminants in
groundwater. The narrow Raman bands hold promise
for simplifying the identification of individual
components in complex mixtures. An inexpensive
computer-controlled portable spectrometer system
coupled to a fiber-optic probe is being developed for
rapid on-site and in situ determination of organic
groundwater contamination. Critical issues pertaining to
durability, repeatability, sensitivity, selectivity, and
universality are being examined, while means for
improvement in these areas are being tested. The
feasibility of utilizing SERS under harsh conditions has
been demonstrated. Substrates have been tailored for
maximum efficiency at particular excitation wavelengths
as a means for increasing the sensitivity of the
technique. Ongoing efforts have refined the state-of-the-
art Raman optrode design and have shown the feasibility
of producing a simple, inexpensive instrument for field
applications. As the technique approaches maturity,
SERS will provide powerful screening capabilities for
numerous organic and inorganic materials. It promises
rapid, reproducible, quantitative detection of trace-level
contaminants in aqueous solutions. For further
information, contact the Principal Investigator, Eric A.
Wachter, Oak Ridge National Laboratory, Health and
Safety Research Division, P.O. Box 2008, Oak Ridge,
TN 37831. Phone: (615) 574-6248 (FTS 624-6248).
Fiber Optic Raman Spectrograph for in Situ
Environmental Monitoring. A small (suitcase-sized)
surface-enhanced Raman spectrometer (SERS) is being
developed to use in field screening for a wide variety of
-------
organic and metallic pollutants in ground and surface
waters. The focus of this contract is twofold: (1) to
demonstrate a small spectrograph with high resolution
(3500 cm"1) and (2)
to demonstrate a micro-optical SERS probe head with
substrates engineered to detect certain critical pollutants
at ppm to ppb levels. The spectrograph will have no
moving parts and will employ fiber-optic sampling, an
ultracompact solid-state diode laser for Raman
excitation, a high-order diffraction grating, holographic
optical filters, and a state-of-the-art charge-coupled
device (CCD) detector. The probe head will be
contained at the sampling end of a fiber-optic probe
over 50 meters long inserted into a well less than 5
centimeters in diameter. The system will identify trace
contaminants in groundwater in real time.
This technique will increase the efficiency of
environmental characterization and mapping, reduce
costs of field sampling and ex situ laboratory analysis,
reduce personnel exposure, and provide site
characterization information. For further information,
contact the Principal Investigator, Michael Carraba, EIC
Laboratories, Inc., Ill Downey St., Norwood, MA
02062, Phone: (617) 769-9540.
In Sim Detection of Organics. The long-term objective
of this research is to develop a fiber-optic-based system
for monitoring contaminant species in groundwater and
to demonstrate it on contaminated groundwater at
Lawrence Livermore National Laboratory (LLNL).
These efforts require the development of optical
indicator reagents that are compatible with fiber-optic
chemical sensors (optrodes). Development of optrodes
for ppb-level detection of trichlorethylene (TCE) and
chloroform (CHC13) is complete and has moved into the
demonstration phase. Carbon tetrachloride (CCl^ and
perchloroethylene (PCE) optrodes are currently being
developed.
The fiber-optic approach has the potential of providing
less expensive measurements of groundwater
contaminants. Also, the reagent indicators and the
chemistry developed in the process of developing the
optrodes will "spin off into other applications. For
example, one chemistry that was developed serves as the
basis for a proposed TCE remediation technique, the
"TCE sponge". Finally, it should be pointed out that
these simple indicators are new and could be used in
other types of contaminant assays. For further
information, contact the Principal Investigator, Mike
Angel, Lawrence Livermore National Laboratory,
Environmental Sciences Division, P.O. Box 808, L-524,
Livermore, CA 94550. Phone: (415) 423-0375 (FTS
543-0375).
Optical Fiber Photothertnal Spectroscopies for in Situ
Monitoring and Characterization. Optical fiber sensors
using thermal lens and photoacoustic spectroscopies for
remote, on-site, real-time optical absorption
measurements of chemical species in groundwater
environments are being developed. Optical fiber sensors
based on photothermal spectroscopies are ideal for
ultrasensitive optical absorption measurements of
actinides and other chemical species in aqueous
environments. An optical absorption spectrum provides
qualitative and quantitative analysis of the species
present in the aqueous environment. The spectra can
also provide complexation information for actinides,
which is important for migration behavior. These
photothermal sensors rely on tunable wavelength for
selectivity and therefore do not require immobilized
agents at the distal fiber end (in the sample area).
Research has demonstrated two optical fiber
photothermal sensors with excellent sensitivity for rare
earth and actinide ions in aqueous solutions. A remote
photoacoustic sensor was demonstrated using a 100-
meter fiber to deliver the tunable laser beam to a glove
box located in a separate room from the laser. Acoustic
signals were returned to the instrument lab via coaxial
cables. An all-fiber thermal lens sensor was
demonstrated using a fiber to deliver the laser light to a
remote sample solution and a second fiber, with a
photodiode attached to the distal end, to measure optical
absorption; electrical cables were not required at the
sample area. For further information, contact the
Principal Investigators, Richard Russo, Lawrence
-------
Berkeley Laboratory, Applied Science Division, M.S.90-
2024, Berkeley, CA 94720, Phone: (415) 486-4258 (FTS
452-4258); and Robert Silva, Lawrence Livermore
National Laboratory, Nuclear Chemistry Division, L-
396, Livermore, CA 94550. Phone: (415) 423-9798
(FTS 543-9798).
Field Measurement of Groundwater Contamination by
Ion Trap Mass Spectrometrv. A transportable ion trap
mass spectrometer for the in situ characterization of soil,
air, or water at chemical waste sites is being developed
and demonstrated. The instrument will have a turnkey
operating system for use by minimally trained
personnel. The approach uses modular design to
produce an instrument that can be readily modified and
repaired in the field. Specifically, this project will
develop a daughter microprocessor system to control
ancillary hardware for sampling and separation and will
develop new software, write macros, and modify
existing software for semi-automated computer control
of the instrument.
The instrument consists of specialized sampling modules
for air, soil, or water samples; a separations module
containing sorbent traps and a megabore capillary
chromatography column; and a detection module, the
Finnigan Ion Trap Detector. Soil or water samples are
purged with helium and the evolved organics are
collected on sorbent traps. A sampling pump is
incorporated for air samples. The full analysis sequence
required 10 minutes. The Finnigan software was
modified through the addition of macros and Forth
routines. The analytical procedure can be selected from
a menu from the instrument's data system. Sampling,
calibration, analysis, and data reduction proceed under
computer control
The detection limit for TCE in water is approximately
20 picograms. Mass spectral identification of 50
picograms of TCE is possible by library comparison of
spectra. A linear calibration curve can be obtained from
10 ppt to 10 ppm organics in water.
Although transportable mass spectrometers are
commercially available for environmental analyses in
the field, the transportable ion trap technology described
here provides several additional benefits, including low
cost. The instrument can be assembled for a parts cost
of about $75K. For further information, contact the
Principal Investigator, Philip H. Hemberger, Los Alamos
National Laboratory, Analytical Chemistry Group, Mail
Stop G740, Los Alamos, NM 87545. Phone: (505) 667-
7736 (FTS 843-7236).
Direct Sampling Mass Spectrometrv. Rapid analytical
technology based upon direct sampling mass
spectrometry is being developed to determine trace
organic pollutants in the environment. This project is
jointly sponsored by DOE, the Department of the Army,
and EPA. Closely related work is sponsored by the
National Cancer Institute (NCI) for analyses of
physiological fluids. Oak Ridge National Laboratory
(ORNL) has developed sampling, sample interface, and
ionization chemistry techniques that are first being
combined with commercial mass spectrometers to
provide rapid laboratory-based methods. Knowledge
gained is used to develop instrumentation optimized for
on-site analysis. Field-sampling and field-sample-
processing methods are being developed to support the
mass spectrometric technologies. The general approach
involves a systematic comparison of the developed
methods using accepted EPA methods to analyze
organics in water, soil, air, and waste. Ion trap mass
spectrometry (TTMS) and glow discharge ionization
quadrupole mass spectrometry (GDMS) are being
investigated. Both GDMS and ITMS are applicable to
the quantitative determination of ppb concentrations of
organics in water and in soil with analysis times of five
minutes or less. This is achieved by purging the water
or soil-water slurry with air or helium and routing the
purge stream directly into the mass spectrometer. Less
volatile organics may be similarly determined by
collection on a suitable solid sorbent followed by
thermal desorption. The method has thus far been
demonstrated for the quantitative determination of
benzene, trichloroethylene, and tetrachloroethylene.
Applicability to semivolatiles has been demonstrated by
-------
the successful determination of nicotine and cotinine in
urine for the NCI and for die determination of military
chemical agents in air for the Army. A method is under
development for the simultaneous collection of samples
for subsequent confirmatory analysis in those cases
where interferences cannot be distinguished by mass
spectrometry or by mass spectrometry/mass
spectrometry alone.
Successful development and validation can reduce costs
and increase sample throughput by up to 90% as
compared to current regulatory analytical methods.
Field-versions of the technology will allow real-time
monitoring of remedial action progress, monitoring of
associated occupational exposure, and screening of
samples prior to shipment to die laboratory for
regulatory analyses. For further information, contact the
Principal Investigators, M.B. Wise, M.R. Guerin, and
M.V. Buchanan, Oak Ridge National Laboratory, P.O.
Box 2008, Bldg. 4500-S, MS-6120, Oak Ridge, TN
37831-6120. Phone (615) 574-4862 (FTS 624-4862)
(Mike Guerin).
Assessment of Subsurface Volatile Organic Compounds
(VOCs) Using Chemical Microsensor Arrays. A new
monitoring instrument that utilizes an array of coated
surface-acoustic-wave (SAW) microsensors is being
developed. Pattern recognition analysis of the
multidimensional sensor output permits determination of
the identity and quantity of target vapors from
difference chemical classes typically found in
contaminated soils and groundwater. The small size,
low cost, low power requirements, high sensitivity, and
large dynamic range of the instrument will facilitate its
use in a variety of applications related to site assessment
and process and control.
The project addresses some fundamental questions: (1)
what is the performance of the SAW microsensor array
instrument in applications relevant to site assessment
and restoration, namely, monitoring volatile organic
chemicals (VOCs) in high humidity environments, (2)
how are the measurements provided by this instrument
related to soil contaminant levels, and (3) how can they
best be utilized in site assessment and restoration
activities? A series of controlled laboratory experiments
will be performed to address diese questions.
The results of this research will demonstrate that
microsensor array instruments can provide rapid and
reliable compound-specific concentrations of volatile
organics in soil vapor. The low projected cost of
manufacture (less than $1000 in production quantities),
the capabilities of continuous, unattended operation, and
the ability to transmit data from remote locations make
the SAW sensor-based monitors a cost-effective and
desirable monitoring approach. For further information,
contact the Principal Investigator, Stuart Batterman,
University of Michigan, Department of Environment &
Industrial Health, 2505 School of Public Health, Ann
Arbor, MI 48109-2029. Phone: (313) 763-2417.
Thin-Layer Detectors: NO2 Detection with Polystyrene
Thin Layers. A solid-state sensor that can be used to
detect NO2 without interference by other species is
being developed. The device incorporates an
interdigitated electrode with a polystyrene thin layer and
operates by simply monitoring the change in
conductance of this thin film as a function of NO2
exposure. Although the film is an insulator in the
absence of NO2 , showing conductance of less than
10 S, upon exposure to NO2 gas, an increase in
conductivity of this highly insulating material occurs
over several orders of magnitude to 10"8-10"9 S. No
interference from ambient gases or water vapor has been
observed, and the effect is very specific to NO2. Upon
elimination of the NO2 gas, the device becomes
completely insulating again, all effects occurring at
ambient temperature and pressure.
The mechanism of the conduction within the film
remains unclear, although the level of conductivity is
related to the amount of residual benzene solvent within
the film. Thus, as the benzene evaporates from the
film, the change in conductivity of the film upon NO2
exposure diminishes dramatically. This effect appears
to be related to a stabilization of NO2 dimer by benzene
within the film. The increased conductivity of the film
-------
in the presence of benzene is attributed to the well-
known self-ionization of N2O4 to NO+ + NO3". For
further information contact the Principal Investigator
Stephen F. Agnew, Los Alamos National Laboratory,
Los Alamos, NM 87545. Phone: (505) 665-1764 (FTS
843-1764).
Antibody-Based Fiberoptics Sensors For in Situ
Monitoring. Sensitive and selective chemical sensors
for in situ monitoring of hazardous compounds in
complex samples are being developed. Special focus is
on a unique fluoroimmuno-sensor (FIS) which derives
its analytical selectivity through the specificity of
antibody-antigen reactions. Antibodies are imirtobilized
at the terminus of a fiberoptic within the FIS for use in
in situ fluorescence assays under field conditions. High
sensitivity is provided by laser excitation and optical
detection techniques. The technique can detect
femtomoles (10 M) of the carcinogen benzo(a)pyrene
and other chemicals of environmental interest. For
further information, contact the Principal Investigators,
T. Vo-Dinh and G.D. Griffin, Oak Ridge National
Laboratory, P.O. Box 1008, MS-6101, Oak Ridge, TN
37831-6101. Phone: (615) 574-6249 (Vo-Dinh) and
(615) 576-2713 (Griffin).
Underground Imaging for Site Characterization and
Clean Up Monitoring. State-of-the-art image
reconstruction techniques (tomography) can be used to
characterize the geology and hydrology of hazardous
waste sites. These methods extend spatial information
of geologic structure and hydrology between boreholes.
Both two- and three-dimensional imaging can be done
using these techniques. High-frequency electromagnetic
(HFEM) tomography is a proven technology for imaging
water content with high spatial resolution, (i.e., submeter
scale for small geologic scale applications (ten meters).
Electrical resistance tomography (ERT) is a newer
technology which has been used in the field with
moderate-scale resolution on larger scale images (meters
on tens to hundreds of meters).
Characterization of the subsurface geology and
hydrology is needed to select the most appropriate
remediation alternative and to demonstrate regulatory
compliance. Design of remedial actions must be based
upon knowledge of the often anisotropic and
heterogenous nature of the subsurface environment and
the natural processes that act upon the waste, as well as
upon protective barriers. Groundwater flow strongly
influences contaminant mobilization and transport and
geologic structure affects the flow of groundwater.
Current subsurface characterization techniques for
addressing these above problems depend heavily upon
drilled boreholes. Drilling is expensive and time
consuming and also creates conduits for contaminant
spread. A special need exists for three-dimensional
noninvasive subsurface characterization technologies.
For more information, contact the Principal Investigator,
William Daily, Lawrence Livermore National
Laboratory, P.O. Box 808, L-156, Livermore, CA
94550. Phone: (415) 422-8623 (FTS 532-8623).
Development of the SEAMIST Concept for Site
Characterization and Monitoring. This project is
developing the Science and Engineering Associates'
Membrane Instrumentation and Sampling Technique
(SEAMIST). The technique permits rapid emplacement
of instrumentation and sampling apparatus in a punched
or drilled hole. The objective of the technique is to
pneumatically emplace an impermeable membrane liner
carrying many instruments into a hole to provide
simultaneous access to the entire hole wall (e.g., many
measurement horizons per hole), elimination of
circulation of fluids within the hole, and isolation of
instruments at discrete locations between the hole wall
and the membrane. The membrane is emplaced by
eversion—it is rolled inside out and then everted using
air pressure. This causes minimal disturbance to the
hole because the assembly does not slide down as with
traditional rigid casings. Instruments such as fiber-optic
sensors, thermocouple psychrometers, gas- and liquid-
sampling systems, and other small instruments are easily
attached to the membrane and carried into the hole with
it.
Using this technique will save 50%-90% of the field
costs, as compared to current monitoring well practices.
10
-------
In addition, the technique is applicable to both vertical
and horizontal wells. For further information, contact
the Principal Investigator, Carl Keller, Science and
Engineering Associates, 612 Old Santa Fe Trail, Same
Fe, NM 87501. Phone: (505) 646-5188.
Site Characterization and Analysis Penetrometer System
(SCAPS). DOE is working with the Department of
Defense on the further development and demonstration
of the SCAPS for use on DOE facilities. The SCAPS,
as developed by the Army Corps of Engineers
Waterways Experiment Station for the Army Toxic and
Hazardous Materials Agency, includes surface
geophysical equipment, survey and mapping equipment,
sensors for contaminant detection, and soil sampling
equipment. Computer systems have been integrated
with the SCAPS in order to provide data acquisition,
data processing, and 3-D visualization of site conditions.
The system is mounted on a uniquely-engineered truck
that provides protective work spaces to minimize worker
exposure to toxic chemicals. The truck also provides
equipment to seal each penetrometer hole with grout.
Real-time sensors that are currently available for
characterization work include those which can determine
the strength, electrical resistivity, and spectral properties
of soils. Two sensors successfully demonstrated to
detect contaminant plumes at DOD facilities are the soil
resistivity unit and a fiber optic contaminant sensor.
The primary advantage of the fiber-optic sensor over
resistivity measurements is based on laser-induced
fluorescence, which presents a problem for contaminants
such as TCE that do not fluoresce; however, colorimetry
and absorption techniques such as the sensors which are
being developed by Lawrence Livermore National
Laboratory and by Fiberchem are tentatively planned to
be demonstrated in conjunction with the penetrometer at
the Savannah River integrated demonstration in FY-91.
Additionally, samplers such as the "Terra Trog"
developed by the Army Corps of Engineers may be
tested in FY-91 at the Savannah River Site. For further
information, contact the Principal Investigator, Stafford
Cooper, Waterways Experiment Station, P.O. Box 631,
Vicksburg, MS 39181-0631. Phone: 601-634-2477.
Design. Manufacture, and Evaluation of a Hvdraulicallv
Installed. Multi-Sampling Lvsimeter. A new lysimeter
sampling device design, approximately 1 inch in
diameter, having multiple sampling zones and capable
of being hydraulically installed at a desired depth in the
vadose zone without drilling will be developed. This
lysimeter will be readily retrievable for reuse and will
provide an inexpensive monitoring technique in
comparison to installation of lysimeters into predrilled
holes. In this project, the hydraulically inserted
lysimeter will be designed and constructed. The effect
of hydraulic insertion on the operation of the lysimeter
will be investigated by comparing hydraulic insertion
with standard boring procedures. The lysimeter should
be commercialized within three years. This new design
is less disruptive to the subsurface, both during
installation and after removal, requiring only a 1-inch-
diameter hole vs. the 4-inch holes commonly drilled for
monitoring wells. Costs are estimated to be under 50%
of that to drill monitoring wells. This project is a
collaborative effort among Bladon International, Inc.,
Institute for Gas Technology, and Timco Manufacturing.
For further information contact the Principal
Investigator, Joe Scroppo, Bladon International, Inc.,
880 Lee Street, Des Plaines, IL 60018. Phone: (505)
883-3636.
Minimally Invasive Three-Dimensional Site
Characterization. Hardware and software are being
developed to permit data acquisition from three
minimally invasive measurement techniques—cone
penetrometer, synergistic electromagnetic mapping
technology and reflection seismology. The software will
permit rapid feedback, comparison, co-calibration, and
data analysis from the combined technology.
Simultaneous application of these three technologies
permits physical and electrical property measurements
to be used to cross-calibrate each data set. The early
acquisition of preliminary data allows field personnel
quickly to adapt their field study strategy to changes in
the perceived site conditions or contamination
distribution.
11
-------
Costs will be saved by rapid feedback of the data to
field personnel, the improved informational quality, and
the lower cost of an integrated system. The minimally
invasive system reduces environmental impact and
reduces risk to field personnel. For further information
contact Principal Investigator, John Gibbons, Applied
Research Associates, Inc., 4300 San Mateo Blvd., N.E.,
Suite A220, Alburquerque, NM 87110. Phone: (505)
883-3636.
High Resolution Shear Wave Seismic Reflection
Surveying for Hydrogeological Investigation. This
technology will enhance the ability to directly determine
aquifers in the characterization and sensing of geologic
and hydrogeologic features. The project will extend the
state-of-the-art of shallow subsurface hydro-geological
characterizations by means of high resolution shear (S)
wave seismic reflection profiling. High resolution
seismic reflection profiling using conventional
compressional (P) wave technology has evolved over the
past ten years to the point where this technique has
become a major component of numerous environmental
investigations. Extension of the existing technology to
include S-wave reflections has the potential for greatly
enhancing the data which can be extracted from the
subsurface. Unlike a P-wave, an S-wave will not travel
through a purely liquid medium, hence its advantage
over current P-wave techniques.
Conventional high-resolution seismic reflection profiling
has proven cost-effective for environmental assessment
by reducing the number of holes and the cost of boring.
S-wave reflection technology will enhance the
information content of the seismic reflection technique
and improve the cost-effectiveness of the technique. For
further information contact the Principal Investigator,
William Johnson, Paul C. Rizzo Associates, Inc., 300
Oxford Dr., Monroeville, PA 15146. Phone: (412) 856-
9700. .
Field Measurements for the Hydrology and Radionuclide
Migration Program (HRMP) at the Nevada Test Site.
The HRMP was begun in 1974 for the purpose of
determining the potential for migration of radionuclides
from underground test areas. HRMP is a multi-agency
research project and is coordinated by the Nevada
Operations Office of DOE. The participants are
Lawrence Livermore National Laboratory, Los Alamos
National Laboratory, Desert Research Institute, and the
U.S. Geological Survey. The present goals of the
program are to learn more about the groundwater rates
and directions of flow on the Nevada test site (NTS),
which is located approximately 80 miles northwest of
Las Vegas, in regional and local systems, to develop
mathematical models of the flow systems, to determine
the effects of nuclear tests on the systems, and to
measure the migration rates of selected radionuclides
under various conditions.
Transport mechanisms for radionuclides from
underground nuclear detonations are studied by
sampling both the contaminated cavity water and
groundwater pumped from the surrounding formation.
Radioactivity in water greater than 9-cavity-radii
distance from the detonation point has been measured
without stressing or pumping the aquifer. A plume of
radioactivity which is being rapidly transported by the
local groundwater has been intercepted. Micro- and
ultrafiltration studies on this groundwater have shown
that radionuclides can be present and mobile in
groundwater systems in colloidal form. Water pumped
from a tritium contaminated satellite well over a 20-year
period drains into a mile-long ditch and has created a
secondary site emphasizing the unsaturated zone.
Current studies along the discharge ditch are
investigating the moisture and tritium front through
shallow alluvium. This project is developing systems
which can measure contaminants such as organics,
tritium, and long-lived radionuclides in wells in depths
from 1400 to 3300 feet. For further information,
contact the Principal Investigator, Jo Ann Rego at
Nuclear Chemistry Division, Lawrence Livermore
National Laboratory, P.O. Box 808, L-234, Livermore,
CA 94551. Phone: (415) 422-5516 (FTS 532-5516).
Depth Profiling in the Water Table Region of a Sandy
Aquifer. The feasibility of using a new multilayered
sampler to investigate organic contaminants in
12
-------
groundwater is being explored. The device passively
collects simultaneous groundwater samples from
multiple levels in the subsurface. In addition, the
project will develop a new device based on experience
with existing sampler.
The sampler, developed at the Weizmann Institute of
Sciences, Rehovot, Israel, was used to detect the
presence of several inorganic and organic species at a
contaminated Brookhaven site. The presence of
microscale heterogeneities in concentration gradients
over a vertical interval of 200 cm was observed for
eight solutes, including metals, organics, and anions. A
planned remediation was modified based on results of
this short sampling event. It is believed that the new
plan will be more cost effective than the original
because the contamination was better defined in the
vertical plane and because an oxygen-depleted zone was
found where it was previously thought to be fully
saturated. For further information, contact the Principal
Investigator, Edward Kaplan, Brookhaven National
Laboratory, Radiological Sciences Division, Building
703M, Upton, NY 11973-5000. Phone: (516) 282-2007
(FTS 666-2007).
Kr81 Counting for Nuclear Waste Sites. A new
technology to date groundwater is being developed. By
combining resonance ionization spectroscopy and mass
spectroscopy, ultralow levels of Kr81 in groundwater can
be detected. From the quantity of Kr8 , the age of the
groundwater can be determined. This information helps
find suitable locations to store nuclear wastes or highly
toxic chemical wastes in groundwater. Several samples
from Europe have been tested and the results are
adequate to search for new waste sites. It is beneficial
to the Department of Energy waste program to find a
geologically safe place to store nuclear wastes and
highly toxic chemical wastes. For further information,
contact the Principal Investigators, C.H. Chen and M.G.
Payne, Oak Ridge National Laboratory, Photophysics
Group, Building 5500, MS-6378, P.O. Box 2008, Oak
Ridge, TN 37831-6378. Phone: (615) 574-5895 (FTS
574-5895).
FUTURE TECHNOLOGY DEVELOPMENT
NEEDS
The OTD activities described here address some, but by
no means all, of the key needs which DOE foresees in
the area of in situ monitoring.
Present site characterization methods are imprecise,
costly, time-consuming, and overly invasive. Improved
site characterization methods will require better
technologies for accurately describing the subsurface
geohydrologic features of a site. For example, more
efficient nonintrusive sampling strategies and practical
models are necessary for understanding and predicting
subsurface transport. Also needed are more reliable
procedures for interpreting characterization data, such as
how clean is "clean".
Traditional hydrologic characterization of the subsurface
environment is highly dependent on data from
groundwater monitoring wells. A thorough
understanding of the subsurface environment requires a
series of hydraulic wells. Interpretation depends greatly
on proficiency of the scientific staff, making subsurface
characterization highly subjective and at times uncertain.
Research is needed to make hydrologic characterization
more precise and more cost effective.
Currently accepted analytical procedures such as those
in the Environmental Protection Agency's (EPA's) SW-
846 do not cover all materials that need to be measured
at DOE sites. DOE is working with the EPA and others
to alleviate such problems with sampling and analyses.
Close coordination with EPA and other regulatory
agencies is needed not only to identify, develop, and
validate appropriate methods, but also to ensure the
acceptance of data generated using these methods.
Intrusive exercises, such as sampling and excavation
during remediation of a site, often involve immediate
hazards to workers in the form of exposure to
radioactive and/or toxic materials. Remote real-time
analyses of ambient levels of potential hazards in the
air, water, and soil during characterization, as well as in
13
-------
the remedial action phase, would help ensure worker
safety and allow continuous operation. Instrumentation
capable of detecting broad classes of hazardous
materials and specific compounds is needed to indicate
cleanup status. Better characterization methods based
on real-time analyses are especially important to confirm
the most effective use of certain in situ remediation
technologies. In the absence of real-time monitoring,
excessive volumes of soil and water must be treated to
guarantee compliance; otherwise, pockets of
contamination may be missed.
Special characterization technologies are necessary for
inactive facilities, underground storage tanks, and
wastewater lagoons. These facilities often contain
significant quantities of radioactive wastes, in certain
cases mixed with heavy metals and/or hazardous organic
compounds that make personnel entry unacceptable.
Thus, the development of advanced robotic samplers,
smart probes, mobile and in situ fiber-optic devices, and
nonintrusive characterization instrumentation (based on
electromagnetic, thermographic, and acoustic principles)
is needed for sampling and chemically characterizing
these sites. The development of such techniques will
significantly reduce radiological exposure to workers
and provide more assurance that the correct remedial
technology has been selected.
Clearly, there are more technology development needs
and more good ideas than there are resources to devote
to these investigations. Priorities must be set to support
those activities deemed most urgent.
OPPORTUNITIES FOR PARTICIPATION
OTD is interested in eliciting broad participation from
qualified organizations who can contribute to its
RDDT&E activities. We are becoming increasingly
aware of the wealth of technological talent and good
ideas in all sectors. OTD has initiated steps during the
past year to increase participation of the private sector
(academia and industry) through competitive
solicitations and through funding of unsolicited
proposals. We have also worked to increase
participation by academia through interagency
agreements for cooperative funding of research and
through establishment of DOE educational consortia.
Several significant technology development activities are
being conducted at DOE sites such as national
laboratories. DOE is funding technology development
activities beyond the United States through direct
contracts, international agreements, and other
mechanisms.
DOE plans to continue this type of support for
technology development in the coming years.
Organizations interested in responding to solicitations
should contact John Beller (for Innovative Technology)
at Innovative Technology Program Coordination Office,
EG&G Idaho, P.O. Box 1625, Idaho Falls, ID 83405-
6902. Dr. Erickson (for applied R&D) at the above
address or Mr. Snipes (for DT&E) at the above address
to be placed on distribution lists. Organizations wishing
to submit unsolicited proposals should contact Larry
Harmon, Director, Division of Program Support (EM-
53), Department of Energy, 12800 Middlebrook Road,
Trevion II Building, Germantown, MD 20874, for
information on submission format and procedures prior
to preparation of a proposal.
REFERENCES
1. United States Department of Energy
Environmental Restoration and Waste
Management. Five-Year Plan, Fiscal Years 1992-
1996, June 1990, DOE/S-0078P.
14
-------
DEPARTMENT OF DEFENSE FIELD SCREENING METHODS REQUIREMENTS IN THE
INSTALLATION RESTORATION PROGRAM
Mr. Dennis J. Wynne
U.S. Army Toxic and Hazardous Materials Agency
The Superfund Amendments and Reauthorization Act
(SARA) and the implementing executive orders under this
legislation require that contamination resulting from Depart-
ment of Defense (DOD) past operations be remediated. In
response to this legislation, the DOD has undertaken a com-
prehensive program to comply with these mandates. Over the
years this program has expanded from a $150 millon effort in
FY 1984 to a $1 billion effort in FY 1991. Some 17000 sites
have been identified at 1808 DOD Installations. Ninety DOD
Installations have been identified on the National Priorities
list by the Environmental Protection Agency. The detection
and remediation of contamination is a long term and resource
intensive effort. Research that allows us to proceed more
quickly in locating contaminants and in pin pointing key soil
and water samples for analysis, assessment, and remediation
purposes can provide a tremendous resource savings to the
ITR Program and, ultimately, the taxpayer. It is noted that
over 30% of the budget is estimated to be totally dedicated to
drilling, sampling and sample testing. Any improvement in
Field Sampling and Analysis will quickly repay the cost of
its associated research and development.
DOD Field Sampling and Analysis accomplishments include
the fielding of a truck-mounted cone penetrometer for more
efficient contaminant plume identification, tracking and
reducing well drilling requirements. Also completed was the
development of a field Analytical Method for the explosives
TNT and RDX in soil and water. Current program efforts
include the development of various contaminant sensors to
be employed in the cone penetrometer system to define
concentrations of contaminants in soil and groundwater as
the penetrometer is advanced through the soil. Future plans
include the concept of placing sampling devices into the
ground with the penetrometer which can be sampled and
analyzed with field instrumentation at regular intervals
thereafter. All these efforts have significant cost reduction
implications and have the interest and funding support of not
only DOD but also DOE.
15
-------
AN OVERVIEW OF ARMY SENSOR TECHNOLOGY APPLICABLE
TO FIELD SCREENING OF ENVIRONMENTAL POLLUTANTS
RAYMOND A. MACKAY
U.S. Army Chemical Researchf
Development and Engineering Center
Detection Directorate
ATTN: SMCCR-DDT
Aberdeen Proving Ground, MD 21010-5423
ABSTRACT
The Army has under development a number
of technologies directed toward the field
detection and Identification of chemical and
biological (CB) agents. This Includes not
only specific sensors, but the technology
required to Integrate these sensors Into
effective field detection systems. Much of
this technology can be adapted to materials
of environmental concern. In particular,
there are technologies 1n various stages of
development which are applicable to vapor
and aerosol clouds, as well as to
contaminated surface water and terrain.
These Include both point sampling and
monitoring systems, as well as remote sensing
systems capable of providing rapid wide area
coverage. This paper will provide an
overview of Army programs applicable to
field screening methods, with particular
emphasis on mass spectrometrlc. Infra red,
and aerosol sampling technologies.
the form of vapors or aerosols. The two
main areas which will be covered are
standoff detection and point detection.
Standoff detection has sometimes been
referred to as remote detection. However,
remote detection 1s defined here as the use
of point detectors which are located at the
site to be monitored, which may be of some
distance from the main monitoring station or
base, and connected to 1t by hard wire or
telemetry. Standoff detection refers to the
use of equipment located at the monitoring
base which can sense chemicals at a distant
location. The point detection technology to
be discussed 1n this paper Is pyrolysls-mass
spectrometry. There will also be some
discussion of aerosol sampling, since this
is pertinent to point detection of
aerosolized particulates, liquid or solid.
It 1s not the aim of this paper to present
detailed experimental results but rather to
provide an overview of the technology and
Its range of applicability.
INTRODUCTION
Technologies which can be utilized for
the detection of chemical warfare agents In
the field may also be applicable to the field
detection, classification and identification
of various substances of environmental
Interest. Although Army detection programs,
particularly those 1n the early stages of
development, focus on biological as well as
chemical detection, and much of the
technology 1s applicable to both. In this
paper, the emphasis will be on chemicals 1n
DISCUSSION
STANDOFF DETECTION: The U.S. Army Chemical
Research, Development and Engineering Center
(CRDEC) 1s currently engaged in an extensive
multi-year exploratory development program
to exploit laser radar for Chemical
Biological (CB) Standoff Detection. At
present, the only near term capability for
the detection of chemical agents at a
distance Is the use of passive infrared
sensors. These sensors can detect only
chemical vapors. Active (laser) infrared
17
-------
(IR) systems employing Differential
Scattering and Absorption L1dar (DISC/DIAL)
are being developed for the detection of
chemical agents 1n all physical forms:
vapors, aerosols, and rains, as well as
liquid surface contamination. In addition,
an ultraviolet (UV) system employing laser-
Induced fluorescence 1s being developed for
the detection of biological clouds
consisting of organisms, toxins and related
materials. The principles of operation of
these systems and the background of their
development will be briefly discussed. The
IR and UV breadboard systems have recently
been used 1n an extensive field test
employing various non-toxic chemicals and
Interferents with excellent results. These
data will be discussed along with the
necessary development efforts required to
adapt the DISC/DIAL technology to practical
field use.
The Army is making a significant
investment in standoff technology because it
1s the only technology known that can provide
rapid wide area surveillance capability
while simultaneously reducing the total
number of detectors required. At CRDEC
there are three phases to the Standoff
Detection program; the XM21 Passive Remote
Sensing Chemical Agent Alarm, along with
technology upgrades; the Laser Radar (LIDAR)
CB Standoff Detection System; and, for the
future, combining these technologies with
other electro-optic systems 1n integrated
sensor suites.
First to be discussed is chemical
detection portion of laser radar project
called IR DISC/DIAL. The objective is to
provide chemical laser Standoff detection
systems for CB defense applications. The
planned systems capabilities are to scan
surrounding atmosphere and terrain, operate
in fixed or mobile mode, detect chemical
contamination in all its physical forms, and
range resolve, quantify and map data. The
purposes of the current program are to
demonstrate concept feasibility, establish
capabilities and limits, complete science
base, determine effectiveness in field
situations and establish basis for rapid
transition to mature development. The IR
DISC/DIAL system can develop data in four
ways (as shown 1n Figure 1):
FIGURE 1
AGENT VAPOR
TOPOGRAPHICAL REFLECTION
(VAPOR)
DIFFERENTIAL ABSORPTION
(VAPOR)
AGENT VAPOR .
V.
NATURAL
AEROSOLS
DIFFERENTIAL SCATTERING
(AEROSOL/RAIN)
AGENT/RAIN/
AEROSOLS -
DIFFERENTIAL SCATTERING
(SURFACE CONTAMINATION)
SURFACE
CONTAMINATION
A0332-EE6 23400"
18
-------
Topographic reflection DIAL; By transmit-
ting different IR frequencies and detecting
their topographic return* chemical vapor
clouds can be Identified by their selective
absorption of some of the IR frequencies.
This measurement detects the presence of the
cloud and Its total concentration times path
length (CD; however, It does not tell how
far away the cloud is or Its density
(concentration).
Aerosol backscatter DIAL; By the same
technique, but with higher laser power, the
normally occurring atmospheric aerosol
begins to reflection IR energy back to the
detector. This distributed reflector can be
"range resolved" by gate timing the
returning signal just as radar systems do.
In this way» average concentrations and
ranges can be developed for many cells
(range lines) down the LIDAR path. By
scanning the system spatially, a map can
then be made of vapor chemical agents.
Agent backscatter DISC: In the same manner,
chemical agent aerosols and agent rains can
be detected by the selective frequencies
that they directly backscatter to the
detector.
Surface reflection; The fourth mode of
detection is the detection of selective IR
frequencies backscattered from agents on
surfaces. This measurement is dependent on
the amount of material located on the
surface of dirt, grass, trees or equipment.
FIGURE 2
Figure 2 shows that, for each of the
detection modes, the return signals are
different so that all measurements can be
made simultaneously. This is important
because there are no significant hardware
design constraints to add aerosol rain and
surface detection to an aerosol backscatter
DIAL system. The first objective of the
DISC/DIAL project was to build a Ground
Mobile Breadboard (GMB) system to demon-
strate the feasibility of DISC/DIAL chemical
detection. The system was mounted in a van
and tested. Based on these tests, the GMB
was upgraded. The current specifications of
the Ground Mobile Breadboard Upgrade (GMBU)
are given in table 1.
The GMBU along with other devices was
then exposed to extensive U.S. Army Dugway
Proving Ground (DPG) field testing. The
goals of these tests were:
(1) Investigate effects of reducing
system size, weight and power on detection
performances. This was because the Army's
near term use was a ground mobile vehicle
application for reconnaissance.
(2) Obtain quantifiable data on vapors,
aerosols, and liquid detection and on
interferences to prove feasibility.
(3) Use more realistic field scenarios
to develop workable use concepts.
BACKSCATTER FROM
"NORMAL" CLEAR AIR
AEROSOL CLOUD
VAPOR CLOUD
HARD TARGET
RETURN
TIME
19
-------
Four C02 TEA Laser
TABLE I. DISC/DIAL Specifications
Transmitter
Lasers
TunablHty Line-Tunable by Grating
Wavelengths 9.2 to 10.8 Microns
Energy (on 10P20) 2.0 J/Pulse
Pulse-to-Pulse ±.3 Percent
Power Stability
Pulsewldth (3dB) 90 ns
Repetition Rate 20 Hz
Beam Divergence 3.5x4.0 MRAD
Mode Multlmode or TEMoo
Timing Jitter 2 NS Pulse-to-Pulse
Receiver
Telescope Diameter 16 Inches
Detector HgCdTe Quadrant
Size 1x1 mm Per Element
Detectivity 4x10 cm/Hz1/2w
Field of View 8 MRAD
Overall Electronic 10 Hz to 7 MHz
Bandwidth
These tests Involved large scale
simulant clouds created by a special 100
meter long spray system as well as aircraft
spray. Also, aerosols were generated by
spray from a high ranger boom, and surfaces
(such as dirt, grass, concrete, trees, or
vehicles) were coated with simulants. The
many accomplishments of these large scale
tests were:
- Demonstrated feasibility of ISC/DIAL
technology
- Demonstrated high sensitivity
- Demonstrated operation 1n motion,
scanning and mapping
- Detected cloud through a cloud
- Detected collocated DMMP and SFg
- Detected DMMP (dimethyl
methyl phosphorate)
- up to 5 Km (range resolved)
- up to 10 Km (column-content)
- 1n presence of all Interferents
(fog, rain, dust and military
smokes)
- on ground by secondary vapor
- at night and 1n reduced
visibility
- 1n calibrated chamber
- Detected SF96 - as an aerosol
- as ground
contamination
on six surfaces
- Detected other volatile and non-
volatile simulants
- Validate emulation and simulation
models
Figure 3 shows a typical GMBU map of a
simulant vapor cloud. Although not evident
in this black and white Illustration, the
range cells are colored to show the average
concentration from 0.1 to 2.0 mg/nrr.
Additionally, this field work was backed
up with an extensive emulation and simula-
tion program which was able to show excel-
lent correlation between predicted and
actual performance. For example, the DMMP
and SFg 1 Km range resolved predicted and
measured values are Identical. Using this
excellent agreement, one can Infer the
following sensitivities to chemical vapors
with strong absorptions 1n the 9-10 micron
region of the Infrared.
Column Content
2 Km 10 Km
10 mg/m 12 mg/m2
Range Resolved
1 Km
0.5 mg/m3
The minimum detectable concentration of
liquid simulants on the ground were measured
at 0.5-5.0 g/m depending on the porosity of
the surface. Also very encouraging 1s the
20
-------
FIGURE 3 GMBU MAP
fact that one four wavelength set (1/20 sec
data) can provide a high amount of informa-
tion about the situation. An example:
between biological simulants and inter-
ferents/backgrounds of UV/LIF are below:
Accuracy of Prediction
(Range Over All Data)
97.2-100 percent
87.4-87.8 percent
66.2-74.1 percent
Information
1 simulant on any
1 of 5 surfaces
Scattering
Signal Level
248 nm
Fluorescence
Signal Level
280-410 nm
1 simulant on
of 5 surfaces
any 5
3 simulants on any
6 of 6 surfaces
This demonstrates that a real time surface
detection algorithm can be developed.
The UV LIF based laser radar was also
successfully tested at DPG for detection of
biological and toxin materials. While not
nearly as far along in development as the IR
system, this system demonstrated significant
detections at ranges up to 1.2 Km. The
system, which measures the laser induced
fluorescence of tryptophane, a compound
occurring in all living material, can sense
the presence of biological/toxin clouds but
cannot as yet uniquely identify the
material. Relative optical discrimination
None
None
None
Small
Small
Strong
Strong
Tryptophane
EG
Egg Albumen
Diesel Exhaust
Auto Exhaust
Road Dust
Trees
Strong
Strong
Strong
Strong
Weak
None
Strong
Other optical concepts based on Mueller
Matrix scattering are currently being
investigated to add additional identifica-
tion capabilities to UV/LIF system.
21
-------
Passive IR. The standoff detection and
Identification of chemical vapor clouds 1s
currently achieved by recording the IR
spectrum 1n the 8-12 micron wavelength
region by means of an Interferometer. This
Is the XM21 Remote Chemical Agent Sensing
Alarm. It Is a tripod-mounted device
weighing approximately 55 pounds, exclusive
of the source power. It scans a 1.5° field
of view (FOV) for 2 seconds, co-adding eight
scans. If the cloud fills the entire FOV,
the sensitivity Is on the order of a
concentration-path length product of 150
mg/m , the precise value depending upon the
strength of the absorption bands. The
interferogram, taken 1n the time domain, 1s
converted to a frequency domain spectrum In
the microprocessor by means of a fast
Fourier transform. A background spectrum of
the FOV must be obtained and stored, and
then subtracted from the sample scan prior
to further signal processing. Because of
the relatively slow scan speed, and the
requirement of the current algorithm for a
background subtract, 1t cannot be operated
from a moving platform.
A lightweight (20 Ibs), fast scan
interferometer is under development. In
addition, recent developments 1n direct
signal processing In the time domain have
both reduced demands on the microprocessor
and relieved the requirement for a
background scan. Since results equivalent
to those on the XM21 can be achieved in a
single scan without a pre-determined back-
ground spectrum, this device can be operated
from a moving platform such as a ground
vehicle or alrframe. Thus, if only vapor
detection 1s required, passive technology
represents an attractive method for rapid
survey of an area, particularly by air.
In summary, CRDEC has demonstrated the
feasibility of IR DISC/DIAL technology for
the detection of chemical agents in all
forms, as well as passive IR for chemical
vapor detection. Prototypes for ground
mobile, fixed site and test facility appli-
cation are beginning to be developed. The
potential exists for modifying these systems
to mount on helicopters, RPVs, and even
satellites, and to add the capability of
detecting biologicals and toxins, as well
as chemicals.
POINT DETECTION: There are two specific
technologies which form the basis of
recently fielded and developmental Army
point detectors; namely, Ion mobility and
mass spectrometery.
Ion Mobility Spectrometrv. This 1s a
technology which operates at atmospheric
pressure. The air sample containing the
vapor(s) to be detected are drawn through a
permselective membrane Into an ionizatlon
region where reagent gas Ions react with the
(polar) compounds to be detected and form
cluster ion species. These are gated Into a
drift tube where the ions migrate under an
applied electric field, and are separated
according to their mobility as measured by
their time of arrival at the collection at
the end of the drift tube. These may be
operated in both a positive and negative
mode. The U.S. Army currently has fielded a
hand-held monitor, the Chemical Agent
Monitor (CAM), and has a point alarm system
(XM22) under development. These relative
low weight, man portable, field hardened
devices are quite sensitive and should be
quite useful for field screening and
monitoring of a wide variety of
environmentally hazardous vapors. Since
this technology and Its applications will be
discussed extensively in the symposium, 1t
will not be considered further here.
Mass Spectrometrv. A mass spectrometer
system which can provide sensitive,
effectively real time detection and
identification of chemicals in the form of
vapors, aerosols, and ground surface
contamination, is currently under
development by CRDEC. Since this system
also has the potential to detect materials
of biological origin, it is referred to as
the Chemical Biological Mass Spectrometer
(CBMS).
The CBMS consists of two major
components, the biological probe and the
mass analyzer chassis. An artist's concept
is shown in figure 4. The biological
sampling probe contains the virtual impactor
and infrared pyrolyzer. The mass analyzer
chassis contains the mass analyzer,
instrument computer, data processing
computer and display, alarm and
communication modules.
The virtual Impactor block of the
biological sampling probe consists of a 1000
l/m1n pump and a four stage virtual Impacter
concentrator. This device separates the
aerosol particles from the air by virtue of
their inertia and directs them onto a quartz
wool matrix. The quartz wool 1s mounted
Inside of the infrared pyrolyzer assembly.
Periodically this assembly 1s heated to
22
-------
• 10 SAMPLER
CHEMICAL/BIOLOGICAL
MASS SPECTROMETER
Figure 4
temperatures near 600 C. As a result, any
biological material collected on the quartz
wool is pyrolyzed. Although the focus is on
biological aerosols, any aerosol particle in
the applicable size range will also be
collected and analyzed in the same way.
This includes liquid or solid chemical
aerosols, or chemicals adsorbed on or
attached to other aerosol particles of
network or anthromorphic origin. These
pyrolysis products are then drawn into a
heated 3 meter long, 1mm O.D. capillary
column and pulled to the mass analyzer
chassis. Any chemical vapors in the air are
also drawn into this capillary and pulled to
the mass analyzer.
The pyrolysis products and/or chemical
vapors enter the mass analyzer by permeating
through a silicone membrane. This membrane
separated the high vacuum mass analyzer from
the ambient pressure sample. After the
sample enters the mass analyzer, it is
ionized using an electron gun and the mass
spectra taken of the ionized components.
The instrument control computer controls
the mass analyzer, the pyrolysis event, and
all other instrument related functions
including temperature settings, electron gun
current, and rf/dc voltages and frequencies.
The data processing computer interprets the
mass spectra and generates the necessary
system responses. The display, alarm and
communications modules are the primary
interfaces to the operator. A block diagram
is shown 1n figure 5.
A QUISTOR (Quadrupole Ion Storage
Device) mass analyzer is used in the CBMS.
(Figure 6) This mass analyzer consists of
two end caps and a ring electrode. An ion
getter pump or molecular drag pump can be
used to produce the required vacuum. An
electron gun is mounted on the sample inlet
side. Selected masses are either trapped
within the QUISTOR or expelled out through
the end caps depending on the voltages and
frequencies applied to the caps and ring.
The masses of the ions that are expelled are
directly correlated to the voltages and
frequencies applied to the rings and caps.
In principle, a mass analysis is made as
follows. First a vapor sample enters the
QUISTOR. This sample is then ionized using
the electron gun. The voltages and
frequencies applied to the rings and end
caps cause these ions to become trapped
within the QUISTOR's Internal electric
fields. The dc voltage applied to the
QUISTOR 1s then changed at a controlled
rate. At specific voltages, certain masses
become unstable and are expelled from the
QUISTOR and are detected at the electron
multiplier. A plot is made of the signal
from the electron multiplier as a function
of the applied voltage. This voltage is
increased until all ions are expelled. The
final mass record is then obtained by
correlating the applied and plotted voltage
to the corresponding masses that should be
expelled.
23
-------
Figure 5
DISPLAY ALARM
AND
COMMUNICATIONS
Figure 6. QUISTOR Schematic
RING
ELECTRODE
IONIZATION
REGION
\ \
V
0
ION STORAGE
REGION ELECTRON
MULTIPLIER
24
-------
FIELD ANALYTICAL METHODS FOR SUPERFUND
Howard M. Fribush, Ph.D. and Joan F. Fisk
U.S. Environmental Protection Agency
Analytical Operations Branch (OS-230)
Washington, D.C. 20460
Abstract
The Analytical Operations Branch (AOB)
of the U.S. EPA is responsible for
coordinating field analytical methods
information transfer for Superfund. With the
assistance of the Environmental Monitoring
Systems Laboratory in Las Vegas (EMSL-LV) , AOB
has initiated a series of projects designed to
facilitate the appropriate use of field
analytical methods throughout Superfund. This
paper will summarize the use of field
analytical methods in the various phases of
Superfund activities, and will describe AOB
efforts in coordinating field analytical
methods information transfer throughout EPA.
In addition, this paper will summarize the
field analytical methods currently used
throughout EPA's Superfund program and
describe the development of comprehensive
document that will compile field analytical
methods and provide guidance to the use of
field analytical methods for environmental
samples.
Introduction
Field analytical methods have been widely
used for the past eight-to-ten years by EPA
organizations under various Superfund
contracts, such as Field Investigation Teams
(FIT), Technical Assistance Teams (TAT),
Emergency Response Cleanup Services
(ERGS), and Ranedial Engineering
Management (REM) contracts. As efforts
to streamline the Superfund site
assessment, site characterization, and
site clean-up processes have developed,
the need to assess field analytical
technologies for their appropriate use in
Superfund decision-making have increased.
The Analytical Operations Branch (AOB) of
the Hazardous Site Evaluation Division
(HSED) has been involved in coordinating
information on field analytical methods
used in support of Superfund. The AOB's
first efforts at coordinating field
analytical methods resulted in a paper
entitled 'Field Monitoring Methods in Use
for Superfund Analyses' (1) and the Field
Screening Methods Catalog (2).
Field analytical methods are used
throughout the Superfund process. In
EPA's Site Assessment, or Pre-remedial
Program, FIT teams under the direction of
EPA Site Assessment Managers (SAMs)
analyze samples in the field for Site
Inspections (SI). The results of the SI
determine whether a site should be added
to the National Priorities List (NPL) of
Superfund hazardous waste sites. In
EPA's Removal Program, a TAT team, under
the direction of an EPA On-Scene
Coordinator (OSC) , will conduct a Removal
25
-------
Assessment, often using field analytical
methods, to determine if an emergency response
(a removal action) is necessary. When a
Removal Action is initiated, an ERGS cleanup
contractor, under the direction of the OSC,
may be dispatched to the site for further
analysis and cleanup. The result of the
removal action is typically a short-term
stabilization of a site, and field analytical
methods are often used to monitor the extent
of the cleanup and determine when to stop the
removal action. In EPA's Remedial Program,
REM contractors, under the direction of
Remedial Project Managers (RPMs), have
conducted field analyses to characterize the
extent of contamination at a site (the
Remedial Investigation), to test remedial
treatment technologies (the Remedial Design),
and for site cleanup activities (Remedial
Actions). In all of these programs, field
analytical methods are often used to identify
critical samples for CLP confirmatory
analyses.
Field analytical methods are typically
not as rigorous as chemical analyses conducted
in a "fixed" laboratory - a laboratory in a
permanent location. Field methods are often
used for screening sites to determine if
contamination is present, and to obtain a
general idea of the extent of contamination.
Further, field analytical methods are most
useful when the contaminants of concern have
already been identified, so that the
appropriate methods, dilutions, calibration
ranges, etc., can be employed. In addition,
field analytical methods are usually designed
to identify only a limited number of analytes.
Recently, however, more sophisticated and more
rugged instrumentation have allowed for more
rigorous analyses in the field; consequently,
field analytical chemistry does not have to be
limited to screening. Even so, it is
generally believed that field analyses provide
less precision and accuracy than analyses
conducted in fixed laboratories. (It should
be noted, however, that despite this
perception, a focused gas chromatographic
analysis is likely to be better than a heavily
quality-controlled QC/MS screen.) In all of
the Superfund activities described in the
previous paragraph, field analyses are used
for the rapid turnaround of sample results.
These results are, in turn, used to expedite
site assessments for NPL listings or for
emergency removal actions, site
characterizations, and ultimate cleanup. Data
quality is not compromised, since field
analyses are usually conducted in conjunction
with confirmatory analyses, such as GC/MS
or ICP/MS analyses using EPA Contract
Laboratory Program (CLP) protocols.
Consequently, field analyses are often
used to identify samples for more
rigorous, CLP-type analyses.
Site Assessment Program
As part of determining whether a
site should be added to the NPL, the Site
Inspection (SI) attempts to make a
determination of "observed release".
This determination indicates that the
site is discharging contaminants into the
environment.
The Site Assessment Program conducts
up to ten percent of its analyses in the
field, and about 75 percent of the
samples are sent to the CLP for full scan
analysis. In the Site Assessment
Program, very little is usually known
about the site and its contaminants;
consequently, it is more cost effective
to use the CLP as a screen rather than
conduct extensive field analyses designed
for analyzing a limited number of target
compounds. Nevertheless, FIT, the Site
Assessment Program's primary contractors,
conduct a limited number of field
analyses to obtain real-time data to
determine worker safety requirements, the
extent of contamination, the presence or
absence of contamination, for the
placement of monitoring wells, and to
select samples for subsequent CLP
confirmatory analysis.
To accomplish these analyses, EPA's
Site Assessment Branch has developed the
Field Analytical Support Project (FASP).
This project has, at this writing,
developed 31 field analytical methods,
called FASP Standard Operating Guidelines
(SOGs), and are designed to be modified
as needed to meet site-specific
conditions (3). These rapid turnaround,
FASP SOGs have been developed by FIT for
water, soil, or oil analyses for
volatiles, polynuclear aromatic
hydrocarbons, pesticides, PCBs, and
metals.
Some EPA Regions have used FASP to
perform preliminary evaluations of new
instrumentation. For example, two
Regions are evaluating Long Path Fourier
Transform Infrared (FTIR) Spectroscopy
26
-------
for the analysis of air samples remote from a
site, and one Region has evaluated the Thermal
Chromatography/Mass Spectrometry (TC/MS)
system for the analysis of solid samples.
According to these latter studies, TC/MS shows
promise as a rapid screen for solid samples
since there is minimal sample preparation.
Ranedial Program
The purpose of the Remedial Program is
to clean, or remediate, a site. This process
can be rather complex, and usually consists of
a Remedial Investigation (RI) phase, a
Feasibility Study (FS), a Record of Decision
(ROD), a treatability study, a Remedial Design
(RD) phase, and a Remedial Action (RA) phase.
The RI consists of data collection activities
undertaken to determine the degree and extent
of contamination within all media. The RI
supports the FS, which determines the risk
that the site poses to human health and the
environment, and identifies the most
appropriate remedial alternatives that can be
used to remediate the site. The ROD is issued
by EPA as the final remedial action plan for
a site. If necessary, a treatability study is
performed to determine the most appropriate
conditions for treatment, the remedy is then
designed (RD), and the site is cleaned (RA).
During all of these phases, the
potential exists for the use of field
analyses. For example, during the three-
dimensional characterization of the extent of
contamination (the RI),
rapid turnaround of sample results may be
necessary to focus subsequent analyses to the
determination of the extent of contamination.
Here, the analyses may be used to optimize
sampling grids for three-dimensional site
characterizations, to determine the location
of monitoring wells and well screen depths, or
to determine the direction and speed of
groundwater plumes. During treatability
studies, rapid turnaround of data may be
necessary to avoid shutting down a treatment
operation to wait for sample results. In the
Remedial Design phase of the remediation,
rapid turnaround of sample results may be
necessary to evaluate the efficiency of a
design. These data may then be used to make
improvements on the design, the net result
being more rapid development of remedial
designs. In removal and remedial actions,
rapid turnaround of data may be required to
determine cleanup levels and to minimize the
costs associated with using expensive cleanup
equipment such as bulldozers. When the field
analyses suggest that a regulatory level
has been reached, CLP confirmatory
analyses can then be performed to confirm
the cleanup level reached.
To accomplish these analyses, EPA's
Hazardous Site Control Division developed
the Close Support Laboratory (CSL)
Program. Because site remediations are
often very complex and typically take
several years to complete, the REM
contractors found it more convenient to
construct temporary, "close-support"
laboratories at the site rather than use
mobile laboratories or portable
instruments for the analytical
investigations. This program has
resulted in the development of 15 field
analytical methods for metals, volatiles,
semivolatiles, and polynuclear aromatic
hydrocarbons in water and soil matrices
(4) . In addition, the CSL program has
developed 16 field protocols for the
determination of physical measurements to
be used during treatability studies.
The Remedial Program conducts about
ten percent of its analyses in the field.
Once EPA has placed the site on the NPL,
Potentially Responsible Parties (PRPs)
are finding that it is more cost-
effective to assume the costs of site
characterizations. Consequently, there
are a growing number of these "PRP-Lead"
sites, requiring fewer analyses by the
EPA. As a result, in many Regions the
Remedial Program is placing increasingly
more resources on overseeing the
analytical activities of the PRPs. This
shifting of focus from "Superfund-Lead"
sites to PRP oversight has also coincided
with the phasing out of the REM contracts
and phasing in of the new Alternative
Remedial Contracts Strategy (ARCS)
contracts. Nevertheless, there are still
many Superfund-Lead remediations in
progress, and the Remedial Program is
planning to use ARCS contractors to
perform analyses in the field.
Removal Program
In addition to the long-term
remedial actions, Superfund legislation
provides for short-term, removal actions.
Removals are performed in emergency-type
situations on unstable sites. A removal
is the cleanup or removal of released
hazardous substances which may present an
27
-------
imminent and substantial danger.
Consequently, removals may be necessary in the
event of a release of hazardous substances, or
to monitor, assess, and evaluate the threat of
release of hazardous substances to prevent,
minimize, or mitigate damage to human health
or the environment.
Due to the nature of these activities,
removals often require a rapid turnaround of
analytical data; consequently, field analyses
are used quite often. The Removal Program
conducts about 30 percent of its analyses in
the field. Under the direction of the OSC,
TAT - the Removal Program's primary technical
contractor - may use field analytical methods
for purposes similar to those of the FIT
teams. If a more in-depth study is required,
the OSC may require the use of field
analytical methods to determine an estimated
extent of contamination. If drums are present
and the contents within the drums are unknown,
TAT may use a Hazard Categorization field kit
to categorize the potential hazard associated
with the contents of the drums. TAT uses this
field kit to perform simple qualitative tests
to determine gross characteristics of the
waste - the compound class, flash point and
other properties, and consequently, determine
the disposal options for the waste.
The Removal Program uses field analyses for
Classic Emergencies (for example, for fires,
spills, train derailments, and explosions), to
determine worker safety requirements, for
designing sampling grids, to estimate
exposure, for monitoring well placement, and
to determine cleanup levels. Across all
programs, the reasons for using field analyses
are for time savings, cost savings, and to
identify critical samples for confirmatory
analyses. Other reasons include being able to
take more samples, ease of acquisition, and
minimal paperwork requirements.
To accomplish these analyses, the
Removal Program established the Environmental
Response Team (ERT). The ERT provides
expertise to the OSCs in the area of
performing field analyses and field analytical
methods development. The ERT has developed a
number of field analytical methods, including
portable gas chromatography methods, x-ray
fluorescence methods for metals, and methods
for the screening and analysis of air samples
(5).
EMSL-LV
The Environmental Monitoring Systems
Laboratory in Las Vegas (EMSL-LV)
supports the Superfund field analytical
programs through both research and
development and through technical support
to the EPA regions. In the Advanced
Field Monitoring Methods Program (AFMMP),
EMSL-LV is developing and validating
field analytical methods. In its
Technical Support Program, EMSL-LV
dispatches field analytical teams to
hazardous waste sites for
characterization studies.
EMSL-LV is working under its
Advanced Field Monitoring Methods Program
(AFMMP) in coordination with the
Analytical Operations Branch (AOB) to
identify, develop, and validate new and
existing field analytical methods and
instrumentation. In addition, the
objectives of AFMMP include the transfer
to and exchange of information with the
EPA regions. EMSL-LV has performed
studies involving immunochemical methods,
soil gas techniques, portable gas
chrcmatographs and associated analytical
methods, X-ray fluorescence, and fiber
optic sensors. In addition, EMSL-LV has
identified a number of new techniques for
study, including Fourier Transform Infra-
Red (FT-IR) , portable supercritical fluid
extractor and solid phase extraction,
field test kits, portable GC/MS, ion
mobility spectrometers, and luminescence
methods.
Development of a Superfund Field
Analytical Methods Catalog
The Analytical Operations Branch
(AOB) is the focal point for coordinating
field analytical method information
transfer for Superfund. In 1988, the AOB
coordinated an effort to compile some of
the field analytical methods used in
Superfund into a document entitled "Field
Screening Methods Catalog".
The AOB is currently designing and
developing a comprehensive compendium
that will contain many of the field
analytical methods described in this
paper for use by all persons involved
with Superfund field analyses. This
compendium will contain developed field
analytical methods, it will contain
instrumentation requirements,
requirements for quality assurance and
quality control, analytical method
28
-------
performance, guidelines for effective
coinnunication, health and safety guidelines,
and evidentiary guidelines. This compendium
is being prepared with the assistance of the
Field Analytical Methods Workgroup, which had
its first meeting on July 19-20, 1990 and the
Field Analytical Methods Management Forum.
The forum is a group of EPA Headquarters and
Regional management representatives who met on
June 27-28 to determine Superfund policies
regarding field analyses in Superfund.
The field analytical methods that will
be a part of the catalog will come from the
sources described in this paper. The methods
will be presented in chapters structured by
fraction, analyte group, and media. In
addition, the methods will be restyled into
SW-846 format for consistency, ease of
reading, and to allow for variations.
Instrumentation requirements will be provided
for each type of method based on available
information and research by EMSL-LV. Quality
Assurance and quality control information will
be designed to facilitate a rapid turnaround
of data appropriate for the generation of
field analytical data, and will be tiered to
allow a variation of requirements for quality.
The compendium will contain a user's guide and
will stress "interactive management" - the
communication between the site manager, the
field analyst, and the sampler. In addition,
an electronic bulletin board will be
established to house the methods for
downloading, facilitate the quick transfer of
technology, information, and ideas. Health
and safety guidelines will be established
based on recent OSHA regulations, and evidence
guidelines for samples and analyses will also
be addressed.
References
1. Fisk, Joan F. Field Monitoring Methods in
Use for Superfund Analyses. Pittsburgh
Conference. February, 1987.
2. Office of Emergency and Remedial
Response, Hazardous Site Evaluation Division.
Field Screening Methods Catalog. Users Guide.
EPA/540/2-88/005. U.S. EPA. Washington, DC.
September, 1988.
3. Site Assessment Branch. U.S.E.P.A.
Hazardous Site Evaluation Division. Field
Analytical Support Project Standard Operating
Guidelines. Unpublished. Washington, DC.
July, 1990.
4. Hazardous Site Control Division.
Compilation of CSL Analytical Methods.
Unpublished. U.S. EPA. Washington, DC.
5. Environmental Response Team. Quality
Assurance Technical Information Bulletin,
Standard Operating Procedures.
Unpublished. U.S. EPA. Edison, NJ.
29
-------
FIELD DELINEATION OF SOILS CONTAMINATION ON HAZARDOUS WASTE SITES
REGULATED UNDER NEW JERSEY'S HAZARDOUS WASTE PROGRAM
Frederick W. Cornell
New Jersey Department of Environmental Protection
Division of Hazardous Site Mitigation
Bureau of Environmental Evaluation and Risk Assessment
401 East State Street, Floor 6W
Trenton, NJ 08625-0413
ABSTRACT
The New Jersey Hazardous Waste
Management Program (HWMP) recognizes
the potential for field analysis
techniques to expedite site
delineation while decreasing site
characterization costs. Although,
field analysis methods produce
accurate, real-time data at a low cost
per sample, the absence of
standardized data quality objectives
and method specific quality assurance
and quality control (QA/QC)
requirements has prevented widespread
use of these technologies. The HWMP
has defined data quality objectives
for each phase of site investigation,
and outlined QA/QC procedures for
several widely available field
analysis methods, including field
x-ray fluorescence spectrometry, field
gas chromatography, colormetric
analysis, and photoionization
surveying. The development of these
method-specific and use-specific
procedures has allowed the HWMP to
routinely recommend the use of field
analysis methods to expedite site
evaluation.
INTRODUCTION
The New Jersey Environmental
Cleanup Responsibility . Act (ECRA)
program requires industrial facilities
that handle hazardous materials to
conduct a site evaluation and develop
a site remediation plan (if necessary)
prior to any real estate transfer or
cessation of industrial operations.
Given the real estate and stock market
activity of recent years it is not
surprising that ECRA subject sites are
often operational facilities. Since
ECRA's enactment in 1984, thousands of
sites have been processed by ECRA.
For larger industrial facilities, site
evaluation has proven to be costly and
time consuming, frequently taking
several years to complete.
Site characterization efforts
typically involve a historical site
survey, site screening, and several
phases of site delineation (1).
Although, initial site screening is
usually conducted using survey
instruments, the remaining delineation
phases generally involve collecting a
limited number of samples for
laboratory analysis and evaluating the
lab results to determine the need for
additional sampling phases. This
typical investigation scheme is time
consuming, requiring months between
phases to allow for sample collection,
data analysis, delineation plan
development, and regulatory review and
interface; however, the phased
approach is necessary to limit
analytical costs. The unfortunate
result of phased investigation is that
remedial investigations frequently
last years and cost hundreds of
thousands of dollars.
These delays in site remediation
may not only render industrial
operations or property transfers
difficult or impossible to conduct,
but also may cause unnecessary
contaminant migration and exposure to
31
-------
human or environmental receptors. In
these situations it is desirable to
implement analytical methods that can
provide the necessary data in a timely
and cost-effective manner. Field
analysis is ideally suited for rapid,
cost-effective site characterization
as it can provide real-time data which
is reliable and inexpensive on a per
sample basis.
FIELD SCREENING AND ANALYSIS METHODS
To develop the standard operating
procedures (SOPs) included in this
paper, efforts were initially directed
at determining the minimum data
quality necessary to make appropriate
technical and regulatory decisions
(this is described in further detail
below). Subsequently, a literature
search was used to identify reliable
methods from the vast number of
commercially available technologies.
Using this information, method
specific SOPs were developed to detail
the minimum requirements a field
delineation plan must meet to receive
agency approval. These SOPs are
designed to encourage the generation
of consistent and reliable data from
user to user and site to site. The
quality assurance and quality control
(QA/QC) requirements in each SOP were
formulated in consistency with the
reliability, accuracy, and limitations
of each method (particularly when
considering field use), while
considering the ultimate use of the
resulting data.
Several instruments and methods
have been evaluated and determined
effective (or potentially effective)
at detecting site contamination at
milligram per kilogram (mg/kg)
concentrations. Although, a single
instrument or method may only be
useful for analyzing one or two
classes of compounds, the use of
several field analysis procedures in
tandem enables site investigation
teams to detect most priority
pollutant compounds at or near
background concentrations. For
example, ambient temperature headspace
analysis is extremely effective in
analyzing volatile organic compounds,
but not polyaromatic hydrocarbons
(PAHs) or metals. Color-metric tests,
on the other hand, are effective at
analyzing aromatic compounds
(including the PAHs), and a field XRF
will detect PCBs and most metals at
concentrations as low as 20-100
milligrams per kilogram. Thus, by
using several field instruments or
methods in tandem a broader suite of
contaminant compounds may be field
analyzed. It should be noted that the
methods cited in this paper are by no
means a comprehensive list of suitable
or potentially suitable field
methodologies. Initial selection for
these SOPs was based on instrument
availability, amenability to field
use, and in-house experience.
DATA QUALITY OBJECTIVES
The New Jersey Hazardous Waste
Management program (HWMP) data quality
designations are based on those
developed by the EPA (2-3) . The EPA
has established five levels of data
quality objectives (DQOs). Two of
these, Level 1 (Field Survey
Instruments) and Level 2 (Field
Portable Instruments), generate real-
time, field data. Level 3 and 4 are
laboratory methods with differing
QA/QC requirements, and level 5 is
laboratory special services. The EPA
has clearly stated the minimum data
quality level required for each stage
of site investigation. Additional
explanation of these data quality
levels may be found in any of the
EPA's Data Quality Objectives manuals,
cited above.
The HWMP data quality standards
have been developed to encourage the
use of real-time analysis methods
during site characterization (4). The
HWMP field data DQOs are: Level 1
(Field Survey Instruments), Level 1A
(Field Analytical Methods), and Level
2 (Field Portable Instruments).
However, unlike the EPA designations,
minimum QA/QC and support documen-
tation (deliverables) requirements are
defined to assure that the data
generated by these methods can be
validated based on technical criteria.
A detailed description of all DQO
levels is provided below.
Data Quality Level 1 instrumen-
tation are intended primarily for
health and safety or initial site
screening. Quality control and
deliverable requirements are limited
to a continuing calibration for
site-specific compounds and the
reporting of values on field/boring
logs. Level one (1) methods are
real-time and at times, erratic.
These methods can be described as
pseudo-qualitative and pseudo-
32
-------
quantitative as the end user can
easily be led to believe that these
instruments are reporting "true
values" or providing selectivity, when
indeed they are not. For example, the
photoionization detector (PID) survey
instrument is commonly thought to be
selective and not sensitive to species
whose ionization potentials (IPs) are
higher than that of the internal
ionization lamp. In practice,
however, species with IPs above the
lamp energy are routinely detected by
PID survey instruments. With respect
to quant itat ion, a PID survey
instrument reports a value often
expressed in mg/kg; however, since
detector response is highly variable
among chemical species this reported
value may not represent site
conditions or correlate with other
site data. For these reasons level 1
data should generally be used to
indicate contaminant presence or
absence, rather than compound identity
or total concentration. The
application of level 1 data should
therefore be limited to health and
safety screening or to guide the
placement of samples being analyzed by
higher DQO methods. Level 1
instruments include field x-ray
fluorescence spectrometers (XRF) with
a remote probe and PID survey
instruments.
Data Quality Level 1A methods
produce fairly precise data; however,
a reduced quality control program is
employed to allow high frequency,
low-cost sampling. Level 1A methods
are suitable for site screening and
site delineation when proper QA/QC
practices are employed. When
delineating using level 1A methods,
minimum deliverable requirements
typically include: calibration data
for site-specific compounds, check
standards data, a non-conformance
summary, a certification statement
signed by the analyst, sample
calculations, isopleth maps, tables
indicating results (raw and
"corrected" based on lab confirmation
data), and chain-of-custody documen-
tation. In addition, lab confirmation
data (10-30% of all samples collected)
must provide "calibration" throughout
the entire analysis range and
confirmation of the "clean" zone.
Level 1A methods include headspace
analysis of volatile compounds and
analysis using colormetric techniques.
Data Quality Level 2 methods
produce precise data when required
QA/QC procedures are employed.
Quality assurance and quality control
requirements are sufficient to allow
rigorous data interpretation, while
providing reasonable field operation
requirements. Level 2 methods are
ideally suited for low-cost, one phase
delineation. Minimum deliverables
requirements will include: an
instrument log, calibration data for
site specific compounds, standards
data, split sample data, raw sample
data, blank data, a certification
statement signed by the analyst, a
non-conformance summary, sample
calculations, isopleth maps, tables
indicating results (raw and
"corrected" based on lab confirmation
data), and custody documentation. Lab
confirmation data (5-15% of all
samples collected) must provide
"calibration" throughout the entire
analysis range and confirmation of the
"clean" zone. Level 2 methods include
field gas chromatography (GC) and
field XRF analysis using a
silicon-lithium detector.
Data Quality Levels 3 and 4 are
"Standard Lab Methods" with varying
deliverable requirements. Methods
which provide these data qualities may
be used for conventional site
characterization activities or to
confirm field instrument results
obtained during site delineation
activities. It should be noted that
the specific QA/QC procedures required
will be dictated by the applicable
regulatory program. Data quality
level 3 methodologies include SW-846
(5) methods and NJ ECRA Deliverables
(1). Data quality level 4 methods
include CLP methods and Scope of Work
(SOW) requirements (6).
Data Quality Level 5 methods are
generally state-of-the-art or non-
approved methods chosen specifically
for a particular site. Level 5
methods are required when "Standards
Lab Methods" are either unavailable or
impractical. Level 5 data may be
accepted to confirm field results or
define a "clean zone".
The goal of any site investigation
is to assure that the information
obtained is sufficient to select and
design an appropriate remedial
technology. Ideally, site character-
ization will provide complete
definition of contamination with
respect to both concentration trends
and actual contaminant load. The
advantages of levels 1, 1A, and 2
33
-------
analysis are rapid site delineation
and low per sample costs allowing high
frequency sampling and a rapid
estimation of concentration gradients;
however, the concentration results
must be assumed to have up to a 150%
error. Level 3 and 4 analysis methods
are not real-time and are more
expensive, limiting sampling
frequency, but reported results can be
assumed to be quite accurate and a
good indicator of actual contamination
present. In summary, the trade-off is
rapid, less expensive site
characterization verses data quality
and accuracy.
At first glance it may appear as
if HWMP has chosen to expedite site
characterization at the expense of
data quality by encouraging the use of
level 1A and level 2 methods. Upon
closer examination, it can be seen
that although the raw data obtained by
field instruments are less accurate
and less precise, the data set is
highly consistent within itself,
clearly indicating trends and
contamination zones. Also, since
field analysis costs are generally per
diem rather than per sample, field
samples may be collected at a greater
frequency, providing the project team
with better site definition and fewer
data gaps. Lastly, all field data are
supported by an independent
calibration or correction factor
provided by the required lab
confirmation samples, discussed above.
Thus, the end product generated is
actually a hybrid of field analysis
data and lab data which, when
combined, may not only be equivalent
in data quality to that obtained by
standard methods, but may actually
provide a more reliable and complete
characterization of site conditions.
SITE INVESTIGATION STRATEGY
The newly developed HWMP DQOs use
a combination of high and low quality
data to produce a data set which is
moderate in both quality and quantity.
These DQOs rely on the ability of
users to calibrate field analysis data
to laboratory confirmation samples,
providing superior site character-
ization at a reduced cost. The net
effect is that most site
investigations may be completed in a
maximum of 1-2 phases or less than one
(1) year. To accomplish this, the
following site investigatory procedure
is recommended (where site contam-
ination is known, step 1A may not be
required).
1. Obtain historical information
(i.e. past or present site
activities).
1A. If the contamination source
is unknown, a sampling
program incorporating site
screening tools (level 1) and
laboratory sample analysis
(level 3/4) should be
implemented. The goal of
this effort is to identify
all contaminants present by
documenting worst-case site
conditions.
2. The information above should
then be used to develop an
open ended, contaminant
delineation plan, including
the use of real-time (Level
1A/2 quality data) methods.
The plan should incorporate
sampling contingencies to
assure site delineation is
completed during this sampling
phase. To provide additional
data reliability, field
instruments should be
calibrated to site-specific
compounds of interest as
defined by previously obtained
information.
3. Upon receipt of the laboratory
confirmation data, the need
for a revised delineation plan
should be assessed. If
required, a phase II delin-
eation plan should incorporate
field analysis methods to
complete site delineation.
4. The complete database should
then be used to develop a site
remediation plan. If in situ
remedial measures are to be
used and system design limits
are being approached, an
increased percentage of
laboratory data may be
required.
DEVELOPMENT OF FIELD SOPS
The development of field SOPs is
considered the most efficient means of
assuring that data collected from site
to site is consistent. These SOPs
were developed by consulting the
literature, instrument manufacturers,
and personnel with extensive field
34
-------
and/or instrumental experience. Each
SOP has 5 technical sections, i.e.
method overview, method requirements
(including QA/QC requirements),
interferences and limitations, data
interpretation and reporting require-
ments, and health and safety
considerations.
The method overview or general
guidance section is intended to
provide the reader with a basic
understanding of the method. This
section details method applications,
including applicable matrices,
detectable compounds, and minimum
detection limits (MDLs). Additional
information is provided for use by the
project manager, including estimated
cost per sample, level of training
required to effectively use the
method, lab method equivalent, and
theory of operation. The theory
section contains instrumental and/or
chemical details aimed at
familiarizing the reader with the
actual science of operation. The last
section of each SOP also provides a
list of references directing
interested readers to a more detailed
explanation of instrumental theory and
use.
The method requirements section
provides four types of information:
sampling considerations, sampling
requirements, field operation require-
ments, and QA/QC requirements.
Sampling considerations include
general information applicable when a
sampling program is being developed.
This section provides guidance with
respect to sample frequency, selection
of lab confirmation samples, and any
other useful information gained
through field experience. As would be
expected, this section is continually
evolving as the experience base grows.
The sampling requirements section
details proper sample collection
procedures when standard field
sampling methods (7) are inappro-
priate. This section also includes
sample handling requirements when past
experience has shown sample
preparation to significantly impact
final results, as is the case with XRF
analysis. The field operation section
contains actual method guidance
intended to supplement or replace
manufacturer's recommendations. This
guidance customizes method procedures
in an effort to meet the goals of the
HWMP regulatory program. The last
section, QA/QC, states all quality
assurance recommendations and require-
ments. The requirements include
analyst "competence" tests, submission
of all raw data, and support
documentation.
The interferences and limitations
section discusses problems which may
be encountered during field use.
These comments are intended to
supplement manufacturer's recommen-
dations by highlighting problems
encountered during previous site
operations. It is likely that this
section will be in constant transition
until a comprehensive database has
been established.
The data interpretation and
submission requirements section
details data manipulation procedures
and regulatory submission require-
ments. Data interpretation require-
ments vary by method and DQO level;
however, all SOPs require the
calculation of "corrected" results,
accounting for discrepancies between
laboratory and field data. Reporting
requirements are standardized for all
field methods and include: scaled
site maps with plotted data, summary
tables indicating all field results
(raw and corrected) and lab reported
values, a calibration plot of lab
split sample data verses field data,
and quality assurance and quality
control documentation (consistent with
the QA/QC requirements stated above).
These requirements are intended to
expedite the required review time by
standardizing report contents and
format, while facilitating validation
of both lab and field data.
CURRENT AND PENDING SOPs
Standard operating procedures have
been completed for four field
instruments and two field analysis
methodologies. Additionally, several
other SOPs are under development. A
listing of all SOPs is provided below.
Level 1 Data Quality
Field Screening Using a Photo-
ionization Survey Instrument.
Field Screening Using an X-ray
Fluorescence Spectrometer
Equipped with a Remote Probe.
*Field Screening Using a Flame
lonization Survey Instrument.
*Field Screening Using a Portable
Infrared Instrument.
35
-------
Level 1A Data Quality
Field Delineation Using a
Colonnetric Test Kit.
Field Delineation Using Ambient
Temperature Headspace Analysis.
*Field Delineation Using a Portable
Infrared Instrument.
*Field Delineation Using a Portable
Ultraviolet Spectrometer.
Level 2 Data Quality
Field Delineation
Fluorescence.
Field Analysis Using a
Chromatograph.
Attachments: 1
2
3
4
*5
Using X-ray
Field Gas
*6,
*7,
PID Detector.
FID Detector.
AID Detector.
ECD Detector.
Analyzing
Extractables
(BNs/PCBs).
Analyzing Water
Samples.
Analyzing Air or
Headspace
Samples.
* - under development
FUTURE DIRECTIONS
Currently, the field SOPs
described above are in widespread use
throughout the HWMP program. Since
these instruments and methods are a
small subset of all currently
available field analysis methods,
similar SOPs will be developed for
several additional methods, including
FID survey instruments, several
spectrometers, and additional field
gas chromatography applications.
The performance of each of these
methods (on NJ regulated sites) will
be monitored using an in-house
database. Upon collection of
sufficient data, the SOPs will be
revised as appropriate. It is
expected that additional field
experience and the associated
understanding of method limitations
and accuracy will lead to wider use of
field analysis methods, making site
evaluation a much less time-consuming
and costly process.
5.
6.
REFERENCES
New Jersey Department of
Environmental Protection (NJDEP) ,
Division of Hazardous Waste
Management (DHWM). March 1990.
Remedial Investigation Guide.
U.S. EPA. March 1987. Data
Quality Objectives for Remedial
Response Activities.
EPA/540/G-87/003, EPA/540/G-87/004
and OSWER Directive 9335.0-7A&B.
U.S. EPA. October 1988. Guidance
for Conducting Remedial
Investigations and Feasibility
Studies Under CERCLA.
EPA/540/G-89/004 and OSWER
Directive 9355.3-01.
NJDEP, Division of Hazardous Site
Mitigation (DHSM), Standard
Operating Procedures: Field
Delineation Series (4.25), 1990.
U.S. EPA. SW-846, Third Edition.
CLP-IFB, most recent
U.S. EPA.
version.
NJDEP/DHWM. February 1988. Field
Sampling Procedures Manual.
BIBLIOGRAPHY
General Field Sampling and Analysis
Myers, J.C. "Converging Technologies",
Hazmat World, June 1989, p24-27.
U.S. EPA. Environmental Response
Team. 1989. Standard Operating
Procedure: Photoionization Detector
(12056).
Siegrist, R.L.; Jenssen, P.O.
"Evaluation of Sampling Method Effects
on Volatile Organic Compound
Measurements in Contaminated Soils",
Environmental Science and Technology,
1990, 24, 1387-1392.
1988. Field
Catalog.
U.S. EPA. September
Screening Methods
EPA/540/2-88/0055.
Keith, L.H. "Environmental Sampling:
A Summary", Environmental Science and
Technology, 1990, 24, p610-617.
Gretsky, P.; Barbour, R.; Asimenios,
G.S. Pollution Engineering, June 1990,
p!02-108.
36
-------
Organic Survey Instruments
Nyguist, J.E.; Wilson, D.L. "Decreased
Sensitivity of Photoionization
Detector Total Organic Vapor Detectors
in the Presence of Methane", Journal
of the American Industrial Hygiene
Association, 1990, 51(6), p326-330.
Gervasio, R. ; Davis, N.O. "Monitoring
in Reduced Oxygen Atmosphere Using
Portable Survey Direct Reading
Instruments (PID and FID)",
Proceedings HMCRI, 1989-90.
Tillman, N.; Ranlet, K.; Meyer, T.J.
"Soil Gas Surveys: Part I", Pollution
Engineering, July 1989, p86-89.
Tillman, N.; Ranlet, K. ; Meyer, T.J.
"Soil Gas Surveys: Part II",
Pollution Engineering, August 1989,
p79-84.
Headspace Analysis
Holbrook, Tim "Hydrocarbon Vapor Plume
Definition Using Ambient Temperature
Headspace Analysis", Proceedings of
the NWWA/API Conference on Petroleum
Hydrocarbons and Organic Chemicals in
Ground Water - Prevention, Detection,
and Restoration, November, 1987.
Roe, V.D.; Lacy, M.J.; Stuart, J.D.
"Manual Headspace Method to Analyze
for the Volatile Aromatics of Gasoline
in Groundwater and Soil Samples",
Analytical Chemistry, 1989, 61,
P2584-5.
Colormetric Analysis
Roberts, R.M., Khalaf, A.A.,
Friedel-Crafts Alkylation Chemistry; A
Century of Discovery, Macel Dekker,
Inc., New York, 1984.
Rohriner, R.L., Fuson, R.C. et al.,
The Systematic Identification of
Organic Compounds, John Wiley and
Sons, New York, 1980.
Hanby, J.D., "A New Method for the
Determination and Measurement of
Aromatic Compounds in Water", Written
Communication, Hanby Analytical
Laboratories, Inc., Houston, Texas,
1989.
X-ray Fluorescence
Piorek, Stanislaw "XRF Technique as a
Method of Choice for On-site Analysis
of Soil Contaminants And Waste
Material", Proceedings 38th Annual
Conference on Applications of X-ray
Analysis, Denver, Vol. 33, 1988.
Grupp, D.J.; Everitt, D.A.; Beth,
R.J.; Spear, R. "Use of a
Transportable XRF Spectrometer for
On-Site Analysis of Hg in Soils", AFI,
November 1989, p33-40.
J.R. Rhodes, J.A. Stout, J.S.
Schindler, and Piorek "Portable X-Ray
Survey Meters for In Situ Trace
Element Monitoring of Air
Particulate", American Society for
Testing and Materials, Special
Technical Publication 786, 1982,
p70-82.
Piorek, S.; Rhodes, J.R. "Hazardous
Waste Screening Using A Portable X-ray
Analyzer", Presented at the Symposium
on ' Waste Minimization and
Environmental Programs within D.O.D.,
American Defense Preparedness Assoc.,
April 1987.
Piorek, S.; Rhodes, J.R. "A New
Calibration Technique for X-ray
Analyzers Used in Hazardous Waste
Screening", Proceedings 5th National
RCRA/Superfund Conference, April 1988.
Piorek, S.; Rhodes, J.R. "In Situ
Analysis of Waste Water Using Portable
Preconcentration Techniques and a
Portable XRF Analyzer", Presented at
Electron Microscopy and X-ray
Applications to Environmental and
Occupational Health Analysis
Symposium, Pennsylvania State
University, October 1980.
Barish, J.J.; Jones, R.R.; Raab, G.A.;
Pasmore, J.R. "The Application of
X-ray Fluorescence Technology in the
Creation of Site Comparison Samples
and in the Design of Hazardous Waste
Treatability Studies", First
International Symposium: Field
Screening Methods for Hazardous Waste
Site Investigations, October 1988.
Piorek, S. "XRF Technique as a Method
of Choice for On-Site Analysis of Soil
Contaminants and Waste Material", 38th
Annual Denver X-Ray Conference, 1989.
Watson, W.; Walsh, J.P.; Glynn, B.
"On-Site X-Ray Fluorescence
Spectrometry Mapping of Metal
Contaminants in Soils at Superfund
37
-------
Sites", American Laboratory, July
1989, p60-68.
Freiburg, C.; Molepo, J.M.; Sansoni
"Comparative Determination of Lead in
Soils by X-Ray Fluorescence, Atomic
Absorption Spectrometry, and Atomic
Emission Spectrometry", Fresenius Z
Anal Chem, 1987, 327, p304-308.
Smith, G.H.; Lloyd, O.L. "Patterns of
Metals Pollution In Soils: A
Comparison of the Values Obtained By
Atomic Absorption Spectrophotometry
and X-Ray Fluorescence", Environmental
Toxicology and Chemistry, 1986, Vol.5,
P117-127.
Jenkins, R "X-Ray Fluorescence
Analysis", Analytical Chemistry, 1984,
56(9), p!099A.
Field Gas Chromatography
U.S. EPA. Standard Operating
Procedure: Sentex Scentograph G.C.
Field Use (SOP #1702), December 1988.
U.S. EPA. Standard Operating
Procedure: Photovac 10S50, 10S55, and
10S70 Gas Chromatography Operation
(SOP #2108), January 1989.
Wylie, Philip, L. "Comparing Headspace
with Purge and Trap for Analysis of
Volatile Priority Pollutants",
Research & Technology, 1988, 80(8),
p65-72.
38
-------
TABLE I. NJDEP/HWMP DATA QUALITY CLASSIFICATIONS
DATA
QUALITY
LEVEL
PURPOSE
OF
SAMPLE
EXAMPLE METHODS
OR
INSTRUMENTS
Health & Safety
Site Screening
Field Use when
excavating.
Portable PID (HNU).
Colormetric Analysis.
Portable FID (OVA).
XRF with a remote probe
(X-met).
1A Site Screening.
Field Use when
excavating.
Site Delineation.
ATH Analysis.
Colormetric Analysis.
Field use when
excavating.
Site Delineation.
Portable GC.
Portable XRF with SiLi
detector.
Mobile Lab (limited QA/QC)
Site Delineation.
Lab Confirmation of
field delineation
samples.
Traditional Site
Characterization.
Laboratory Analyzed
Samples, without
QA/QC documentation,
i.e. 600 Series.
Mobile Lab.
Traditional Site
Characterization.
Lab Confirmation of
field delineation
samples.
Laboratory Analyzed
Samples, with full
QA/QC documentation,
i.e. CLP-IFB.
Traditional Site
Characterization.
Lab Confirmation of
field delineation
samples.
Laboratory Special
Services.
Mobile Lab.
39
-------
PLENARY SESSION DISCUSSION
LLEWELLYN WILLIAMS: There was a reference in the first or second paper
to our concern about the acceptance of field screening and field analytical data
by a regulatory group. How do we deal with, or how do we encourage the
acceptance of field screening and field analytical method data in the regulatory'
arena?
DENNIS WYNNE: Pan of it. I think, is encouraging the risk-taking among our
managers. What we have dealt with in the past is a tendency to rely almost
exclusively on the tried and true methods of the contract lab program (CLP).
What we are trying to do in the Superfund Program is to wean people off
inordinate use of the CLP by saying that that is for a specific intended purpose.
Il's not intended for all uses. If you focus on thedataquality objectives approach,
there's not a need lo over rely on the CLP. because you often have for some uses
a gold plated version that isn't needed for some of the basic uses. The field
screening methods would be more appropriate. Some of the ways to do that is by
the work groupapproach and trying things toencourage managers to use it. We're
trying to focus on things like streamlining the remedial investigations and
feasibility studies, and you really can't tell what you find if you're using fixed
labs exclusively. There's downtime while the data are being sent out. analyzed
and rev icwed. In some cases I think what we're trying to do is look where the time
being spent on the program. Trying to shorten those times where we can. trying
to encourage the user community to come together in work groups, being able to
provide guidance through training programs are ways we can gel more people
familiar with field screening and thereby limiting some of the conservativism
that we deal with in some of the managers who tend to rely on contract lab
programs.
Another pan of it. I think, would be toemphasize that some of the field analytic
methods can provide you with as much accuracy as you get through fixed labs.
We need to emphasize those points, so people aren't always assuming field
methods are sort of the poor cousin of the fixed labs.
1 guess from a PRPperspective. which the Army is. our approach has been to use
field screening as a powerful tool to guide the traditional quality control in lab
data. We've been forced that way because of our negotiations with the regions.
because of the requirement for a lot of this data to eventually stand up in court.
So we see it as minimizing the requirement for that extreme case of chain of
custody and total reliability of the data because of extreme quality control. We'll
minimize the number of samples we really have to take, because of this powerful
tool, the field screen.
HOWARD FRIBt'SH: 1 think that your continued use of field analytical
methods and analyses is going to force it to be accepted, for one thing. Another
thine, the acceptance seems to be more fragmented. That is. it seems to be more
accepted to say in the Removal Program, and less accepted but somewhat
accepted in other programs. I think that without field analytical methods and
analyses, we're really sampling blind, and there is no reason what 90^ of the
samples that get sent to the CLP should be nonhits. when 909t could be hits.
Another way is to document all this just like we're trying to do with the catalog
and the user's guides. I think it will be accepted much more than it is now.
N ABIL YACOUB: My question has two pans: 1) would that manual encompass
methods developed by the Army and other entities? 2) w ould the methods incl ude
those for matrices other than w ater. because in the real w orld, y ou have a problem
with soils and sludges and such.
HOWARD FRIBUSH: Yes. it will include other matrices. As long as the
methods have been used for Superfund activities, and they have been shown to
work, and they've been field tested, there is no reason why they can't be included.
That's why channeling performance information one of these methods, back to
EMSL-Las Vegas for an ongoing evaluation is so important. In the future
updates, we can either delete some of the methods, or it might help us to combine
some of the methods. And as far as your first question: 1 would say that we're
definitely open to including methods developed by the Army, especially if
they're users in Superfund activities, for example in the Federal Facilities
Program. If there is performance information, we would like to know that.
MICHAEL C ARR ABB A: I have more of a comment or a suggestion directed
at both the Environmental Protection Agency, as well as the Department of
Energy.
If you look at the Chemical Sensors session, there are six talks: two representa-
tives from the Federal Government, and four representatives from small busi-
ness. My comment is that the Environmental Protection Agency, as well the
Department of Energy, is grossly underutilizing the small business innovative
research program to bring forth some of these field screening technologies, such
as in the area of chemical sensors or optical spectroscopy. If you look at the
current solicitations for 1991 for the Department of Energy, we've been hearing
about this great problem in environmental restoration and field screening. There
arc no topics in there for small business, and a lot of the innovation that we're
going to need in the future, particularly for the DOE and EPA, is going to come
from small business with new and innovative ideas. This is not the case for the
Department of Defense, who is actually doing a pretty good job in using the SBIR
program lo fulfill these needs.
EDGAR SHL'LMAN: I noticed in the user's guide that is presently out, that
there is a heavy emphasis on fieldable methods, and very little on the man-
portable type of instruments or methodology. Could you comment on what the
future direction is relative to the man-portable type of instruments for field
screening? And also perhaps to other panelists in terms of their judgment as to
the value of smaller dev ices for field screening?
HOWARD FRIBUSH: 1 think that the catalog in the user's guide is intended to
be comprehensive, and there is no reason that the smaller survey instruments.
such as organic vapor analyzers, or portable radionuclide analyzers couldn't be
included. In fact, since they are used a lot, especially in the Site Assessment
Program, and the Removal Program, they should be included and will be
included.
Up a stage to the man portable instruments, we now have portable GC/MS. Those
certainly will be included. I think the short answer to your question is that we
w ant everything that is used in Superfund typically to be included into the catalog
and user's guide.
EDGAR SHULMAN: I guess I was looking toward your judgment in terms of
the v alue of small devices. Would the priority in the future be toward encouraging
people to actually get much smaller devices? I know you are talking about man-
portable GC/MS. but they really are not man portable right now. They're
fieldable. you still need a truck or something similar.
HOWARD FRIBUSH: Are you talking about field kits, or are you talking about
survey instruments?
EDGAR SHULMAN: I'm talking about survey instruments, try-ing to encour-
age their research and development community, in terms of an agenda for
research. Maybe that's what I'm looking for. Where should the priorities be put,
from the R&D community, relative to the kinds of methods that are envisioned
for the future.
LARRY REED: You've made a good point. I think looking at the present catalog
we have out. There is a bias that was introduced when we were gathering existing
information, a large pile which was available as part of the Field Analytical
Support Program. This program was developed in part for field investigation
teams, and the Site Assessment Program nationwide. There had been a focus to
look at the bigger equipment and the more refined type of equipment. I think that
was done just to get the catalog out, what information is in use — was available
and useful. I think what we are going to try and do is balance it now by more of
the technologies, try to focus more on portable kinds of instruments, also. I know
40
-------
DISCUSSION
in particular when I'm looking at the future of the Site Assessment Program, as
the field investigation team contracts start to expire this year, we're going to be
looking at two phases of the shifting of the equipment, the larger field analytic
support equipment, and then the portable equipment. We want to make sure that
that equipment will be transferred to the people who are going to be doing the Site
Assessment work and in looking at the next generation of it. That's a good point
you make. I think we'll try to balance it out.
HOWARD FRIBUSH: I just wanted to say that the survey instruments have a
definite use in Superfund. They are used quite often to determine the health and
safety requirements for workers, and also to identify hot spots. So since they have
a definite use in Superfund, they will be included in the next update.
LLEWELLYN WILLIAMS: I was just reminded thai the EPA is not 100%
Superfund. There are a fair number of other programs out there for which field
screening technologies will have a place, and in many of those applications, I
think some truly portable measurement instruments are going to be very, very
important.
CHRIS LIEBMAN: I though that the key to the compendium and the success
of the compendium was really dependent on people submitting their methods to
the working group, so that we can see that they are included in the compendium.
I think it's important to point out that if survey instruments are not currently in
the compendium, that largely reflects the fact that people had not submitted
methods. I think if you are unhappy with what is in the compendium, to change
that, make your submissions.
DOUG PEERY: In putting the catalog together, of course you're addressing
purely programs that the EPA is addressing, in dealing with private clients who
rely on these things for their own information, you get locked in. We also have
to respond to that. Is there going to be a flexibility in this catalog whereby we.
as the person developing the procedure, can go through steps and prove that the
procedures are applicable and usable, and not be locked in or having to reply.
Maybe taking the USATHAMA Procedure and Methodology Proof Program,
making it simpler, and integrating the two, so that it can be done very quickly and
easily and economically would be one way. Is there a procedure or a thought to
adding something along that line?
HOWARD FRIBUSH: I think that is a really good idea, and a really good
statement. This is something that the work group has not yet addressed. I think
that is a good topic for a future item at our next work group.
Originally, we had talked about EMSL-Las Vegas doing some of that validation.
When we look at all the methods that we have, I think it might be more
appropriate to have EMSL-Las Vegas look at the performance information. But
for new methods, I think that that is an area for future consideration and I
appreciate the comment.
COLLEEN PETULLO: I notice that DOD, DOE and EPA are all developing
innovative technologies, or supporting innovative technologies to be developed.
We all march to a different drummer in terms of QA. How is that all being
coordinated?
LLEWELLYN WILLIAMS: There are a number of ways in which attempts are
being made to harmonize Quality Assurance, not the least of which is the
interagency ad hoc committee on QA for environmental measurements that's just
been established. We are looking very hard at both QC and QA requirements,
both from a process standpoint and from an operations standpoint, to see if we
can get more uniform application of QA/QC procedures, agency wide, as well as
across the agencies. We're well aware that there have been concerns in the past
with respect to dealing with each of our individual Regions, as separate
autonomies, and that a DOE or a DOD may have a difficult time in getting the
same kind of response to the same situation going from Region to Region. This
is part of what we're hoping can come out of the interagency work is to get more
uniform application and uses.
COLLEEN PETULLO: Is there one form of QA program plan that you're kind
of leaning to at this point?
LLEWELLYN WILLIAMS: When you say a form of QA program plan, I'm
not quite sure what you mean.
COLLEEN PETULLO: EPA tends to be more a laboratory type QA versus
field operational, and DOE tends to be more field operational, and I'm just
curious as to how you're going to get all this all melted together.
LLEWELLYN WILLIAMS: I think there is much we can learn from the
approaches of other agencies. We will attempt to accommodate and utilize the
best that the other agencies can offer, and provide a focused program that
everyone can buy in on and live with.
41
-------
A FiberOptic Sensor for the Continuous Monitoring of
Chlorinated Hydrocarbons
P.P. Milanovich1, P.P. Daley2, K. Langry1, B.W. Colston1 Jr.,
S.B. Brown1, and S.M. Angel1
Environmental Sciences Division,
Environmental Restoration Division
Lawrence Livermore National Laboratory
Livermore, CA 94550
Abstract
We have developed a fiber optic chemical sensor for use
in groundwater and vadose zone monitoring. The sensor is
a result of modification of previous work in which we dem-
onstrated a fluorescence based sensor for the non-specific
determination of various volatile hydrocarbons. The prin-
ciple of detection is a quantitative, irreversible chemical
reaction that forms visible light absorbing products. Modifi-
cations in the measurement scheme have lowered the detec-
tion limits significantly for several priority pollutants. The
sensor has been evaluated against gas chromatographic
standard measurements and has demonstrated accuracy and
sensitivity sufficient for the environmental monitoring of
trace levels of the contaminants trichloroethylene (TCE) and
chloroform.
In this paper we describe the principles of the existing
single measurement sensor technology and show field test
results. We also present the design of a sensor which is
intended for continuous, sustained measurements and give
preliminary results of this sensor in laboratory experiments.
Background
This sensor technology is an outgrowth of research
initially sponsored by the U.S. Environmental Protec-
tion Agency. Here, a fluorescence based probe for the
remote detection of chloroform was conceived, devel-
oped and demonstrated in the mid-1980's.1 The sensi-
tivity and accuracy of the probe proved insufficient for
many monitoring applications and research was dis-
continued. However, in DOE sponsored research one
of us (SMA) invented a new concept sensor that has
demonstrated significantly improved sensitivity and
accuracy for both TCE and chloroform.2 This sensor is
currently under evaluation in monitoring well and
vadose zone applications.
Principles of Operation
The basic components of the sensor technology are
the chemical reagent, electro-optic measurement de-
vice, and the sensors. For the latter, we have developed
two versions, one for single and one for continuous
measurements. A brief desrciption of the components
follows.
Chemistry. The chemical basis of this technology is
the irreversible development of color in specific re-
agents upon their exposure to various target molecules.
The primary reagent is an outgrowth of the work of
Fujiwara3 who first demonstrated that basic
pyridine,when exposed to certain chlorinated com-
pounds, developed an intense red color. This red color
is due to the formation of highly conjugated molecules
as shown below. We and others have since demon-
43
-------
strated that this and closely related reactions can be
used to detect trace amounts of these same com-
pounds.4
H H
H R H
(Red)
Sensors The single measurement sensor (Fig 1) is
comprised of the terminus of two optical fibers and an
aliquot (20 ul) of reagent in a small capillary tube. The
fibers are sealed into one end of the capillary tube and
reagent is placed into this capillary to a length of ap-
proximately 5 mm. A porous teflon membrane is
placed over the open end of the capillary to prevent
loss of the reagent. Target molecules, TCE for example,
readily pass through the membrane and produce color
in the reagent. This color results in decreased transmis-
sion of light at 540 nm. The measurement of the time
history of the color development provides a quantita-
tive measure of the target molecule concentration.
Since the reaction is non-reversible, the reagent must be
replenished for every measurement. This is readily
accomplished through the use of easily replaceable,
disposable capillaries.
Electro-optics. The readout device is shown highly
schematically in Fig 3. Here the emission of a minia-
ture tungsten-halogen lamp is collected by suitable
optics, chopped with a tuning fork and directed into an
optical fiber. The fiber transmits this light with high
efficiency to the sensor where it passes through the
chemical reagent, reflects off the teflon membrane, and
is collected by a second optical fiber. This latter fiber
transmits the reflected light to an optical block where it
is divided into two beams by a long pass dichroic
mirror. These resulting beams are optically filtered at
540 nm and 640 nm, respectively, and their intensity is
ultimately measured with silicon photodiodes using
phase sensitive detection techniques.
Figure 3. Sensor readout device
Computer
-: M :-•
-.J---_TCE Chloroform
Figure 1. Schematic of the single-measurement sensor
Figure 2 shows a sensor that has been designed for
continuous operation." It is essentially identical to the
single measurement version with the exception of the
addition of two micro-capillary tubes. These are used
to supply new reagent to the sensor either continuously
or on demand.
n«»a»m
-*-TCE Chloroform
Figure 2. Schematic of the continuous-measurement
sensor
Since the colored product absorbs strongly at 540
nm and is virtually transparent at 640 nm, the ratio of
540 to 640 gives a nearly drift free measure of 540 nm
absorption. The sensors are calibrated in two ways (1)
in the headspace above standard TCE solutions of
known w/w concentration or (2) in vapor phase using
calibrated dilutions (v/v) of dry TCE vapor. Figure 4
shows the time dependent transmission of sensors
exposed to TCE standard solutions and a resulting
calibration curve.
Results and Discussion
Groundwater monitoring. The sensor has been
evaluated against contractor sample and analysis of 40
monitoring wells located within the boundary of LLNL.
These wells are sampled quarterly with subsequent
chemical analysis performed by EPA standard 624
purge and trap gas chromatography (GC). We ob-
tained concurrent samples during the quarterly con-
tractor sampling and used our fiber sensor to make
-------
0.90
0.70
o
| 0.50
a>
I
5 0.30
0.10
3.0
9.0
15.0
Time (min)
21.0
0 ppb
27.0
0)
O
0.60
0.50
0.40
0.30
0.20
0.10
0.00
100
200
300
400
500
600
[TCE] (equilibrium vapor phase over given
ppb in stirred water solution at 25°C)
Figure 4.
Standard
Sample transmission ratio curve, and working standard curve for dual-wavelength absorption sensor.
curve obtained from % transmission at a fixed time following iniation of exposure
duplicate TCE concentration determinations. Figure 5
shows a diagram of the laboratory measurement appa-
ratus. Samples were sequestered with no head space
into 250 ml Pyrex bottles. These were immediately
returned to the laboratory and divided in half. The
fiber sensor was then introduced into the resulting
headspace through a gas tight valve and a measure-
ment was initiated after stiring the sample for 5 min-
utes.
Optical fiber
(to spectrometer)
Capillary to pump
(when operating in
continuous mode)
Gas-tight valve
Sensor
_JL^
^ <=> -4
^
t
**"»',
•
-" Water sample
Magnetic stirrer
Table 1 below shows the comparison of some of the
contractor measurements with the fiber sensor. All
fiber sensor values are the average of the duplicate
samples. There is excellent agreement between the GC
and fiber sensor determinations with nearly all values
within the variance of the GC.
Vadose zone monitoring. LLNL site 300 was chosen
as the location for initial vadose zone evaluation of the
fiber sensor. The vadose zone was accesssed at several
locations through existing dedicated soil vapor moni-
toring points. The samples were drawn at nominally
450 cc/min through copper tubing to a remote mobile
laboratory. The lab contained both the fiber sensor
apparatus and a portable GC. The instruments were
connected to the sample stream in series as depicted in
Fig 6 below. Both devices were calibrated for TCE
Figure 5. Schematic of vessel used for laboratory
headspace measurements
Figure 6. Schematic of vadose zone sampling and
calibration apparatus. Sample air is drawn with a pump
on board the GC
45
-------
Table 1. Representative data from field calibration study, compiled from TCE measurements
from monitoring wells and piezometers at LLNL.
Well
MW352
P418
MW271
MW217
MW365
Date
2/13/90
2/13/90
3/7/90
3/5/90
3/6/90
[TCEKppb)
Fiber GC
44
54
86
106
27
58
72
160
86
22
Well
MW357
P419
MW364
MW458
MW142
Date
2/13/90
2/13/90
3/7/90
3/6/90
3/6/90
[TCEKppb)
Fiber GC
78
61
59
33
94
84
66
74
20
140
15-H
10
20 30
Elution time (min)
40
measurements with precision gas mixtures prior to
sampling. The fiber sensor tracked the GC very well
through a wide range of concentrations. Figure 7 is a
particularly interesting result. Here both instruments
were compared in a nearly contamination free location.
It is clear that the GC was at its limit of detection,
whereas the fiber sensor readily made a successful
measurement. Estimates of TCE concentration for this
location was <10 ppb.
Continuous measuring sensor. The above described
sensor has demonstrated adequate sensitivity and
accuracy to represent a viable new environmental
monitoring technology. However, the current design,
Time (min)
Figure 7. Results of (above) GC (SRI Instruments 8610,
PID detector, 6' x 1 /8" silica gel column), and
(below) fiber sensor measurement of extremely
low TCE levels in soil gas (estimated to be -150
ppbv/v, i.e.: 150 umoles TCE per mole air).
1.0-1
E
M
| 0.6-
0.4-
0.2-
'0
20
30
Time (min)
60
Figure 8. On-demand measurement of 10 ppm TCE
(i.e.: headspace measurement over water containing 10
ppm TCE) with continuous sensor system
-------
which incorporates an irreversible chemical
reaction, requires the sensor to be refurbished
subsequent to each measurement. This liabil-
ity limits its application somewhat in envi-
ronmental monitoring.
The sensor shown in figure 2 represents the lowest
risk mitigation of this liability. Preliminary resits with
prototypes of this sensor are very promising. Figure 8
shows typical on-demand measurements obtained with
this sensor in laboratory testing. We anticipate that this
sensor will become an integral component in a down-
well monitoring instrument currently being developed
atLLNL.
Acknowledgements
This work is supported by the DOE Office of
Technology Development (OTD) and performed
under the auspices of DOE contract W-7405-
Eng-48 and the Center for Process Analytical
Chemistry. The authors are indebted to Dr.
Lloyd Burgess and the Center for Process Ana-
lytical Chemistry, Univ of Washington for col-
laboration that led to the design and demon-
stration of the continuous sensor. The authors
also wish to thank Dr. F. Hoffman of LLNL for
many helpful discussions.
References
1. F.P. Milanovich, D.G. Garvis, S.M. Angel, S.K.
Klainer, and L. Eccles, Anal. Inst, 15,
137(1986).
2. S.M. Angel, M.N. Ridley, K. Langry, T.J. Kulp
and M.L. Myrick, "New Developments and
Applications of Fiber-Optic Sensors," in
American Chemical Societry Symposium
Series 403, R.W. Murray, R.E. Dessey, W.R.
Heineman, J. Janata and W.R. Seitz,
Eds.,(American Chemical Society, Washing
ton, D.C.,1989) pp 345-363.
3. K. Fujiwara, Sitzungsber. Abh. Naturforsch.
Ges. Rostock, 6,33(1916).
4. S. M. Angel, P. F. Daley, K. C. Langry, R.
Albert, T. J. Kulp, and I. Camins, LLNL UCID
19774, "The feasibility of Using Fiber Optics
for Monitoring Groundwater Contaminants
VI. Mechanistic Evaluation of the Fujiwara
Reaction for the Detection of Organic Chlo
rides", June, 1987.
5. R. J. Berman, G. C. Christian and L. W. Bur
gess, Anal. Chem, 62,2066(1990).
47
-------
Chemical Sensors for Hazardous Waste Monitoring
M.B. Tabaccoj Q. Zhou, and K. Rosenblum
GEO-CENTERS, INC.
7 Wells Avenue
Newton Centre, MA 02159
M.R. Shahriari
RUTGERS UNIVERSITY
Fiber Optic Materials Research Program
Piscataway, NJ
ABSTRACT:
A family of novel fiber optic
sensors is being developed for on-
line monitoring of chemical species
in gases and liquids. The sensors
utilize porous polymer or glass op-
tical fibers in which selective che-
mical reagents have been immobiliz-
ed. These reagents react with the
analyte of interest resulting in a
change in the optical properties of
the sensor (absorption, transmis-
sion, fluorescence). Using this ap-
proach, low parts per billion level
detection of the aromatic fuel va-
pors, benzene, toluene and xylene,
and hydrazines has been demonstrat-
ed, as have sensors for ethylene
vapor. Also relevant to groundwater
monitoring is the development of a
pH Optrode System for the pH range
4-8, with additional optrodes for
lower pH ranges.
INTRODUCTION
The functional operation of
optical fiber chemical1 sensors in-
volves the interaction of light
which propagates through the fiber,
with a reagent that in turn selec-
tively interacts with the environ-
ment to be sensed. Typical optical
properties including evanescent ab-
sorption and fluorescence, and che—
miluminescence can be exploited in
these sensors. The reagents are
normally immobilized into a membrane
or porous polymer matrix and then
coated either on the tip or side of
the fiber.
One of the problems encounter-
ed with fiber optic chemical sensors
based on evanescent absorption is
their characteristic low sensitivi-
ty. This results from the limited
depth of penetration of the evanes-
cent field of the light into the
reagent cladding as well as the ef-
fect of internal reflections [1-4].
Figure 1 illustrates the prin-
ciple of detection used in fiber
optic chemical sensors. In the fig-
ure, porous glass and porous polymer
approaches are compared to conven-
tional evanescent chemical sensors.
In the porous fiber, the analyte
penetrates into the pores and inter-
acts with the reagent which is pre-
viously cast (immobilized) into the
pores. The porous fiber has a large
interactive surface area (due to the
large surface area provided by the
pores), resulting in dramatically
enhanced sensitivity in the optrode.
Another advantage of a porous glass
fiber is the small sensing region
(about 0.5 cm in length and 250 mi-
crons in diameter). Additionally
the sensor is an integral part of
the fiber waveguide. This latter
feature minimizes the complications
associated with the physical and
optical coupling of the sensor probe
to data transmission fibers. In
addition, multiple fiber sensors can
be deployed using a single analyti-
cal interface unit. These sensors
are expected to be less expensive
than conventional fiber optic chemi-
cal sensors based on materials cost
and ease of fabrication. Porous
49
-------
fiber sensors for the measurement of
humidity, pH, ammonia, ethylene, CO,
hydrazines, and the aromatic fuel
constituents benzene, xylene and
toluene have been successfully dem-
onstrated by GEO-CENTERS, and by
Rutgers University [5-12].
Fabrication of Porous Glass Optical
Fiber
Porous glass optical fibers
are fabricated by the Fiber Optic
Materials Research Program at
Rutgers University, using the meth-
odology described below [5].
The material used in the fiber
is an alkali borosilicate glass with
the components Si02> B203 and alkali
oxides. This type of glass is a
well characterized system, produc-
ible at a low cost. Most important-
ly it exhibits the phenomenon of
liquid/liquid immiscibility within a
certain temperature range. The
above composition is melted in an
electrical furnace at 1400°C and
cast into rods with a 20 mm diameter
and 0.5 m in length. The rods are
drawn into fibers at about 700 "C by
a draw tower equipped with an elec-
trical furnace. Fibers with a 250-
300 micron diameter with a 5-10 cm
length are then heat treated in a
tube furnace at 600°C for about 3
hours. The heat treated glass be-
comes phase separated, with one
phase silica rich and the other bo-
ron rich. The boron rich phase is
leached out of the glass by placing
the fiber in a bath of hydrochloric
acid. The fibers are subsequently
washed with distilled water and
rinsed with alcohol. Figure 2 il-
lustrates the processing steps for
fabricating porous fibers.
Subsequent to fiber prepara-
tion, the porous segment is cast
with the sensing reagent (indicat-
or) . This is done by dissolving the
reagent in a solvent at a predeter-
mined concentration and soaking the
porous fiber in the solution. The
reagent is then dried into the pores
by air drying or in a low tempera-
ture oven. Alternatively, the glass
surface can be treated with a silan-
izing reagent to facilitate chemical
coupling to the sensing reagent.
Fabrication of Porous Polymer
Optical Fiber
As an alternative to chemical
immobilization or physical adsorp-
tion in porous glass, porous polymer
optial fibers can also be used to
create fiber optic chemical sensors.
Sensors using these fibers have been
demonstrated for ethylene, CO, NH3,
pH, and humidity detection. The
principle of porous polymer fiber
sensors has the same basis as porous
glass sensors. Consequently high
sensitivity is achieved. In this
approach the indicator is dissolved
directly into the monomer solution
before forming the polymer fiber;
therefore, the indicator is strongly
bonded to the polymer network. In
fact, the porous polymer approach
provides the advantage of both chem-
ical bonding and physical entrapping
of the indicator. Also, the pore
size and the amount of indicator can
be precisely controlled by changing
the composition of the monomer solu-
tion, resulting in very good sensor-
to-sensor reproducibility. This
fabrication process is additionally
quite suitable for mass production.
This reduces the cost of optrodes.
The porous polymer fibers are
prepared by a heterogeneous copoly-
merization technique. The basic
principle behind this technique is
the polymerization of a mixture of
monomers which can be crosslinked in
the presence of an inert and soluble
component solvent. Subsequent to
polymerization, the inert solvent
which is not chemically bound to a
polymer network, is easily removed
from the polymer leaving an inter-
connected porous structure.
Monomer starting solutions are
prepared which contain the cross-
linker, initiator, inert solvent and
chemical indicator. The mixture,
including the indicator, is injected
into a length of glass capillary,
(typically 500 microns in diameter).
The filled glass capillaries are
sealed such that they are virtually
free of air, and polymerization is
initiated and completed in a low
temperature oven. After polymeriza-
tion, the uniform and transparent
polymer fibers are pulled out of the
capillaries. Finally, the fibers
50
-------
are washed in an organic solution to
remove any remaining inert solvent.
A combination of parameters
determines the final physical prop-
erties of the cross—linked polymer
network. These include the solvent
properties, amount and type of inert
solvent, as well as the quantity of
cross-linking agent employed.
Results and Discussion
Porous glass and porous poly-
mer optrodes have been designed and
demonstrated for aromatic fuel va-
pors (benzene, toluene, xylene),
hypergol vapors (hydrazine and
UDMH), for NH3, CO and ethylene.
Similarly, optrodes have been demon-
strated for the chemical parameters
pH, humidity and moisture content.
A pH Optrode System is cur-
rently under development which is
applicable to a variety of field
screening and contamination monitor-
ing tasks. Porous glass pH optrodes
have been fabricated which are oper-
ational in the pH 4-8 range. A
unique co—immobilization technique
was developed to tailor the sensor
pH sensing range to a specific ap-
plication. Optrodes are fabricated
by first silanizing the porous fiber
surface to facilitate the attachment
of the sensitive indicator material.
Spectral transmission scans are con-
ducted in order to identify the
wavelength region of maximum sens-
itivity to pH. The sensor interro-
gation wavelength is selected based
on these spectral scans.
Optical intensity versus time
measurements as a function of pH,
have been made for each optrode at
the interrogation wavelength. The
sensitivity and linearity is deter-
mined by plotting optical intensity
at equilibrium, versus pH. Figure 3
shows the response of the optrode
with an immobilized indicator. The
sensor is operational between pH 4
and pH 6.5, with greatest sensitivi-
ty and linearity between pH 4.5 and
pH 6. Saturation of the sensor re-
sponse occurs at pH values above 7
and less than 4.
A second indicator, which is
structurally very similar to the
first indicator, has been tested
with the intent of increasing sensi-
tivity at higher pH values. The
response of this indicator is pre-
sented in Figure 4. The data indi-
cates good linearity and sensitivity
above pH 7.
A mixture of the two indica-
tors was immobilized in a porous
glass fiber. The results with this
sensor are shown in Figure 5. The
data indicates both excellent sensi-
tivity and linearity across a pH
range extending from 4 to 8. The
co—immobilization of these two indi-
cators represents a unique approach
to sensor design and demonstrates
that sensing range can be tailored
to meet specific requirements.
The reversibility of these
sensors has been evaluated. This is
accomplished by cycling a test solu-
tion, into which the pH optrodes
have been immersed, between pH val-
ues of 4.5 and 7.
Figure 6 depicts the variation in
optical transmission of the pH opt-
rode as a function of time. The
data indicate that the sensor is
fully reversible and peak to peak
reproducibility is better than 90%.
The spikes in the response curves
are artifacts associated with the
test setup. Similar results have
been obtained using porous polymer
optical fiber.
Fuel Vapor Optrodes
GEO-CENTERS, INC. has design-
ed, fabricated and evaluated porous
fiber optrodes for detection of aro-
matic fuel constituent vapors. A
xylene optrode with sensitivity <50
ppb has been demonstrated. Response
time, reproducibility, linearity,
and selectivity have been determin-
ed. Benzene and toluene optrodes
have also been demonstrated. Labo-
ratory results indicate that there
are highly sensitive optrodes, with
near real time response. They are
additionally capable of selective
detection of target species.
51
-------
With these optrodes (as well
as the hypergol, ethylene, and CO
optrodes) the rate of change of the
optical transmission is directly
proportional to analyte concentra-
tion. An example of xylene optrode
response to different xylene concen-
trations is presented in Figure 7.
Each curve corresponds to a differ-
ent xylene concentration. A plot of
the slopes of the data in Figure 7
versus xylene concentration is shown
in Figure 8. This data demonstrates
good sensor linearity from low part
per billion to low part per million
concentrations.
Hypergolic fuel optrodes have
been developed to detect vapors for
NASA and U.S. Air Force operation
applications.
The principle of operation and sen-
sor response is similar to that of
the xylene optrodes. The hypergolic
fuel optrodes can be configured as
personal dosimeters for industrial
hygiene applications or as portable
detection instruments. Figure 9
shows a typical optrode response as
a function of time for different
concentrations of hydrazine. The
slope of the optical intensity ver-
sus time curve may be correlated to
the hydrazine vapor concentration.
Conclusions
Sensors utilizing optical
waveguides offer many advantages for
hazardous waste monitoring applica-
tions including size, near real time
response, and low manning and exper-
tise requirements. Additionally,
porous glass and polymer optical
fibers offer significant advantages
in these applications because their
large interactive surface area dra-
matically improves sensitivity.
They also provide a continuous opti-
cal path. This minimizes mechanical
and optical coupling losses. Addi-
tionally, sensor interfaces can be
developed that allow multi-sensor
operation. These chemical optrodes
can be applied in a variety of envi-
ronmental monitoring scenarios, as
well as to developmental bioreac-
tors, control of process streams,
and industrial hygiene. A family of
fiber optic optrodes offers the pos-
sibility of effectively having a wet
chemistry laboratory that can be
brought to the field.
References
1. J.F. Giuliani, H. Wohltjen,
and N.L. Jarvis, Opt. Lett. 8,
54 (1983).
2. A.P. Russell and K.S. Fletch-
er, Anal. Chem. Actal. 170.
209 (1985).
3. D. S. Ballantine and H.
Wohltjen, Anal. Chem. 58, 883
(1986).
4. C. Zhu and G. M. Hiefttse,
Abstract 606, paper presented
at the Pittsburgh Conference
and Exposition on Analytical
Chemistry and Applied Spectro-
scopy, Atlantic City, N.J.,
(1987).
5, M.R. Shahriari, Q. Zhou, G.H.
Sigel, Jr., and G.H. Stokes,
First International Symposium
on Field Screening Methods for
Hazardous Waste Site Investi-
gations, Las Vegas, NV (1988).
6. M.R. Shahriari, G.H. Sigel,
Jr., and Q. Zhou, Proc. of
Fifth International Conference
on Optical Fiber Sensors, Vol.
2 Part 2, 373, (January 1988).
7. M.R. Shahriari, Q. Zhou, and
G.H. Sigel, Jr. Opt. Lett. 13,
407 (1988).
8. M.R. Shahriari, Q. Zhou and
G.H. Sigel, Jr., "Detection of
CO Based on Porous Polymer
Optical Fibers", Chemical,
Biochemical and Environmental
Fiber Sensors, V. 1172, SPIE
Sept. 6-7, 1989.
9. M.B. Tabacco and K. Rosenblum,
"Aromatic Hydrocarbon Optrodes
for Groundwater Monitoring
Applications", GEO-CENTERS,
INC. Technical Report GC-TR-
89-1912. April 1989.
52
-------
10. M.B. Tabacco, K. Rosenblum,
and Q. Zhou, "Optrode Develop-
ment for Environmental pH Mon-
itoring" , GEO-CENTERS, INC.
Technical Report GC-TR-89-
1989, August 1989.
11. M.B. Tabacco, K. Rosenblum,
and Q. Zhou, "Personal Hydra—
zine Vapor Dosimeter", GEO-
CENTERS, INC. Technical Report
GC-TR-90-2071, February 1990.
12. M.B. Tabacco, Q. Zhou, and K.
Rosenblum, "Development of
Trace Contaminant Vapor Moni-
tors", GEO-CENTERS, INC. Tech-
nical Report GC-TR-90-2138,
August 1990.
53
-------
a) Evanescent (Internal Reflection), RFS
Chemical Reagent
b) Evanescent (Internal Reflection), Side Coated FOCS
c) Porous Fiber (In-Line Absorption or Luminescence)
Figure 1.
Schematic Diagram Comparing Basic
Sensor Designs
100
80
>
1 60
1 40
Q.
0
20
0
4
Sei
x
^
\
\
v
^*»
^•^
— ••
5 6
PH
Figure 3.
isor Response With Bromocresol Green
Indicator As A Function Of pH
7
Composition Design
Na2BeO,3 Si02
Melting And Casting
Fiber Drawing
Heat Treatment
Leaching
Surface Treatment
Figure 2.
Processing Steps For Producing Porous
Glass Fibers
CH2
ii z
(CH3)
o-c
S;
360
300
£" 240
I
Z 180
0" 120
60
0
(
**• -.
•^--i
' ,
<
\
I
\
\
56789
pH
Figure 4.
Sensor Response With Bromocresol Purple
Indicator As A Function Of pH
-------
Figure 5.
Sensor Response With Co-immobilized
Indicators As A Function Of pH
50 100 150 200 250
Time (seconds)
300
Figure 6.
Optrode response time as a function of pH
105
X
«
E
c
i
o
in
0
(
• ^"^"^
^
^r
* i
)
^s>
S
^^
^,
^
2
Xylene Concentration in ppm
Figure 8.
Calibration Curve for Xylene Optrode
Based onPorous Glass Fiber
234
Time in Minutes
Figure 7.
Response Curves for Porous Glass Xylene
Sensor. Xylene Concentrations Range from
2 ppm to -40 ppb
55
-------
x 10
Average Slope (sec
20
15
5
0
Y = 2.55 + 0.1
^
^^
S
13X R = 0.99
„,--
^
*^
I^*3*"
; 20 40 60 80 10
Hydrazlne Concentration (ppb)
Figure 9.
Optrode Response to Hydrazine Vapor
at 32% relative Humidity and 24 °C
-------
Rapid, Subsurface, In Situ Field Screening
of Petroleum Hydrocarbon Contamination
Using Laser Induced Fluorescence Over Optical Fibers
S H Lieberman and G A Theriault
Naval Ocean Systems Center
Code 522
San Diego. CA 92152
(619)553-2778
S S Cooper P G Malone and
RSCMsen
US Army Waterways Experiment Station
Vicksburg, MS39180
(601)634-2477
PWLurk
U S Army Toxic and Hazardous Materials
Agency
Aberdeen Proving Ground, MD 21010
(301)671-2054
ABSTRACT
A new field screening method is described that couples a fiber
optic-based chemical sensor system to a truck mounted cone
penetrometer. The system provides the capability for real-time,
simultaneous measurement of chemical contaminants and sol
type to depths of 50 meters. Standard sampling rates yield a
vertical spatial resolution of approximately 2-cm as the
penetrometer probe is pushed into the ground at a rate of 1-m
mirf1
The system employs a hydraulic ram mounted in a truck with a
20-ton reaction mass to push 1 meter long, threaded, steel pipes
into the ground. The first section of pipe is terminated in a
60-degree cone and includes strain gauges for measurement of
tip resistance and sleeve friction. A sapphire window mounted
in the side of the pipe, approximately 60-cm above the probe
tip, provides a view port for a fiber optic-based ftuorometer
system. The soil sample is excited through the sapphire window
by light transmitted down the probe over a 500 micron d iameter,
60 meter long fiber coupled to a pulsed nitrogen laser located
at the surface. Fluorescence generated in the sol sample is
carried back to the surface by a second fiber where it is
dispersed using a spectrograph and quantified with a time-
gated, one-dimensional photodiode array. Readout of a fluores-
cence emission spectrum requires approximately 16
mail-seconds. A micro-computer based data acquisition and
processing system controls the fluorometer system, acquires
and stores sensor data once a second, and plots the data In
real-time as vertical profiles on a CRT display.
Results are presented from the first field tests of the system at a
POL (Petroieum-OH-Lubricant) contaminated hazardous waste
site. Initial results from a series of more than thirty pushes
indicate that the system is useful for rapid characterization, in
three-dimensions, of the boundaries of a POL contaminant
plume at concentrations equivalent to sub-parts-per-thousand
of diesel fuel marine. Vertical fluorescence profiles show sig-
nificant small scale vertical structure on spatial scales of a few
cm. This vertical micro-structure appears to correlate with sol
characteristics estimated from point resistance and sleeve fric-
tion. Field ami laboratory calibration of the fiber optic sensor
system using different fuel products is presented and discussed.
Sensor performance is characterized as a function of sol mois-
ture content
Introduction
Defining the location and extent of subsurface chemical con-
tamination is a difficult task. Detailed site investigations require
installation of many monitoring wells and subsequent analysis
of discrete sol and groundwater samples. Effective site char-
acterization is often limited by the ability to select optimal
locations for monitoring wells. Furthermore, the ability to
resolve horizontal and vertical features in the distribution of
chemical contaminates is a function of limitations imposed by
the spacing between wells and the vertical spacing between
samples.
At present, locations for monitoring wells are usually based on
information gleaned from site historical data, ground water
hydrology, and/or indirect chemical screening such as soil gas
measurements. Because of uncertainties in the information
available, well placement is at best an inexact science. Histori-
cal data is often incomplete or inaccurate. Knowledge of
groundwater hydrology at the site may not provide the level of
detal required to understand site characteristics. Interpreta-
tions of sol gas measurements may be complicated by erratic
movement of vapor in the sol due to impervious layers and
changes in atmospheric temperature and pressure. Conse-
quently, many wells are not property positioned and, therefore,
yield information of marginal utlity.
Accurate delineation of the boundaries of contaminant plumes
and defining small scale vertical structure in the distribution of
contamination has important implication with respect to site
remediation. The more precisely the area of contamination is
defined, the less likely ft is that "dean" material will be unneces-
sarily removed or subjected to costly remediation procedures.
Improved techniques for in situ, subsurface, field screening
would have several benefits. Knowledge of the distribution of
chemical contamination in sols and groundwater could be used
to more effectively guide the placement of monitoring wells and
thereby, greatly reduce the number of wells required. Field
screening methods that provide real-time chemical information
at closely spaced intervals could be used to rapidly delineate
small scale horizontal and vertical structure in contaminant
plumes. In addition to increasing the effectiveness of site char-
acterization there should also be a significant cost savings
57
-------
Figure 1. Photograph ofpenetrometer truck developed for use with the fiber optic fluorometer system. The data acquisition system
and fluoromcter system are located in the rear compartment. The hydraulic system used to push the penetrometer probe into the soil is
in thefoward compartment.
associated with the reduced requirement for monitoring wells
and associated analytics.
Towards this goal of improving capabilities for rapid site char-
acterization, we have equipped a truck-mounted cone
penetrometer system (Fig. 1) with a fiber optic based, laser-in-
duced fluorometer system. Cone penetrometers have been
widely used for determining soH strength and soil type from
measurements of tip resistance and sJeeve friction on an instru-
mented probe (1). The probe is normally pushed into the
ground at a rate of approximately 2-cm sec using hydraulic
rams working against the reaction mass of the truck. For a 20
ton vehicle, the standard (35-mm diameter) peoetrometer rod
can be pushed to a depth of approximately 50-m in normally
compacted soils. In order to extend the measurement
capabilities of the penetrometer system to chemical con-
taminants of environmental concern, it is possible to use the
penetrometer system as a platform for insertion of other sensors
into the soil. To date, use of penetrometers for direct sensing
of chemical constituents in soils has been limited to resistivity
measurements (2) and sensors for measuring radioactivity (3).
This report describes the development of an optical based
sensor for direct in situ screening of chemical contaminants.
The system employs optical fibers to make remote laser-in-
duced fluorescence measurements through a window in the
probe tip. The system can be used to characterize contaminant
plumes that contain compounds that fluoresce when exposed
to ultra-violet light. In its present configuration, which uses a
nitrogen (N2) laser (337 nm) excitation source, the system is
selective for polycyclic aromatic hydrocarbon compounds
which are components of POL products. Coupling the optical
fiber sensor with the cone penetrometer provides a capability
for direct, real-time sensing of petroleum hydrocarbon com-
pounds in soils that has not previously existed.
System Description
A schematic diagram of the fiber optic fluorometer system is
shown in Fig. 2. The system was adapted from a design original-
ly developed for in situ fluorescence measurements in seawater
(4-5). The penetrometer system uses two silica clad silica
UV/visible transmitting optical fibers. One fiber is used to carry
excitation radiation down through the center of penetrometer
pipe and a second fiber collects the fluorescence generated in
the soil sample and carries it back to the detector system.
Excitation and emission fibers are isolated from the sample at
the probe tip by a 6.35-mm diameter sapphire window mounted
flush with the outside of the probe approximately 60-cm from
the tip. Although different fibers from several sources have been
employed, the fibers used in studies reported here were 500-^m
in diameter and 60-m in length, unless otherwise noted. At-
tenuation was specified by the supplier to be about 100 dB/km
at 337 nm (this corresponds to 25% transmission at 337 nm for
a 60 m fiber.
Excitation radiation is provided by a pulsed N2 laser (Model
PL2300, Photon Technology, Inc) that operates at 337 nm with
a pulse width of 0.8 nsec and a pulse energy of 1.4 mJ. The
beam is coupled into the excitation fiber using a 2.5-cm quartz
lens. Because of asymmetry in the beam dimensions, 6-mm x
9-mm at the laser aperture, coupling losses into the fiber are
somewhat greater than what would be expected for a conven-
tional Gaussian resonator type laser. No attempt has been
made to reshape the beam to improve coupling. Instead, we
take advantage of the non-symmetrical beam shape by using a
separate length of optical fiber to intercept a portion of the laser
beam that would not normally be coupled into the excitation
fiber. This auxiliary fiber is coupled to a photodiode that is used
to provide an optical trigger for time gating the detector. Opti-
cal triggering of the detector eliminates problems associated
with laser jitter that are experienced with electronic triggering of
58
-------
• ELECTRICAL SIGNAL
' FIBER OPTIC CABLE
Figure 2. Schematic of laser induced fiber optic fluorometer system.
the detector.
A photodiode array detector system is used to quantify the
fluorescence emission spectrum brought back to the surface
over the second 60-m fiber. The detector system consists of a
Model 1420 Intensified Photodiode Array Detector (EG&G
PARC) coupled to a quarter-meter spectrograph which houses
a 300 line/mm diffraction grating. The 1024 element array con-
sists of 25 micron wide diodes centered at 25 micron incre-
ments. For the 300 line/mm grating the dispersion of the
spectrograph translates to a spectral resolution of 0.45 nm per
pixel at the array surface when a 25 micron input slit is used.
The resolution may be increased to 0.075 nm per pixel by using
an 1800 line/mm grating. Readout of an emission spectra
requires approximately 16 msec. Because the detector can be
readout quickly it is possible to add spectra from multiple laser
shots in order to improve the signal to noise ratio of the meas-
urement. Typically, 10 laser shots are used per sample interval.
Control and readout of the detector is performed by a Mode)
1460 optical multichannel analyzer (OMA) (EG&G PARC).
Measurements are initiated by an electronic signal from the OMA
that fires the laser. The laser pulse then triggers an optical
trigger (Model 1303, EG&G PARC) which sends an electronic
signal to a fast pulser (Model 1302, EG&G PARC). The fast
pulser implements an appropriate delay and gates the detector
"on" fora period of 20 nanoseconds. Fast-gating of the detector
activates it only during the time period when the fluorescence
signal is present, thereby minimizing any contribution to the
signal from background light and detector noise.
Incrementing the delay of the detector gate for successive laser
pulses also permits determination of fluorescence decay times.
Other studies have shown that differences in fluorescence
decay times are useful for discriminating compounds of environ-
mental interest (eg., pdycyclic aromatic hydrocarbons) that
cannot be resolved based on differences in their fluorescence
emission spectra (5). At present, fluorescence lifetime meas-
urements are not performed routinely with the penetrometer
system because additional measurement and processing time
would be required. In the future, however, fluorescence decay
measurements could easily be implemented via software con-
trol to take advantage of "dead time" that is currently not utilized
when the push is halted every meter in order to install the next
section of pipe.
An Intel 386 based microprocessor host computer is used to
automate the overall measurement process. The host computer
controls the OMA system and stores fluorescence emission data
received from the OMA and data from strain gauges on the
probe tip. A representative fluorescence spectrum obtained
59
-------
16000
14000-
12000-
10000-
8000
6000-
4000
2000-
300 350 400 450 500 550 600 650 700
Wavelength (nm)
hle=OUA .iai soil
Figure 3. Fluorescnece emission spectum measured for
contaminated soil using fiber optic fluorometer system.
from contaminated soil at the first test site is shown in Figure 3.
The host computer is also used to generate real-time depth plots
on a CRT of the chemical fluorescence measurements and soil
characteristics as interpreted from the strain gauge data. Under
normal operating conditions, fluorescence measurements are
made at a rate of approximately once a second. For the stand-
ard push rate of 2-cm sec"1 this corresponds to a vertical spacial
resolution between measurements of 2-cm. Because each
fluorescence measurement consists of intensities measured at
1024 wavelength points, a push to a depth of only 10 meters will
generate more than 500,000 data points. In order to simplify
data presentation a window (approximately 50 nm wide) is set
in the spectral region anticipated to contain the maximum
fluorescence intensity. The average fluorescence intensity in
the spectral window is then plotted as a function of depth, in
real-time, as the probe is pushed into the soil. A typical data plot
is shown in Figure 4. The entire fluorescence emission
spectrum is stored on a fixed disk to facilitate post-processing
of the data
Characterization and Calibration of Sensor Response
Initially, there were several practical concerns about the viability
of using an optical fiber system to make in situ measurements
in soil in conjunction with the cone penetrometer. Issues of
concern included: (1 )Would the sapphire viewing window retain
contaminant after exposure and thereby exhibit a memory ef-
fect? (2)Could the optical fiber withstand the necessary han-
dling required to thread it through the penetrometer pipe during
insertion and removal? (3)Would the constant flexing of the fiber
during measurement significantly alter the attenuation charac-
teristics of the fiber and thus, invalidate quantitative measure-
ments? Experience gained to date, suggests that none of these
issues appears to be a problem. Inspection of data in Fig. 4
shows that when the probe was pushed through layers of soil
containing relatively high concentrations of contaminant,
fluorescence intensities rapidly approached background levels
as soon as the probe moves out of the contaminant zone. This
suggests that the high pressures acting on the window as the
probe is forced through the soil are effective in removing any
residual contamination that might be adsorbed on the window.
Field experience to date demonstrates that the fibers can
withstand the normal handling required for operations with the
penetrometer. No fiber failures have occurred during the more
than 80 cone penetrometer tests (CRTs) that have been made
so far. Finally, measurements in the field showed that there was
no measurable difference in the amount of laser energy trans-
mitted through the 60-m excitation fiber depending on whether
the fiber was laid out on the ground with no bends or threaded
through 50 meters of penetrometer rod with a 180 degree bend
approximately every meter (as was normally the case). It ap-
pears that as long as the minimum fiber bend radius, for which
total internal reflection is maintained for all modes, is not ex-
ceeded there is no significant variation in throughput loss.
Response of the fiber optic fluorescence sensor has been
calibrated both in the laboratory and in the field using different
fuel products added to soils. We have elected to use fuel
Sleeve Cone Soil Class Fluorescence
Friction Resistance s (relative)
(tons sq ft) (tons'sq ft) &
2 ? -O T3
I I I I
012345 0 100 200 012345 0 10002000
C-3O-90
Figure 4. Example of real-time display showing vertical profiles of soil characteristics and chemical fluorescence measurements.
60
-------
2 4 6 8 10 12 14 16
Diesel Fuel Marine (parts-per-thousand)
18
20 40 60 80 100 120 140 160 180 200
DFM CONCENTRATION (parts per thousand)
Figure 5. Laboratory calibration curves for DFM in soil as
a function of soil moisture content.
Figure 6. Field calibration of penelrometer fiber optic sen-
sor using dieslefuel marine in sand. Inset shows response is
linear below Wppt.
products rather than pure compounds because fuel products
contain a representative mixture of the compounds that may
fluoresce in environmental samples. Obviously, there is no way
to be sure that the distribution of compounds that respond to
our measurement system in the field is an exact match to the
product we select for calibration. In fact, in many cases there
will undoubtedly be a mismatch between the distribution of
compounds in the product used for standardization and the
mixture of compounds present at environmental "dump" sites.
These sites may contain a potpourri of products that have had
time to undergo degradation and loss of more volatile com-
ponents. However, at sites such as tank farms that contain
recent or ongoing fuel leaks, it may be possible to get a good
match between the product used to calibrate the sensor and the
product in the ground. Therefore, it should be stressed that the
utility of the system, in its present form, is for rapid delineation
of hydrocarbon contaminant plumes in order to guide the place-
ment of monitoring wells. With these qualifications with respect
to calibration in mind, data is presented which shows that the
fluorescence sensor appears to be at least a semi-quantitative
sensor for in situ screening of petroleum hydrocarbons.
Laboratory results (Fig. 5) show that measured fluorescence
intensities increased linearly as a function of diesel fuel marine
(DFM) added to uncleaned beach sand. Added quantities of
DFM ranged from 500 to 20,000 parts-per-million (ppm) for this
experiment. Standards were generated by adding known
quantities of fuel product to weighed samples of "clean" soil and
tumbling the mixture overnight in tightly sealed glass containers.
Figure 5 also shows that the measured response did not change
significantly when the water content of the soil was varied from
0 to 10%. Other calibrations using jet fuel (JP-5) in sand also
showed that the fluorescence response did not change when
the water content of the soil sample was varied from 0 to 25%.
This suggests that the response of the fluorescence sensor
should be relatively insensitive to changes in soil moisture
content as the probe moves through the vadose zone into the
saturated zone.
The penetrometer fluorescence sensor was also calibrated in
the field by placing a cylinder over the sapphire window and
filling it with "clean" beach sand (Fisher Scientific) containing
added quantities of DFM. Results (Fig. 6, inset) show linear
response (^ = 0.99) for concentrations in the range of 1000 -
10000 ppm. This is similar to laboratory results discussed
above. Figure 6 shows that for higher concentrations, fluores-
cence intensities appear to approach a saturation val ue at about
10% DFM in sand (weight/weight). This appears to set an upper
limit on the concentrations that can be quantified with this
system. We believe this saturation effect arises because the
fluorescence response of the sample is to a large extent a
surface phenomena. At high concentrations of fluorophore, the
surface of the soil particles become saturated with product and
therefore, the fluorescence approaches a limiting value. The
lower limit of detection for the system configuration described
in this report is approximately 100 ppm (two times noise) using
10 laser shots. Detection limits can be improved, at the expense
of analysis time, by increasing the number of laser shots that are
stacked for each sample interval. Efforts are currently in
progress to determine the effect of soil type on fluorescence
response and to evaluate the "depth of view" of the fluorescence
measurement (ie., how far into the sediment adjacent to the
sapphire window does the measurement penetrate).
Results of initial field tests
Initial field tests of the fiber optic fluorometer equipped
penetrometer were conducted at a hazardous waste site in the
southeastern United States. The site, which dates back to the
1940's, had been used for several decades as a disposal area
for mixed petroleum wastes. In the mid-1980's a ditch was dug
around the site and a recovery system installed. A map of the
site showing locations of the CRTs is given in Figure 7. Figure
8 shows representative results from a transect paralleling the
recovery ditch (CRTs 30-37). The depth of sampling in this study
was limited to 30 ft by a hard limestone layer. Inspection of the
fluorescence profiles ind icates that hydrocarbon related fluores-
cence was detected at locations 30,32,33,35 and 36 but not at
location 34 or 37. These results illustrate how it is possible to
rapidly delineate the horizontal extent of the contaminated area
by making a series of CRTs at the site. Each CRT required
61
-------
Figure 7. Map showing locations of cone penetrometer
tests (CPTs) at test site.
approximately 20 minutes to complete. Detailed inspection of
the vertical structure in fluorescence profiles at the locations with
the highest fluorescence intensities (CPTs 30,32 and 36) shows
marked similarities. Highest intensities were observed at a
depth of approximately 15 feet with a secondary maximum at
about 10 feet and background levels at the surface and at the
bottom of each profile. Similarity in the vertical structure ex-
hibited by the fluorescence profiles at the three locations and
the covariance with measured soil characteristics supports the
hydrogeological consistency of the data. The observation that
CPTs 34 and 37 showed no measurable fluorescence suggests
that at this site naturally occurring organic material did not
contribute to measured fluorescence signals. In order to
facilitate interpretation, fluorescence and soil property data from
individual CPTs can be combined with position information and
transformed (Dynamic Graphics, Inc) into a 3-dimensional
gridded file for visualization on a minicomputer system. Figure
9 shows an example of. a 3-dimensional representation of the
fluorescence data from the CPTs at the sites indicated on the
map in Figure 7. For this example, fluorescence intensities have
been converted into diesel fuel equivalents using the linear
portion of the calibration curve presented in Figure 6.
Conclusions and future efforts
Efforts to date suggest that use of a fiber optic based fluorometer
system in conjunction with a cone penetrometer may be useful
for rapid delineation of subsurface petroleum hydrocarbon con-
tamination at hazardous waste sites. Laboratory and field
calibration of the fluorometer system using fuel products (diesel
fuel marine and JP-5) indicates that the fluorometer system is
quantitative for direct determination of these products in soil
Fluorescence Soil Class Fluorescence Soil Class Fluorescence Soil Class Fluorescence Soil Class Fluorescence Soil Class Fluorescence Soil Class Fluorescence
| (relative) | (relative) | (relative) | (relative) | (relative) | (relative) ; (relative)
its! mi nil nil ml ml ml
012345 0 10002000 012345 0 10002000 012345 0 10002000 ^ 0123*5 0 10002000 ( 01 2345 0 10002000 ^12345 0 10002000 ( Q1 2345 0 10002000
. -
•
j;::
u-
.
u-
il.il
NO
DATA
P1 — i~t~
HO
DATA
r . i — i—
]_ HOLE 2 _ |
C-3S-90
L--J L--J |_»°-J
Figure 8. Test data showing the use of the fiber optic fluorescence sensor for locating the boundaries of a hydrocarbon plume.
62
-------
Figure 9. Eample of 3-dimensional visulalion of soil contamination based on CPTdata. The volume shown represents areas that
had fluorescence intensities equivalent to 1000 ppm or more dieselfuel marine. The lines on the upper surface represent cultural fea-
tures (ditches and roads) present at the site.
(sands) for concentrations in the range of 100 ppm to 10000
ppm. At present, the greatest utility of the system is for rapid
screening for POL contamination in order to more precisely
locate contaminated zones, and thus significantly reduce the
number of monitoring wells required for site characterization.
The accuracy of converting measured fluorescence intensities
to concentration units will depend on how closely the product
used for sensor calibration emulates the product in the soil.
Experience in the field indicates that the optical fiber system is
rugged enough to withstand normal deployment procedures
with the penetrometer system and that the sapphire viewing
window appears to be self-cleaning, thereby avoiding memory
effects.
Efforts currently planned, or in progress, include: (1) rigorous
intercomparison of penetrometer field measurements with con-
ventional sampling and standard analytical methods, (2) char-
acterization of the effect of different soil types and
characteristics on system calibration, (3) enhancing the
capabilities of the sensor system for measuring compounds that
are excited at higher energies by replacing the N£ excitation
source with a Nd-YAG operating at the third and fourth har-
monics (355 and 266 nm).
References
1. Olsen, R.S. and J.V. Farr. "Site Characterization Using the
Cone Penetrometer Test." Proceedings of the ASCE Con-
ference on Use of In-situ Testing in Geotechnical Engineering.
Amer. See of Civil Eng., New York, N.Y. 1986.
2. Cooper, S.S., P.G. Malone, R.S. Olsen and D.H. Douglas.
"Development of a computerized penetrometer system for Haz-
ardous waste Site Soils Investigations." Rept. No. AMXTH-TR-
TE-882452, U.S. Army Toxic and Hazardous Materials Agency,
Aberdeen Proving Ground, MD. (1988), 58 pp.
3. Campanula, R. G. and P. K. Robertson, "State-of-the-art in
in-situ testing of Soils: Developments since 1978," Department
of Civil Engineering, University of British Columbia, Vancouver,
Canada, 1982.
4. LJeberman, S.H., S.M. Inman and G.A. Theriault. "Use of
Time-Resolved Ruorometry for Improving Specificity of Fiber
Optic Based Chemical Sensors." In: Proceedings SPIE Op-
toelectronics & Fiber Optic Devices & Applications, Environ-
ment and Pollution Measurement Systems. Vol 1172. (1989),
p. 94-98.
5. Inman, S.M., P.J. Thibado, G.A. Theriault and S.H. LJeberman,
Development of a pulsed-laser, fiber-optic-based fluorimeter:
determination of fluorescence decay times of polycyclic
aromatic hydrocarbons in sea water," Anal. Chim. Acta, 239,
(1990), p. 45-51.
63
-------
DISCUSSION
The following is a panel discussion in which questions were posed to the first three
authors of papers in the Chemical Sensors Session.
DICK GAMMAGE: Most of the data you showed was for sand. You're going
to have different quenching problems, different degrees of quenching for
different soils. Can that throw you out at all? Also, I thought the original intent
of this device was to be able to lower it directly into groundwater and take in water
measurements. And I'm wondering why your focus seems to be totally on the
headspace ai this stage?
STEPHEN LIEBERMAN: I'll talk about this soil type question. That's a good
question. It's something thai has been on our mind. We actually have a laboratory
study going right now where we're going to evaluate the effect of soil type on the
response of the sensor. One of the other considerations with soil type, and this was
something we visually observed, is if you have a sand the sample volume is going
lo be different than if you have a very fine grain clay or something like that. We
have not parameterized that or really documented what that effect is yet. but we
are looking at that. That's kind of one of the drawbacks of rushing some of the
stuff out in the field, just to see if you can get that fiber down there without
breaking in and some of ihose very basic questions. But we haven't ignored that.
FRED MILANOVICH: The answer to the second pan is a quite complicated
answer. The experience we've had is that headspace measurement is far and away
more reproducible. And since this is a result-driven technology, we want
something that works. When we designed the continuous probe the reagent is
now in contact with a membrane. When we wet it on the other side with water,
we have problems. In the original probe there was an air space, and you could
stick that probe into the water. With the membrane being teflon, the wetting
phenomenon was different than what has been exposed to the pyridine. So some
work would have to be done there. But I don't see there's a great liability lo stay
with headspace.
JOHN SCHABRON: How often do you have to recalibrate the probe? I guess
now that you can introduce solution into it. you can calibrate it more frequently.
Could you also address the issue that, with the two diodes, the red and the green,
you're not compensating for the difference in output of the two diodes as you
would if you had a single lamp and a monochromalor with two different
wavelengths.
FRED MILANOVICH: The calibration issue is a function again of a lot of
factors. If you make enough reagent and it's stored cold, you can go with the
calibration. We've gone months with the calibration. But if you mix a new
reagent, open a new bottle of pyridine. chemistries are different. So you'd have
to recalibrate.
MARY BETH TABACCO: Basically we found that you can adjust the output
from those two diodes to make them match, make one greater or one less. The
ability for the ratio to remain constant isn't dependent on the output from the
ditxles. In the graph that I showed you. the green output was lower. In fact, the
system electronics that we've built, the green is just about the same output value.
By adjusting the current lo the LED. you adjust the output value.
DeLYLE EASTWOOD: As some of you may know, there is a fiber optic
committee chaired by Dr. Tuan Vo-Dinh of Oakridge which is working on
developing the calibration standards, fluorescence standards and standards for
terminology. and collecting a data base for fiber optic chemical sensors. We use
the term fiber optic chemical sensors because as some of you know. Optrode is
a registered trademark. Dr. Vo-Dinh is giving a presentation on that at the
Pittsburgh conference in Chicago, Monday. March 4.1 will also chair a meeting
on luminescence at that conference.
There's been a lot of previous work on classification and identification of oils.
some of which is in the literature, and is the basis for a couple of ASTM methods.
My question is. do you plan to use another laser and fiber to measure BTX?
STEPHEN LIEBERMAN: Yes, but I'm not sure we're going to get down to
BTX. We did have plans to use a different excitation source. That should be
coming on line, should at least be available to us about the end of this month. That
will give us the 266 and 355 excitation. But I think benzene and others are even
excited at lower wavelengths. The thing we're bucking there is the transmission
down the fiber. As you know, the attenuation dramatically increases as you go
down in the UV. So right now the 337 is kind of a nice compromise between what
we can get down there and a wavelength that will excite some of the 2,3,4-type
ring compounds. But if we could get the energy down there, it would be real nice
to try to go 200 or so. But I don't see that happening right now. I think 260 is going
to be pushing it. Even at that, we're going to be brute forcing the energy down
there. So I think we may be approaching the damage threshold of the fiber, versus
what we can get out the other end.
GORMAN BAYKUT: I have a question about telling compounds apart in a
mixture. You gave an example of a mixture of three compounds. If you have a
high concentration of some compounds with a very low concentration of another
compound, do you have any problems with determining them just using the
slopes?
STEPHEN LIEBERMAN: We have not actually done experiments where
we'vejuggled concentrations of these different compounds and really determined
what the range of concentrations we're able to discriminate. Obviously that's a
concern. We've done a little bit of work using Lifetimes as a way to discriminate
different metal ions that complex w ith a particular indicator molecule. We've had
some success fining biexponential curves to those compounds. But again, we
haven't really pushed the limit by having tremendous differences in concentration.
Our current thinking is it's going to take a combination of techniques and maybe
a smart pattern recognition-type techniques. We may be looking a neural
networks as a way. But obviously, there's going to be some point in the
differences in concentration that you're going to be able to determine.
FRED MILANOVICH: In these experiments we actually prepared the solution
so that they'd give a similar initial intensities.
BRIAN PIERCE: I have four questions: (1) These indicators in your porous
meter, are these reactions reversible? (2) What are the polymers you're using in
your porous polymer monitor or sensors? (3) Have you considered waveguide
configurations? (4) How is it possible to construct these 3-D visualizations from
the finite number of points that you've sampled? What kind of assumptions go
into that?
MARY BETH TABACCO: We're working with both reversible and irrevers-
ible systems. The pH Optrodes. the ammonia sensors are all fully reversible.
Right now. for some of the other vapors sensors forhydrazines.carbon monoxide.
we have irreversible indicator systems. But as 1 mentioned, in e case of the
irreversible systems we've demonstrated that by monitoring the slope you can
look at real time changes in concentration. For example, with ethylene, we've
cycled concentrations from 100 ppb to 100 ppm and you basically can monitor
the change in the slope to pull out real time information.
Your third question was about waveguiding. And no. we've not considered that
approach here.
Concerning the actual polymers we're working with, we're using a variety of
polymer systems, both hydrophilic and hydrophobic. These are
methylmethacrylate systems with bis-acrylamide cross-linkers. The actual
formulation varies depending on the sensor. We have applied for a patent for the
pH Optrode under development. But as I mentioned, it is kind of a witch's brew
at this point.
STEPHEN LIEBERMAN: By the 3-D visualization I assume you mean the
fancy three-dimensional figure of field data. I'm not quite sure if I understand the
question. There's actually a lot of data points here that represents about 30
64
-------
DISCUSSION
pushes. We're firing that laser about once a second as we're pushing it into the
ground. So we're getting a point in the vertical about every two centimeters. Now
obviously you have to be careful in any kind of three-dimensional visualization
— it only represents reality as good as those contouring algorithms. I think the
proper way is to first plot out your raw data in cross-section or by profile. You
have to make sure that the visualization you generate by the more sophisticated
computer program reflects the reality of what you saw in those individual
profiles.
MARTY HARSHBARGER-KELLY: What is the software package you're
using on that Macintosh for data manipulation and who's the software
manufacturer?
FREDMILANOVICH: The software package is Lab View. It's all icon driven.
so no words are typed to do all that interfacing, just moving icons around. I
believe the software manufacturer is National Instruments.
BERT FISHER: Your instrument is measuring polyaromatic hydrocarbons, so
it's a bit misleading to say that you're measuring product, because you're
measuring some chunk of that. Also, this really shou Id be able to look at historical
spills. Have you looked at weathered materials, because the PAH's will hang
around. And my comment on the three-dimensional visualization is, it's a lot like
doing geology. You have great resolution in the vertical and you accept the
horizontal on faith. So it's like doing stratigraphy.
STEPHEN LIEBERMAN: As to your question regarding weathered product,
I showed you data from a Jacksonville site that has a rather checkered past. Those
deposits go back 30 or 40 years. Now in geological terms that may not be your
idea of a weathered product, but it's not a fresh product. Actually there's some
work Iknow out of the petroleum people that shows that those PAH spectra don't
seem to change very much as a function of time, at least with the PAH
components, but we don't have any real evidence. This is also sort of a brute force
method here. We're taking this thing out on the field and we're sticking it in the
ground. We don't know very well what's down there or what we're even looking
at. Personally I think it would be much nicer to go to some sites where we have
some more recent leaks from a tank farm or something like that where we could
put ourselves to a better test of whether wecan discriminate for instance JP5 from
diesel fuel. Hopefully we would also have information on how old the product
is and how long it's been in the ground.
BERT FISHER: That really was my concern, in that you would be seeing stuff
where there in fact was no product, but you were looking at a tremendous amount
of PAHs that had been hanging around for many years.
STEPHEN LIEBERMAN: That may be the case in that example.
PETER KESNERS: As I understand your apparatus, there's a membrane
permeation front on it. What sort of membrane types have you investigated? Do
you think it's feasible to measure pyridine in water with other membranes with
the sensor working the other way around?
FRED MILANOVICH: That's a real interesting plot. Our concern with the
membrane is to keep pyridine out of the water, so we have solicited help
anywhere we can. The current membrane that works the best is plumber's tape,
simple expandable teflon plumber's tape. And that's a result of trial and error
from attempts too numerous to mention. Probably 40 or 50 membranes have been
tried and plumber's tape is the best. We do have some proprietary technology
from companies that we aren't able to speak about yet that could exceed the
plumber's tape.
TODD TAYLOR: It seems to me that the cal ibration curve that you showed on
the screen is going to depend on quite a few things in addition to the soil type. It
seems to me it's going to depend on the water content, because water is going to
affect the amount of oxygen quenching going on in the soil. It's going to depend
on the oxygen concentration. Surface soils are known to contain a lot of humic
materials, and those materials naturally fluoresce. Their fluorescence, in fact,
depends on metal concentration in the soil. So it seems to me there are quite a few
factors which may be involved in looking at the fluorescence of the soil. And the
last question is not really a question. It's more the fact that I think that you have
a lot more work to do in characterizing your system.
STEPHEN LIEBERMAN: In the previous graph, I did show that we have
looked at varying the water content over from all the way dry to up to 10% in the
data I showed you and 25% with JP5. And seeing, somewhat surprisingly to me,
no real significant change in the response of the sensor. And so I think at least as
a first cut we have addressed that. As to the question of humics, we've also
considered that question. In the case of the Jacksonville data, we showed the fact
that we could leave the area that historically was the site where the contamination
was and get down the background fluorescence, at least at that site. I don't think
we have a problem with background fluorescence due to the humic substances,
although we have done some other tests where we've measured humic substances.
We' ve looked at their spectra characteristics and also looked at their decay times.
The decay times for the humic substances appear to be much shorter than what
we're seeing for the petroleum products. So if we do run into a case where we are
getting background fluorescence due to naturally occurring organics, there's at
least some hope that we may be able to resolve that based on their emission
curves.
I agree with you, there's tons of problems out there that need to be addressed and
looked at in more detail. Our approach has been one to let's push this thing out
in the field and see what happens. Let's fill in some of these questions later, when
we get some handle on what we are seeing. But I think that the true proof of this
thing, and this is where we stand right now, is going to be to do some of these
profiles and then rigorous validation of it: to collect samples and analyze them
by the more conventional methods. Obviously that needs to be done. And that's
going to be the thrust of our effort now.
65
-------
SPECTROELECTROCHEMICAL SENSING OF CHLORINATED HYDROCARBONS
FOR FIELD SCREENING AND JN. SHU MONITORING APPLICATIONS
Michael M. Carrabba, Robert B. Edmonds and R. David Rauh
EIC Laboratories, Inc.
111 Downey Street, Norwood, MA 02062
and
John W. Haas, m
Oak Ridge National Laboratories, Health and Safety Division
P.O. Box 2008, Oak Ridge, TN 37831-6383
ABSTRACT
The detection and identification of chlorinated hydrocarbon
solvents (CHS) have been demonstrated by combining the
principles of spectroscopy and electrochemistry. The
successful observation of die CHS is highly dependent on
the analysis procedure. The procedure is based on a photon
induced electrochemical reaction which is detected by
surface enhanced Raman spectroscopy (SERS) on
electrodes. The results and methodology of the technique
will be discussed.
INTRODUCTION
The importance of techniques to sense and monitor
chlorinated hydrocarbon solvents (CHS) are becoming
increasingly more important with the intensifying presence
of groundwater contaminations. Our research and
development effort is aimed at producing a commercial,
low cost, field portable instrument for the field screening/in
situ monitoring of contamination from chlorinated organic
solvents based on spectroelectrochemical fiber optic
probes. Some of the advantages of this technique for
monitoring a contamination site are cost, small size of
sampling probe, real-time analysis, the capability of sensing
in adverse environments, and the ability of using a central
detection facility. The technique has an advantage over
current fiber optic chemical sensing methods for
chlorinated organics in that the sensing only takes place
when the electrochemical device is turned on. This should
enable long term monitoring of a well to be accomplished
with only one probe.
Our monitoring system for chlorinated organic solvents is
based on the principle of combining spectroscopic,
electrochemical and fiber optic techniques (Spectro-
electrochemical Fiber Optic Sensing (SEFOS)). SEFOS is,
in principle, a generic technique which can be adapted to
many different sensing applications. With the SEFOS
technique, we use electrochemical methods to reduce the
chlorinated organic solvents into reactive intermediates.
The reactive intermediates can then react with the
"trapping" reagent and spectroscopic changes, such as
surface enhanced Raman spectra, are used to sense the
chlorinated organics at levels far below their detection
means by electrochemical methods alone. Previous work
(1) has shown the usefulness of using surface enhanced
Raman spectroscopy (SERS) for the detection of
groundwater contaminations and the technique has also
been successfully applied to fiber optics (2). However,
these past experiments have mainly been restricted to
aromatic hydrocarbons.
In this manuscript we will discuss some of the fundamental
aspects of using SERS for the examination of the following
chlorinated hydrocarbons or organochlorides: carbon
tetrachloride, 1,2-dichloroethane (DCE), chloroform and
trichloroethylene (TCE). Our interest in these compounds
stems from their existence in the groundwater at the
Department of Energy hazardous waste sites.
EXPERIMENTAL
The Raman spectroscopy system for conducting the SERS
experiments at EIC has been previously described (2). The
system used at Oak Ridge National Laboratory (ORNL) is
shown in Figure 1 and, with the use of an optical fiber for
excitation, represents a first step toward a remote fieldable
Raman system. Of note in the optical system is placement
of the laser line pass filter (BP) after the optical fiber to
remove interfering Raman scattering from the fiber itself
67
-------
(3). Both research groups employed high-resolution
spectrometers and diode array detectors for measuring
Raman scattering from similar spectroelectrochemical
cells. As shown in Figure 2A, each cell was fabricated from
a 3 x 6 x 3 cm quartz cuvette with O-ring joints fused into
three sides and the top. Electrodes were fed into the cell
through O-ring joints and consisted of Pt counter, Ag/AgCl
reference, and copper working electrode. The working
electrode was placed about 2 mm from the Oarge) face of
the cell between the two electrodes. This orientation
minimized the path length of incident and scattered light
through the sample solution and simplified alignment of the
electrode in the optical system. For transport/concentration
studies, a membrane could be sandwiched between die
spectroelectrochemical cell and a second cuvette with
matching O-ring joint fused into the bottom (Figure 2B).
The spectroelectrochemical procedures were first
developed at EIC and then used at ORNL. Electrochemical
roughening of polished copper electrodes, consisting of
high purity 1.0 mm copper wire, was achieved with an
oxidation/reduction cycle (ORC) from -0.6 to +0.2V in a
0.1M KC1 electrolyte at 25 m V/sec. Saturated solutions of
the chlorohydrocarbon solvents (CHS) in distilled water or
100 ug/ml solutions of CHS in 0.1 M KC1 were cycled
several times under the same conditions and optimum SERS
spectra were acquired at -0.2V on the cathodic sweep. All
cycling occurred under laser illumination at 625 nm at EIC
or 647 nm Krypton illumination at ORNL. The use of the
slightly different wavelengths for illumination and Raman
spectroscopy did not produce significantly different results
at the two labs.
RESULTS AND DISCUSSION
Our results confirmed previous experiments (1) which
indicated that carbon tetrachloride was not observable on
Ag substrates. In addition, we were unable to observe the
chlorinated hydrocarbons on Ag or Au substrates.
However, when we examined the chlorinated hydrocarbons
with a Cu electrode, we were able to observe the SERS
spectra of carbon tetrachloride (Figure 3) as well as the
SERS spectra of TCE, DCE and chloroform (Figure 4).
The best SERS spectra were obtained when the ORC cycle
was stopped during the reduction step at the potential of
zerocharge for Cu(-0.2V) (4). The observation of the SERS
spectra was also highly dependent on illumination during
the cycling. Previous work by Thierry and Leygraf (5) has
indicated the importance of illumination during the
electrochemical roughening of Cu electrodes to produce
Raman active sites.
The vibrational features in Figure 4 indicate that a reaction
is occurring on the electrode surface (see Table 1 for
vibrational assignments). From the spectra, it appears that
ring formation is occurring due to an electrochemical and/or
photochemical process. However, in our experiments no
SERS spectra of the CHS were observed unless the
electrode was illuminated during the reduction step and thus
a strictly electrochemical reaction can be ruled out
This "photo" induced result indicates the possibility of a
photoelectrochemical process. Copper oxides are known
to be p-rype semiconductors which eject electrons under
illumination (Equation 1) (6). The band gaps for the two
possible copper oxides are 2.0-2.6 eV (620-477 nm) for
Cu2O and 1.7 eV (730 nm) for CuO. These electrons can
then electrochemically reduce the chlorinated hydrocarbon
solvents.
Cu2O + X -> Cu2O(hV)
This electrochemical reduction is similar to a reaction
scheme for the electrochemical reduction of chloroform
which has been determined by Fritz and Kornrumpf (6) to
be:
CHC13 + 2e -> CHC12 + Cl
CHC12- + CHC13 -> CH2C12 + CC13
CC13 -> :CC12 + Cl
The formation of the dichlorocarbene during the
electrochemical reduction process would tend to form a ring
type structure (6). This ring type structure is indicated in
our SERS spectra with the strong band at 1380 cm'1.
A preliminary observation has indicated that the SERS
spectrum is only observable for a finite amount of time. The
result is either due to the degradation of the electrode or the
sample. If the electrode was replaced with a new SERS
surface and then placed in the same solution, the spectrum
was still not observable. This indicates that the chlorinated
hydrocarbons were being consumed during the experiments
in the small volume (10 ml) of analyte. Confirmation of this
result would indicate that the SERS on Cu surfaces is a
method which is capable of both sensing and removing the
chlorinated hydrocarbons from the solution.
To determine the cause of the disappearing SERS signal, a
series of SERS/GC experiments which determined the TCE
concentration before and after the SERS experiments were
performed. Saturated samples of trichloroethylene (TCE)
in 0.1M KC1 and distilled H20 were cycled in a sealed glass
SERS cell to prevent the possibility of outgassing of the
TCE. Samples of the saturated TCE solutions were
collected both before and after the electrochemical cycling.
These samples were analyzed on a Hewlett-Packard Model
HP 5730A Gas Chromatograph. Chromatograms were
recorded and the magnitudes of retention peaks were
examined for the TCE peak in the experiments. Large
spikes at the 45 second retention time were due to impurities
in the distilled water. The chromatograms showed that a
large amount of TCE was consumed during electrochemical
cycling. Figure 5 represents a typical "before" and "after"
chromatogram.
68
-------
Analysis of "before" and "after" chromatograms showed an
average consumption of 66% of the trichloroethylene
during the electrochemical cycling and SERS experiments.
This is consistent with our observation that a film was being
formed on the roughened copper surface of our working
electrode. The formation of a film also indicated the carbene
may be originating a radical induced polymerization.
Methods for determining the exact structure of the products
formed during electrochemical cycling are currently under
investigation.
CONCLUSION
The observation of a "photo" induced SERS process in the
analysis of the chlorinated hydrocarbon solvents has future
implications for environmental sensors. Previous to this
work it was thought that the CHS type compounds were not
observable by the SERS technique. Upon completetion of
our fundamental experiments, future work will concentrate
on the analytical applications of the process and the
development of field portable Raman instrumentation.
ACKNOWLEDGMENT
This work was conducted in part under a collaborative
research agreement (CR-90-003) between EIC and ORNL
(Martin Marietta Energy Systems). Financial support for
this work was derived in part from the Office of Health and
Environmental Research Division of the Department of
Energy under the Small Business Innovative Research
program.
REFERENCES
1. Carrabba, M.M., R.B. Edmonds and R.D. Rauh,
"Feasibility Studies for the Detection of Organic Surface
and Subsurface Water Contaminants by
Surface-Enhanced Raman Spectroscopy on Silver
Electrodes", Anal. Chem., 52,2259 (1987).
2. Carrabba, M.M., R.B. Edmonds, PJ. Marren and R.D.
Rauh, Proceedings of the First International Symposium
on Field Screening Methods for Hazardous Waste Site
Investigations, Las Vegas, Nevada, October 1988, p33.
3. Carrabba, M.M. and R.D. Rauh, "Apparatus for
Measuring Raman Spectra Over Optical Fibers", U.S.
Patent Application 07/442,235 (1989).
4. Bunding, K., J. Gordon, and H. Seki, "Surface-Enhanced
Raman Scattering by Pyridine on a Copper Electrode",
J. Electroanal. Chem., 184.405 (1985).
5. Thierry, D. and C. Leygraf, "The Influence of
Photoalteration on Surface-Enhanced Raman Scattering
from Copper Electrodes", Surf. Sci, 149.592 (1985).
6. Fritz, H. and W. Kornrumpf, "An Improved Cathodic
Generation of Dichlorocarbene", Liebigs Ann. Chem ,
2,1416(1978).
Figure 1. Experimental setup for "photo" induced SERS
experiments at ORNL. F = optical fiber, O - microscope
objective, CL- collimating lens, FL = focusing lens, P =
right angle prism, BP = laser line pass filter, BR = laser line
rejection filter, C = spectroelectrochemical cell.
(A)
0=0=0
Figure 2. Diagram of spectroelectrochemical cell. (A)
Top view showing 3 electrode ports and O-ring joint
opening in the top of the cell. (B) side view showing sample
reservoir attached to the top for membrane concen-
tration/transport studies. Only 2 of the 3 electrode ports are
visible. In both diagrams the arrows point along the optical
axis as shown in Figure 1.
69
-------
4OO 60O 8OO
WAVENUMBER
10OO 12OO
TCE "before"
TCE "after"
UL
Time (minute*)
Time (minutes)
Figure 3. The SER spectnim of a saturated solution in
water on a Cu electrode of carbon tetrachloride. The
spectrum has been smoothed for clarity.
Figure 5. Gas chromatograms of TCE solution before
and after the SERS experiment. Retention time for the TCE
peak was 2 minutes.
B
BOO BOO 1OOO 12OO
WAVENUMBER
14OO
Figure 4. The SER spectra of saturated solutions in water
on a Cu electrode of (A) trichloroethylene, (B)
1,2,-dichloroethane and (C) chloroform.
70
-------
Table 1
Major Raman/SERS Peak Positions (cm"1) and Vibrational Assignment for the Chlorinated Hydrocarbon Solvents
CCl,
Raman
227s
319s
462s
762 w
787 w
SERS
220 w
261 w
288 w
521 w
791m
1051 w
1089 w
CHC13
Raman
689s
760m
1218 w
SERS
526m
670 w
783s
1021 w
1056m
1151 w
1234 w
1313m
1352m
1381s
1465 w
1550 w
1581 w
DCE
Raman
656s
674m
755s
882 w
944 w
1055 w
1209 w
1306 w
1433 w
SERS
521m
782s
965 w
1024 w
1058m
1101 w
1148w
1239 w
1312m
1379s
1464 w
1509 w
1582m
TCE
Raman
628s
780m
842 w
930 w
1247m
1585s
SERS
524m
781s
862 w
963 w
1018 w
1055m
1105w
1167w
1237 w
1312m
1358m
1379s
1463 w
1505 w
1580s
Vibrational Assignment
Cu-C?
Cu-C?
Cu-C?
"chain expansion"
symmetric CC14 str.
CCl str., Cu-C stretch?
CCl str. - secondary CA
CCl str. - primary CA
symmetric CC13 str.
CCl str. - primary CA
CCl str. - primary CA
CCl str. - primary CA
CC skeletal str.
CC skeletal str., ring
"breathing"
in-plane CH deformation, CC
str., ring "breathing"
CC str., ring "breathing"
CC str., ring "breathing"
ring "breathing" - cyclopropane
type
CH2 twist and rock
CH2 twist and rock, in-plane
CH deformation
CH2 in-phase twist, CH2 twist
and rock, in-plane CH
deformation
CH deformation
ring str.,
CH2 deformation
CH2 deformation
symmetric C=C str. - cyclo
C=C str. - cyclobutene
C=C str. CA, 3 or C=C couple
str. - polyene
s - strong intensity, m - moderate intensity, w - weak intensity
CA = Chloroalkane, str. = stretch
71
-------
DISCUSSION
ARTHUR D'SILVA: In the E.I.C. experiments at what wavelength did you
measure the fluorescence?
MICHAEL CARRABBA: We're looking at the complete spectrum, in this case
a very simple proof of concept. We weren't trying to develop a highly skilled
system as the Livermore people have developed, or as the people at GEO-
Centers. We're proving the concept here. We just monitored the intensity under
the total fluorescence band.
ARTHUR D'SILVA: What is the excitation wavelength?
MICHAEL CARRABBA: The excitation wavelength was 514 nanometers. We
added an argon-ion laser. We believe we could use just about any of the
wavelengths from 488 up to about possibly 600, but we really didn't try the 600.
EDWARD POZIOMEK: In the experiment where you described the photon
induced reaction, did you utilize a base?
MICHAEL CARRABBA: In the electrochemical experiment you don't need
the base. We use it as our bench mark, and then put the electrodes in. I believe we
don't need the base, and that's probably the important point.
EDWARD POZIOMEK: If you had the opportunity to solve a technology
barrier, which one would you go after first in this area to move it faster?
MICHAEL CARRABBA: The implication of the dichlorocarbene, going after
a double bond, could be quite lucrative in the future. And we believe we can make
probe systems that have been coded right onto an optical fiber and a very simple
sensor. That's where I think we'd pursue it at this point. Basically we'd use some
particular dyes that when the dichlorocarbene attacks the double bond it breaks
the conjugation and the fluorescence disappears or new fluorescence appears.
That's the direction that we're working on right now.
72
-------
SURFACE ACOUSTIC WAVE (SAW) PERSONAL MONITOR FOR TOXIC GASES
N. L Jarvis, H. Wohltjen, and J. R. Lint
Microsensor Systems, Inc.
6800 Versar Center
Springfield, VA 22151
ABSTRACT
A demonstration model 4-sensor Surface Acoustic Wave
(SAW) Personal Monitor for Toxic Gases was designed and
built, with emphasis on minimizing the overall system
size, weight, power consumption, and complexity. The
completed demonstration unit contained four 158 MHz SAW
delay lines, supporting RF electronics, microcomputer
(microcontroller), a miniature pump, valve, gas transfer
lines, and a small scrubber to provide a clean, dry, air
source to establish sensor baseline frequencies. The
demonstration unit weighs approximately 2 pounds. The
projected size of the follow-on unit is expected to be 6" x
3" x 1". Unlike previous SAW vapor sensor arrays, which
utilized coatings that interact reversibly with specific
classes of toxic organic vapors, this SAW Personal Monitor
takes advantage of sensor coatings that react irreversibly
with toxic chemicals. Thus it can more easily and
effectively determine total exposure to a given toxic gas.
The following toxic inorganic gases were selected for study
with the demonstration system: HCI, NO2, SO2, NO2, H2S
and NH3. Coating materials were selected that react
irreversibly with each gas. The coatings were applied to
the SAW sensors and their performance evaluated for
exposure to a single gas. The results show that suitable
materials are available for use as dosimeter coatings for
SAW sensors. Thus the potential exists for developing an
effective SAW Personal Monitor for detecting and
monitoring each of the above gases, except NOg, at
concentrations well below the OSHA "action levels".
INTRODUCTION
In all areas of environmental monitoring, as well as
industrial hygiene, there is a need for smaller, more
sensitive, and inexpensive personal monitors (e.g.,
dosimeters) for toxic gases and vapors. For example,
personnel involved in field screening must be concerned
with their personal health and safety when working at a
field site, and may often require accumulated exposure data
for various toxic gases. SAW sensor technology, however,
is not limited to use in a Personal Monitor (e.g., a toxic gas
monitor that can be worn on clothing). The same sensor
technology could be extended to the development of small,
hand-held or in-situ monitors for a variety of field
screening applications.
There are a number of techniques currently being used to
acquire toxic exposure data, however, each have their
limitations. In the future large numbers of more effective
monitors will be required for the rapid and reliable
detection and/or monitoring of toxic gases and vapors at
ever lower concentrations, in response to increasingly
stringent state and federal health and environmental
regulations. Chemical microsensors have demonstrated the
sensitivities and physical properties needed to meet the
size, cost, and performance requirements of a new
generation of personal monitors, and should ultimately
find a wide range of applications within the industrial,
medical, and environmental communities (1 - 13).
Of the chemical microsensors that have been investigated to
date, SAW devices, which measure changes in mass when a
chemically specific surface coating adsorbs or reacts with
an appropriate gas, are the best characterized and the most
promising for rapid development. SAW devices have been
shown to respond in just seconds to selected vapors at
concentrations down to the parts per billion range for
specific organic chemicals. Because of their solid state
construction and compatibility with integrated electronics,
they can be easily incorporated into very small,
lightweight instruments, small enough to be worn on
clothing. The primary challenge remaining in the
development of SAW based microinstruments is the
development of more selective and sensitive SAW coatings
for specific gases and vapors. Other technical areas to be
addressed are the miniaturization of supporting electronic
components and the development of computer software to
facilitate sensor operation, data analysis, and data
reporting.
73
-------
OBJECTIVE
2. SAW Sensitivity and Selectivity
The objective of the present study was to demonstrate the
feasibility of developing a miniaturized Surface Acoustic
Wave (SAW) Personal Monitor with the size, sensitivity,
selectivity, reliability, and low power consumption
appropriate for wearing on clothing. To achieve this
objective it was necessary demonstrate that: (1) the SAW
sensors and necessary support electronics can be
sufficiently miniaturized; (2) chemically selective SAW
coating materials are available or can be developed for the
detection of a wide range of toxic gases; and (3) the SAW
sensors and their coatings can be sufficiently sensitive to
specific toxic gases to meet the requirements of field
screening, personal safety, and related monitoring
applications.
SAW SENSOR INSTRUMENTATION
1.
SAW Sensor Operating Principles
SAW devices are mechanically resonant structures whose
resonance frequency is perturbed by the mass or elastic
properties of materials in contact with the device surface.
Rayleigh surface waves can be generated on very small
polished chips of piezoelectric materials (e.g. quartz) on
which an interdigital electrode array is lithographically
patterned. When the electrode is excited with a radio
frequency voltage, a Rayleigh wave is generated that
travels across the device surface until it is "received" by a
second electrode. The Rayleigh wave has most of its energy
constrained to the surface of the device and thus interacts
very strongly with any material that is in contact with the
surface. Changes in mass or mechanical modulus of a
surface coating applied to the device produce corresponding
changes in wave velocity. The most common configuration
for a SAW vapor/gas sensor is that of a delay line
oscillator in which the RF voltage output of one electrode is
amplified and fed to the other. In this way the device
resonates at a frequency determined by the Rayleigh wave
velocity and the electrode spacing. If the mass of the
coating is altered, the resulting change in wave velocity
can be measured as a shift in resonant frequency. SAW
vapor/gas sensors are similar to bulk wave piezoelectric
crystal sensors, except they have the distinct advantages of
substantially higher sensitivity, smaller size, greater ease
of coating, uniform surface mass sensitivity, and improved
ruggedness. Practical SAW sensors currently have active
surface areas of a few square millimeters and resonance
frequencies in the range of hundreds of MHz. However,
SAW devices having total surface areas significantly less
than a square millimeter and resonant frequencies in the
gigahertz range are possible using modern
microlithographic techniques. Such devices would
ultimately increase device sensitivity as well as decrease
size. Most of the SAW vapor sensors reported in the
literature employ two delay line oscillators fabricated side
by side on the same chip, with one delay line used to
monitor the toxic chemical and the other to act as a
reference to compensate for changes in ambient
temperature and pressure.
A 158 MHz SAW device having an active area of 8 mm2'
will give a resonant frequency shift of about 365 Hz when
perturbed by a surface mass change of 1 nanogram. This
sensitivity is predicted theoretically and has been
confirmed experimentally. The same device exhibits a
typical frequency "noise" of less than 15 Hz RMS over a 1
second measurement interval (i.e. 1 part in 107). Thus,
the 1 nanogram mass change gives a signal to noise ratio of
about 24 to 1. For vapor or gas sensing applications, the
objective is to have the chemical selectively adsorb onto
the mass sensitive surface of the device. Chemically
selective coatings are used for this critical operation.
3. Selective Coatings
The operational behavior of a Surface Acoustic Wave device
can be very sensitive to changes in density, elastic
modulus, and viscosity of the surrounding medium;
however, SAW devices are not inherently sensitive to the
chemical properties of the medium surrounding the device.
When coated with a chemically selective thin film they can
exhibit remarkable sensitivity to small quantities of a
chemical vapor or gas. The development of such selective
coatings for toxic chemicals can take two directions, (1)
coatings that will selectively and reversibly adsorb a
selected vapor or gas by matching "solubility"
characteristics; and (2) coatings that react chemically and
irreversibly with a selected vapor or gas. SAW
selectivities in excess of 10,000 to 1 for certain toxic
chemical agents have been demonstrated using the
"solubility" approach. Much greater selectivities should
be possible using chemically reactive coating/vapor (gas)
combinations.
SAW INSTRUMENTATION DEVELOPMENT
1. Miniaturization of SAW Sensor Array and RF
Electronics
Ultimate miniaturization would be achieved by going to
hybrid circuitry, where the sensors and support RF
electronics could be reduced in size to a few cm2 or less.
Hybridization, however, will require a major engineering
effort and was beyond the scope of this study. The emphasis
was therefore on the selection and arrangement of the
discrete components and electronic packages to minimize
the size of the demonstration unit. The basic design of the
system is essentially the same as used in previous SAW
Vapor Monitors. The four coated SAW dual delay line
devices were mounted in small, gold 1C packages. The lids
of each package were modified with short, 1/16" ID, gold
plated gas inlet and outlet tubes to provide the toxic gases
access to the sensors. A fifth SAW dual delay line, sealed to
prevent exposure to the ambient environment, was place in
a separate package. In the demonstration unit, this fifth
device was used as a reference for all other sensors to
compensate for changes in temperature and pressure. The
output of the 4 SAW Sensor Array was integrated with a 4
channel frequency interface card to generate the measured
74
-------
frequency differences, Af, and with an onboard
microcomputer (microcontroller) for data analysis.
2. Instrument Configuration
The system was designed with three circuit cards: a sensor
card, a four channel frequency interface card, and a
microcomputer card. The entire instrument will fit in an
enclosure 4-3/4" x 8" x 3", allowing room for the
necessary pumps, valves and gas transfer lines. The
system was designed for either battery operation or with a
120 VAC 50-60 Hz power supply. 1/8" Swagelok
bulkhead fittings on the enclosure provided gas inlet and
outlet to the system. Except for the stainless steel
Swagelok finings on the front of the enclosure, all surfaces
in contact with the gas up to the SAW devices are either
Teflon or gold.
The four channel microcomputer controlled frequency
counter measures and reports the frequency of each SAW
sensor every two seconds while controlling the solenoid
valves by means of a solid state relay. For laboratory
evaluation of the demonstration model SAW Personal
Monitor for Toxic Gases, the counter output is provided on
a 9600 baud RS-232C serial communications line. For
better control and monitoring of the demonstration model,
and it's subsystems, all communication with the unit was
through the FtS-232 line and a personal computer with a
serial communication port. In a follow-on program, a
different communication scheme will be devised so that the
user will have the option of entering all instructions
directly on the instrument. Also, all concentration data
and/or signals will be presented on visual (LCD) displays
or by audio alarms mounted on the instrument enclosure.
There will still be the option of communicating with the
SAW Personal Monitor via a personal computer to retrieve
data stored in memory.
In the demonstration unit, the onboard Octagon SB S-150
microcomputer was programmed to control operation of
the system, but not for analysis of the sensor array data.
Development of a sensor array data analysis program is
planned for the follow-on effort. With the demonstration
unit, the performance of each SAW sensor, and it's coating,
was evaluated individually against a specific toxic gas.
There are a number of experimental variables that also
require computer control and or analysis. For example,
due to the possible adsorption/desorption of ambient gases
(especially water vapor) on the coatings, the computer
must continually determine the actual baseline for each
sensor, by intermittently providing clean, dry (filtered)
air to the sensors. The computer must also store
calibration data for each sensor and provide total exposure
values on demand and/or activate an alarm when certain
values are exceeded. Figure 1 provides a pictorial layout
of a SAW Array Personal Exposure Monitor.
SAW COATING SELECTION
1 . Selection of Candidate Coatings
A series of candidate materials was selected for screening
as coatings for the SAW devices. They were selected on the
basis of their known reactivity with the toxic gases chosen
for evaluation. The coatings selected for screening against
the reactive gases are given in Table 1 .
Table 1. Candidate Coating Materials for
SAW Sensors
Candidate Coating
Diphenylbenzidine
2,4, Dinitrophenylhydrazine
o-Toluidine
Triethylenediamine (TEDA)
Na[HgCl2] (hydrate)
Pb(C2H302)2 ' 5H20
CuSO4 • 5H2O
K[Ag(CN)2]
Ninhydrin
Reactive Gas
NO2
NO2
NO2
S02
SQ2
H2S
H2S
H2S
2 .
Polyvinylpyridine (PVP)
Coating of SAW Devices
HCI
Each of the above coatings was applied to two 158 MHz Saw
devices. Each SAW device to be coated was inserted into a
suitable connector mounted on a circuit board that
contained the necessary electronics to operate the device
and provide frequency signals to an external data aquisition
system. Prior to coating, each dual 158 MHz SAW device
was ultrasonically cleaned in isopropanol or chloroform,
dried in a stream of compressed dry, zero air, and
positioned in the coating apparatus. In all but a few
instances, the coatings were applied by a spray deposition
technique developed by Microsensor Systems. The primary
requirement is that the coating material must be soluble in
a volatile solvent. Zero air was used to generate a fine mist
of the specific coating solution. A mask was placed over the
SAW device so that only the interdigitated delay lines were
coated.
The quantity of coating material deposited on each delay
line was closely monitored by the computer data system
which reported the mass of material deposited as an
increase in frequency, Af. The amount of coating material
applied was held closely to 250 KHz + 50 KHz. The
frequency shift, Af, corresponds to coating thickness,
assuming uniform surface coverage. Once the coatings
were applied, the SAW devices were covered and stored in a
low humidity (< 10% RH) environment until ready for
testing. As the candidate coating materials given in Table 1
are generally hygroscopic, it can be assumed that a certain
amount of water will be associated with each coating and
must be considered in subsequent gas interactions.
75
-------
SAW ARRAY
PERSONAL EXPOSURE MONITOR
PICTORIAL LAYOUT
(PHASE
REPLACEABLE
SAW SENSORS
(4 UNDERNEATH
SCREW-ON LID)
MICROCOMPUTER
BOARD
6VIS I 2 Ah
4x2 X I 807.
RECHARGEABLE
BATTERY PACK
AMBIENT
VAPOR
INLET
Figure 1. Pictorial Layout of SAW Array Personal Exposure Monitor
76
-------
3. Screening and Selection of Coatings for SAW Test
and Evaluation
The following criteria were established to define a
successful candidate material: (1) that a coating give a
frequency shift equivalent to a 100:1 signal to noise ratio
when exposed to the toxic gas at a concentration of
approximately 100 ppm for 1 minute or less; and (2)
that the coating react irreversibly with the test gas. With
a baseline noise level of approximately 15 Hz, a 100:1
signal to noise ratio would be equivalent to a frequency
shift on the order of 1500 Hz. Thin film coatings showing
less response would not have sufficient sensitivity nor
capacity to be useful in field monitoring applications.
A calibrated cylinder of each of the test gases (NO2, SO2,
HCI, H2S, NHa) in air was obtained from the Scott
Specialty Gas Co. The concentration of each gas source
was:
Toxic Gas Source Concentration
HCI 103.3 ppm
NHs 106.5 ppm
H2S 100.6 ppm
NQ2 108.0 ppm
SQ2 102.5 ppm
By simple dilution of the compressed gas with clean, dry,
zero air, a steady state concentration at any value less than
100 ppm could be easily prepared. A constant gas flow
rate of 200 cc/min was maintained. A valve was arranged
so that clean air, or a known concentration of the specific
test gas, could be alternately delivered to the sensor. A lid
with 1/8" gold gas inlet and outlet tubes was placed over
the device and was connected to the output of the gas
dilution chamber. The frequency output of the dual delay
lines could be monitored using a small frequency counter.
In the tests, a coated SAW device was first exposed to clean,
dry air at 200 cc/min to obtain a steady baseline
frequency. The valve was then turned to expose the sensor
to a known concentration of the toxic gas, at the same flow
rate, for a pre-determined period of time. The sensor was
then exposed once again to clean, dry air to establish a new
baseline. If the clean air baseline, after exposure to the
toxic gas, was significantly different from the initial clean
air baseline, it was assumed the change in frequency was
due to an increase in coating mass resulting from the
irreversible reaction with the challenge gas. If there was
no significant change in SAW frequency, the device was
exposed to higher gas concentrations for longer periods of
times. If there was still no permanent change in baseline,
it was assumed there was no reaction and that the coating,
in its present form at least, was ineffective. All tests were
performed with dry air, unless otherwise specified in the
text.
The results of the initial screening tests are given in Table
2. They show that for each toxic gas there was at least one
coating that gave an acceptable response. However, in
several instances there were rather unexpected results.
For example, NC-2 did not appear to react at all with 2,4
Dinitrophenyl hydrazine unless there was a relatively high
moisture content (= 80% RH) in the carrier gas. It was
also surprising that H2S did not react readily with the lead
acetate coating, even though we have observed this surface
reaction in a previous study. Copper sulfate seemed
unreactive initially, however, after repeated cycling it did
react to give a very large and permanent frequency shift.
The reaction, or lack of it, in each case may depend to a
large extent upon the amount of water present in the film.
Table 2. Results of Initial Coating Screening
Test
(Thickness of all coatings approx. 250 Hz)
Coating
Diphenylbenzidine
2,4, Dinitrophenyl
hydrazine
o-Toluidine
TEDA
Na[HgCl2]
Pb(C2Hs02)2*'
CuS04"**
Ninhydrin
CoCl2
PVP
Af Stable
Conc./Time (Hz) Reaction
NC-2 50 ppm/60 s. 900 No
NO2 50 ppm/60 s. 2,800 Yes
NC>2 50 ppm/60 s. <100
SO2 50 ppm/60 1,000 Yes
SO2 50 ppm/60 s.
H2S
H2S 50 ppm/60 s. 2,000 Yes
50 ppm/60 s. 100
50 ppm/ 20 s 2,700 Yes
HCI (known to react)
Reacted only in presence of high RH
* * Reacted in a previous study, but now
Reaction occurred after repeated H2S exposure
Based on the results of Table 2, the following coatings were
selected for more careful evaluation. 2,4 Dinitrophenyl-
hydrazine was not used for NO2- Rather TEDA was used for
both SO2 and NO2-
Toxic Gas
HCI
NO2andS02
H2S
Coating Material
Polyvinylpyridine (PVP)
Triethylenediamine (TEDA)
Copper sulfate (CuS04)
Cobaltous chloride (CoCl2)
TEST AND EVALUATION OF SAW SENSORS AS MONITORS FOR
TOXIC GASES
1.
Coating of SAW Sensors
The coating procedure used was the same as described
above. Both SAW delay lines on each device were coated
simultaneously, and the amount deposited was measured
and recorded. The identification number of each device and
the coating mass (in terms of frequency shift, Af) are
given in Table 3. The coatings applied are very thin, on the
order of a micron or so in thickness, on the average.
2. Evaluation of SAW Sensors as Monitors for Toxic
Gases
The frequency difference, Af, of each SAW device being
tested was input to a Apple Macintosh computer where the
data was collected and displayed. The test system evaluated
77
-------
only one sensor at a time against a single toxic gas. Even
though each of the coating materials being tested could very
likely react with more than one gas, binary gas mixtures
and interference studies were not included in this
preliminary investigation. Interference studies will be a
part of the follow-on study, using multiple sensor arrays
and other techniques to address the problem of sensor
specificity.
The gas dilution chamber was again used to deliver known
concentrations of each test gas to the SAW sensors at a
constant flow rate of 200 cc/min at ambient pressure, and
a constant "baseline" frequency established for each SAW
device by exposing it to a clean, dry air stream. Once a
constant baseline frequency was established, the sensor
was exposed to a predetermined "dose" of the selected toxic
gas. The size of the dose could be varied from 10 to 100
ppm over any selected time interval. After exposure to the
toxic gas, the sensor was again exposed to clean, zero air
until a new baseline frequency was established. The
difference between the initial baseline and the final
baseline was taken as the frequency shift due to the
irreversible reaction of the toxic gas with the coating
material. The magnitude of this frequency shift could be
correlated with the amount of toxic gas interacting with the
sensor.
The intent of the tests was to quickly look for order of
magnitude changes in frequency and general reproduci-
bility of performance when exposed to moderate changes in
gas concentrations; i.e., to identify coatings that could be
used in a more comprehensive follow-on development
program. This study did not include a careful
characterization of each coating reaction. In any event an
accurate characterization of the surface reactions would be
difficult without a more careful control of trace water,
both in the hygroscopic coating materials and the gas
delivery system.
3.
Exposure of NHs to CoCl2 Coated SAW Sensor
The SAW devices were at ambient temperature and thus
subject to the room temperature fluctuations (= 25° +/-
1°C). Although a reference SAW device was used to
compensate for both temperature and pressure changes,
the compensation is not exact, and may have caused some
small, random drift in device background frequency. These
slow changes occurred in cycles of many minutes and thus
did not adversely effect the measurements. Even though a
number of the coating materials have a small volatility,
the signal drift reflected "apparent" increases as well as
decreases in weight. Thus volatility did not have a
measurable effect on the measurements. Once a device was
equilibrated with the laboratory environment (temper-
ature and pressure) the slow baseline drift was usually on
the order of ± 50 Hz. In addition to temperature changes
and the possibility of volatility, the baseline drift may also
be due in pan to changes in gas flow rate (due to changes in
flow through the non-precision needle valve used to set the
flow rate). Even with the small observed background
drift, the following data show that system performance was
excellent and clearly able to detect and monitor changes in
SAW frequency upon exposure to the challenge gases.
Sensor drift will be corrected for in the follow-on
Personal Monitor development program.
An example of data for the exposure of ammonia to the
CoCl2 coated SAW devices is shown in Figures 1. An
exposure of 20 ppm NHs for 20 seconds was selected for
testing the CoCl2 coated sensors. When the NHs was
introduced, there was a large initial decrease in SAW
frequency followed by a rapid increase. Each point on the
curve corresponds to a 2 second time interval. After 20
seconds, when the clean air at 200 cc/min was again
introduced, Af continued to increase through a small
maximum and then level off to a new, higher, baseline
value. The initial negative "spike" in the Af vs time plot
may be due in part to disruption and re-establishment of a
constant gas flow rate, while the subsequent increase in Af
most probably results from both adsorption and reaction of
the NHs witn tne CoCl2 coating. The maximum may result
from a more gradual desorption of non-reacted NHs from
the coating. The equilibrium frequency shift values for all
devices are shown in Table 4.
160000
0)
-
CT
o
150000 -
140000 -
130000
.
V
200 400 600 800 1000 1200 1400
Time (seconds)
Figure 1. Frequency Shift (Hz) vs. Time for Repeat
Exposure of CoCIa Coated SAW Device
(9024-11) to 20 ppm NHs for 20 Sec.
Table 3. Thickness of SAW Device Coatings
Coating Material
PVP
CuSO4
CoCl2
TEDA
Coating Thickness (KHz)
Device Number Side "A"
9024-1
9024-2
9024-3
9024-7
9024-8
9024-9
9024-10
9024-1 1
9024-12
9024-4
9024-5
9024-6
255
198
198
149
150
196
136
1 12
106
149
178
300
76
-------
Table 4. Frequency Shifts for CoCl2 Coated SAW
Devices Upon Repeated Exposure to
20 ppm NH3 for 20 seconds
158000
Device Number
9024-10
(Coating 112 KHz)
9024-11
(Coating 136 KHz)
9024-12
(Coating 106 KHz)
Exposure
Frequency Shift
a. - d. (dose optimization test)
e,
f.
a.
b.
c.
a.
b.
c.
d.
e.
f.
1,200 Hz
OHz
4,000 Hz
4,000
1,000
2,600
2,000
1,200
1,600
2,000
Hz
Hz
Hz
Hz
Hz
Hz
Kz
OHz
From the data in Table 4 it is evident that CoCl2 coated
SAW devices show large (Kilohertz), irreversible shifts
in frequency when exposed to small doses of ammonia, and
that with continued exposure the coatings saturate as
expected. Even allowing for the variation is response of
the different sensors, the sensitivity of the CoCl2 coatings,
i.e., those with some residual capacity, is on the order of 5
to 10 Hz/ppm/sec. Considering that the background noise
level of the SAW sensors is on the order of 15 Hz, a ten
seconds exposure of a sensor to 1 ppm NHs would give a
signal of better than 50 Hz, at least three times the
background noise. Thus the CoCl2 coatings have more than
enough sensitivity to detect ammonia at concentrations
below the OSHA Exposure Limit of 50 ppm NHs for an 8
hour weighted average.
4. Exposure of CuSCU Coated SAW Sensor to H2S Gas
The test procedure was essentially the same as described
above. Typical results are shown in Figure 2 for device
9024-7. H2S shows a decrease in SAW frequency with
exposure rather than an increase in Af as observed with
the reaction of NHs with the CoCl2 Also, there was no
initial "spike" in Af when the challenge gas was introduced.
Upon repeated exposure, the frequency shifts became
progressively smaller, due to saturation of the reactive
sites of the CuSO4 coating.
The Af values for the CuSCM coated sensors 9024-7 and
9024-8 are given in Table 5. SAW device 9024-9
apparently became defective during the coating process.
SAW device 9024-7 was exposed five times to 20 ppm of
H2S for 20 seconds. With the initial dose of H2S, Af
decreased by 1,400 Hz. The second exposure decreased Af
by only 400 Hz. Subsequent doses caused essentially no
further change in Af. Thus the CuSC>4 coatings were
essentially saturated by a single 20 ppm dose of H2S for
20 seconds.
5
~
—
en
157000 -
156000 -
155000
]?
-,
0 200 400 600 800 1000 1200 1400
Time (seconds)
Figure 2. Frequency Shift (Hz) vs. Time for Repeat
Exposure of CuSO4 Coated SAW Device
(9024-7) to 20 ppm H2S for 20 Sec.
Table 5. Frequency Shirts for CuS04 Coated SAW
Devices Upon Exposures to 20 ppm H2S
for 20 seconds
Device Number
9024-7
(Coating 149 KHz)
9024-8
(Coating 150 KHz)
9024-9
(Coating 196 KHz)
Exposure Fjeguencv Shift
a.
b.
c.
i'..
e
a.
1,400 Hz
400 Hz
100 Hz
0 Hz
OHz
1,400 Hz
(device defective after coating)
Thus the CuSCvj coated SAW devices, like the CoCl2 coated
devices, do give large (KHz), irreversible shifts in
frequency when exposed to small doses of an appropriately
reactive gas, and that with continued exposure the coatings
saturate as expected. The sensitivity of a newly prepared
CuSO4 coating is on the order of 3 to 4 Hz/ppm/sec. With
background noise on the order of 15 Hz, a ten second
exposure to 1 ppm H2S would give a signal of around 30 to
40 Hz, equivalent to a signal to noise ratio of 2:1. The
detection limit of this coating is thus also is well below the
OSHA Exposure Limit of 20 ppm H2S for an 8 hour
weighted average.
5. Exposure of TEDA Coated SAW Sensor to SO2 Gas
The procedure used to test the TEDA coated SAW sensors
with S02 was the same as described above. Typical results
are shown in Figure 3 for device 9024-6. The results for
device 9024-5 were similar. SAW device 9024-4 was
reserved for testing with N02, which was expected to react
with TEDA in much the same way as SO2. A rather
unexpected behavior was observed when the TEDA coated
devices were initially exposed to SO2- For the first few
79
-------
exposures of 20 ppm SO2 (20 seconds), the coatings did
not respond significantly. After several repetitions,
however, the coatings did begin to respond with positive
shifts in Af with the continuing exposure. Thus it appears
there was a "conditioning" period, after which the coatings
began to respond. The "conditioning" must be associated
with some chemical change in the coatings upon exposure to
the test gas, or to the zero air, most likely involving
associated water. As each device, after being coated, was
covered with a close fitting lid (but not hermetically
sealed) and stored in a = 10% RH environment, they must
have adsorbed some water vapor (or perhaps another
ambient gas) which was subsequently desorbed from the
coatings by the dry (< 1% RH) zero air and/or the dry
sample (S02) air. This "conditioning" or "ageing" effect
was not further explored at this time, but will of necessity
be investigated in the follow-on study in order to provide
coatings that behave predictably and reproducibly.
280000
3
270000 -
260000 -
250000
0 200 400 600 800 1000 1200 1400
Time (seconds)
Figure 3(a). Frequency Shift (Hz) vs. Time for Repeat
Exposure of TEDA Coated SAW Device
(9024-6) to 20 ppm SO2 for 20 Sec.
(First exposure, a)
268500
£ 267500 -
266500 -
265500 -
264500
0 200 400 600 800 1000 1200 1400
Time (seconds)
Figure 3(b). Frequency Shift (Hz) vs. Time for Repeat
Exposure of TEDA Coated SAW Device
(9024-5) to 20 ppm SOa for 20 Sec.
(Exposures d to h)
After the initial induction period, the frequency shift vs
time plot in both Figure 3(b) shows an increase in the
SAW baseline with each 20 second dose of SO2, after the
initial "spike" in Af. Device 9024-5 was allowed to stand
in the test apparatus for approximately two hours with
continuous exposure to zero air, before the run. Even so,
it wasn't until exposure f that the device began to respond.
Somewhat similar behavior was observed for device
9024-6, however the conditioning period was much
shorter. For both device 9024-5 and 9024-6, once the
coatings became reactive, the shifts in frequency were
regular and irreversible.
The frequency shifts are given in Table 6. The data clearly
show the induction period during which there was no effect
of SC-2 exposure, and the subsequent increases in Af when
reaction began to occur. If we assume an average response
of 1,200 Hz for device 9024-5 and 1,800 Hz for device
9024-6, the sensitivities are approximately 3 and 4.5
Hz/ppm/sec, respectively. The coating on device 9024-6
was a third again the mass of the coating on 9024-5 (300
KHz to 178 KHz), thus one would expect the sensitivity to
SO2 to be a third again as high, which was observed. Thus
the two coated devices had essentially equivalent
sensitivities.
Table 6. Frequency Shifts for TEDA Coated SAW
Devices Upon Repeated Exposure to
20 ppm SO2 for 20 seconds
Device Number
9024-5
(Coating 178 KHz)
9024-6
(Coating 300 KHz)
Exposure
a.
b.
c.
d.
e.
f.
g.
h.
a.
b.
c.
d.
e.
f.
Frequency Shift
OHz
OHz
OHz
OHz
OHz
800 Hz
1,400 Hz
1,000 Hz
OHz
OHz
200 Hz
1,600 Hz
2,000 Hz
1,800 Hz
With sensitivities of about 3 to 4 Hz/ppm/sec, depending
upon coating thickness, and a background noise level of 15
Hz for the SAW devices, the sensors should ultimately
detect concentrations of S02 as low as 1 ppm within 10
seconds at a signal to noise ratio of about 2:1. With this
sensitivity, these coatings should easily detect SC>2 at or
below the OSHA Exposure Limit of 5 ppm SO2 for an 8 hour
weighted average.
-------
6. Exposure of TEDA Coaled SAW Sensor to NO2 Gas
It was anticipated that TEDA would respond to NO2 in much
the same manner as to SO2; however, the data for the one
available sensor showed quite different behavior. First, no
conditioning period was observed. The first 20 second dose
of 20 ppm NO2 gave a relatively small but definite
increase in SAW frequency which apparently saturated the
sensor, as no further increase in Af was observed with
additional exposure to N02- The frequency shift data are
given in Table 7. The baseline shift of approximately 350
Hz for an exposure of 20 ppm NO2 for 20 seconds, is
equivalent to about 1 Hz/ppm/sec, well below the
sensitivity to S02- With a sensitivity of approximately 1
Hz/ppm/sec, and a background noise level of 15 Hz, the
TEDA coated sensors would have to be exposed to 1 ppm NO2
for over 30 seconds to give a 2:1 signal to noise ratio. In
addition, the film apparently has a very low capacity for
N02 (i.e., saturating at a very low exposure
concentration). TEDA is therefore of only marginal utility
as a dosimeter coating for NO2-
Table 7. Frequency Shins for TEDA Coated
SAW Devices Upon Repeated Exposure
to 20 ppm NO2 for 20 seconds
Device Number Exposure
9024-4 a.
(Coating 149 KHz) b. - g.
Frequency Shift
350 Hz
OHz
7. Exposure of PVP Coated SAW Sensors to HCI Gas
Device 9024-1 was given 5 separate exposures to 20 ppm
of HCI for 20 seconds, over approximately a 30 minute
period, with no apparent reaction of the HCI with the PVP.
We know from previous studies that surface films of PVP
do react with HCI, thus the lack of response must be
similar to the "conditioning" period observed for S02 gas
on TEDA. To accelerate the reaction, the PVP coated device
9024-1 was exposed to a higher concentration of HCI
(100 ppm) for 2 minutes. The result was a very large
increase in Af, over 30,000 Hz in the 2 minute period, as
shown in Table 8. A second large dose (100 ppm over a
60 second period) further increased Af by only 4,800 Hz,
indicating that the PVP coating was approaching saturation.
The estimated sensitivity, based on the 30,000 Hz shirt is
about 3 Hz/ppm/sec.
Device 9024-2 was exposed to repetitive doses of HCI at a
concentration of 25 ppm for 20 seconds. The results given
in Table 8 indicate no conditioning period was needed. The
very first exposure gave an increase of about 900 Hz and
appeared to be stable with time. Subsequent exposures also
increased Af, until the film began to saturate. Sensitivity
based on the initial exposure is about 2 Hz/ppm/sec.
Device 9024-3 did require a conditioning period when
exposed to 25 ppm HCI for 20 seconds. HCI exposures
were increase to 50 ppm for 30, 60 and 90 seconds,
before an increase in Af was observed. With the final
exposure, a frequency increase of approximately 6,400 Hz
was observed.
Table 8. Frequency Shifts for PVP Coated SAW
Devices Upon Repeated Exposure to HCI
Frequency
Device Number Exposure Shift
9024-1 a.(20 ppm 20 sec) 0 Hz
(Coating 255 KHz) b.(20 ppm 20 sec) 0 Hz
c.(20 ppm 20 sec) 0 Hz
d.(20 ppm 20 sec) 0 Hz
e.(20 ppm 20 sec) 0 Hz
f.(100 ppm120 sec) 30,000 Hz
g.(100 ppm 60 sec) 4,800 Hz
9024-2 a.(25 ppm 20 sec) 900 Hz
(Coating 198 KHz) b.(25 ppm 20 sec) 600 Hz
C.(25 ppm 20 sec) 400 Hz
d.(25 ppm 20 sec) 600 Hz
e.(25 ppm 20 sec) 400 Hz
f.(25 ppm 20 sec) 200 Hz
9024-3 a. (25 ppm 20 sec) 0 Hz
(Coating 198 KHz) b.(25 ppm 20 sec) 0 Hz
c.(25 ppm 20 sec) 0 Hz
d.(50 ppm 30 sec) 0 Hz
e.(50 ppm 60 sec) 0 Hz
f.(50 ppm 90 sec) 6.400 Hz
The sensitivities of the PVP coated SAW devices were in the
range of 1 to 3 Hz/ppm/sec. Device 9024-1, with the
greatest apparent sensitivity (3 Hz/ppm/sec), had the
highest, coating mass, as would be expected. Thus the
results for the three devices are consistent. With a
sensitivity of 1 to 3 Hz/ppm/sec, a sensor would have to
be exposed to 1 ppm HCI for 10 to 30 seconds to give a 2:1
signal to noise ratio. The PVP films do appear to have a
high capacity for HC1, as evidenced by the 30,000 Hz shift
for device 9024-1. Considering that the OSHA Exposure
Limit is 5 ppm HCI for an 8 hour weighted average, the
PVP coating should be considered a good candidate for
further development as a coating for monitoring acid gases.
CONCLUSION
In the evaluation of the various SAW coatings it was found
that for each toxic gas, except NO2, a relatively large,
easily measured SAW response was observed when an
appropriate coating was exposed small concentrations.
The measured sensitivities show that each toxic gas studied
(except NO2) could be detected by a SAW sensor well below
the "action level" set by OHSA, when monitored for a
period of one minute or less. The candidate coatings, toxic
gases, and the respective OSHSA exposure limits, are:
OS'HA Exposure
Limit - 8 hour
Candidate Coating Toxic Gas Weighted Ave.
polyvinylpyridine (PVP) HCI 5 ppm
triethylenediamine (TEDA) NO2 and SO2 5 ppm
copper sulfate (CuSO4) H2S 20 ppm
colbaltous chloride (CoCl2) NHs 50 ppm
The study thus successfully achieved it's objective of
demonstrating that: (1) the SAW sensors and necessary
support electronics can be appropriately miniaturized;
81
-------
(2) a number of successful coatings are readily available
and others can certainly be identified in the literature, or
developed, for additional toxic gases; and (3) SAW sensors
are sufficiently sensitive to meet OHSA requirements, at
least for the toxic gases selected for this demonstration
study. A number of technical problems and/or potential
limitations of the technology were identified and
approaches suggested for their solution. Based on the
results of this program, we conclude that a prototype
Surface Acoustic Wave Personal Monitor for Toxic Agents
could be readily developed in a follow-on program. In
addition to use as a Personal Monitor, such a small,
sensitive and rugged solid state instrument could possibly
find other applications in the field screening for toxic
chemicals. In all applications however, the usefulness of
SAW sensors will increase with the continued development
of more sensitive and selective device coatings.
ACKNOWLEDGEMENT
This research was supported by the Department of Health
and Human Services, Public Heath Service, Small Business
Innovation Research (SBIR) Program, under Phase I Grant
No. 1R43 ES5039-01A1.
REFERENCES
1. H. Wohltjen and R.E. Dessy, "Surface Acoustic Wave
Probe for Chemical Analysis I. Introduction and
Instrument Design", Anal. Chem., 51(9), 1458-
1464 (1979).
2. H. Wohltjen and R.E. Dessy, "Surface Acoustic Wave
Probe for Chemical Analysis II. Gas Chromatography
Detector", Anal. Chem., 51(9), 1465-1470 (1979).
3. H. Wohltjen and R.E. Dessy, "Surface Acoustic Wave
Probe for Chemical Analysis III. Thermomechanical
Polymer Analyzer", Anal. Chem., 51(9), 1 470-
1475 (1979).
4. H. Wohltjen and H. Ravner, "The Determination of the
Oxidative Stability of Several Deuterated Lubricants
by an Electronic Gas Sensor", Lubrication
Engineering, 39(11), 701-705 (1983).
5. (Invited) H. Wohltjen, "Chemical Microsensors and
Microinstrumentation", Analytical Chemistry,
56(1), 87A-103(1984).
6. H. Wohltjen, "Mechanism of Operation and Design
Considerations for Surface Acoustic Wave Vapor
Sensors", Sensors and Actuators, 5 (4), 307-325
(1984).
7. H. Wohltjen, W. R. Barger, A. W. Snow, and N. L.
Jarvis, "A Vapor Sensitive Chemiresistor Fabricated
with Planar Microelectrodes and a Langmuir-Blodgett
Organic Semiconductor Film", IEEE Trans, on Electron
Devices. ED-32, No. 7, 1170-1174 (1985).
8. W.R. Barger, J.F. Giuliani, N.L. Jarvis, A.Snow, and H.
Wohltjen, "Chemical Microsensors- A New Approach
for the Detection of Agro Chemicals", Environ. Sci.
Health, B20(4), 359-371 (1985).
9. W.R. Barger, A.W. Snow, H. Wohltjen, and N. L.
Jarvis, "Derivatives of Phthalocyanine Prepared for
Deposition as Thin Films by the Langmuir-Blodgett
Technique", Thin Solid Films, 133, 197 206 (1985).
10. A.W. Snow, W.R. Barger, M. Klusty, H. Wohltjen, and
N.L. Jarvis, "Simultaneous Electrical Conductivity and
Piezoelectric Mass Measurements on Iodine-Doped
Phthalocyanine Langmuir-Blodgett Films", Langmuir,
2, 513-519 (1986).
11. D.S. Ballantine, S.L. Rose, J.W. Grate, and H.
Wohltjen, "Correlation of SAW Coating Responses with
Solubility Properties and Chemical Structure Using
Pattern Recognition", Anal. Chem. 58, 3058 (1986).
12. G. S. Calbrese, H. Wohltjen, and M.K. Roy, "Surface
Acoustic Wave Devices as Chemical Sensors in
Liquids", Anal. Chem. 59, 833 (1987).
13. J. W. Grate, A. W. Snow. D. S. Ballantine, Jr., H.
Wohltjen, M. H. Abraham, R. A. McGill, and P. Sasson,
"Determination of Partition Coefficients from Surface
Acoustic Wave Vapor Sensor Responses and
Con-elation with Gas-Liquid Chromatographic
Partition Coefficients", Anal. Chem. 60, 869 (1986).
82
-------
DISCUSSION
WILLIAM BOWERS: You showed some data on individual sensor responses
for single exposures. Have you done any interference effects on some of these?
I am glad to see you're going to resonators now.
N. LYNN JARVIS: We did no interference studies in this particular program.
You could probably tell that many of the coatings used would respond to more
than one vapor. These were not selective coatings in that sense. Selectivity is
much more difficult to get. That's why we end up using an array of sensors to get
the selectivity. Resonators are much, much nicer.
MICHAEL CARRABBA: When you put the coating on these SAW devices,
and the coating goes over electrodes, is the area on the whole surface sensing the
weight or is it just the area between the electrodes, or the area on the electrodes'?
N. LYNN JARVIS: The whole area surface senses the weight. The wave will
cover most of the surface. Most of the surface is sensitive and you get a response.
PHILLIP GREENBALM: Have you tried attaching antibodies to these? And
if not, do you think that would be a problem?
N. LYNN JARVIS: We have not and you could certainly attach them. The
problem is that antibodies are very large, and you're trying to attack very small
molecules with the antibody. You may get a very small signal i.e., the change in
weight is very small. Sensitivity might be fairly low in this case. It would not be
a way we would probably choose to go with these particular sensors. There are
probably better sensors for that.
MAHADEVA SINHA: Are these things disposable once you use them? After a
certain while do you throw them out?
N. LYNN JARVIS: Yes. In this system, once a sensor is used up. we propose to
it throw it away and plug in a new one.
MAHADEVA SINHA: You talked about the reversibility of some of the
reactions. What did you mean by that?
N. LYNN JARVIS: There are two ways you can go with a coating on a SAW
dev ice. You can use coat ings where the vapors absorb onto the coating, depending
on solubility characteristics and other factors. They will absorb when the vapor
is present. When the vapor challenge is removed, it desorbs again from this
polymer and is removed. So it's a completely reversible system with certain
vapor coating combinations. You can use a coating where there is no chemical
reaction. However, if you have a chemical reaction, then it is completely
irreversible, which is what we're looking for in this particular application. In
some applications you want reversibility; in some you don't, depending on the
intended use.
EDWARD POZIOMEK: In your last viewgraph and also in your comments
you mentioned the possibility of the wide applications to environmental
measurements, and you said something about putting a SAW down a well.
Perhaps you could comment on the state of this SAW technology for use in
liquids, because the applications presented here were for vapors or for gases.
N. LYNN JARVIS: If we put a sensor in a well, it would have to be within the
well headspace to be monitored, not the liquid. The technology for SAWs in
liquid is very poorly developed, and is just barely beginning. We know of no
really effective way to monitor using a SAW in solution.
83
-------
ARRAYS OF SENSORS AND MICROSENSORS
FOR FIELD SCREENING OF UNKNOWN CHEMICAL WASTES
W.R. Penrose, J.R. Stetter, M.W. Findlay, and W.J. Buttner
Transducer Research, Inc., Naperville, IL 60540
Z. Cao, Illinois Institute of Technology,
Department of Chemistry, Chicago, IL 60616
Abstract
The high cost of laboratory-based analysis has
driven the development of rapid screening
methods for hazardous chemicals in unknown
wastes. Screening methods permit the "triage"
of samples into those that (a) contain no
regulated wastes, (b) definitely contain
regulated chemicals, or (c) are ambiguous.
Only the last category requires detailed
analysis.
The requirements of portability and ease of use
place extraordinary demands on the designers of
analytical instruments. In this paper, we will
discuss several approaches to obtaining
qualitative analytical data from multiple
sensors or highly-selective sensors. These
are: (a) a sensor with a selectivity 1000-
10000 times greater for chlorinated or
brominated compounds than for unsubstituted
ones; and (b) pyrolysis-EC, which uses
catalytic pyrolysis, arrays of electrochemical
sensors, and pattern recognition methods to
identify pure chemicals and mixtures. Two
applications of the latter are described, the
rapid identification of chemical vapors, and
the grading of grain according to "odor".
Introduction
The high cost of laboratory-based analysis has
driven the development of rapid screening
methods for hazardous chemicals in unknown
wastes. A screening method is one that can be
done on-site, by non-chemists, inexpensively
and safely. On the other hand, a screening
method is less likely to provide the definitive
data that a full laboratory analysis, perhaps
requiring GC/MS or ICP, might give. In the
case where no information is available, however,
even limited information can be of value,
especially if it is used to supplement data
gathered from other sources. For example, a
suite of simple screening methods may be used for
the "triage" of unknown samples into positive,
negative, and ambiguous groups. Often, the
nature of the chlorinated compounds may be known
from purchase or production records, so that only
the ambiguous category may require detailed
analysis. Screening methods may also be useful
for confirming conclusions that have already been
drawn from independent data, for example, that a
collection of similar barrels do indeed contain
the same materials.
The will ingness to accept reduced certainty for
the sake of economy and practicality opens the
door to a wide variety of useful techniques that
can be used in the field. In this paper, we will
describe two such methods.
A unique semiconductor sensor has been found that
is very sensitive to chlorinated and brominated
organic compounds (1-3). It shows no detectable
response to hydrocarbons, oxygen- or nitrogen-
containing organic compounds, or fluorocarbons.
A second method that has given us promising
results has been catalytic pyrolysis of chemical
vapors combined with electrochemical detection.
Compounds that are not normally thought of as
electrochemical analytes, such as chloroform or
cyclohexane, can be partially oxidized on a hot
platinum surface (4). The volatile products
always include some that give a response on a
porous-electrode electrochemical sensor. We have
confirmed over several years that the products of
the pyrolysis are reproducible for most organic
and some inorganic compounds when the conditions
are kept reasonably constant (5). We have also
85
-------
confirmed the critical requirement that the
products are independent of analyte
concentration, at least at concentrations of
below 200 ppm. We call this method pyrolysis-
EC.
The present embodiment of pyrolysis-EC is an
instrument we call the CPS-100. This device
uses an array of electrochemical gas sensors
with different, but overlapping, selectivities.
The incoming gases are pyrolyzed over noble
metal catalysts heated at controlled
temperatures. The operation of the instrument
is orchestrated by a fairly powerful computer
which can perform pattern analysis on the
resulting data. In this paper, we report the
results of a study on pattern recognition of
odors in spoiled grain. The unique properties
of neural networks have been shown to have
significant potential for handling low-quality
information. On reflection, this unique
application is not so different from the
problems encountered in classifying and
handling hazardous wastes.
A simplified implementation of pyrolysis-EC has
also been tested that uses a single sensor and
a single catalytic filament. This drastically
simplified system was still capable of
distinguishing many organic chemicals. With
fewer parts and lower power consumption, this
simplified configuration may be suitable for
selective hand-held vapor monitors.
Experimental Methods
Organochlorine sensor. The sensor was made by
mounting a coil of platinum wire on a threaded
base. A separate platinum wire is also mounted
on the base and located axially within the
coil. A mixture of lanthanum oxide, lanthanum
fluoride, and a binder was applied to the coil.
The coil was slowly heated with an electric
current until a reaction occurred, forming the
active material. The sensor is used by heating
it to 550 °C with an electric current;
conductivity is measured between the heating
coil and the separate platinum electrode. When
the sensor contacts the vapor of a chlorinated
organic compound, the conductivity increases.
A simple circuit can be used to provide a
voltage output which is proportional to the
concentration.
Permeation device. The permeation sampler
consisted of a bundle of 0.025" o.d.
dimethyl si licone tubing (Silastic, Oow-Corning)
(Figure 1). The bundle could be placed in an
aqueous sample containing dissolved organic or
organochlorine compounds. A continuous flow of
air was circulated through the lumen of the
tubing, and organic material diffusing inward
through the silicone membrane entered the gas
phase. In a typical experiment, two permeators
were used to provide separate reference and
sample signals (Figure 2).
Pyrolysis-EC. The CPS-100 Toxic Gas Monitor has
been described in several earlier publications
(5-11); its configuration is diagrammed in Figure
3. The four sensors had platinum or gold working
electrodes. For the grain odor experiments, the
sensors were biased at differing oxidizing
potentials, since reducing potentials gave very
low or poor signal s. A single rhodiurnpyrolysis
filament was operated at 25, 450, 750, and 850 °C.
The combination of four sensors and four
temperatures gave an array of sixteen data points
per analysis.
The apparatus for simplified pyrolysis-EC
consisted of a single platinum filament and a
single platinum-electrode gas sensor. A control
circuit maintained the catalyst at any one of
four presel ected temperatures. The f i 1 ament was
enclosed in a Teflon-lined chamber of small
volume through which the analyte gas was pumped
at about 50 cc/min. The gas then passed through
a short tube to the sensor. The experiments were
controlled, and data gathered, by a commercial
datalogger (Onset Computer Corp., N. Falmouth,
MA).
Gas samples. Accurate samples of test compounds
in vapor form were made by injecting measured
volumes of the 1 iquids into 40-1 iter Tedlar gas
bags and filling with air pumped through a
charcoal/Purafil filter. A flowmeter together
with a stopwatch was used to determine the volume
of air being pumped into the bag. Samples of
permanent gases were made from standard mixtures
obtained from commercial sources. Volumes of the
standard mixtures and air were calculated and
pumped into a sample bag, using the flowmeter and
stopwatch to determine the volumes.
Samples from grain odors were generated by
heating a sample of grain to 60 °C and flushing
with a measured volume of air. The effluent air
was passed through an ice trap to collect a "non-
volatile" fraction and a liquid nitrogen trap to
collect the "volatiles". The two fractions were
run separately and in duplicate. Grain samples
were obtained from Drs. L. Seitz, and 0. Saur of
the USDA Grain Marketing Research Laboratory.
86
-------
Results and Discussion
Organochlorine sensor. Typical responses of
the sensor to different vapors in air are shown
in Figure 4. The sensor was exposed to 100 ppm
concentrations of chlorobenzene, benzene, and
n-hexane. Only chlorobenzene caused a
response. Of a series of compounds
investigated, only HC1, and compounds
containing carbon-chlorine and carbon-bromine
bonds, gave a response (Table I). The response
to concentration is essentially linear over at
least four orders of magnitude.
Combined with the permeator device, the highly-
selective organochlorine sensor was shown to
respond rapidly to dissolved material. Figure
5 shows the response to chloroform in water at
concentrations that dip below the part-per-
million level. This sensor can be used to
measure an organochlorine in groundwater, for
example, without any sample preparation. Many
sites, especially military bases, and areas
such as Rockford, Illinois, where there is a
large concentration of machine shops, have
serious problems with chlorinated C2 compounds
in the groundwater. In these cases, the nature
of the compounds is generally known, and
selectivity is not a concern. Nevertheless,
the sampling procedure, sample preparation, and
gas chromatography to determine these compounds
is involved and expensive. The availability of
a simple probe that can just be inserted into
a groundwater sample will greatly reduce the
number of laboratory analyses that need to be
done. The silicone material is chemically
resistant, and can be left in place for years.
Particulates cannot enter the system. Lastly,
and importantly, the permeator is very
inexpensive.
Pyrolysis-EC: Grain Odors Only a few organic
compounds will react directly with amperometric
sensors under field conditions. On a typical,
platinum-electrode sensor, we can detect
alcohols, epoxides, and formaldehyde. We also
detect many permanent gases, such as carbon
monoxide and hydrogen sulfide. Among these
gases that do react, there is no inherent
selectivity. The use of different sensors and
controlled pyrolysis, however, gives us extra
degrees of freedom that can be used to achieve
selectivity.
The grain odor problem is very instructive,
even to an audience that is concerned with
identifying individual hazardous compounds.
Sensor-array-based methods, including the
pattern-analysis methodologies used, treat
mixtures no differently than single compounds;
both give characteristic patterns which can be
identified against a pattern made from the same
mixture. The individual components of a mixture
need not be identified.
Grains are presently classified by odor by a
panel of trained inspectors. The results are
necessarily subjective. More importantly, the
subjective opinion is the standard; there is no
point in telling a customer that a sample of
grain is acceptable because a machine says so.
If it smells bad, it smells bad. On the other
hand, trained inspectors frequently disagree to a
greater or lesser extent on both the category and
degree of an odor (Table II). Attempts to
identify specific compounds associated with the
odors, using GC or GC/MS, have produced masses of
data, but limited results (12, 13).
The data obtained on the CPS-100 was subjected to
two different kinds of analysis. The first was
an established method called k-nearest neighbor
(KNN, ref. 5). The 16 data points acquired by
the CPS-100 were treated as a vector in 16-
dimensional space. Each known sample of grain
produced a vector which could be associated with
a particular odor category. The vectors from the
unknown samples were tested against this library
of known vectors by calculating the scalar
distance between the unknown vector and each
known vector in the library. All vectors were
first normalized to constant length, to remove
the concentration-dependent part of the
information. The shortest distance is the
identification (Figure 6).
The second method is the neural network (for
general references, see 14, 15). This is a
recently-developed method that has received so
much "hype" that we were at first suspicious of
it. However, its performance has been
outstanding in this application, the moreso
because we used a commercially-available packaged
method (NeuroShell, Ward Systems Group,
Frederick, MD), without really understanding the
internal mechanics of the method. This is a very
important feature of a method which may be used
in the field by operatives with differing
technical backgrounds.
Figure 7 shows the CPS-100 data, in histogram
form, for "good" wheat samples. The patterns are
very similar, in contrast with data showing some
extreme samples (one "sour" (S3) and one "insect"
(13) odor) (Figure 8). A experiment using the
older KNN method was run using a dataset derived
from three grades of wheat samples. A library of
vectors was prepared by averaging the signals for
all runs on each sample of wheat. The scalar
distances were calculated between all possible
pairs of the original data set and each of the
averaged vectors. A summary of the
identifications is shown in Table III. We were
very (pleasantly) surprised to find that those
samples that are "misclassified" by the KNN
87
-------
algorithm are also those that the human
inspectors did not agree on! Sample 42, for
example, was voted "good" by two inspectors and
"musty" and "COFO" by the other two. (COFO
means "commercially objectionable foreign
odor".)
Although KNN has shown good performance in past
applications (5, 6, 8-11), it has some serious
practical disadvantages. The greatest is that,
when the sensors become aged or drift for other
reasons, the complete training set must be
remeasured.
A larger data set had been gathered by the time
the work was begun with the neural nets. This
data set had a peculiarity built into it: one
of the sensors in the array went bad halfway
through the measurements and was replaced. The
data taken after that point gave noticeably
different histograms.
The data set was arbitrarily divided into two
groups. One group was used to "train" the
neural network, a process requiring up to 150
hours on a 386-type computer. The actual
classification process took seconds. Two tests
were run on the optimized neural net. First
was a test to confirm that the optimization
process was complete. This was done by using
the training set itself as unknowns. The rate
of correct classification was 100%. Second,
random, linearly-distributed errors were added
to the data, followed by classification. The
net tolerated 5% error without missing a
correct classification. Added error of 10% and
15% caused a small amount of degradation (Table
IV).
Having confirmed the robustness of the neural
net, it was challenged using the reserved
dataset. The net had not seen these numbers
before; nevertheless the rate of correct
classification was 65% (Table IV). This is
low, although substantially better than random.
Because the test conditions had changed during
the measurements, we added another element to
the data vectors to differentiate the
measurements made before and after the sensor
was changed. The numbers were arbitrary, 100
for the old sensor and 200 for the new. Using
these 17-element vectors, the neural net was
retrained. Now, the rate of correct
classification of the reserved dataset jumped
to 83%.
Pyrolysis-EC: Simplified Version This work is
the result of a project to determine whether a
greatly-simplified form of pyrolysis-EC would
be useful for situations requiring limited
selectivity. Figure 9 is a diagram of the
patterns obtained for representative compounds
in a typical experiment. The temperature of
the catalyst is programmed for two minutes at
room temperature, two minutes each at
temperatures of 500, 600, 700, and 800 °C, and
finally two minutes at room temperature again.
The patterns that are obtained are distinct for
many compounds. If your field problem is simply
confirming the identity of the contents of a
number of similar barrels of an unknown chemical,
the pyrolysis-EC approach may in itself be
sufficient, although most practitioners would
feel more comfortable if it supplemented other
field screening methods.
A table of distances for this limited
configuration is shown in Table V. The smaller
the number, the more similar the two compounds
will appear for a given configuration of the
experimental apparatus. This configuration gives
very good identification of ethylene oxide in the
presence of all but alcohols.
The pyrolysis-EC method has several advantages
that are especially conducive to field work. It
is suitable for portable instrument use; the
components are shock-resistant and will operate
in any orientation. They compact and
lightweight, and the power requirements are
small. They are also inexpensive.
Conclusions
1. A sensor has been developed and
characterized that can identify chlorinated or
brominated compounds in the vapor phase or, with
the use of a permeable membrane, in dissolved
form.
2. A combination of catalytic pyrolysis and
electrochemical detection (pyrolysis-EC) can be
used to distinguish unknown compounds with a
modest degree of selectivity that may be adequate
for many field applications.
3. Pyrolysis-EC data, combined with k-
nearest neighbor and neural network
classification methods, has been used effectively
for such varied tasks as the classification of
stored grains by odor, or the classification of
waste chemicals by functional group (11).
4. The neural net can be made to adapt
dynamically to instrument drift. In effect, it
learns from experience.
4. Errors made by the classification
methods correspond in a general way to errors
made by human experts faced with similar
ambiguities in the data.
88
-------
Bibliography
Stetter, J.R., and Cao, Z. "Gas sensor
and permeation apparatus for the
determination of chlorinated
hydrocarbons in water". Anal. Chem. 62,
(1990), 182-185.
Stetter, J.R., and Cao, Z. "A real-time
monitor for chlorinated organics in
water". Proc. 1990 EPA/AWMA Int'l.
Symposium on "Measurement of Toxic and
Related Air Pollutants", Raleigh, NC,
April 3 - May 4, 1990.
Cao, Z., and Stetter, J.R. "A selective
solid-state sensor for halogenated
hydrocarbons". Case Western Reserve
University, Edison Sensor Technology
Center, Proc. Third Int'l. Meeting on
Chemical Sensors, Cleveland, OH,
September 24-26, 1990.
Stetter, J.R., Zaromb, S., and Findlay,
M.W. "Monitoring of electrochemically
inactive compounds by amperometric gas
sensors". Sensors and Actuators 6,
(1984), 269-288.
Stetter, J.R. Penrose, W.R., Zaromb, S.,
Christian, D., Hampton, D.M., Nolan, M.,
Billings, M.W., Steinke, C., and
Otagawa, T. "Evaluating the
effectiveness of chemical parameter
spectrometry in analyzing vapors of
industrial chemicals". Proc. Second
Annual Technical Seminar on Chemical
Spills, Environmental Protection
Service, Environment Canada, Toronto,
Canada, February 5-7, 1985.
Stetter, J.R., Jurs, P.C., and Rose,
S.L. "Detection of Hazardous Gases and
Vapors: Pattern Recognition Analysis of
Data from an Electrochemical Sensor
Arrays," Anal. Chem. 58, (1986) 860-
866.
Stetter, J.R., Zaromb, S. and Penrose,
W.R. "Sensor array for toxic gas
detection". U.S. Patent no. 4,670,405,
1987.
Stetter, J.R., Penrose, W.R., Zaromb,
S., Nolan, M., Christian, D.M., Hampton,
D.M., Billings, M.W., and Steinke, C. "A
portable toxic vapor detector and
analyzer using an electrochemical sensor
array". Proc. DIGITECH/85 Conference,
Instrument Society of America, Boston,
MA, May 14-16, 1985.
9. Stetter, J.R., Zaromb, S., Penrose, W.R.,
Otagawa, T., Sincali, A.J., and Stull, J.O.
"Selective monitoring of hazardous
chemicals in emergency situations". Proc.
1984 JANNAF Safety and Environmental
Subcommittee Meeting, Laurel, Maryland.
10. Stetter, J.R., Zaromb, S., Penrose, W.R.,
Findlay, M.W., Otagawa, T., and Sincali,
A.J. "Portable device for detecting and
identifying hazardous vapors". Hazardous
Materials Spills Conference, April 9-12,
1984, Nashville, TN.
11. Findlay, M.W., Stetter, J.R., and
Pritchett, T. "Sensor array based monitor
for hazardous waste site screening". Proc.
HAZMAT 90 Central Conference, Environmental
Hazards Management Institute, Durham, NC,
March 13-15, 1990.
12. Weinberg, D.S. "Development of an Effective
Method of Detecting and Identifying Foreign
Odors in Grain Samples," Final Report,
Volume I, USDA Contract # 53-6395-5-59,
SoRI-EAS-86-1208, Dec., 15, 1986.
13. Ponder, M. C. and Weinberg, D.S.
"Development of an Effective Method of
Detecting and Identifying Foreign Odors in
Grain Samples," Literature and Equipment
Survey USDA, Contract # 53-6395-5-59, SoRI-
EAS-85-727, Aug., 5, 1985.
14. Nelson, M.M., Illingworth, W.T. A
Practical Guide to Neural Nets. Addison-
Wesley Publishing Company, Reading, Mass.,
1990.
15. Caudill, M., "Neural Network Primer", AI
Expert, Miller Freeman Publications, 1990.
89
-------
Table I. Sensitivities of the organochlorine
sensor to several halogenated compounds.
Vapors
C«HTCI
dH?Br
C,H,I
CJfcF
C.H.C1
C.H.BF
C.H.I
C.C1F.
Concitntration
(ppn)
125
US
125
62.5
61. S
62.5
125
12.5
R4t«pon*«
( X 10'*«hO/PPB)
0.024
0.016
0.003
0.005
0.029
0.020
0.00]
0.022
Table IV. Summary of the accuracy of the neural
network algorithm for identifying vapors drawn
from the wheat samples.
Sorghum Accuricy of
Dati Set Identification
1. Orlglnjl Diti
2. 5X Error tdded
3. 10% Error Added
4. 15X Error Added
100X
100X
98%
92%
Wheit Simples Accuricy of
Qltl Sfit JfJepf 1f1cit1on
1. Totil Oiti
2. Tnln on S5X of
Diti set
3. Add cliinnel for
Test Conditions
100%
65X
83X
Table II. Subjective odor characterization of the
grain samples used in our study.
OKRL INSPECTORS
SAMPLE '
NO.
F41
F42
F67
F78
F128
F30
F39
F69
FB9
N53
N166
N168
OS
CI
QUO
OKO
OHO
OKO
13
12
11
13
12
S3
82
US
OKO
OKO
HI
OKO
OKO
13
C3
12
13
S3
S3
S3
ICF
OKO
HI
OKO
M2
OKO
13
12
12
12
32
S3
SI
KM
OKO
C2
OKO
OKO
OKO
13
Cl
H3
S3
S3
S3
32
FUIS
CONSENSUS
OK
OK
OK
OK
OX
INSECT
INSECT
INSECT
INSECT
S3
S3
S3
nve.
INTENSITY
0.5
0.7
0.2
0.5
0.0
3.0
2.0
1.0
2.7
2.8'
J.91
2V
Table V. Distance matrices for a series of
organic compounds. Table V-A is several
concentrations of ethylene oxide; the
concentrations are shown as the numbers in the
symbols, e.g., ET0100 - 100 ppm. Table V-B shows
the distances among the series of thirteen
compounds. The Abbreviations are:
ISO - Isopropanol
KER - kerosene
STY - styrene
ETG - ethylene glycol
CHX - cyclohexane
ETE - ether
CLO - chloroform
FORM - Formaldehyde
ETO - Ethylene Oxide
ACE - acetone
XYL - xylene
WL- hilothne
ETA - ethanol
TABLE V-A
Distance for Ethylene Oxide
rroioo ET040 rroio
ET0100
FT040
ET020
ETO:
ETOS
ETOl
0.00
0.31
0.28
0.22
0.25
1.02
0.31
0.00
0.07
0.21
0.18
0.80
0.2>
0.07
0.00
0.21
0.16
0.82
ET05
0.22
0.21
0.31
0.00
0.09
0.85
ETO5
0.2S
0.18
0.16
0.09
0.00
0.80
1.02
0.60
0.62
0.85
0.80
0.00
TABLE V-B
Table III. KNN classification of the USDA grain
samples.
Average of Known Vectors
Good Insect Sour
(128. 42. 67, 41) (30, 39, 89) (53, 166, 168)
CHX
ISO
ACE
JTI
XYL
KER
CLO
STY
FORM
HAL
CT3
ETO
ETA
CHX
0
1.57
0.19
1.76
1.03
1.07
0.69
1.44
1.74
2.09
1.52
1.73
1.95
ISO ;
1.57
0
1.43
0.44
0.76
0.63
1.49
0.46
0.11
0.73
0.4
0.63 3
0.81 ]
kCE m XYL
1.19 1.76 1.02
L.42 0.44 0.76
0 1.59 0.85
.59 0 0.82
.85 0.82 0
.91 0.74 0.34
.55 1.53 0.98
.27 0.41 0.45
.56 0.2 0.85
.93 0.59 1.37
.35 0.3 0.55
.55 0.35 0.75
.77 0.36 0.98
KIR
1.07
0.62
0.91
0.74
0.34
0
1
0.49
0.76
1.06
0.54
0.75
0.9S
CLO
0.69
1.49
0.55
1.53
0.88
1
0
1.26
1.55
1.93
1.33
1.44
1.62
STY
1.44
0.46
1.27
0.41
0.45
0.49
1.26
0
0.45
0.9]
0.13
0.37
0.63
FORM
1.74
0.31
1.56
0.3
0.85
0.76
1.55
0.45
0
0.61
0.34
0.41
0.56
HAL
3.09
0.73
1.93
0.59
1.37
1.09
1.93
0.93
0.61
0
0.64
0.79
0.73
ETC
1.52
0.4
1.35
0.3
O.SS
0.54
1.33
0.12
0.34
0.84
0
0.3
0.55
ETO
1.73
0.62
1.55
0.25
0.75
0.75
1.44
0.37
0.41
0.79
0.3
0
0.27
ETA
1.9»
0.61
1.77
O.Ji
0.96
0.99
I.*'
0.61
0.5«
0.73
o.ss
0.27
0
128,128.42,
67,67,41,41,
41,41
89
168
42
30, 30, 30, 39,
39, 89, 89, 89
168, 168
42
30
53, 166, 166,
166, 168, 168
90
-------
Vent
Sensors
Figure 1. Permeation apparatus used to extract Figures. Configuration of the CPS-100 Toxic Gas
°i"ganochlorines from water. Analyzer, fitted with four electrochemical
sensors and two catalyst filaments.
Permeation
apparatus
3-way
valve
Permeation
apparatus
Ci"rier a
In blank water
Exhaust
Fl9ure 2. Experimental apparatus for selective
analysis of aqueous chlorinated hydrocarbons
Using a separate reference permeator.
RESPONSE OF C6H5CI, C6H6 and C6H14
100 PPM, SEH5O» *OO t 10-M
40 60 80
TIME (WIN)
100
120
Figure 4. Response of the organochlorine sensor
to chlorobenzene, benzene, and hexane.
91
-------
Low cone. CHCI3 in Water
Oun.100. FFI-UOcc/mnl
Figure 5. Response of the organochlorine sensor
to decreasing concentrations of chloroform.
,
It 12 13 14 21 22 23 24 31 32 33 34 41 42 43 44
CHANNEL
I IB
JI | B^. 4
Figure 7. Histogram of normalized responses of
the CPS-100 to four samples of "good" grain.
5*6»oigwq1 900d/tOHout/123co(o
Data vectors are normalized
to vectors of unit length.
U, is unknown compound,
P, and P2are
known pattern vectors.
Scalar distance between
vectors U, (unknown)
andn P( and U, and
P3are calculated and
compared (D, and D,)
Figure 6. Schematic representation of the KNN
pattern recognition method in 3-dimensional
space. PI and P2 are library patterns for
known compounds, and Ul is the vector for an
unknown. The distances from Ul to PI and P2
are calculated and compared.
100
60
20
-9D-
~H
[
"H
n
1
1
UH
CHANNEL
Q] OK (Conlrol) B§ 101(53) (
l 1?3 ICOFO1
Figure 8. Normalized responses of the CPS-100 to
"good" (OK), sour (S3), and COFO grain.
-------
Chloroform
Isopropanol
Elliylene Oxide
Acetone
Elhanol
Halothane
Figure 9. Responses of the simplified
pyrolysis-EC apparatus to six different
chemicals. In this experiment, the catalyst
filament was programmed in 2 minute steps at
room temperature, 500, 600, 700, and 800
degrees, and room temperature again.
DISCUSSION
GORMAN BAYKUT: My question is about the chemical analysis with these
sensors. I'm not talking right now about the wheat vapor. But in terms of real
chemical analysis, you must know the compounds you are going to analyze,
otherwise you can't do the analysis because you need training. You can't analyze
the unexpected compounds, am I right?
WILLIAM BUTTNER: The way the CPS 100 Program was originally
envisioned, you had to install the library vectors of potential compounds. If you
were going to look at TCE, there had to be a library vector associated with the
TCE. On the other hand, these arrays are not totally selective in response. The
response to TCE was similar to PCE, that is, tetrachloroethane. You could
therefore identify classes of compounds. But you are right. You have to have
some ideaof the type of vapors present. Atotally unknown situation will still give
some ambiguity in your analyses.
GORMAN BAYKUT: But I think even though your software is powerful, you
need a training period for every compound. How about the mixtures? If you
analyze the mixtures will there be a problem?
WILLIAM BUTTNER: Mixtures are a problem for this type of system. Certain
types of mixtures are well behaved. Gasoline, for example^ is a mixture of many
types of compounds, but it behaves as a single class.
GORMAN BAYKUT: I'm referring to the cracker. You have a thermal cracker
in front of the electrochemical sensor areas. Sometimes you have a mixture of
two or three compounds, or five, or seven and they react in the cracker. You get
different answers, and the correlation is not linear.
WILLIAM BUTTNER: What you're referring to are the reaction products of
the thermal catalysis that result from mixtures being exposed to the sensors. Yes,
you are right. There is frequently a nonlinear response. The reaction products
frequently do react with each other. That's a comment relevant to many field
screening techniques. In some mixtures that factor is a little less significant. If
you do generate very reactive compounds, for example from chlorinated com-
pounds TCE, you do get a nonlinear response. That is a problem. This instrument
was designed to look at single vapors, maybe not necessarily positively identi-
fied, but single vapors.
STEVEN KARR: I wondered if you've given any thought to applying fuzzy
logic algorithms to this problem as opposed to neural networks?
WILLIAM BUTTNER: The neural network was a six-month program that we
tried on the SBIR (we've just finished Phase I). To stay within the time
constraints, we stuck to simple systems. We are investigating other neural
network software packages and other identification algorithms. We will certainly
consider fuzzy networks.
EDWARD POZIOMEK: Have you tried any real-world environmental samples
with the system.
WILLIAM BUTTNER: I had a program through Savannah River to monitor for
TCE emissions out of their stripping tower, as part of their groundwater clean up.
Initially the results were very encouraging. The analyses that I measured were
compared back to groundwater samples as measured at an independent labora-
tory. They were comparable in value. The unfortunate thing is that these
amperometric sensors did not behave truly reversibly to chlorinated compounds,
and that after u period of time their response factor, their sensitivity, would
degrade and ultimately their response would die completely. For that reason it
was determined that these types of sensor systems would not be applicable for
the problems associated with Savannah River Laboratory. This was before this
chlorine selective sensor was developed. It could potentially have application
down there.
93
-------
REAL-TIME DETECTION OF ANILINE IN HEXANE
BY FLOW INJECTION ION MOBILITY SPECTROMETRY
G.E. BURROUGHS
National Institute for Occupational
Safety and Health, 4676 Columbia Parkway,
Cincinnati, OH 45226
G.A. EICEMAN and L. GARCIA-GONZALEZ
Chemistry Department
New Mexico State University
Las Cruces, NM 88003
DISCLAIMER: Mention of company names or
products does not constitute endorsement
by the National Institute for
Occupational Safety and Health.
ABSTRACT
Ion mobility spectrometry (IMS) with a
conventional "Ni ion source exhibits
chemical behavior that should be
advantageous in detection of molecules
with high proton affinity such as
aromatic amines in common organic
solvents. Since IMS instrumentation can
be considered a continuous-sampling point
sensor, IMS may be adapted for industrial
process monitoring or area environmental
monitoring. However, quantitative aspects
of IMS are not well established and
possible interferences may limit the
usefulness of IMS. In order to
characterize IMS behavior as an effluent
sensor, a flow injection IMS device was
evaluated in which an IMS was used as a
detector for a heated injector port. An
IMS drift tube was used with an acetone
doped reaction region and a membrane
inlet. Five microliter replicate samples
were introduced and vaporized in the
inlet at 15 - 90 second intervals and
drawn into the IMS. Detection limits
were ca. 0.5 mg L-1 for 5 ul aliquots (2
ng per sample). Sampling intervals could
be reduced to 15 seconds for all
concentrations below 40 mg L-1 above which
however a working range could be
considered to approximately 100 mg L-1.
Precision was 10 - 25% RSD and was
largely concentration independent. Since
the IMS alone in a vapor stream shows ca.
1-2% RSD, the bulk of variance was from
the inlet and inlet-IMS interface. Four
solvents (benzene, methylene chloride,
ethyl acetate, and acetone) were
evaluated as interferences. All solvents
at some concentrations affected the peak
area for aniline, although the causes
arose through different mechanisms. The
use of IMS as a flow sensor for aniline
in organic solvents should presently be
restricted to samples free of compounds
with strong proton affinities and
solvents which do not exhibit strong
dipoles.
INTRODUCTION
Ion mobility spectrometry (IMS) with a
conventional "Ni ion source exhibits
chemical behavior that should be
advantageous in detection of molecules
with high proton affinity such as
aromatic amines in common organic
solvents. Since IMS instrumentation can
be considered a continuous-sampling point
sensor, IMS may be adapted for industrial
process monitoring or area environmental
monitoring. However, quantitative
aspects of IMS are not well established
and possible interferences may limit the
usefulness of IMS. Among the attributes
of an acceptable "field screening method
for hazardous waste and toxic chemicals"
are sensitivity, specificity, accuracy,
precision, speed, and portability. Also,
to be worthwhile, it should be applicable
to the screening of analytes or classes
of compounds which have a reasonably high
toxicity. The optimum value of a
real-time field technique would be in the
screening of substances with acute
toxicity, thereby assisting in the
elimination of short term exposures. The
purpose of this work is to investigate
95
-------
such quantitative aspects of IMS as
sensitivity, accuracy, and precision;
interference is examined as a comparison
of response to solvents of varying proton
affinity; and speed of analysis is an
additional experimental parameter.
In IMS, vapors are drawn into a reaction
region where analyte is ionized through
proton or electron transfers from a
reservoir of charge, the reactant ions.
The reactant ions originate in beta
emission from a 63Ni radioactive foil and
the reactant ions exhibit near thermal
energies. Consequently, product ions
usually experience little fragmentation
and exist principally as M+, MH , or MjH"1".
lonization in the reaction region is
based on competitive charge exchange, and
unequivocal response occurs when the
target analyte has a proton affinity
larger than that for any component in the
sample matrix. When this is not assured,
response can become confusing even for
simple mixture (1) . Thus, the primary
basis for selectivity of IMS as a
detector is based upon differences in
proton affinities of constituents
following vaporization into a flowing air
stream. Product ions are injected into a
drift region where ions acquire a
constant velocity in a weak electric
field. Differences in ion velocities are
due to differences in cross-sectional
areas, and this serves as a useful,
second level of selectivity in IMS.
However, response in IMS is fundamentally
governed by the original step of product
ion creation; thus, if a product ion is
not formed in the ion source, regardless
of cause, a peak corresponding to that
substance will not be observed in the
mobility spectrum.
Flow injection analysis (FIA) is a type
of continuous analytical technique where
discrete, reproducible aliquots of sample
are introduced into a flux, allowed to
interact with other components of that
flux or with forces exerted on that flux,
and are subsequently monitored by a
detector having some inherent specificity
for the resultant species. Reviews of
flow injection analysis by Betteridge (2)
and by Ranger (3) date the origins of
this technique to the early to mid-1970's
as an adaptation or subcategory of
"continuous flow analysis" as described
by Skeggs (4) . This type of analysis has
the advantages of being simple, accurate,
reliable, reproducible, and can be
accomplished with a small amount of
simple equipment. All of these
attributes are desirable in any
real-time, field screening method. The
disadvantages of FIA methods come from a
dependence on detector selectivity in the
absence of any separator techniques, as
will be seen later.
Chemically, the high proton affinities of
aniline and other aromatic amines suggest
that ion mobility spectrometry may be a
technically acceptable technique for
monitoring of these substances by flow
injection technique. Development of a
field screening method for these
compounds would be worthwhile based on
toxicity, the primary toxic effects of
this class of compounds on man including
methemoglobin formation and cancer of the
urniary tract (5). Environmentally,
"aromatic amines constitute a family of
serious pollutants due in part to a high
degree of toxicity toward aquatic life
(6) . Particular attention has been given
to the effects of aniline, aniline
derivatives, and aromatic amines on fish
(7,8), Daphnia magna (9,10) and microbes
in estuarine water (11)." (Eiceman et al)
Commercially, they are important as
intermediates in the manufacture of
dyestuffs and pigments, but are also used
in the chemical, textile, rubber, dying,
paper industries and other (5).
EXPERIMENTAL
Instrumentation
The introduction of a flow injection
stream to an IMS detector was
accomplished using the instrumentation
and procedures described below. A block
diagram of the flow injection IMS
apparatus is shown in Figure 1 and was
comprised of a heated injector taken from
a gas chromatograph, an Airborne Vapor
Monitor (Grasby Analytical, Ltd.,
Watford, UK) as the IMS detector, a
pressurized source of air and supporting
electronics to control injector
temperature. Air flow through the
injector port was ca. 5 ml/min and the
injector temperature was 100°C. Both the
injector block assembly and the IMS
instrument were placed inside a
laboratory hood, and there was a distance
of less than 1 cm between the injector
exhaust and the IMS inlet. Digital
signal averaging was used to acquire
mobility spectra with an Advanced Signal
96
-------
Processor (ASP) (Grasby Analytical, Ltd.)
into an IBM XT microcomputer. Also,
signal was routed from an output voltage
on the ASP to a Hewlett-Packard 3380A
recording integrator so peak areas for
the aniline product ion could be recorded
versus time and integrated. The window
of observation for drift times for the
aniline peak was ca. 0.1 - 0.2 ms wide
and was centered on the drift time for
aniline, 8.74 ms. Other parameters for
signal collection through the ASP board
were: number of waveforms, 32; points per
spectrum, 512; and scale expansion, 0.25.
The integrator parameters were:
attenuation and threshold, each 9; chart
speed, 1 cm/min; area rejection, 10000;
and peak width 0.5.
Reagents and materials
The following solvents were obtained in
high commercial purity and used without
further treatment: aniline (Aldrich
Chemical Co., Milwaukee, WI, 99.5%+),
hexane (Chromopure, Burdick & Jackson
Co., Muskegon, MI), acetone (Chromopure,
Burdick & Jackson Co.), benzene (B&J
Brand, Chromopure, Burdick & Jackson
Co.), ethyl acetate (Fisher Scientific,
Pittsburgh, PA), and methylene chloride
(Fisher Scientific) .
Procedures
In general, 5 ul aliquots of liquid
sample were delivered with a 10 ul
syringe (Hamilton Co., Reno, NV) to the
heated injection port during continuous
signal processing with the IMS. An
interval of 15 to 90 seconds was
permitted for the air to sweep vapors
from the inlet before another injection
was made. Several parameters were
examined to determine optimum operating
conditions and access the reliability of
IMS as a flow injection detector. The
particular details of each of these
studies were:
Clearance study and response curve - Five
microliters of aniline in hexane at
concentrations from 0 to 100 ppm
(volume/volume liquid) were delivered in
five replicates at different intervals
from 15 to 90 seconds. Peak areas were
determined for the aniline product ion in
the preparation of a quantitative
response curve. The effect of injection
interval also permitted the determination
of memory effects in the IMS under a
range of concentrations.
Chemical interferences - In the study of
chemical interferences in aniline
determinations, 5 ul of 5 ppm aniline in
hexane were co-injected with 0 to 4 ul of
pure interfering solvent. These
interfering solvents were methylene
chloride, benzene, acetone, and ethyl
acetate. Five replicate determinations
were made at 60 second intervals.
RESULTS AND DISCUSSION
General
The reactant ion peak (RIP) with acetone
reagent ion chemistry and the mobility
spectrum for aniline in the hand-held IMS
are shown in Figure 2. The mobility
spectrum for aniline contained a single
symmetrical peak at 8.74 ms drift time,
consistent with previous findings for
aniline with water-based chemistry in the
ion source (12). Residual amounts of
reactant ion at 6.97 ms in aniline
mobility spectrum demonstrated that the
ion source was not saturated and that
comparable behavior may be anticipated at
vapor levels lower than this. This
mobility spectrum was generated using 5
ul of a 5 ppm solution (25 ng absolute
mass) and the peak height relative to the
RIP was reasonable considering the high
proton affinities of aniline.
Previously, aniline was shown with IMS/MS
to yield a protonated molecule, MH"1"
product ion (12) although the ambient
temperature drift tube and alternate ion
chemistry used here may favor the
existence of a MH^S ion where S is an
acetone solvent molecule, but this has
not been unequivocally established.
Clearance Behavior. Standard Deviation.
and Response Curve
The hand-held IMS used in this work would
be suited for field use due to its size
(40cm x 15cm x 8cm), weight(2.6 kg), and
ability to operate continuously in
hostile environments unattended. The IMS
itself is battery powered and could be
interfaced with a battery powered lap top
computer for data acquisition, providing
a portable system. However, this IMS
could be expected to exhibit memory
effects from the ambient temperature
drift cell and membrane-equipped inlet.
At high concentrations of aniline, slow
clearance from repetitive determinations
might occur. In Table 1, peak areas and
percent relative standard deviations
(%RSD) from repetitive determinations are
given for solutions between 0 and 100 ppm
at injection intervals from 15 to 90
seconds. The %RSD ranged from 13 to 125,
but showed a median of 21% Previous
97
-------
experience with this IMS as a detector in
FIA methods had yielded reproducibility
of peak heights of 8 to 10 %RSD and this
large variance was suspected to be due to
the placement of the FI-IMS in the fume
hood. Turbulence in a fume hood has been
associated with position and movement of
the user as well as amount and location
of equipment in the hood (13). This
turbulence likely affected yields in the
interface between the inlet and IMS and
this large RSD was suggestive that
mechanical improvements in interface
between the IMS and injection port are
needed. A straightforward leak-tight
connection was not employed in these
studies due to the flow characteristics
for this IMS and the eminent rupture of
the membrane inlet if pressure
differences developed between the inlet
and ion source regions.
The anticipated memory effect from slow
clearance of the aniline from the IMS was
evident in the peak areas given in Table
1. In general, peak areas with 90 second
injection intervals were the lowest for a
given concentration level. Injection
intervals less than 90 seconds caused an
accumulation of aniline in the IMS and
peak areas increased for example as much
as 100% at 30 second intervals with the
100 ppm concentration. This was
manifested in the signal for continuous
monitoring as a rising baseline and in
the mobility spectrum as a persistent
product ion. Memory effects here were
dependent upon concentrations, as
expected, and at concentration below 20
ppm, injection intervals of 15 seconds
could be employed with reasonable
differences in absolute areas.
A plot of peak area versus concentration
of aniline in hexane for 5 ul injections
at 90 second intervals is shown in Figure
3 and resembled previous response or
calibration curves in IMS (14). Such
curves are comprised of narrow linear
ranges (in this instance between 5 to 20
ppm) , a shallow but mostly linear
response at concentrations above the main
linear region and a nearly linear plot
with shallow slope below the linear
region. This behavior is due to the
nature of the kinetics of reactant ion
formation from the beta emitting ion
source and, thus, to the limited
reservoir of charge available to analyte
vapors.
Chemical Interferences
The existence of solvents with a range of
proton affinities in industrial waste
streams constitutes a potential
compromise on the integrity of IMS
response in flow injection determinations
through two mechanisms. Conceivably,
large levels of such solvents might
compete for charge resulting in reduced
peak areas for aniline at given vapor
levels. Alternately, solvents may cause,
at ambient cell temperatures, ion-solvent
clusters which lead to shifts in drift
times for product ions. This will cause
a decline in certainty regarding peak
identity or may cause the peak to fall
outside a window of observation in the
signal processing software.
Four solvents with low and medium proton
affinities were selected for interference
studies and mobility spectra for
individual solvents are shown in Figure
4. Methylene chloride gave little
response in positive polarity IMS as
expected due to a low proton affinity.
For the same reason, benzene showed a
weak response with an acetone reactant
ion chemistry and the product ion had a
drift time shorter than that for the RIP.
Acetone formed cluster ion, with drift
times longer than that for the RIP,
through ion-molecule interactions in the
IMS drift region as described by Preston
and Rajadhax (15). Only ethyl acetate
(EtOAc) showed significant competition
with the reactant ion, due to large
proton affinities of EtOAc relative to
acetone, with the obvious result of a
product ion. Of these solvents, only
benzene has been mass identified as M
(16) though acetates are known to form
MH+ and MjH"1" product ions (17).
The influences of these solvents on IMS
response to a 5 ul injection of 5 ppm
aniline in hexane are shown in Figure 5
as a plot of peak height for aniline in
various ratios of four solvents in a
binary mixture with hexane. All solvents
affected the peak area for aniline
although the causes arose through
different mechanisms. In Figure 6,
mobility spectra are shown from egual
mixtures of hexane and solvent for 5 ppm
aniline and these can be compared
directly to spectra for individual
solvents (Figure 4) and for aniline
(Figure 2) . For EtOAc, the product ion
dominated the ion chemistry when aniline
98
-------
was present even though proton affinities
favored aniline. Ethyl acetate at high
concentrations relative to aniline
appropriated virtually all the charge
except that remaining with the RIP. The
ion-molecule chemistry for acetone as an
interference also followed this pattern
and aniline was not detected with high
levels of acetone. Thus, the rise in
peak areas in Figure 5 represented a
false positive by acetone for aniline
since acetone product ion intensity
intruded upon the drift time window used
to monitor aniline. In such a situation,
only inspection of the mobility spectrum
could avert an error in monitoring on
analyses. A product ion for aniline was
evident with methylene chloride due to
the low proton affinities of methylene
chloride. However, the increase in
response for aniline in positive
polarities from addition of methylene
chloride to hexane (Figure 5) was
unprecedented in IMS and conclusions
cannot be made pending IMS/MS studies.
Benzene, with proton affinities between
methylene chloride and acetone or EtOAc,
exhibited a type of intermediate
behavior. A product ion for aniline was
observed in the presence of benzene, but
the benzene was at a level sufficient to
effectively compete for protons from the
RIP and a benzene product ion was also
observer (Figure 6) . These spectra and
trends suggest that an IMS will be
sensitive to common solvents at low
levels even with an alternate reactant
ion chemistry, a membrane inlet, and low
(<1%) levels of solvents other than
hexane. However, if the solvent
composition is known and reasonably
constant, calibrations presumably could
be prepared in that matrix. These
findings for simple compositions argue
for standard addition techniques with
flow injection IMS determinations.
CONCLUSIONS
Ion mobility spectrometry has never been
widely regarded as a quantitative
instrument, but as a detector for flow
injection determination, IMS exhibited
suitable response curves, standard
deviations, and response times. This was
accomplished under the demanding
situation of a fast transient vapor level
in FIA methods. The linear range is a
weak aspect to quantitative IMS and
alternative configurations to
conventional 63Ni sources should be
sought. Reactant ion chemistry based on
acetone was not wholly successful in
discriminating chemically against common
organic solvents. Consequently, until
improved source chemistry is found,
standard addition should be considered
the method of choice for quantitative FIA
with IMS for aromatic amines.
REFERENCES
1. Eiceman, G.A., Blyth, D.A., Shoff,
D.B., Snyder, A.P. Anal. Chem., 1990, in
press.
2. Betteridge, D. Anal. Chem. 1978,
50, 832A-845A.
3. Ranger, C.B. Anal. Chem. 1981 53,
20A-32A.
4. Skeggs, L.T. Am.J. Clin. Pathol.
1957 13, 451.
5. Beard, R.R., Noe, J.T. "Aromatic
Nitro and Amino Compounds," in Pattv's
Industrial Hvoiene and Toxicology. Vol.
2A, G.D. Clayton and F.E. Clayton,
editors, Wiley-Interscience, New York,
1981.
6. National Research Council, "Aromatic
Amines: An Assessment of the Biological
and Environmental Effects," No.
PB83-133058, Washington, DC. 1981.
7. Bradbury, S.P., Henry, T.R., Nieme,
G.J., Carlson, R.W., Snarski, V.M.
Environ. Toxicol. Chem. 1989, 8, 247-261.
8. Newsome, L.D., Johnson, D.E.,
Cannon, D.J., Lipnick, R.L. "Comparison
of Fish Toxicity Screening Data and QSAR
Predictions for 48 Aniline Derivatives,"
QSAR Environ. Toxicol., Proc. Int.
Workshop, 2nd, Kaiser, K.L., Editor,
Reidel, Dordrecht, Netherlands, pp.
231-250, 1987.
9. Kuehn, R., Pattard, M., Pernak,
K.D., Winter, A. Water Res., 1989, 23,
495-499.
10. Gersich, R.M., Milazzo, D.P. Bull,
Environ. Contam. Toxicol., 1988, 40, 1-7.
11. Hwang, H.M., Hodson, R.E., Lee, R.F.
Water Res. 1987, 21, 309-316.
12. Karpas, Z. Anal. Chem. 1989, 61,
684-689.
13. National Research Council, Committee
on Hazardous Substances in the
Laboratory, "Prudent Practices for
Handling Hazardous Chemicals in
Laboratories," Washington, DC, 1981.
14. Leasure, C.S., Eiceman, G.A. Anal.
Chem., 1985, 57, 1890-1894.
15. Preston, J.M., Rajahyax, L. Anal.
Chem., 1988, 60, 31-34.
16. Kim, S.H., Betty, K.R., Karasek,
F.W. Anal. Chem, 1978, 50, 1754-1758.
17. Eiceman, G.A., Shoff, D.B., Harden,
C.S., Snyder, A. P. Internal. J. Mass
Spectrom. Ion Processes, 1988, 85,
265-275.
99
-------
fro* Plot* of AniliM Product Ion lotwwitr
AniliM
Concentration
(PP»>
PEAK AREA
-------
CH
6 6
EtOAc
Drift Time (msl
4. Ion mobility spectra for solvents
expected to be encountered in
analysis of non-aqueous streams for
aniline. Mobility spectra were
obtained in positive polarity with
acetone reagent ion chemistry.
Spectra were obtained with solvent
vapors permitted to deplete reactant
ion intensity ca. 50% from
background levels.
w/CH?l
I
7.77 ™ w/Acetone
| """ w/EtOAc
I 15.01 5
Drift Time (ms)
6. Mobility spectra for mixtures of 5
ppm aniline in 50 : 50 mixtures of
individual solvents with hexane.
Aniline in hexane exhibited a single
product ion with drift time of 8.74
ms as shown in Figure 2.
0.5 5 25
Percent in Hexane
5. Effect on peak height for aniline at
5 ppm in binary solvent mixtures of
hexane with other common solvents
with vol/vol percentage from 0» -
50% Curves were normalized to the
peak Sit of aniline in hexane
solution.
101
-------
DISCUSSION
STEVEN HARDEN: The question I have is with respect to orthonitrophenol
and the sensitivity of the IMS system to that particular kind of material. Did you
ever do a calibration run to determine what that sensitivity might be under various
conditions?
PETER SNYDER: The answer to that question is no, we have not on pure
orthonitrophenol. However, the signals—the amount of signal that we see from
the other point of view, looking at it from the organism's point of view, and
knowing how much organism we have. It seems like there is still plenty of
analyte, given the relatively short time of detection, and knowing that the signal
is still a bit spread out. The signal is not in one, or say two, or maybe three at the
most, peaks. We see it at about seven, eight, nine 10 peaks, until it finally clears
down.
So I'm not trying to skirt the question. It's just that no, we haven't done it to see
how sensitive the CAM itself is, or the ion mobile spectrometer 20MP. However,
I suspect that it has to be very sensitive, since 200, even 50 cells is a good
response, and the response is spread out, so if we can find ways of compacting
it, it'd be that much better.
MAHADEVA SINHA: What are the vapor pressures for the orthonitrophenol
when it gets combined with the glucose. Do you get any response?
PETER SNYDER: Yes, we've done many, many blanks. We always do a blank
before and after.
First of all, the vapor pressure of orthonitrophenol is 5.54 torr at ambient
temperature. That doesn't should like much, but relatively speaking, that's a lot
for the CAM. And the controls — we have done ONP by itself, with buffer,
without buffer, and then just organisms themselves. Organisms do produce some
peaks, but that's just right after the reactant ion peak. But it just happens to tail
off, and there is no signal in the area that the ONP shows. So we have been pretty
lucky in that respect.
The ONP has very negligible vapor pressure by itself. Even if you get a bottle of
the dry powder, and just stick the CAM in the bottle, you see no response at all.
That should be the most amount, the dry powder, and if anything's going on it
would show. But even in the solution, there's no problem.
Orthonitrophenylacetate is a different story. There is hydrolysis going on and
over a couple of hours, you can see orthonitrophenol being produced.
102
-------
DETECTION OF MICROORGANISMS BY ION MOBILITY SPECTROMETRY
A.P. Snyder, M. Miller
and D.B. Shoff
U.S. Army Chemical Research,
Development and Engineering
Center, Aberdeen Proving
Ground, MD 21010-5423
G.A. Eiceman
New Mexico
State University
Las Cruces, NM
88003
D.A. Blyth & J.A. Parsons
GEO-CENTERS, INC.
c/o U.S. Army Chemical
Research, Development
and Engineering Center
Aberdeen proving Ground,
MD 21010-5423
ABSTRACT
A relatively new concept is explored
where the potential for ion mobility
spectrometry is investigated for the
detection and determination of living
microorganisms. The hand-held,
NATO-fielded Chemical Agent Monitor
(CAM) embodies the analytical device.
Advantage is taken of the inherent
enzymes found in microorganisms and an
exogenous, tailored substrate was
provided in order to initiate the
desired biochemical reaction. The
substrate was ortho-nitrophenyl-beta-
D-galactopyranoside, and the product,
ortho-nitrophenol, can be detected in
the negative ion mode of the CAM and
signals the presence of bacteria.
Detection limits of approximately 10E4
E. coli bacterial cells in 5 min. and
3300 E. coli cells in 15 minutes were
realized. The results suggest a new
application of the CAM in the
screening of bacterial contamination
in community water and wastewater
testing situations.
KEYWORDS: ion mobility spectrometry;
microorganisms; E. coli; enzymes;
ortho-nitrophenol; Chemical Agent
Monitor; ortho-nitrophenyl-galacto-
pyranoside; fecal coliforms.
INTRODUCTION
Detection and identification of
microorganisms is a challenge in view
of the required sensitivity, selec-
tivity, and time of response of the
detection technique. Table 1 lists
these requirements for a number of
methods. It appears that analytical
instrumentation techniques broadly
fall in the detection limit range of
10E6 bacterial cells with an instru-
mental response time of approximately
1.5 hr. The colorimetric and fluoro-
metric enzyme assay procedures fare
better and can be characterized by
10E3-10E5 bacterial cell limits of
detection in a 0.25-4 hr response
time domain.
ion mobility spectrometry (IMS) is a
straightforward, analytical vapor
detection technique. Neutral analyte
vapors enter the device and are
ionized, usually by a Ni ring. The
ions are electrically gated and
"drift" through an antiparallel flow
of buffer gas (air or nitrogen). The
ions are focussed by an electrical
field about the heated, cylindrical
drift region and are registered by
a Faraday cup detector. The entire
process, from vapor sampling to the
detection event, takes place at
ambient or near-ambient pressure, and
thereby atmospheric pressure ioniza-
tion chemistry characterizes the ion
formation process. Ions are parti-
tioned primarily according to their
mass and shape and 'are characterized
by their corrected drift times (typi-
cally in msec) or ion mobilities. In
the negative mode, IMS is very simi-
lar with respect to an electron
103
-------
capture detector in terms of the
detection event and sensitivity.
The detection of bacteria by IMS
originated from the concept of aug-
menting the hand-held Chemical Agent
Monitor (CAM) with capabilities for
biological detection, more specifi-
cally, that of viable microorganisms.
The Hypothesis was that the ion-mole-
cule chemistry that characterizes the
atmospheric pressure-based IMS tech-
nique, embodied by the CAM device,
could be used to detect a targeted
volatile product of the biochemical
reaction between an rn vivo bacterial
enzyme and a tailored organic
substrate. This proved to be an
interesting challenge because
parallels could be drawn with that of
standard, well-established microbio-
logical and clinical bacterial evalu-
ation procedures in the process of
devising the CAM detection of micro-
organisms .
EXPERIMENTAL
Ortho-nitrophenol (ONP) and ortho-
nitrophenyl-beta-D-galactopyranoside
(ONPG) were obtained from Aldrich
Chemical Co., Inc., Milwaukee, WI and
Sigma Chemical Co., St. Louis, MO,
respectively. The beta-galactosidase
enzyme and ONP-acetate were obtained
from Sigma Chemical Co., St. Louis,
MO. Pure E. coli suspensions (ATCC
11303) or Bacillus globigii (ATCC
9372) were prepared by growth in a
nutrient broth solution for 48 hr
which was supplemented with 0.5%
lactose sugar for induction of the
beta-galactosidase enzyme. The
bacterial growth was centrifuged
and the pellet was washed three times
with a sterile 0.1M phosphate-
buffered saline solution (0.7% NaCl)
at pH 7.4 (PBS). Approximately Ig of
human fecal matter was suspended in
10 ml of distilled water. Strips of
Whatman 15 filter paper (Whatman
International, Ltd., Maidsgone,
England) were baked at 150 C over-
night in glass vials and used for
bacterial determination experiments.
Two microliters of the E. coli or
fecal matter suspensions were used
for filter paper experiments and 0.1
ml was used for bulk volume liquid
experiments. Two microliters of a
2.0 mg/ml ONPG solution in PBS were
used for the filter paper experiments
and 1.9 ml of the same ONPG solution
was used for bulk volume microbial
determinations. The fecal bacterial
experiments were conducted at room
temperature (25 C) while the pure E.
CQp.i experiments were carried out at
38 C.
After selected 'incubation periods at
the given temperatures, the headspace
of the bottle was sampled with the
hand-held CAM by removing the cap and
immediately placing the vial opening
at the inlet of the CAM unit.
The hand-held CAM (Graseby Ionics,
Ltd., Watford, England) device was
used as the analytical detection
technique which was designed speci-
fically for air sensing in military
field applications (15). Signals
from the CAM were processed by using
a Graseby Ionics, Ltd., advanced
signal processing (ASP) board and
software with an IBM-PC/AT. Details
of the CAM unit are as follows:
drift gas, nitrogen or air; ion
source, 10-mCi Ni; drift region
length, 7 cm; drift field, 230 V/cm;
drift gas flow, 300 mL/min; reaction
region length, 3 cm; drift tube tem-
perature, ambient; shutter width,
0.1 msec (16). A schematic and de-
tails of the operation of the hand-
held CAM ion mobility spectrometry
unit can be found elsewhere (17).
For the fecal bacteria experiments
the data were captured and displayed
by the ASP software while for the
pure E. coli determinations, the ion
mobilTty signals were captured and
displayed by a Nicolet 4094A oscillo-
scope and Hewlett/Packard 7470A
plotter.
RESULTS AND DISCUSSION
A number of constraints were
realized in that for a system such as
the CAM to be a realistic analytical
method for biological detection, only
minimal logistic burdens to the
collection, processing and introduc-
tion of the sample to the hand-held
IMS unit would be tolerated. There-
fore the question was posed: How can
the CAM be used as it is intended
(i.e. - a vapor detector) in the
detection and possible identification
of extremely complex entities such as
microorganisms? The microbiological
104
-------
literature provided constructive in-
sights into this problem in the form
of constitutive enzymes (enzymes that
are always present in a bacterium)
that are secreted at significantly
different quantitative levels depen-
ding on the organisms. This is a
property of living active cells and
not of dead or dormant microorgan-
isms. Conventional clinical proce-
dures used in the detection and iden-
tification of organisms rely on
tailored substrates (i.e. - compounds
that mimic the enzyme's natural
substrate) to interact selectively
with the secreted enzymes of bac-
teria. The enzyme-catalyzed products
of natural substrates are usually
spectroscopically-silent and as such,
tailored compounds substitute a por-
tion of the natural substrate with a
compound such that when it is re-
leased, it becomes spectroscopically
active (e.g. - colorimetric or
fluorimetric properties). This con-
cept was then related to the proposed
CAM detection of bacteria, except
that the product would have to dis-
play a relatively high vapor pressure
and the CAM must respond to the
product.
Enzyme Substrate and product
Previous investigations in this
laboratory (13) have shown that
bacteria such as Bacillus subtilis
(BG) , the yeast Saccharomyces
cerevisiae, Serratia marcescens (SM)
and E. coli produced at various rates
the 3-hydroxyindole (indoxyl) as a
highly fluorescent and blue colori-
metric product from the reaction of
indoxylacetate, indoxylglucoside and
indoxylphosphate with their respec-
tive esterase, glucosidase and phos-
phatase enzymes. 4-methyl-umbelli-
feryl-beta-D-galactoside reacted with
the beta-D-galactosidase enzyme in E.
coli and SM to produce the
fluorescent 4-methylumbelliferone
product (13). The indoxylacetate
probe (13) was the most sensitive
where as little as 500 BG cells/ml
could be detected in under 15
minutes. Modification of these
substrates, with extensive biochemi-
cal IMS experimentation underscored
the role of the organic substrate as
the heart of the project. A number
of important requirements concerning
the substrate must be satisfied in
order to ensure a successful
approach. Requirements of the sub-
strate include that it (a) is water
soluble, (b) is recognized by a tar-
geted enzyme, (c) displays rapid
enzyme-substrate kinetics (i.e. -
favorable association constant), (d)
has minimal/negligible spontaneous
hydrolysis and (e) that it gives a
minimal/negligible response to the
CAM. Requirements for the product
include (a) a low association
constant with biological material,
(b) a relatively low water solubil-
ity, (c) favoring the gaseous phase,
and (d) being "CAM-active". Alter-
nate compounds were sought. instead
of ester compounds, established
microbiological colorimetric indica-
tors were analyzed. ONPG displays an
acetal functional group that joins
ONP and the beta-D-galactopyranoside
sugar monomer (Figure 1) and is a
standard microbiological indicator
for the detection of all (total)
fecal coliform bacteria (18, 19).
Fecal coliform bacteria belong to the
Enterobacteriaceae and are comprised
of E. coli (4xlOE8 cells/g feces) ,
Klebsiella sp. (5xlOE4 cells/g) ,
EnterobacTer (10E5 cells/g) and
Citrobacter (10E6 cells/g) (20) .
These bacteria, with E. coli as the
predominate species, are tound in
fecal matter, and the latter three
genera are also associated with
plants and soils. E. coli, however,
can only be found in the environment
through fecal contamination (21).
Figures 1 and 2 pictorially display
the enzyme-substrate biochemical and
detection events of the ONP product
by the CAM. Figure 3 shows a CAM
response of a phosphate-buffered
saline solution of ONP in the
negative ion mode. The main peak at
6.2 msec,consists largely of
O (H 0) clusters and the shoulder
t£ ti?e ¥eft of the peak is
characteristic of the chloride ion.
The peaks at 9.1 msec represent the
ONP monomer at different concentra-
tions and the low intensity peak at
11.7 msec represents the dimer ion
(22). Thus, a favorable analytical
situation has been established in
that a compound has been found that
not only has established roots in the
microbiological detection and
identification arsenal as a
colorimetric indicator but also
105
-------
responds to ion mobility spectrometry
through well established ion-mole-
cule, gas-phase reaction chemistry.
CAM-Bacterial Trials
A buffered solution of the ONPG
substrate produced no response from
the CAM unit. When an aliquot of
pure beta-D-galactosidase enzyme was
added to the ONPG solution, a yellow
color appeared within seconds and the
CAM ASP registered this event in the
negative ion mode in a fashion
similar to that in Figure 3. Bac-
terial tests followed. one was from
a pure culture of E. coli and the
other bacterial source was of fecal
origin. Microliter volumes of
bacterial sample and buffered ONPG
were spotted on a strip of sterile
filter paper and the latter was
inserted into a vial. The vial was
secured with a screw cap in order to
contain any ONP product that was
released into the vial headspace.
For the fecal suspension, 2
microliters of a Ig feces/10 ml
distilled water was used. Since an
approximate concentration of E. coli
in human fecal matter (20) is~~about
4xlOE8 cells/g, the actual applied
amount of bacteria was approximately
8xlOE4 cells. Figure 4 portrays the
results of this study. Position A
in Figure 4 represents a background
CAM response of the bacterial
inoculation without ONPG substrate.
The bacterium does provide distinct
ion mobility peaks which are most
likely due to inherent bacterial
volatile compounds. A blank
consisting of the buffered ONPG
solution produced only the negative
background ion mobility signal
(Position B in Figure 4). Position C
in Figure 4 represents the CAM
response of the vial headspace after
the buffered ONPG substrate was added
to the bacterial spot on the filter
paper and was acquired 40 min. after
substrate addition. Note that in
addition to the background ion
mobility signal and the three peaks
representing the bacterial volatile
products, a new peak appeared at 9.1
msec which matched that of ONP
(Figure 3). Figure 5 shows a
replicate experiment where frame A
represents the ONP response 15 min.
after an ONPG solution was added to a
fecal inoculation on a filter paper
strip. Frame B shows that at 45 min,
the ONP signal grew considerably.
An inoculated dose of 10E4 E. coli
cells from a pure suspension on a
filter paper strip produced a peak in
five minutes (Figure 6). At the same
E. coli inoculation, Figure 6 also
shows the CAM "response to the
production of ONP after 10, 15 and 20
minutes. The background shows
essentially no peak in the 9.1 msec
time window and the reaction
consisted of 2 ul of phosphate buffer
added to 2 ul of ONPG. This
indicates that the spontaneous
hydrolysis of ONPG at 38 C is minimal
and intense ONP signals can be
observed over a relatively short
period of time resulting from the
bacterial enzymatic reaction at the
relatively low amount of 10E4 E. coli
cells. Figure 7 shows similar data
except that the amount of inoculated
E. coli was 3.3xlOE3 cells. Indeed,
within 20 min., a clear ONP signal
was observed at 9.1 msec. This
experiment was repeated (Figure 8)
and in 15 min. a discernible ONP peak
was observed. A bulk 2.0 ml volume
suspension consisting of ONPG and
fecal matter (a total of 4xlOE6 fecal
bacterial cells) took 2 hr for a
response from CAM while the yellow
ONP color in the suspension was
observed prior to the CAM detection
event. The longer dwell time is to
be expected because the relatively
large volume of water had a small
surface area for the ONP to partition
into the gas phase as opposed to
microliter amounts which rapidly
diffuse across a strip of filter
paper .
Other Enzyme/Substrate Complexes
ONP-acetate can be cleaved by an
esterase and this compound was used
in the determination of the lipase
enzyme in Bacillus globigii. Table
2 presents the amount of bacteria
used to generate an ONP ion mobility
peak after a 15 min. incubation time.
One thousand cells of B. globigii
produced an ONP signal comparable to
that of Figure 6E. However, with the
ONPG substrate, no signal was
observed with 10E5 cells. The
absence of an ONP signal is due to
106
-------
c —
approximately 3.:
bacterial cells.
the fact that B. globigii, as well as
most other bacilli, do not contain
the beta-galactosidase enzyme and as
such ONP is not produced. The
opposite situation occurs with E.
coli. As Table 2 indicates, E. coli
provides a positive biochemical
reaction with ONPG, but not with
ONP-acetate.
Comparison to Other Techniques
For the E. coli fecal coliform ONPG
test, the CAM unit was observed to
provide an ONP signal in 15 min. with
,3xlOE3 E. coli
It is of interest
to compare these response time/
inoculation figures of merit with
that of established and potential
microbiological, clinical and
analytical instrumentation
techniques. Table 1 provides a list
of a number of these methods
including total number of bacteria
and the time needed for a reliable
analysis of bacterial presence. The
CAM concept of bacterial detection
via inherent enzyme biochemical
reactions which yield tailored
volatile products appears to be a
competitive technique in the
determination of microbial presence.
CONCLUSIONS
A major step in the chemical detec-
tion and identification of viable
(i.e. - living) microorganisms was
presented in terms of analytical
techniques. The ion-molecule
chemistry associated with IMS was
shown to be a promising avenue for
the monitoring of bacterial presence
by taking advantage of available sub-
strate-induced accessible enzymes.
The hand-held ion mobility spectrom-
eter CAM unit displayed detection
sensitivity levels for E. coli fecal
coliforms and response times similar
or better than that of most commer-
cially-available methodologies and
analytical instrumentation techni-
ques. This suggests a potential
application of IMS for screening of
bacterial presence in community/local
water and wastewater testing
protocols.
ACKNOWLEDGEMENT
The authors wish to thank Ms. Linda
Jarvis for the preparation and
editing of the manuscript.
REFERENCES
1.
4.
Newman, R.S. , and O'Brien, R.T.,
"Gas Chromatographic Presumptive
Test for Coliform Bacteria in
Water," Appl . Microbiol. Vol.
1975, ~
30,
Bachrach, u. and Bachrach, z.,
"Radiometric Method for the
Detection of Coliform Organisms
in Water," Appl. Microbiol. vol.
28, 1974, pp. 169-171.
Wilkins, J.R., Young, R.N. and
Boykin, E.H. , "Multichannel
Electrochemical Microbial
Detection Unit," Appl. Environ.
Microbiol. vol. 35~T"T978, pp.
214-215.
Cady, P., Dufour, S.W., Shaw, J. ,
and Kraeger, S.J., "Electrical
Impedance Measurements : Rapid
Method for Detecting and
Monitoring Microorganisms," J.
Clin. Microbiol. vol. 7, 197?,
pp. 265-272.
Fraatz, R.J., Prakash, G. and
Allen, F.S., "A Polarization
Sensitive Light Scattering System
for the Characterization of
Bacteria," Am. Biotechnology Lab.
Vol. 6, 1988, pp. 24-28.
Libby, J.M., and Wada, H.G.,
"Detection of Neisseria
meningitidis and YersTnia pestis
with a Novel Silicon-Based
Sensor," J. Clin. Microbiol . vol.
27, 1989, pp. 1456-1459.
Shelly, D.C., Quarles, J.M., and
Warner, I.M., "Preliminary
Evaluation of Mixed Dyes for
Fingerprinting Non-Fluorescent
Bacteria," Anal . Lett. , vol.
14(813), 1981, pp. 1111-1124.
Steinkamp, J.A. , Fulwyler , M.J.,
Coulter, J.R., Hiebert, R.D.,
Homey, J.L. and Mullaney, P.F.,
"A New Multiparameter Separator
107
-------
for Microscopic Particles and
Biological Cells," Rev. Sci.
Instrum. Vol. 44, 1973, pp.
1301-1310.
9. Graham, K., Keller, K. , Ezzel,
J. and Doyle, R., "Enzyme-Linked
Lectinosorbent Assay (ELLA) for
Detecting Bacillus anthracis,"
Eur. J. Clln. MicroDio.1. vol. 3,
1984, pp. 210-212.
10. Feng, P.C.S. and Hartman, P.A.,
"Fluorogenic Assays for
Immediate Confirmation of
Escherichia coli ," Appl.
Environ. Microbiol. Vol. 43,
1982, pp. 1320-1329.
11. Warren, L.S., Benoit, R.E. and
jessee, J.A., "Rapid Enumera-
tion of Fecal Coliforms in Water
by a Colorimetric beta-Galacto-
sidase Assay," Appl. Envi ron.
Microbiol. Vol. 35, 1978, pp.
136-141.
12. Godsey, J.H., Matteo, M.R.,
Shen, D., Tolman, G. and Gohlke,
J.R., "Rapid Identification of
Enterobacteriaceae with
Microbial Enzyme Activity
Profiles," £. Clin. Microbiol.
Vol. 13, 1981, pp. 483-490.
13. Snyder, A.P., Wang, T.T. and
Greenberg, D.B., "pattern
Recognition Analysis of In
Vivo Enzyme-Substrate
Fluorescence Velocities in
Microorganism Detection and
Identification," Appl. Environ.
Microbiol. Vol. 51, 1986, pp.
969-977.
14. Berg, J.D. and Fiksdal, L.,
"Rapid Detection of Total and
Fecal Coliforms in Water by
Enzymatic Hydrolysis of
4-Methylumbelliferone-beta-D-
Galactoside," Appl. Environ.
Microbiol. Vol. 54,
1988, pp.
2118-2122.
16. Eiceman, G.A., Shoff, D.B.,
Harden, C.S., Snyder, A.P.,
Martinez, P.M., Fleischer, M.E.
and Watkins, M.L., "Ion Mobil-
ity Spectrometry of Halothane,
Enflurane, and Isoflurane
Anesthetics in Air and
Respired Gases," Anal. Chem.
Vol. 61, 1989, pp. 1093-1099.
17. Eiceman, G.A., Snyder, A.P.
and Blyth, D. A., "Monitoring
of Airborne Organic Vapors
using Ion Mobility Spectrom-
etry," Intl. J. Environ.
Anal. Chem. Vol. 38, 1990,
pp 415-425.
18. Paik, G., "Reagents, Stains,
and Miscellaneous Test
Procedures, in Manual of
Clinical Microbiology, Third
Edition, E.H., Lennette, A.
Balows, W.J. Hausler, Jr. and
J.P. Truant, eds., American
Society for Microbiology,
Washington, DC, 1980, p. 1006.
19. Colilert Most Probable Number
Method Product Brochure, Access
Medical Systems, Inc.,
Branford, CT 06405, 1989.
20. Olivieri, V.P., "Bacterial
Indicators of Pollution," in
Bacterial Indicators of
Pollution, W.O. Pipes, ed., CRC
Press, Boca Raton, FL, Chapter
2, 1982.
21. Stratman, S., "Rapid Specific
Environmental Coliform
Monitoring," Am. Lab, vol. 20,
1988, pp. 60-64.
22. Snyder, A.P., Shoff, D.B.,
Eiceman, G.A., Blyth, D.A. and
Parsons, J.A., Anal. Chem.,
1991, in press.
15.
CAM Chemical Agent Monitor;
Commercial brochures from
Graseby Ionics, Ltd.: Watford,
England, 1988.
108
-------
TABLE 1. COMPARISON OF MICROORGANISM DETECTION BY IMS TO OTHER
TECHNIQUES
Total number
of bacteria
8°c
10i
105
io7
107
10-
10
11
1
2.7xl04
10?
IO4
IO5
5xl07
IO5
10^
0.5
3
9
4
4.25
0.5
0.25
0.25
3.3x10-
Time
(hr) Technique
8.5 gas chromatography
1 radiometry
1.5 electrochemical
0.5 organism growth
0.25 polarized light
scattering
0.4 light-addressable
potentiometric sensor
1 excitation-emission
matrix
3-laser flowthrough
cytometry
enzyme-linked
lectinosorbent assay
H2/CO2 evolution
gfucuronidase enzyme
extracellular enzyme
aminopeptidase enzymes
extracellular enzymes
extracellular enzymes,
nutrients
0.25 CAM
Response
Reference
ethanol metabolite 1
CO metabolite 2
H« metabolte 3
electrical impedance 4
Mueller matrix 5
redox potential 6
fluorescence 7
fluorescence 8
lectin-conjugate 9
visual/ gas bubbles 10
fluorescence 10
colorimetric 11
fluorescence 12
fluorescence 13
fluorescence 14
vapor metabolite this
study
TABLE 2. ENZYME/SUBSTRATE BIOCHEMICAL REACTIONS PROBED IN MICROORGANISMS
PRESENT LIMIT
ORGANISM
E. coli
E. coli
Bacillus subtilis
B. subtilis
ENZYME
PROBED
/3-galactosidase
Lipase
/3-galactosidase
Lipase
SUBSTRATE
ONPG
ONP acetate
ONPG
ONP acetate
OF DETECTION
(Bacterial Cells)*
3.3 x IO3
6 x IO5"
IO5**
IO3
*Within 15 minutes
**No signal observed at the given concentration
109
-------
coll + e«l«cto»« + ortho-nlfro(>h»nol
•ruym* (ONP)
FIGURE 1. PICTORIAL REPRESENTATION OF THE £. COLI/EETA-SAIACTOSIDASE BIOCHEHICAL
REACTION KITH THE OHP6 SUBSTRATE.
HAND-HELD VAPOR DETECTOR
or0*
C3
-KO,
-^
Small lon« trav«t («al»r than larg* loni In «n •:*ctrlcal Br«dl»nl
FIGURE 2. PICTORIAL REPRESENTATION OF THE ONP DETECTION EVENT WITH THE CAK
HAND-HELD MONITOR. REFERENCE 17 PROVIDES DETAILS OF THE OPERATION
OF THE CAM.
IMS
SCAN
NUMBER
(SEC)
6.2
11.7
KSEC
FIGURE 3. ION MOBILITY SPECTRUM OF ONP IN THE NEGATIVE MODE. THE PEAK AT 6.2
MSEC REPRESENTS THE BACKGROUND ION SIGNAL AND THE PEAKS THAT LIE AT
9.1 MSEC REPRESENT ONP AT DIFFERENT RELATIVE CONCENTRATIONS.
110
-------
32
S.2
MSEC
9.1
FIGURE 1. ION MOBILITY SPECTRUrt IN THE NEGATIVE ION MODE (A) OF AN INOCULATION
OF 8x10^ FECAL BACTERIAL CELLS ON A-FILTER PAPER STRIP. (B) OF ONPfi
SOLUTION ON A FILTER PAPER, (C) AFTER 10 KIN. FROM AN ONP6 SOLUTION
ADDED TO AN INOCULATION OF SxlO11 FECAL BACTERIAL CELLS ON A STRIP OF
FILTER PAPER. A PEAK AT 9.1 RSEC, DUE TO ONP, ONLY APPEARS VHEK BOTH
OHP6 AND BACTERIAL CELLS ARE PRESENT.
J
w**wvrK'»»
MiU^uv-'
T~
6.2
KSEC
9.1
FIGURE 5. (A) 15 HIK. AND (B) 15 HIM. ION MOBILITY SPECTRA OF A REPLICATE FECAL
BACTERIA EXPERIMENT (REFER .TO FIGURE 1C FOR DETAILS).
FIGURE 6.
ION MOBILITY SPECTRA OF ONP LIBERATED FROM THE REACTION OF lO4 £. £flU
CELLS AND OKP6 KITH AN INCUBATION AT 58°C FOR (B) 5 HIN (SHADED AREA),
(C) 10 HIK, (D> 15 «IN, 20 HIK. FRAME A REPRESENTS THE ION MOBILITY
SPECTRUM OF A BLANK CONSISTING CF TWO HICROLITERS OF BUFFER AND OKP6
SOLUTIONS ON A PIECE CF FILTER PAPER,
6.2 9.1
nstc
111
-------
6.2 9.1
RSEC
FIGURE 7. ION MOBILITY SPECTRA OF ONP LIBERATED FROM THE REACTION OF J.JxlO3
£. COL1 CELLS AND ONP6 KITH AN INCUBATION AT 38°C FOR (B) 5 MIN,
(C) 10 MIN, (D) 20 MIN. FRAME A REPRESENTS THE ONP6 BLANK. NOTE
THAT ONLY FRAME D SHOWS A CLEAR ONP RESPONSE OVER BACKGROUND.
6.2 9.1
KtC
FIGURE 8. REPLICATE EXPERIMENT OF FIGURE 7 EXCEPT THAT SPECTRUM D WAS TAKEN
AT 15 MINUTES. NOTE THAT ONLY FRAME D SHOKS A CLEAR ONP RESPONSE
OVER BACKGROUND.
112
-------
DATA ANALYSIS TECHNIQUES FOR
ION MOBILITY SPECTROMETRY
Dennis M. Davis
Analytical Research Division, Research Directorate
U.S. Army Chemical Research, Development and Engineering Center
Aberdeen Proving Ground, MD 21010-5423.
ABSTRACT
The past several years have seen
the advance of ion mobility
spectrometry (IMS) as an analytical
technique. Most of these advances
have been made in the hardware
development end of the problem, the
result being that portable IMS
devices have begun to appear in the
marketplace. The other end of the
problem, the signal processing and
data analysis techniques, has not
been addressed to the same degree.
Recent attempts at applying data
analysis techniques to IMS data have
been made, and the results are
encouraging. Data processing
algorithms ranging from those which
perform simple tasks to those
performing more difficult tasks have
been developed. Among the algorithms
which will be discussed are
algorithms for measuring the peak
areas of selected peaks of interest
in biological studies, and linear
discriminant analysis for detecting
and identifying industrial chemicals
at, or near their maximum exposure
limits.
INTRODUCTION
When dealing with environmental
issues, there are two points of
emphasis that must be considered.
These two points of emphasis are the
protection of individuals in the
workplace, a task regulated by the
Occupational Safety and Health
administration (OSHA), and the
protection of the environment in
which we live, a task regulated by
the Environmental Protection Agency
(EPA). These two points of
emphasis, while dealing with the
same general problem, are typically
at different ends of the
concentration range of chemical or
biological contamination or
exposure. The concentration ranges
for which one must monitor an
individuals exposure to chemical and
biological contaminants is usually
in the low parts-per-million, ppm,
range to tens of thousands of ppm
[1-3], and is set by Federal law
[3]. The concentration range which
is monitored for environmental
compliance is usually parts-per-
billion, ppb, to low ppm. A useful
method for the monitoring both
concentration ranges at the same
time is ion mobility spectrometry,
IMS.
Ion mobility spectrometry is
based upon the flow, or drift, of
molecular ions through a gas of
uniform temperature and pressure. A
weak electric field is uniformly
applied to the gas in the drift
region of the IMS, causing the ions
to move along the field lines.
These ions continue to drift until
their movement is impeded by
collisions with neutral gas
molecules. Since the electric field
is still being applied to the gas,
113
-------
the ions are accelerated once again
and the process of acceleration and
collision is repeated until the ions
strike the detector. IMS is similar
to Time of Flight mass spectrometry
in that the electric field causes the
ions to drift, but it differs in that
Time of Flight mass spectrometry is
performed under vacuum and there are
few, if any collisions to retard the
ions. The average velocity, vd, of
the ions is determined by millions of
the accelerations and energy-losing
collisions. The time required for an
ion to traverse a known distance in
the drift region of the spectrometer
is the drift time, td.
The average velocity of the
ions, also called the drift velocity,
is related to the strength of the
applied electric field through the
equation
vd = Id /
= KE
(1)
where vd is the drift velocity, ld is
the length of the drift region of the
spectrometer, td is the drift time of
the ion, E is the electric field
strength, and K is a constant of
proportionality. This constant K is
also called the "mobility" of the
ion. The mobility of the ion is
directly dependent upon both the
molecular ion being studied, and the
neutral gas through which the ion
must drift. A more useful constant
which is used in IMS work is the
"reduced mobility" of the ion. The
reduced mobility of the ion, the
mobility of an ion through a gas at
standard temperature and pressure, is
related to the measured mobility of
the ion through the equation
K0 = K (273.15/T) (P/760)
(2)
where T is the absolute temperature
of the gas in the drift region, P is
the total pressure of the gas and the
ions in the drift region, and Ko is
the reduced mobility of the ion.
Because it is often difficult to
measure the temperature and pressure
within the drift region of the
spectrometer, a common practice which
is used in determining the identity
of ions is to measure the ratio of
the reduced mobility of the ion of
interest to that of a known species.
This known species is usually the
reactant ion for the study. If the
neutral gas is air, the reactant
ions are H3O+ when dealing with
positive ions, and 02"when dealing
with negative ions. The ratio of
the reduced mobilities are related
to measurable quantities through the
equation
(K01/K02) = (Ki/K2) = (td2/tdl) (4).
The only parameters which are needed
in the analysis is the ratio of the
drift times for the ions.
The equation for calculating the
mobility of an ion through a gas has
been shown to dependent on the
first-order collision integral
[4,5], which is proportional to the
transport cross section. This
implies that the mobility of an ion
is dependent on the size of the
ions, the shape of the ion, and the
distribution of charge on the ion;
this results in the possibility of
more than one ion having the same
mobility.
In an ion mobility
spectrometer, Figure 1, the sample
is introduced through a sample inlet
probe. This inlet probe contains a
semi-permeable membrane, which
allows only a portion of the sample
to enter the ionization chamber.
The portion of the sample which does
not enter the ionization chamber is
vented through the exhaust. The
carrier flow gas, which is input
directly into the ionization chamber
and the sample are then exposed to
the ionizing source, a 63Ni source
in this work. The ions and the gas
molecules are then allowed to mix
and react in the ionizing chamber.
Typical ion reaction schemes which
take place in the ionization chamber
are shown in Table A. A driving
pulse of known shape and duration is
then applied to the bipolar gating
grid, allowing the mixture to enter
the drift region of the
spectrometer. While in the drift
region, the ions are subjected to an
applied electric field (200 V/cm in
our studies), which causes the ions
to begin their acceleration and
collision process. After the ions
have traversed the drift region,
114
-------
TABLE A
TYPICAL ION REACTION SCHEMES
Typical Positive Ion Reactions
(X is the species to be detected)
Typical Negative Ion Reactions
02~+ AB -> 02 + AB~
02~+ AB -> 02 + A + B~
O2~+ AB -> (AB'O2)~
(AB is the species to be detected)
they strike the collector electrode
The signal is then processed to
produce the ion mobility spectrum.
For those who wish, a more detailed
description of ion mobility
spectrometry can be found elsewhere
[6].
The past several years have
seen the advance of ion mobility
spectrometry as an analytical
technique, with the utility of IMS as
an analytical tool for the rapid
detection of airborne vapors in the
atmosphere being previously
demonstrated [7-10], and computer
techniques for pre-processing IMS
signals have also been presented [11-
12].
EXPERIMENTAL
Equipment
Data were collected on an IMS
spectrometer [Airborne Vapor Monitor
(AVM) from Graseby Analytical,
Watford, Great Britain] and stored on
an IBM Personal Computer. The data
transfer is accomplished using a
Graseby Analytical Advanced Signal
Processing (ASP) board and its
associated software. Each spectrum
consisted of 640 data points, which
was collected at a sampling frequency
of 30 kHz. The other operational
parameters of the AVM are shown in
Table B.
Vapor Generation
The vapors being used in the
linear discriminant data set are
generated with a Q5 vapor generator,
shown in Figure 2 . The Q5 generator
has 16 component parts. These parts
are: (1) an equilibrator assembly,
(2) an air supply (or nitrogen
supply) stopcock, (3) a constant
pressure regulator (stabilizer) for
the air supply, (4) two sampling
bubblers filled with solvent (the
bubbler is not shown Figure 2), (5)
a flowmeter (manometer) for the air
supply, (6) a constant pressure
regulator (stabilizer) for the
diluent air supply, (7) stopcocks
for the stabilizers, (8) a stopcock
shut off the flow of air from the
equilibrator to the mixing chamber,
(9) a flowmeter (manometer) for the
diluent air supply, (10) a mixing
chamber, (11) a reservoir, (12,13)
sampling stopcocks, (14) a reservoir
exhaust stopcock, (15) a charcoal
trap on the exhaust of the reservoir
(not shown in Figure 2) , and (16) a
charcoal canister on the sampling
line after the SAW device (not shown
in Figure 2) .
The equilibrator assembly is the
liquid test reagent container of the
dilution apparatus. Dry air, under
a constant controlled pressure,
flows into the equilibrator. This
air stream passes over the surface
of the test reagent, and becomes
115
-------
TABLE B
OPERATIONAL PARAMETERS FOR THE AVM
Number Of Waveforms To Be Summed - 32
Number Of Samples Per Waveform - 640
Gating Pulse Repetition Rate - 40 Hz
Gating Pulse Width - i80 US
Delay To Start Of Sampling - o us
Sampling Frequency - 30 KHz
Gating Pulse Source Is
** External **
saturated with the reagent vapor.
The equilibrator is maintained at a
constant temperature of 25 °C by
partial immersion in a constant
temperature water bath. Included in
the equilibrator is a porous alumina
cylinder (from Thomas Scientific,
Swedesboro, N.J.) to produce a
greater surface area for the liquid-
vapor equilibration. The dry air-
test vapor mixture flows from the
equilibrator assembly to the mixing
chamber where it is diluted with dry
air to the required concentration of
milligrams test vapor per liter of
dry air.
The flow of air through the
equilibrator is controlled by an in-
line stopcock, a constant pressure
regulator, and a flowmeter. The
stopcock is located at the inlet of
the equilibrator, and acts as the
shutoff valve for the air supply,
from the flowmeter to the
equilibrator. The constant pressure
for the air supply is maintained by
bubbling the dry air through a
constant level of fluid, e.g. water,
in the stabilizer. By raising or
lowering the level of the fluid in
the stabilizer, the air pressure
controlled. The level of the fluid
is raised by adding fluid to the
stabilizer, and lowered by draining
fluid through the stabilizer stopcock
located on the bottom of the
stabilizer. Changing the pressure of
the air supply in this way increases
or decreases the flow of the test
vapor through the dilution apparatus.
Excess air passing through the
stabilizer is vented to the
laboratory hood. The flowmeter, or
manometer, consists of an inner
glass tube, which is graduated in
millimeters, and outer glass tube
through which the air flows, a glass
capillary tube of predetermined bore
size, a cover to seal the capillary,
and a bulb type bottom filled with
colored water, which is connected to
the constant pressure regulator.
The capillary is calibrated such
that the flowrate through the
capillary is known for any water
height. Thus, the flowrate is
determined by the height of the
water in the inner tube, and the
capillary calibration data. The
flowmeter measures the flow rate of
the dry air-test vapor mixture in
milliliters per minute. The flow
rate of the diluent air is
controlled in the same fashion as
the equilibrator air supply with a
larger inside diameter capillary
tube. The flowmeter for the diluent
air is measured in liters per
minute. The nominal concentration
of the test vapor can be calculated
using the equation
C = {(f * p)/([F + f]*P)}
(5)
where C is the nominal concentration
of the test vapor in parts-per-
million by volume, f is flow rate of
air through the equilibrator, F is
the flow rate of the diluent air, p
is the vapor pressure of the test
reagent at the temperature of the
experiment, and P is atmospheric
pressure. Thus, the concentration
116
-------
of the test vapor may be easily
changed by varying either the flow
rate of air through the equilibrator,
or by changing the flow rate of the
diluent air. In practice, it works
best to change the flow rate of the
diluent air, when possible, because
the efficiency of the vapor
generation in the equilibrator
decreases at higher flow rates.
The dry air-test vapor mixture
from the equilibrator and the diluent
air are passed into the mixing
chamber located at the entrance of
the reservoir. The dilute test vapor
is thoroughly mixed by a swirling
circular motion of the air in the
mixing chamber before entering the
reservoir. The reservoir is the
container for the diluted test vapor,
from which samples are taken for
concentration analysis and for
testing purposes. There is a
charcoal canister located on the
exhaust of the reservoir. This
canister serves as a scrubber to
remove test vapors passing from the
reservoir to the atmosphere in the
laboratory hood.
Pre-Processinq of Spectra for Linear
Discriminant Analysis
The pre-processing and data
processing procedure used in the
linear discriminant analysis is shown
in Figure 3. The first pre-
processing step is to determine if
the spectrum has been collected in
the positive (+) or negative (-)
mode. This knowledge is important
since the Graseby ASP board does not
differentiate between the two types
of spectra, i.e. the ASP board
converts all spectra to positive
values. The determination of the
operating mode under which the
spectrum was collected is made by
reading the data file header which
includes a single character which is
used to designate mode. A
preliminary discrimination is made
based on the mode; a spectrum
collected in the negative mode has
no chemical semblance to a spectrum
collected in the positive mode. Once
the mode has been determined, it is
necessary to determine the time at
which the reactant ion peak (RIP)
appears. The reactant ion for the
AVM. O2~in the negative mode and
H3CT~ in the positive mode, is the
species which transfers the charge
to the chemical species being
analyzed. The location of the RIP
must be determined for each
spectrum, if possible, because the
location is affected by changes in
temperature, pressure, and relative
humidity. If no RIP is found, then
one must assume the RIP is located
at the same time as the RIP for the
previous spectrum. After
determining the time at which the
RIP appears, the spectrum is
normalized to create a dimensionless
X-axis. To do this, each value on
the X-axis was divided by the value
position of the reactant ion peak.
For negative ion spectra collected
at, or near, sea level, the peak
position, with the maximum intensity
between 6.0 and 7.0 milliseconds
drift time was used for the
identification of the reference ion
peak. For positive ion data, a
value between 6.5 and 7.5
millisecond drift time was used as a
window in which to find the
reference ion peak. This reference
window is easily adjusted for
spectra collected at other altitudes
or pressures by multiplying the
window values by the ratio of the
operating pressure to atmospheric
pressure at sea level. This new
spectrum also appears as a pseudo-
"Reduced Mobility" spectrum which
has a dimensionless X-axis
corresponding to a Ratio of Drift
Times, TR. Only the data in the
range 0.5 to 3.0 along the TR axis
are used. A cubic spline is then
applied to the spectra such that
every spectrum has the same data
spacing with respect to the Ratio of
Drift Times axis. The IMS data
files used in this study have data
points every 0.005 TR.
LINEAR DISCRIMINANT ANALYSIS
Traditionally, much of the
effort associated with the analysis
of the IMS spectra has been left to
the chemist. In an effort to aid in
the preliminary identification, a
117
-------
personal computer (PC) based spectrum
identification package has been
developed. This package, written in
Microsoft Fortran, uses a linear
discriminant function for its
identification, and consists of three
separate programs. These programs
are: IMSDISC, a program which reads
selected data files from the PC and
builds a discrimination data set;
TRAIN, a program which analyzes the
discrimination set and calculates the
linear discriminant function that
best isolates the data of interest
from the interferant data; and
IMSIDENT, a program which reads the
data to be analyzed and identified
and calculates its linear
discriminant value.
Linear discriminant analysis, one
of the most basic forms of pattern
recognition used by scientists, is
used as a supervised learning
technique. In supervised learning
techniques, the computer learns to
classify the samples being analyzed
based on knowledge about the samples;
in this study, the samples either
belong to the class of chemicals you
wish to identify, or they do not.
The goal of the learning is to
develop a classification rule, the
linear discriminant function, which
allows the validity of the
classification to be tested and
ultimately to properly classify
unknowns .
The linear discriminant function
has the general form
n
g(x) = w0 + X Wi xi (6)
where wo is the threshold vector, wi
is the weight vector, X£ is the
response vector, and g(x) is the
response function. The discriminant
function, g(x) is determined by
choosing those variables xj[ with
characteristics which differ between
the groups being classified. These
variables are then linearly combined
and weighted such that the groups are
as statistically different as
possible. This linear combination of
variables is calculated using the
perceptron convergence criteria.
The perceptron [13-15] is a
pattern recognition procedure which
consists updating the weight vector
by considering only those patterns,
or spectra in this work, which have
been misclassified in the training
set. Each misclassified pattern is
considered in turn, with a fraction
of each misclassified spectrum being
added to the weight vector. This
procedure is continued until all of
the spectra are classified
correctly, or until it is determined
that the procedure fails to converge
to a satisfactory solution.
In this software package, the
three programs are run separately,
but are still inter-related. The
first program, IMSDISC, uses a file
called NAMES. NAMES is simply the
file that contains the names of the
individual data files to read, and a
value that tells the program whether
the file is to be treated as the
sample or as an interferant. The
data from the individual data files
is then treated such that all the
files are compatible with respect to
time spacing between data points,
delay to start of data sampling, and
number of data points. To
accomplish this, IMSDISC uses a
spline function to interpolate and
fit the data. After the data has
been treated to fill the
compatibility requirement, the
discriminant threshold is set to
zero by multiplying all interferant
spectra by negative 1, (-1). The
sample spectra are left unaltered.
The data is then stored in a
discriminant data file.
The second program, TRAIN,
develops a linear-discriminant based
on the perceptron convergence
criteria. TRAIN prompts the
operator for the name of the input
discriminant file that was created
with the program IMSDISC. It reads
the data from the discriminant data
set, accepts input for the values of
a scaling factor, between
0.000000001 and 0.1, and the number
of iterations to perform using this
scaling factor. In practice, it is
generally necessary to use a series
of decreasing scaling factors and
iterations to calculate the linear
discriminant function which best
differentiates the samples and the
interferants. After the linear
118
-------
discriminant function has been
calculated, the linear coefficients
are written to a file on the computer
disk for use by the last program.
These first two programs, IMSDISC and
TRAIN, are the time consuming
programs and are run only when a new
compound is to be added to the
database.
The third program in this
package, IMSIDENT, uses the linear
discriminant values created with the
program TRAIN. Thus, it is dependent
on the first two programs in the
package. IMSIDENT can be used in one
of two possible configurations; the
first configuration is as a stand-
alone program, and the second is that
it can be incorporated into a data
collection program for real time
identification of an unknown
environment. In the stand-alone
configuration, the program prompts
the operator for the name of the data
to analyze. The program reads the
data, and performs a spline
interpolation to make the data
compatible with the discriminant data
sets. Next, the program reads a file
named COEF.FIL that contains the
names of the coefficient files. The
linear discriminant value is then
calculated. i'f the linear
discriminant value is positive, an
alarm message is generated which
notifies the operator that the
spectrum has been identified. No
message is generated if the
discriminant value is negative. The
results of the identification process
are then written to a file named
ALARM.RPT for later use, and the
program then prepares to read the
next data file to be analyzed.
In the second configuration, the
program functions as a real time
monitor. The name of the data file
to be analyzed is passed from the
data collection program to the
IMSIDENT package rather than
prompting the operator for the name
of the data file to analyze. The
spline interpolation is then
performed on the data, and the linear
discriminant value is calculated. If
the discriminant value is positive,
the alarm message is generated; no
message is generated if the
discriminant value is negative. The
results of the identification
process are written to a file named
ALARM.RPT for later use.
DISCUSSION
The program package was
developed for use with the Graseby
Ionics Advanced Signal Processing
(ASP) board, the Graseby Airborne
Vapor Monitor (AVM), and a Zenith
286 PC. Using this hardware and the
linear discrimination package, it
has been possible to identify and
semi-quantitate the presence of 15
common chemical vapors in air.
These compounds, most of which are
of industrial importance, and the
levels at which the Occupational
Safety and Health Administration
(OSHA) have determined them to be
hazardous are shown in Table C, with
the ion mobility spectra of these
compounds shown in Figures 4 through
21. When the software is used in
the stand-alone configuration (i.e.,
separate from the data collection
routines) and using the Zenith 286
PC, the presence of these compounds
can be determined and the compound
identified in less than ten seconds.
This includes the time necessary to
perform the spline interpolation and
the calculation of the discriminant
value for the data; however, this
does not include the time required
to create the discriminant
functions.
The results shown in Table D
are from the evaluation of a series
of files used to determine the
presence of N-Methyl Formamide. The
"All Clear" report indicates that
the IMSIDENT program does not find
any similarities between the N-
methyl formamide test spectrum and
the spectra of the fifteen compounds
stored in the database. The report
of an alarm indicates that the
program did find similarities in the
spectra, and the magnitude of the
discriminant is a measure of the
amount of similarity.
It is not really surprising
that there are a number of false
positive alarms indicating the
presence of diethyl ether. Older
119
-------
versions of the AVM used an acetone
dopant within its detection system,
whereas newer versions of the AVM use
water vapor in the atmosphere as the
dopant. This dopant in the older
AVM's results in the presence of an
acetone reactant ion. This reactant
ion is the ionic species which is
responsible for transferring the
ionic charge to the chemical compound
being studied. All of the spectra
used in the discrimination functions
were recorded using water as the
reactant ion. Thus, the discriminant
functions have not been trained to
eliminate the possibility of alarming
on a spectrum which has an acetone
reactant ion peak, and an alarm is
reported. Examination of two
representative spectra for which an
alarm was reported, shows the
similarity of the IMS spectrum for
the diethyl ether, the lower trace
in Figure 22 (ETHER in Table D) and
N-methyl formamide background
spectrum, the upper trace in Figure
22 (\AVM\DATA\nmfoOOOO.ACQ in Table
D). The location of the reactant
ion peak does not appear at the same
time as does the diethyl ether peak,
however the ba.nd shapes are similar.
If the discriminant function is
trained to ignore the acetone
reactant ion peak, one does not get
an alarm. Results of identification
procedure with the acetone reactant
ion peak being ignored is shown in
Table E.
TABLE E
File "ALARM.RPT" for
N-Methyl Formamide Analysis
with Acetone Reactant ion Ignored
ALL CLEAR FOR FILE \AVM\DATA\nmfoOOOO.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0001.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0002.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0003.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0004.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0005.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0006.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0007.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0008.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0009.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0010.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0011.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0012.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0013.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0014.ACQ
ALL CLEAR FOR FILE \AVM\DATA\nmfo0015.ACQ
120
-------
LITERATURE CITED
1. Threshold Limit Values and
Biological Exposure Indices for 1989-
90, American Conference of
Governmental Industrial Hygienists,
Cincinnati, OH, (1989) .
2. NIOSH Pocket Guide to Chemical
Hazards. U.S. Department of Health
and Human Resources, National
Institute of Occupational Safety and
Health, Washington, D.C., (1985).
3. Code of Federal Regulations, 29
CFR 1910, Subpart Z - Toxic and
Hazardous Substances, 1910.1000 - Air
Contaminants. 19 January 1989.
4. E. W. McDaniel and E. A. Mason,
The Mobility and Diffusion of Ions in
Gases. Chapt. 2, John Wiley and Sons,
New York, 1973.
5. H. E. Revercomb and E. A. Mason,
Anal. Chem.. 1975, 47. 970.
6. Plasma Chroroatocrraphv. Carr, T.
W., ed., Plenum Press, New York,
1984.
7. G. A. Eiceman, A. P. Snyder, and
D. A. Blyth, Inter. J. of Environ.
Anal. Chem.. 1989, 38. 415.
8. G. A. Eiceman, M. E. Fleischer,
and C. s. Leasure, Inter. J. of
Environ. Anal. Chem., 1987, 28. 279.
9. c. s. Leasure, M. E. Fleischer,
and G. A. Eiceman, Anal. Chem..
1986, 58., 2141.
10. J. M. Preston and L. Rajadhyax,
Anal. Chem., 1988, 60, 31.
11. D. M. Davis and R. T. Kroutil,
Anal. Chim. Acta. 1990, 232. 261.
12. D. M. Davis and R. T. Kroutil,
in P. Jurs (Ed.), Computer-Enhanced
Analytical Spectroscopy, Vol. 3,
Plenum Press, New York, 1990, (in
press).
13. F. Rosenblatt, Principles in
Neurodvnamics: Perceptrons and the
Theory of Brain Mechanisms, Spartan,
New York, (1962).
14. R.O., Duda and P.E. Hart, Pattern
Classification and Scene Analysis,
Wiley, New York, (1973).
15. Y.-H. Pao, Adaptive Pattern
Recognition and Neural Networks,
Addison-Wessley, New York, (1989).
121
-------
CARRIER FLOW C6TB-, I 1 DRIUINC PULSE
CfiRRIERFLOWCAIR) _| |_ TO CRATING GRID
I ^-GATING
O . GRID
DRIFT
REGION
o
ft
>
•
1J
* O
9 o o o
* o o
i
1 IONIZING
X
I
SAMPLE
IONS
OUTPUT SIGNAL
AMPLIFIER
MICROPROCESSOR
SVSTEM
REGION
TO PUMPS
AIR IONS
1 SAMPLE IONS
COLLECTOR
ELECTRODE
Figure 1. Schematic diagram of an ion mobility spectrometer.
122
-------
CAPILLARY
>IR SUPPLY CONNECTION
KHNO FLOWNETEH
NITROSEN SUPPLY
SAMPLING LINE
VACUUM SUPPLY
Figure 2. Schematic diagram of the Q5 vapor generator.
123
-------
LINEAR DISCRIMINANT
ANALYSIS
ACQUIRE DATA
DETERMINE
MODE(+ OR-)
J_
LOCATE
REACTANT ION
PEAK (RIP)
NORMALIZE
SPECTRUM WITH
RESPECT ID RIP
IF NO PEAK
PRESENT,
ASSUME
LOCATION FROM
LAST SPECTRUM
COMPARE
RESULTANT
SPECTRUM WITH
THE REFERENCE
SPECTRA
IF DISCRIMINANT
VALUE < 0 ; ALL
CLEAR
IF DISCRIMINANT
VALUE > 0 ; SET
ALARM FOR
THAT SPECIES
Figure 3. Block diagram showing the steps taken when performing a linear
discriminant analysis on an ion mobility spectrum.
124
-------
in
•4J
•H
c
D
3)
L
03
L
-M
•H
I]
L
-------
in
-M
•H
c
D
L
ID
L
4J
•H
£!
L
-------
01
-M
•-I
C
D
31
L
05
L
4J
•H
U
L
-------
in
4->
•H
c
D
31
L
ID
L
4-1
•H
U
L
•H
n
L
-------
III
•p
•H
c
D
L
ID
L
•M
•H
a
L
-------
in
+j
•H
c
D
31
L
ID
L
-P
•H
12
L
-------
in
-M
•H
c
D
31
L
ID
L
4J
•H
J]
L
-------
in
-u
•H
c
D
Dl
L
ID
L
+J
•H
SI
L
-------
in
4J
•H
c
D
31
S.
IU
L
4J
•H
12
L
-------
Arbitrary Units
ro.
CJ
H- CO
in •
ID en
n
o H>
2 ®
D. •
in -4
CO
(0
(0
A)
ro
H*
•
CO
03
Figure 22. Typical IMS spectra analyzed using linear discriminant analysis.
Spectra show the similarities often encountered in IMS spectra. Spectrum A
is diethyl ether, and spectrum B is an acetone reactant ion spectrum.
134
-------
DISCUSSION
DREW SAUTER: Perhaps you could explain certain aspects that have hindered
adoption of ion trap mass spectroscopy, basically ion molecule reactions. One of
the things I've run into, and others have, is that in certain limited scenarios, you
can probably define your ion molecule chemistry.
PETER SNYDER: Yes
DREW SAUTER: But the truth of the matter, and correct me if I'm wrong, is
that you can have unknown reacting ions in the sample. In an unknown situation,
it would seem that you could actually get spectra that were sample dependent.
Basically, would you see IMS being more useful as a sort of screening tool on
relatively limited scenarios, as opposed to a tool that could offer more general
analysis capabilities?
PETERSNYDER: Well, I can't disagree with that when you just talk about IMS
by itself. Because of the potential complicating responses that can occur if your
environment is not controlled, anything can happen.
DREW SAUTER: What I mean though is in the real environmental world, a lot
of samples have a lot more than one compound, and not only that do they have
a lot more than one compound, recognizing that you can separate things by GC's,
they tend to have different concentrations.
PETER SNYDER: Yes.
DREW SAUTER: Hence if they have different concentrations, and there's ion
molecule reactions going on. you have them going on with some rate constant.
They're producing different populations of ions, and hence a different sample
dependent spectra. That strikes me as a significant drawback, despite all the
grand things that you've shown.
PETER SNYDER: You have to consider what IMS is based on? IMS is based
on ion molecule reactions, and that can be broken down into proton affinity and
electron affinity by and large. So then you have to look at what kind of
compounds are responding.
DREW SAUTER: But there's al so a concentrat ion term that you showed in your
graph.
PETERSNYDER: Yes, absolutely. Concentration is very important. I guess the
difficulty in response comes then when you get to phosphonate compounds or
phosphoryl compounds that are very sensitive to proton affinity. They get that
proton very nicely, and by and large to the exclusion of many other compounds,
even in their presence, or at relatively high concentrations. Ammonia, probably
would take exception too. That might be a complicating factor.
But in most cases, phosphoryl compounds really come through, and that's one of
the strengths of the chemical agent monitor, in terms of looking for phosphoryl-
oased nerve agents.
As you go down to amines, esters, ketones and alcohols, the relative proton
affinities are not as wide.
STEVE HARDEN: I'd like to just comment on that before we get on to the next
question, and say that yes, you have indeed hit upon one of the problems with ion
mobility spectrometry for analyzing real-world mixtures.
The reason the Army has developed it for their purposes is that the compounds
they're interested in either have such extreme proton affinities, or extreme
electronegativities, and that the sensitivity is very high for those compounds. So
it works for our purposes, and it may not work for some environmental purposes
that you mentioned, because of this mixture problem.
It also points out one of the needs and requirements in this unknown analysis, or
analysis of unknowns, for preparation of sample you mentioned the GC/MS
system. We'll hear some more about that in our next paper.
But one can also point out that in some of the data (in this paper) for some
compounds that do have a high electronegativity, can be picked out using these
techniques that we were talking about, and we can then point out the fact that yes,
indeed, that material was present.
That little bump on the side of that peak was, I think, the mustard, which is an
Army compound of interest. The bump was on the side of a peak of phenol,
phenol being in much greater concentration.
In previous sensitivities and single processing techniques, we can bring it even
more if we used preparation of samples. However, you do separate samples at the
expense of complexity of instrumentation, and that's one reason why the Army
hasn't pursued that to this particular point. So we have.
HERB HILL: Fora long time now we have been using ion mobility spectrometers
as a chromatographic detector, basically because we feel that there really are
problems with interferences, except for very specific cases.
I'm really excited to see us beginning to talk about the use of, what 1 call
chromatographic filters on the front end of IMS, for field monitoring. We've
done studies, for example, treating IMS as a chromatographic detector, and you
can see that the interferences under conditions like that are no worse than you
would have with a flame ionization detector, an electron capture detector. The
quantitative value of IMS is acceptable in any range. It's as good as any of the
standard chromatographic detectors that we have. We've published papers in
which we've put interfering species in, compared them to an FID, and ECD, an
IMS, and you see that the quantitative value of the data is fine, it's good in IMS.
When you add the chromatograph controls on the front end, you can do dioxins.
We do ligands in blood analysis, we do a variety of very small, minute trace
compounds in very, very complex mixtures, as well or better, than you can be a
lot of techniques.
And it should apply very well to field analysis for portable, if you put a portable
GC on the front end of that.
PETER SNYDER: Yes, you' re absolutely right. And the literature that you have
published over the past decade and a half, attests to that. There's many different
sample matrices that Professor Hill has looked at with very good resolution,
depending upon the column characteristics. There has been a lot of good
information coming out of that, using an IMS as a detector.
So basically the newer innovative topic we're looking at here, is using the hand-
held version of the IMS, to see how far we can go with that.
135
-------
ION MOBILITY SPECTROMETRY AS A FIELD SCREENING TECHNIQUE
Lynn D. Hoffland and Donald B. Shoff
Analytical Research Division,Research Directorate,
U.S. Army epical Research,Development and Engineering Center,
Aberdeen Proving Ground, MD
1. INTRODUCTION
Ion Mobility Spectrometry (IMS),
also called Plasma Chromatography, is
used to detect trace quantities of
organic va,pors in gaseous mixtures.
Several researchers over the past 15
gears have demonstrated the utility of
mobility detection for a variety of
organic compounds. 1~H Quantities as
low as
10-10 grams of nitrosamines have been
reported^^.
IMS is a conceptually simple
technique that relies on the drift time,
or time of flight, of molecular, or
cluster, ions through a host gas as a
means of differentiation. This differs
from classical mass spectrometry in that
there is little, if any, fragmentation
and the ions are not mass analyzed.
Detailed theory can be found
elsewhere.13'15 The ions are
differentiated by charge and by
mobility. The reduced mobility KQ
(corrected for standard pressure and
temperature) is expressed as
Kg = 42.51 D
where D is the sealer diffusion
coefficient of Pick's law. This reduced
mobility KQ is catalogued and identified
for each ionic species present.
2. EXPERIMENTAL
The work was performed on a MMS-290
Ion Mobility Mass Spectrometer (POP,
Inc.). shown in figure 1 and an Air
Vapor Monitor made by Graseby Ltd.
The PCP, Inc. MMS-290 spectrometer
used in these experiments consists of an
ion mobility spectrometer followed by a
quadrapole mass spectrometer coupled to a
Nicolet signal averager with a computer
interface for storage, data manipulation
and display.
There are four modes of operation
for the -MMS-290. In the total ion mode
the MMS-290 acts as an ion mobility
spectrometer. Ions are gated into the
drift region and detected by the
electrometer. All ions detected are
averaged, stored and displayed. In the
integral ion mode the mass spectrometer
is the detector instead of the
electrometer. Again, all ions are
detected, averaged, stored and displayed.
There is no mass analysis in this mode.
It is used to check that the ion
distribution is not changed by traveling
the extra distance through the mass
spectrometer. The third mode is the mass
spectrum. The shutter grid is held open
to allow a continuous stream of ions into
the mass spectrometer which is mass
analyzing the ions. This provides a mass
spectral scan of the total ion flux. The
last mode of operation is the tuned ion
mode where the MMS-290 is operated as in
the integral ion mode but the mass
spectrometer is only detecting one mass
ion at a time. This shows which mass
ions are associated with each mobility
peak.
The Airborne Vapor Monitor (AVM)
used in these experiments consists of an
IMS described above with a membrane inlet
and internal electronics for signal
processing and alarm. It operates in
both positive and negative ion mode, has
137
-------
no internal display but can be
interfaced to a personal computer for
display and storage of the IMS spectra.
The AVM has only an electrometer, it has
no mass spectrometer to mass analyze the
ions, and it operates as the total ion
mode of the MMS-290.
Air, or the sample gas, is drawn
into the ionizing region and is ionized
by 60 keV Beta rays from a radioactive
Ni63 source. A potential exists
between the ionizer and the collector
forcing the ions in the direction of the
shutter grid. The closed shutter grid
neutralizes all ions reaching it. The
shutter is pulsed open for approximately
0.1 millisecond (msec) and a cross
section of the ions flow into the drift
region. The shutter closes again
isolating a short pulse of ions that
travel down the drift region propelled
against the drift gas flow by the
potential on the collector. The ions
are differentiated by their charge in
the electric field and their mobility in
the drift gas (velocity V*)
Vd = K E.
The IMS differentiates the ions because
by the time that they reach the shutter
grid the ion molecule reactions have
equilibrated and in the drift region no
more reactions take place.
As the separated ions reach the
collector, they are detected by a fast
electrometer, and a current is generated
directly proportional to the number of
ions. The resultant spectrum is
depicted in figure 2 **>.
The highest KQ ions (C+) are
usually smaller or more compact followed
by the slower ions B+ and A+, in time.
Both positive and negative ion
formation of reactant and product ions
are multistep processes. Good, Durden,
and Keburle*7 have determined the
mechanisms involved in positive reactant
ion formation:
+ e-
+ 2 N
OH
(H20)2H+
The size of the resultant reactant ion
water clusters depend upon the relative
humidity but generally water chemistry
dominates the positive ion mode. The
water ion may cluster directly with the
sample molecule M or, as is the case
more often, the sample molecule
abstracts the proton from the water
cluster and then may attract more or
less water molecules depending on the
humidity. At high concentrations the
sample molecules may form dimers with a
proton. whenever there is some other
molecules present with a higher proton
affinity than water they may replace the
water in the above mechanisms i.e.
Acetone or NH-,. So, in figure two peak
C may be the reactant ion, B the
hydrated monomer, and A the protonated
dimer.
Negative reactant ion formation as
summarized by Spangler and Collins-1**
include the following:
e~(thermal) + 0,
"here n = 1,2.
The sample molecule can cluster with the
O,- or abstract the O2 from the CO2-
As can be easily seen, the chemistry can
be quite involved before any products
are formed.
138
-------
The operating parameters for the
MMS-290 were;
Cell length
Operating voltage
Electric field
Carrier gas
Drift gas
Cell Temperature
Pressure
Drift distance
15 cm
3000 volts
200 volts/cm
200 ml/min
500 ml/min
40 °C
Entered Daily
10 cm
The AVM was operated as received
from Grase&y Analytical, Ltd. (Watford,
Herts, UK) . Signals from this IMS were
processed with a Graseby Analytical,
Ltd., advanced signal averaging (ASP)
board installed in an IBM PC/AT
computer. Known or approximate operating
conditions were;
inlet flow
Drift tube temperature
membrane temperature
reaction region
drift region
field gradient
The samples were generated using a
0-5 apparatus, (where a saturated vapor
stream is mixed with a high volume
diluent dry air stream). By varying the
quantities of both streams the
concentration of sample in the diluted
vapor stream was controlled. The
resulting diluted vapor stream was
sampled by either the IMS /MS inlet or
the AVM. All samples were used as
received from the manufacturer. The
concentration of the saturated vapor
500 ml/min.
ambient
70 °C
2.2 cm
"3.8 cm
~ 200 V/cm
stream was calculated from vapor pressure
data or from the Antoine equation.
3. RESULTS AND DISCUSSION
The data following are an example of
the power of this detection system to
high concentration vapors of acetic acid.
The acetic acid was used "as is", and, as
will be shown, was contaminated with
acetic anhydride(as is often the case).
The target concentration for acetic acid
139
-------
detection with the AVM was the Time
Weighted Average (TWA) of 10 ppmiy, the
Short Term Exposure Limit (STEL) of 15
ppmi9, up to the Immediately Dangerous
to Life or Health (IDLH) level of 1000
ppm20. Figures 3-6 show the response of
the AVM for these three concentrations.
The identity of the peaks in the
above data was determined with the
IMS/MS in the following manner. First,
the reduced mobility is calculated for
each peak. Since the reduced mobility is
a factor of pressure and temperature and
these vary in the AVM and between the
AVM and IMS/MS, a drift time ratio is
calculated by dividing the specie
mobility by the reactant ion mobility
(both are under the same temperature and
pressure). Then, the IMS/MS is operated
in the total ion mode and the integral
ion mode to check that there is no
effect between the different inlets of
the AVM and the IMS/MS and that the mass
spectrometer entrance of the IMS/MS does
not change the specie (figures 7 and 8) .
The first thing noticed is that the
pinhole inlet of the IMS/MS is much more
sensitive than the membrane of the AVM.
The membrane is required, however, to
keep too much water and contaminants
from spoiling the sensitive IMS cell.
So, allowing for the difference in
sensitivity, the mobility spectrum of
the IMS/MS is compared with the AVM to
correlate the mobility peaks between the
two instruments. Once confirmed, the
mass spectrum is taken to determine what
mass species are the major contributors
to the ion mobility spectrum (figure 9).
Then, each mass is scanned in the tuned
ion mode to determine to what peak in
the mobility spectrum each mass
contributes (figure 10). As can be
seen, in this low concentration, the
masses 55,73,83,101,and 129 are all
hydrates and "nydrates" of the H
"reactant ion" and the masses 79, 97,
125, and 153 are hydrates and "nydrates"
of the H+ acetic acid monomer. The
concentration is then increased and the
analysis series is repeated. As the
concentration increases the mass
spectrum becomes more complicated but
assignments can be made bases upon past
experience. Since, at this time, we do
not have the capability there is no
secondary mass fragmentation for
confirmation of these species. Tables 1
indicates the assignments for each mass
fragment in the mass spectrum. Table 2
is a list of the mobility ratios and the
assignments for each mobility peak seen
at the various concentrations.
CONCLUSION
This example of acetic acid illustrates
the potential of this hand held ion
mobility spectrometer to differentiate
between regulated concentrations of
hazardous chemicals. In support of
another program this work has been
extended to identification of these
regulated concentrations (TWA, STEL, and
IDLH) of 15 other solvent chemicals.
Although limited in scope, by extending
this data base the AVM could be used as a
field screening device and as a safety
device for field personnel.
140
-------
TABLE 1
AMU
55
73
79
83
97
101
125
129
153
Specie
H+(H20) 3
H+(H20)4
m H+(H2O)
H+(H2O)3+^2
m H+(H2O)2
H+(H20)4+N2
m H+(H2O)2+N2
H+(H20)4+2N2
m H+(H2O)2+2N2
Comment
reactant ion
reactant ion
monomer hydrate
reactant ion
monomer hydrate
reactant ion
monomer "nydrate"
reactant ion
monomer "nydrate"
Mobility Ratio
1.00
1.08-1.09
1.18-1.24
1.34-1.35
1.47-1.48
TABLE 2
Assignment
fi+(H20)x(N2)y
m »+(H20)x(N2)y
m2 H+(H20)X(N2) y
m n H+(H20)x(N2)y
n2 H+(H20)x(N2)y
Reactant Ion
Acid Monomer
Acid Dimer
Acid Anhydride
Anhydride Dimer
141
-------
References
1. Cohen, M. J., and Karasek,F. W, ,
"Plasma chroma tography - A new dimension
for gas chroma tography and mass
spectrometry", J. Chromatogr. Sci. £,
(1970), 331.
2. Karasek,F. W., and Kane,D. M. ,
"Plasma chroma tography of the n-alfcyl
alcohols", J. Chromatogr. Sci., 10,
(1972), 673.
3. Karasek,F. W. , Tatone,0. S., and
Denney,D. W. , "Plasma chromatography of
the n-alfcyl halides", J. Chromatogr. 87,
(1973), 137.
4. Karasek,F. IV., Tatone,O. S. , and
Kane,D. M. , "Study of electron capture
behavior of substituted aromatics by
plasma chromatography", Anal. Chem, 45,
(1973), 1210.
5. Karasek,F. W . , "Plasma
chromatography", Anal. Chem. 46 , (1974),
710A and references.
6. Karasek,F. IV., Denney,D. W. , and
Dedecker,B. H. , "Plasma chromatography of
normal alkanes and its relationship to
chemical ionization mass spectrometry",
Anal. Chem, 46_, (1974), 970.
7. Karasek,F. W.f and Denney,D. W. ,
"Detection of aliphatic N-nitrosamine
compounds by plasma chromatography",
Anal. Chem., 46, (1974), 1312.
8. Karasek,F. W., Malcan,A., and
Tatone,O. S., "Plasma chromatography of
n-alkyl acetates", J. Chromatogr., 110,
295.
9. Karasek,F. tf., Kim,S. H. , and
Rokushika,S. , "Plasma chromatography of
alfcyl amines", Anal. Chem., 50, (1978),
2013.
10. Spangler,G. E. , and Lawless,?. A.,
"lonization of nitrotoluene compounds in
negative ion plasma chromatography",
Anal. Chem., 50, (1978), 884.
11. Shumate,C., St. Louis,R. H. ,
Hill, Jr.,H. H., "Table of reduced
mobility values from ambient
pressure ion mobility spectrometry",
J. Chromotogr., 373, (1986), 141.
12. Karasek, F. W. and Denney,
D.IV., "Detection of Aliphatic N-
Nitrosamine Compounds by Plasma
Chromatography", Anal. Chem.46,No.
9, (August 1974), 1214-1312.
13. McDaniel,E. W. , and Mason,E.
A., The Mobility and Diffusion of
Ions in Gases, John Wiley and Sons,
New York, (1973) .
14. McDaniel,E. W., Cermak,V.,
Dalgarno,A., Ferguson,E. E., and
Friedman,!.., Ion-Molecule Reactions,
John Wiley and Sons, New York
(1970).
15. Loeb,L. B., Basic Processes of
Gaseous Electronics (2nd edition),
UniversityofCalifornia Press,
Berkeley (1960).
16. Spangler,G. E., and Cohen,M.
J., "Instrument Design and
Description"in p. 15, Plasma
Chromatography, Ed. Timothy W. Carr,
Plenum Press, New York, 1984,1-42.
17. Good,A. I., Durden,D. A., and
Kebarle,P., "Ion-molecule reactions
in pure nitrogen and nitrogen
containing traces of water at total
pressures 0.5-4 torr. Kinetics of
clustering reactions forming
H+(H20) ", J. Chem. Phys., 52,
(1970), 212.
18. Spangler,G. E., and Collins,C.
I., "Reactant Ions in Negative Ion
Plasma Chromatography", Anal. Chem.
43, (March 1975), 2.
19. Threshold Limit Values and
Biological Exposure Indices for
1988-1989,AmericanConferenceof
Governmental Industrial Hygienists
20. NIOSH Pocket Guide to Chemical
Hazards, U.S. Department' of Health and
Human Services, 1985.
142
-------
FIGURE 1
IMS/MS
ATMOSPHERIC PRESSURE (760 TORR)
MOBILITY REGION FOR 0 NO. 1
lOc
-------
TYPICAL [ION ARRIVAL TIME SPECTRUM]
T 2x10-«A
MILLISECONDS
FIGURE 2
144
-------
in
4J
•H
c
31
L
01
S-
4J
•H
n
-------
Figure 4: AVM Spectrum (Acetic Acid 10 pom)
in
4-»
•H
c
DJ
L
to
4-«
•H
J3
L
-------
Figure 5:AVM Spectrum (Acetic Acid 15 ppm)
m H+(H2O)x(N2)y Acid Monomer
in
4J
•H
D
31
ID
-4J
•H
JQ
-------
Figure €:AVM Spectrum ( Acetic flcic? I
000 ppm)
m
•n-m-(H20)x(N2)y Acid Anhydride
01
4J
•H
c
\
D m2 H+(H20)x(N2)y Acid Dimer
3)
L
01
L
•H H
JQ
L.
-------
c:
a
UJ
c_>
-
•t.13 IO.3S,
ST) 22. 6B 2B.81 3-t . 3S •* 1 12
Tine tnsi
Figure 7:IMS/MS Spectrum "Total Ion Mode" ( Acetic Acid 80 ppb)
149
-------
G
c?
or
r~
r- —
r— .
'Of-
UJ
1/.M
10.35 16.SO 22.BS 28.61 3-j . 36 VI.12
TIME IMSI
Figure 8:IMS/MS Spectrum "Integral Ion Mode"
(Acetic Acid 80 ppb)
150
-------
p
H3C-C-OH
a
a
cu
r-
•2.
ru
97
73
55
79 83
101
125
i
129
153
10.0 'jQ.D
7O.D 1DO.D 13O.O 1 GO.0 13O.O
ness
Figure 9:IMS/MS Spectrum "mass spectrum mode"
(Acetic Acid 80 ppb)
151
-------
m/e 153
m/e 125
m/e 97
m/e 79
Total Ion
ID.35 IB.SO 22.BS 26.Bl 3*1.36 'r 1 12
TIME (MSI
Figure 10:IMS/MS Spectra "Tuned Ion Mode" (Acetic Acid 80 ppb)
152
-------
HAND-HELD GC-ION MOBILITY SPECTROMETRY FOR ON-SITE ANALYSIS
OP COMPLEX ORGANIC MIXTURES IN AIR OR VAPORS OVER WASTE
SITES
Suzanne Ehart Bell
Los Alamos National
Laboratory
MS K484
Los Alamos, NM 87545
G.A. Eiceman
New Mexico State University
Department of Chemistry
Box 30001, Dept. 3C
Las Cruces, NM 88003
ABSTRACT
Ion mobility
spectrometry (IMS) was
formally introduced
approximately 21 years
ago, and has been used as
a detector for chemical
warfare agents. IMS
research and development
outside the military has
recently been the subject
of renewed interest.
Military IMS units are
small, rugged, and
portable which makes them
ideal candidates for
inclusion in portable
airborne vapor monitoring
systems. The strengths of
IMS are low detection
limits, a wide range of
application, and
simplicity of design and
operation. The gentle
ionization processes used
in IMS impart a measure of
selectivity to its
response. However,
atmospheric pressure
chemical ionization with
compounds of comparable
proton affinities leads to
mobility spectra for which
interpretive and
predictive models do not
exist. An alternative
approach for the analysis
of complex mixtures with
IMS is the use of a
separation device such as
a gas chromatograph (GC)
as an inlet. The
attractions of GC-IMS over
GC-mass spectrometry (MS)
for field use include the
small size, low weight,
and low power demands of
GC-IMS.
Parameters in GC-IMS
which required examination
before further development
or field application
included three major
concerns. The first was
selection of an optimum
temperature of the IMS
detector and evaluation of
the effect of IMS
temperature on mobility
spectra. The second was a
study of the stability
and reproducibility of
chromatographic retention
and mobility behavior.
The final issue was the
153
-------
development of suitable
data reduction methods.
Results suggest that an
IMS cell temperature of
ca. 150° to 175°C provided
mobility spectra with
suitable spectral detail
without the complications
of ion-molecule clusters
or fragmentation. A
commercially available,
portable IMS unit was
configured as a GC
detector to evaluate the
possibility of using the
unmodified unit as the
basis for a portable
prototype. Significant
fluctuation in peak
heights were observed (ca.
+/- 12%), but mobilities
varied slightly ( ca. 1 %)
over a 30 day test period.
Neural network pattern
identification techniques
were applied to data
obtained at room
temperature and at 150°C.
Results showed that
spectral variability
within compound classes
was insufficient to
distinguish related
compounds when mobility
data was obtained using
the commercial room
temperature IMS cell.
Similar but less severe
difficulty was encountered
using the 150°C data.
Incorporation of retention
indices as a referee
parameter was useful in
eliminating false
positives.
INTRODUCTION
Background
The detection of
trace levels of hazardous
organic volatile compounds
in complex mixtures
represents an analytical
and sampling challenge.
Waste site sampling
requires ppb detection
limits in samples
comprised of complex
matrices and mixtures of
from ten to hundreds of
analytes. Other
considerations include the
time of sampling and time
of analyses, delays in
analysis, labor costs,
labor training, and
cost/sample ratio. The
time and expense of
complete laboratory
analyses can force that
fewer samples be taken
with the attending risks.
Technical aspects make the
translation of widely
accepted laboratory
instrumentation (GC-MS and
GC-FTIR) difficult or
unsatisfactory due to cost
and complexity.
Certainly, gas
chromatography with some
advanced detector will be
required for chemical
resolution of complex
mixtures of organic
compounds over waste
sites. Proven detectors
such as mass spectrometry
and infrared spectrometry
allow necessary
specificity of detection
but represent cumbersome
and intricate
instrumentation not easily
configured for field use.
These instruments often
require highly skilled
operators as well. The
high power consumption of
portable GC/MS and GC/IR
systems certainly limits
their use in many field
situations. Other
detectors which have been
154
-------
common to portable GC
units lack specificity and
necessitate a reversion to
dual column or dual
detector methods for
confirmation of peak
assignments. The
development of a hand-held
GC-IMS combines the
separation power of GC in
combination with a
multidimensional detector.
The release of the
civilian counterpart of
the military IMS units was
a logical starting point
for development of a
portable GC-IMS.
Ion Mobility Spectrometry
Ion mobility
spectrometry (Figure 1) is
based on the ionization of
vapors in air at
atmospheric pressure. The
differentiation of ions
occurs by measurement of
gaseous ionic mobilities
(1). A typical IMS
instrument is divided into
two regions. The first is
the reaction region
containing an ion source
(typically 63Ni). Ion
separation occurs in the
second (drift) region of
the spectrometer, where
separation is based on the
size-to-charge ratio of
the ions. The ion shutter
that separates the two
regions injects ions from
the reaction to the drift
region using period pulses
of the shutter field. The
drifting ions are detected
at the end of the drift
tube by a detector plate.
In IMS, ionization
occurs through collisional
charge transfer between a
reservoir of charge, i.e.
the reactant ions, and
neutral analytes, M. The
most abundant reactant
ions generated from a
beta-emitting source in
air are (H2O)n*H+ and
(H2°)n*°2~- These ionic
clusters co-exist at near
thermal energies in the
reaction region. Product
ions experience little or
no fragmentation and exist
commonly as M+ and MH+ or
M~ and M*C>2~ depending on
proton or electron
affinities of the neutral
species. Ions formed in
the reaction region are
injected into the drift
region by the ion shutter.
In the drift region, ions
move at particular drift
times (td) through an
electric field, E, of ca.
200 V/cm. For a drift
region with a given
length, L (cm) , the drift
time is related to
velocity (vd, cm/s) and
ion mobility (K, cm /V*s) )
through equations 1 and 2 :
d =
(1)
/ td
(2)
= vd /E
K
Ions strike a flat plate
detector and a mobility
spectrum or plot of
detector current (in pA or
nA) versus td (usually in
ms) is produced.
Consequently, the basis
for selectivity in IMS is
differences in drift times
for ions governed by ion
mobilities. Drift times
are dependent on
temperature and pressure
and are normalized to
reduced mobility
constants, K0, that are
related to molecular
155
-------
properties through the
Mason-Shamp equation. In
general, the equations for
mobility constants are
considered well-
established for small
spherical ions but
extrapolations to large
organic molecules may be
tenuous. Practically
speaking, direct
quantitative predictions
of Ko values for organic
molecules are presently
impossible. Mobilities
are inversely proportional
to collisional cross
sections. Thus, IMS is an
ion separator based on
size/charge rather than
mass/charge as found in
mass spectrometers.
Ion mobility
spectrometry offers
advantages such as low
power, simple and rugged
construction, ppb
detection limits, and
mobility spectra
representative of
individual constituents.
Disadvantages
traditionally ascribed to
IMS include significant
memory effects,
irreproducible behavior
and complex response to
mixtures (2). These
difficulties can be
circumvented with the
addition of a GC as an
inlet and with the
reconfiguration of the
drift tube (3,4).
Furthermore, hand-held IMS
instruments are currently
available in military-
hardened form with battery
operation (5). The
military IMS cells are
attractive for use in
portable GC units and were
used as a starting point
for the study of GC/IMS
parameters.
Objectives
Several areas of GC-
IMS have not been
addressed and must be
understood for practical
advances in field
applications of GC/IMS.
The first area is
optimization (or
influence) of IMS
temperature on GC/IMS
performance and on the
mobility spectra obtained
from the IMS. Second is
the evaluation of the
effect of concentration on
reduced mobility and
mobility patterns. Third
is the evaluation of a
commercially available
portable IMS as a GC
detector, and the final
area is the preparation of
a suitable software peak
identification program.
Each of these has served
as the basis for an
objective in the work
described below.
RESULTS AND DISCUSSION
Effects of Temperature on
Ion Mobility
The successful
development of a portable
GC-IMS requires that the
optimum IMS temperature be
determined. This data had
to be determined
empirically, since little
foundational theory was
available. Typically, low
temperature mobility
behavior shows
considerable ion
clustering and complexity,
while higher temperatures
encourage ion
156
-------
fragmentations. An
intensive study was
undertaken to determine
the optimum operating
temperature for the IMS
since a wide variety of
analytes are expected to
be encountered. A
representative set of 43
compounds was selected
from seven different
chemical classes, shown in
Table 1. The temperature
effect study was conducted
on a Tandem Ion Mobility
Spectrometer (TIMS, PCP
Inc., West Palm Beach,
Fla.) which allowed
heating of the inlet and
drift tube.
Confirmational mass
spectral studies were
conducted on an MMS-160
IMS/MS (PCP, Inc., West
Palm Beach, Fla).
There are four basic
processes that can occur
when a compound is
introduced into the IMS.
First, there may be no
detectable reaction, such
as when a species that is
active only under positive
polarity is introduced
into an IMS operating in
negative polarity.
Second, clusters may form
between the analyte and
various ions such as
N2+, or NH4+. Such
clusters appear as peaks
in the spectrum. The
third possibility is the
formation of cluster ions
which subsequently undergo
equilibria reactions while
in the drift tube. The
magnitude of the
equilibrium constant will
determine the effect on
the resulting mobility
spectrum. If the
equilibrium is slow
relative to transit time,
no significant effects
will be seen. If the
equilibrium is fast
relative to the transit
time, the ions arriving at
the detector can differ
significantly from the
original ions produced,
and peak broadening may
result. Finally,
fragmentation may occur,
and the resulting spectra
may exhibit such behaviors
as a generalized increase
in the baseline or a
series or numerous small
peaks. The exact
manifestation will depend
on the degree of
fragmentation. The IMS
portion of a portable
GC/IMS should operate
isothermally to reduce
power consumption and
complexity. It is thus
essential to select the
cell temperature such that
clearly resolved, sharp,
and reproducible peaks are
produced. Peak broadening
and fragmentation patterns
will be difficult, if not
impossible, for a data
reduction system to
classify. It is also
desirable that the cell
operating temperature be
as low as possible to
minimize power
requirements. The other
factor that must be
considered for temperature
selection is memory
effect. Higher
temperatures encourage
rapid clearing of the cell
and promote cleaner
operation. Thus, 3
factors must be balanced
in selecting the optimum
IMS temperature: clearing
time, mobility behavior,
and power requirements.
157
-------
The effect of IMS
cell temperature on
mobility behavior was
studied by analyzing the
43 target compounds using
nine different cell
temperatures from 50 to
250°C. The results showed
that while all compounds
behaved differently, a
general pattern was
discernable. At the lower
temperatures (ca. 50 to
150°C), many compounds
experienced drift tube
reactions, and peaks were
either very broad or moved
as the concentration in
the drift tube changed.
At the midrange
temperatures (ca. 100-
200°C), drift tube
equilibria decreased, and
stable ion/molecule
clusters were observed.
At the higher temperatures
(ca. 200-250°C),
fragmentation became
prevalent. Figures 2 and
3 show two examples of
compound classes and their
behavior over the
temperature range studied.
The aromatics (figure 3)
are not dramatically
affected by temperature
changes, although benzene
and ethylbenzene do show
evidence of drift tube
reactions at 75 through
150°C. The alcohols
(figure 4) show greater
variability with
temperature than the
aromatics, but the general
pattern of drift tube
reactions-clustering-
fragmentation is evident
in the ethanol and n-
propanol.
Members of the
chemical classes of
ketones, alcohols,
halocarbons, and esters
were examined by IMS/MS at
three temperatures to
confirm the data obtained
using the TIMS. At 50°C,
ion cluster formation
dominated mobility spectra
and the formation of dimer
and solvated ions was
evident. At elevated
temperatures (150° and
225°C), these ions were
not observed or present at
low levels. At 225°C,
fragmentation was
prevalent rendering
mobility spectra less
informative than those
from lower temperatures.
Compilation of the
TIMS and IMS/MS data leads
to several observations
cogent to the design of a
hand-held GC/IMS. First,
a portable GC/IMS will
require the use of a
heated IMS cell to obtain
distinctive and
informative mobility
spectra. If the
instrument is to be used
as a monitor for a wide
range of compounds, the
optimum temperature range
appears to be 150-200°C.
Second, the cell
temperature can be set to
optimize the response of
selected compound classes.
For example, the
halocarbons showed greater
spectral detail at higher
temperatures than did the
rest of the target
compounds. If the GC/IMS
is to be used as an in-
situ monitor for
halocarbons, the IMS cell
temperature could be set
at 225°C. Finally, the
variations in behaviors
with temperature might be
useful as an added
discriminator in GC/IMS
applications. For
158
-------
example, acetone and
isopropanol have similar
chromatographic retention
indices on many GC
columns. At lower IMS
cell temperatures,
isopropanol and acetone
both exhibit drift tube
equilibrium reactions, and
their spectra have many
similar features that
might confuse pattern
recognition software. At
175°, the spectrum of
isopropanol begins to show
distinct stable peaks,
while acetone still shows
drift tube reactions up to
ca. 225°. Thus, the
selection of cell
temperature could be used
to help discriminate
between these two
compounds.
Stability and
Reproducibility of IMS
Graseby Analytical
(United Kingdom),
manufactures a portable
IMS that is used by
western military
establishments for
detection of chemical
warfare agents. This IMS
(abbreviated as AVM for
airborne vapor monitor)
was coupled to a GC to
evaluate three parameters.
The GC used was a Hewlett-
Packard (Palo Alto, CA)
5730 equipped with a
Supelco (Supelco Park, PA)
SPB-5 30 meter capillary
column. Nitrogen was used
as the carrier gas, and
makeup gas was air. The
AVM operated in a water
chemistry mode. The
effect of concentration on
mobility behavior was
examined first to
determine if IMS mobility
patterns were
significantly influenced
by analyte concentration.
The stability and
reproducibility of the IMS
response over an extended
period was evaluated as
well. These findings were
then used to determine if
it would be practical to
use an essentially
unaltered AVM as the IMS
cell for a portable
prototype GC-IMS. These
findings were also used to
isolate and identify those
features of the AVM that
could be modified to
improve its performance as
a GC detector.
' The effect of
concentration on mobility
was studied, by injecting a
series of dilutions of
each of the target
compounds into the GC-AVM.
Review of the data
obtained led to several
unanticipated findings.
First, the AVM spectra of
many of the positive mode
compounds were very
similar. The data
obtained at 50°C using the
TIMS did not show these
similarities. As the
concentration of the
target analyte decreased,
the similarities between
the spectra generally
increased. Product ions
were often shoulders off
the reactant ion peak as
opposed to the separate
product peaks usually
observed using the TIMS.
Finally, a clear linear
relationship between peak
height and concentration
was not obtained over the
concentration range
studied. As a result, no
definitive statement
159
-------
regarding the effect of
concentration on mobility
was possible.
The reproducibility
of AVM was evaluated over
a 1 month period. Peak
heights, drift times, and
mobilities were monitored
for positive and negative
background spectra. The
spectra of known amounts
of positive and negative
mode standard compounds
(ethylbenzene and CCl^,
respectively) were also
examined. The results of
the study are shown in
Table 2. The variability
of intensity of the
reactant and product ions
showed drift over the 30
days, but reduced
mobilities varied
slightly. Any attempt at
quantitation using only
mobility spectra patterns
and relative abundances
would be difficult using
the AVM as configured.
Table 2 also shows that
the larger ions exhibit
more reproducible
behavior, as shown by the
decrease in relative
standard deviations with
decreases in mobility.
This fact was exploited in
neural network pattern
identification studies
which followed.
Evaluation of Neural
Networks for
Identification of
Compounds
Neural networks have
in the last 10 years
become very popular for
pattern recognition in
many disciplines. A
network consists of a
series of interconnected
nodes (called neurons or
perceptrons) in which
mathematical weighting,
summation, and submission
to a function are
performed. The output of
each neuron is-then sent
on to another neuron where
a similar operation takes
place. The network itself
can consist of a variable
number of neurons in a
layer, and variable
numbers of layers. The
network is trained by
submitting to it target
vectors consisting of
input and the target
output desired. In this
work, the factors included
in the training vector
were retention index and
mobility peak data. The
target output was the name
of the compound possessing
these GC-IMS
characteristics. The
network takes each
training vector and
adjusts the weights
applied in each neuron to
get the correct value
output. The next training
vector is submitted using
the previously obtained
weighting factors, and the
resultant error is used to
adjust the weights again.
This repetitive process
continues until the
weights are adjusted so
each training vector
submitted to the network
yields the correct output.
Training sets may consist
of hundreds of facts, and
the training process
itself may take hours.
Once the network is
trained, however, response
is rapid. For this
reason, neural networks
are well suited for use
in a portable instrument.
160
-------
For this study,
neural networks were used
with both the TIMS data
(150°C) and the AVM data.
The training vectors
consisted of retention
indices, reduced
mobilities, and in some
cases, the percent
relative abundance of the
mobility peaks. Aspects
of network structure,
training, and failures
were examined with both
data sets. The network
was unable to train on the
AVM data for the alcohols.
Many of the alcohol
spectra were very similar,
and the network was unable
to distinguish between
them even with the
retention index included.
The network was able to
train successfully using
the TIMS alcohol data.
The difficulty with the
AVM data may arise from
operating the cell at
ambient temperature and
from using a membrane in
the inlet.
A network was trained
using data from all the
positive mode compounds
obtained at 150°C.
Approximately 10% of the
initial test data was set
aside as a test set. The
network was trained using
the remaining 90% of the
original data set. The
trained network was able
to identify ca. 95% of the
test set. Failures were
associated with similar
compounds, i.e., within
compound classes. A
typical problem was
differentiating
ethylbenzene from the
xylenes. This problem was
successfully addressed by
using the retention index
of the test compound to
determine the correct
identification. For
example, if the network
yielded both ethylbenzene
and o-xylene as potential
identifications, the
retention index of the
test compound was compared
to the retention index of
the standard target
compounds. In all cases
of multiple
identifications, this
approach eliminated the
false positives. In no
instances were false
identifications seen
across compound classes,
i.e., never was a ketone
mistakenly identified as
an alcohol when the
retention index cr iteria
was used.
CONCLUSIONS
The findings
demonstrate that GC-IMS is
a viable field monitoring
technique, and holds
promise of evolving into a
genuinely portable and
powerful field screening
device. Elevated
temperature cells,
operating without
membranes, will be
required for such devices.
Commercial portable IMS
units such as the AVM
cannot, as currently
configured, be used as
detectors for GC-IMS.
While these devices work
well for specialized
applications, use of the
AVM as a generalized
detector is not possible
without modifications.
Neural networks can be
successfully used to
identify compounds when
161
-------
chromatographic data is
included in the training
process and mobility data
obtained at elevated
temperatures is used.
When the pattern
recognition process fails
to identify a compound,
retention index can be
used to obtain the correct
identification. Neural
networks are system
specific. The network can
not be trained using data
obtained on different GC-
IMS system. Aspects of
the chromatographic and
mobility behavior (via
temperature) can be
modified to suit specific
applications or can be set
to cover a broad range of
target compounds. The
small size and low power
requirements of GC-IMS
combined with the ability
to tune the instruments to
different applications
gives GC-IMS an advantage
over many other portable
techniques.
REFERENCES
1. G.A. Eiceman, Critical
Reviews in Analytical
Chemistry 1990, in press.
2. M.M. Metro and R.A.
Keller, J. Chrom. Sci.
1973, 11, 520.
3. H.H. Hill, Jr.,
Critical Reviews in
Analytical Chemistry 1990,
21,
4. C.S. Leasure, V.J.
Vandiver, G. Rico, and
G.A. Eiceman, Analytica
Chimica Acta 1985, 175,
135.
5. D.A. Blyth, "A Vapour
Monitor for Detection and
Contamination Control",
Proc. Internl. Symp.
Protection Against
Chemical Warfare Agents,
Stockholm, Sweden June 17-
19, 1983, pp. 65-69. b)
Commercial brochures from
Graseby Ionics, Ltd. and
Graseby Analytical, Ltd.,
Watford, Herts., UK.
ACKNOWLEDGEMENTS
Financial support to
NMSU by KRUG Life Sciences
for NASA through project
no. 50,016 is gratefully
acknowledged as is
financial and professional
assistance from Los Alamos
National Laboratory to
Suzanne Bell.
162
-------
Table I. Listing of analytes studied using GC-IMS.
Positive Mode
ALCOHOLS
Methanol
Ethanol
n-Propanol
i-Propanol
n-Butanol
i-Butanol
s-Butanol
t-Butanol
AROMATICS
Benzene
Toluene
Ethylbenzene
o-Xylene
m-Xylene
p-Xylene
Styrene
ESTERS
Methyl Methanoate
Methyl Ethanoate
Methyl Propanoate
Methyl Butanoate
Methyl Pentanoate
Ethyl Methanoate
Ethyl Ethanoate
KETONES
Acetone
2-Butanone
3-Metyl-2-Butanone
2--Pentanone
3-?entanone
ALDEHYDES
Propanal
Butanal
3-MethyIbutana1
Pentanal
Hexanal
Negative Mode
HALOCARBONS
Methylene Chloride
Chloroform
Carbon Tetrachloride
Trichloroethene
1,1,1-Trichloroethane
Tetrachloroethene
1,2-Dichloroethane
1,1,2,2-Tetrachloroethane
CHLORINATED AROMATICS
Chlorobenzene
o-Dichlorobenzene
2-Chlorotoluene
163
-------
Table 2
AVM Reproducibility Study
Description Mean* Rel. Std. Dev. (%)
Reactant Ions
Peak Height
Positive Mode
Negative Mode
Reduced Mobility
Positive Mode
Negative Mode
Product Ions
Peak Height
Positive Mode
Positive Mode
Negative Mode
Reduced Mobility
Positive Mode
Positive Mode
Negative Mode
6911
2109
1.87
1.60
935
679
1687
1.64
1.39
2.22
11.2
22.2
2.01
2.18
8.65
8.22
8.77
1.19
0.98
0.99
*: Mobilities reported in cm2 V -1 s ^ and peak heights reported
in millivolts.
**
: Ethylbenzene had 2 product ions.
164
-------
Ion Mobility Spectrometer
Reaction Region Drift Region Vent
Drif
Car
Ni63
Gas
ier Gas
Repeller
Shutter
Detector
Figure 1. Schematic of ion mobility spectrometer
Acetone 2-Butanone
KO
50 100 '50 200 250 50 100 <«> 200 250
Temper aluie
3-Methyl-2-Butanone
Figure 2. Behavior of selected ketones over the 9
temperatures studied. Legend for Figures 2 and 3: P:
that moved over the course of the elution. The P marks the
extremes of the mobility. X: Distinct stable peak.
Extremes of a drift tube reaction broadened peak.
Approximate center of the peak associated with a dri
reaction.
165
-------
Toluene
SO 100 '50 200 250 50 100 150 200
Temperature
200
Z50
1.75
Ethylbenzene
Sfyrene
Figure 3. Behavior of selected aromatics over the 9
temperatures studied. See figure 2 for key.
DISCUSSION
COLLEEN PETULLO: Did you use the same IMS in the IMS-MS study or
were several used?
SUZANNE BELL: The IMS-MS instrument was different than the heated
instrument we used in New Mexico State. That's simply because we didn't have
an IMS-MS available, so we simply used one that PCP was gracious to rent us
for a week.
COLLEEN PETULLO: But you only used one in the study at any given time,
right?
SUZANNE BELL: Right. The nine temperatures and 43 compounds were all
run on one instrument. The IMS-MS was on another instrument, and then the GC/
IMS was yet another instrument.
COLLEEN PETULLO: How long would it have taken you to train the neural
networks if you would have programmed it for the 43 compounds?
SUZANNE BELL: I would assume it would take eight to ten hours, at the worst.
The training time gets longer as you get more and more similar data. If we gave
it. for example, 25 examples of benzene spectra over a wide concentration range,
that would let the network generalize but you pay the price in training lime. It
could take hours or weeks to train the computer.
COLLEEN PETULLO: You had mentioned that you didn't do this because of
time constraints.
SUZANNE BELL: Right.
COLLEEN PETULLO: How many did you ultimately program?
SUZANNE BELL: We ultimately trained 23 in the combined data set. This was
about half.
166
-------
Remote and In Situ Sensing of Hazardous Materials
by Infared Laser Absorption, Ion Mobility
Spectrometry and Fluorescence
Dr. Peter Richter
The Institute of Physics, Technical University of Budapest
1111. Budapest, Budafoki u't 8. HUNGARY
ABSTRACT
Three instruments will be described that were
developed at the Technical University of Budapest
for the sensing of hazardous materials. A remote
sensing infrared differential absorption lidar
based on the coherent detection of backscattered
CO2 laser light has been built. The lidar can be
used for the detection of a wide range of molecular
pollutants in the atmosphere from ranges of a few
kilometers along a path to a topographic target.
Results of field measurements to detect molecular
pollutant clouds from km ranges will be presented.
The experiments were carried out on NHs and
DDVP but detection of more than 80 air-polluting
components such as Freons, SC>2, etc. is also
potentially possible. In addition, an ion mobility
spectrometer will be discussed which has been
developed for in situ measurements of impurities
in air. The impurities are identified with the help
of a dynamic dual-grid cell. Upon evaluation of
the frequency-ion current spectrum, the detection
of several impurities (e.g., NHs, DDVP, HF etc.) was
demonstrated. The instrument can operate either
in a stand alone or a remote controlled mode and
can be connected to a central computer. A
fluorescence detector for the detection of surface
contamination will also be discussed. Based on
chemical indicator reactions, UV excitation and
fluorescence detection via fiber optics, a mobile
instrument for detection of pesticide
contamination and control of decontamination has
been built. Reliability detection of concentrations
of 0.1 mg/cm2 for DDVP was achieved with a
measurement time of less than 5 sec. Applications
of the instruments and methods will also be
discussed.
INTRODUCTION
Sensing hazardous materials is a task that should
be approached using techniques that are
appropriate not only for the materials to be
detected but also for the measurements required.
A variety of sensing techniques are available to
accomplish this end. In this presentation, three
different methods and instruments that have been
developed at the Technical University of Budapest
will be described. As will be noted, these
instruments are applicable for different specific
purposes. The sensing techniques that will be
discussed are as follows:
A remote sensing lidar to measure pollutant
clouds in the atmosphere from km ranges;
An ion mobility spectrometer for in situ
measurement of air samples; and
An UV fluorescence detector to measure
surface contamination without direct
surface contact.
REMOTE SENSING LIDAR
Lidars are laser radars sensing backscattered laser
light from long ranges making use of the special
characteristics of laser light. Differential
absorption lidars measure light intensities at two
wavelengths corresponding to absorption maxima
and minima of the absorbing atmospheric
component along the beam path. Due to their
broad tunability range in the infrared region
around the 10 u.m wavelength where several
molecular pollutants have characteristic
absorption spectra, systems based on CO 2 laser
sources are of major importance [1]. In the group
of more than 80 detectable pollutants, some of the
more important ones are: NH3, C2H4, 03, SC>2, SFg,
C2H3C1, as well as pesticides such as DDVP (2,2
dichlorovinyl dimethyl phosphate).
167
-------
Two major problems associated with this technique
were eliminating the disturbances due to the open
path and keeping the system compact and
transportable. These problems were solved by the
development of the system, the optical part of
which is shown schematically in Fig. 1. Electronic
separation of the signals at the two wavelengths
allow the measurements to be simultaneous and
coincident, thus avoiding, for example, problems
due to turbulence and differential backscattering.
Using the internal amplification of the
backscattered light by the lasers and heterodyne
detection, make application of small CW lasers and
a transmit-receive telescope of diameter 15 cm
only possible. Topographic backscattering makes
long path absorption measurement possible. The
system used in the field tests is shown in Fig. 2 and
the results of a field test using stationary
topographic backscattering from 500m range with
an artificial cloud of NH3 is shown in Fig. 3. It is
the time dependence of the differential absorption
signal
K>.2.t)
E(t) =ln
that is displayed where ICXi^.t) are the
normalized detected light intensities at the two
wavelengths at time t. The column content along
the beam path cL (molecular concentration c times
the path length L) is given by
where A (T is the absorption cross section
difference of the molecule for the two
wavelengths.
The temporal variations of E in Fig. 3 are due to
the concentration changes in the cloud blown
across the beam path. The time resolution is 1 sec.
Due to the atmospheric window around A, = 1 0 |4.m ,
the reference range of the system is about 3 km
(material dependent) and is not significantly
influenced by the visibility conditions.
The measurement wavelengths and sensitivities
for some specific molecules are as follows:
(cL)min
(ppb)(km) (mg/m3)(km)
NH3 10.33 10.32
C2JU 10.53 10.59
O3 9.49 9.59
SO2 9.02 9.02
SF6 10.51 10.50
C2H3C1 10.61 10.50
8
8
22
710
1.5
34
8.6 x 10 '3
9 x 10 '3
4.2 x 10 '2
1.7 x 10 '2
9 x 10 '3
8.4 x 10 ~2
This system can be used in a stationary mode when
with a scanning attachment it can monitor either
large area (~ 30 km2) pollution distribution
(immission), or emission from certain selected
sources. When coverage of a larger area is
necessary, it can be used from a flying platform as
well.
ION MOBILITY SPECTROMETRY _
A simple and cost-effective technique for the in
situ detection of air pollutants is through the use
of ion mobility spectrometry. Here the air sample
is ionized by a radioactive source in a chamber and
the ions produced are moved by the use of an
electric field. The arrival time and current of the
ions characterize the products and their
concentration. However, as the predominant
charge carriers in the chamber are ion clusters
consisting of fragments of water, Nitrogen as well
as the molecule to be detected (e.g., NH3, HF,
CH3COH3, C2H5OC2H5, HCN, different pesticides), the
selectivity of the system requires the application
of sophisticated hardware and software solutions.
12].
The structure of the chamber is shown in Fig. 4.
Ambient air is drawn in across a semi-permeable
membrane allowing a portion of its component
gases and vapors to be introduced to relatively dry
air in the ionizing region. An alternating voltage
with frequencies sweeping from 0-30 kHz is
connected to a dual grid of transversal Venetian
blind type in front of the collecting electrode.
Recombination on the grid is dependent on the
mobility of the ions; therefore, evaluation of the
ion current as a function of grid frequency
improves the selectivity of the system. In Fig. 5.
ion currents are shown as a function of grid
frequency for clean air and air with NH3
Automatic evaluation of these curves are carried
out by a microprocessor taking derivatives of the
ion current curve at five characteristic
frequencies that correspond to f=OHz, f(Imin).
I(f2=OV) , f (jf = o) and f = fmax = 30 kHz.
With the help of an algorithm, these values are
compared with sets of stored data that had been
determined empirically.
Many materials can be monitored in the low ppm
region. The system shown in Fig. 6 can be used in
a network through a RS232 line that is also
supported by its low mass and power consumption
(2kg, 1W).
SURFACE CONTAMINATION DETECTOR _
Determining the contamination of surfaces of
ground areas as well as equipment and personnel
and the verification of the effectiveness of
decontamination from hazardous materials are
important considerations in assessing the extent of
f l =
168
-------
residual chemical activity such as in the
application of pesticides or in setting clean up
goals for site remediation.
With the technique described here, the
monitoring is based on the fluorescence analysis
of chemical compounds produced in a reaction
where a non-fluorescent compound, indole in an
alkaline peroxidase solution is oxidized by the
agent to be detected to give highly fluorescence
indoxyl [3].
To detect trace impurities, fluorescence techniques
show an inherent advantage compared to methods
based on absorption. Namely while the extinction
shows a logarithmic dependence on light intensity
given by
E = a c L = In
Io
Io -
the fluorescent light intensity F exhibits
an approximately linear relationship given by
F= QF Ia = QF Io (1 - e'ccL) = QF Io o c L ,
where IQ is the incident light intensity, Ia the
absorbed light intensity, and Q F is the quantum
efficiency of the fluorescence. Therefore, with
fluorescence the sensitivity can be improved by
increasing the exciting light intensity IQ. Also
surface contamination often appears in thin,
sometimes discontinuous layers or droplets where
the additional selectivity provided by the
wavelength discrimination of the fluorescent
light from the backscattered light can be
exploited.
In the chemical reaction described above, the
material to be detected plays the role of the
catalyst; therefore, the quantity of the fluorescent
material can be controlled to a certain extent by
the amount of reagent added.
The advantages of this method compared with
those requiring probe sampling are, that this
method operates without physical contact, is not
influenced by the surface type, and is highly
selective. The application of this method consists
of the following steps:
• spraying the contaminated area,
• illuminating it with UV light, and
• detection of the frequency shifted
fluorescent light and evaluating
the detector signal.
This system (shown in Fig. 7) consists of the
following three units:
• a spray unit to store and pump the
chemical reagents;
• an optoelectronic unit housing the
Mercury vapor light source, the
photomultiplier detector, the
spectral filters matched to the
compound to be detected and the
electronics using lock in detection;
and
• a sensor head unit (containing
optical elements and controls)
connected with 3 m long hoses,
cables and fiber optic bundles to
the other units.
Experiments carried out with DDVP and a reagent
containing NaBOs and indol in water solution
showed that the response time was less than 5 sec
after spraying and the detection limit was at 100
fig/cm^. Time duration of the fluorescence can be
adjusted by proper selection of concentration of
the reagents. This system can be used either in a
stationary mode or on a moving vehicle to monitor
large ground surfaces.
REFERENCES
1. Richter P., Proc. SPIE 883, 162 (1988)
2. Brokenshire J., Pay N., International
Laboratory, Oct. 1989.
3. Diehl W., Proc. 2nd Int. Symp. Protection
Chemical Agents, Stockholm, Sweden, 173
(1986)
169
-------
Telescope
m
Chopper
\
Lasers
b.s.
V
Attenuator
rA A . ,
Hm
—' V Detector
Lens
m
Figure 1. Arrangement of the differential absorption lidar.
Figure 2. The lidar system.
170
-------
30 60 90 120 150 180 210 240 270 t
1/DIV START:! SEC
AVO5 1989.09.06. 15.23 "E"
Figure 3. Time evolution of differential absorption signal
for an artificial NHa cloud.
insulator rings
metal housing
source
electrode
dual grid
collector
electrode
Figure 4. Structure of the ion mobility spectrometer chamber.
171
-------
l[pA]
80 _
60 -
40 -
20 -
air+NHg
/,
I
100 Hz 300 Hz 1kHz 3kHz
I
I
-»- f
10kHz 30kHz
Figure 5. Dependence of ion current on grid frequency for
clean air and air with 0.2 ppm
Figure 6. The ion mobility
spectrometer sensor.
Figure 7. The surface contamination
fluorescence detector.
172
-------
THE DEPARTMENT OF ENERGY'S
ROBOTICS TECHNOLOGY DEVELOPMENT
PROGRAM FOR ENVIRONMENTAL RESTORATION
AND WASTE MANAGEMENT
A. C. Heywood
Science Applications International Corporation
Pleasanton, California 94566
S. A. Meacham
Oak Ridge National Laboratory
Oak Ridge, Tennessee 37831-6305
P. J.Eicker
Sandia National Laboratories
AlbiK^uerque, Hew Mexico 87185
In August 1989, the new Office
of Environmental Restoration and
Waste Management (ER&WM) in the
Department of Energy (DOE)
published an ER&WM Five-Year Plan
which established DOE's agenda and
commitment to correct existing
environmental problems, ensure
compliance with applicable Federal,
State, and local requirements, and
effectively execute DOE's waste
management programs. The plan
includes a section covering the
applied research and development
needed to support the five-year
plan. In November 1989, DOE Issued
a draft Applied Research.
Development, Demonstration.
Testing, and Evaluation (RDDT&E1
Plan for ER&WM which expands on the
applied research and development
section of the five-year plan. The
RDDT&E plan provides guidance to
the new ER&WM Office of Technology
Development (OTD) for its mission:
"to manage and direct programs and
activities to establish and
maintain an aggressive national
program for applied research and
development to resolve major
technical issues and rapidly
advance beyond current technologies
for environmental restoration and
waste management operations." The
development and application of
robotics technology for the
resolution of identified problem
areas at DOE sites is a major
element of the RDDT&E program plan.
The OTD has established a
Robotics Technology Development
Program (RTDP) to integrate
robotics RDDT&E activities and to
provide needs-oriented, timely, and
economical robotics technology to
support environmental and waste
operations activities at DOE sites.
DOE laboratories, private industry,
and universities have existing
robotics technology that provides a
strong foundation for initiating an
aggressive RDDT&E program to
support ongoing and emerging ER&WM
functions.
A major objective of the ER&WM
Program's five-year RTDP is the
application of robotic technology
in the resolution of DOE's
identified problem areas. The
thrust of the application is to
reduce exposure of personnel to
hazardous substances and radiation
while increasing productivity. An
additional goal is to integrate all
such activities to obtain the most
economical approach to resolving
site-related waste problems using
robotic technology and to
demonstrate robotic technologies
that can be applied to major site-
specific waste clean-up efforts.
The Robotics Five-Year Program
Plan provides the focus and
direction for the near-term (less
than five years) and guidance for
the lona term (five to twenty
173
-------
years) R&D efforts associated with
resolution of site-specific waste
problems. The goals include: (1)
supporting the ER6WM Program and
being responsive to the ER6WM Five-
Year Plan, (2) focusing near-term
robotic R&D efforts to be
responsive to application
requirements, (3) ensuring that
robotic applications are responsive
to site requirements and scheduler
needs, (4) integrating all robotic
activities to obtain the most
economical approach to resolving
site problems while reducing
personnel exposure, and (5)
providing guidance for the Office
of Energy Research long-range (>5
year) robots research program.
Program Focus and Objectives
The Program currently
addresses a number of important
issues facing the ER&WM activities
at the DOE sites. Among the areas
included are:
• underground storage tanks
(material characterization
and remedial actions),
• buried waste retrieval,
• waste minimization,
• contaminant analysis,
automation,
• decontamination and
decommissioning,
• basic and applied research
and development required to
support the above areas.
The objectives of the Program
are to develop, test, evaluate, and
make available robotic technologies
that:
• allow workers in waste
operations and remediation
to be removed from hazards,
• increase the speed and
productivity with which
ER&WM operations can be
carried out when compared
to alternative methods and
technologies,
• increase the safety of
ER&WM operations, and
• provide robotic and remote
systems technologies that
have lower life cycle costs
than other methods and
technologies.
In addition to developing
robotics technology, the program
promotes the availability of the
technology and supports its
deployment and use in ER&WM
activities at DOE sites. The
program further serves as a bridge
between the ER&WM robotics RDDT&E
and the basic robotics research
carried out by DOE's Office of
Energy Research, providing guidance
for the basic research program and
integrating its results in applied
research and advanced development
projects.
Program Organization
In order to execute the
Program, the Program has been
structured as shown in Fig. 1.
Since the Program is an element of
the DOE ER6WM Applied RDDT&E
program, it is administered by the
ER&WM OTD through the Program
Manager (RPM).
To ensure that the Program
responds to the needs of the DOE
complex, RPM is assisted by an
Operations Review Group (ORG).
This group is familiar with the
ER&WM issues facing the DOE
complex. RPM also receives
assistance from a Technical Review
Group (TRG) of robotics and
automation experts from the DOE
laboratories and sites,
universities, industry, and other
federal agencies. A Program and
Budget subcommittee of the TRG also
assists the RPM.
The Robotics Applications
Coordinators (RAC) develop robotics
program plans focused on each of
the major ER&WM issues.
The RAC is responsible for
coordinating the flow of technical
information relevant to the
applications area among those
groups having an interest in the
area. RAC is also responsible for
keeping the other groups in the
relevant applications areas
apprised of the results of RTDP
174
-------
r sir-
?: (v»«t)
Coordinator
ST~
7 (east*
^Coordinator
5?
Burled
waste*
Contaminant
/ Analyst*
Automation
•: Waste
Minimization
Coordinator
D«contam.
and
; Dftcomm.
Coordinator
Wasta
Faculties
t Operations
Coordinator
C
i
II
n
Ficurc 1
\TDP Oreamzanon
175
-------
funded activities. The
coordinator, with the approval of
the RPM also convenes occasional
conferences on the applications
area.
The coordinators function as
the advocate for the technologies
applicable to their particular
problem area. To facilitate the
application of the best technology
with a high probability of success
to the particular problem area, the
coordinator actively solicits
proposals from the entire robotics
and automation community for
routing to the RPM. A thorough
familiarity with the ERSWM problems
and issues is required of the
coordinators. This familiarity
will be maintained through site
visits, personal contacts, and
symposia where appropriate.
Applied research is funded
through the applications center
that has identified the
technological need. This helps
insure that the applied research is
responsive to the needs of the
group sponsoring the research.
Coordinators who put together a
team approach with industry, labs,
•universities, or other agencies are
most favorably reviewed.
The R&D Coordinator (RDC)
reports to the RPM and is
responsible for coordinating the
flow of technical information other
than applied research. The RPM is
familiar with all aspects of the
RTDP and is able to identify areas
of future need in robotics and
ancillary systems which are not
being addressed in the applied RSD
areas. He is responsible for
coordinating with universities,
industry, DOE laboratories, and
other federal agencies to bring
proposals for need advanced
technology to the TRG and RPM.
Program Planning
A comprehensive technical
program plan has been developed
during the first year of funding.
This initial plan development is a
significant effort since the plan
is based on the needs of the
environmental restoration and waste
management operations as identified
by the eight DOE field offices and
the sites they administer. A major
portion of the initial plan
development is assessing and
understanding those needs. The
technical program plan covers a
five-year period with primary
emphasis on the one-year plan and
secondary emphasis on the two- and
three-year projections. The plan
covers technical work, budget
requirements, and schedules and is
tied closely to the requirements
and schedules of individual site
environmental restoration and waste
management projects.
FY 1990 Accomplishments; The RTDP
accomplished a number of
significant activities in FY 1990,
which facilitated a fast start for
robotics technology development and
established a sound basis for
program activities over the next
five years.
Program Planning: Five priority
DOE sites were visited in March
1990 to identify needs for robotics
technology in environmental
restoration and waste management
operations. This 5-Year Program
Plan for the RTDP was prepared on
the basis of the needs identified
at the DOE sites, and provides a
needs-based road map for detailed
annual plans for robotics
technology development.
Initiating Interactions with the
Robotics Technoloov Community: In
July 1990, a forum was held
announcing the robotics program.
Over 60 organizations (industrial,
university and federal laboratory)
made presentations on their
robotics capabilities.
Technology Demonstrations: To
stimulate early interactions with
the ERSWM activities at DOE sites,
as well as with the robotics
community, the RTDP sponsored four
technology demonstrations related
to ERSWM needs. These
demonstrations integrate commercial
technology with robotics technology
developed by DOE in support of
areas such as nuclear reactor
maintenance and the civilian
reactor waste program.
176
-------
Rapid, swing-free movement of
simulated waste containers was
demonstrated using control
algorithms developed at Sandia
National Laboratories (SNL) with
technology in computer control of
large gantry bridges at Oak Ridge
National Laboratory (ORNL). This
technology decreases the time for
materials movement and increases
safety by eliminating the potential
for collisions of swinging
payloads.
A scaled waste tank
remediation demonstration at SNL
integrated sensors and advanced
computer control into a commercial
gantry robot. The extensive use of
models for robot system control
allowed graphical programming of
the system complete with operator-
supervised path planning to
increase speed of repetitive waste
removal tasks.
A teleoperated vehicle with
advanced sensing technologies for
mapping of buried waste sites was
demonstrated at a small buried
waste site at ORNL. Navigation
technologies were coupled with the
sensing information (from
radiation, gas, and subsurface
large object sensors) to
automatically map subsurface
materials.
A team consisting of LLNL,
SNL, LANL, SAIC, and IBM demon-
strated a robotic system for
loading powder into a furnace in a
Pu production line, and then
transferring the product to the
next operation in a mock up
facility. This robotic system
eliminates the need for operator
hands-on transfer operations and
reduces the generation of operator-
associated waste materials such as
wipes, protective clothing, gloves,
and transfer bags.
SITE VISITS/NEEDS
In March 1990 RTDP planning
teams visited five DOE sites.
Additional site visits will be
conducted in the future to expand
the planning basis.
The purposes of these visits
were (1) to understand the needs
and requirements of the highest
priority environmental restoration
projects and waste management
operations at the sites, (2) to
obtain information for use in
planning the program, and (3) to
describe the RTDP to personnel at
the site and discuss development of
the program plan. Emphasis was
placed on both technical and
schedular (i.e., compliance dates)
needs and requirements.
The results of these visits
are documented in a Site Needs and
Requirements Document. This
document summarizes the findings at
each site and highlights priority
needs.
APPROACH TO NEEDS DIRECTED
TECHNOLOGY DEVELOPMENT
The visits to five DOE sites
led to selection of six areas of
need for robotics technology to
support ER&WM activities. These
need areas are:
• Remediation of waste
storage tanks,
• Retrieval of buried wastes,
• Automation of contaminant
analyses,
• Waste minimization,
• Decontamination and
decommissioning,
• Waste Facilities Operations
Plans for development and
application of robotics technology
are based on the need areas listed
above. In addition, the plans
reflect other aspects of needs at
the sites such as regulatory
compliance dates, planned remedial
actions, and established schedules.
The fundamental approach to
developing robotics technology to
meet these needs couples available
and emerging technology with
advanced technology. Near-term
needs can be met by integrating
177
-------
available commercial technologies
with emerging technologies
available in RSD laboratories. At
the same time, development of
advanced technology will proceed to
meet intermediate and long-term
needs. In addition, attention will
be given to development of cross-
cutting technology which will be
applicable to multiple need areas.
Technology development will be
keyed to integrated demonstrations
at the DOE sites to further couple
the robotics technology development
to the site needs and to the
deployment of remedial actions
technology.
The DOE sites are evaluating
alternative approaches to remedial
actions. The robotics technology
developed for each application must
meet the needs, and match the
approach selected by each site.
The plans described for robotics
technology development are based on
reference concepts, selected as
reasonable and likely concepts from
the alternatives, which form the
basis for identifying needed
technology development, estimating
schedules, and estimating budgets.
The robotics technology
development plans are also keyed to
demonstrations of technology at the
DOE sites. Wherever possible,
demonstration of the robotics
technology is integrated with
larger integrated remediation
technology demonstrations.
CROSS-CUTTING AND ADVANCED
TECHNOLOGY DEVELOPMENT
Near-term applications of
robotics to ER&WM activities is
necessarily focused on existing
technologies that can be readily
adapted to the specific cleanup
tasks and environments. As the DOE
cleanup activities progress and
evolve, a larger body of robotic
technology will be needed for
application to ER&WM projects. A
technology development program
targeted at relevant cross-cutting
and advanced technology development
will make possible a more rapid
insertion of beneficial technology
into these activities. This
technology development will be
focused on high payback projects
that offer safer, faster, or
cheaper approaches to cleanup
goals.
An advanced technology
development program including a
long term research and development
component is a means to effectively
incorporate the expertise of the
universities, national laboratories
and other basic research
organizations into the nation's
cleanup projects. Also, this
offers educational training
opportunities consistent with the
DOE emphasis on developing the next
generation technical work force.
Needs identified at DOE sites
indicate that cross-cutting and/or
advanced technology development in
the areas listed below would be
highly beneficial to application of
robotics in ER&WM activities.
Mechanical Subsystems
Manipulators
End-Effectors
Mobile Systems
Control Subsystems
Computing, Graphics and
Modeling
Man-Machine Interfaces
Communications
Telerobotic Operations
Motion Planning and
Control
Sensor Subsystems
Environmental Sensors
Servo Mechanical Control
Sensors
Imaging & Vision Systems
Multi-Sensor Integration
Cross-cutting and advanced
technology developments need to
focus on near-term, mid-term, and
long-term implementations. By
investing in a sustained long-term
development program, emphasizing a
balanced evolution in technology
development with implementations
continually encompassing technology
advances, steady progress may be
assured toward the technology
required for the more complicated
or demanding tasks of the decades
to come. Development of advanced
robotics technology that is
commonly applicable to many
environmental restoration, waste
178
-------
management, and waste minimization
activities can lead to higher
efficiency, increased reliability,
and reduced life cycle costs in
these operations.
Participants in this program
are the following whom we wish to
thank for their contribution.
SAIC - Science Applications
International Corporation
LANL - Los Alamos National
Laboratory
SNL - Sandia National
Laboratories
LLNL - Lawrence Livermore
National Laboratory
ORNL - Oak Ridge National
Laboratory
T-12 - Oak Ridge Y-12 Plant
RF - Rocky Flats Plant/EGiG
Rocky Flats
SR - Westinghouse* Savannah
River Company
'.VHC - Westinghouse Hanford
Company
?NL - Pacific Northwest
Laboratory
£GSG - EGSG Idaho
INEL - Idaho National
Engineering Laboratory
WMC - Westinghouse Materials
Company of Ohio
WINCO- Westinghouse Idaho
Nuclear Company, Inc.
Fernald Feed Materials production
Center
179
-------
Field Robots for
Waste Characterization and Remediation
William L. Whittaker
Field Robotics Center
Carnegie Mellon University
Pittsburgh, PA 15213
(412)268-6559
David M. Pahnos
Field Robotics Center
Carnegie Mellon University
Pittsburgh, PA 15213
(412) 268-7084
Abstract
field operations for waste characterization and remediation
offer real opportunities and compelling motivations for
advanced robot work systems. The application of field
robotic technology can enhance the quality of data collected
al waste sites through standardization, verification, and
repeatability of methodology. It can increase the coherence
of data by enabling dense data collection, advanced
correlational databasing, and the collection of previously
unavailable data, such as position tagged data or
inteipretable 3D subsurface images. Held robots can operate
were humans are precluded, in pipes, tanks, abandoned
mines, and sea and river bottoms or where humans perform
inefficiently in protective clothing and breathing apparatus.
Thus, field robots can greatly increase the knowledge base
gained during site investigations; this knowledge will
expand remediation options performed by human and open
the way for the use of field robots in remediation activities.
Moreover, the development and use of field robotic
technologies in the service of national efforts to characterize
Mid remediate nuclear and hazardous waste will eventually
lave profound effects on large commercial industries and
open new world markets for robotic technologies.
Introduction
Hazard has been the historical justification for the use of
field robots; operations surrounding accidents at Chernobyl
"id Three Mile Island have world impact, preclude humans
and call field robots to action. Less reactive than these crises
are the innumerable nuclear, deep sea, military, and space
operations that are inhospitable to humans and are
significant both strategically and fiscally. The ultimate
opportunities, however, for field automation are those
immense and inefficient industries like construction, mining,
timbering, hazardous waste management, subsea and outer
space that dwarf the economics of manufacturing.
Characterization and cleanup of the nation's weaponry
complex alone is now estimated at 100 billion dollars: efforts
of this magnitude require new technologies. As a growing
technology, the potential of field robots to apply sensing and
analytical capabilities and to perform precise, repetitive, and
dangerous tasks is virtually untapped in the world.
Field robots work in environments as they are encountered,
not idealized or altered to accommodate automation. While
an assembly process can be structured into a limited number
of predictable actions, a robot working in an unstructured
environment encounters new situations that it has not been
explicitly programmed to deal with.
Field robots are thus challenged to perform goal-driven tasks
that defy pre-planning in unpredictable and changing
environments. In order to explore, work, and safeguard
themselves and the environment, field robots must sense
complex phenomena in a dynamic world. As these robots
move towards autonomy, they must plan and implement
their work tasks.
181
-------
Robots are quickly becoming mobile in natural terrains,
perceptive, self-navigating, and competent in the field.
Within the next few years, a number of robotic performance
niches in waste characterization and remediation will be
exploited where humans are precluded from the scene or
where robots offer superior capabilities. Areas of
opportunity include reconnaissance, surveying, subsurface
imaging, soil gas sampling, perimeter monitoring, fast
analytical screening, accident response, remote sampling
and manipulation, remote coring, and excavation.
Automated Characterization
Perhaps the most frustrating aspect of waste characterization
is the paucity of reliable data that scientists and engineers
have to work with following an investigation. Field sampling
is expensive, time consuming, and labor intensive. Although
methodologies are standardized, human judgement and
sometimes intuition are broadly applied when deciding
where to sample or survey and how to interpret data once
collected. This is particularly true for the selection of
boreholes, the interpretation of geophysical data, and the
selection of soil and soil gas sampling points. Analytical
instruments and techniques have improved greatly over the
past several years, but the results are only as good as the
choice of sampling points, which often are too few and
chosen poorly.
Field robots can deploy screening instruments far more
rigorously, sampling hundreds or thousands of times per
acre, achieving total site coverage. They can create a three-
dimensional data base by analyzing air, soil gas, and the
subsurface: they can screen organics on the fly and create 3D
images of buried waste from radar data, sampling at
centimeter resolution. Held robots can survey a site and
layout a precise grid; take samples, position tag, package,
and label them; position tag instrument data, store the data in
a single spatially correlated data base, and present multiple
types of data to users in a straightforward visual format.
Quality of Data
Capable field robots can greatly increase the quality of data
from a waste site by obtaining verifiable data with a high
degree of repeatability, and they can advance the process of
data collection to a higher standard than is possible using
present methodologies. Ultimately, field robots can also help
ensure that the right samples are sent to analytical
laboratories.
Standardized Data
Most waste sites have long lives; the time from preliminary
assessment to the remedial action can stretch into years, and
monitoring can take place for decades after. Throughout the
life of a site, scores of scientists, engineers, technicians, and
workman perform tasks, and as a site transitions from
assessment to investigation to remediation, the cast of actors
changes.
Although methodologies are standardized, no two
investigations at a site are performed in exactly the same
way; indeed, no two investigators can be relied upon to bring
the same experience, judgement, and skill to a site or to
collect data in exactly the same way, thus making it difficult
to achieve standardization.
Moreover, because waste sites vary greatly in topography,
soil types, geology, and the nature of contaminants, it is
difficult to achieve standardization across a range of sites,
partly because humans perceive the sites differently.
The use of field robots to collect and screen data can
significantly improve standardization. Robots can be relied
upon to treat data in the same way in each investigation.
Robots eliminate human variables and collect far greater
quantities of data. The data thus become more reliable, and
data from different sites can be compared legitimately.
Ultimately, a single, complete data base can follow a site for
its entire life. Created during the preliminary assessment, a
three-dimensional computer data base can be an interactive
repository in which each new set of data is entered.
Verifiable and Reoeatable Data
Field robots can verify data taken previously at a site and
repeat the collection and screening process precisely.
Because robots process and store data at the time of
collection, the chain of custody can be maintained more
reliably and securely. Repeatable outcomes translate into
defensible conclusions and reduce uncertainly when
182
-------
planning remedial actions and issuing a record of decision.
Field robots can become an important tool in the process.
Relevant Data
Two ways to increase the relevance of data are to collect it in
quantities great enough to yield high statistical reliability and
collect several types of data at the same time. Reid robots
can build dense data bases. They are also capable of
deploying a range of sensors that humans cannot; e.g., three-
dimensional laser rangefinders, infrared sensors, sonar,
radar, etc. In addition, they can deploy analytical instruments
smultaneously and determine their position accurately in
global coordinates.
UK site investigation robot (SIR) under development at
Carnegie Me lion's Field Robotics Center collects ground
penetrating radar data (GPR) at two centimeter intervals,
accumulating in excess of a 400,000 data points per acre.
GPR data are inherently three-dimensional and can be
processed into 3D images, if the data are dense enough. A
human cannot attain the positioning accuracy or deploy the
sensor with enough precision to collect dense data, as a robot
can. The result is not just more data but new and better data.
Further, the robot can be configured to collect additional
types of data or samples simultaneously, e.g., organics in air
or soil.
hJeipretable. Usable Data
Investigators are often confronted with data that do not
asily yield to interpretation or, at worst, require the
investigator to make a guess as to what the data show. Field
robots can process data, making it easier to visualize and
understand.
GMU's Site Investigation Robot provides a visual image that
is not only quantitatively better but qualitatively better than
standard GPR data bases. The user is provided with an image
defined accurately in x, y, and z, making the data more
Weipretable, even to a novice.
Dta bases become more usable when one is able to see
oxidations among data in new ways. The availability of
multiple types of data superimposed on a computer-
derated site map will enable investigations to gain a whole
site profile in a single visual image. This kind of user power
will not only speed the investigation process but give
entirely new insights to investigators.
Accessible Data
Finally, when data are accessible to many people over time,
the likelihood of good use being made of the data increases
significantly. Data collected by field robots can be stored on
central file servers, available to all who need to determine
what is known about a site or who have new data to add to
the file.
When Humans Are Precluded
Some investigations and remedial activities preclude
physical human access, such as the interiors of pipes, tanks,
and ducts; abandoned mines; and river, harbor, and sea
bottoms. Field robotic technologies offer the best access to
collect data and to perform remedial activities.
Generations of competent pipe crawlers have been
developed and are in service in petroleum and natural gas
industries. In-tank inspection robots and remediation robots
are needed at DOE complexes. One such robot is being
developed by RedZone Robotics to inspect tanks containing
nuclear waste. At CMU's Field Robotics Center, we are
developing autonomous navigation and vision systems for
underground mining equipment and autonomous navigation
systems for walking machines and wheeled vehicles to
traverse rough terrain. Others have significant experience
with competent sub-sea robots and have demonstrated their
capabilities and utility.
Another class of sites precludes humans because of health
and safety concerns, e.g., high-level waste, mixed waste,
transuranics, unbreathable atmospheres, unknown waste,
and accident response. These sites present high-motivations
for robots to perform not only reconnaissance and sampling
activities but forceful manipulation and heavy work to a high
degree of precision. These activities include excavation,
loading, haulage, and packaging of diffuse materials;
removal of sludges and mixing of materials; removal of
debris; barrel handling; boring on gassy landfills, and the
handling of explosive materials or operations in explosive
environments.
183
-------
Field Robotic technologies have now progressed to the point
where the robotics community can begin to build competent,
rugged, and reliable systems to meet the performance needs
of waste characterization and remediation programs.
Integrated Characterization and Remediation
Systems
Robotic technologies can fulfill the need to better integrate
characterization and remediation systems. An excellent
example of this is the case of trenched transuranic wastes.
Conditions preclude most invasive means of characterizing
the volume and position of the waste, and having a human
onboard of an excavator is precluded during the remediation.
The work can, however, be performed by robots in a
coordinated sequence. A site investigation robot (SIR), using
ground penetrating radar, can produce measurements of
buried waste in x an y to a reasonable accuracy (7 to 14
inches), which would allow a robotic excavator to trench on
both side of the waste to install steel sheeting. The excavator
would have the SIR's position data and subsurface map
available to it to guide it through the digging process, along
with active sensing of its own.
The SIR also surveys the z axis, determining the depth of the
waste and the distance from the soil surface to the waste.
Through a sequence of iterative sensing and excavation, the
clean overburden could be removed, leaving 4 inches of soil
covering the waste. The excavator could then remove the
waste autonomously.
In this scenario, robots working together can perform the
tasks more efficiently and with greater accuracy than human
operators. Five years ago sensing and control in both robots
to the degree of accuracy described above would have been
wishful thinking; two years ago it was beyond the reach of
the technology, today it is within reach, and although it is not
yet ideal for selectively finding and excavating, deeply
buried hot spots, it is likely the safest, most cost effective
approach to retrieving radioactive, trenched wastes that can
be expected in the next several years.
Future Opportunities
Commercial applications for capable field robots will
number in the hundreds. Among them are significant field
robotic applications that are achievable in the near term with
evolutionary extensions to our current technology base.
Moreover, there are significant opportunities, some of which
are unique to the U.S., e.g., robotic timbering, surface
mining, and large-scale agriculture.
Federal agencies should not miss opportunities to develop
and apply robotic technologies in programs where they have
a legitimate interest and obligation to protect human health,
increase productivity, and decrease costs. Because robotic
technologies are extensible to many applications, there
should be a coordinated effort by Federal agencies to 1)
focus performance-based research to move the technologies
forward; 2) apply the technologies in Federal programs
where they will produce high-leverage results, sufficient to
pay for the investment; and 3) ensure that programs will be
sufficiently stable over time to attract world-class
researchers to the field.
There is an opportunity to reduce significantly the total
cleanup costs of chemical and nuclear waste sites through
the programmatic development of robots to perform site
investigation, data collection, and remedial activities. The
core technologies have reached a stage of development to
begin the task of putting together integrated, teleoperated
and semi-autonomous systems for this purpose. The
opportunity is to alleviate a major national problem and, at
the same time, to develop and apply new technologies that
will impact the world.
184
-------
DISCUSSION
BRIAN PETERS: You mentioned American leadership. What about the position
of the Japanese in this area? They're well known for corporation robotics on
tdomobile assembly lines, for example.
WILLIAM WHITTAKER: The Japanese are a significant force in this arena.
hiticularly, they have programs that have matured, driven in a strategic way, top
down,over several years, and they look very good. They look extremely good in
construction. They have lesser presence in subsurface and in space. Consider, if
you will, that we enjoy a 20- or 30- year history in space, and they're just building
ieirfirstrockets. But to bring it to terms here, I look for the United States to drive
Ibis agenda because we are the ones who pioneered some of the nuclear
tdmologies, and we are the ones that have the volumes and the programs to go
atelhis.
Resistance, if you looked at the navigation technologies, there aren't a lot of
places in Japan that have enough roads to drive something like that. And so if you
tola the agenda in the program, I think that is enough to really focus operations
lot I actually have a video tape of condensed Japanese technology that I just
pal together this week. After this session I'll be happy to show that.
GREGG DEMPSEY: On your remote vehicles that stand completely alone,
0% run on telemetry or whatever) is the technology such that if there's an
incident out on a site or something, and you lose communication, can the
machine actually turn itself around and come back?
WILLIAM WHITTAKER: Yes, that technology is available. However, I think
it's important to know that it's in very select pockets of seasoned research groups,
and very select pockets of small organizations that can move fast to put it
together. Specifically, that kind of technology source is from the DOD Strategic
Computing Initiatives and DARPA's Road Following Programs, which were
funded at the hundred million dollar level over a number of years, going back
three or four years.
GREGG DEMPSEY: I remember when the robots went into Three-Mile Island
there were problems with the camera lenses darkening up because of the radiation
exposure. Has that problem been solved to any great extent?
WILLIAM WHITTAKER: In the first deployment in November of 1984, it's
true that the cameras didn't function well. And that's because we were using a
CCD technology. It was small, and it was very new! But within a month that was
straightened up. And with the years that have gone by, particularly out of military
and space initiatives, rad hardened CCD's are a known technology. It's very
straightforward now.
GREGG DEMPSEY: So we have technology that can operate in the thousands
of roentgens per hour now?
WILLIAM WHITTAKER: Yes.
185
-------
SPACE TECHNOLOGY FOR APPLICATION TO TERRESTRIAL HAZARDOUS
MATERIALS ANALYSIS AND ACQUISITION
Brian Muirhead Susan Eberlein
James Bradley William Kaiser
NASA/Jet Propulsion Laboratory
California Institute of Technology
Pasadena, CA
ABSTRACT
In-situ and remote measurements of elemental, molecular
and mineralogical composition of materials has been part
of the space science program since its beginnings. There is
a great deal of commonality between space science missions
and terrestrial hazardous materials screening in the types
of measurements, methods and instrumentation used.
There are also strong parallels between the hostile
environments of space and those of a hazardous material
This paper discusses the measurements, methods and
nstrumentation used on past, present and future space
missions for in-situ and remote analysis of materials.
Specific instrumentation discussed includes gas
chromatographs, mass spectrometers, imaging
spectrometers, X-ray and gamma-ray spectrometers.
Work sponsored by the National Aeronautics and Space
Administration's Sample Acquisition, Analysis and
Preservation technology program is discussed, including
concepts and hardware for multi-spectral remote sensing.
Instrument data analysis and interpretation, material
acquisition and processing. Some new concepts for micro
Sensors for making various chemical measurements are
also discussed. Possible applications of space technology to
terrestrial hazardous materials field acquisition and
analysis are presented.
NTRODUCTION
In-situ and remote measurements of elemental, molecular
and mineralogical composition of materials has been part
of the space science program since its beginnings. Two of
the best known surface science missions were the Viking
mission to the surface of Mars and the Soviet Venera series
to Venus. The Galileo spacecraft is carrying a probe to
sample Jupiter's atmosphere and the National Aeronautics
and Space Administration (NASA) has just started a project
to make a variety of in-situ measurements of the comet
Kopff. NASA is currently working on technology to enable
robotic and human missions to the Moon and Mars. Such
missions will include a wide variety of in-situ and remote
science and engineering measurements. There is a great
deal of commonality between space science missions and
terrestrial hazardous materials screening in the types of
measurements, methods and instrumentation used as well
as in the hostile nature of the environment in which these
measurements are made. NASA is very active in the design,
development and utilization of the instruments. Table 1
contains a listing of some science data requirements and
associated instrument(s) that are used and/or under
development within NASA for its past, present and future
missions.
NASA has established a technology program called Sample
Acquisition, Analysis and Preservation (SAAP) to address
the specific needs of in-situ science and engineering
measurements. SAAP is intended to develop critical and
significantly enhancing technologies for remote
identification, acquisition, processing, analysis and
preservation of materials for in-situ science, engineering
characterization and earth return. Although the technology
being developed in the SAAP program is not currently
being applied to specific missions, the SAAP program will
broaden the base of technology available for future
missions. Specifically, SAAP is developing concepts and
hardware for multi-spectral remote sensing, instrument
data analysis and interpretation, material acquisition and
containment [1,2,3,4]. Some new concepts for
microsensors for making various chemical measurements
are also under development. There are many possible
applications of space technology to terrestrial hazardous
materials field acquisition and analysis.
SPACE INSTRUMENTS, MEASUREMENTS AND
APPLICATIONS
There is very high scientific value to direct surface
measurements, independent of whether a sample is
returned to a laboratory. In particular, the analysis of
volatiles is probably best done in-situ due to the potential
for loss or chemical change after prolonged storage. For
space applications, in-situ measurements may be a
necessity because of the limitations on sample return.
187
-------
Table 1. SCIENCE DATA REQUIREMENTS vs INSTRUMENT TYPES
Required Data
Example Instruments
Elemental Composition
Mineralogical Composition
Water Detection and Mapping
Atmospheric Composition
Subsurface Structure
Seismometry
Volatiles
Imaging
Exobiology
Magnetic Fields
Gamma-ray Spectrometer, a-p-x Spectrometer
XRF, a-Backscatter
Visible-Infrared Spectrometer
Mossbauer Spectrometer, DSC, XRD
Neutron Spectrometer, Electromagnetic Sounder
GCMS, Laser Spectrometer
Electromagnetic Sounder, Active Seismometer
Passive Seismometer
DSC-EGA, Visible-Infrared Spectrometer
Camera, Imaging Sectrometer
Viking Biology Instrument
Magnetometer
Although terrestrial applications do not face the same
limitations, major advantages in speed and accuracy can be
gained by employing field analysis prior to selecting
samples for laboratory study.
Below are listed some of the characteristics of a few
instruments that have been flown by NASA or are being
proposed for NASA future missions. The constraints on
mass and power, combined with the need to function in a
hostile environment, place severe requirements on these
instruments. The technology developed to meet these
requirements could benefit the production of similar
instruments for terrestrial applications.
.AM,.
- ELECTRON
MULTIPLIER
ION SOURCE
HOUSING
r
ION
PUMP
r
1
1
MAGNET
L" = 2 6 CM
ELECTRIC
SECTOR
Re=47CM
Rm- 38 CM
L'e = 1 7 CM-i
= 1 7 CM
Figure 1. The mass spectrometer for the Viking Lander
GCMS. The electric sector has a radius of 4.7 cm.
Chemical Analyzers
The prime example of a chemical analyzer is the Biology
Experiment on the Viking Landers. The experiment
included a GC-MS system for analysis of organic
compounds in Martian soil [5]. The GC-MS part of the
system had a mass of 16 kg, measured 28 cm x 38 cm x 27
cm and consumed 25 to 125 W when active. When the
system was presented with a soil sample it could sift a soil
sample into a pyrolysis tube, seal the tube to a GC inlet,
perform a controlled heating on the sample, and perform a
mass spectral analysis of the GC effluent with
exceptionally high sensitivity. The mass spectrometer also
had a direct inlet for analysis of the Martian atmosphere.
Figure 1 shows a diagram of the mass spectrometer.
Currently under development for the Comet
Rendezvous/Asteroid Flyby mission is the Cometary Ice and
Dust Experiment (CIDEX) instrument that incorporates a
3-column GC system for evolved gas analysis over a
sample temperature range of -90 to +1000 C. The
instrument also includes an x-ray fluorescence
experiment in a 15 kg package that uses an average of
about 22 W. The system will analyze comet dust for
organic materials and elemental composition.
New GC-MS systems have been proposed that combine the
analytical speed of microbore GC columns with the
exceptionally high sensitivity of a focal-plane mass
spectrometer equipped with an integrating focal plane
detector. Such a flight system would be comparable in size
and mass to the Viking Lander GC-MS, but with analytical
cycle times of a few minutes and the ability to analyze GC
peaks separated by a few hundred milliseconds. Such a
system could measure dynamic processes or determine
planetary atmospheric composition while descending on a
probe or parachute. The robust, portable nature of such an
instrument would make it a good candidate for deployment
in terrestrial field screening activities as well. A gas
chromatogram from a laboratory prototype is provided in
Figure 2.
188
-------
10000 -i
>; 8000-
(0
I
I 6000-
Q
4000
2000
0-
4 5
8
lOuil
9
A
50 100 150 200 250 300 350
FRAME NUMBER
Figure 2. Chromatogram of a mixture of EPA priority
pollutants. Each 50 mg frame contains a time-integrated
mass spectrum from mass 25 to 500 amu. (Peak 1 is air
and peak 9 is toluene.)
Elemental Analyzers
Gamma-ray spectrometers have been used in orbiting
spacecraft to obtain elemental maps of atmosphere-free
bodies such as the moon. The Mars Observer spacecraft
wil contain a gamma-ray spectrometer for elemental
mapping of the Martian surface through its thin
atmosphere. The recently built and proposed gamma-ray
systems for elemental analysis have tended to follow
commercial technology by use of cooled germanium
detectors. These detectors use radiators aimed into cold
space to achieve the required temperatures. The detected
elements are those with naturally radioactive isotopes or
which are excited by cosmic rays. Long counting times are
needed. Related instruments may be useful in the remote
determination of radioactive isotope composition at
terrestrial sites.
New. high efficiency x-ray fluorescence analyzer systems
have been proposed for lunar and Martian landers that use
new toroidal focussing crystals to achieve many orders of
magnitude increase in x-ray flux from microfocus x-ray
lube sources to achieve rapid and high sensitivity analyses
[6]. With the use of uncooled mercuric iodide x-ray
detectors, such an x-ray fluorescence system might have
""ass of 4 kg. consume 10 W, and occupy a volume of about
35cm x 25 cm x 25 cm. The same microfocus x-ray
source could be used in a high-efficiency, toroidal-
fccussing powder x-ray diffractometer for identification of
minerals. Both instruments can work in an atmosphere of
w x-ray absorption density, such as that on Mars, or in
vacuum.
VISIBLE AND NEAR INFRARED REMOTE SENSING
Imaging spectrometers play a major role in both Earth
observation and planetary exploration. The Airborne
Visible/Infrared Imaging Spectrometer (AVIRIS) images
with 20 m x 20 m spatial resolution in 224 spectral
channels from 400 to 2450 nm wavelengths [7,8]. The
data, obtained from NASA ER-2 aircraft at 20 km altitude,
is spectrally and radiometrically calibrated to provide
information for disciplines such as ecology, geology,
oceanography, inland waters, snow hydrology and
atmospheric science. An AVIRIS type instrument might be
used for aircraft tracking of ocean oil spills, smoke
plumes, or other indicators of chemical contamination.
In addition to visible and near infrared imaging
spectrometers, NASA has developed a portable backpack
point spectrometer (Portable Instantaneous Display and
Analysis Spectrometer - PIDAS). At a mass of about 30
kg, PIDAS obtains and records with integrating detectors,
reflectance spectra in 830 bands from 400 to 2450 nm.
The instrument, developed at JPL, has been used to support
geological and ecological disciplines, and can be calibrated
for identification of a wide range of materials. The
instrument field of view is 10 to 30 cm when hand held.
NASA is currently, working to develop an adaptive, reliable
and compact imaging spectrometer system for autonomous
site and sample selection and analysis of materials. This
system will provide wide area as well as close-up
identification of minerals which is enabling for surface
science and engineering missions.
The key element of the SAAP remote sensing subsystem is a
multi-spectral imager based on the solid-state acousto-
optic tuneable filter (AOTF). This device operates on the
principle of acousto-optic interaction in an anisotropic
medium and acts as a controllable narrow band filter. The
current breadboard version can collect spectral images at
4 nm spectral resolution in the visible range (0.5 and 0.8
microns). It has been implemented with a 1000x1000
fiber optic bundle between the foreoptics and the AOTF.
The fiber optic cable enables the mounting and articulation
of the foreoptics, remote from the main spectrometer body.
Figure 3 shows the current breadboard hardware.
By altering the pass band sequentially, only the desired
spectral bands are collected. Each pixel has a spectral
signature associated with it and classification is
accomplished on the basis of elemental content and spatial
location. Figure 4 shows a set of spectrometer images of a
rock containing the rare earth mineral neodymium taken
in the range of 783-710 nanometers. The absorption
characteristics of this mineral at around 750 nanometers
is evident in the dark spot in the right-center of the second
row of images. Figure 5 shows the complete spectral
signature of neodymium as taken by the AOTF
spectrometer.
Although the current instrument operates in the visible
region, the AOTF technology will also allow construction of
tunable filters for the infrared and ultraviolet regions of
the spectrum, with a total range between 0.35 and 25
microns. This may provide a new class of tunable spectral
analyzers for a variety of space and earth applications.
189
-------
CO
O
Figure 3. AOTF Spectrometer Breadboard
-------
- ;Cl • lul ;SJk ;Ui liB -. lul
^ii ^^ ^ *& -' wv :'& ^ • *«•
iJl iJl iA lA xfll Jl
1200
1150
LL?
3 1100
>
z
9. 1050
—
m 1000
i—
z
950
900
560 nm to 528 nm
Figure 4. AOTF Spectrometer Output in
710-783 nm
The completed imaging spectrometer will be capable of
collecting high resolution images at hundreds of discrete
wavelengths. Processing of such a large amount of
information (>1 gigabit per scene) will strain
computational systems without some means of data
reduction. Hierarchical analysis schemes, in combination
with neural nets, have been shown to produce several
orders of magnitude reductions in total computation time
and are discussed below.
INTELLIGENT DATA ANALYSIS
Spectral data from a variety of instruments is used in
many areas of chemical analysis. The proceedings of the
First International Symposium on Field Screening Methods
for Hazardous Waste Site Investigation [9] report on the
use of fieldable instruments for mass spectroscopy, x-ray
fluorescence spectroscopy, infrared spectroscopy and
Raman spectrosopy. For any of these instruments, the
spectral data produced is complex, requires a highly
trained chemist to assist in the interpretation process,
and often requires extensive computer work for proper
analysis. In many cases the data analysis and interpretation
step presents a significant bottleneck which prevents the
most efficient utilization of the instruments.
Work done within the SAAP program has concentrated on
the the analysis of visible and near infrared spectra for
mineral determination [10]. The developing system
incorporates a number of data analysis methods and
algorithms which will transfer readily to use with other
types of spectral data. Application of these approaches to
the instrumental analysis required for field screening of
toxic waste will improve the speed and efficiency of the
analysis step. Table 2 shows a comparison for speed and
accuracy of four classification methods. The first matched
0.7 0.6
WAVELENGTH (urn)
0.5
Figure 5. Neodynium Absorption Spectrum
from AOTF Spectrometer
filter is a brute force approach using full dimensionality
of all patterns, and requiring the most computation. By
reducing the dimensions used for matching, or performing
the matching in several steps (e.g. a grouping step and a
finer classification step), the computation is reduced. The
hierarchy of neural network pattern classifiers combines
these approaches. Images consist of 32-band spectra for
all pixels, and are classified as one of 28 known minerals
in each case.
Neural networks are trained to recognize spectra or
classes of spectra by presenting many examples of each
spectrum, complete with noise and normal variation in
features. Following training, new variants of the spectra
contained in the training set may be identified with a high
degree of accuracy. During the training procedure, the
network extracts the common features among the training
examples representative of each type of spectrum, and
learns to recognize these as important identifying factors,
while the noise is discarded. Thus new spectra are
classified based on the presence of the diagnostic features
specific to a type of compound, without significant
interference due to normal variation, noise, and
background contamination. The major components in
mixture spectra may also be identified, if the mixing
process does not obscure the critical features.
The neural network spectrum classifiers currently used
within the SAAP system work hierarchically, placing
spectra into progressively more detailed classes. This
approach allows either a rough estimate of mineral
composition, or a very detailed analysis and identification.
The final analysis step includes an assessment of the
classification accuracy. This allows the system to identify
those spectra which were poorly classified, and which may
represent mixtures or other unexpected spectra. Since the
191
-------
Table 2. COMPARISON OF 4 SPECTRAL CLASSIFICATION METHODS
METHOD
Single Matched Filter
Reduced Dimension Matched Filter
Two Step Matched Filter
Hierarchy
DATASET TOTAL OPERATIONS
Mars
AISA
Mars
AISA
Mars
AISA
Mars
AISA
16,226,560
5,017,600
8,113,280
2,508,800
6,374,720
1,971,712
4,858,284
1,006,099
ACCURACY
80%
80%
69%
89%
Note: Mars dataset is a simulated multispectral image derived from a Viking Lander image.
AISA dataset is a real multispectral image taken by the Airborne Imaging Spectrometer.
final application of this spectral analysis system requires
almost complete automation of the analysis process, the
results of the spectral analysis are integrated into an
automated decision making procedure. The decision making
is goal-driven: specific classes of minerals may be
searched for and analyzed in great detail, while other less
important compounds are discarded at an early step in the
analysis procedure.
The goals of the existing (planetary) spectral analysis and
decision making system include identifying interesting and
uninteresting areas on the basis of spectral information,
and identifying samples which should be acquired for more
detailed analysis. Similar goal driven systems could be
designed with the objectives of finding specific types of
chemical compounds or determining which samples will
prove most informative regarding chemical distribution in
an area. The hierarchical goal driven architecture allows
the system to analyze many samples rapidly, and to provide
the user with information regarding which samples are
most important for further examination.
Application to field screening for hazardous waste:
Two aspects of the work done for spectral data analysis in
planetary exploration will be of interest for the field
screening of hazardous waste. The neural network based
spectral analysis approach will be useful for the analysis
of IR, XRF, Raman, and mass spectra, if networks are
trained with real spectra gathered under the anticipated
field conditions. The hierarchical analysis architecture
that incorporates goal driven decision making may be
adapted to assist field workers in making rapid decisions
regarding the areas requiring special attention during a
field screening operation.
Although special neural network pattern recognition
systems will be required for each type of instrument data,
the basic algorithms developed for the analysis of
visible/near IR mineral spectra should transfer readily to
the analysis of other spectra. A hierarchical, neural
network based spectral identification system will have
several applications:
1. Unknown identification.
A network based hierarchy can replace a library search
procedure with favorable results for the identification of
unknown spectra. Progress is being made in the
implementation of hardware network pattern matchers
which will allow the equivalent of very large library
search procedure to occur in microseconds.
2. Searching for specific compounds.
A hierarchy of networks is particularly well suited to the
search for specific compounds. A spectrum is presented to
the hierarchy, and is progressively classified until it
becomes apparent that the spectrum does not represent the
desired compound (or until the desired compound is found).
A negative result is usually determined fairly quickly,
since at each step of the hierarchy, a large group of spectra
may be eliminated since they are not potential matches.
3. Searching for classes of compounds based on specific
features.
This is a variant of the hierarchical search for a specific
compound, with the difference that a positive result may
occur when a given branch point of the hierarchy is
reached, rather than only at the end of the search. The
192
-------
hierarchy is designed so that the groups of spectra that
represent important classes are together within a branch
of the hierarchy. The selection of critical spectral features
for identifying a class is ensured by using specific spectral
bands for training the networks. Extensive knowledge of
the chemistry is required at the training step for optimal
results.
4. Extracting major components from mixtures.
Identification of spectra of mixtures presents problems for
traditional library search and match techniques. Since
mineral spectra generally derive from mixtures of pure
minerals, this problem is being addressed in the work
wttiin the SAAP program. The neural network approach
has the advantage of basing results on important features
which are extracted from the anticipated data in advance,
rather than on complete spectral matching. This allows
identification of major components in many mixtures.
Situations where mixing causes masking or shifting of
critical spectral features require special treatment.
SYSTEM CONCEPTS
h-situ analysis systems can range from single
instruments placed on the surface to multi-purpose,
mobile units looking for specific materials or unique
materials units. An autonomous space exploration system
nil require the functions of planning, analysis, execution
control, reflex action, data processing and interpretation,
in order to operate in real time in a hostile environment.
for an in-situ analysis subsystem, the spectrum of
possible architectures can be characterized by two
extremes. At one end is a set of disjoint, self-contained
elements working more or less independently to perform
the required functions. At the other extreme is a fully
integrated system with many interdependent relations
between the elements. The former case is probably more
comparable to the terrestrial applications, where several
independent instruments are operated by humans. This
system design causes some problems for space systems
since it is not efficient in terms of mass or power and
compromises science due to uncoordinated measurements.
Multi-instrument data fusion and corroboration is an
important consideration in this system design.
An extreme example of the latter case is a multi-purpose,
foctory-like system, implementing a set of processes that
nay vary significantly depending on the desired outcome or
product. Physical material, not just data, must move
Between the elements. Current requirements and desires
brcoordinated measurements as well as mass, power and
volume limitations make an integrated design approach the
logical basis for technology requirements, but this
approach clearly pushes technology. Technology developed
fa such an integrated system could be applicable to the
automation of sample gathering and analysis in extremely
tostile earth environments, in cases where human
fraction must be remote and limited for safety reasons.
Technology will be validated in the laboratory and then
toegrated into the series of evolving SAAP testbeds. The
representative environment provided by the testbed will
be used to verify technologies and demonstrate overall
SAAP operational capability. By the end of September
1992 an initial laboratory testbed will be constructed to
demonstrate sample identification and acquisition. By the
end of 1995 a fully functional system testbed will be in
operation which will transition into a complete self-
contained transportable testbed for end-to-end "field"
operations. A preliminary system conceptual design of a
SAAP platform with a full complement of subsystem
components except for a regolith deep core drill is shown
in Figure 6. This configuration can be considered a
preliminary model for the full-up system testbed; no final
payload or mission configuration has been selected.
SAMPLE ACQUISITION
The capability to acquire physical samples robolically,
without human intervention, would be significantly
beneficifial in many hazardous waste screening
applications. The principal requirement driving sample
acquisition for planetary exploration is to obtain samples
of weathered and unweathered materials from accessible
rocks or outcrops. These samples must not be significantly
altered either mechanically or thermally during
acquisition. Conceptual designs and early experimental
work have been completed to help understand the
mechanical, controls and automation issues for sample
acquisition in the hostile environment of a planetary
surface. Effort has focused on mechanical designs to
achieve functional capability and is now proceeding to
include testing of control and automation methodologies.
Laboratory validation at the component level will be
followed by further development and verification at the
system level in a series of SAAP testbeds.
Various techniques have been studied for sample
acquisition including sawing, coring and chipping. Of
these, core drilling represents an efficient way of
obtaining surface and subsurface samples that are easy to
handle by a preparation or storage subsystem. Terrestrial
coring processes, however, require direct human
supervision and utilize high power and introduce large
volumes of fluid to aid the cutting process by cooling the
bit and removing cuttings of rock and/or soil.
SAAP has developed the means for core drilling low
porosity, high compressive strength rocks without the use
of coolant. High velocity diamond matrix core barrels are
used under the control of robotic manipulators. Under
study are various control approaches and a variety of
sensors modalities including, position, force, vision,
spectral, temperature and vibration. Progress in this area
should improve the prospects for remote robotic
acquisition of solid samples from hazardous areas on earth
as well.
In addition to tools, work is underway to identify and
develop end effector and manipulator technologies
necessary for the sample acquisition operations.
Preliminary studies of end effector and manipulator
dexterity versus reliability, mass, power and performance
have been made for some mission scenarios. The current
193
-------
6 DOF ARM
7 DOF ARM
SAMPLE PREPARATION SYSTEM
MULT-
SPECTRAL
IMAGER
TOOL/INSTRUMENT
BOX
SAMPLE CANISTER
X-RAY FLUORESCENCE
SPECTROMETER
-REDUCED DEGREE OF
FREEDOM END EFFECTOR
X-RAY DIFFRACTOMETER
DIFFERENTIAL SCANNING
CALORIMETER
•GAS CHROMATOGRAPH/
MASS SPECTROMETER
Figure 6. SAAP Preliminary System Conceptual Design
state-of-the-art in end effectors consists of either very
limited capability industrial vise-type grippers, or
extremely complex anthropomorphic designs being studied
in research laboratories. In general, the fewer degrees of
freedom the better for simplicity. However, to achieve
high inherent reliability, mechanical redundancy at each
degree of freedom will be required. Concepts that provide
adaptability or flexibility and involve trade-off's of
degrees of freedom with redundancy will be studied
further.
ADVANCED CONCEPTS
NASA is interested in developing new sensing device
technology for in-situ science investigations. Currently
available instruments for in-situ science investigations
are often incompatible with mission requirements due to
their excessive mass, volume and power consumption.
Science capabilities may be significantly extended by the
development of sensing device systems which represent
smaller payloads. The sensing device development is
directed to enable compact, low-mass, low-power
consumption instruments for a variety of mission
requirements. The advanced technology of silicon
micromachining for device fabrication will be employed to
implement highly capable, sensitive, and robust
instruments while retaining compact structure and low
mass attributes.
The development of silicon micromachined gas
sensors will be based on the compact gas chromatography
(CaC) instruments recently demonstrated in silicon
micromachined structures. The key components of the
compact GC systems include a silicon micromachined gas
dispersion column, integral gas metering valves and
silicon thermistor gas detectors, fabricated entirely on a
single silicon wafer. The successful operation of this
prototype time-of-flight GC system indicates the range of
opportunities for unique instruments of this type. In this
task, specific gas detector applications will be identified
and instrument requirements will be formulated. Gas
sensors and instruments will fabricated and tested for
operation in the Martian atmospheric environment
Finally, with results of device testing, complete
instruments will be designed for specific mission
applications.
CONCLUSION
This paper has discussed some of the measurements,
methods and instrumentation used on past, present and
future space missions for in-situ and remote analysis of
materials. Work sponsored by NASA's Sample Acquisition,
Analysis and Preservation technology program included
concepts and hardware for multi-spectral remote sensing,
instrument dafa analysis and interpretation, and material
194
-------
acquisition, and new concepts for micro sensors for making
various chemical measurements. Much of the technology
under development in the SAAP program has application to
terrestrial hazardous waste materials acquistion and
analysis.
REFERENCES
1) Moreno, C., Editor, In-situ Science Investigation
System Catalog, Version 1.0, JPL Document, June 5, 1990
2) B. Muirhead and G. Varsi, (1990), Next-
Generation In-situ Science Concepts and Technology, IAF
90-444, 41st Congress of the International Astronautical
Federation, Oct. 6-12, 1990.
3) Muirhead, B., et al, (1989), " Sample Acquisition,
Analysis and Preservation Technology Development",
Presented at the 2nd International Conference on Solar
System Exploration, Pasadena, CA.
4) Plescia, J., Editor, Sample Acquisition, Analysis
and Preservation Instrument Technology Workshop,
Proceedings, Johnson Space Center, November 14-16,
1988.
5) D.R. Rushneck, A.V. Diaz, D.W. Howarth, J.
Rampacek, K.W. Olson, W.D. Dencker, P. Smith, L
McDavid, A. Tomassian, M. Harris, K. Bulota, K. Biemann,
Al. LaFleur, J.E. Biller, and T. Owen, (1978), Viking gas
chromatograph-mass spectrometer, Rev. Sci. Instr.
49(6), 817-834.
6) D.M. Golijanin and D.B. Wittry, (1988).
Microprobe x-ray fluorescence analysis - new
developments in an old technique, Microbeam Analysis-
1988, D.E. Newbury, Ed., San Francisco Press. 1988,
397-402.
7 ) W.M. Porter and H.T. Enmark, (1987), A system
overview of the Airborne Visible/Infrared Imaging
Spectrometer (AVIRIS), Proc. SPIE. 834.
8 ) G. Vane, M. Chrisp, H. Enmark, S. Macenka, and J.
Solomon, (1984), Airborne Visible/Infrared Imaging
Spectrometer: An advance tool for earth remote sensing,
Proc. 1984 IEEE Int'l Geosciences and Remote Sensing
Symposium, SP215, 751-757.
9 ) Field Screening Methods for Hazardous Waste Site
Investigation, First International Symposium Proceedings,
October 11-13, 1988.
10) Eberlein, S., Yates, G. (1990). "Neural Network
Based System for Autonomous Data Analysis and Control",
In "Progress in Neural Networks". Volume 1, pp 25-55,
Ablex Publishing Corp.
ACKNOWLEDGEMENTS
The Sample Acquisition, Analysis and Preservation
Program is part of the Exploration Technology Program
within the NASA Office of Aeronautics, Exploration and
Technology. This project is the joint effort of the Jet
Propulsion Laboratory, Johnson Space Center and Ames
Research Center, with JPL as the lead center.
The research described in this paper was carried out by
the Jet Propulsion Laboratory, California Institute of
Technology, under a contract with the National Aeronautics
and Space Administration.
DISCUSSION
BRIAN PIERCE: My question concerns the fiber optic bundle. You said
infrared. Do you mean the near infrared or closer to the mid i.R.?
SUSAN EBERLEIN: Right now the fiber optic bundle that we've actually
wwked with has only been in the visible range. We're looking this year in the near
infrared of 1.2 to 2.5 microns. In the long-term maybe more, but I gather that as
you go further into the infrared you get more trouble with your fibers.
BRIAN PIERCE: Yes, that's right. You also mentioned very intriguing hard ware
neural networks. What do you mean by that?
SUSAN EBERLEIN: What I mean by hardware neural networks is micro
silicon chips where the connection weights for the neural network matrices are
actually in the resistances in the chips. JPL is fabricating some of these. They are
still in the early stages, and not as precise as we need them. Some othercompanies
are working on making them commercially as well. If in fact they turn out to be
a viable technology that can be space qualified, they offer very, very rapid
processing for specific problems.
195
-------
DEVELOPMENT OF A REMOTE TANK INSPECTION (RTI)
ROBOTIC SYSTEM
Chris Fromme
RedZone Robotics, Inc.
2425 Liberty Avenue
Pittsburgh, Pennsylvania 15222
(412) 765-3064
Barbara P. Knape
RedZone Robotics, Inc.
2425 Liberty Avenue
Pittsburgh, Pennsylvania 15222
(412) 765-3064
Bruce Thompson
RedZone Robotics, Inc.
2425 Liberty Avenue
Pittsburgh, Pennsylvania 15222
(412) 765-3064
ABSTRACT:
RedZone Robotics, Inc. is developing a Remote Tank
Inspection (RTI) robotic system for Westinghouse Idaho
Nuclear Company to perform remote visual inspection of
corrosion inside high level liquid waste storage tanks. The
RTI robotic system provides 5.8 m (19 ft) of linear extension
inside the tank to position a five degree-of-freedom robotic
arm with a reach of 1.8 m (6 ft) and a payload of 15.9 kg (35
Ib). The primary end effector is a high resolution video
inspection system. The RTI Intelligent Controller provides a
standardized, multi-tasking environment which supports
digital servo control, I/O, collision avoidance, sonar
mapping, and a graphics display. The RTI robotic system
features an innovative, standardized, and extensible design
with broad applicability to remote inspection,
decontamination, servicing, and decommissioning tasks
inside nuclear and chemical waste storage tanks.
I. APPLICATION
Westinghouse Idaho Nuclear Company (WINCO) will
use the RTI robotic system at the Idaho Chemical Processing
Plant (ICPP) to perform remote visual inspection of corrosion
inside high level liquid waste (HLLW) storage tanks. The
ICPP tank farm consists of several HLLW storage tanks that
are 15.2 meter (50 ft) in diameter with a capacity of
1,135^00 liters (300,000 gallons). The domed roofs of the
tanks are buried 6.1 m (20 ft) below ground level. The bottom
of the tanks are located approximately 12.5 m (41 ft) below
ground level. The tanks will be drained of liquid prior to
inspection,-however a 30 cm (1 ft) layer of caustic sludge will
remain on the bottom of the tanks. The only access to the
tanks is through 25 cm (10 in) and 30 cm (12 in) diameter riser
pipes which extend from ground level down into the tank
roof dome. Accessible risers are typically located 0.8 m (2.5
ft), 3.6 m (12 ft), and 6 m (20 ft) away from the tank wall.
Currently, the RTI system will only be deployed through the
30 cm (12 in) tank risers. Cooling coil arrangements line the
tank walls and the tank floor.
The primary mission of the RTI robotic system is to
perform remote visual inspection of the interior walls of the
tanks for corrosion which may have been caused by the
combined effects of radiation, high temperature, and caustic
chemicals present. Due to the location and limited number of
accessible risers inside a tank, the intent is to inspect only a
pie-shaped portion of the tank to qualify the typical
condition of corrosion inside the tank. Thus the application
does not require a robotic arm with a long reach.
n. SYSTEM OVERVIEW
The RTI robotic system features a vertical deployment
unit, a robotic arm, and a remote control console and
computer. One of the major design constraints for the RTI
system is that the in-tank components are inserted through a
25.4 cm (10 in) diameter riser. This criteria lead to the
design of compact, electric actuators for the robotic arm,
which provide high torque and absolute position feedback.
The RTI robotic system is initially lowered by a facility
crane into the top of the riser. The vertical deployment unit
then provides another 5.8 meters (19 ft) of servo controlled
extension inside the tank. The RTI robotic system transmits
minimal loading to the riser pipe since it is self-supporting
via a support structure that rests on the ground above the
riser. Figure 1 provides an illustration of the RTI robotic
system installed inside a tank.
A five degree-of-freedom robotic arm provides 1.8
meters (6 ft) of articulated reach to accurately position a
high resolution video inspection camera to examine the tank
walls. The arm has sufficient dexterity to position the
camera normal to the curvature of the tank wall. The
controller provides coordinated end point motion so that the
operator can easily jog the arm inside the tank. A graphics
display is provided at the control console to give the
operator a sense of how the arm is positioned inside the
tank. The robotic arm also positions a pressurized spray
nozzle to wash down the tank walls prior to inspection. In
addition, the end of the arm has an interchange flange to
allow the robotic arm to carry a gripper instead of the
inspection camera. Another camera system is mounted at the
top of the robotic arm to provide the operator with an
overview of the arm operating inside the waste tank. The
RTI robotic system is capable of manual recovery to retrieve
the system in event of motor failure.
197
-------
LUTING CAGE
ID'S" NOMINAL
911- NOMLNAL
Z-AXIS ACTUATOR WITH BRAKE
AND HOMLVG LIMIT SWITCH
S-5" NOMINAL
411" TO 511-ADJUSTABLE
TETHER MANAGEMENT SYSTEM
19-OF CABLE
COUNTER WIND ELIMLVATES NEED
FORSUTRING
SUPPORT STRUCTURE
0 • ir ADJUSTABLE FOOT PADS
GUIDE SLEEVE
167- NOM. RETRACTED/ 35V NOW. EXTENDED
HORIZONTAL
REACH 6TT
(WITHGRIPPER)
SHOULDER ROTATE (±180°)
.TILT MOUNTED OVERVIEW CAMERA
REMOTE FOCUS, ZOOM AND IRIS
ATT VARIABLE INTENSITY LIGHT
'NARSENSO:
SHOULDER PITCH (±90°)
EL BOW PITCH C±120->
WRIST ROLL a ISO-)
WRIST PITCH (±120-)
CRTPPER(t!-4-.0-70LBS)
25T SOM. RETRACTED/ 431 r NOM. EXTENDED
41 tT TO TANK BOTTOM -
Figure 1. RTI Robotic System Deployed Inside HLLW Waste Tank
198
-------
The RTI system is radiation and environmentally
hardened to assure reliable performance in the tank
environment. The design criteria requires that all in-tank
components be capable of withstanding a 20 psi washdown of
10% nitric acid and 10% oxalic acid, radiation field of 100
Rad/hr for a total accumulated dose of 10,000 Rad, and
operating temperatures of 4 to 49 °C (40 to 120 °F) at 100
percent humidity. The RTI system uses sealed components
such as connectors, video equipment, sensors, and actuators to
preclude the intrusion of decontamination fluids. Bearing
and wear surfaces are stainless steel and non-stainless
components are anodized or coated with epoxy paint to
prevent damage from caustic decontamination fluids.
The RTI's control system uses RedZone's standardized
Intelligent Controller for Enhanced Telerobotics to provide a
high speed, multi-tasking environment on a VME bus.
Currently, the robot is controlled in a manual, joint jog mode
or a coordinated end point motion control mode. Control
capability is available to develop a pre-programmed,
automated or teach/playback mode of operation. The
control system incorporates sensing and software safeguards
to prevent an operator from inadvertently colliding with the
tank wall. Collision prevention is implemented in software
and backed up with four proximity sensors. A sonar range
finding sensor is used to establish the orientation of the RTI
robotic system inside the tank.
HI. MECHANICAL DESIGN
The major components of the RTI mechanical system are
the support structure, vertical deployment unit, robotic arm,
accessories, and strongback. These assemblies are described
in the sections that follow.
A. Support Structure
The support structure rigidly supports the vertical
deployment unit at ground level. It consists of the alignment
guide sleeve and support stand assembly. The support stand
is a four legged structure that spans the riser pipe and
bunker. Its leg pads provide 1 foot of vertical adjustment and
allow the stand to be levelled. A facility crane is used to
position the support structure over the riser and to insert the
alignment guide sleeve into the riser pipe. The guide sleeve
follows the inclination of the riser pipe to guide the vertical
deployment unit during insertion. The objective is to avoid
loading the riser pipe if it is not absolutely vertical.
B. Vertical Deployment Unit
The vertical deployment unit provides 5.8 m (19 ft)
of servo-controlled vertical extension, at speeds of up to 7.6
cm/sec (3 in/sec), to position the robotic arm inside the waste
tank. The vertical deployment unit consists of a telescoping
tube assembly, cable management system, drive motor, and
junction box. The telescoping tube assembly contains a fixed
outer tube and an inner extending tube to minimize the
overall retracted height of the system. With the inner tube
extended, the wrist flange of the arm can reach the tank
floor. An adjustable hard stop is provided to safely reduce
the extent of vertical travel. The outer tube is a 20 cm (8 in)
square stainless steel tube and the inner tube is a 15 cm (6 in)
square tube. The vertical deployment tubes are designed for
deployment through 30cm (12 in) risers. However, the
robotic arm is designed to pass through a riser as small as
25 cm (10 in). The inner extending tube is supported and
guided along the upper tube by stainless steel linear bearings
and rails. The rails are mounted along the length of the
inner tube and the bearing blocks are attached to the inside
of the outer fixed tube.
An electric motor drives the lower tube, Z-axis, by a
dumb-waiter arrangement of a drive chain and pulley. The
motor package includes an integral gear reducer, brake and
resolver. The motor's output shaft is directly coupled to a
drive sprocket which drives a steel chain attached to the
upper section of the inner tube. The chain moves within the
gap between the upper and lower tubes. The drive sprocket
was designed so it can be driven from either side. In the
event of a motor failure, an identical backup motor package
can be quickly mounted in order to drive the telescoping tube
assembly. Due to the relatively large gear ratio and large
travel of the chain, absolute position feedback on the
vertical deployment was avoided. Instead, a resolver is
attached directly to the motor shaft and a limit switch is
used to home Z-axis position at start-up.
After insertion into the riser, the top flange of the
vertical deployment unit is bolted to the guide sleeve. On
top of the vertical deployment are located the cable
management drum and a junction box. Cabling is payed out
from a spring loaded cable drum which has a large diameter
so that only two wraps are required to pay out the 5.8 m (19
ft) of cable length. This design obviates the need for
electrical slip rings. The vertical deployment junction box is
connected to the control console with 30.5 m (100 ft) of cable.
The junction box contains some pneumatic and valve
equipment and terminal strips but no circuitry. Its main
purpose is to serve as a termination point for cables routed
down the vertical deployment unit to the robot arm.
At the base of the vertical deployment unit is a
mounting flange for the robotic arm. Cables are routed
internal to the inner tube and exit the tube at its bottom. At
the bottom of the outer fixed tube, a spray ring is mounted to
spray decontamination fluid on the inner tube as it retracts
upward. This minimizes the spread of contamination inside
the telescoping tube assembly.
C. Robotic Ann
The RTI robotic arm mounts to the bottom of the
lower extending tube. The arm is a five-degree-of-frcedom
revolute arm consisting of shoulder rotate, shoulder pitch,
elbow pitch, wrist roll and wrist pitch axes. The primary
function of the robotic arm is to position the WINCO
inspection camera system mounted to the wrist flange. The
arm has sufficient degrees of freedom to position the
inspection camera normal to the curvature of the tank wall.
Coordinated end point motion control allows the operator to
move the inspection camera in/out and along the curvature ol
the tank wall. An overview camera is packaged between
the shoulder rotate and pitch joints to rotate with the arm,
allowing a continuous view of the end of arm. A spray nozzle
is attached to the robot wrist so that the robot can wash
down the tank wall prior to corrosion inspection.
The robotic arm weighs approximately 100 Kg (220 Ib)
and has an overall length of 2.5 m (8 ft). The arm has a 1.6 m
(64 in) length to the wrist mounting flange, providing the
199
-------
robot with a 1.8 m (6 ft) reach when positioning the
inspection camera. The last three joints of the arm, elbow
pitch, wrist roll and wrist pitch, are clustered in close
proximity to provide dexterous manipulation. All axes are
electrically driven, feature absolute position feedback, and
are actively servoed to hold position. Upon loss of power,
the controller automatically shorts the motor leads to
provide dynamic braking. Gravity will backdrive the arm
into a nearly vertical position so the RTI system can be
removed from the riser in a manual recovery mode. Table 1
provides performance characteristics of the arm.
Table 1. Performance Specifications
Max
Description TraselVelocity
Shoulder Rotate
Shoulder Pitch
Elbow Pitch
Wrist Pitch
Wrist Rotate
±180°
±90°
±120°
±120°
±180°
1.0 rpm
1.0 rpm
2.4 rpm
55 rpm
5.5 rpm
Reach of Arm 6 feet
Coordinated End Point Motion
2.5 ips
Key: ips = inches/sec, rpm = rev/minute, ° = degrees
The five joints of the robot arm are driven by three
different sized actuator packages as specified in Table 2.
The three actuators are similar in concept and design but
provide differing torque and speed characteristics. The
capabilities of these actuators were optimized to meet the
goal of providing a 15.9 Kg (35 Ib) payload for the robot.
The actuators are designed into a compact, pancake-style
package. In the case of the shoulder pitch it was necessary
to keep the actuator small enough to fit sideways, in profile,
through the 25 cm (10 in) riser. Frameless DC high torque
brush motors were used as they offer the smallest size,
highest torque and lowest speeds available. Each motor is
coupled to a pancake type Harmonic Drive gear reducer,
providing a single step reduction of up to 200:1. These drive
components are integrated with slim line ball bearings and a
resolver to produce compact servo-actuators capable of large
torques. The integral resolver is directly coupled to the joint
output allowing precise, absolute, servo control of the arm.
Table 2. Mechanical Characteristics of Actuator Packages
Robot Joints
Shoulder
Rotate&Pitcl
Elbow Pitch
Wrist
Roll & Pitch
Actuator
Size
Heavy
Medium
Light
Dimensions
9.0" dia x 4.5'
351bs
6.5" dia x 3.5'
18Ibs
5.2" dia x 3.0'
81bs
Max
Torque
(in-lbs)
8400
2500
800
Max
Speed
(RPM)
1.1
2.4
55
The actuators and links are constructed of aluminum,
which is anodized on all exterior surfaces. The actuators are
environmentally sealed to protect them from the
decontamination solution. Since the actuators are not
equipped with brakes, they experience a 100% duty cycle
when the arm is loaded, causing the motors to heat up
significantly. Analysis of the system indicates that the
actuators' capabilities are thermally limited. That is, the
maximum payload of the arm is dictated by the motors
maximum winding temperature of 155 °C (311 °F) and not by
the maximum mechanical torque of the actuators. To
increase the actuator and arm payload capabilities an air
line is run into the actuators to provide cooling for the
motors. Cabling to each of the joints and tooling is routed
along the I-beam shaped linkages of the arm. Submersible,
molded connectors are provided on each motor.
D. Accessories
Accessories for the RTI robotic arm comprise the
quick change interface, decon spray nozzle, gripper,
overview camera system, sonar sensor, and proximity sensors.
A description of each accessory is provided below:
• A manual quick change interface is provided at the
wrist mounting flange to change end effectors
(inspection video system and gripper). The interface
consists of an electrical connector, pneumatic connectors,
and a common mounting plate.
• A decontamination spray nozzle is mounted directly
above the wrist flange to wash down the tank walls. It
has an adjustable flowrate of up to 15 liters (4 gallons)
per minute.
• A pneumatic parallel jaw gripper is provided with a 10
cm (4 in) stroke and adjustable gripping force of up to 482
kPa (70 psi).
• The overview camera system consists of an
environmentally sealed color camera with a zoom and
focus lens. The camera is mounted inside a cut-out
section of the robot shoulder linkage. A rotary actuator
provides the ability to pitch the camera along the
robot arm while zooming in for close views. Remote
control of the camera, rotary actuator, and light
intensity is provided at the control console.
• A miniature sonar detector is used to determine the
relative orientation of the robot inside the tank. The
sonar detector is mounted on the shoulder of the robot
arm to calibrate shoulder rotation to distance of the
tank wall. Since the risers are not located on the center
line of the tank, radial extensions from the riser to the
tank wall vary in length. An applications software
package automatically controls the sonar sensing and
rotation of the shoulder axis. The software processes
the data to identify the location of the wall closest to
the riser. Once distance to the tank wall is known as a
function of shoulder rotation, distance of the robot's end
of arm to the tank wall can be calculated based on
forward kinematics. Distance to the wall is displayed
on the graphics monitor and also used for software
collision avoidance. The accuracy of this information is
dependent the combined accuracy of the robot, sonar
detector, data processing, arm dimensions, and the
assumed location of the riser.
200
-------
• For impending collision detection, four photoelectric
proximity sensors are mounted on the leading edge of
the robot arm linkages to detect close proximity to the
tank wall.
E. Strongback
The Strongback fixture rigidly supports the RTI
robotic system during shipment. It is designed to attach to
the bed of a semi-trailer truck. The Strongback consists of a
tubular framework to cradle and support the full 10.7 m (35
ft) horizontal length of the RTI system. For additional
protection, the robotic arm is housed inside a reinforced cage
before it is placed onto the Strongback. A facility crane is
used to pivot the RTI robotic system vertical from the
Strongback during deployment at a riser.
IV. CONTROL CONSOLE
The RTI control console is the remote station from
which an operator can control and monitor the robotic arm to
perform visual inspection of the tank. The control console
will be located on a desk top inside a trailer located up to
30.5 m (100 ft) from the RTI mechanical system. The console
consists of two side-by-side 48 cm (19 in) racks which
maximize the useful working and viewing area to the
operator. The racks are encased in structural foam and
housed together in one self-contained shipping container. A
removable front cover protects the monitors and control
panels during shipment. All cables enter the control console
through external chassis mounted connectors.
The control console is composed of industrial grade
components, rated for operation in indoor, industrial
environments. The inspection and overview camera each
have their own display monitor and camera control panel.
VCR's are provided to record the video output signal of the
cameras.
As shown in Figure 2, the control console displays the
following equipment to the operator:
• Operator Control Panel
• 8-inch Color Monitor to display Overview Camera
• 9-inch Black & White Monitor to display B&W
Inspection Camera
• Two Super VHS Recorders
• Overview Camera Control Panel
(camera, zoom, pan, & lights)
• Inspection Camera Control Panel
(camera, zoom, pan & tilt, & lights)
• Control Panel (B&W Camera focus & iris)
• 20-inch Color (Video & Graphics) Monitor to display
inspection cameras or computer graphics
The control console also contains the following components
within its cabinet:
• Intelligent Controller Rack
• Servo Amplifier Rack
• Power Box
• Fan Panels
Figure 2. Operator Control Panel (Front View)
A. Operator Control Panel
The operator control panel provides the operator
with a complete interface to drive the RTI system. All
devices and accessories arc operated from the control panel
with the exception of the cameras which have independent
control panels. The operator control panel is wired directly
to the digital I/O boards of the controller. The controller
acknowledges operator commands by illuminating activated
switches. The controller performs safety checks of operator
commands before executing them.
The operator control panel provides switches for speed
selection and jogging of each individual axis. To prevent
accidental activation, each "Axis Jog" toggle switch must bo
held down continuously by the operator to jog the axis. The
axis will move at the selected speed (slow, medium or fast).
Once released, the toggle switch returns to its neutral "off"
position. In the event of a controller failure, the robot can be
driven in an open-loop mode by hooking up a battery power
supply directly to the motor amplifiers. An emergency stop
pushbutton is provided on the operator panel.
201
-------
The operator must depress a pushbutton to select
coordinated end point motion. A 4-position joystick is
provided to jog the end point of the arm towards or away
from the tank wall and clockwise or counterclockwise along
the tank wall. Consistent orientation of the end point is
maintained. In coordinated motion control, the Z-axis, wrist
roll and wrist pitch axis jog keys are also active. Depending
on the orientation of the robot arm, wrist roll and pitch
control the relative pan & tilt of the inspection camera
mounted at the end point.
Controls are also provided to open and close the gripper
and to control the decon spray ring and spray nozzle. The
operator control panel provides an up/down arrow and enter
key so the operator can make selections of menu commands
displayed on the graphics monitor.
B. Intelligent Controller
The design of the RTI controller is based on
RedZone's Intelligent Controller for Enhanced Telerobotics,
a proprietary, standardized platform for computation and
communications for the control of a wide variety of multi-
axis robotic systems. The Intelligent Controller is housed
inside a 12-slot VME bus chassis inside the control console.
The Intelligent Controller performs the following functions,
in a multi-tasking environment, for the RTI robotic system:
• Translation and execution of all operator commands
originating from the operator control panel.
• Digital servo control of all movement including
individual axis joint control and coordinated end point
motion of the robot arm.
• Execution of automatic routines; system self check,
power-up, sonar map control, and shut-down sequences.
• Safety monitoring of proximity sensors, joint overtravel,
joint and velocity tracking errors and overtorque
conditions.
• Continuous monitoring of potential collision states.
• Logging significant events in a data file.
• Displaying on the graphics monitor: plan view and side
view of robot arm inside tank, distance and orientation
of end point to wall, absolute position of each axis, error
message & diagnostics, and menu prompting of routines.
The computational devices of the RTI Intelligent Controller
consist of the following boards:
• 68020 CPU Boards (2) with floating point processors.
• RGB Video Board to interface the controller to the
graphics display monitor.
• Resolver to Digital Boards (2) to transform resolver
output to the digital signal used to compute current
position and velocity of each axis.
• Digital to Analog Board to convert the digital control
signal generated by the CPU to an analog control signal
to drive the motor amplifiers.
• Timer Interface Board to measure time-of-flight of
sonar echoes generated by the sonar ranging module.
• SCSI Interface Board to interface to the removable
cartridge disk drive.
• 44 MByte Removable Cartridge Disk Drive to provide
portability with hard disk performance. All software
resides on the disk drive.
• Digital Input Boards (2) to provide 64 opto-isolated
channels that are interrupt driven to the controller.
Digital I/O serves as primary interface between CPU
and operator control panel.
• Digital Output Board to provide 32 dry reed relay
outputs. Allows CPU to control devices and indicator
lights on each switch.
Control of robot motion is achieved by a control law
implemented in software on the main CPU. Motion control
boards are not required as servo control is flexibly
implemented in software. The CPU reads resolver inputs,
computes forward and inverse kinematics, and generates a
digital control signal. This digital control signal is then
converted into an analog input to the motor amplifiers. The
CPU performs all of the control calculations for robot motion,
interprets user commands from the operator control panel,
and maintains the graphics display. Two CPU boards allow
the computational load to be distributed by running the
motion planner on one board, and the remainder of the
software modules on the other. This results in stiffer motion
control and faster updating of the graphics display.
V. SOFTWARE
Under separate contract to the Department of Energy
(DOE), RedZone is developing an Intelligent Controller for
Enhanced Telerobotics to provide a standardized, multi-
tasking, VX Works ™ environment for software
development. The RTI system uses the hardware and
software architecture defined by the DOE Intelligent
Controller architecture. All software is written in the C-
language and resides on the disk drive. Figure 3 is a block
diagram of the major software modules of the system. The
software is organized into five main modules: the task
executive, the motion planner, the motion controller, the
data processor, and the graphics module. Communication
between (and in some cases within) these modules is
performed using RedZone's proprietary Robotic
Communications Protocol (RCP) which is the heart of the
Intelligent Controller. RCP provides both intra-cpu and
inter-cpu communications as well as global variables,
functions calls and semaphores between modules. Below,
each module is described in detail.
A. System Control
The system control module is the "front-end" of the
RTI controller. It contains four sub-modules: digital
input/output drivers, task executive, health monitor, and
data logger. The digital input and output drivers provide a
standardized software interface to the digital I/O boards.
The task executive's main function is to monitor the state of
the operator panel and of the robot. It directs action based
on these inputs. The data logger records events, errors, and
change of state into a file. The log is maintained on the
hard disk to help understand and troubleshoot failure or
accident scenarios.
202
-------
B. Motion Planner
The motion planner module provides a collection of
high level path generating modules, collision detection
modules, and kinematics utilities that operate with a
nominal cycle time of 10 milliseconds. The path generating
modules include joint space profile generation, cartesian
space profile generation, and control for sonar mapping.
Cartesian space points are transformed via inverse
kinematics into joint space goals to generate a smooth
trajectory for each joint in motion. The sonar map utility
automatically controls the arm while the sonar mapping
sequence is in progress. The collision avoidance module
monitors the proximity of the arm to the tank wall. The
kinematic module contains the mathematical model of the
arm, including link lengths and axes of rotations. Forward
kinematics are used to compute the end point position of the
arm based on axis joint positions for collision avoidance
checks. Inverse kinematics are used to compute the axis joint
positions necessary to achieve a desired end point position
for coordinated motion control.
Monito
153
A
Operator Console ^i,
f /jran
1
" ' -graphic-
driver
A
graphics
Graphics module
, f.
lOOHz
System am
module
T
1
—DID
rol drivi
1
I Task
'
f
-
r
r
n control & status bits
Executive L.
Monitor |
Data
I^RfilT
v ^-^ \ ^
collision
avoid.
T
Cart, space Joint space ^
profile profile
generation generation
t „
forward
kinematics
1KHz
inverse
kinematics
/
S S^ Motion planner
/ ^
Soft Limits
contro
^ module
1 Interpolation
^|r
j servo bw
1 •
•driver
J
Arm ^
il
-<
vsolvur
iriver -
Data pnxressor
. "
:>nar Map
Control
i •
sonar
m.i . >inK
t
1
i
r
Figures. Software Organization
1. Jog Control. Robot motion is initiated whenever
the operator holds down an axis jog toggle switch or the
coordinated motion joystick. The controller responds to the
switch transition state. An acceleration ramp is
immediately generated to ramp up to the preselected speed
range. The motion control module then generates new,
incrementally small, position goals for the joint every 10
milliseconds.
2. Coordinated End Point Motion. The operator's
primary objective is to position the robot's inspection video
camera relative to the tank surface. It is often difficult and
tedious to position the end-of-arm while jogging individual
axes. To facilitate easier positioning of the camera,
coordinated end point motion is provided in two axes while
maintaining a consistent orientation of the tool faceplate:
horizontal extension of the arm to the tank wall and
following the curvature of the wall at a constant distance.
Coordinated motion for the RTI robotic system is constrained
in the cylindrical world frame of the tank. Control is
simplified by requiring the arm to be in a preferred
orientation. Should the operator choose to deselect
coordinated motion and jog in joint mode, a resume function is
available to allow the operator to return to his former
position and resume coordinated motion.
3. Collision Avoidance. The collision avoidance
software consists of a real-time background program that
continuously checks the position of the arm to avoid a
collision with the tank. The computer checks for penetration
by the arm into a safety zone that extends from the tanks
walls and floor. If the robot enters the safety zone, the
computer executes an interrupt of the current motion and
warns the operator of the condition. Once the robot arm is in
the software collision state, the software only allows the
operator to jog arm motion away from the tank surface.
Proximity sensors are also provided to detect an impending
collision and initiate an emergency stop. A manual override
button is provided so the operator can override collision
avoidance so that the RTI can touch the tank wall or floor.
C. Motion Control
The motion control module reads the joint absolute
position from the resolver-to-digital driver every
millisecond. The servo law, an enhanced PID control, uses
commanded and actual position read from the resolvers to
calculate a command output to send the power amplifiers.
Robot motion is controlled in a position controlled mode, not
a rate controlled mode, as commonly used on robotic
manipulators. Position control provides stiffer motion
control with more damping. It also allows an easy upgrade
to programmed operation at a later date. Execution of the
motion control task is triggered by a clock interrupt to ensure
precise timing. The motion control module also enforces soft
stop limits and performs linear interpolation on the
commanded positions.
D. Sonar Data Processor
The sonar data processor module reads and
processes the sonar data to map distance to the tank wall as
a function of shoulder rotation. Radial extensions from the
RTI to the tank wall vary in length, since the RTI system is
inserted through a riser that is offset from the tank center.
The sonar sensor produces a digital pulse each time it is
203
-------
fired. The length of the pulse is proportional to the time
from transmission of the sonar signal to the return of the first
echo. The sonar driver measures this time-of-flight which
is converted into distance and recorded in an array with the
corresponding shoulder rotation angle. The sonar mapping
module performs pre-processing of the signal to remove
erroneous data and compensate for the wide beam width of
the sonar. Signal processing of the sonar signal is performed
to derive a circular model of the tank from the raw data.
E. Graphics Module
The graphics display on the large color monitor
provides the operator with a physical sense of the robot
arm's position inside the waste tank. Objects are portrayed
as two-dimensional diagrammatic models. A plan view
shows the orientation of the arm inside the tank and a side
elevation view shows the robot arm configuration to the
tank wall. The monitor displays robot joint angles, as well
as the distance and orientation of the end of the arm to the
tank. These views and information will greatly enhance the
operator's efficiency in operating the robot within the tank.
The graphics software module continuously reads the current
position of all axes and uses the kinematic model to compute
and display the configuration of the arm. The graphics
display module also provides menu commands, status
information, and messages to the operator.
VI. CONCLUSION
RedZone Robotics will deliver the RTI robotic system to
WINCO in April 1990. The RTI robotic system will then
become one of the first robotic systems deployed to remotely
inspect hazardous waste tanks. The initial mission of the
RTI will be remote visual inspection of corrosion inside the
ICPP waste tanks. WINCO is currently planning additional
development of the RTI robotic system including advanced
tooling to sample the sludge and inspect the bottom of the
tank, supervisory control to provide enhanced force control of
the tooling, and a programmed mode of operation.
The RTI robotic system provides a 15.9 Kg (35 Ib)
payload, 1.8 m (6 ft) reach, five degree of freedom robotic
arm that can be inserted through a 25 cm (10 in) diameter
opening. The vertical deployment unit provides 5.8 m (19 ft)
of servo controlled extension. The robotic arm can
manipulate a variety of tools: inspection viewing systems,
gripper, spray nozzle, or other specialized end of arm
tooling. The arm can be flexibly mounted on a variety of
platforms or even a mobile base. Its compact, high torque,
electric, servo-controlled actuators can be re-configured with
different linkages to customize a rcJbotic arm of any
configuration and degrees of freedom. The RTI robotic system
is radiation and environmentally hardened to assure
reliable operation in hazardous environments. The
Intelligent Controller provides a multi-tasking environment
to support digital servo control, I/O, collision avoidance,
sonar mapping, and a graphics display. The controller,
based on the standardized DOE architecture, is extensible to
servo control almost any multiple axis application. In
conclusion, the RTI robotic system and its components offer an
innovative, standardized, and extensible design with broad
applicability to remote inspection, decontamination,
servicing, and decommissioning tasks.
REFERENCES
Griebenow, Bret & Martinson, Lori, "Robotic System for
Remote Inspection of Underground Storage Tanks,"
Proceeding of 1990 American Nuclear Society Winter
Meeting. Washington D.C., Nov. 1990.
204
-------
AUTOMATED SUBSURFACE MAPPING
Jim Osborn
Field Robotics Center
Carnegie Mellon University
Pittsburgh, PA 15213
412-268-6553
Abstract
Non-invasive imaging of the underground is an essential
component of hazardous waste site investigations, yet,
despite advances in sensor technology, high quality maps of
the subsurface are difficult to obtain. Subsurface mapping
depends on the spatial correlation of individual sensor
measurements taken at multiple locations. Current manual
data collection techniques, however, are suboptimal for
precisely positioning subsurface imaging sensors and, in
general, are quite inefficient. Use of the sensors also requires
considerable experience on the operator's part to acquire and
interpret sensor data. In short, locating and identifying
buried objects and geological features is a process that relies
heavily on human adeptness and expertise. Thus by applying
automation and computer vision technologies to the
problem, subsurface mapping can be improved.
In our Site Investigation Robot (SIR) project, prototypical
robots are used to position ground penetrating radar (GPR)
equipment with the accuracy needed to generate three
dimensional subsurface maps. Estimating its site location by
a combination of dead reckoning and inertial measurements,
a rough terrain mobile robot deploys a gantry mechanism to
scan the ground with the GPR antenna. Radar data are
digitized and stored in three dimensional arrays for spatial
correlation and image enhancement on a color graphics
workstation. We have also applied basic image processing
and visualization techniques to assist in the interpretation of
these subsurface maps. Control of the robots and access to
the software are through user-friendly interfaces, which
facilitate the subsurface mapping process.
Introduction
For years, robotics and automation have increased
productivity in manufacturing industries through
standardization and repeatability. Core robotic technologies
have now progressed to the point that robots are moving into
the field and offering similar benefits performing tasks in
unstructured settings. One class of these field robots is
emerging to meet one of the most important challenges now
facing the world: the clean up of hazardous waste sites.
One of the cost drivers in remediation of a site is the lack of
information about the site itself. A detailed and costly
investigation is required to develop a knowledge base of site
geology, hydrology, chemistry, the extent of contamination,
etc., that can be used to select appropriate remediation
technologies and effectively plan the cleanup effort. Much of
this expense can be attributed to inefficiencies in manual
data acquisition techniques, lack of standard data collection
procedures, and the cost of insuring and protecting the
personnel who conduct the investigation. As an alternative,
automation offers the prospect to collect large quantities of
data in a form that supports more complete assessments and
at a significantly lower cost.
205
-------
Most investigations include efforts to locate buried objects
that are potential sources of contamination (such as drums),
identify and measure the extent of contaminant plumes, and
determine the morphology of geological formations that
affect pollutant migration. Commonly used methods to
generate such information include resistivity measurements,
acoustic techniques and ground penetrating radar. While
each has unique advantages, no single method alone
provides complete information, and all have limited utility
owing to the inaccuracies and inefficiencies of manual
sensor deployment. Ideally, the data resulting from the
application of these non-invasive techniques can be used to
construct an accurate graphical representation of the
geometry of buried structures - a map of the subsurface.
In this paper we present the Site Investigation Robot, a
system for automated subsurface mapping with ground
penetrating radar (GPR), as one aspect of a program to
automate hazardous waste site characterization. The Site
Investigation Robot is a mobile robot that collects and
spatially registers GPR data and recovers them to its base
station where they are correlated, enhanced and displayed so
that inferences about the shape and location of buried
structures can be made. This program's broader goal is to
develop robotic systems to make the data acquisition process
faster and more complete and to apply advanced data
processing techniques that will make these data more
accessible and easier to interpret.
System Overview
The Site Investigation Robot consists of a robot and
controller, data acquisition system, and a body of subsurface
mapping software to manage, process and visualize data
collected during investigation missions. The present
configurations of these subsystems are described below;
future enhancements planned for each are described in the
section that follows.
Robot
The Site Investigation Robot prototype is pictured in Rgure
1. We have employed an existing mobile robot, the
Terregator (short for terrestrial navigator), a driveriess,
outdoor vehicle built for autonomous driving and
exploration research, for the data acquisition aspect of this
project. Terregator is a rugged, six-wheel, skid-steer
locomotor scaled and powered to negotiate moderately
rough terrain and steep slopes.
On both the right and left sides of the base locomotor, three
wheels are linked together with chains and driven by a low-
speed, DC motor through a harmonic gear unit. This
drivetrain, in conjunction with off road floatation tires, gives
Terregator excellent tractive characteristics to overcome
obstacles and grades. For position feedback, each motor is
coupled to an incremental rotary encoder. Theoretically, this
arrangement gives the Terregator open loop positional
accuracy in the sub-millimeter range; in practice, tire
deflections, vehicle/ground surface interaction and other
non-linearities limit Terregator's dead-reckoning ability to
distances on the order of centimeters.
To position subsurface imaging sensors, a single-axis gantry
mechanism is attached to Terregator's frame forward of the
generator such that the direction of motion is perpendicular
to the mobile robot's path. The mechanism consists of a
buggy that is pulled along parallel fiberglass T-beams by a
chain belt driven by a DC motor. The GPR antennas are
suspended from the buggy with threaded rods for height
adjustment. A rotary encoder directly coupled to the motor
allows the antenna to be positioned accurately to one
centimeter over the entire two-meter length of the gantry.
Limit switches at each end of the gantry ensure safe
operation and provide a convenient way to identify the
antenna's limits of travel.
A 3kW, 120 VAC gasoline generator and ventilated, shock-
isolated electronics enclosure are mounted atop Terregator's
base to provide power for the locomotion, computation,
sensing and communications. Raw generator output is tied in
to the base locomotor's 90 VDC power supply; the generator
output is also conditioned by an uninteruptible power supply
(UPS) for more sensitive devices, including telemetry
equipment, onboard computers and disk drives, safety logic,
sensors and interface electronics. Substantial auxiliary
power is available for mission specific payloads, such as
GPR equipment.
206
-------
At the heart of the Terregator is a VMEbus card cage that
bouses a 68020 CPU card with 4 Mbyte onboard memory,
SCSI and ethernet ports. The system CPU functions as a
multi-tasking controller, coordinating and sequencing
locomotion and gantry motions, GPR data acquisition,
communications with the base station and system
monitoring functions. Other boards in the card cage include
a serial interface card, two 2-axis motion control cards, and
a sensor interface card with eight channels of analog-to-
digital (A/D) conversion, four channels of digital-to-analog
(D/A) conversion and 16 bits of digital I/O. All connections
to these boards are made through an intermediate patch panel
that facilitates the addition of new sensors and other
peripherals to the basic system. For development purposes, a
single board workstation and disk are located on the
equipment deck above the electronics enclosure and
interfaced to Terregator's CPU via an ethernet cable. The
organization of these components is shown graphically in
Rgure 2.
Controller
The Site Investigation Robot is intended for use by persons
who are much better versed in the practices of field
screening, data collection and analysis than they are in
operating a robot. It is thus essential to hide the complexities
of controlling the robot from its users and make interactions
with the SIR as simple and straightforward as possible. This
motivated us to develop a control architecture that allows
SIR users to command and monitor the robot at a high level
while masking the details of implementing expressed user
intentions.
The SIR command interface presents the user with a set of 2-
Dsurface maps of the site, that show the size, spatial location
and orientation of boundaries, known man-made structures
(e.g., buildings and roads) and natural features (e.g., trees
and surface water bodies) in a consistent, user defined site
coordinate system. These maps are created with a simple
CAD package, developed specifically for this purpose, at the
outset of a site investigation, and can be updated and edited
as the investigation proceeds. To initiate a data acquisition
ran, the user first displays a map of the site on the base
station computer by recalling a file that contains a CAD
description of a particular region of interest. Site boundaries
are indicated by straight line segments while all known
objects and other obstacles to the mobile robot are shown as
polygons. Using the computer mouse, the user then draws a
bounding box (a rectangle that encloses part of the map)
around the area of the site from which data is to be collected.
A set of routines to plan a path that covers all of the obstacle-
free ground surface within the bounding box are then
invoked. First, the dimensions of the bounding box and all
obstacles it contains are adjusted using dimensional
parameters of the SIR. In this algorithm, the robot's effective
turning radius is calculated by finding a circle within which
all parts of the skid steered locomotor will remain when it
turns in place. All sides of the bounding box and all included
polygonal obstacles are 'grown' by an amount equal to the
radius of that circle. Should the transformed bounding box
be found to intersect a site boundary, which is a pathological
case for the current path planner, the initial bounding box is
rejected arid the user instructed to redraw it. Once an
acceptable bounding box is found, the robot can be modelled
as a single point travelling through a more constricted space,
thus simplifying subsequent path planning.
Planning paths for the Site Investigation Robot is a departure
from traditional mobile robot path planning in the objective
is to cover as much of the ground surface as possible, rather
than finding the shortest route between two points. The SIR
path planning problem is constrained by the mobility
characteristics of the Terregator mobile robot. Terregator
can faithfully execute straight line motions of specified
length by dead reckoning, in which the wheel encoders are
used to measure distance travelled; it can also make accurate
turns in place, using a gyroscope to measure the angle of
rotation. However, the indeterminacy of Terregator's skid
steering makes following an arc of specified curvature
difficult even on hard, flat surfaces. For this reason, we have
limited all driving to straight line motions and point turns.
This is acceptable given the data acquisition protocol
described below.
SIR's path planner examines the resulting free space in the
transformed bounding box and finds a way to cover it such
207
-------
that the number of turns are minimized. If obstacles are
present the user selected area is divided into smaller
obstacle-free areas, and a path is planned for each. Since
there are often multiple ways to perform the subdivision,
solutions are not always unique. Furthermore, there is no
way to guarantee that the resulting path is optimal. However,
once a path is found, it is overlaid on the site map for
validation. This affords the user the opportunity to draw
smaller bounding boxes and specify point-to-point moves
that connect the subregions of the map.
The final path description is translated into a sequence of
driving commands (straight lines and rotations) that are
placed in a queue and transmitted to the robot via a wireless
modem. Using a software joystick, the robot is then
leleoperated to its starting point and set on its route. While
driving, the robot transmits its location back to the base
station which is displayed as an icon on the site map. Other
status information is similarly relayed so that the user can
supervise the data acquisition mission.
Subsurface Mapping Software
The Site Investigation Robot deploys and supports a
commercial ground penetrating radar set (Geophysical
Survey System, Inc. SBR.-3) to acquire subsurface data. A
data acquisition run is comprised of combinations of
Terregator drive motions and gantry movements in which
the basic procedure is to move the antenna from one limit to
the other and then drive forward some incremental distance.
At regular intervals through the antenna's travel, a series of
radar pulses are transmitted into the ground and the energy
reflected to the receiving antenna amplified, filtered and
digitized. These signals are stored adjacently in a buffer until
the antenna has completed a full scan. The result is a two
dimensional data array, in which the columns are individual
GPR waveforms, stored on disk as an image along with the
mobile robot's site coordinates. More details on the
principles of GPR are presented in the Appendix.
Every row of pixels in the GPR image contains data acquired
at a constant time delay relative to the transmitted pulse. That
time delay is converted into a distance from the antenna by
the speed of electromagnetic wave propagation in the
imaged subsurface media based on measured and/or inferred
electrical parameters. Since the position of the mobile robot
and the position of the antenna relative to the mobile robot
are measured for every recorded GPR waveform, it is
possible to assign three spatial coordinates to each pixel in
the image. It is this position tagging that makes it possible to
spatially correlate and visualize GPR data in three
dimensions.
Each recorded waveform spans a depth range that is
governed by the wavelength of the transmitted energy and
the electrical properties of the subsurface medium. Generally
speaking, there is a trade-off in depth of penetration and the
physical dimensions that can be resolved. The 500 MHz
antenna used in this work can image structures buried to
depths of 3 meters with 5-10 cm resolution in the best of
conditions (e.g. dry, sandy soils); lower frequencies
penetrate deeper at the sacrifice of resolution. GPR
performance is poorer in materials with high conductivity
and high dielectric constant - conditions associated with high
moisture content - due to attenuation of the radar energy. In
saturated soils and clays, imaging potential may be limited to
depths of only one meter,
This data acquisition procedure is repeated until the robot
has covered its entire planned route. Once the robot returns
to its base station, all acquired images are transferred from its
onboard disk to mass storage devices connected to the base
station computer for archiving and processing. Acquired
GPR data are arranged in volumes, each containing a set of
parallel subsurface sections stored as images. Individual
sections are stored as files that also contain other parameters,
including location of the scan, date and time of acquisition,
and radar gain and time base settings. These files are
organized in a Unix file system such that each subdirectory
corresponds to a unique site volume. Each subdirectory also
contains an additional site index file that is used to retrieve
and store individual images. Figure 3 shows an example of a
site map from which nine volumes of the subsurface would
be scanned.
Since the intuition of experienced field screening personnel
is still required to apply the appropriate processing steps and
208
-------
choose parameter values to make sense of the images, we
have developed a set of programs to process GPR data
acquired by the Site Investigation Robot that are called by
the user through a common menu-driven interface. This
software package, known as gpr-shell, includes routines for
reading and writing data files, applying time domain filters
to individual records, displaying of 2D subsurface sections
as color or gray-scale images, scaling and windowing
images, spatial correlation all GPR records in a subsurface
volume, and a variety of image enhancement functions. To
facilitate processing, Gpr-shell also provides command line
completion, prompting, and on-line help. It also provides the
user with an 'on-line lab notebook', in which the steps and
parameters used to process each image are automatically
recorded for future reference.
In order to transform raw GPR data scans into high
resolution images, several processing steps have been
implemented, as illustrated hi Rgure 4. (We have yet to
identify a single methodology or set of parameters that can
be successfully employed to generate interpretable
subsurface maps from all GPR data, however, the following
steps are generally taken.) First the signal is deconvolved
with the return signal from a pulse transmitted into air.
Deconvolution is a matched filter operation that removes the
effects of the secondary pulses from the return signal and
effectively transforms a return from the transmitted pulse
into the return that would have been caused by an ideal
impulse function. The resulting signal is then low pass
filtered to remove noise components introduced by the
deconvolution.
The waveform recorded at each grid point is actually a
composite of all radar reflections within the antenna's
conical beam pattern due to the poor focusing of the GPR
antenna. However, since the spacing between surface grid
points is accurately measured, we are able to correlate all of
the measurements and synthetically focus the antenna. A
process known as 'migration' is applied to convert the
deconvolved and filtered data into a representation of the
subsurface. Migration is very similar to the synthetic
aperture focusing techniques used for high resolution pipe
location, in that its underlying principle is data from adjacent
scans tend to reinforce one another.
A three dimensional array of GPR data is constructed by
sampling data from vertical sections in the scanned volume.
The value in each cell, or voxel (for volume element), is then
added to all array locations equidistant from the transmitter
and within the antenna beam. This effectively 'spreads' each
part of the return signal over surface that is a locus of points
with the same time of flight from the antenna. By applying
this algorithm cell in the array, the recorded signals
originally associated with individual voxels constructively
interfere with one another. This reinforcement indicates the
presence of an impedance discontinuity at the corresponding
subsurface location and emerges in the migrated image.
Migration can thus be used to effectively focus the
transmitted radar beam. (We note, however, that its success
requires a good estimate of the soil's dielectric constant,
which determines the speed at which GPR waves travel
through the subsurface, and the antenna beam pattern and
soil conductivity, both of which influence attenuation.)
Once a volume of data has undergone 3-D migration, vertical
and horizontal sections are extracted from it as individual
images. These images are then enhanced by a number of
image processing operations, including 2D low- and high-
pass filters of varied bandwidths, edge detectors and region
growing operators, depending on the image features of
interest.
Figure 5 through 7 show the results of these processing steps.
All three are images of a small metallic drum containing
water buried in sand. Figure 5 is a vertical section of raw data
and Figure 6 is the same image after deconvolution and
migration. In this case, the barrel cross section is best seen by
the thresholding of the image after it is finally processed by
the 2-D high pass filter (Figure 7).
Future Enhancements
A number of enhancements to our current system are
planned to increase its ability to operate on waste sites, ease
its use, and improve the quality of the subsurface maps it
generates.
209
-------
For sites with very rough terrain and/or numerous obstacles,
improving the mobility of the base locomotor will result in a
greater percentage of ground surface that SIR can cover.
This can be accomplished with suspension, greater ground
clearance, replacing the wheels with tracks, etc. An even
more significant performance increase can be realized by
improving SIR's position cognizance, regardless of its
mobility characteristics. The most promising technologies to
provide a more accurate measurement of the robot's location
on the site are inertial navigation units (INS) and global
positioning (GPS) receivers, both of which can be deployed
onboard and readily interfaced to the robot controller. By
providing a position estimate that is independent of the
robot's dead reckoning, the robot can be navigated with a
closed loop path tracking control scheme, a paradigm in
which the robot's actual (measured) position is used to
correct for deviations from the planned path that may result
from wheel slippage or other controller disturbances. Path
tracking control using combined INS and GPS has
successfully guided our NavLab mobile robot at speeds
exceeding 20 km/hr, more recently, the same controller has
been ported to an off-road dump truck.
More accurate GPR antenna positioning can also be
achieved by replacing the gantry mechanism with a tnulti-
degree-of-freedom robot arm. Our concept for such a sensor
deployment arm (SDA) is a long reach mechanism able to
position and orient sensor payloads weighing up to 10 kg.
over a 2 meter x 2 meter area, adjusting to any undulations
of the terrain. The principal advantage of an SDA is greater
integrity of the sensor position measurements - complete a
3D data array can be collected with a common frame of
reference, eliminating the possibility of positioning errors
between adjacent scans due to motions of the mobile base,
which arc typically an order of magnitude less accurate than
manipulator movements.
There appears to be a synergy between the Site Investigation
Robot and geographical information systems (GIS), another
emerging technology for waste site investigations.
Geographical information systems are software tools for
cataloging; manipulating and displaying any form of data
that can be related to a cartographic map. GIS applications
include land use management, record keeping of legal
boundaries, roads and utility networks, agriculture, and
many others. A GIS can be also linked to a relational data
base to provide a powerful tool for site investigation. Many
available GIS packages include routines to enter previously
digitized terrain maps and survey data which would aid in
the development of site maps for the SIR user interface. The
other attractive feature of GIS is simplified storage and
retrieval of data: entry of acquired position tagged data into
the GIS data base can be automated and its recall reduced to
the simple positioning of a cursor in the display window.
Two advances in subsurface mapping software are currently
being pursued. One is the development of more general three
dimensional migration algorithms that will account for the
non-homogeneous nature of the subsurface nature of the
subsurface medium. This will entail assigning permittivity
and conductivity values to each voxel in the scanned
subsurface volume in order to better model GPR wave
propagation. Techniques to measure and/or infer these
parameters will have to be developed to make the best use of
this algorithm. In addition faster processing engines and
techniques will be required to achieve results in useful time
frames. The second advancement will be the application of
three dimensional enhancement and rendering techniques to
subsurface maps. Such techniques exist in the domains of
medical imaging and geological exploration, but have yet to
be adopted for GPR.
Finally, our goal is to integrate these hardware and software
elements into the more complete system for waste site
characterization, as shown in Figure 8.
Summary
Subsurface mapping is a discipline that has advantageously
adopted technologies from the domains of robotics and
computer science. In this research, we have successfully
implemented registration of sensor position and automated
acquisition of sensor data using a robot, and thereby created
opportunities to apply processing techniques to create 2-D
and 3-D subsurface maps of higher quality than previously
attainable. This and other spatially correlated information
that the Site Investigation Robot generates can be used to
210
-------
more effectively characterize waste sites and ultimately
lower the expense of site cleanups.
More generally, robotics and automation can benefit waste
site characterization in a number of ways.
• The enormous data requirements will be satisfied
faster and at lower cost when data arc acquired by
robots.
• The quality of those data will be enhanced
through standardized, repeatable measurement
techniques.
• By automatically indexing measurements by
position in a geographical information system,
opportunities for numerical modeling, graphical
visualization and straightforward data correlation
are created.
The Site Investigation Robot is an example of an emerging
class of robots dedicated to the solution of hazardous waste
problems. We view the SIR be the first in a family of robots
for environmental applications. Systems that follow will
have additional perceptive capabilities and self-reliance to
perform detailed site assessments.
Acknowledgments
This research is sponsored through a cooperative agreement
with the U.S. Environmental Protection Agency and a grant
from the Ben Franklin Technology Center of Western
Pennsylvania. We also acknowledge RedZone Robotics,
Inc., for its participation in the Site Investigation Robot
project.
Appendix: Principles of GPR Sensing
Ground penetrating radar works by transmitting an
electromagnetic pulse into the earth which spreads as a
conical wavefront as it travels further from the antenna.
When the radar wave reaches a discontinuity in electrical
impedance of the subsurface, an echo is relumed, the
strength and the phase of which indicate the magnitude and
sign of the change. Mathematical descriptions of these
interactions in all but the simplest of cases defy closed form
solutions; even finite element methods are too cumbersome
for practical modeling of the GPR phenomenon. Fortunately,
modeling the physics using geometrical optics can produce
meaningful results. With this simplification, the transmitting
antenna is treated as a light source from which rays emanate
and are reflected to the receiving antenna. The distance to the
point of reflection (assuming a direct reflection) can thus be
estimated with time-of-flight measurements, i.e., the latency
of the echo relative to.tbe transmitted pulse.
A difficulty with the optical assumption is the poorly
focused radar beam. Commercially available GPR antennas
are designed to limit beam spread of the transmitted wave to
an elliptical cone, however, for a single return, the exact
location of an echo within this volume cannot be determined.
To resolve this ambiguity, the antenna is scanned in a line
over the ground surface to create an ensemble of return
signals. Latency of echoes are lowest when the antenna is
directly over an object and increase as the antenna moves
away. By combining recorded echoes from points along the
scan line, distinctive curves are generated which are then
interpretedby GPR experts to identify subsurface features.
In practice, there are several factors that complicate the radar
return. Time of flight measurements on return echoes depend
on knowledge of the propagation velocity of the transmitted
pulse, which is not a constant but instead depends on
electrical permittivity (or equivalently, dielectric constant)
of the subsurface material. This introduces uncertainty in the
measurements, which is currently resolved either by
calibration in the field or simply by estimation of subsurface
permittivity. Both the transmitted and reflected radar waves
are attenuated due to losses in the media that are governed
primarily by its conductivity, another parameter requiring
estimation. Geometric dilution of the wave energy as the
beam spreads with distance travelled is a further
complication since the exact shape of the antenna beam
pattern within the subsurface medium cannot be determined.
Finally, the difficulties of controlling the shape of the
transmitted pulse at GPR operating frequencies (one
hundred megahertz to over one gigahertz) introduce
additional return signals that confuse the main return echo
and must be removed.
211
-------
Figure 1. Site Investigation Robot prototype
VME Bus Computer
•:
Tirn^-z!
Js.
•n~^on
\ \ 4UBP | —
Optical Disk
1
i_
o
_
g
6
e
o
1
-------
PARKING LOT SUE MAP
ildjZ
test pa
test pa
test pa
ch 3
:ctiZ
;ch 1
leslpa
test pa
test pa
ch4
\
ch5
I
:ch6
1
test pa
est pa
le§tpa
ch 9
ch 8
:ch 7
Pzxking Cuib
Parking Cuib
figure 3. Site map with nine scanned subsurface volumes
Data Acquisition
Rgure 4. Ground penetrating radar processing steps
213
-------
Time
(IIS)
0.0
1.0
2.0
3.0
4.0
5.0
:,.'•
7.0
BjO
JjQ
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
19.0
Depth
(mm)
0.0
74.9
149.9
224.8
299.8
374.7
449.7
524.6
599.6
674.5
749.5
824.4
899.4
9743'
1049.3'
1124.2
1199.2'
1274.1'
1349.1'
1424.0'
< \iliu-M;ip: -I.W.I i
Volume: icsl volume I Slice Niimher: 5 (400 mm)
Antenna: 3102 I-rcq: 500 MH/. Pcnn: 4.00 Time Slcp: 50 ps.
X (mm)
0 149 299 449 599 749 899 10-19 1199 1349 I49S I64K I79S 194S
I I I I I I I I I I I I I
Figure 5. Vertical section of burled drum (raw GPR data)
Tune Ucplh
UK! (nun)
0.0
74.9
149.9
224.8
299.8
374.7
449.7
524.6
599.6
674.5
749.5
24.4
12.0 899.4
13.0 974.3
14.0 1049.3
15.0 1124.2
16.0 1199.2
17.0 1274.1
18.0 1349.1
19.0 1424.0
Figure 6
:
:
i
;
•
7.0
-
::
Volume: Icsl volume 1 Slice Number r> (4()()mm)
Aiuaina: 3H)2 l-'req: 500 MH/. Perm: 4.00 Time Step: SO ps.
X (mm)
0 149 299 449 599 749 899 1049 1199 1349 1498 1648 1798 1948
I I I I I I I I I I I I
-
Image of buried drum (Figure 5) after deconvolution and migration
214
-------
, i 7;i IS.'I
Volume: lesl volume 1 Slice Number: 5 (400mm)
Antenna: 3102 Freq: 500 MHx, Perm: 4.00 Time Step: 50 ps.
Figure 7. Drum image from Figure 6 following 2-D high pass filter and thresholding
215
-------
CITE IK'
Motion planner
Motion controller
Site posrtion
Data acquisition
Self monitoring
SIR System Architecture
: .v-fsa
Display site map
Display acquired data regions
Select acquired data
Overlay scalar data arrays
Show robot posrtion and path
Select robot path
ooooooooooo
OODOOOOOOQO
CO OOOOOO O O O
O O O L ^Q o D
Position and time tagged :
Sensor specific data
Site maps
Data annotation
Processing results
*B-*s
— -
KAal FILE SYSTEM
Variable length GPR data
Processing history
Processing results
VHSU&US&TOfi
2D Radar display
3D GPR 'stacks'
Solid visualization
Depth map displays
•:7'- ._••..-: .
Deconvolution
2D migration
3D migration
Depth mapping
Parameter Estimation
Bd&QE PHQCESSIH3
Thresholding
Segmentation
Frequency analysis
Rltering
Object tagging
Figure 8. Site Investigation Robot system architecture
DISCUSSION
BRIAN PIERCE: My first question has to do with using ground penetrating
radar, as just one example, or using a magnetometer as another type of sensing
device. And the second question has to do with the use of a pair of robots or a team
where you could take advantage of forward scattering using the ground penetrating
radar. Right now it seems to me you're just using back scattering in a monostatic
configuration.
JAMES OSBORN: That's certainly correct. If you recall the viewgraph that
Ann put up. they are actually going to pursue the magnetonietry type of sensing.
In fact, there is really a whole class of sensors that can be put on it. Each one has
unique requirements. In particular, some of the magnetic techniques can't be near
these very metallic robots. So you've got to come up with long deployment
booms. The idea of doing bistatic radar soundings is an interesting one. I can
think of a couple ways that do that. One is to have a multiple arm system on a
single mobile base. And the other is to actually go w ith two mobile bases. I would.
at this time, say the preferred way would be the former (two arms) because of the
ability to register a manipulator and/or affect your position with much higher
accuracy than you could a mobile robot.
CHRISTOPHER FROMME: There are some excellent available technologies
for registering line of sight over short ranges, like the distance between the two
of us right now. So the idea of a pair of robots working in unison and precision
may have some merit.
DOUGLAS LEMON: Is this technology resident in the university or is it in the
RedZone Robotics Company, and who has funded this?
CHRISTOPHER FROMME: The project isfunded by EPA.And the technology
is currently in the university, although we have had some collaboration from
RedZone. in particular to turn out that robot controller that drives the system. So
we are getting some collaboration from RedZone. but the project is resident at
CMU.
DOUGLAS LEMON: Do you expect this technology to eventually be
commercially available? Is that where you're headed?
CHRISTOPHER FROMME: Yes. If not, then it doesn't make any sense to do
it.
216
-------
A QUALITY ASSURANCE SAMPLING PLAN FOR EMERGENCY
RESPONSE (QASPER)
John M. Mateo, Quality Assurance Officer
and Christine M. Andreas, Assistant
Quality Assurance Officer, Roy F. Weston,
Inc.^REAC, GSA Raritan Depot, 2890
Woodbridge Avenue, Building 209 Annex,
Edison, NJ, 08837-3679
William Coakley, Quality Assurance
Coordinator, USEPA, Environmental
Response Team, GSA Raritan Depot,
2890 Woodbridge Avenue, Building 18,
Edison, NJ, 08837
Abstract
Integration of critical elements into a com-
prehensive Quality Assurance Sampling
Plan (QASP) is crucial to implementation
of an effective plan. How can a project
manager ensure consideration of all these
elements? Utilizing a software package
called QASPER, a project manager is
prompted to consider elements necessary
to generate a comprehensive Quality As-
surance Sampling Plan for Emergency
Response.
QASPER is a PC-based software package
which compiles generic text and user
provided, site-specific information into a
draft QA/QC Sampling Plan for the
Removal Program. QASPER addresses
the nine sections of a QA/QC Sampling
Plan, as specified in OSWER Directive
9360.4-01, Removal Program QA/QC In-
terim Guidance, Sampling QA/QC Plan,
and Data Validation Procedures (revised
April, 1990). Sections include: Initial data,
background information, data use objec-
tives, QA objectives, approach and sam-
pling methodologies, project organiza-
tion and responsibilities, QA
requirements, deliverables, and data
validation.
QASPER was created to facilitate the
timely assembly of a comprehensive sam-
pling plan for emergency response ac-
tions. By thorough consideration and
attention to the necessary requirements
of QA/QC sample planning through an
automated process, it is anticipated that
reliable, accurate and quality data can be
generated to meet the intended use.
The On-Scene Coordinators (OSC) or
the Technical Assistance Team (TAT)
contractors are the primary users of
QASPER. These individuals will have
access to the site specific information and
the sampling objectives which charac-
terize a particular hazardous waste site
investigation. They are also responsible
for assembling the information into an
acceptable plan for implementation.
217
-------
The system, however, is applicable to many
regulatory programs that require the com-
pletion of QASPs.
Features of QASPER are numerous.
QASPER is self contained, no other
software is required for support. ASCII
outputs are generated so that files may be
uploaded to other word processing pack-
ages for further manipulation. Database
files on all previous sampling plans are
retained. Consistency and comprehensive-
ness of sampling plan creation efforts are
maintained throughout office, region or
zone, therefore, sampling plans are created
more efficiently. Redundant data entry is
minimized by integrating repetitive infor-
mation throughout the plan after one entry.
The user is provided access to standardized
generic text with the capability to overwrite
and edit. QASPER allows for flexible data
entry throughout the plan. QASPER runs
on an IBM PC or 100% compatible, with a
hard drive, 640K RAM and a printer (for
hardcopy output).
Introduction
The U.S. Environmental Protection Agen-
cy (EPA) has divided the Superfund
cleanup program into short-term and long-
term remedial activities. Short-term inves-
tigative and mitigative efforts, typically
addressing imminent threat, are referred to
as "Emergency Response Actions" under
EPA's Removal Program. To ensure ade-
quate and comprehensive response, suffi-
cient time must be allocated for thorough
planning; however, planning is often
regarded as a luxury in an emergency
response scenario.
The EPA has taken a number of steps to
establish planning criteria for emergency
response actions which are sufficiently
detailed to ensure that data generated will
be of known quality to serve its intended
purpose and are commensurate with the
emergency response timeframes. The
first of these steps was the establishment
of data quality objectives (DQOs) for the
Removal Program. Second, EPA also
established a minimum framework for an
acceptable Quality Assurance Sampling
Plan. Both of these guidelines are set
forth in OSWER Directive 9360.4-01
released April 1990 (Publication No.
EPA/540G-90/004).
This paper will describe the Removal
Program DQO's, define the framework
of the QASP, and describe a third, in-
novative step EPA has taken in creating
a software package which facilitates the
timely assembly of both into a com-
prehensive plan ready for implementa-
tion in an emergency response. The
majority of this paper will describe the
features of the software program.
Removal Program Data Quality Objec-
tives
The quality of data is determined by its
accuracy and precision against
prescribed requirements or specifica-
tions, and by its usefulness hi assisting the
user to make a decision or answer a ques-
tion with confidence. OSWER Directive
9360.4-01 guides the user in defining data
quality within a framework that also in-
corporates the intended use of the data.
The guidance is structured around three
quality assurance objectives, each as-
sociated with a list of minimum require-
ments. The three QA Objectives,
hereafter referred to as QA1, QA2 and
QA3 are described as follows:
QA1 is a screening objective to afford a
quick, preliminary assessment of site
contamination. This objective for data
218
-------
quality is for data collection activities that
involve rapid, non-rigorous methods of
analysis and quality assurance. These
methods are used to make quick, prelimi-
nary assessments of types and levels of pol-
lutants. The primary purpose for this
objective is to allow for the collection of the
greatest amount of data with the least ex-
penditure of time and money. The user
should be aware that data collected for this
objective have neither definitive identifica-
tion of pollutants nor definitive quantita-
tion of their concentration level.
QA2 is a verification objective used to
verify analytical (field or lab) results. A
minimum of 10% verification of results is
required. This objective for data quality is
for data collection activities that require
qualitative and/or quantitative verification
of a "select portion of sample findings"
(10% or more) that were acquired using
non-rigorous methods of analysis and
quality assurance. This quality objective is
intended to give the decision-maker a level
of confidence for a select portion of
preliminary data. This objective allows the
user to focus on specific pollutants and
specific levels of concentration quickly, by
using field screening methods and verifying
at least 10% by more rigorous analytical
methods and quality assurance. The results
of the 10% of substantiated data gives an
associated sense of confidence for the
remaining 90%. However, QA2 is not
limited to only verifying screened data. The
QA2 objective is also applicable to data that
are generated by any method which satisfies
all the QA2 requirements, and thereby in-
corporates any one or a combination of the
three verification requirements.
QA3 is a definitive objective used to assess
the accuracy of the concentration level as
well as the identity of the analyte(s) of in-
terest. This objective for data quality is
available for data collection activities
that require a high degree of qualitative
and quantitative accuracy of all findings
using rigorous methods of analysis and
quality assurance for "critical samples"
(i.e., those samples for which the data are
considered essential in making a
decision). Only those methods that are
analyte specific can be used for this
quality objective. Error determinations
are made for all analytes of the critical
sample(s) of interest.
Quality Assurance Sampling Plan
Framework
There are nine sections to a Removal
Program QA Sampling Plan. Section 0.0
addresses basic information require-
ments such as site name, relevant work
order numbers, primary personnel
names and titles, etc. Section 1.0 solicits
information about the location of the
facility, type of facility, type and volume
of materials to be addressed, sensitive
adjacent environments, and action
levels. Section 2.0 addresses data quality
objectives (DQOs), i.e., regarding
decisions the data will support. Section
3.0 addresses the linkage of DQOs with
matrix and parameters. The project
manager must decide which parameter
will be assessed, by matrix, for which in-
tended data use, at which QA objective
(QA1, QA2, or QA3). Section 4.0 ad-
dressed the Sampling Approach and
Methodologies, including documenta-
tion requirements. This section will in-
clude a discussion of sampling design,
type of equipment, fabrication and
whether equipment decontamination
will be employed, standard operating
procedures, numbers of field samples
and control samples needed to achieve
the stated QA Objectives. It also in-
cludes a timetable for sampling activities.
219
-------
Section 5.0 addresses information about
what personnel are assigned which respon-
sibilities, and which laboratories will be
analyzing which samples. Section 6.0 dis-
cusses the requirements necessary to
achieve the quality assurance objectives
identified in Section 3.0. Section 7.0 ad-
dresses the types of deliverables to be
produced and what they will contain. Sec-
tion 8.0 addresses the degree of data valida-
tion necessary to achieve the identified QA
Objective.
for
Quality Assurance Sampling Plan
Emergency Response (OASPER)
QASPER is a PC-based software package
which compiles generic text and user
provided, site-specific information into a
draft QA/QC Sampling Plan for the
Removal Program. QASPER addresses
the nine sections of a QA/QC Sampling
Plan, as specified in OSWER Directive
9360.4-01, Removal Program QA/QC In-
terim Guidance, Sampling QA/QC Plan,
and Data Validation Procedures.
The site manager (On-Scene Coordinator)
or contractors are the primary anticipated
users of QASPER.
These individuals will have access to the site
specific information and the sampling ob-
jectives which characterize the site inves-
tigation. It is their responsibility to
assemble that information into an accept-
able sampling plan for implementation.
QASPER has a database of standard
generic text which is utilized in an
electronic "cut and paste" process with user
provided site specific information to create
a draft QA Sampling Plan. This approach
enables the user to focus on critical infor-
mation while the software manages both
the presentation and correlation of that
information with other essential data.
Perhaps the best way to illustrate this
process is to "walk through" QASPER.
The user should progress in a sequential
manner, starting with section 0.0 because
the plan database will build on previously
provided information. This feature
avoids the need for redundant input of
data which must appear in several sec-
tions of the completed plan. It is possible
to skip sections, or avoid certain input
requirements (e.g., when information re-
quested is not yet known to the user).
This allows the user to create those por-
tions of the database at times that are
convenient to the user. However, it may
not be possible to complete certain sec-
tions (most notably the DQO sections:
3.0, 6.0, and 8.0) without providing cer-
tain information in preceding sections
(e.g. Section 2.0).
Figure 1. Main Edit Plan Menu
Section 0.0 identifies certain information
required to complete the title page of a
Sampling Plan; some of information will
also be utilized elsewhere throughout
the completed plan. If the user chooses
not to enter the information requested,
220
-------
the completed plan (through the Output
menu) will be assimilated as if that informa-
tion was not requested. Should the user
wish to add alternate information currently
not requested by QASPER, this would be
accommodated through the Edit menu
after the plan has been compiled from the
database (through the Output menu).
Section 1.0 solicits background information
about the site. The user is first prompted to
geographically locate the site, characterize
its size and operating status, i.e., operation-
al or abandoned. The user is requested to
provide information about the type of
facility. (This information request is cur-
rently limited to one response per
category). For sites with multiple facility
types, the user may enter this data through
the Edit Text menu after the file has been
compiled. Next, the user is requested to
provide information about the materials
handled, the surrounding environs and
populations. Responses to these requests
are facilitated by pop-up menus of standard
responses. In the last three parts of this
section, the user provides the information
requested by typing onto free-form test
screens. Although there is room for multi-
ple page responses under each information
request, one to several paragraphs should
be sufficient.
Section 2.0 requests information regarding
the objective and purpose of the sampling
event. How does the user expect to utilize
the resultant data? Several standard
responses are provided and may be ac-
cessed by the arrow keys and or selected by
the "Return" key. The user may input an
alternate "objective" or "purpose" by select-
ing the "Other" category and specifying the
other use. The return key is utilized to
mark or unmark each item. A critical con-
sideration for any data collection event is
whether the data will be evaluated against
an existing database or action level.
Specification of the contaminants of con-
cern and their respective actionable
levels will help determine appropriate
analytical methods and quality assurance
needs later in the plan. Multiple selec-
tions are permissible from the screen.
Selections under the "Purpose" group
will be carried forward to other sections
of the plan. This section, therefore, re-
quires input in order to enable the user
to complete portions of Sections 3.0,4.0,
6.0, and 8.0.
Figure 2.
Menu
Section 3.0 QA Objectives
In Section 3.0, the user will select among
various parameters to identify the class
of compounds to be investigated. This
parameter selection will initiate the
DQO logic for a parameter, in a matrix
(next menu), for a given purpose (sub-
sequent menu), at a selected Quality As-
surance Objective (subsequent menu).
At the end of the logic path, the user will
be brought back to the parameter menu
to make another selection, if ap-
propriate. QASPER remembers the last
logic path, therefore if the user wishes to
select the same parameter, same matrix,
same purpose, and a different QA Objec-
221
-------
live, he/she need only move the highlight on
the last option.
Section 4.0 of the system solicits informa-
tion about the proposed sampling rationale
and how sampling will be conducted. There
are five subsections which address the fol-
lowing:
1. Sample Equipment
The user is requested to identify sampling
equipment that will be utilized, what
material it is made of (fabrication), and
whether it is to be dedicated and/or decon-
taminated. The user must identify the sam-
pling tools which will be used to collect
samples from the various matrices. This
process is initiated by first selecting a matrix
from among those previously identified in
Section 3.0. Next, the user will identify the
type(s) of equipment to be used in the
various matrices selected. The emphasis
here is on the equipment which will be
utilized to obtain the sample from the en-
vironment and transfer it to the sample con-
tainer. Most of the equipment in the menu
has a corresponding Standard Operating
Procedure (SOP) available in Subsection
4.3.
Figure 3. Sampling Equipment Decon-
tamination Sequence Menu
The user is also requested to identify the
equipment fabrication, or material of
construction. This is important so that
the quality of the sample is not com-
promised, inadvertently, by the materials
it comes in contact with during sample
collection. This is usually critical for low
concentration investigations, or situa-
tions of incompatibility between sample
contaminants and sampling device
fabrication. If the equipment is not dedi-
cated, QASPER will import generic text
describing decontamination procedures
and solicit additional information about
the user's preference for the decon-
tamination sequence and chemicals (e.g.
solvents) of choice. The user will high-
light, or select, the decontamination
steps from a menu in the order he/she
wishes the sequence to be conducted in
the field. A manifestation of that se-
quence will be compiled in the plan out-
put.
2. Sampling Design
In this section, the user will indicate the
sampling design or grid proposed to
achieve the sampling event objective. It
is expected that the user will detail where
and how many samples will be collected.
A basis for the sampling scheme would
be described herein, and a sampling map
would be referenced. QASPER will
print a blank page with the name of the
site and the title, "Sampling Location
Map", for incorporation of this map.
3. Standard Operating Procedures
There are three sections to the SOP sub-
section, addressing standard text for
Sample Documentation, Sampling, and
Sample Handling and Shipment.
QASPER allows the user to choose exist-
ing generic text from the database, or
222
-------
write new text to describe how sample
documentation will be achieved. If the user
selects "Write own Text", a free form edit
screen of several pages will appear to
receive the user's narrative.
Figure 4. Available SOPs Menu
QASPER enables the user to choose from
an inventory of standardized SOP texts to
prepare a description of how the sampling
event will be conducted. There are several
approaches for incorporating Sampling
SOPs:
-The user may import only the titles of
SOPs into the compiled plan. This reduces
the bulk of the final plan document and may
be appropriate where all users of the plan
would have access to a repository of the
actual SOP texts.
-The user may import title and text into the
compiled plan. This allows the final plan to
be a "stand alone" document.
-The user may import any portion of the
generic titles and text available through
QASPER and/or modify and add SOPs to
the QASPER database.
4. Schedule of Activities
The user is requested to provide a
timetable for the sampling activities.
This usually begins with the procurement
process for laboratory services and may
end with delivery of the final report. A
tabular presentation will be created
when the plan is compiled.
5. Tables
QASPER presents a summary table of
each parameter, matrix, purpose, and
QA objective as compiled in Section 3.0.
The user will select by means of the high-
light bar and "return" key to initiate a
method selection for each parameter,
identification of level of sensitivity, num-
ber of samples to be collected and QC
samples needed to address the relevant
QA objective. This information will be
assimilated by QASPER into Field Sum-
mary and QA/QC Summary Tables.
FigureS. Field QA/QC Summary Tables
Menu
In Section 5.0, the user is requested to
identify what personnel will be perform-
ing what tasks or responsibilities for the
sampling event. Likewise, the user is re-
223
-------
quested to provide the name of the lab and
a city or state descriptor for an address.
Labs will be characterized as either CLP,
commercial, EPA or field under the space
for lab type. Parameters may be identified
by class of compound.
Section 6.0 of the plan database receives
standardized text regarding QA require-
ments, based on the QA Objectives
selected in Section 3.0. The user has the
opportunity to view and edit the text in
Section 6.0, since this is where the informa-
tion will appear in the final compiled plan.
There are also options for deleting generic
text or writing unique text (requirements).
The menu will indicate which QA Objective
requirements are being imported (e.g.
QA1, QA2, and/or QA3).
In Section 7.0, QASPER contains an inven-
tory of standardized descriptions of the
types of deliverables which may be
prepared under a sampling event. The user
need only select the appropriate
deliverables, and the resultant plan will
contain the appropriate text.
Figure 6. Deliverables Menu
Section 8.0 contains the requirements for
validating the data generated under the
plan. The text in this section will be auto-
matically imported at the time the QA
Objective(s) is selected.
After completing review and/or
modification of Sections 0.0-8.0, the user
may proceed to the output menu to com-
pile the plan for eventual printing or
sending to diskette.
Features of OASPER
If contained, requires no other software
for support
-Generates ASCII outputs - file and
hardcopy. Files may be uploaded to
other word processing packages for fur-
ther manipulation
-Creates (draft) hard copy QA/QC Sam-
pling Plan document ready for approval
signatures and implementation
-Retains database files on all previous
sampling plans for future manipulation
(e.g. recreating documents, searching for
similar sampling plans by location,
facility type, contamination, etc.)
-Capable of transmitting (compiled)
sampling plan or database via diskette or
modem
-Improves consistency and comprehen-
siveness of sampling plan creation efforts
throughout office, region, or zone
-Improves efficiency for creating and
reviewing QA/QC Sampling Plan docu-
ments
-Repetitive use of information
throughout the plan without the need for
redundant data entry
224
-------
-Provides the user access to standardized
generic text with overwrite capability for
editing
-Flexible data entry throughout
Requirements
QASPER runs on an IBM PC or 100%
compatible, with a hard drive, 640KRAM
and a printer (for hardcopy output).
Conclusion
QASPER is a PC-based software package
which compiles generic text and user
provided, site-specific information into a
draft QA/QC Sampling Plan for the EPA
Removal Program. It is envisioned that this
tool will primarily facilitate the timely as-
sembly of comprehensive QA Sampling
Plans in emergency response scenarios and,
indirectly, educate users on the correlation
of data quality objectives and sampling ac-
tivities.
Mention of trade names or commercial
products does not constitute EPA endorse-
ment or recommendation for use.
References
U.S. Environmental Protection Agency,
Quality Assurance/Quality Control
Guidance for Removal Activities, Sam-
pling QA/QC Plan and Data Validation
Procedures, Interim Final EPA/540G-
90/004, April 1990.
225
-------
A RATIONALE FOR THE ASSESSMENT OF ERRORS IN SOIL SAMPLING
J. Jeffrey van Ee*
Exposure Assessment Division
Environmental Monitoring Systems
Laboratory
Las Vegas, Nevada 89193
*0irect questions to this author.
Clare L. Gerlach
Lockheed Engineering & Sciences
Company
Las Vegas, Nevada 89103
ABSTRACT
Considerable guidance has been provided on
the importance of quality assurance (QA),
quality control (QC), and quality assessment
procedures for determining and minimizing
errors in environmental studies. QA/QC
terms, such as quality assurance project
plans and program plans are becoming a part
of the vocabulary for remedial project
managers (RPMs). Establishment of data
quality objectives (DQOs) early in the
process of a site investigation has been
stressed in EPA QA/QC guidance documents.
Quality assessment practices, such as the
use of duplicates, splits, spikes, and
reference samples, are becoming widely
accepted as important means for assessing
errors in measurement processes. Despite
the existence of various forms of guidance
for hazardous waste site investigations,
there have been no clear, concise, well-
defined strategies for precisely how these
recommended QA/QC materials can be utilized.
The purpose of this paper is to familiarize
field scientists with an approach to these
questions:
How many and what type of samples are
required to assess the quality of data
in a field sampling effort?
How can the information from these
quality assessment samples be used to
identify and control sources of error
and uncertainties in the measurement
process?
The primary audience for this paper is
assumed to be RPMs who have concerns about
the quality of the data being collected at
Superfund sites but have little time to
investigate the complexities of the
processes used to assess the quality of
data from the total measurement process.
The approach outlined in this document for
assessing errors in the field sampling of
inorganics in soils may be transferrable,
with modification, to other contaminants in
other media.
This presentation is a summary of "A
Rationale for the Assessment of Errors in
the Sampling of Soils" by J. Jeffrey van
Ee, Louis J. Blume, and Thomas H. Starks,
1990.
An in-depth treatment of the statistical
approach is outlined in the Rationale (1),
and it is recommended reading.
INTRODUCTION
This document expands upon the guidance for
quality control samples for field sampling
as contained in Appendix C of EPA's Data
Quality Objectives for Remedial Response
Activities - Development Process (2). That
report outlines, in greater detail,
strategies for how errors may be assessed
and minimized in the sampling of soils with
emphasis on inorganic contaminants.
Basic guidance for soil sampling QA, which
includes a discussion of basic principles,
may be found in EPA's Soil Sampling Dualitv
Assurance Users Guide developed at the
Environmental Monitoring Systems
227
-------
Laboratory, Las Vegas (3). The Users Guide
is intended to be revised on a periodic
basis. It is anticipated that some of the
guidance provided in this document will
eventually be incorporated into the Users
Guide.
The sampling and analysis of soils for
inorganic contaminants is a complex
procedure from experimental design to the
final evaluation of all generated data.
Sources of error abound but they can be
successfully mitigated by careful planning
or isolated by intelligent error assessment.
Error (or variability) can be either bias or
random. Biased error is indicative of a
systematic problem that can exist in any
sector of soils analysis, from sampling to
data analysis. The first step in analysis
of variability (or error) is to establish a
plan that will identify errors, trace them
to the step in which they occurred, and
account for variabilities to allow direct
action to correct them. In anticipation of
errors, it is essential to ask two
questions:
1. How many and what type samples are
required to assess the quality of data
in a field sampling effort?
2. How can the information from these
samples be used to identify and
control sources of error and
uncertainty in the measurement?
Error assessment should be understood by the
field scientist and the analyst. To aid
scientists in the estimation and evaluation
of variability, the Environmental Monitoring
Systems Laboratory-Las Vegas (EMSL-LV) has
developed a computer program called ASSESS.
ASSESS can trace errors to their sources and
help scientists plan future studies that
avoid the pitfalls of the past.
BACKGROUND
Superfund and RCRA site investigations are
complicated by: the variety of media being
investigated, an assortment of methods, the
diversity of investigators, the variety of
contaminants, and the numerous risks to and
effects on human health and the environment.
Many phases exist in Superfund site
investigations. An initial phase, generally
described as a preliminary investigation,
consists of collecting and reviewing
existing data and data from limited
measurements using practically any
available method. The next phase,
generally described as site
characterization, uses selected methods and
prescribed procedures to characterize the
magnitude and extent of the contamination.
Later phases include an examination of
remedial actions, which involve an
assessment of treatment technologies, and
continued monitoring to assess the degree
of cleanup at a site. A final phase may
require long-term monitoring to
substantiate that no new or additional
threats occur to affect human health and
the environment. Throughout Superfund site
investigations QA/QC procedures change as
data quality objectives vary and different
phases proceed.
RANDOM ERRORS
Random errors can result in variations from
the true value that are either positive or
negative but do not follow a pattern of
variability. During the measurement
process, random errors may be caused by
variations in:
1) sampling
2) handling
3) transportation
4) preparation
5) subsampling
6) analytical procedures
7) data handling
The greatest source of error is usually the
sampling step. In the Comprehensive
Environmental Response, Compensation, and
Liability Act of 1980 (Superfund, or
CERCLA) and the Resource Conservation and
Recovery Act (RCRA), site investigations,
analytical, and data handling variability
are checked by the CLP protocol. When more
than one laboratory is involved, handling,
transportation, subsampling, and
preparation can be checked at Level IV.
All analyses are performed in an offsite
Contract Laboratory Program (CLP)
analytical laboratory following CLP
protocols.
228
-------
But how can the analyst know that the sample
in the jar is representative of the
surrounding samples at the site? How can
the field analyst know that the more (or
less) contaminated soil didn't stick to the
auger or split-spoon?
It is strongly recommended that the
traditional approaches used in mitigating
the error in the last six steps be applied
to sampling itself, i.e., use of duplicates,
splits, spikes, evaluation samples, and
calibration standards. A certain amount of
random error is inherent in samples
themselves. In fact, the total variance
equals the measurement variances plus the
population variances, as defined by the
equations:
where at = total variability
am = measurement variablity
a = population variability
and
and
where as = sampling variablity
(standard deviation)
Qh = handling, transportation
preparation variability
0SS= preparation variability
(subsampling variability)
laboratory analytical variability
between batch variability
aa =
a =
NOTE: It is assumed that the data are
normally distributed or that a
normalizing data transformation has
been performed.
We can address the variance in measurement;
the population variance, however, is a true
picture of the complexity of the soil.
BIAS ERROR
Some sources of error are systematic, that
is, in a given situation conditions exist
that consistently give positive or
consistently give negative results. This
skewing of data can be introduced early in
a sampling regime, e.g., by a sampling
device that alters the composition of the
soil matrix. It can occur in the middle of
the sampling regime, e.g., by the
preferential handling of a sampler who
isn't trained in the intricacies of sample
handling and preparation. Or it can be
introduced in the later, analytical stages,
where it is easier to trace because of
interlaboratory comparisons and frequent
calibration checks. The pervasive quality
of an early bias error is its resistance to
detection and the fact that other
variabilities are added throughout the
process until, finally, the reported data
may be significantly non-representative of
the true value. Bias errors can be traced
to:
faulty sampling design
skewed sampling procedure
systematic operator error
contamination
degradation
interaction with containers
displacement of phase (or chemical
equilibria)
inaccurate instrument calibration
PREVENTION
To avoid both random and bias errors (or at
least to be able to pinpoint their
occurrence and estimate their extent), it
is wise to plan a study well, anticipating
possible sources of error. The inclusion
of quality assurance samples used for
quality assessment and quality control can
help isolate variability and identify its
effect.
An effective technique is to concentrate
duplicate sampling early in the study and
send the samples off for rapid CLP
analysis. Dependent on the results, it may
not be necessary to include as many quality
assessment samples after these samples
demonstrate reliability in the sampling
process. Early detection of sources of
error can help the field scientist
customize the remainder of the study to
meet the specific needs of the project.
QUALITY ASSESSMENT SAMPLES
A Remedial Project Manager (RPM) must ask:
how many samples are needed to adequately
characterize the soil at this site? The
229
-------
key word is "adequately." By determining
the data quality objectives (DQOs) in
advance, the RPM can assure adequate
sampling at a site. Too little sampling, as
well as too much, is a waste of time and
money. The extent of QA/QC effort is
dependent on the risk to human health, the
nearness of action levels to detection
limits, and the size, variability, and
distribution of contamination. Ultimately,
the number of quality assessment samples is
determined by the DQO for the site. Table
1 explains various types of quality
assessment samples and their uses.
SOME STATISTICAL CONCERNS
Confidence in quality assessment sample data
can be expressed as an interval or as an
upper limit. All confidence levels/limits
are based on the number of degrees of
freedom and the limits get lower (or the
intervals get smaller) as the number of
degrees of freedom increases. For example,
if 15 samples are taken at a site, and each
split is extracted twice at a CLP
laboratory, and 2 injections of each
extraction are made into an Inductively
Coupled Plasma/Mass Spectrometer (ICP/MS),
the total number of degrees of freedom
associated with this experimental design
would be calculated as:
15 samples X 2 preparations splits = 30
X 2 CLP extractions = 60
X 2 injection replicates = 120
120 degrees of freedom for the whole
process. But, if only the population
variability in the field samples (which
includes the sampling error) is being
estimated, the number of degrees of freedom
is 15-1, or 14. There are 15 independent
samples but one degree of freedom is lost
with the estimation of the mean. Therefore,
there are 14 degrees of freedom for the
sampling variance estimate. As another
example, to estimate the variability in the
extraction step, one has 30 independent
pairs of numbers, each pair associated with
one extraction. Thus, there are 30 degrees
of freedom associated with the extraction
error.
Obviously, the confidence associated with
any particular sampling is directly related
to the number of samples taken. In Table 2
(also Table 3 of the Rationale Document) or
in a statistics manual, guidance is given
for the number of quality assessment
samples that must be used with the routine
site characterization samples. These
tables assume that data are normally
distributed. The tables will show the user
the confidence interval associated with the
degrees of freedom. Then, decisions may be
based upon the requirements of the DQOs. A
synopsis of this targeted approach can be
seen in Figure 1. The total measurement
error is comprised of error in the sampling
(a), subsampling (ass), handling (ah),
batch (ab), and analysis (aa) steps. Each
is addressed in the regime depicted in
Figure 1.
SAMPLE COLLECTION CONSIDERATIONS
If Level IV CLP analysis is performed on
the soil, we can assume that very little
error occurs in the analytical stage. This
focuses our attention on sources of error
in the sampling, handling, and preparation
steps. The two major considerations in
collection of environmental samples are:
1. Will the collected data give the answers
necessary for a correct assessment of
the contamination or a solution to the
problem?
2. Can sufficient sampling be done well and
within reasonable cost and time limits?
ASSESS
The EMSL-LV has developed an easy-to-use
program to calculate the necessary
statistics, as described in the Rationale
(1), from the generated data for an
accurate determination of precision and
bias. ASSESS is a public domain, FORTRAN
program that is available from EMSL-LV and
written for personal computers. It may be
applied
to cases where no field evaluation samples
are available as well as cases where they
are. ASSESS is user-friendly and its use
will greatly aid both field scientists and
RPMs in decision-making based on soil
studies.
230
-------
TABLE 1
QUALITY ASSESSMENT SAMPLES AND THEIR USES
• ALLOW STATEMENTS TO BE MADE CONCERNING THE QUALITY OF THE MEASUREMENT SYSTEM
• ALLOW FOR CONTROL OF DATA QUALITY TO MEET ORIGINAL DQOs
• SHOULD BE DOUBLE-BLIND:
Types of Samples Description
Field Evaluation
(FES)
Low Level Field
Evaluation (LLFES)
External Laboratory
Evaluation (ELES)
Low Level External
Laboratory (LLELES)
Field Matrix Spikes
(FMS)
Field Duplicates
(FD)
Preparation Splits
(PS)
• SHOULD BE SINGLE-BLIND:
Field Rinsate
Blanks (FRB)
Preparation Rinsate
Blank (PRB)
Trip Blanks (TB)
Samples of known concentration are introduced in the field
as early as possible to check for measurement bias and to
estimate precision
Low concentration FES samples check for contamination in
sampling, transport, analysis, detection limit
Similar to FES but without exposure in the field, ELES can
measure laboratory bias and, if used in duplicate, precision
Similar to LLFES but without field exposure, LLELES can
determine the method detection limit, and presence of laboratory
contamination
Routine samples spiked with the analytes of interest in the
field check recovery and reproducibility over batches
Second samples taken near routine samples check for
variability at all steps except batch
Subsample splits are made after homogenization and are used
to estimate error occurring in the subsampling and analytical
steps of the process
Samples obtained by rinsing the decontaminated sampling
equipment with deionized water to check for contamination
Samples obtained by rinsing the Blanks sample preparation
apparatus with deionized water to check for contamination
Used for Volatile Organic Compounds (VOC), containers filled
with American Society for Testing and Materials Type II water
are kept with routine samples through the sampling, shipment,
and analysis phases
• MAY BE NON-BLIND: AS IN THE INORGANIC CLP PROTOCOL
231
-------
TABLE 2
Some 95 Percent Confidence Intervals for Variances
Degrees of Freedom Confidence Interval
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
30
40
50
100
0.27s2
0.32s2
0.36sJ
0.39s2
0.42s2
0.44s2
0.46s2
0.47s2
0.49s2
0.50s2
0.52s2
0.53s2
0.54s2
0.54s1
0.56s2
0.56s2
0.57s2
0.58s2
0.58s2
0.59s2
0.60s2
0.60s2
0.61s2
0.62s2
0.64s2
0.67s2
0.70s2
0.77s2
< a2
< a2
< a'
< o2
< a2
< a2
< o2
< a1
< o2
< a2
< a2
< a2
< a2
< o1
< a2
< a2
< a2
< a2
< a2
< a1
< a2
< a2
< a2
< a2
< a2
< a1
< a2
< a1
< 39.21s2
< 13.89s1
< 8.26s2
< 6.02s2
< 4.84s2
< 4.14s2
< 3.67s2
< 3.33s2
< 3.08s2
< 2.88s2
< 2.73s2
< 2.59s2
< 2.49s2
< 2.40s1
< 2.32s2
< 2.25s2
< 2.19s2
< 2.13s2
< 2.08s2
< 2.04s2
< 2.00s2
< 1.97s2
< 1.94s'
< 1.91s'
< 1.78s'
< 1.64s2
< 1.61s2
< 1.35s2
232
-------
Figure 1.
QUALITY ASSESSMENT SAMPLES
DUPLICATES AND SPLITS
EVALUATION SAMPLES
BLANKS
SAMPLE TAKING
PREPARATION
ANALYSIS
SOURCES OF ERROR
III!
| ROUTINE | | FIELD | , , , , , —
1 QAMPI F 1 InilPtTTATFl 1 FFS 1 I FFC 1 1 F
1 1 1
_ 1 I . I „„ 1 „ ....__.....
1 1 j |
I I II
1 1 1 II
III II
| ROUTINE | JPREP. SPLIT] 1 FD 1 1 FES j 1 FES |
| SAMPLE |-| SUBSAMPLE | | SUBSAMPLE | 1 — , — 1 ' — , — 1 NO PREPARATION
i it ii i
1 1 1 IELESI |ELES|
»]
r-1
| PRS |
1
1
1
III II
| RS | I PS | 1 FD | | FES | | FES | ' |ELES| |ELES| | FRB | | PRB |
? oss.Oa a, ah. ass. aa os.ob.oh aa oa.ok.os oa.oh
ACKNOWLEDGEMENT
This work is based on the in-depth treatise,
"A Rationale For the Assessment of Errors in
the Sampling of Soils" by J. Jeffrey van Ee,
Louis Blume, and Thomas Starks.
NOTICE
Although research described in this article
has been funded wholly by the United States
Environmental Protection Agency under
contract number 68-03-3249 to Lockheed
Engineering & Sciences Company, it has not
been subjected to Agency review and
therefore does not necessarily reflect the
views of the Agency, and no official
endorsement should be inferred. Mention of
trade names or commercial products does not
constitute Agency Endorsement of the
product.
REFERENCES
(1) van Ee, J.J., L.J. Blume, T.H. Starks,
A Rationale For the Assessment of
Errors in the Sampling of Soils, U.S.
EPA, 1990, 600/4-90/013.
(2) U.S. EPA. 1987. Data Quality
Objectives for Remedial Response
Activities - Development Process.
EPA/540/6-87/003.
(3) U.S. EPA. 1989. Soil Sampling Quality
Assurance Users Guide (2nd. Edition).
Environmental Monitoring Systems
Laboratory, Las Vegas, Nevada. EPA
600/8-89/046.
233
-------
DISCUSSION
REX RYAN: You did an admiral job of explaining the strategy of breaking down
what we call a "nugget effect" by using ANOVA techniques. I was a little bit
shocked that you didn't discuss the amount of variance distance contributes
within a sampling program. I was also surprised that you didn't discuss variograms
or any of those kind of issues that would affect a sampling team's success in
determining what is in fact going on at a site.
JEFFREY VAN EE: The two methods go together. The method I've described
is useful in pinpointing sources of variability in the measurement process if you
want to make changes. But the points that you're making address the larger
question of where your samples are located and whether they're going to be
representative of the site, assuming that the measurement variability is relati v'ely
low. That certainly needs to be looked at how representative are your sampling
locations to the contamination throughout the site.
REX RYAN: In your experience which do you think is larger—which in fact
could—in your professional judgment be a larger contribution to total variabil-
ity: the problem of extending samples in distance or trying to replicate samples
at the same location?
JEFFREY VAN EE: I don't think I have enough data to answer that question.
I can pose a few questions for all of you to consider. Let's say that we're sampling
volatile organics or a contaminant that varies with depth. This approach would
be useful in determining whether the sampling of that contaminant is being done
well. If you take a field duplicate sample and you go down, say, four inches and
your contamination is in the first two inches of the surface, then this method will
allow you to see that kind of variability from how the samples actually collected.
This method w ould also allow you to look at the loss of volatile organics. By the
time the samples get to the lab, it's more difficult with volatile organics and we
need to do some more research to see if this approach is applicable. But those are
some of the questions that can be answered by using this approach.
Both methods have been used together—at a site in Region VII. and they both
yielded very useful information. The CEO Statistical Approach again, looks at
the question of how many samples you need to collect to characterize a site and
then our method looks at whether those samples are being collected properly,
handled properly, those kinds of questions.
NABILYACOUB: I have aquestion about a statement you made about a second
sample collected at about an inch and a half and two inches from the original
which relates directly to this concern. Would this be a measure of the effect of
sample handling, the performance of the laboratory, containers, etc.? I beg to
differ because we are introducing here a variable that might bias the results.
Would you consider this sample as a split sample? If not, would you consider a
split sample more representative of the effect of these operations rather than this
end?
JEFFREY VAN EE: You need to use a combination of samples together. We are
assuming, (although we can disprove it,) that the spatial variability in those two
inches is insignificant. We can disprove it by the introduction of other samples
throughout the process. Once we collect a field duplicate, we could split that
sample and then analyze it separately to get a handle on errors down the line:
handling in the subsampling of the core or analytical errors. If we do come back
with this analysis and see that we do indeed have tremendous differences in
moving two inches away and we compare that to the GEO Statistical Approach
then we've got some real problems in characterizing that site.
A lot really depends how the contaminant was distributed at the site. If the
contaminant was uniformly distributed at a site, then I would expect the spatial
variability to be low. If we have leaking drums, we might just happen to hit on
that area, and if we move two inches over we would get a dramatically different
result. But the more samples we collect, the more field duplicates we collect.
presumably we will get a more representative idea of where the variability is. If
we were to rely on just one field duplicate or a few. then we would really be prone
to some of the misjudgments that you're alluding to.
ROY KAY: As I understand it, the objective of sampling and population
comparison within samples, is to provide a cost effective means of reducing the
total sampling costs while maintaining a high level of accuracy. Am I correct
there so far?
JEFFREY VAN EE: Yes.
ROY KAY: Has there been any cost evaluation information developed on the
relative cost of going through the process of designing and multiple batching
your samples versus simply expanding randomly the samples that you take—
particularly if you're starting from an nonhistorical, time-zero point of view?
JEFFREY VAN EE: I think a lot depends on the objectives that you establish
for that site. You need to look at the economics of collecting more samples, what
type of samples, versus the kind of action that you're going to be taking. It you
know that you're going to be cleaning up the site in large pan then taking a lot
of samples may not be appropriate.
But if the cost of that clean-up is significant, if the cost of disposing of the
contaminant is significant, then you will want to pay more attention to how
accurately you can characterize the site. And then, of course, you want to know
whether the data that you're getting represents the site or whether it is more
representative of variabilities in the measurement process.
I'm not sure I really answered your question well. It's a difficult question to
answer, because it varies depending upon the site.
ROY KAY: I'm looking at a situation where in a time-zero, first evaluation of
a site, there are certain theoretical things that you had assumed, like if you have
an explosion of some kind, it would naturally be expected to disperse contami-
nants. Whereas a leaking drum would expect to leach in a continuous fashion and
probably in all geometric dimensions. That is. of course, is a seat-of^the-punts
guess in each individual case. But lacking historical experience on that particular
site, do the sampling techniques dial in on the proper variables and reduction of
their influence faster than simply expanding the sampling population?
JEFFREY VAN EE: In a situation like that I would weigh more QA samples.
as well as more samples, period, early on in the process. You can hopefully back
off as you learn more about the site. Now that's assuming you don't have
historical information on how well that particular contractor performs out in the
field, or how well that particular sampling method performs.
Let me demonstrate very quickly another value that comes out of this process.
Say you're out sampling the site and you're concerned about the change of the
contaminant overtime, you may have different labs involved, and you may have
different sampling crews involved. If you do not have a rigorous QA program
instituted, then when the data comes back out of the lab. it's very difficult for you
to say whether that data reflect the pollutant changing over time or whether it's
your measurement process changing over a period of time. So, at some point.
you've got to pay your dues and you've got to start developing that data. We have
a tremendous amount of data right now on how well the contract labs perform.
but we don't have enough data on how well those samples are transported to the
lab and how well they're prepared. Say there's a rainfall event during your
sampling study, how do you know that the data you collect after that significant
event is comparable to the data before that event?
J ANINE ARVIZU: Could you describe some of the programmatic applications
of the program and whether or not there were any good real world experiences
learned?
JEFFREY VAN EE: The philosophy I'm espousing today is relatively simple
and it's relatively new. My hope is that more people will pick up on it whether
they're in RCRA or Superfund Programs. I think we really do need to demon-
strate where the variability is throughout the measurement process. Right now
I'm simply advocating that we try it. How well it's used remains to be seen. We
have applied it to a Superfund site in the middle part of the country and we looked
at the spatial variabilities. As a result of our efforts using GEO Statistics, we
saved about 6 million dollars in the sampling effort at this particular site. We w ere
able to demonstrate that the sampling method that they were using, while it was
crude, was sufficient to meet data quality objectives. We were able to tell them
that they could back off on a number of samples that they're taking in certain
areas, because the measurement variability was relatively low. They weren't
getting a lot of variability in the compositing of the samples. We have had a few
success stories, but not nearly enough. We can just hope with time there will be
more stories like that.
234
-------
A REVIEW OF EXISTING SOIL QUALITY ASSURANCE MATERIALS
Kaveh Zarrabi, Chemist
Amy Cross-Smiecinski, Quality Assurance Officer
Thomas Starks, Senior Statistician
Environmental Research Center
University of Nevada, Las Vegas
4505 S. Maryland Parkway
Las Vegas, Nevada 89154
ABSTRACT
Assessment of the quality of environmental data
often depends on the availability of quality
assurance (QA) materials to measure errors at
various stages of the measurement process. A
rigorous approach has been developed to
evaluate the quality of data from the sampling
of metals in soils. "A Rationale for the
Assessment of Errors in the Sampling of Soils"
was written for application to hazardous waste
site investigations. The rationale described
is based primarily upon duplicate and split
samples and QA materials known as performance
evaluation materials. The rationale depends,
in varying degrees, on performance evaluation
materials being readily available for use in a
hazardous waste site investigation.
Unfortunately, early experiences in testing the
rationale indicate that inadequate numbers,
types, and volumes of performance evaluation
materials and other types of soil QA materials
exist to fully implement the rationale.
In order to begin to answer questions as to the
necessity of, and alternatives to, soil QA
materials, it is necessary to know the current
availability and the state of research and
development of soil QA materials. The intent
of this paper is to provide such information -
what materials are available and what is being
done to provide more materials.
INTRODUCTION
SCOPE
Millions of dollars are spent in designing and
implementing monitoring and remediation programs
for hazardous waste sites. It is the Agency's
responsibility to ensure that the data resulting
from these programs are of adequate qua!i ty to be
defensible in a court of law as well as to be
considered scientifically sound.
Quality assurance (QA) materials are an important
part of many environmental sampling and analysis
programs today. Results from the analyses of
hazardous waste site samples are often accepted
or rejected solely on the basis of data obtained
from QA samples analyzed for Agency programs
ranging from water quality monitoring to
hazardous waste remediation. It is alarming that
only a 1 imited supply of these QA materials is
available for soil sampling and analysis (Table
1). What does a project manager do when no QA
materials exist? It is the intent of this report
to discuss the need for soil QA materials in many
environmental programs'1'3 and to demonstrate the
limited availability of these materials. An
alternative to the use of manufactured QA
materials is briefly described as are approaches
for increasing the supply and variety of the most
commonly needed soil QA materials. This report
does not purport to have the answer to the
scarcity of soil QA materials, but simply to
point out the problem and explore some solutions
with the hope that more attention will be given
to the issue.
RESEARCH
Research in the area of QA materials has been
limited. In fact, the bulk of the information
gathered for this report came from catalogs,
personal communications, and internal reports.
The following examples were obtained through a
literature search. Recently, Taylor[3] published
a comprehensive book, Quality Assurance of
Chemical Measurements. The book discusses the
basic concepts of quality assurance and provides
details on evaluation samples, traceability, and
235
-------
reference materials. Sewardt41 of the National
Institute of Standards and Technology (NIST),
formerly the National Bureau of Standards
(NBS), published a book which contains 25
papers describing national and international
programs for the development of reference
materials. The selection criteria, use of
statistics, and steps for certification of
standard reference materials are discussed.
Reports of 15 panel sessions reviewing the use
of and needs for reference materials are
included.
Calit5] of NIST, in another NBS monograph,
examines the general use of standard reference
materials and their role in the measurement
system. Further, procedures for certification
of standard reference materials are discussed,
and examples of several selected industries are
given in which standard reference materials
have made a significant contribution. Steger
compiled the information on all of the
available certified reference materials through
the Canadian Certified Reference Material
Project. Taylor published a handbook for
standard reference material users. The
preparation and analysis of reference materials
has been discussed and documented by several
programs. C8'9'10>11«12'13] In other studies, the
design and stability of reference
materials have been evaluated.
Another search of "Chemical Abstracts" from the
year 1979 to the present resulted in just five
more references. Studies in which the QA
materials were used range from proficiency
samples discerning between immunoinhibition and
electrophoretic measurement to soil and
geological reference materials.
SOIL QA MATERIALS
DEFINITIONS
The uses of QA materials have been predefined
for the purposes of this paper in the EPA
report referenced in the abstract: "A
Rationale for the Assessment of Errors in the
Sampling of Soil."m Briefly summarized, there
are two basic uses of QA materials: quality
assessment or evaluation (QAS) and quality
control (QC). QAS samples are intended to aid
in evaluating data quality and can be used in
QC. QC samples are used specifically on a
real-time basis to detect and correct problems
before a large body of erroneous or out-of-
control data is generated. The main difference
between the two uses becomes evident when the
data generated from them is interpreted. QAS
data are usually analyzed at the end of
studies, whereas QC data is analyzed as it is
generated; hence, the quality is "controlled."
QAS and QC samples exist in several types such as
reference materials and performance evaluation
materi al s. Reference materi al s are defi ned as
having "one or more properties which are
sufficiently well established to be used for the
calibration of an apparatus, for
the assessment of a measurement method, or for
assigning values to materials."" Reference
materials are typically used as QC samples but
can be used as QAS samples. Originally, soil QA
materials began existence as reference materials
and are slowly evolving as important components
of QA programs.
Performance evaluation materialst2'171 often are
associated with an analytical program in which
participants submit results to a central
authority who "grades" the data either in
comparison to the pooled results of all of the
participants or against a "referee" laboratory in
order to judge the overall performance or
accuracy of the laboratory. Performance
evaluation materials are, therefore, examples of
QAS samples.
Whether the data is used on a real-time basis
(QC) or at the end of a study (QAS), the overall
effect of a QA sample is to evaluate measurement
system performance. The sample may be used to
evaluate a whole system, from sampling through
data validation, or a part of the system; such as
extraction efficiency.
An important issue for soil sampling and
analytical QA is how closely soil QA samples
represent the routine samples of interest. A QA
sample should be similar to the routine samples
for the analytical parameter in order for a true
correlation to exist between the two. Analytes
spiked onto potter's clay or sand probably do not
accurately mimic environmental samples visually
or analytically and, therefore, test only the
recoverability of the analytes from the clay or
sand in combination with the competence of the
analysts. In the chemical analysis of natural
soil samples, it is especially important that a
QA sample be of a similar soil type as that of
the samples being analyzed to eliminate the
effects of various matrices effects on analytical
measurements and final results.
This paper deals with three basic types of QA
soil samples which are non-blind, single-blind,
and double-blind soil QA samples. Non-blind QA
samples are used for internal quality control and
for calibration. Single- and double-blind QA
samples are used in quality assessment and
external quality control. All three types of
blind QA materials have been successfully
236
-------
utilized to control and evaluate laboratory
measurements.c18'191
Non-blind QA Samples
These samples are not blind to the analyst.
The identity and reference values of the sample
are known. Reference materials and laboratory
control samples are examples of non-blind
samples.
Single-blind QA Samples
Single-blind QA samples are used principally as
a reference point in analyses, the data from
which serve as a guide to acceptance or
rejection of routine sample data. A single-
blind QA sample is known to be a QA sample, but
its composition is not known to the analyst.
A performance evaluation material is an example
of a single-blind QA sample.
Double-blind QA Samples
Double-blind QA samples are used as a basis for
acceptance or rejection of routine sample data
and for quality assessment. The difference
between single- and double-blind QA samples is
that the double-blind QA sample is intended to
be indistinguishable from a routine sample.
Visually, the QA sample resembles the routine
sample in container type, number system, soil
texture and soil color. Analytically, the QA
sample resembles the routine sample in
interferences, coanalytes, etc. This minimizes
bias in processing the sample batch. A double-
blind QA sample is even more difficult to
compose or develop because, in addition to
having the same or similar chemical make-up,
the sample must appear to be of the same soil
type. For example, if the soil being sampled
for analysis is a Hagerstown silt loam (a fine
textured medium brown soil with a neutral pH),
an acidic red-colored sand would not be an
appropriate double-blind sample. Spiked field
samples and field duplicates are examples of
double-blind QA samples. Manufactured double-
blind QA materials are rare.
Use of Single-blind and Double-blind QA Samples
Quality assurance samples are used to detect
bias and to estimate precision in the
measurement system. The advantage of double-
blind QA samples is that they are treated
exactly like the routine samples in the
analytical laboratory and hence should be
exposed to the same types and levels of errors
in the preparation and analytical processes.
Unfortunately, it is often difficult to employ
double-blind QA samples for studies of
environmental pollution. Difficulties in using
double-blind QA samples usually arise for one of
two reasons. The first reason is that the nature
of the pollutant may make it impossible to carry
out the drying, grinding, sieving, homogenizing,
and subsampling (to obtain a laboratory sample)
of routine samples outside the analytical
laboratory. This series of preparatory steps is
essential for obtaining homogeneous soil QA
materials. Such treatment produces QA samples
that look different from the routine samples,
provided the routine samples did not go through
the same process before entering the laboratory.
The second most probable reason is that an
appropriate soil QA material is not available,
and there is insufficient time prior to field
sampling to characterize the soil QA material for
double-blind samples. It should be noted that no
matter how many soil QA materials are available,
it is unlikely that a soil QA material exists
that is appropriate for double-blind samples
unless the material actually comes from the site
under investigation.
If it is not possible to employ double-blind QA
samples in an investigation, an alternative
procedure has been suggested based upon single-
blind samples and additional field duplicate
samples.[" The additional field duplicate
samples in this alternative procedure allow the
estimation of total measurement error (i.e., the
precision of the measurement system) and the
estimation of the variance contributions of
several of the possible sources of error.
Depending on where they are incorporated into the
sampling and analytical scheme, the single-blind
samples provide means for detecting bias from
sample handling, preparation, and analysis.
Unfortunately, the single-blind QA samples may
miss some of the bias in the laboratory, owing to
special handling by the chemist, to which a
double-blind sample would not have been
subjected. A research study by Rumley[20]
evaluated the effects of favorable treatment of
samples and of alteration of results to reduce
bias on indices of performance in external
quality assessment (EQA) schemes. He concluded,
in fact, that EQA schemes can be affected by
giving favorable treatment to single-blind
samples.
Since there will always be a need for single-
blind soil QA samples, and the need will often
involve situations requiring rapid response, it
seems imperative that an extensive inventory of
soil QA materials be prepared and maintained for
future environmental pollution studies. Double-
blind soil QA samples should be employed where
practicable, and facilities should be available
to produce such samples in an expeditious manner.
237
-------
AVAILABILITY
The establishment and expansion of monitoring
and enforcement programs by federal agencies
requires the use of many QA samples. Federal
agencies such as the U.S. EPA and the Food and
Drug Administration (FDA) established
repositories of QA materials out of the
necessity to support their own programs. The
private sector, although originally interested
in producing standards for calibration of
different instruments, produces QA samples in
various media and for specific environmental
programs (e.g., RCRA) in a limited variety.
Although a listing of many of the soil QA
materials described that are available today
(Table 1) may appear sizable, many analytes are
not represented. At this time, the authors are
unaware of any sources of soil QA materials for
volatile organic analytes. The natural
variability of soils, however, is the factor
that makes a large number of QA materials
necessary. The same factor limits the ability
to manufacture sufficient materials to provide
realistic and/or blind QA materials for all
hazardous waste sites that are being
investigated. This deficiency makes it
difficult to plan and implement many soil
sampling and analysis QA programs.
SOIL QA MATERIALS NEEDED
AGENCY NEEDS
Clearly there is a need for more sources of
soil QA materials. This leads to certain
questions. Which types are most often needed?
Which materials should be manufactured first?
A survey1211 of U.S. EPA officials shows that
all 10 Regions share an interest in a national
QA material program for Superfund analyses,
primarily for use by the Contract Laboratory
Program and by Potentially Responsible Parties
(PRPs).
Although each Region has specific needs, there
is some agreement on analytes. Most interest
is in materials containing Target Compound
Listt213 analytes. Special requests include
tetrachlorodibenzo-p-dioxin (TCDD) and
pentachlorodibenzo-p-dioxin and -furan
(PCDD/PCDF) isomers; explosives (RDX); benzene,
toluene, and xylene (BTX), solvents; and
polycyclic aromatic hydrocarbons (PAHs) in
sediment. The number of QA materials needed
per year and their concentrations vary among
the Regions (Table 2).c211 One Regional
official commented that site-specific QA
materials are needed.1 The value of the soil
QA materials distributed by the EMSL-LV CLP
Performance Evaluation Program has been
proven, ' but as demonstrated in Table 1, this
program offers a limited variety of samples and
analytes. The EMSL-LV program would need
additional resources in order to be able to
provide a wider variety of materials.
INDUSTRIAL POLLUTANTS
ble
]
Industrial organic chemicals presently comprise
the highest volume of hazardous waste produced,
followed by wastes from general chemical
manufacturing, petroleum refining, and explosives
(Table 3)12 . According to the Comprehensive
Environmental Response Compensation Liabil ity
(Act) Information Systems database, the most
abundant pollutants on the National Priority List
(NPL) of Superfund sites are from the industrial
and general organic chemicals industries,
petroleum refining, and explosives industries
(Table 3). The pollutants found most often on
these NPL sites (Table 4) are Pb, As, Cd, Cr, Hg,
Cu, and cyanides for inorganic pollutants, and
trichloroethylene (TCE), other chlorinated
solvents, and BTX for organic pollutants. The
highest volumes of organic pollutant/waste are
volatile organic compounds (VOCs) , while heavy
metals comprise the greatest volume of inorganic
wastes. It would seem that soil QA samples
containing the pollutants specified by the users
(e.g., Regional users) and/or those most commonly
found at the NPL sites should be the first to be
produced.
SUGGESTED RESEARCH
Supplying Blind Soil QA Materials
At this time, preparing and stocking complete
(adequate analytes) and realistic (double-blind
as well as single-blind) soil QA materials is not
feasible due to the tremendous natural
variability of soils. On the other hand, as
stated previously, variety of QA samples from
present sources is limited (Table 1).
Two general approaches, that overlap somewhat in
their philosophy, are presented for manufacturing
both single-blind and double-blind QA materials.
These are: industry-specific QA materials in
which a limited number of soils are produced that
contain analytes specific to polluting
industries; and site-specific QA materials in
which soils found at hazardous waste sites are
prepared to contain analytes or analyte
combinations commonly found at hazardous waste
sites. Either approach would require a rigorous
multi -laboratory characterization study. As one
example, soils naturally rich in particular
metals could be obtained and processed for either
industry- or site-specific QA materials
238
-------
representing mining industry wastes for sites
with similar soil characteristics.
Industry-specific Materials
Using historical industry data as well as NPL
data, information such as geographic location,
contaminant types, and concentrations can be
mapped and evaluated for any general geographic
trends. This information can then be
correlated with 10-15 general soil-types1 '
to narrow the choices of industry-specific
soil/analyte combinations. The next step would
be to collect and homogenize the selected
soils. During homogenization some of the soils
would be spiked with contaminants for
characterization and distribution. This would
result in samples that could be used for non-,
single-, and perhaps double-blind, blank, or
contaminated soil QA materials. The materials
could then be stored at distribution centers to
fill user requests for various industry-
generated hazardous waste sites.
Site-specific QA Materials
Relying on NPL site data in combination with
geographically related soils, a set of site-
specific soil QA samples could be developed.
In this approach, the selected soils could be
collected for spiking and processing, as
described in the previous section; or, using
site-specific soil/analyte combinations, the
materials could be collected from actual
hazardous waste sites, with blanks being
obtained from nearby uncontaminated soils of
similar composition. The artificially composed
materials and the materials obtained from waste
sites could be used during the investigation
and remediation of sites having similar soils
characteristics, or they could be stored and
used throughout the study of the site from
which they were obtained.
Site-specific QA materials have been
successfully manufactured and used for
treatability studies for similarly
characterized sites,c283 as single-blind QA
samples with routine samples, and for
integration of QA data (site comparison
soils)1293 among several projects on a large (21
square mile) site for the duration of the site
investigation and remediation.
A disadvantage in preparing site-specific soil
QA materials is that often they cannot be used
;as double-blind samples because their visual
characteristics may be altered by the
processing that is employed to prepare QA
materials. The site-specific approach is very
successful, however, when the site is fairly
dryt151 and sieving is not necessary.
CONCLUSION
Increased public interest in environmental issues
has led to new legislation at both the state and
federal levels. As a result of these laws, many
contaminated sites have been or will be
evaluated. A large number of these sites have
been grossly contaminated by a variety of
hazardous chemicals at different concentrations.
A parallel increase in the number of sites added
to the National Priority List (NPL) and the
number of contaminants regulated by RCRA and
Superfund Amendment Reauthorization Act (SARA)
(CERCLA) and other federal and state regulations
demands a comprehensive suite of quality
assurance samples111 or a mechanism to produce
such on short notice. The QA samples should
represent the variety of contaminants at
appropriate concentrations and natural soil
characteristics to provide a true comparison to
real world samples. The authors of this report
recommend that the rationale document113 described
previously be consulted to determine whether the
information and conclusions presented there pose
serious problems for the investigator. If the
quality of environmental data cannot be
adequately assessed because suitable QA materials
do not exist, then more effort clearly needs to
be made to increase the supply of soil QA
materials.
Future research should include a preliminary
study comparing approaches for producing
realistic soil QA materials. It is felt that
such a study may show that the site-specific
approach produces the most useful soil QA
materials. A multi-laboratory pilot study would
evaluate the advantages and disadvantages of each
approach and should lead to a long-term plan for
providing a supply of soil QA materials.128'293
NOTICE
Although the research described in this paper has been funded wholly
or in part by the United States Environmental Protection Agency inder
Cooperative Agreement No. CR 814701 with the Environmental Research
Center of the University of Nevada, Las Vegas, it has not been
subjected to Agency review and therefore does not necessarily reflect
the views of the Agency and no official endorsement should be
inferred.
REFERENCES
1. U.S. EPA. 1989. "A Rationale for the
Assessment of Errors in the Sampling of
Soils." EPA 600/X-89/203, Environmental
Monitoring Systems Laboratory-Las Vegas,
NV.
239
-------
2. Hertz, H.S. 1988. "Quality Assurance,
Reference Materials, and the Role of a
Reference Laboratory in Environmental
Measurements." Proceedings, The
International Symposium on Trace Analysis
in Environmental Samples and Standard
Reference Materials. Honolulu, HI, pp.
5-8, January 6-8.
3. Taylor, J.K. Quality Assurance of
Chemical Measurements. Lewis Publishers,
Inc., Chelsea, MI, 1987, pp.159-163.
4. Seward, R.W., editor. Standard Reference
Materials and Meaning/Measurement. NBS
SP 408. National Bureau of Standards,
Gaithersburg, MD, 1973.
5. Cali, J.P. The Role of Standard
Reference Materials in Measurement
Systems. National Bureau of Standards
Monograph 148. NBS, Gaithersburg, MD,
1975.
6. Steger, H.F. Certified Reference
Materials Report 80-6E. Canada Centre
for Mineral and Energy Technology, Ottawa
Canada, 1980.
7. Taylor, J.K. Handbook for SRM Users.
NBS SP 260-100. National Bureau of
Standards, Gaithersburg, MD, 1985.
8. U.S. EPA. 1984. "Quality Assurance
Support: Project Plan for the Superfund
Standards Program." Tr-506-112A
(Internal Report). Project Officer J.G.
Pearson.
9. Bowman, M.S., G.H. Faye, R. Sutarno, J.S.
McKeague, and H. Kodema. 5o/7 Samples
SO-1, SO-2, SO-3, and 50-4: Certified
Reference Material. Report 79-3. Canada
Centre for Mineral and Energy Technology,
Ottawa Canada, 1979.
10. Stoch, H., and E.J. Ring. The
Preparation and Analysis of Reference
Materials and the Provision of
Recommended Values. Progress Report No.
5, Report No. M. Council for Mineral
Technology, Randburg, South Africa, 1983.
11. Holynska, B., J. Jasion, M. Lankosz, A.
Markowitz, and W. Baran. "Soil SO-1
reference material for trace analysis."
Fresenius Z Analytical Chemistry,
322:250-254, 1988.
12. Campana, J.E., D.M. Schoengold, and L.C.
Butler. "An environmental reference
material program: Dioxin performance
evaluation materials." Chemosphere 18(1-
6):169-176, 1989.
13. Inn, K.G.W., W.S. Liggett, and J.M.R.
Hutchinson. "The National Bureau of
Standards Rocky Flats Soil Standard
Reference Materi al." Nuclear Instruments
and Methods in Physics Research 223:443-
450, 1984.
14.
Jorhem, L., and S. Slorach. "Design and
use of quality control samples in a
collaborative study of trace metals in
daily diets." Fresenius Z Analytical
Chemistry, 322:738-740, 1988.
15. Thiers, R.E., G.T. Wu, H. Reed, and L.K.
Oliver. "Sample stability: A suggested
definition and method of determination."
Clin. Chem. 2212:176-183, 1976.
16. McKenzie, R.L., ed. "NIST Standard
Reference Materials Catalog 1990-1991."
NIST Special Publication 260, January,
1990.
17. U.S. EPA. "Annual Summary Report FY89,
Quality Assurance in Support of Superfund."
EPA600/X-90/033, Environmental Monitoring
Systems Laboratory-Las Vegas, NV, February,
1990.
18. Frank, D.J. "Blind sample submission as a
tool for measurement control." Institute
of Nuclear Materials Management, 14(3):112-
117, 1985.
19. Glenn, G.C., and T.K. Hataway. "Quality
control by blind sample analysis."
American Journal of Clinical Pathology
72(2):156-162, 1979.
20. Rumley, A.G. "External Quality Assessment
(EQA): The effect and implications of
favourable treatment of EQA samples."
Medical Laboratory Sciences, 41:295-298,
1984.
21. Bleyler, R. Viar, and Company. "Survey of
Quality Control for Superfund Programs."
April, 1989.
22. Gaskill, A. "News and Views:
environmental reference standards."
Environmental Lab, Z: 12-15, 1990.
23. Butler, L.C. Personal communication. U.S.
EPA, Environmental Monitoring Systems
Laboratory-Las Vegas, NV, 1990.
24. Krieger, J. "Hazardous waste management
database starts to take shape." Chemical &
240
-------
Engineering News, pp. 19-21. February 6,
1989.
25. McCoy, D.E. '"301" Studies provide
insight into future of CERCLA.' The
Hazardous Waste Consultant, March/April
1985, McCoy and Associates, Lakewood,
Colorado, Vol. 3/2: 18-24, 1985.
26. U.S. Department of Agriculture. Land
Resource and Major Land Resource Areas of
the United States. U.S. Soil
Conservation Service, Agriculture
Handbook 296, 1981.
27. U.S. Geological Survey. The National
Atlas of the United States of America.
Department of the Interior. Washington,
D.C. pp. 85-88, 1970.
28. Esposito, P., J. Hessling, B. B. Locke,
M. Taylor, M. Szabo, R. Thurman, C.
Robers, R. Traver, and E. Barth.
"Results of treatment evaluations of a
contaminated synthetic soil." JAPCA 39:
294-304, 1989.
29. Barich III, J.J., G. Raab, R, Jones, J.
Pasmore. "The Application of X-ray
Fluorescences Technology in the Creation
of Site Comparison Samples and in the
Design of Hazardous Waste Treatability
Studies." First International Symposium
Field Screening Methods for Hazardous
Haste Site Investigations, Symposium
Proceedings. Las Vegas, NV, pp. 75-80,
October 11-13, 1988.
241
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS*
8
SUPPLIER
Environmental
Research Associates
5540 Marshall St.
Arvada, CO 80002
USA
1-800-372-0122
QA
MATERIAL
Sludge
CLP-priority
pollutant in soil
Hydrocarbon
fuel in soil
Total petroleum
hydrocarbons
(TPH) in soil
Benzene,
toluene, ethyl
benzene and
xylene (BTEX)
in water/soil
DESCRIPTION
Certified QC standards in a
sludge matrix for volatile
(Benzene & TCE), semi-
volatiles (5 BNA), pesticides/
PCB, and metal analysis (11
metals)
Certified QC standards in soil
matrix for Superfund volatiles
(6 to 8 VOCs), semi-volatiles,
trace metals, and cyanide
analysis
Standards of gasoline, No. 2
diesel, heating oil, and crude
oil in a soil matrix
Standardized 50 g QC soil
samples, one specifically
designed for analysis of TPH
in soil in the presence of fatty
acids in screw top bottles
QC set containing two
standard concentrates and
one soil matrix
TYPE & CONCENTRATION
RANGE
Volatiles (5-500 ug/kg)
Semi-volatiles (300-30,000 ng/kg)
Pesticides/PCBs (10-10,000
"g/kg)
Trace metals (1-5,000 mg/kg)
Volatiles (5-500 ug/kg; Sealed
ampoule containing VOCs in
methanol to be spiked into 10 g
of soil)
Semi-volatiles (300-30,000 ug/kg)
Pesticides/PCBs (10-10,000
ug/kg)
Trace metals (1-5,000 mg/kg)
20 g QAS containing unleaded
gasoline (5-500 mg/kg)
No. 2 diesel fuel, heating oil or
crude oil (10-5,000 mg/kg)
Standard 1 - 50 g (100-2000
mg/kg) level
Standard 2 - the presence of fatty
acids (100-2000 mg/kg)
Ampulated 5-500 ug/kg in
CH3OH to be spiked onto 10 g
soil
APPLICATION
40 CFR 503
Evaluation of
laboratory
performance -
especially for
CLP-type
analysis
Evaluation of
specific analysis
for Underground
Storage Tanks
(UST program)
UST program
UST program
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
QA
MATERIAL
DESCRIPTION
TYPE & CONCENTRATION
RANGE
APPLICATION
Fisher Scientific
711 Forbes Avenue
Pittsburgh, PA 15219
USA
(412) 562-8300
Solid waste
Real world samples,
homogenized for consistency
and tested for accuracy
Fly ash (4 metals)
Waste water treatment
media (3 metals)
Diatomaceous earth filter cake (4
metals)
Circuit board coating sludge
(5 metals)
Electroplating tank bottoms
(5 metals)
Raw sludge, chrome plating
process (4 metals)
Incinerated sludge (5 metals)
Municipal incinerator ash (8
TCLP metals, 4-4000 ppm)
PAH-contaminated soil
(14 PAH and PCPs, 20-1200
ppm)
Custom Orders
SW846
Water treatment
facilities
SW846
Waste from
electronic
industries
Waste from
electroplating
Waste from
electroplating
Waste from
incinerators
SW 846,
Methods 3050,
6010
SW 846,
Methods 3540,
3550
As required
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
National Institute of
Standards and
Technology
Chemistry Bldg. B-
311
Gaithersburg, MD
20899 USA
302-975-6776
QA
MATERIAL
Ore, minerals,
and refractories
Solid organics
DESCRIPTION
QC reference materials for
critically important material
balance in mining and
metallurgical industries
QA materials for analysis of
materials for constituent of
interest
TYPE & CONCENTRATION
RANGE
Copper ores (5 metals,
0.03 ppm to 0.84%)
Fluorospar (CaF2) (97.4 to
98.8%)
Iron ores (Fe, 58 to 90.8%)
Bauxite ores (Al, 21.1 to
28.8%)
Powdered lead-based paint
(Pb, 12%)
Trace mercury in coal
(Hg,0.13ng/g)
Lead in refinery fuel
(5 varieties, 11.0 to
780.0 ug/g)
APPLICATION
Mining and
metallurgical
processing
Lead-based paint
analysis
Heavy metals in
fuel
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
National Institute of
Standards and
Technology
Chemistry Bldg. B-
311
Gaithersburg, MD
20899 USA
302-975-6776
QA
MATERIAL
Trace elements
Urban dust
Diesel
particulate
matter
PAH in solid
matrices
Polychlorinated
biphenyls in
sediments
Organics in
marine
sediments
DESCRIPTION
Trace elements in solid
matrices (12 to 42 elements)
Urban dust QA materials for
analysis of organic
constituents
QA materials for analysis of
diesel particulate matter and
its organic constituents
QA materials with variety of
PAHs on solid matrices
QA materials of sediments
contaminated by PCBs
QA materials made of marine
sediment contaminated by
organics
TYPE & CONCENTRATION
RANGE
Urban particulate
(1.0-860 ug/g)
Coal - bituminous
(0.1-100 ug/g)
Coal - fly ash, 4 varieties
(0.2-200 ug/g)
Coal - subbituminous
(0.1-20 ug/g)
Estuarine sediment
(0.5-375 ug/g)
Buffalo River sediment
(0.1-555 ug/g)
10 g
100 mg/ampoules
6 varieties, 1.0-4000 ug/g
In preparation
In preparation
APPLICATION
Evaluation of
laboratory
performance
especially for
analysis of trace
elements in
variety of
matrices
Air pollution
Air pollution
SW 846 or
similar analytical
programs
SW 846 or
similar analytical
programs
General
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
Canada Centre for
Mineral and Energy
Technology
555 Booth Street
Ottawa, Canada
K1A OG1
United States
Geological Survey
Geochemistry
Branch
P.O. Box 25046 MS
973
Denver Federal
Center
Denver, CO 80225
U.S. Environmental
Protection Agency
RREL, Releases
Control Branch
Edison, NJ
08837-3079 USA
201-321-4372
QA
MATERIAL
Soil Samples
SO-1, SO-2,
SO-3, SO-4
GXR-1-6
Synthetic Soil
Matrix/I
Synthetic Soil
Matrix/II
DESCRIPTION
Compositional Reference
Materials
Jasperoid soils, Cu millhead
tailings, B horizon soil
30% clay, 25% silt, 20% sand,
20% topsoil, 5% gravel
High organic, low metal
Low organic, low metal
TYPE & CONCENTRATION
RANGE
Clayey soil, sandy podzolic B
horizon with a high organic
content, a calcareous till, and a
chernozemic A horizon
Chemical and physical soil and
mineral properties
Organic: 400-8200 mg/kg
Metal: 10-450 mg/kg
Organic: 40-820 mg/kg
Metal: 10-450 mg/kg
APPLICATION
General
analytical and
earth science for
agricultural,
forestry, and
environmental
applications,
especially for
mining and
metallurgical
operations.
General
analtyical and
earth science for
agricultural,
forestry, and
environmental
applications,
especially for
mining and
metallurgical
operations.
Soil treatability
studies
Soil treatability
studies
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
U.S. Environmental
Protection Agency
RREL, Releases
Control Branch
Edison, NJ
08837-3079 USA
201-321-4372
U.S. Environmental
Protection Agency
EMSL-LV, QAD
P.O. Box 93478
Las Vegas, NV
89193-3478
702-798-2114
FTS 545-2214
U.S. Environmental
Protection Agency
EMSL-LV, QAD
P.O. Box 93478
Las Vegas, NV
89193-3478
702-798-2114
FTS 545-2214
QA
MATERIAL
Synthetic Soil
Matrix/Ill
Synthetic Soil
Matrix/IV
Dioxin
performance
evaluation
materials
Base-neutral-
acid PEMs
Pesticide PEMs
DESCRIPTION
Low organic, high metal
High organic, high metal
Real World samples
contaminated by dioxin and/
or selected matrices fortified
by dioxin
Sand fortified with selected
BNAs
••„
Real world samples
contaminated with toxaphene
and other pesticides or
selected soil fortified by
selected pesticides & PCBs
TYPE & CONCENTRATION
RANGE
Organic: 40-820 mg/kg
Metal: 500-22,500 mg/kg
Organic: 40-8200 mg/kg
Metal: 500-22,500 mg/kg
Kiln ash, XAD Resin, filter
paper, florisil, clay, sand (20 ppt
to 6 ppb)
TCDD/PCDF soil
Times Beach soil
Times Beach & PCDD/PCDF
soil
Times Beach & Region 9 soil
Low level BNA (400 ppb)
Medium level BNA (15 ppm)
High level BNA (75 ppm)
Mixed level BNA
Toxaphene soil
Pesticide soil 1 (4-40 Hg/kg)
Pesticide soil 2 (4-40 ug/kg)
Pesticide soil 3 (30-100 ug/kg)
+ PCB 1016
Pesticide soil 4 (30-60 Ug/kg)
+ PCB 1266
APPLICATION
Soil treatability
studies
Soil treatability
studies
SW 840, 8280
SW 846, 8250,
8270
SW 846, 8080
-------
TABLE 1. LIST OF GOVERNMENT AND PRIVATE SOURCES FOR SOIL/SOLID QA MATERIALS (Continued)
SUPPLIER
U.S. Environmental
Protection Agency
EMSL-LV, QAD
P.O. Box 93478
Las Vegas, NV
89193-3478
702-798-2114
FTS 545-2214
QA
MATERIAL
Inorganic PEMs
DESCRIPTION
Selected soil samples fortified
with metals and cyanide
TYPE & CONCENTRATION
RANGE
LCS metals (1 ppm-200,000 ppm)
LCS, cyanide (4-8 ppm)
APPLICATION
SW 846, 6010
Information containedin this table was obtained in September 1990 and may not include some sources of QA materialsTaespite
the authors' efforts to be accurate and complete.
DISCLAIMER: Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
-------
TABLE 2. SOIL AND WATER PE SAMPLES NEEDED BY THE 10 REGIONS OF THE U.S. EPA™.
Region
I
II
III
IV
V
VI
VII
VIII
IX
X
Analytes
VOA, BNA, PEST/PCB
soil blanks for VOA and BNA
Dioxin
Unspecified
«TCE 25 ppb
toluene
vinyl chloride
phenols
napthalene
pentachlorophenol
2 or 3 mixes for each fraction; e.g.
5 analytes
7 analytes
3 analytes (determine in workgroup)
•VOA and BNA from CLP-TCL
PEST/PCBs
Metals
VOA and BNA
case by case; not routine enough to predict
levels or analytes
PCBs
Pest/Herb
PCP
TCE and solvents
dioxin congeners, tetrachloro-specific isomcrs
Complete TCL (grouped aromatics, PAH,
etc.)
EDB
RDX explosives
TCDD only
PCDD/PCDF
chloroform, carbon tetrachloride
BTX
chlorinated hydrocarbons
VOA and BNA
Heavy metals
include most common and possibly some
more difficult compounds
PAH (sediment)
Levels
same as CLP PE
no detectable levels
isomer specific; not only 2,3,7,8
Unspecified
100 ppb
100 ppb
100 ppb
50 ppb
100 ppb
2 x (CRQL)"
5 x (CRQL)
10 x (CRQL)
-1.5 ppb
-CRQL
-CRQL
•CRQL
100-80,000 ppm (soil)
300-10,000 ppm (soil) ,
300-30,000 ppm (oily matrix)
low ppb (water)
Low (10 x CRQL)
Med (50 x CRQL)
100 ppt; 1 ppb
Ippb
1 ppb; 5 ppb; 10 ppb (soil)
10 ppt (water)
10 ppb (soil)
20 ppb
wide variety
high for
high soils;
low for drinking water
asbestos needed but don't expect it
in this effort
low and high (within DOT
regulatory limits)
# PE samples/year
100/type/year
100, or if replace MS/MSD* 1/50
samples
unknown
15-20
15-20
15-20
15-20
up to 100 if convenient and
flexible schedule
200 water, 200 soil
50
20
1500 soil
50 water
50
-30;
contractor!, would like 2
50-75/matrix/analyte set
if replace MS/MSD, 1 per data set
* Matrix spike/Matrix spike duplicate
*Soil samples not requested.
"Contract required quantitation limit
249
-------
TABLE 3. VOLUME OF WASTE GENERATED BY INDUSTRIAL ACTIVITIES PER
YEAR'24'.
Standard Industrial
Classification
2869
2800
2911
2892
2821
4953
2879
2865
2816
2812
Category
Industrial organic chemicals
General chemical
manufacturing
Petroleum refining
Explosives
Plastic materials/resins
Refuse systems (commercial
TSDR* facility)
Agricultural chemicals
Cyclic crudes/intermediates
Inorganic pigments
Alkalis/chlo rine
Hazardous Waste Volume,
Millions of metric tons
60-80
40-50
20-30
10-15
6-10
5-8
5-8
5-8
3.5-5
2.5-4.5
* Transportation, storage, disposal, or recycling
250
-------
TABLE 4. MOST FREQUENTLY REPORTED SUBSTANCES AT 546 NPL SITES'25'.
Rank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Substance
Trichloroethylene
Lead
Toluene
Benzene
Polychlorinated biphenyls
(PCBs)
Chloroform
Tetrachloroethylene
Phenol
Arsenic
Cadmium
Chromium
1, 1, 1-Trichloroethane
Zinc and compounds
Ethylbenzene
Xylene
Methylene chloride
Trans- 1,2-Dichloroethylene
Mercury
Copper and compounds
Cyanides (soluble salts)
Vinyl chloride
1,2-Dichloroethane
Chlorobenzene
1, 1-Dichloroethane
Carbon tetrachloride
Percent of Sites
33
30
28
26
22
20
16
15
15
15
15
14
14
13
13
12
11
10
9
8
8
8
8
8
7
251
-------
DISCUSSION
JANINE ARVIZU: Have you considered as one of your options for preparation
of these materials, reconstruction of some simulated soils from stockpiles of
individual soil constituents (clays and gravels) and so forth? Based on compo-
sitional analysis of the soils, would you be able to reconstruct QA materials on
a site-specific basis?
AMY CROSS-SMIECINSKI: Yes, we have considered this possibility and
have tried to locate large stockpiles of various types of soils. Most of the sources
of soils that we have found are not extensive. They're small volumes and the
people who distribute them are apprehensive about sending out large quantities.
They are used mostly for a routine soil sample analysis.
JANINE ARVIZU: I'm curious as to how you would envision addressing the
problem of accurately dealing with active soils, (e.g., biologically active soils or
natural soils that have absorptive properties) and being able to accurately
determine the recovery of analytes from those types of materials?
AMY CROSS-SMIECINSKI: In another study we have in the poster session,
we have looked into various types of soil, specifically volatile organic preserva-
tives, to prevent those kinds of degradations and activities. But then it's
something that would be a real problem for any type of soil QA material.
LLEW WILLIAMS: I might just comment on something we've been wanting
to try to see if we can get better representative spiking into QA materials. I think
this has always been a concern that spiked materials frequently don't reflect in
recoveries for instance. The same analytes, if they were naturally in a waste
material, we may get fifteen percent (15%) recovery, we spike them and then we
get ninety percent (90%) back.
One of the things that we're looking into right now and some of you who have
the facilities might want to play around with it a little bit, too, is looking at the
concept of using super critical fluid to put analytes back into matrices, rather than
taking them out. If the concept is a good one to reach down into the pores and draw
analytes out of a matrix, it's possible to be able to release the pressure and put
analytes deeply into a matrix in a way that they may better assimilate natural
materials.
JANINE ARVIZU: Your concerns about double blind QA samples for soils, I
think are really legitimate. Have you considered the introduction of single blind
QA samples with every analytical batch as an alternative to having a double
blind? Would it serve some of the same purposes?
AMY CROSS-SMIECINSKI: We believe it does and it has. Single blind QA
samples have been used this way for some time, particularly in the dioxin
program. But we feel that the double blind QA samples, although they're very
hard to manufacture, would be the most realistic type of soil QA samples at this
point.
252
-------
EVALUATION OF EMISSION SOURCES AND HAZARDOUS
WASTE SITES USING PORTABLE CHROMATOGRAPHS
R. E. Berkley
Environmental Protection Agency
Atmospheric Research and Exposure Assessment Laboratory
Research Triangle Park, NO
ABSTRACT
Portable gas chromatographs (PGC) cap-
able of direct detection of ambient con-
centrations of toxic organic vapors in
air were operated in field studies while
simultaneous data were taken for compar-
ison by the Canister/TO-14 Method. Sam-
ples were obtained downwind of Superfund
hazardous waste sites, highways, chem-
ical plants, and in locations where
there was concern about odors or nasal/
respiratory irritation. In some cases
two PGCs equipped identically were used
side-by-side or upwind/downwind. In ot-
hers, different columns were used side-
by-side to analyze a larger group of
compounds. Reasonable agreement between
methods was found, even though sampling
techniques were not equivalent. Such
agreement suggests that both methods
were free of sampling errors, and that
the data were substantially accurate.
This paper has been reviewed in accor-
dance with the U. S. Environmental Pro-
tection Agency's peer and administra-
tive review policies and approved for
presentation and publication. Mention
of trade names or commercial products
does not constitute endorsement or re-
commendation for use.
INTRODUCT I ON
Toxic organic compounds are usually pre-
sent in ambient air at such low levels
(typically about one ppB) that they can-
not be analyzed without preconcentra-
tion. In the TO-14 Method, six-liter
»ir samples are collected in passivated
canisters and stored pending analysis.
Just prior to analysis they are cryogen-
ically preconcentrated (1). Use of a
portable gas chromatograph (PGC) equip-
ped with a photoionization detector
(PID) sensitive enough to detect organic
compounds at sub-ppB levels without pre-
concentration offers an alternative sam-
ple collection method which produces
data on-the-spot in near real-time.
PID detectors are no longer novel. In
1984-5 Verner (2) and Driscoll (3) re-
viewed more than a decade of PID use in
gas chromatography. There have been
several reports since 1980 describing
analyses of airborne organic vapors with
them. However, none of the instruments
were portable, and sample preconcentra-
tion was always required because those
PIDs were not significantly more sensi-
tive than other kinds of detectors
(4-8). Then Leveson and coworkers dev-
eloped a 10.6 electron-volt PID of sig-
nificantly greater sensitivity and in-
corporated it into a PGC (9). The light
source was an electrodeless discharge
tube which was excited by a radio-fre-
quency oscillator to produce an intense
emission line. The chromatograph was
claimed to detect benzene without pre-
concentration at 0.1 ppB (10-13). How-
ever, the lamp is restricted to low-tem-
perature operation because heating it
would decrease sensitivity by broadening
the emission line. For Leveson1s PGC
(Photovac Model 10A10), Berkley estim-
ated a benzene detection limit equival-
ent to 0.03 ppB. The smallest sample
actually analyzed, one microliter con-
taining 1.6 picogram of benzene, produ-
253
-------
ced a 2.3 volt-second peak at maximum
gain. A linear response to benzene was
observed over a wide concentration range
(0.5 to 130 ppB), and injections as
large as one milliliter could be made
without significant loss of chromato-
graphic resolution. Similar sensitivity
to other aromatic compounds and to
chloroalkenes was also observed (14).
Such an instrument obviously should be
useful for air monitoring, but few re-
ports of it have appeared. Lipsky an-
alyzed vinyl chloride from landfills
(15), and Hawthorne analyzed indoor air
in a "research house" (16). Jerpe est-
imated a benzene detection limit of 20
picograms using a Model 10A10 PGC to
which an external capillary column and
constant-volume sample loop had been
connected (17). Users of the Model
10A10 PGC experienced difficulty with
battery endurance, baseline drift, and
on-site data interpretation. These pro-
blems were mostly resolved by the later
series of Model 10S- PQCs. Since PGCs
can be more easily transported than
large numbers of canisters, they more
readily produce large volumes of data in
the field. Their disadvantages are that
(a) at present they are limited to low
resolution chromatography, (b) they id-
entify, by retention time only, the lim-
ited number of compounds which they can
detect at low ppB levels, and (c) they
require a skilled operator.
It is difficult to be certain that pre-
concentrated samples are not being spoi-
led by sampling errors. Although sample
integrity during storage in passivated
canisters has been demonstrated in the
absence of highly reactive compounds
(18), artifact formation can be caused,
for example, by HCl (19). We have eval-
uated PGCs in both laboratory and field
operation (20, 2t). Because PGCs are
not affected by breakthrough of analytes
from a preconcentration trap, by chem-
ical reactions between collected com-
pounds, or by sample degradation during
storage, use of them in parallel with
the Canister/TO-14 Method could identify
such problems, should they ever occur,
if the two methods could be shown to
consistently produce similar results
under field conditions. That requires
much parallel use over a long period of
time at a variety of sites under differ-
ent ambient conditions using many kinds
of operating parameters. Herein are re-
ported an accumulation of comparative
data obtained during the past two years.
EXPERIMENTAL
Spherical 6-liter electropolished can-
isters (SIS, Incorporated) were used to
collect air samples and store PGC calib-
ration standards. Canisters were clean-
ed by heating to 90°C while evacuating
through a liquid nitrogen trap to a fin-
al pressure below 10 micrometers (mer-
cury equivalent) for two hours. Samp-
ling for direct comparison of canister
and PGC data was done by holding a can-
ister with its inlet less than 10 centi-
meters from the end of the PGC probe and
opening the valve to fill it during the
time the PGC sample pump was running.
Another method of comparison was to per-
form consecutive PGC analyses while
time-integrated canister samples were
being collected. For time-integrated
measurements, evacuated canisters were
fitted with pre-calibrated mechanical
flow controllers, and air was sampled at
25 mi 11i1iters/minute for two hours. Air
samples collected in canisters were
transported to a laboratory, cryogen-
ically preconcentrated, and analyzed
using a modified Hewlett-Packard Model
5880A gas chromatograph equipped with
flame ionization and electron capture
detectors. A Hewlett-Packard Model
5970A mass selective detector was used
for some samples. Calibration was based
on 41 organic compounds cited in the
Canister/TO-14 Method (1).
Microprocessor-controlled PGCs (Photovac
Model 10S70) were used. They were
equipped with constant-temperature col-
umn enclosures and 0.53 millimeter ID X
10 meter fused-silica wall-coated open-
tubular (WCOT) columns, a 1.67 meter
section of which was backflushable pre-
column. Chemically-bonded stationary
liquid phases were used, either CPSilSCB
or CPSil19CB (Chrompak). A KCl/Alumina
porous-layer open-tubular (PLOT) column
of the same size and configuration was
used for extremely volatile compounds.
Ultrazero air (less than 0.1 ppM carbon)
was the carrier gas. An IBM-compatible
laptop computer, using vendor-provided
software via an RS-232 interface, con-
trolled chromatograph operation and data
storage. Chromatographic peaks were id-
entified and quantitated using retention
times and response factors stored in
nonvolatile memory of the PGC micropro-
cessor. The calibration library was
created by analyzing mixtures of anal-
ytes (10 ppB) produced by flow-dilution
of commercially-prepared standards as
described above. Compounds with ioniza-
-------
tion potentials greater than 10.6 elec-
tron-volts were not detected by PGCs at
ambient (below 10 ppB) levels. Before
beginning to sample, a stable baseline
Mas observed, and the library was recal-
ibrated with a single-compound standard
(approximately 10 ppB) which had been
certified by GC/FID analysis. Chloro-
benzene or tetrachloroethylene were used
as calibrants with the WCOT columns, and
vinyl idene chloride with the PLOT col-
umn. During sampling, automatic recal-
ibration was performed every 4 or 5 runs
using the single compound standard, af-
ter which the microprocessor corrected
the retention time and response factor
for the calibrant, then corrected pro-
portionally the retention times and res-
ponse factors of other compounds. Samp-
les were taken every 15 minutes. Air
was drawn into the sample probe (3 met-
ers long X 2 millimeter ID stainless
steel tubing) for 45 to 60 seconds. Then
the sample was injected for 7 to 15
seconds, after which the sample loop was
removed from carrier flow to minimize
peak tailing. The precolumn was back-
Hushed by the carrier stream except
iihile calibrated compounds were passing
through it. Calibration runs differed
from sample runs only in that the loop
received calibration mixture instead of
an air sample. PGCs were sheltered from
drafts and direct sunlight inside a ve-
hicle or building, and a stainless steel
Sample probe was extended through a win-
dow or a sampling port. External re-
phargable 12-volt batteries (Johnson
Controls GC12800 or PP12120 Gel-Cell,
and Sears Die-Hard Marine) were used to
supply power.
RESULTS AND DISCUSSION
|n comparing Canister and PGC data it is
important to remember that samples col-
lected by the two methods are not equi-
Sfalent. A PGC analyzes only one of 50
to 70 milliliters of air which enter the
probe during sampling, whereas a repre-
sentative sample of the entire six lit-
ers collected by the canister is anal-
yzed. If the air is well-mixed and dev-
oid of reactive or corrosive materials,
then canister and PGC data should resem-
ble each other, and generally do. How-
sver, if a heterogeneous plume is samp-
Jed, or if highly reactive materials en-
ter the canister, then PGC and canister
tfata could differ significantly even
though the "same" air was sampled.
Complaints about episodes of stench at
Marcus Hook, PA were investigated at the
request of EPA Region III. A PGC was
operated in a van at several sites, and
canister samples were taken for compari-
son. The results are shown in TABLE 1.
The PGC twice failed to recognize small
benzene peaks which eluted in the tail
of the large initial peak. The CPSilSCB
column eluted compounds so close toget-
her that resumption of backflush always
interfered with some peak, no matter
when it occurred. In this case toluene
was missed. Trichloroethylene, reported
by the PGC, was never found in the can-
isters. That peak was undoubtedly due to
some other compound which had a similar
retention time. For other compounds,
agreement between the two methods was
reasonab1e.
TABLE 2 shows samples taken at hazardous
waste sites near Wilmington and New Cas-
tle, Delaware. Concentrations at the
Superfund remediation sites were low,
typical of sub-ppB background levels in
remote areas, showing that buried waste
was not emitting significant quantities
of these compounds into the air. Rela-
tive agreement between PGC and canister
data seemed to improve with increasing
concentration. PGC data for tetra-
chloroethylene at the waste lagoon were
not reported because of a persistent co-
eluting peak. Samples taken by both
methods near the waste incineration
plant show toluene and higher homologues
at significant levels. High levels of
benzene and chlorobenzene were found by
both methods downwind of the Standard
Chlorine plant. For compounds found by
both methods, agreement was reasonable
over a wide range of concentrations.
Under Project 02.01-12 of the US-USSR
Environmental Agreement, samples were
taken at a roadside site about 12 kilo-
meters from Vilnius, Lithuania. Two
PGCs were operated while time-integrated
canister samples were collected. A mo-
bile laboratory stood about 20 meters
from the highway on ground about 2 me-
ters below it. Daytime traffic volume
was moderate-to-heavy without stop-and-
go congestion and subject to a 100 km
per hour speed limit. No industrial ac-
tivity was visible in the immediate vi-
cinity. Two identically equipped PGC's
were compared side-by-side and then up-
wind/downwind. During side-by-side op-
eration inside the mobile laboratory,
the sample probes extended to about 18
meters from the roadway and one meter
above it. TABLE 3 compares colocated
255
-------
and upwind/downwind PGC analyses with
time-integrated canister data. During
colocated sampling canisters were placed
3 and 10 meters downwind of the highway.
Sampling was done during nonturbulent
movement of air across the site and
while traffic density was fairly con-
stant. Average levels of benzene, tol-
uene, ethylbenzene, m,p-xylene (reported
as one compound) and o-xylene found by
the PGC's were in reasonable agreement
with data from the canisters. PGC data
for toluene, and sometimes m,p-xylene,
exceeded average concentrations found in
the 10 meter canisters, even though the
PGCs were farther from the highway.
This discrepancy may have occurred be-
cause the PGCs often sampled the plumes
of passing vehicles. When the PGCs were
deployed across the highway from each
other, PGC-1 was inside a van parked 12
meters downwind while PGC-2 remained
upwind in the mobile laboratory. Canis-
ters were again placed 3 and 10 meters
downwind of the highway. Scheduling
constraints allowed only a half hour of
PGC sampling to be compared to the can-
isters, but downwind PGC results agreed
substantially with canister data.
At a Superfund remediation site in
northwest Georgia, airborne emissions
produced strong odor but contained low
levels of compounds which could be det-
ected by the PI Ds. Two PGCs equipped
with CPSil19CB columns were operated
side-by-side while canister samples were
taken for comparison. Data are shown in
TABLE 4. Toluene and xylenes were con-
sistently seen by both methods at sim-
ilar levels. Some styrene was also
seen. These compounds probably came
from trucks and earth-movers on the
site. The CPSil19CB columns provided
better resolution than CPSilSCB columns,
but benzene peaks smaller than one ppB
were missed because the PGC peak-recog-
nition algorithm could not find them on
the tail of the large initial peak.
Compounds which can be analyzed without
concentration by a PGC are those to
which the PID is sensitive and which can
be separated from each other by an iso-
thermal column at low temperature (50°C
maximum). The number of compounds which
can be analyzed can be increased by op-
erating two PGCs side-by-side with dif-
ferent columns. An example is shown in
TABLE 5. The site was about 40 meters
downwind of a dry cleaning plant. PGC-1
was equipped with a KCl/Alumina PLOT
column and used to analyze vinyl chlor-
ide and vinylidene chloride. Since the
PLOT column had very low bleed, the PGC
could be operated at maximum gain
(1000). PGC-2 equipped with a CPSilSCB
column was calibrated for the usual list
of compounds. Traces of vinyl chloride
and vinylidene chloride were found by
PGC-1 but not found in the canisters.
These concentrations were below detec-
tion limit (approximately 0.2 ppB) for
the Canister/TO-14 Method. PGC detec-
tion limits for vinyl chloride and vin-
ylidene chloride were 0.005 and 0.010
ppB, the amounts which would have pro-
duced 5 millivolt-second peaks. The
integration algorithm does not process
smaller peaks. Canister and PGC data
showed tetrachloroethylene at elevated
concentrations. They did not agree
closely, probably because the plume was
poorly mixed. To measure the extent of
agreement between PGC and canister data
a criterion for evaluation is needed'
The absolute difference between results
was chosen because it does not change
drastically with concentration. For
each compound, the averages of absolute
differences are shown in TABLE 6. For
the CPSilSCB column these differences
(from data in TABLES 1, 2, and 5) range
approximately from 1 to 2 ppB. Appar-
ently, absolute differences do increase
slightly with increasing concentration
Supposing they did not, then at about
100 ppB, relative differences would be
5%. At 10 ppB they would be approx-
imately 10%, and at one ppB, 100%. A
difference of 100% seems large, but sup-
pose one method reported one ppB of tol-
uene while the other reported two ppB.
That difference would arouse little
concern; the data would be considered
similar because both results are
"small". Detection limits for the
Canister/TO-14 Method (about 0.2 ppB)
prevent making such comparisons at sig-
nificantly lower concentrations. For
data taken with CPSil19CB columns
(TABLE 4),. agreement was much better,
because those columns retain compounds
longer and resolve them better, so peaks
are more likely to be identified and in-
tegrated properly. Agreement for ben-
zene and styrene was poorer than for
other compounds because benzene was lost
in the tail of the initial peak on every
run, while styrene was crowded by an ar-
tifact peak produced by column bleed.
PGC performance could most readily be
improved by using a column with better
resolution and less bleed, perhaps a
thicker-phase CPSilSCB, which would pro-
256
-------
vide better resolution of early-eluting
compounds and sufficient space between
later peaks to accommodate the minute-
long baseline disturbance which erupts
when backflush resumes. Improvement of
resolution will ultimately be limited by
flow system configuration. Another ad-
vantage of using a column with less
bleed would be that operation at higher
gain could result in lower detection
limits.
CONCLUSIONS
Portable gas chromatographs can rapidly
produce reasonable estimates of ambient
background concentrations of many vol-
atile nonpolar and semi-polar organic
air pollutants which ionize below 10.6
electron-volts. Because they process
data immediately, they are useful for
evaluation of hazardous waste sites,
chemical spills, and other sources of
airborne organic vapors. PGC data gen-
erally agree well with data from the
Canister/TO-14 Method, which provides
further indication that the latter is
generally valid for sampling atmospheres
not contaminated with highly reactive
compounds, even when analyses are de-
layed. Combined Canister/PGC analyses
should be used at uncharacterized sites
or where highly reactive compounds are
suspected. Positive interferences could
affect either PGC or canister data, but
negative interferences might be less
likely to influence PGCs because they do
not store or preconcentrate samples.
Furthermore, when analyses using dif-
ferent sampling methodologies produce
similar results, a preponderance of ev-
idence is created that sampling errors
did not occur and that data are sub-
stantially correct. Comparison of can-
ister and PGC sampling should be exten-
ded to include additional classes of
compounds, especially polar compounds.
REFERENCES
1. Compendium of Methods for the Deter-
mination of Toxic Organic Compounds
in Ambient Air. Environmental Pro-
tection Agency, Atmospheric Research
and Exposure Assessment Laboratory,
Research Triangle Park, NC 27711.
EPA-600/4-84-017. June 1988.
2. Verner, P. J. Chromatogr. 1984,
300, 249-264.
3. Driscoll, J. N. J. Chromatogr.
Sci., 1985, 23, 488-492.
4. Driscoll, J. N.; Atwood, E. S.; He-
witt, G. F. Ind. Res. Dev. , 1982,
24, 188-191.
5. Cox, R. D.; Earp, R. F. Anal.
Chem., 1982, 54, 2265-2270.
6. Rudolph, J.; Jebsen, C. Int. J.
Environ. Anal. Che., 1983, 13, 129-
139.
7. Nutmagul, W.; Cronn, D. R.; Hill, H.
H., Jr. Anal. Chem., 1983. 55,
2160-2164.
8. Langhorst, M. L. J. Chromatogr.
Sci. , 1981 , 19, 98 - 103.
9. Leveson, R. Ger. Offen. DE 3031358,
3-19-83. Leveson, R. C. US-
4398152, 8-9-83. Leveson, R. C.;
Barker, N. J. CA 1158891 A1. 12-20-
83.
10. Barker, J. J.; Leveson, R. C. Am.
Lab., 1980, 12, 76.
11. Leveson, Richard C.; Barker, Ni-
cholas J. Proc. of the Annu. ISA
Anal. Instrum. Symp.. 27th, St.
Louis, MO, Mar. 23-26 1981. Paqes
7-12.
12. Collins, M.; Barker, N. J. Am.
Lab., 1983, 15, 72.
13. Clark, A. I.; Mclntyre, A. E.; Les-
ter, J. N.; Perry R. Intern. J. En-
viron. Anal. Chem., 1984, 17, 315-
326.
14. Berkley, R. E. Evaluation of Photo-
vac 10S50 Portable Photoionization
Gas Chromatograph for Analysis of
Toxic Organic Pollutants in Ambient
Air. EPA/600/4-86/041. PB87-132858.
15. Lipsky, D. Proceedings of the APCA
Mid-Atlantic States Section Confer-
ence, Wilmington, DE April 18-19,
1983. Paper D.
16. Hawthorne, A. R.; Matthews, T. G.;
Gammage, R. B. Proceedings, 78th
Annual Meeting - APCA, Detroit. Ml
June 16-21, 1985. Paper 85-30B.
17. Jerpe, J.; Davis A. J. Chromatogr.
Sci., 1987, 25, 154-157.
18. Oliver, K. D.; Pleil, J. D.; McClen-
ny, W. A. Atmos. Environ., 1986,
20, 1403-1411.
257
-------
19. Gholson, A. R.; Storm, J. F.; Jayan- 21. Berkley, R. E.; Varns, J. L.; Mc-
ty, R. K. M.; Fuerst, R. G.; Logan, Clenny, W. A.; Fulcher, J. Proceed-
T. J.; Midgett, M. R. JAPCA 1989, ings of the 1989 EPA/AWMA Symposium
39, 1210-1217. on Measurement of Toxic and Related
Air Pollutants, AWMA, Pittsburgh.
20. Berkley, R. E. Field Evaluation of PA, 1989, pp. 19-26.
Photovac 10S50 Portable Photoion-
ization Gas Chromatograph for Anal-
ysis of Toxic Organic Pollutants in
Ambient Air. EPA/600/D-88/088.
TABLE 1. MOBILE PGC AND CANISTER SAMPLING AT MARCUS HOOK, PENNSYLVANIA
April 25, 1990. PGC in van with probe one meter above roof on upwind side.
CPSilSCB column. Concentrations are parts per billion by volume.
Tri- Tetra-
chloro- chloro- Chloro- Ethyl- m,p-
Benzene ethylene Toluene ethylene benzene benzene Xylene o-Xylene Styrene
Market Street at Railroad Overpass. 77°C.
PGC
CAN
PGC
CAN +
PGC
CAN +
Rt.
PGC
CAN
Rail
PGC
CAN +
6
4
7
6
10
ND
.8
.86 2
.7
.59 2
.2
1 3 at Trai 1 er
1
road
4
4
ND
.6
Station
.83 1
.7
ND
ND
. 18
ND
.20
ND
Park
ND
ND
*
15.9
*
15.6
*
20. 1
Street,
*
3.7
SW Parking Lot
.99
ND
*
7.6
4.90
0. 1
ND
0. 1
ND
0. 1
Trai ner ,
ND
0.4
. 77°C.
ND
0.6
0.53
ND
0. 13
ND
ND
ND
PA. 77°
ND
ND
ND
ND
ND
I .4
ND
2.3
0.47
3.4
C.
0.03
0.5
ND
1 .0
6
2
7
7
13
0
1
0
3
ND
. 1
.82
.5
. 14
.0
.84
.8
. 63
.4
ND
ND
ND
1 .4
ND
ND
ND
ND
ND
ND
ND
1 .9
0.36
2.5
ND
4. 1
ND
0.8
ND
1 .7
+ An appreciable concentration of hydrocarbons (not calibrated) was observed
in the canister sample.
* Toluene detection by PGC prevented by incorrect placement of valve time.
ND Not detected. Peak was absent or smaller than 5 millivolt-second.
258
-------
TABLE 2. PGC AND CANISTER DATA AT HAZARDOUS WASTE SITES IN NORTHERN DELAWARE
April, 1989.
mounted i
column .
Samples taken
n a van with probe
Concentrations
Benzene
April 5,
* PGC
CAN
1989
0.59
0.93
Tri-
ch 1 oro-
ethy 1 ene
Grantham
ND
ND
are
at Superfund haza
one meter above
rdous waste sites.
roof on
parts per bi 1 1 ion
Tetra
_
chloro- Chi
To 1 uene ethy 1
Lane
1
ND
.00 <0
by vol
oro- Ethy
upwi
ume .
1-
ene benzene benzene
INT
. 15
0
.60
ND
ND
ND
nd s i de
m.p-
Xy 1 ene
ND
0.62
PGC was
CPSil
Styrene
ND
ND
5CB
o-Xy 1 ene
ND
0.38
Army Creek
* PGC
CAN
Delaware
* PGC
CAN
April 6,
PGC
CAN
PGC
CAN
0.39
0.75
Sand
ND
0.45
1989
ND
1 .00
ND
0.50
ND
ND
& Gravel
ND
ND
0
0
ND
.69 <0
ND
.27
INT
. 15
I NT
ND
0
0
<0
.54
ND -
-------
TABLE 3. COLOCATED AND UPWIND/DOWNWIND PGC AND CANISTER OPERATION IN USSR
Vilnius June 1989. Colocated: PQCs in mobile laboratory. Probes 2.5 cm
apart 18 m from highway. Canisters sited on same side of road as PGCs and
filled continuously between 1631 and 1815. Data not shown if either PGC
recalibrating. Upwind/downwind: PGC-1 in van 12 meters downwind of roadway
with probe extended 1.5 meters above roof. PGC-2 in mobile laboratory.
Canisters downwind of road and filled continuously between 1100 and 1300.
CPSil19CB columns. Concentrations are parts per billion by volume.
Start
Time
COLOCATED
1631
1701
1716
1731
1801
Benzene
Toluene
Ethyl-
benzene
m, p - X y 1 e n e
DATA
(D
0.8
June
(2 )
0.8
1, 1989
2.5
(2)
2.8
1.3 1.0
0.6 0.7
1.1 1.0
0.7 0.8
2.7 2.7
2.1 2.1
2.4 2.2
2.3 2.6
ND
ND
ND
ND
ND
(2}
ND
ND
ND
ND
ND
(D
0.5
(2)
ND
0.7 ND
0.7 ND
0.9 ND
0.5
N'D
Average PGC values during the canister sampling oeriod
0.9 0.8 2.4 2.5 0.0 0.0 0.7 0.0
Canister sample values (distance from roadway in meters)
(3) (10) (3) (10) (3) (10)
(3)
2.1
(10)
1.1
3.1 1.3
0.4 0.2
UPWIND/DOWNWIND DATA June 2, 1989
(1) (2) (1) (2)
1229 1.3 4.5
1244 6.3
1259 3.3 0.4 7.4
(D
ND
ND
(2)
ND
ND
1.2 0.5
(1)
2.5
3. 1
1244 6.3 2.0 ND ND
1259 3.3 0.4 7.4 1.3 ND ND 3.1 ND
Average portable chromatograph values during the canister sampl
2.3 3.3 5.9 1.6 0.0 0.0 2.8 0.0
Canister sample values (distance from roadway in meters)
(3) (10) (3) (10) (3) (10) (3) (10)
2.1 1.1 3.1 1.3 0.4 0.2 1.2 0.5
o-Xy1ene
( 1 ) ( 2 )
ND 0.4
ND
ND
ND
ND
0. 1
0. 1
ND
ND
0.0 0.1
(3) (10)
0.5 0.2
( 1 ) ( 2 )
ND
ND
ND ND
ing period
0.0 0.0
(3)
0.5
(10)
0.2
ND
Not detected. Peak was absent or smaller than 5 millivolt-second.
260
-------
TABLE 4. SIDE-BY-SIDE PGC AND CANISTER DATA AT LAFAYETTE, GEORGIA
June 6, 1990. Shaver's Farm Superfund Site. PGCs in van were moved to several
sites. CPSil19CB columns. Concentrations are parts per billion by volume.
Tri- Tetra-
chloro- chloro- Chloro- Ethyl- m,p-
Benzene ethylene Toluene ethyiene benzene benzene Xylene o-Xylene Styrene
PGC-1
PGC- 2
CAN
PGC-1
PGC- 2
CAN
PGC-1
PGC- 2
CAN
PGC-1
PGC -2
CAN
PGC-1
PGC- 2
CAN
PGC-1
PGC- 2
CAN
ND
ND
0.7
ND
ND
0.3
ND
ND
0. 1
ND
ND
3.0
ND
ND
0. I
ND
ND
0. 1
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
1 . 10
1 .75
1 .0
0.99
0. 18
0.7
0.59
ND
0.4
0.32
0. 10
0. 2
0.08
ND
0. 1
0. 19
ND
0. 3
ND
ND
0. 1
ND
ND
0.2
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
0.67
0.62
1 .0
0.46
0.69
0.8
ND
ND
0.2
ND
ND
0. 1
ND
ND
ND
ND
ND
0.2
1 .31
1 .84
1 .6
0.28
0.58
0.8
ND
ND
0.4
0.04
0.07
0.2
ND
ND
ND
ND
ND
0.4
0. 12
ND
0.8
ND
ND
0.4
ND
ND
0.2
ND
ND
0.2
ND
ND
0. 1
ND
ND
0.3
*
18.03
5.9
*
13.68
4.5
*
ND
0.2
*
ND
0.2
*
ND
ND
*
ND
0.4
* PGC-1 was not calibrated for styrene because of a persistent interfering
peak probably caused by column deterioration.
ND Not detected. Peak was absent or smaller than 5 millivolt-second.
261
-------
TABLE 5. TANDEM PGC DATA AND CANISTER DATA IN
RESEARCH TRIANGLE PARK, NORTH CAROLINA
March 23, 1990. PGC-1 analyzed vinyl chloride and 1 , "i-dich loroethy 1 ene with
KCl/Alumina PLOT column. PGC-2 analyzed other compounds with a CPSil5CB column.
PGCs in car with probes 1.5 meters above the on upwind side, 40 meters downwind
of dry-cleaning plant. Concentrations are parts per billion by volume.
1,1-Di- Tetra-
Vinyl- chloro- chloro-
chloride ethylene Benzene Toluene ethylene m,p-Xylene Styrene o-Xylene
PGC
CAN
PGC
CAN
0.01
ND
ND
ND
0.02
ND
0.01
ND
ND
0.63
ND
0.76
ND
0.56
ND
0.62
1 .41
3.38
4.09
3. 15
ND
0.35
ND
0.29
ND
0.20
ND
ND
ND
0. 20
ND
< 0. 20
ND
Not detected. Peak was absent or smaller than 5 millivolt-second.
TABLE 6. AVERAGE ABSOLUTE DIFFERENCES BETWEEN PGC AND CANISTER DATA
Absolute values of differences between PGC and canister results for each
compound were averaged. CPSilSCB data taken from TABLES 1, 2, and 5.
CPSil19CB data taken from TABLE 6, in which two PGC values for each analysis
were averaged. CPSilSCB is methyl si "ii cone. CPSil19CB is 7% cyanopropyl-
silicone, 1% pheny1si1icone, 85% methyIsi1icone. and 1% vinylpolysiloxane.
Phase thicknesses 2 micrometers. Differences have dimensions of Darts per
billion by volume.
Compound
Benzene
Tr ich loroethy lene
Toluene
Tetrach loroethy lene
Ch lorobenzene
Ethy Ibenzene
m,p-Xy lene
o-Xy 1 ene
Styrene
UO 1
CPSi 15CB
1 .49
2.03
0.75
1 .27
1 . 17
0.97
1.69
0.40
1 .53
umn
CPSi 1 19CB
0.72
*
0. 15
0.05
*
0.20
0.30
0.31
3.69
No data were available for these compounds.
262
-------
DISCUSSION
Q)WARDFURTAUGH:WealsohaveaPhotovaclOS70.Iuseditinasmoking
lounge and at a gain of 100, there was a monster peak occurring near the retention
time of toluene. Any suggestions what it could have been? The Photovac people
Ait I've talked to haven't been able to shed much light on it.
RICHARD BERKLEY: Indoor air is a pretty tough thing to deal with. You
normally see ambient background levels of things like toluene as a result of single
photon absorptions. In indoor air you can have ppm concentrations so you can
teethings which are ionized in double photon absorptions. You can, for example,
calibrate things like carbon tetrachloride and chloroform, which you can't see at
iBatambient background levels. So, in indoor air all bets are off, and I have seen
some horrendous things in indoor air which are probably relatively high levels
ofthingsthat the instrument normally can't see. If it didn't have the retention time
of toluene, and if you were using a constant temperature column accessory the
dunces are very good that it was not toluene. It may be a much larger level of
something else.
TOM SPITTLER: We just did an air study of our building in Boston using the
Photovac. And we found 1 to 2 ppb of benzene and toluene inevery place because
it's > very well ventilated building. But in the smoking room we found about 100
ppbof toluene and about 50 of benzene. There wasn't any question, the retention
tees matched beautifully. We took samples back and confirmed them on GC/
MS. You get benzene and toluene in all smoking rooms. I'm not sure why your
(leak wasn't exactly there, but I bet anything that's what it was. A question
tough: you were using canisters and the Photovac with what? Occasional
sampling or regular sampling? How often did you sample with the Photovac in
onfcr to cover the period of time you were drawing the canister sample?
RICHARD BERKLEY: In most cases what I did was take a canister grab
sample by holding the canister within ten centimeters of the tip of the Photovac
Make and opening the can so that it filled during the same time, during the minute
or so that the Photovac pump was running. These samples are necessarily
nonequivalent. In the canister you get six liters, and you take a representative
sample of that to analyze it. The Photovac takes a milliliter or something that
happened to be flying through at the moment when it decided to inject. These are
not equivalent, but if the same air is being sampled they ought to resemble each
other.
TOM SPITTLER: Yes, I agree. I think it's really a nice correlation.
RICHARD BERKLEY: So, in all cases expect variances while taking those
{tab samples. Variances were seen with two-hour integrated canister samples,
and we were taking Photovac runs during the time.
TOM SPITTLER: You just averaged them then?
RICHARD BERKLEY: Well, the canister samples were shown as dotted and
•lashed lines because they were the time-integrated samples. If you could go back
and compare those slides, you'd find that all those dotted and dashed lines were
a the same level on all of those slides. We couldn't quite figure out how to show
4e continuity there.
TOM SPITTLER: No, I thought it was really nice data. This afternoon a couple
of guys from the Regional Lab up in Boston are going to show some Photovac
versus canister standards and calibrated by different techniques. They are
actually directly comparable samples, and you see the same basic kind of
correlation. It may be a little tighter now because they're sampling exactly the
same way and they're sampling the same known mixture of air.
RICHARD BERKLEY: Something I forgot to mention and it'll be important
to some people, we are using canisters to hold our calibration standards, and
we're preparing the standards the same way we prepare the standards for the
method that is used to analyze the canisters. There is no independence on that
point. These two methods are locked together, and if we make a mistake on one,
we make a mistake on the other. What's independent here is sampling methodology,
and I should have said that.
JOSEPH EVANS: My question pertains to detection limits. I notice that you're
down measuring at very low levels (1,2 ppb). Your worst agreement was at those
levels. When you got to the higher levels you had much better agreement. And
I was wondering about how close you were to your detection limits for the two
different methods?
RICHARD BERKLEY: Well, there are two limits to talk about here. One of
them is detection limit, and for single photon ionizations, compounds that ionize
well below 10 eV, such as benzene, its homologs and the chloroethylenes. We
measured detection limits by extrapolation, three times the baseline noise, using
an old 10A10 with a gain turned all the way up, and it appeared that the absolute
detection limit was somewhere in the neighborhood of 18 femtograms. That
would translate out to down in the neighborhood of 1/100 ppb in a 1 mL sample.
That's just a detection limit.The instrument in fact will refuse to process any peak
that is smaller than five millivolt seconds. And of course, when we did that
detection limit we were only extrapolating — our smallest sample was 1.6
picograms, and it produced a peak of about 2.3 volt seconds. We were no where
near this extrapolated detection limit with any sample we actually delivered to
the instrument. So, we're just guessing. But, we do have a substantial basis to
guess that a 5 millivolt second peak is way above that. And all you have to do if
you want to really get tough about what the detection limit is, is to run a sample
on a blank library, then shift to a calibrated library and calculate how much it
would take to make a five millivolt second peak, assuming linear response.
JOSEPH EVANS: What levels were your calibration standards?
RICHARD BERKLEY: We generally try to use between 10 and 20 ppb. If
something is very convenient to prepare like chlorobenzene or tetrachloroethylene,
we like to use one of them on one instrument and the other one on the other
instrument, because there is some tendency, if the calibration gas valve is a little
bit weak, to have some carryover contamination, usually no more than 0.5 ppb.
You do need to look at that for your standard, whatever your standard compound
JOSEPH EVANS: You were actually measuring below your lowest calibration
standard?
RICHARD BERKLEY: We're using a single point calibrations. We did a lot of
work on this thing early on and found that we were getting pretty consistent linear
responses from as low a sample as we could inject all the way up to higher than
we could inject.
263
-------
HIGH SPEED GAS CHROMATOGRAPHY FOR AIR MONITORING
Levine, S.P. (A,*), Ke, H.Q. (A), Mouradian, R.F. (A)
Berkley. R. (B) and Marshall, J. (C)
A) Department of Environmental and Industrial Health,
University of Michigan, Ann Arbor, Michigan 48109-2029
(B) U.S. EPA, AREAL/MRB (MD-44), 79 TW Alexander Dr.,
Research Triangle Park, NC 27709
(C) HNU Systems, 160 Charlemont, Newton Highlands, MA 02161
(*) Author to whom correspondence should be addressed.
Abstract
Gas chromatography has the potential
to be a much faster method of
separation than is usually realized.
If column operating conditions are
optimized for speed and injection
band width is minimized, some simple
separations can be completed in a few
seconds. In the work described here
the system was evaluated using common
organics including alkanes,
aromatics, alcohols, ketones and
chlorinated hydrocarbons.
Quantitative trapping and reinjection
was achieved for all tested
compounds. Limits of detection (LOD)
for many compounds, based on a 1 cm3
gas sample, were less than 1 ppb, but
for one carbon-chlorocarbons the LOD
when using a flame ionization
detector was inadequate. By using
the cold trap inlet with a low dead
volume detector and a high speed
electrometer, the efficiency
available from commercial capillary
columns can be better utilized and
retention times for some routine
separations may be reduced to a few
seconds.
Introduction
Gas chromatography (GC) is often used
for routine, repetitive analysis of
simple mixtures. For some of these
applications, the use of 2 to 5 m
capillary columns operated at linear
velocities of 100 to 200 cm/s offers
the possibility of greatly decreased
analysis times. This potential for
high speed analysis has been
documented in the literature (1-7).
Under optimal conditions, a 0.25 mm
i.d. column should be capable of
achieving 5000 to 7000 effective
plates with retention times of 5 to
10 seconds (4,8). Although this
number of plates is low compared to
most capillary systems, it is
comparable to the number of plates
achieved by many packed column
systems with retention times of
several minutes or more. Therefore,
some routine GC separations that are
currently performed using packed
columns or non-optimized open tubular
columns could be performed much
faster with a capillary system that
is optimized for speed.
While the theoretical potential of
capillary columns for high speed
analysis is well known, limitations
in commercially available equipment,
especially inlet systems, have
prevented general application of high
speed techniques. With most
commercial instruments, the major
265
-------
factors that limit analysis speed are
the width of the initial band
produced by the inlet system and the
response time of the electrometer.
Efficient separation with retention
times of 5 to 10 seconds and a column
diameter of 0.25 mm requires an
initial band width of about 20 ms or
less and an electrometer response
time of about 5 ms. For purposes of
comparison, most capillary GC systems
produce injection band widths of 50
to 500 ms and feature electrometer
response times of 150 ms or longer.
In response to the requirement for
narrow injection bands, a number of
experimental inlets have been
described (5, 9-13). Our group has
described a prototype cold trap that
was used as a vapor collection device
and which may also serve as a
focusing system for rapid analysis of
simple mixtures (14-15). The design
reported by our group, which expanded
on the innovative work of Hopkins and
Pretorius (16), featured a cold trap
that was cooled by a continuous flow
of cold nitrogen, and was resistively
heated using a current pulse. This
design was a marked improvement over
that reported earlier, which had a
number of unrecognized serious flaws
that prevented reliable and/or
quantitative operation (17-18). More
recently, van Es et al described a
fast GC system that utilized a
similar inlet (19). In their design,
a 50 micron capillary column was used
for the separation.
Experimental Section
The design and operation of the cold
trap is given in detail elsewhere
(14,15), and is shown schematically
in Figure 1.
Operating conditions and
chromatographic equipment. All
chromatograms were collected
isothermally at column temperatures
of 35 to 60 °C using a 5 m long, 0.25
mm i.d. fused silica column with a
0.1 micron bonded methyl silicone
stationary phase (Quadrex). The
carrier gas was hydrogen, which was
supplied at a flow rate of 2.5 to 3
ml/min to produce linear velocities
of 85 to 102 cm/s. The injector and
detector were heated to 225 °C. A
flame ionization detector (FID) was
used in all experiments. To minimize
the effective dead volume, the column
was moved close to the base of the
flame. Both a Varian 3700 or an HNU
301 GC were used.
For trap recovery studies, test
mixtures were prepared either
without solvent or in high purity
carbon disulfide provided by The Dow
Chemical Company. The injection
volume was 2.5 uL in all cases and
the split ratio ranged from about
50:1 to 500:1 depending on the sample
concentration. For vapor studies,
samples were injected in humidified
or laboratory air in volumes of
0.025-1.0 cm3.
Results and Discussion
Design Considerations. A number of
design considerations were found to
be important in determining the
durability and performance of the
system. The choice of trap material
and dimensions affects durability and
reinjection performance. An ideal
material would have high electrical
resistivity, low chemical activity, a
low coefficient of thermal expansion,
would be highly malleable and would
not work harden. A number of
materials, including stainless steel,
nickel, platinum, Monel 400, and an
alloy of thirty per cent copper -
seventy per cent nickel were
evaluated for use as trap tubes. The
work reported here was done using a
trap made of Monel 400. Stainless
steel, which was used in some early
studies (17, 18), is the least
desirable choice because of its
tendency to work harden and become
brittle. For a trap made of hard-
tempered Monel 400 with an internal
diameter of 0.25 mm, a wall thickness
of 0.18 mm provided a good
combination of strength and
performance.
266
-------
Trapping and Reinjection Efficiency.
Cold traps have been used in GC for
many years (19-23). Since the short,
open tubular trap used in these
experiments may be less efficient
than some other designs (23), a
careful evaluation of trapping
efficiency was necessary.
In order to test trapping and
reinjection efficiency, samples were
injected without using the cold trap
and average peak areas were
calculated for each compound. In
addition to comparing peak areas
obtained with and without trapping,
the FID response was monitored during
the entire process to allow any
breakthrough of the sample to be
detected. At temperatures of -100 °c
or colder, each of the tested
compounds was guantitatively trapped
and reinjected. Peak area
reproducibility for all compounds was
very good with coefficients of
variation ranging from 1 to 5 per
cent, or less in all cases in which
trapping was used.
Compounds tested were (given in order
of increasing boiling point):
isoprene, pentane, dichloromethane,
acrolein,chloroform,methanol, hexane,
carbon tetrachloride, acrylonitrile,
2-butanone,benzene,propanol, heptane,
i-octane, toluene, n-butanol,
tetrachloroethylene, octane, m- & o-
xylene, nonane, 4-ethyltoluene, and
1,3-dichlorobenzene. Detailed results
are given elsewhere (15). Trapping
efficiency was also measured for 1%
solutions of aromatics prepared in
carbon disulfide. The trapping
efficiencies obtained in those
experiments were not significantly
different than those measured without
solvent. These materials can be
effectively trapped and reinjected at
temperatures of -100 °C. However,
trapping behavior is not easily
predicted on the basis of boiling
point or freezing point, and in most
cases an effective temperature must
be experimentally determined for each
type of sample. Highly volatile
materials, which may be gases at room
temperature, and low volatility
materials, which may be difficult to
revaporize, have not yet been tested
and may be difficult to trap and
reinject with this system.
Limit of Detection (LOD). For
monitoring volatile organics in
ambient or workplace air, the LOD of
the method must be very low. As of
early November, 1990, the LCD's for
pentane, hexane, heptane, octane,
benzene, toluene, xylene,
ethylbenzene, 4-ethyltoluene, 1,3,5-
/1,2,4-trimethylbenzenes, and
chlorobenzene have been measured and
been shown to be in the range of 0.2
- 5 ppb, with the most recent results
all being <1.0 ppb. (The drop in LOD
has occurred as a result of improved
methodology as work has proceeded
over the past few months. There has
not been time to re-do some of the
earlier work.)
All of these values were determined
based on an injection of a maximum of
1 cm3 of air, and the use of an FID.
The LOD was calculated based on a
definition of three times the
standard deviation of the noise.
One of the major factors contributing
to the reduced LOD was the
optimization of the custom-designed,
high speed electrometer supplied for
this project by HNU Co. A filter
setting of 12 Hz was found to be
optimal for GC peaks in the retention
time range of 5-10 seconds.
Note that these LOD's are not
achievable for one carbon-
halocarbons. LOD's in the sub-20 ppb
range for certain halocarbons will
only be achievable with the use of an
electron capture detector (BCD).
Unfortunately, an BCD has, of
necessity, a certain internal volume
that may significantly spread peaks,
and reduce the advantage of the Fast-
GC method. This may require assays to
be performed on a 30-60 second basis,
rather than on a 5-10 second basis.
267
-------
In addition, it is important to
remember that the Fast-GC technique
trades chromatographic resolution for
speed. Although the cost of this
trade is reduced by tuning the column
for high speed, low retention time
use (8,14), the separation of
components of complex mixtures may
not always be possible.
Further, the limitations imposed by
the use of an isothermal GC method
(necessitated by the short analysis
times) limit the ability to monitor
compounds of widely differing boiling
point simultaneously. While this
might be overcome by flow-programming
methods, the extent to which such
strategies will allow effective
ambient air monitoring is unknown at
this time.
Acknowledgements
The authors acknowledge Lauri
Mendenhall and George Capps of
Prototype Design Inc. for engineering
and technical assistance in the
development of the capacitor
discharge power supply and
temperature measurement devices.
This research was supported by U.S.
EPA (AREAL/MRB) cooperative agreement
CR-817123-01-0. Earlier work leading
to this stage had been supported by
the Centers For Disease Control,
National Institute for Occupational
Safety and Health Grant R-01-OH02303,
the U.S. EPA (OER) R814389-01, and
The Dow Chemical Company Health and
Environmental Studies Laboratory.
References
1. D. H. Desty. capillary columns:
Trials tribulations and triumphs.
Advances in chromatographv. Vol 1.,
J. C. Giddings and R. A. Keller eds.,
Marcel Dekker, NY, 1965, pp. 199-228.
2. D. H. Desty, A. Goldup and W. T.
Swanton. Performance of coated
capillary columns. Gas
Chromatography. N. Brenner, J. E.
Callen and M. D. Weiss eds., Academic
Press, New York, 1962, pp. 105-135.
3. J. C. Sternberg. Extra column
contributions to chromatographic band
broadening. Advances jn
Chromatoqraphy Vol 2., J. c. Giddings
and R. A. Keller, eds., Marcel
Dekker: N.Y. 1966, pp. 203-270.
4. G. Caspar, R. Annino, c. Vidal-
Madjar and G. Guiochon. Influence of
instrumental contributions on the
apparent column efficiency in high
speed gas Chromatography. Anal. Chem.
50: 1512-1518 (1978).
5. G. Caspar, P. Arpino and G.
Guiochon. Study in high speed gas
Chromatography. i Chromatogr. sci .
15: 256-261 (1977)..
6. A. van Es, J. Janssen, R. Bally,
C. Cramers and J. Rijks. Sample
introduction in high speed capillary
GC ; Input band width and detection
limits. HRC&CC. 10: 273-279 (1987).
7. C. P. M. Schutjes, E. A.
Vermeer, J. A. Rijks and C. A.
Cramers. Increased speed of analysis
in isothermal and temperature-
programmed capillary GC by reduction
of the column inner diameter. J.
Chromatoqr. 253: 1-16 (1982).
8. R. Villalobos and R. Annino. The
computer aided optimization of
capillary columns for minimum time
analysis and minimum detectability.
HRC&CC. 12: 149-160 (1989).
9. R. L. Wade and S. P. Cram.
Fluidic logic sampling and injection
system for gas Chromatography. Anal.
Chem. 44: 131-139 (1972).
10. R. Annino and J. Leone. The use
of coanda wall attachment fluidic
switches as GC valves. J. Chromatogr.
Sci. 20: 19-26 (1982).
11. C. P. M. Schutjes, C. A.
Cramers, c. Vidal-Madjar and G. J.
Guiochon. fast fluidic logic
injection at pressures up to 25 bar
in high-speed capillary GC J.
Chromatogr. 279: 269-277 (1983).
12. R. J. Jonker, H. Poppe and J. F.
K. Huber. Improvement of speed of
separation in packed column GC. Anal.
Chem. 54: 2447-2456 (1982).
13. R. Tijssen, N. van den Hoed and
M. E. van Kreveld. Theoretical
aspects and practical potentials of
rapid gas analysis in capillary GC.
Anal. Chem. 59: 1007-1015 (1987).
268
-------
14. Mouradian, R.F., Levine, S.P.,
Sacks, R.D. and Spence, M.W.
Measurement of Organic Vapors at Sub-
TLV Concentrations Using Fast Gas
Chromatography. Amer Ind Hya Assoc
J. 51:90-95 (1990).
15. Mouradian, R.M., S.P. Levine and
R.D. Sacks. Limits of Detection and
Recoveries for Fast-GC. i..
Chromatoar.. 28: 643-648 (1990)
16. B. J. Hopkins and V. J.
Pretorius. Rapid evaporation of
condensed GC fractions. «L=_
Chromatoar. 158: 465-469 (1978).
17. B. A. Ewels and R. D. Sacks.
Electrically heated cold trap Inlet
system for high-speed GC. Anal. Chem.
57: 2774-2779 (1985).
18. L. A. Lanning, R. D. Sacks, R.
F. Mouradian, S. P. Levine, and J. A.
Foulke. Electrically heated cold trap
inlet system for computer-controlled
high-speed gas Chromatography. Anal.
Chem. 60: 1994-1996 (1988).
19. A. van Es, J. Janssen, C.
Cramers and J. Rijks. Sample
enrichment in high speed narrow bore
capillary gas Chromatography. HRC&CC.
11: 852-857 (1988).
20. G. Schomburg, H. Husmann and F.
J. Weeke. Aspects of double-column GC
with glass capillaries involving
intermediate trapping. J. Chromatogr.
112: 205-217 (1975).
21. J. A. Rijks, J. Drozd and J.
Novak. J. Versatile all-glass
splitless sample-introduction system
for trace analysis by capillary GC.
J. Chromatoar. 186: 167-181 (1979).
22. D. Kalman, R. Dills, C. Perera
and F. DeWalle. On-column cryogenic
trapping of sorbed organics for
determination by capillary GC. Anal.
Chem. 52: 1993-1994 (1980).
23. J. W. Graydon and K. Grob. How
efficient are capillary cold traps?
J. Chromatoar. 254: 265-267 (1983).
24. Mouradian, R.F., S.P. Levine,
H.Q. Ke and H.H. Alvord. Measurement
of Volatile Organics at Parts Per
Billion concentrations Using a Cold
Trap Inlet and High Speed Gas
Chromatography. submitted to J_. Air
Haste Manag. Assoc.
269
-------
Figure 1
Fast-GC system;
A: syringe or gas sampling
loop injection port;
B: silica transfer line;
C: low dead volume unions;
electrical contacts;
trap tube;
upper chamber cold trap;
lower chamber cold trap;
baffle;
capillary column
flame ion. detector;
K: capacitor power supply
270
-------
DISCUSSION
BANK WOHLTJEN: How much energy did your capacitive discharge heater
we?
STEVEN LEVINE: It's running about 30 to 70 volts discharge with a few tens
rfamps.
HANK WOHLTJEN: How big are the capacitors? Are they a tenth of a farad
ursomething like that?
SIEVEN LEVINE: All the details of the design is in that paper in Analytical
Chemistry.
HANK WOHLTJEN: You mentioned electric cooling of the trap. What do you
Jhiai you'd use for that, a refrigerator or a thermal electric?
STEVEN LEVINE: It would have to be a thermoelectric cooler. We are
investigating that at this moment.
JOHN SNYDER: I was curious as to the diameter of the columns you're using.
STEVEN LEVINE: They're just 0.25 mm columns. They're very traditional
columns. They're not megabore. They're not ultra small.
JOHN SNYDER: You also spoke about the dead volume in the detectors. Are
you modifying traditional detectors or are you making your own detectors?
STEVEN LEVINE: We have a 90 pi dead volume BCD from HNU Systems at
this point that we're working with. We feel that size is probably too big.
271
-------
SCREENING VOLATILE ORGANICS BY DIRECT SAMPLING ION TRAP AND
GLOW DISCHARGE MASS SPECTROMETRY*
Marcus B. Wise, G.B. Hurst, C.V. Thompson, Michelle V. Buchanan, and Michael R. Guerin
Analytical Chemistry Division
Oak Ridge National Laboratory
Oak Ridge, Tennessee 37831-6120
ABSTRACT
Two different types of direct sampling mass
spectrometers are currently being evaluated in our
laboratory for use as rapid screening tools for volatile
organics in a wide range of environmental matrices.
These include a commercially available ITMS ion trap
mass spectrometer and a specially designed tandem
source glow discharge quadrupole mass spectrometer.
Both of these instruments are equipped with versatile
sampling interfaces which enable direct monitoring of
volatile organics at part-per-billion (ppb) levels in air,
water, and soil samples. Direct sampling mass
spectrometry does not utilize chromatographic or other
separation steps prior to admission of samples into the
analyzer. Instead, individual compounds are measured
using one or more of the following methods: spectral
subtraction, selective chemical ionization, and tandem
mass spectrometry (MS/MS). For air monitoring
applications, an active "sniffer" probe is used to achieve
instantaneous response. Water and soil samples are
analyzed by means of high speed direct purge into the
mass spectrometer. Both instruments provide a range of
ionization options for added selectivity and the ITMS
can also provide high efficiency collision induced
dissociation MS/MS for target compound analysis.
Detection limits and response factors have been
determined for a large number volatile organics in air,
water, and a number of different soil types.
INTRODUCTION
Direct sampling mass spectrometry for the
measurement of trace levels of volatile organics in
environmental matrices has a wide range of important
field screening applications. These include the
measurement of volatiles in waters, soils, oily wastes, stack
emissions, and ambient air, among others. In addition, real-
time "sniffing" capability provides a convenient means of
detecting soil gas emissions, leaking waste containers, and
probing the atmosphere in enclosed storage facilities.
Because of their small size, relative simplicity,
ruggedness, and low power consumption, conventional
quadrupole mass spectrometers and quadrupole ion trap
mass spectrometers are especially attractive for
transportable field screening applications. In fact, several
commercial quadrupole based instruments are currently
available for field monitoring applications and recently,
several different research groups have been developing and
demonstrating transportable ion trap mass spectrometers for
on-site GC/MS applications (1-3).
This paper describes the use of an ion trap mass
spectrometer and a tandem source glow discharge mass
spectrometer for the direct measurement of ppb levels of
volatile organics in air, water, and soil. Because these
instruments do not use chromatographic separation prior to
admitting a sample into the mass spectrometer, the response
time is virtually instantaneous and accurate quantification of
target analytes can be accomplished in less than 2 minutes.
Although the tandem source quadrupole mass spectrometer
is somewhat limited in its ability to handle complex samples,
the ion trap mass spectrometer has the capability of
selective ion storage and multiple stages of collision induced
dissociation for much greater specificity.
Laboratory-based instruments arecurrently being used
to develop and validate methods for direct air monitoring
and the screening of water, soil and waste samples. A
transportable ion trap mass spectrometer for field use is
under construction in our laboratory and will be initially
tested in 6-9 months.
273
-------
EXPERIMENTAL
Instrumentation
Ion Trap Mass Spectrometer
All ion trap experiments were performed with a
Finnigan MAT Corporation ITMS ion trap mass
spectrometer. Our instrument is equipped with a
specially designed vacuum chamber which is
electropolished on the inside and pumped to high
vacuum with two air cooled 330 L/sec turbomolecular
pumps. The vacuum chamber and analyzer cell are
maintained at a constant temperature of 120° C by
means of infrared heating lamps which help to minimize
the adsorption of contaminants on the analyzer surfaces.
This instrument is also equipped with the necessary
hardware and software to perform electron impact (El)
and chemical ionization (CI), as well as selective ion
ejection, and collision induced dissociation multiple-step
(tandem) mass spectrometry experiments (MS/MS).
Control of the instrument and data acquisition are
performed with an IBM AT compatible computer using
software provided by the manufacturer.
The standard chromatographic interface provided
with the ITMS instrument has been replaced with a
custom designed interface developed in our laboratory.
This interface consists of a short length (14 inches) of
110 micron ID uncoated fused silica capillary tubing
which is maintained at atmospheric pressure at one end
and high vacuum at the other end. The high vacuum
end of the capillary is inserted directly into the ITMS
analyzer cell and the atmospheric pressure end is
connected to a quick-coupling device which allows rapid
switching of sampling modules for different monitoring
applications. The gas flow rate through the capillary
restrictor is approximately 0.5-1.0 mL/min. Because the
samples are introduced directly into the ion trap cell,
the manifold pressure is maintained at a lower pressure.
This is believed to help reduce deterioration of the
electron filament and the electron multiplier. For
example, even when sampling water-saturated air for
extended periods of time, the electron filament lifetime
has been approximately 6 months and the multiplier
lifetime has been in excess of 12 months.
ITMS Air Sampling Probe
For direct air monitoring experiments, a special
sampling system has been developed as shown by the
diagram in Figure 1. This system consists of an 1/4 inch
OD teflon transfer line which is connected at one end
to the air sample generation system and at the other
end to a sampling "cross" arrangement which allows
helium to be mixed with the air sample prior to entering
the ITMS. The helium is necessary as a buffer gas in
the ITMS to collisionally cool ions, thus reducing loss of
ions from the trap and improving the overall
performance. A pulsed valve is used to meter helium
into the air stream providing approximately an order of
magnitude increase in sensitivity relative to a fixed-ratio,
continuous mixing of helium with the air. A vent port also
located on the inlet "cross" of the sampling system allows
the gas stream to be continuously sampled at a high flow
rate, thus decreasing the response time for the mass
spectrometer. The other port of the inlet "cross" is
connected to a short section of uncoated fused silica
megabore capillary which is used as an "open/split" interface
with the ITMS by inserting 1 inch of the microbore capillary
restrictor into the other end of the megabore tubing.
Approximately 2 L/min of air is drawn through the
megabore tubing by means of a small sampling pump;
however, a metering valve located between the pump and
the splitter can be used to reduce the pumping speed if
desired. This combination of active pumping and the use of
the open/split capillary interface minimizes the dead volume
in the inlet system leading to a response time of only a few
seconds.
Purge Device for Water and Soil samples
For the measurement of volatile organics in water and
soil samples (slurries), the air sampling probe is simply
replaced with a high speed needle sparge purge device as
shown in Figure 2. This device accepts standard 40 mL
VOA vials which mount directly on the needle sparger. A
pressure regulator and a precision needle valve control the
flow of helium purge gas through the sample and the
purged components exit through a 10 inch length of
megabore capillary tubing. Normal helium flow rates vary
from 100 to 200 mL/min which efficiently purges the volatile
components from a room temperature sample in less than 5
minutes. The purge device connects directly with the
capillary restrictor interface in an open-split configuration
with a split ratio of approximately 100:1. The bulk of the
sample is diverted to the vent port. As an added feature
for screening applications, the vent port is capable of
accepting resin cartridges for trapping of components that
would normally be vented. This enables the collection of an
archived sample which may be sent back to a central
laboratory for confirmatory analysis by GC/MS.
Tandem Source Quadrupole Mass Spectrometer
The tandem-source quadrupole mass spectrometer
(TSMS) is a prototype instrument constructed using an
EXTREL C-50 quadrupole mass spectrometer as the basic
system. This instrument was configured with 3/4" diameter
rods for high transmission efficiency and a 300 watt RF
power supply for a maximum mass range of 500 amu.
Control of the instrument is provided by a Dell 325
computer using software written in our laboratory. An axial
El source was purchased with this instrument for testing
purposes and for generating conventional 70 eV electron
impact spectra.
In order to produce a versatile instrument for
environmental monitoring applications, the configuration of
the standard C-50 mass spectrometer was extensively
274
-------
modified. In addition to the axial El source which was
purchased with the spectrometer, a glow discharge
ionization source was designed and constructed for this
instrument. This source is housed in a differentially
pumped vacuum chamber which is separated from the
rest of the mass spectrometer by a 1.5 mm diameter
vacuum conductance limit as shown in Figure 3. The
glow discharge source is typically maintained at a
pressure of 0.25 torr while the analyzer is maintained at
2 x 10"5 torr. Ions generated by glow discharge
ionization pass through a lens assembly into the high
vacuum portion of the instrument where they enter the
lens assembly of the axial El source and are
subsequently focussed into the mass analyzer.
Air samples can be introduced into the tandem
source quadrupole mass spectrometer by two different
methods, either through the differentially pumped glow
discharge source chamber, or directly into the electron
impact source by means of a simple capillary restrictor.
Both inlet systems have been designed so that they are
directly compatible with the same sampling devices used
with the ion trap mass spectrometer. Thus, essentially
the same apparatus and experimental conditions are
used for direct purging of water and soil samples
regardless of the mass spectrometer used. The only
difference is the ability of the glow discharge ionizer to
sample air directly without the need for the air sampling
pump and open/split interface used with the ITMS.
Dynamic Sample Generator
A dynamic sample generation apparatus is used to
produce known concentrations of volatile organic
analytes in an air stream. This apparatus was used for
the determination of instrumental detection limits for
real-time air monitoring experiments. It basically
consists of a variable speed syringe pump and a dilution
air manifold. The syringe pump continuously meters
small amounts of organic compounds into a controlled
stream of air. Concentrations of the analytes can be
easily varied by adjusting the speed (metering rate) of
the syringe pump and/or by changing the flow rate of
dilution air through the manifold. Turbulent mixing of
the organic compounds and the dilution air occurs in the
manifold line which provides a homogeneous
concentration at the sampling ports.
." Components of the dynamic sample generator
include a Razel Instruments model A-99 syringe pump
equipped with a 5 mL syringe, a 100 psi air supply line
equipped with an on/off toggle valve and a precision
metering valve, a 1.5 m x 6 mm Teflon line (dilution
manifold), and two 1/4 inch Swagelock sampling ports.
The apparatus produces continuous and stable
generation of organic concentrations in air and also
; allows rapid changes in concentration without having to
wait excessively to reach a steady-state concentration.
Air containing the desired concentration of individual
organic compounds is typically generated by metering a
(1:1) water/methanol solution containing approximately 400
ug/mL of the organic compound into the dilution air stream
using the syringe pump. The flow rate of the syringe pump
can be continuously varied from 8.47 x 10"4 mL/min to
0.0503 mL/min. The dilution air flow is typically adjusted
for a rate of 25 L/min through the manifold. As this air
flows rapidly past the syringe pump needle, it quickly
vaporizes the volatile organics and the solvent. Liquid flow
from the syringe, however, must be maintained low enough
to prevent condensation in the system. By knowing the
concentration of the organic in the liquid solution, the flow
rate out of the syringe, and the flow rate of the dilution air,
the concentration of the organic compounds in the air can
be readily calculated. This assumes that there is minimal
adsorption of analytes on the walls of the manifold and
complete vaporization of the liquid into the dilution air.
Operating Conditions
Ion Trap Mass Spectrometer
Most of the ion trap data presented in this paper was
generated using electron impact ionization conditions. Scan
functions for the acquisition of mass spectra were written
using the scan function editor program supplied with the
commercial software. Typically, for optimum sensitivity the
electron ionization time was 50 msec. Low mass cut-off was
60 amu, preventing the storage of ions due to water and air.
The mass scan range was approximately 50 to 200 amu
which enabled the detection of major ions for each of the
volatile organic compounds. In order to improve the signal-
to-noise ratio, 16-25 microscans were averaged per displayed
scan. Axial modulation was used for all experiments in
order to achieve optimum instrument performance. Helium
buffer gas was admitted into the system exclusively through
the sample transfer line.
Tandem Source Quadrupole Mass Spectrometer
The glow discharge ionization source is specifically
designed for high sensitivity direct air monitoring
applications. Air is admitted into the ionization region
through a metering valve at a flow rate of 0.5-1.0 standard
mL/min while a 160 L/min roughing pump maintains the
pressure in the ionizer at a constant 0.25 torr. Coaxial
ionization electrodes are used for the discharge and consist
of a 1 cm diameter x 2 cm long hollow cathode with a 20
gauge wire anode. A potential difference of approximately
600 volts is sufficient to strike and maintain a discharge in
the source. Ionization of organic compounds in this source
is the result of ion molecule reactions which produce proton
transfer and charge exchange reaction products. Conditions
within the glow discharge source can be adjusted to
optimize either proton transfer or charge exchange
reactions. The proton transfer reactions provide high
sensitivity for compounds which have proton affinities
greater than that of water (which is the primary proton
275
-------
transfer reagent). Charge exchange on the other hand,
is a much more universal ionization method and
produces fragmentation spectra which are similar to
electron impact ionization spectra. By operating the
glow discharge source at low pressures, the formation of
water cluster ions which often hamper API mass
spectrometers is nearly eliminated, improving sensitivity
and decreasing the complexity of the spectra.
Direct sampling using the electron impact ionization
source of the quadrupole mass spectrometer is
accomplished by means of a 1 meter length of 110
micron ID uncoated fused silica capillary tubing. A
simple on/off valve between the capillary and the source
allows the restrictor to be isolated when not in use. The
conditions in the ionizer include an electron current of
0.5 to 1.0 milliamps and an electron energy of 17 to 20
eV. The use of lower electron energies helps to
minimize fragmentation, thus concentrating ion current
in fewer ions.
Samples and Chemicals
Individual samples of 31 different volatile organic
compounds from the USEPA Target Compound List
were obtained from Ultra Scientific Company as
solutions of the neat compound dissolved in methanol at
a concentration of 10,000 ppm. Solutions for use in the
dynamic sample generation system were prepared from
the methanol stock solutions using ultra-pure water and
spectroscopic grade methanoL In order to verify the
proper calibration and performance of the dynamic
sample generation system, certified standards of volatile
organics hi nitrogen were purchased from Scott
Specialty Gases.
Water samples were prepared using distilled water
containing 0.15 g/L of sodium chloride and 0.17 g/L of
sodium sulfate. A series of concentrations of individual
volatile organics from approximately 1 ppb to 200 ppb in
water was prepared by injecting a known concentration
of a methanol solution into water and then carefully
pipetting the water standard into a 40 mL pre-cleaned
VOA vial. The vials were capped with Teflon lined
septa until used. Most samples were prepared at
approximately pH 7; however, samples of benzene,
trichloroethylene, and tetrachloroethylene were also
prepared at pH 2 and pH 10.
A total of 5 different soil samples were examined as
part of this study including 2 soils provided by the U.S.
Army Toxic and Hazardous Materials Agency
(USATHAMA), 2 local soils, and a potting soil. These
represent a range of soil types including clay, sand, and
high humic content. The soil samples were prepared by
injecting a pre-weighed 5 gram sample of soil in a 40
mL VOA vial with a known quantity of the volatile
organic in methanol and allowing it to sit for a short
period of time. Slurries of the soil samples for direct purge
experiments were prepared by adding 25 mL of water to the
sample and allowing them to sit for at least 1 hour prior to
analysis.
RESULTS AND DISCUSSION
Volatile Organics in Air
The primary objective of the air monitoring study was to
optimize the experimental conditions and determine the
real-time detection limits for a representative sample of
volatile organic pollutants. This sensitivity assessment was
performed using standard electron impact ionization on
both the tandem source quadrupole mass spectrometer and
the ITMS. This enables comparison of our results with
other mass spectrometer systems which are commercially
available and use electron impact ionization. For all ITMS
experiments, the electron ionization time was 50 msec.
Mass scan ranges were selected as appropriate for each
compound although the lower mass cut-off was normally at
least 40 amu or higher. This prevented water, nitrogen, and
oxygen ions from being stored in the ion trap simultaneously
with the analyte ions, thus minimizing the effects of space
charge and unwanted ion-molecule reactions. Future studies
will involve a comparison of sensitivities for chemical
ionization and electron impact ionization.
Using the ITMS instrument, sensitivities for the 31
volatile organics were determined. However, pumping
problems with the tandem source quadrupole mass
spectrometer restricted experiments to the determination of
detection limits for only 3 compounds: benzene,
trichloroethylene, and tetrachloroethylene. For both
instruments, response curves (instrument response vs.
concentration in air) were prepared for each of the
compounds studied. The range of concentrations examined
was generally between 4 and 200 ppb. A typical experiment
involved the acquisition of a background level signal,
followed by the acquisition of spectra for a series of
decreasing concentrations in air generated with the dynamic
sample generator. Instrument response vs. time produced a
"stair-step" curve as the concentration of organic was
reduced to successively lower levels. Each concentration
level was maintained for several minutes to ensure that a
steady state concentration was reached before further
reducing the level.
Ion Trap Mass Spectrometer
An electron impact mass spectrum of a mixture of
volatile organics in air is shown in Figure 4. This mixture
contained carbon disulfide, benzene, chloroform, toluene,
and ethyl benzene at concentrations of approximately 1 to
10 ppm. As shown in this figure, space-charge-induced
peak broadening and mass shifting are not significant.
A typical "stair-step" air monitoring response curve
acquired with the ITMS is shown in Figure 5. This is a
276
-------
reconstructed plot of the ion current for m/z 83 as"seen"
by the ITMS instrument vs. time for a sample of
chloroform in air. As the concentration of the
chloroform was decreased to lower values over a period
of time, the response of the ITMS decreased
proportionally. This same type of plot can be generated
in real-time continuous monitoring applications, allowing
changes in the concentration to be readily visualized.
As shown in Figure 5, the response time of the ITMS to
changes in concentration was very fast (less than 15
seconds) and the time required for the sample generator
to reach steady state at a new concentration was
typically less than 3 minutes.
In addition to the continuous plotting of the ITMS
total ion response, it is also possible to monitor the
actual mass spectrum in real time in order to detect
changes in specific ion intensities. This is especially
useful whenever multiple components are present in a
sample. All of the information which is generated in
real-time may be stored on a hard disk as a temporal
series of mass spectra, allowing response curves for any
ion in the mass range to be reconstructed, plotted, and
integrated. An example of a post-processed mass
spectrum of chloroform in air is shown in Figure 6.
An important feature of the response curves
generated with the ITMS is the pseudo-sinusoidal
waveform superimposed on the curve. This is not
actually noise, but is actually an effect due to the pulsed
valve addition of helium into the air stream. Maxima
correspond to the optimum helium/air ratio and minima
correspond to the least effective helium/air ratio. By
synchronizing the pulsing of the helium valve with the
acquisition of the spectral scans, this effect should be
nearly eliminated.
The experimentally determined detection limits for
the 31 volatile organic compounds in air are presented
in Table 1. As shown in this table, the detection limits
are generally in the low ppb range which is comparable
to the sensitivity of some commercially available API
mass spectrometers. Exceptions to this include
bromoform, chloroethane, and chloromethane.
However, because chloromethane and chloroethane are
extremely volatile (boilding points of 24°C and +12.3°C,
respectively), it is likely that these compounds were lost
during preparation of the standard. Bromoform, on the
other hand, is less volatile than most of the compounds
examined, with a boiling point of +150.5°C. Bromoform
probably condenses on the walls of the vapor generating
system at room temperature and never reaches the
ITMS inlet. With proper sample preparation techniques
and a shorter, heated sampling line, detection limits for
chloromethane, chloroethane, and bromoform would
probably be more comparable to the other compounds
studied. This is a reasonable assumption since these
compounds are chemically very similar to other
halogenated hydrocarbons that have been successfully
measured and would be expected to have similar ionization
efficiencies under electron impact ionization conditions.
The detection limits which are reported for volatile
organics in air, were calculated using the RMS (root mean
square) variation in the signal measured with no sample
present (a blank). This is an accurate determination of the
analytical detection limit and represents the lowest
concentration of a compound in air that can reliably be
observed with the current sampling interface and ITMS
operating parameters. For these calculations, the lowest
reliably measured signal is defined as the average of the
blank signal plus three times the RMS variation in this
signal. From the lowest reliably measured signal, the
detection limit can be calculated from a calibration curve
relating signal to concentration. Linear least squares
calibration curves were constructed for the 31 volatile
organics studied. Due to space charging effects
encountered with a few compounds, a quadratic model was
necessary to describe a better fit for the data.
Tandem Source Quadrupole Mass Spectrometer
Detection limits for benzene, trichloroethylene, and
tetrachloroethylene in air were also determined using the
tandem source quadrupole mass spectrometer. Various
concentrations of the individual compounds were generated
using the .dynamic sample generator as previously described.
One signal averaged mass spectrum (n=36) was acquired
and stored for each concentration. Signal averaged
background samples were also acquired and subtracted from
the mass spectra of the actual samples. Experimental
difficulties arising from a high hydrocarbon background in
the instrument complicated these low-level analyses. The
background problem was due to backstreaming of diffusion
pump oil and condensation on the ionization source.
Linear regressions of the data were calculated and both
data and regression were plotted for each compound. Due
to the nature of the signal averaging experiments, an
accurate detection limit could not be determined for the
three compounds using the same RMS noise calculation
method as the ITMS. Rather, the detection limit was
determined by calculating the standard deviation of the
linear regression plot and then determining the
concentration at which the signal is equal to the standard
deviation4 as shown in Figure 7. The regression curve for
benzene in air is shown in Figure 8 and the calculated
detection limit was determined to be approximately 11 ppb.
Based on the linear regression curves for trichloroethylene
and tetrachloroethylene, detection limits for these
compounds were determined to be approximately 42 and 29
ppb respectively.
Although the electron impact ionization was used
predominantly for this study, earlier experiments with the
glow discharge ionization source indicate that the detection
limits are very similar to or slightly better than those
277
-------
achievable with the electron impact ionization source.
In fact, the tandem source configuration of the
quadrupole mass spectrometer is unique and provides
extra versatility in terms of sample introduction and
ionization options relative to a conventional electron
impact ionization quadrupole. For example, air may be
sampled and ionized directly with the glow discharge
source or it may be sampled through a capillary
restrictor and ionized with the axial electron impact
ionization source. Since both ionization sources are
simultaneously installed on the spectrometer, switching
between ionization modes or sample inlet systems is a
simple matter of opening the appropriate valve and
turning on the electronics for the selected source.
The advantages of the glow discharge source
relative to the electron impact ionization source are that
it is more rugged for long term operation, the response
time is virtually instantaneous, and the source is very
tolerant of high oxygen and water saturated
atmospheres. Primary advantages of the axial electron
impact ionization source are ease of operation and the
ability to produce library searchable mass spectra. A
major problem with the electron impact source is that
the filament assembly is very susceptible to oxidation
and burn-out if exposed to large amounts of oxygen or
water. For example, when performing direct air
monitoring experiments with the electron impact source,
the filament must be replaced every 3 to 4 weeks.
Volatile Organics in Water and Soil
The sample handling apparatus and methods for the
determination of volatile organics in water and soil
slurries are identical for both the ITMS and the TSMS
experiments. Volatile organics are purged from a water
or soil slurry directly into the mass spectrometer without
any preconcentration such as trapping on a resin
cartridge. In the simplest case, conventional electron
impact ionization spectra are continuously acquired over
a mass range of approximately 40-200 amu in order to
observe the response for ions corresponding to the
purged volatile organics. As shown in Figure 9, the
purge profiles for a particular ion can be reconstructed
as a plot of response versus purge time. At a helium
purge flow of 200 mL/min, purging is normally 90% or
more complete after 3 minutes. The area beneath a
purge profile correlates well with the concentration of
the analytes in the sample as shown in Figure 10.
Quantification is accomplished simply by integrating the
area of a reconstructed purge profile for the ions
corresponding to the target analytes. A typical
calibration curve for benzene in water from 1 to 100
ppb is shown in Figure 11. Using carefully prepared
standards, correlation coefficients of better than 0.998
are possible. Quantitative reproducibiltiy of less than
10% at the 95% confidence level can also be achieved
for water samples without the use of internal standards.
A series of experiments were conducted in which the
detection limits, relative response factors, and standard
spectra were generated for a series of volatile organics in
water. In addition, studies with benzene, trichloroethylene,
and tetrachloroethylene were also conducted in order to
examine the effects of pH and soil type on the purge
efficiency of water samples and soil slurries relative to
solutions of volatile organics in pH-7 water. Data for these
samples were acquired simultaneously using both the ITMS
and the TSMS instruments in order to compare detection
limits and quantification accuracy.
The detection limits for 21 different volatile organics in
pH-7 water using the ITMS and electron impact ionization
are shown in Table 2. These range from approximately 3
ppb for benzene to approximately 60 ppb for dichloro-
ethane and appear to be routinely achievable using the
direct purge method. For comparison, the detection limits
for compounds purged into the TSMS are also typically less
than 200 ppb, although they are generally not quite as good
as can be achieved with the ITMS. Accurate detection
limits for acetone, 2-butanone, and 4-methyl-2-pentanone
have not yet been established due to much lower purge
efficiencies.
The matrix effect experiments which were conducted for
benzene, trichloroethylene, and tetrachloroethylene
appeared to show essentially the same purge efficiency at
pH-2, pH-7, and pH-10. Similar results for these
compounds were also obtained for a potting soil leachatc
with a high humic content. These results suggest that
accurate quantification maybe achieved without the need
for extensive sample preparation or the use of internal
standards for many water samples. An exception to this
may be water samples which contain a high surfactant
concentration, although comparative data have not yet been
generated.
As opposed to the water samples, differences in the
purge efficiencies for volatile organics in soil slurries are
more pronounced. As shown in Table 3, the relative purge
efficiency for benzene, trichloroethylene, and
tetrachloroethylene ranges from approximately 25% to 90%
relative to pH-7 water. The least efficient purging was from
the soils which had a high clay content and the most
efficient purging was from soils having the highest sand
content. Although the general trend exhibited by these
results is probably reasonable, the actual purge efficiencies
are probably better than the data indicate. For example,
comparative purge profiles for benzene, trichloroethylene,
and tetrachloroethylene in pH-7 water and a potting soil
slurry are very similar as shown in Figure 12.
Apparent differences in purge efficiency most likely
reflect inefficient stirring and sample purging using a single
needle sparger. Further studies have also shown that there
was probably significant loss of volatiles from the soil
278
-------
samples during the preparation step using our soil
spiking procedure. Improvements in the purging of soils
samples could probably be achieved by simultaneously
stirring samples to ensure more homogeneous sparging.
Further, the use of an internal standard would be useful
to help minimize quantitative errors due to differences
in purge efficiency.
CONCLUSIONS
The results of these studies have demonstrated the
feasibility of using direct sampling mass spectrometry for
the real-time detection of trace organic compounds in
air, water, and soils. Detection limits for both the
tandem source quadrupole mass spectrometer and the
ion trap mass spectrometer are generally in the range of
5 to 200 ppb for water and soil samples without any
sample preparation or preconcentration. The detection
limits for volatile organics in air using the ITMS range
from approximately 1 to 45 ppb for the 31 volatiles
studied which is approximately 1,000 times lower than
the threshold limit values (TLV's) for these compounds.
These detection limits are comparable to those that can
be achieved with API mass spectrometers. Detection
limits for the compounds studied using the TSMS are
slightly worse than those obtained with the ITMS;
however, they also are well below the published TLV's.
This suggests that the ITMS or TSMS could indeed be
useful for field monitoring of stack emissions and soil
gas emissions at hazardous waste sites.
Although it is not likely that significant
improvements can be made in the detection limits
achieved with the TSMS, modification and optimization
of the sampling interface for the ITMS will probably
result in even better detection limits than reported in
this document. In addition, the ITMS instrument also
has the capability of chemical ionization which can be
used to selectively enhance certain target analytes
relative to other compounds in a sample stream.
Both the TSMS and ITMS have excellent detection
limits for volatile organic compounds in air, water, and
soil; however, experience with the two different mass
spectrometer systems suggests that the ion trap mass
spectrometer overall is a more useful instrument for
continuous air monitoring. Specifically, the ITMS is
highly reliable, easier to operate, and more stable than
the tandem source quadrupole mass spectrometer.
Further, the ion trap mass spectrometer has the
capabilities of controlled chemical ionization, selective
fan storage, and collision induced dissociation (CID)
tandem mass spectrometry (MS/MS). These features
are especially important in helping to identify individual
components in a complex sample, especially since no
chromatographic separations are performed -on the
sample prior to entering the mass spectrometer.
Without these features, the TSMS is restricted to
monitoring samples that typically have fewer than 10-15
components. Finally, due to the simplicity of the ion trap
analyzer assembly, this type of instrumentation lends itself
to downsizing, portability, and remote operation better than
the TSMS.
While the results of this study have been quite
successful and demonstrate the potential of the
instrumentation for screening of environmental samples,
much work remains. Especially important is the
development of methods for the identification and
quantification of compounds in complex mixtures. This
work will involve a thorough examination of chemical
ionization reactions, the generation of MS/MS spectra of
commonly encountered organic pollutants and potential
interferences, and the development of computer programs
to process this information in real time.
ACKNOWLEDGEMENT
'Research sponsored by the U.S. Army Toxic and
Hazardous Materials Agency under Interagency Agreement
1769-A073-A1 under U.S. Department of Energy Contract
DE-AC05-84OR21400 with Martin Marietta Energy
Systems, Inc.
REFERENCES
1. McClennen, W.H., Arnold, N.S., Sheya, S.A.,
Lighty, J.S., Meuzelaar, H.L.C., "Direct Transfer
Line GC/MS Analyses of Incomplete
Combustion Products from the Inceneration of
Medical Wastes and the Thermal Treatment of
Contaminated Soils", Proc. 38th ASMS Conf. on
Mass Spec. All. Topics, Tucson, AR, 1990, 611-
612.
2. Hemberger, P.H., Alarid, I.E., Cameron, D.,
Leibman, C.P., Cannon, T.M., Wolf, M.A.,
Kaiser, R.E. "A Transportable Gas
Chromatograph/Ion Trap Detector for Field
Analysis of Environmental Samples", Int. J. Mass
Spectrom. Ion Proc.. In press.
3. Wise, M.B., Buchanan, M.V., Guerin, M.R.,
"Rapid Environmental Organic Analysis by
Direct Sampling Glow Discharge Mass
Spectrometry and Ion Trap Mass Spectrometry",
Oak Ridge National Laboratory TM-11538, Oak
Ridge Tennessee, 1990.
4. Hubaux, A.; Vos, G., Anal. Chem.. 235, 1967,
849-855.
279
-------
Table 1
Detection Limits for Volatile Organics in Air using Direct Sampling FTMS
Compound Detection Limit (ppb)
1,1,1-Trichloroethane 2
1,1,2,2-Tetrachloroethane 3
1,1,2-Trichloroethane 20
1,1-Dichloroethane 16
1,1-Dichloroethene 6
1,2-Dichloroethene 3
1,2-Dichloropropane 45
2-Butanone 48
4-Methyl-2-Pentanone 17
Acetone 22
Benzene 5
Bromodichloromethane 4
Bromoform > 80
Bromomethane >280
Carbon Disulfide 25
Carbon Tetrachloride 16
Chlorobenzene 2
Chloroethane >209
Chloroform 3
Chloromethane >268
Cis-l,3-Dichloropropene 6
Dibromochloromethane 12
Ethylbenzene 2
Methylene Chloride 12
Tetrachloroethylene 8
Toluene 3
Trans-l,3-Dichloropropene 7
Vinyl Acetate • 44
Vinyl Chloride 5
O-Xylene 4
280
-------
Table 2
Detection Limits for Volatile Organics in pH-7 Water using Direct Purge ITMS
Compound
1,1,1-Trichloroethane
1,1,2,2-Tetrachloroethane
1,1,2-Trichloroethane
1,1 -Dichloroethene
1,2-Dichloroethane
1,2-Dichloroethene
Benzene
Bromoform
Carbon Disulfide
Carbon Tetrachloride
Chlorobenzene
Chloroform
Cis-l,3-Dichloropropene
Ethylbenzene
Methylene Chloride
Styrene
Tetrachloroethylene
Toluene
Trans-l,3-Dichloropropene
Vinyl Chloride
Xylenes (total)
Detection Limit (ppb)
12
28
18
33
27
21
3
15
18
16
5
20
6
4
60
5
5
4
15
5
4
Table 3
Purge Efficiency of Volatile Organics in Soil Slurries Relative to pH-7 Water
Soil Sample Soil Type
THAMA 1 Clay 29
THAMA 2 Sand/Clay 51
Local 1 Sand/Clay 61
Local 2 Sand/Clay/Humic 46
Potting Sand/Humic 91
Relative Purge Efficiency (%)
Trichloroethvlene Tetrachloroethvlene
20
48
45
42
77
19
46
61
42
53
281
-------
ITMS DIRECT AIR INLET
VENT/AUX SAMPLING PUMP
t
MICROBORE
MEGABORE CAPILLARY RESTRICTOR
AIR INLET.
PULSED SOLENOID VALVE
TO ION TRAP CELL
METERING VALVE
HEUUM INLET
TO SAMPLING PUMP
Figure 1 Air sampling interface for ITMS.
TEFLON
HELIUM
PURGE
GAS -
AN
Lll
•
t
9
•
A
•t
SF
ME
;
ER
\
"f*~* * — ^" *
SPLIT (
WAI \/t=N
i
•
i
X
3APILLARY
X3B33
ION
MC
Mo
i —
' — i
FPUMP |
Figure 2 Device used for direct purge of volatiles from water and soil samples.
282
-------
1.5 mm Vacuum Conductance Limit
Aa*Iyz»r Vacuum Chamber
Clew Discharge
Senrcv Vacuum
Chamber
To Reo(h Pump
Figure 3 Diagram of the tandem source quadrupole mass spectrometer.
500 -
(Cone. Low PPM)
0
100
Mass (AMU)
110
120
Figure 4 ITMS electron impact mass spectrum of ppm levels of VOCs in air.
283
-------
ITMS Response to Chloroform in Air
170
160
•53
140
130
120
no
•oo
90
80
70
60
50
4-0
30
20
•0
0
170ppb
56ppb
28ppb
14 ppb
blan
0.2 0.4
0.6 0.8
(Thousands)
Elapsed Time (seconds)
Figure 5 ITMS response for m/z 83 at various concentrations of chloroform in air.
INT
Chloroform in Air
83
56 61 65 69 73 81
85
8?
100
105
100 120
Figure 6 ITMS post processed mass spectrum of chloroform in air.
80
Mass (AMU)
284
-------
t>
<
130
120 -
110 -
100
90
80
70
60
50
40
30
20
10
0
Estimate of Tetrachloroethylene Detection Limit
Tandem-Source Quadrupole Detection
*95% Confidence Limits
^Calibration Curve
I 11 I I I ! I I
I I I I I I
20 40 60 80 100
Concentration (ppb)
120 UO
160
Figure 7 Graphical determination of detection limits for the TSMS instrument.
BENZENE
Tandem Source Ouadrupola Air Monitor
i
S.
'c
800
700 h
600
500
400
300
200
100
0
-100
I I I I L
20 40 60 80
Concentration in Air (ppb)
100
120
140
Figure 8 Linear regression curve for benzene in air using the TSMS instrument.
285
-------
RESPONSE FOR m/z 78
0:38
1:18 2:05
TIME (min)
2:45
3:25
Figure 9 Reconstructed purge profile for 100 ppb of benzene i
in water.
SOLUTION PURGE PROFILES OF AQUEOUS
VINYL CHLORIDE STANDARDS (ppb = ng/ml)
z
LLJ
H
Z
LLJ
LU
DC
40 ppb
20 ppb
10 ppb
2 ppb
7:09
T-T- | i i i i r-r-r
9:32 11:54 14:17 16:39 19:01
TIME (min)
Figure 10 Direct purge profiles for 4 different concentration of vinyl chloride in water.
286
-------
RESPONSE FOR m/z 78
0
20 40 60 80
CONCENTRATION (ppb)
100
Figure 11 Response curve from 1 to 100 ppb for direct purge of benzene from water.
287
-------
VOLATILE ORGANICS PURGED FROM PH-2 WATER
770
385 -\
840-
50 ppb each compound
BENZENE
m/z78
z
O
TRICHLOROETHYLENE
m/z 130
1000-
500-
TETRACHLOROETHYLENE
m/z 166
3 4
TIME (MEM)
VOLATILE ORGANICS PURGED FROM POTTING SOIL SLURRY
50 ppb each compound
BENZENE
m/z 78
a 630
I 315H
UJ
TRICHLOROETHYLENE
m/z 130
420
210 H
TETRACHLOROETHYLENE
m/z 166
Figure 12 Comparison of VOC purge profiles for pH-7 water and potting soil.
288
-------
DEVELOPMENT AND TESTING OF A MAN-PORTABLE
GAS CHROMATOGRAPHY/MASS SPECTROMETRY SYSTEM
FOR AIR MONITORING
Henk L.C. Meuzelaar, Dale T. Urban and Neil S. Arnold
Center for Micro Analysis & Reaction Chemistry, University of Utah
214 EMRL, Salt Lake City, UT 84112
ABSTRACT
A fully man-portable, GC/MS system based on the
combination of an automated vapor sample inlet, a
"transfer-line" gas chromatography module and a
modified Hewlett Packard model 5971A quadrupole MS
system is described. The current prototype weighs
approx. 70-75 Ibs and uses 150-200 W of battery power.
Th&mass spectrometer and computer are carried in front
of the operator by means of a shoulder harness whereas
battery pack, carrier gas supply and roughing vacuum
system are carried as a backpack. Air samples can be
malyzed using a special automated air sampling inlet.
TTie man-portable GC/MS system is designed to be
Supported by a vehicle transportable "docking station".
BACKGROUND
In situations involving severely contaminated hazardous
sraste sites, industrial accidents or natural disasters, as
srell as special military or law enforcement operations,
mobile laboratories may be of little use because of
Emited site access, restrictions due to contamination or
terrain constraints. Under such conditions, man-portable
analytical instruments may offer the only acceptable
means of carrying out on-sitc analyses.
Obviously, man-portability puts severe constraints on
Wight, size and power requirements as well as on
fflggedness and user-friendliness. Consequently, the man-
fortability requirement may also function as a convenient
benchmark for the development of analytical equipment
for a variety of special operational environments ranging
fiom remotely operated devices (e.g., robotic vehicles,
femes or probes) space stations and operating rooms.
All of the above environments require a high degree of
miniaturization, reliability and ease of operation.
The past decade has witnessed impressive progress in
miniaturization of mass spectrometric systems. Besides a
broad range of commercially available benchtop
instruments, including the Hewlett Packard MSD (Mass
Selective Detector) and Finnigan MAT ITD (Ion Trap
Detector), several specialized MS instruments have been
developed for applications where transportability is a
prime requirement. Well known examples include the
Bruker Franzen MM1 system, originally developed for
military applications involving chemical agent detection,
and the Viking Spectratrak system primarily designed for
environmental applications.
As shown in Figure 1 most commercially available
miniaturized systems are characterized by a combination
of relatively low weight (typically 100-300 Ibs, excluding
power source) and modest power requirements (600-1800
W range). In spite of these marked advances in system
miniaturization, however, man-portability and some of the
other abovedescribed applications require even more
stringent size, weight and power limitations.
This prompted us to undertake a study aimed at obtaining
maximum power and weight reduction using the Hewlett
Packard MSD as a starting point. Although the project is
still under continuing development, some preliminary
results and conclusions are starting to take shape, as will
be discussed in the following paragraphs.
SYSTEM DESIGN CONCEPTS
An overview of the selection criteria for the main
system modules and components is given in Table I.
289
-------
Automated Vapor Sampling Inlet Module
Transfer Line Gas Chromatography Module
Transfer line gas chromatography (TLGC) is defined here
as a form of GC in which the column connects two
environments, viz. an atmospheric environment at
ambient pressure and the vacuum environment of the MS
ion source region. In other words, column inlet and
outlet pressures are more or less fixed and, consequently,
optimization of column flow requires suitable adaptation
of column length and/or diameter. This sets TLGC apart
from the more widely used short column gas chromato-
graphy (SCGC) technique in which column inlet
pressures can usually be adjusted while column length is
kept below 5 meters or so.
Although most TLGC applications reported thus far do
use short to very short column lengths, optimum GC
conditions for a 500 p.m i.d., ambient inlet transfer line
columns connected to a vacuum detector (e.g., MS) may
dictate column lengths in the 50-100 m range (see Figure
2). In view of the abovedescribed distinctive differences
between TLGC and SCGC we feel justified in adding yet
another term to the already baffling jargon of the
chromatographer.
When sampling condensable and potentially labile vapors
from air, the main challenge is to avoid compound losses
through irreversible adsorption and/or decomposition in
the transfer line section. To this end, a novel, automated
vapor sampling method was recently designed at the
University of Utah Center for Micro Analysis & Reaction
Chemistry (1,2). The most characteristic property of this
sampling method, illustrated in Figure 3, is the absence
of any valves or other mechanical obstructions in the path
of the molecules between the ambient environment and
the ion source. Only quartz walls and/or surfaces coated
with inert stationary phases (e.g., poly-dimethylsilicones)
are seen by sample molecules on their way to the ion
source.
A second advantage of the new sampling technique is the
potentially very short switching time. Sampling times as
short as 60 msec have been used already (2) and 20 msec
or less may be achievable in the near future. This
enables "injection" of a narrow sample plug into the
TLGC column, thereby minimizing peak broadening due
to sample injection and allowing repeat GC analyses at 6-
60 sec intervals (3). All air flows in the inlet are
sustained by means of a Graseby Ionics miniaturized dual
air pump (max. capacity 2 x 500 ml/min, max power
consumption 1 W) whereas rapid switching of air flows is
performed with a Skinner micro valve (5 msec response
time).
The GC oven module consists of a simple heated
aluminum cylinder which houses the capillary GC
column, e.g., a 29 cm long, 50 \im i.d. fused silica
capillary coated with a 0.2 pirn thick layer of poly-
dimethylsilicone (DBS) and providing a continuous He
flow of approx. .02 ml/min.
At present the oven is used in isothermal mode only. A
temperature programming option as described by Arnold
et al. (4), which would allow a broader range of
compounds to be analyzed in a single GC run and also
help protect the column from oxidative degradation, has
not yet been implemented in the present prototype. A
direct consequence of the rapid GC run time is the need
for very high temperature programming rates, e.g, 10-20
C/sec. This requires significantly larger power supplies
than necessary for isothermal operation.
A small (2 ft3) compressed gas cylinder with flow
controller provides more than 36 hours of He or N2
carrier gas flow. The theoretical relationship between
inner column diameter, max. resolving power, column
length and retention time is depicted in Figure 2.
Obviously, the use of a 50 urn i.d. column (primarily
selected to keep gas flows as low as possible) has the
advantage of allowing very rapid separations, although
limiting maximum achievable resolving power.
Quadrupole MS Module
A Hewlett Packard Model 5971 MSD (Mass Selective
Detector) was modified extensively in order to reduce
system weight and power requirements and increase
overall manoeuverability. The original housing was
completely discarded and the relative positions of the
electronic boards were changed to enable convenient
operation of the air sampling inlet. The new
configuration is shown in Figures 4 and 5. Most
importantly, the original AC and DC power supplies were
removed and replaced by a battery powered 12 V DC
supply with DC/DC converters for the various DC
voltages required for mass spectrometer, computer and
sampling inlet operation. Total power consumption of
the modified MS system was determined to be 43 W (see
Table I).
Vacuum System
The vacuum system of the HP model 5971 MSD was
completely reconfigured to provide operating pressures in
the lO^-lO"5 torr range while minimizing roughing
vacuum requirements. The original 60 I/sec diffusion
290
-------
pump was exchanged for an Alcatel Model 5010 MDP
JMolecular Drag Pump) with a max. pumping speed of 8
sec'1 for N2 and a roughing vacuum requirement of < 30
millibar. This enabled us to replace the original rotary
pimp (power requirement approx. 160 watts; weight 14
Jbs) with a simple vacuum buffer capable of maintaining
iroughing vacuum of better than 10 millibar for up to 12
iours at the specified GC column flows. The vacuum
assembly configuration can be seen also in Figs. 4 and 5.
Micro Computer Module
A Toshiba model 5200, 20 Mhz, 80386 lap top is used to
control all GC/MS functions by means of a standard PC
interface and software available from Hewlett Packard.
In addition the PC system controls the operation of the
air sampling inlet. The only modification of the Toshiba
SlOO consisted of removing the built-in, relatively heavy
DC and AC power supplies and connecting the unit
directly to the specially constructed DC power supply
down in Figures 4 and 5.
SYSTEM INTEGRATION
Mechanically, the various components described thus far
were integrated by means of a specially designed
Shoulder harness and backpack frame, as shown in Figure
5. The aluminum backpack frame carries the two
batteries as well as the vacuum reservoir whereas the
entire mass spectrometry assembly with MDP and PC is
Suspended from the shoulder straps and stabilized by two
ilp straps. Due to the difficulty of typing in detailed
(omputer commands during field use, especially when
tearing gloves, a beach ball type mouse was installed to
fflable direct communication with a single (gloved) hand.
Alternatively, one could envisage the use of a built-in PC
Computer card (without display screen or keyboard)
icmotely controlled by a second, more completely
outfitted PC using standard PC software such as Carbon
Copy® or PC Anywhere®.
The most simple remote control option would be to use
in umbilical cord carrying a twisted pair cable in addition
to AC power. The latter option would eliminate the
heavy (28 Ib) battery pack, thus resulting in greatly
reduced overall size and weight. Finally, as also shown
kFigure 4, a special transportable "docking station" (still
under construction) enables vacuum system regeneration,
fettery recharging and carrier gas refills at 6-10 hour
intervals.
PRELIMINARY TEST DATA
TLGC/MS curves generated with a 100 cm long, 100 jim
i.d. capillary column, coated with 0.25 jim
polydimethylsilicone (DB5, Supelco) while sampling a
mixture of 10 ppm vapor components in air for 1 sec at
30 sec intervals are shown in Figure 6. Obviously, a
highly useful level of chromatographic separation is
achieved with the very short transfer line. Also the
narrow peak shapes (half height width < 1 sec) illustrate
the efficiency of the rapid sampling air inlet. Overall
peak height reproducibility (approx. + 10%) is influenced
by the limited resolution of the sampling time due to
manual operation.
From the selected ion profile (tropylium fragment ion at
m/z 91) in Figure 6 the minimum detectable
concentration in direct air sampling mode appears to be
approx. 1 ppm. Although this is 1-2 orders less than the
minimum concentrations detected by means of ion trap
type MS systems when using the automated vapor
sampling inlet (2), the MSD system has not yet been
fully optimized for operation under the present vacuum
and flow conditions. However, since it may be
anticipated that some of the most promising applications
will require detection limits in the lower ppb range, a
suitable adsorption/desorption module is currently under
development in our laboratory.
Figure 7 illustrates the performance of the automated air
sampling TLGC/MSD system with polar compounds
under similar experimental conditions as in Figure 6.
Note the rapid separation of a mixture of ketones into its
components and the relatively minor degree of peak
tailing due to the heated, all quartz vapor sampling inlet.
Finally, Figure 8 shows selected ion chromatograms for
several chemical agent simulants, demonstrating the fast,
repetitive (17 sec interval) analysis capability of the short
(29 cm) narrow bore (50 jxm i.d.) capillary column used
while maintaining adequate chromatographic resolution.
Although it is tempting to envisage the use of man-
portable GC/MS instruments for military reconnaissance
purposes, e.g., when venturing into contaminated regions
with high levels of background interferents, it should be
pointed out here that the current sensitivity of the MSD
based TLGC/MS system is insufficient for such appli-
cations. Partially, this is due to the relatively low sample
mass flow through the narrow bore capillary columns
used. In principle, this could be corrected by closing up
the MSD ion source thereby increasing the residence time
291
-------
of the vapor molecules in the source which would result
in increased ionization efficiencies.
Additionally, the use of rapid absorption/desorption
methods for sample preconcentration should be
considered. Assuming a 10 second absorption interval at
10 times normal flow, followed by a 1 second desorption
interval at normal flow, it should be possible to obtain a
100 times enrichment factor without sacrificing analysis
speed. Basically, the 10-15 seconds necessary for
chromatographic separation is then being used to collect
and preconcentrate the next sample.
Finally, we are investigating the use of rapid (10-20
C/sec) temperature programmed heating in order to
broaden the range of compounds that can be analyzed in
a single chromatographic run. The feasibility of this
approach has been demonstrated by Arnold et al. (4). A
second, important advantage of rapid temperature
programming is that the initial "air peak" passes through
the column at low temperature, thereby considerably
reducing the likelihood of oxidative degradation of the
column. This then allows programmed heating of the
column to high temperatures (e.g., 300 C) thus enabling
separation and detection of large polar molecules such as
underivatized trichothecenes, as demonstrated by
McClennen et al. (5). Many commercially available, air
sampling mass spectrometry and ion mobility
spectrometry systems use silicone membrane interfaces,
thereby the detection of large, polar compounds.
CONCLUSIONS
The feasibility of constructing a fully man-portable
"transfer line" GC/MS system with automated vapor
sampling capability has been demonstrated. In its present
form, the system weighs 72 pounds, consumes 160 W of
electrical power and can operate continuously for 6-10
hours. Application of novel battery technologies, further
integration of the microcomputer module and use of
alternative vacuum pumping strategies is expected to
reduce overall system weight to less than 50 Ibs.
Without vapor preconcentration, practical detection limits
appear to be in the low ppm range. Development of
rapid temperature programming capabilities is being
considered in order to facilitate detection of relatively
nonvolatile species and to increase the range of
compounds that can be analyzed in a single run. The
ultralow power and weight requirements of the technique
would seem to offer promise for a broad spectrum of
field applications ranging from hazardous waste sites and
industrial or natural disaster areas to reconnaissance
drones, space stations, interplantary probes and
autonomous vehicular robots.
REFERENCES
1. McClennen, W.H., Arnold, N.S., Meuzelaar,
H.L.C., Apparatus and Method for Sampling. U.S.
Patent 4,970,905.
2. Arnold, N.S., McClennen, W.H., Meuzelaar,
H.L.C., "A Vapor Sampling Device for Rapid,
Direct Short Column Gas Chromatography/Mass
Spectrometry Analyses of Atmospheric Vapors",
Anal. Chem., in press.
3. McClennen, W.H., Arnold, N.S., Sheya, S.A.,
Lighty, J.S., Meuzelaar, H.L.C., "Direct Transfer
Line GC/MS Analyses of Incomplete Combustion
Products from the Incineration of Medical Wastes
and the Thermal Treatment of Contaminated
Soils", Proc. 38th ASMS Conf. on Mass Spec.
All. Topics, Tucson, AR, 1990, 611-612.
4. Arnold, N.S., Kalousek, P., McClennen, W.H.,
Gibbons, J.R., Maswadeh, W., Meuzelaar, H.L.C.,
"Application of Temperature Programming to
Direct Vapor Sampling Transfer Line GC/MS",
Proc. 38th ASMS Conf. on Mass Spec. All.
Topics, 1990, 1401-1402.
5. McClennen, W.H., Meuzelaar, H.L.C., Snyder,
A.P., "Biomarker Detection by Curie-point
Pyrolysis in Combination with an Ion Trap Mass
Spectrometer", Proc. 1987 CRDEC Conf., 271-
277.
ACKNOWLEDGEMENTS
The authors acknowledge Jean-Luc Truche and
John Fjeldsted (Hewlett Packard Corp.) for their valuable
ideas and continued technical support and thank William
H. McClennen and Pavel Kalousek (University of Utah,
Ctr. Micro Analysis & Reaction Chemistry) for their
expert technical advice and assistance. This work was
financially supported by Hewlett Packard Corporation
(University of Utah Instrumentation Grant) and by the
Advanced Combustion Engineering Research Center.
Funds for this Center are received from the National
Science Foundation, the State of Utah, 23 industrial
participants and the U.S. Department of Energy.
292
-------
TABLE I: PRIMARY SYSTEM COMPONENT SELECTION CRITERIA
Automated Vapor Sampling Inlet Module
fully automated
only inert quartz and fused silica materials
ultrashort sample "injection" pulse
Transfer Line GC Module
interferent rejection
rapid analysis capability
Hewlett Packard 5971A Mass Selective Detector
low power requirements (43 W)
lightweight (7 kg)
Alcatel 5010 Molecular Drag Pump
low power consumption (17 W)
high backing pressure up to 40 mbar (no backing pump
needed)
light weight (2.35 kg)
Toshiba 5200, 20 mhz, 386 Computer
low power consumption (40 W)
high speed, capable of running existing MSD software
"o
O
1E5
1E4
1000
100
N>
CO
CO
200
150
Sf
B 100
O
50
FINNIGAN ITD
BRUKER MM-1
O
INCOS 500
D
HP MSD 5971A
V
SPECTRATRAK 600
MAN-PORTABLE
10 100 1000
Transfer Line Length (cm)
1E4
100
1E-2 1E-1
1 10 100
Retention Time (s)
1000
1E4-
500
1000
1500
2000
2500
POWER (Watts)
Figure 1. Power requirements and weights of typical
miniaturized GC/MS systems (note that man-portable
system includes power and carrier gas sources).
Figure 2. Theoretical relationships between internal
column diameter (in \am), maximum achievable resolving
power, column length and retention time for a compound
with capacity factor k=5.0. (Triangles indicate points of
minimum plate height operation.)
-------
To Transfer Line
and Detector
Sampling
Mode
Inert
Vacuum Carrier
Gas
Sample He He +
Sample
Control
System
J Vacuum
Inert
Vacuum Carrier
Gas
To Transfer Line
and Detector
Separation
Mode
Figure 3. Operating principle of automated vapor sampling inlet developed at University of Utah (US patent no.
4,970,905).
HP MSD ANALYZER
RF GENERATOR
QUADRUPOLE
DETECTOR
HP HARDWARE
INTERFACE
VACUUM SYS
CEN
PORTA
MSDD
INLET (
TEM
MOLECULAR DRAG PUMP
VACUUM RESERVOIR
VAPOR INLET
SAMPLE PUMP
GC OVEN HEATERS
CARRIER GAS
CENTRAL DATA SYSTEM
PORTABLE 386 COMPUTER
SAMPLE IN
"DOCKING STATION"
POWEF SYSTEM
24 VDC BATTERY
DC/DC CONVERTER
REFILL CARRIER GAS
RECHARGE BATTERY
EVACUATE RESERVOIR
Figure 4. Block diagram of man-portable GC/MS system and docking station interface.
294
-------
Figure 5. Schematic outline of GC/MS man-portable
system with operator. A) vapor inlet/transfer line GC
column; B) MSD analyzer; C) control electronics; D)
portable 386 computer; E) molecular drag pump; F)
vacuum hose; G) vacuum reservoir; H) carrier gas, and I)
24VDC battery.
12000-1
10000
3 8000
—
B
a
6000
3 4000
o
2000
1.20 1.40 1.60 1.80 2.00 2.20
Time (minutes)---
2.40 2.60 2.BO 3.00 3.20
Figure 6. Selected ion chromatogram profile of an alkylbenzene mixture at m/z 91 obtained by TLGC/MS using
the automated vapor sampling inlet in combination with a 100 cm long, 100 [xm i.d., DBS coated fused silica
capillary column. (1) toluene; (2) ethyl benzene; (3) m-xylene; (4) o-xylene. Approximate vapor concentrations:
lOppm. Arrows indicate air sampling events at 30 second intervals. Note that o-xylene (peak 4) elutes after the
next sampling event.
3944
i.2o '' i.« " i.to i.bo
Time (minutes)— >
Figure 7. Total ion chromatogram (TIC) for a mixture of 4 ketones. 1) acetone; 2) methyl ethyl ketone; 3) ethyl
acetate; 4) 3-pentanone; 5) methyl iso-butyl ketone.
295
-------
« 100 -
m/z 79, DiVEVEP
z 111, DEEP
6 sec
Time (s)
Figure 8. Selected ion chromatograms of 4 chemical agent simulants (DMMP=dimethyl methyl phosphonate,
DEEP=diethyl ethyl phosphonate, DlMP=diisopropyl methyl phosphonate, DEM=diethyl malonate). Arrows
indicate air sampling points (17 sec interval). Note separation of all 4 simulants within 6 sec. Star symbol (*)
indicates "pseudo" peak due to effect of eluting air on MS system.
296
-------
DISCUSSION
IALPH SULLIVAN: With these high flow rate systems, how did you go about
calibrating it and how do you introduce the gas to it to know what you have in
the system?
HENK MEUZELAAR: You make diluted air — the flow rate doesn't have to
fe above 100 mL per minute, or even 50 per minute. So, if you have a dilution
tjrstem that can give you that kind of output, you can just calibrate it with a
calibrated dilution system.
AUDIENCE PARTICIPANT: Could you repeat that?
HENK MEUZELAAR: All right. What I said is the high flow of the outer tube,
tie first sampling tube, can be as little as 50 or 100 mL per minute. So, if you have
tvapor dilution system that can give you a couple hundred mL output you can
4) a loose coupling for such a system and get very good results. If you have a
tapor dilution system that just puts out a few mL per minute it would be more
difficult to do that. You could do it from a bag if you could fill a bag and keep it
at atmospheric pressure for several minutes, you could obtain a sample without
changing the pressure or the concentration in the bag.
BILL McCLENNY: I was wondering what the prospects would be for using
some type of preconcentration that involved a cold trap, using thermo electric
cooling or something of that sort, and what that would add to the power
requirements for this unit?
HENK MEUZELAAR: I think almost any type of absorption, desorption, or
preconcentration by any method I know that would keep the high response
characteristic intact, would certainly require power because you would have to
desorb for a relatively short period of time. And the only way to make gain is to
absorb for let's say 60 seconds and flush desorb in one or two seconds. That's
going to require power. We are currently looking at a number of different
methods. The power requirement is just needed, for a second, or maybe even less
than that. I think it's a doable thing, but it certainly will add to the power
requirement.
297
-------
ON-SITE MULTIMEDIA ANALYZERS:
ADVANCED SAMPLE PROCESSING WITH ON-LINE ANALYSIS
S. Liebman
GEO-CENTERS, INC.
c/o U.S. Army Cml
Rsch, Dev & Engr Ctr
Attn: SMCCR-RSL
Aberdeen Proving Ground,
MD 21010-5423
M. B. Uasserman
U.S. Army Cml Rsch,
Dev & Engr Ctr
Attn: SMCCR-RSL
Aberdeen Proving Ground,
MD 21010-5423
E. J. Levy and S. Lurcott
Computer Chemical Systems, Inc.
Rt. 41 and Newark Rd., Box 683
Avondale, PA 19311
ABSTRACT
The need for on-site chemical analysis
of air, water, and soils has led to
development of two highly automated
prototype instruments in the field of
trace organic analysis: EPvA, the
Environmental £yroprobe Analyzer and
CHAMP, the Chemical Hazards Automated
Multiprocessor. In the EPyA unit, a
purge and trap module permits routine
determination of target chemicals in
water and hazardous wastes. A thermal
desorption module permits controlled
thermal desorption of air sampling
cartridges, as well as dynamic
headspace/pyrolysis analyses of
solids. CHAMP is based on supercriti-
cal fluid extraction (SFE) with liquid
C0£ mobile fluid for solid samples in
amounts from milligrams to over
several grams in six individually
heated extractors. Specialty
interfaces, such as TRANSCAP. provide
on-line analysis by chromatographic
and/or spectral detectors.
Both benchtop, microprocessor-based
systems are newly designed for in-
field operation, as well as laboratory
or plant sites. Highly automated
instruments such as EPyA and CHAMP
operating with external expertise
provided by artificial intelligence
(AI) software, illustrate the Inte-
grated Intelligent Instrument (I3)
approach which is focused on multi-
media analyses for hazardous
materials.
INTRODUCTION
Advantages of precision, accuracy, and
reproducibility are realized with the
use of automated instruments to per-
form thermal and nonthermal sample
processing with on-line chromato-
graphic and/or spectral analyzers.
New.engineering designs are required
to bring this analytical power on-site
to the field, mobile lab, or plant to
provide rapid, validated information
to analysts. Two prototype analytical
systems are described to meet these
needs; the Environmental Pyroprobe
Analyzer, EPyA (1) and CHAMP, the
Chemical Hazards Automated Multi-
Processor (2). The prototypes are
designed for compactness with inte-
grated specialty separation and/or
detector units that are important to
the hazardous waste field for on-site
use. Figure l(a,b) shows the bench-
top units, each about 2'x2'x3' and
weighing ca. eighty pounds. The pur-
pose of this report is to describe the
ongoing development of specialty in-
strumentation that is based on proven
analytical methodologies in trace
organic analysis.
I. Thermal Sample Processing - EPyA
The thermal analyzer system, EPyA. is
the result of over fifteen years of
engineering design and manufacture of
microprocessor-based instrumentation
used throughout the world for trace
organic analysis of vapors, liquids
and solids. Studies in the 70's and
299
-------
80's developed purge and trap modules
for water analyses and thermal desorp-
tion methods for rapid analyses of air
sampling cartridges that contained
treated charcoals, porous polymers,
Ambersorb, Tenax, etc. (3a). Figure 2
shows a test air mixture with 40 ppb
levels of typical solvents (benzene,
toluene, chloro-benzene, heptane, o-
dichlorobenzene, and dodecane) sampled
for 90 sec at 0.5 mL/min on a Tenax
sorbent bed (100 mg) which was then
thermally desorbed for GC/FID analysis
(3b-e). Figure 3 shows an analysis
conducted for gasoline/fuels using a
cryofocusing concentrator module and
on-line capillary GC/FID detection
(4). Figure 4a,b,c give results from
other studies (4) using remote air
sampling cartridges for analyses of
outside air, laboratory air, and paint
shop air (all 500 ml samples) with
GC/FID analysis.
Recently, the thermal desorption/cyro-
trapping module was used in trace
particulate analysis of a mlcroencap-
sulated pesticide, Diazinon, in an air
sampling cartridge with on-line analy-
sis by GC-MS (5) (Figure 5). A
corresponding dynamic headspace/py-
rolysis method using the Pyroprobe Pt
coil pyrolyzer on a few micrograms of
a microencapsulated sample also
provided trace detection and
identification of the Diazinon core,
which gives a parent ion at m/z 304
and a base peak at m/z 179. Clearly,
thermal desorption, rather than C$2
solvent stripping, proved to be the
optimum analytical method which is now
used throughout the world in the
industrial R&D, forensic, and
environmental fields. However, some
thermally sensitive samples required
additional effort for reliable
analyses. A more effective method
than solvent extraction was needed,
both for analyzing thermally labile
materials, as well as to eliminate
solvent wastes. The traditional
Soxhlet solvent extraction method has
further disadvantages of hour or day-
long extraction times and off-line,
more labor intensive, multistep
analyses for complex environmental
samples.
II. The Nonthermal Sample Processing
Analyzer - CHAMP
The nonthermal multiple sample proces-
sing system, CHAMP, using supercriti-
cal fluid (SF) technology (6) permits
the conduct of trace organic analysis
on diverse samples, including cart-
ridge sorbent beds (7), soils, coals,
or hazardous waste solids. Six
individually heated sample extractors
may contain up to five grams or more
of material to be treated near or at
supercritical fluid conditions in the
2, 4, or 6 mL extractor vessels.
Automated SF extraction (SFE)-capil-
lary GC analysis of gasoline from
charcoal filters may be routinely
analyzed with either single or
multiple SFE units. Analytes re-
quiring well-established capillary GC
methods use the automated SFE system
configured for GC separation.
Alternatively, in Figure fc, the SFE-
SFC analysis is shown of a phosphonate
chemical in soil (ca. 500 mg) with
detection by FID at estimated ppb
levels. The SFE was conducted at 3000
psi, 100°C with C02, which is a
nontoxic, safe and inexpensive mobile
fluid. The SFC was conducted with a
Nucleosil CN microbore column at 120°C
and pressure programming from 2000 to
6000 psi at 300 psi/min with a FID
unit.
As with the thermal processing
analyzer, EPyA. it is necessary to
have a variety of detection systems
for adequate analytical sensitivity
and specificity. The SFE process has
been used with FID, ultra-violet and
mid-infrared (ir) spectrometers using
fiber optic monitors (FOM) (6,8).
Figure 7 represents the recent on-line
SFE-SFC analysis of a polyolefin/
naphthalene mixture. An ion trap MS
detector (ITD) was used to detect the
molecular ion from naphthalene (m/z -
128) (9). Other detectors show
similar potential for trace on-line
analyses with highly specific and
sensitive responses to hazardous/toxic
substances, e.g., fluorescence/uv with
fiber optic technology and advanced
data analysis with applied AI (10).
Both EPvA and CHAMP incorporate new
design engineering features that
emphasize compact, transportable
systems. Sample processing,
integrated with separation and
300
-------
detection units are controlled by
microprocessors with programmable,
interactive software. External AI
software will provide guidance in the
use of the total system.
III. Applied Artificial Intelligence
- Expert System Networks
The I3 approach combines data
generation using highly automated
modular/interfaced systems with
external intelligence for development,
data analysis, interpretation and
validation. Development of a
proprietary expert system network for
SF technology, MicroEXMAT, has been
reported using CCS SF hardware and
methods (11). Currently, a multi-
variate experimental design based on a
Box-Behnken central composite is
linked explicitly in the network via
an expert system, EXBOXB. Further
integration of MicroEXMAT into a full
laboratory information management
system (LIMS) was also outlined
previously (12). Applications to EPyA
and CHAMP are being developed. The
recent ACS Symposium on expert systems
applied to the environmental field
(13) indicates the growing importance
of AI in analytical chemistry.
IV. Summary
Newly designed instrumentation for
multimedia (air, water, solids)
environmental trace organic analysis
is described for on-site applications.
The automated prototype units feature
advanced sample processing with
interfaces for on-line analyses with
chromatographic and/or spectral
detectors. Thermal sample processing
is provided by EPyA. including modules
for purge and trap/thermal desorption,
dynamic headspace, and pyrolysis.
Nonthermal multi-sample processing is
conducted with CHAMP based on super-
critical fluid extraction and
specialty interface units. Analyses
of low ppb levels of vapors, aerosols/
particulates, gasoline, and soils
illustrate the proven capabilities of
the integrated modular systems. A
developing expert system network,
MicroEXMAT, encodes expertise to guide
analysts in analytical strategy,
instrumental configurations, and
aethod development for the proposed
on-site analyzers.
REFERENCES
1. Manufactured by CDS Instruments,
Division of Autoclave Engineers,
Oxford, PA.
2. Manufactured by Computer Chemical
Systems, Inc., Avondale, PA.
3. (a) Michael, L.C., Pellizzari,
E.D., Norwood, D.L., Environ. Sci.
Technol.. 25, 150-155 (1991), "Appli-
cations of the Master Analytical
Scheme to Determination of Volatile
Organics in Wastewater Influents and
Effluents."
(b) Applications Laboratory,
Chemical Data Systems, Inc., Oxford,
PA.
(c) Liebman, S.A., Ahlstrom,
D.H., Sanders, C.I., First FACSS
Mtg., Atlantic City, NJ, Nov 1974.
"Automatic Concentrator/GC System for
Trace Analysis."
(d) Ahlstrom, D., Kilgour, R.,
Liebman, S., Anal Chem.. 47. 1411
(1975), "Trace Determination of Vinyl
Chloride Monomer by a Concentrator/GC
System."
(e) Liebman, S.A., Wampler,
T.F. , Levy, E.J., EPAInternat.
Sympos. on Recent Advances in
Pollutant Monitoring of Air, Raleigh,
NC, May 1982, "Advanced
Concentrator/GC Methods for Trace
Organic Analysis."
4. Applications Lab., CDS
Instruments/ Division of Autoclave
Engineers, Oxford, PA.
5. Liebman, S.A., Smardzewski, R.R.,
Sarver, E.W., Reutter, D.J., Snyder,
A.P., Harper, A.M., Levy, E.J.,
Lurcott, S., O'Neill, S., Proc. Poly-
meric Materials Science and Engineer-
ing. 5.2, 621-625, Amer. Chem.
Soc., Los Angeles, CA, September 1988.
6. (a) Liebman, S.A., Levy, E.J.,
Lurcott, S., O'Neill, S., Guthrie, J.,
Yocklovich, S., J. Chromatogr. Sci..
27, 118-126 (1989), "Integrated
Intelligent Instruments: Supercritical
Fluid Extraction, Desorption, Reaction
and Chromatography."
7. Raymer, J.H., Pellizzari, E.D.,
Anal. Chera.. 59, 1043, 2069 (1987),
"Toxic Organic Compound Recoveries
Using SF C02 and Thermal Desorption
Methods."
8. Liebman, S.A., Fifer, R.,
Griffiths, P.R., Lurcott, S., Bergman,
B., Levy, E.J., Pittsburgh Conf..March
1989, Atlanta, GA, Paper No. 1545,
"Detection Systems for Supercritical
Fluid/GC Instrumentation: Flame
301
-------
lonization Detector (FID) and Fiber
Optic Monitor (FOM) Units."
9. Liebman, S.A., et al., Pittsburgh
Conf., March 1990, NY, Paper No. 546,
"New Applications of I3 in Trace
Organic Analysis."
10. (a) Siddiqui, K.J., Eastwood, D.,
Lidberg, R.L., SPIE. 1054. 77-90
(1989), Fluorescence Detection III:
Soc. Photo-Optical Instrument. Eng. ,
Bellingham, WA, "Expert System for
Characterization of Fluorescence
Spectra for Environmental
Applications."
(b) Eastwood, D., Lidberg, R.L.,
Simon, S.J., VO-Dinh, "An Overview
Advanced Spectroscopic Field Screening
with In-Situ Monitoring Instrumenta-
tion and Methods, private communica-
tion.
11. Liebman, S.A., Fifer, R.,
Morris, J., Lurcott, S., Levy, E.J.,
Intelligent Instruments and
Computers. May/ June 1990, pp 109-120,
"An Expert System Network for
Supercritical Fluid Technologies."
12. Liebman, S.A., Snyder, A.P.,
Wasserman, M., Brooks, M.E.,
Watkins, J., Lurcott, S., O'Neill, S.,
Levy, E.J., Internat. Conf. on Anal.
Chem., University of Cambridge, UK,
July/Aug, 1989, "Integrated
Intelligent Instruments in Materials
and Environmental Sciences."
13. Hushon, J.M., Ed., ACS Sympos.
Series 431, Amer. Chem. Soc.,
Washington DC, 1990, "Expert Systems
for Environmental Applications."
(a)
EPyA
CHAMP
(b)
SPECIALTY INTERFACES TO
FTIR AND MS, MS/MS SYSTEMS
ANALYTICAL
PYROLYSIS MODULE
—PYROPROBE*
TRANSCAP INTERFACES
TO GC/SFC, FTIR, MS SYSTEMS
Figure 1. (a) EPyA, the Environmental Pyroprobe Analyzer
(b) CHAMP, the Chemical Hazards Automated Multiprocessor
(c) TRANSCAP Interface to Finnigan TSQ MS/MS
302
-------
CARTRIOQE SAMPLING FOB LOW PPB LEVELS
OF HALOCAHBOMS, ALIPHATIC!. AND AROMATIC*
TEST AIR MIXTURE ANALYSIS
TENAX CARTRIDGE
CHLOHOMItZtNl «» frt
•INZCNE
J
--»_>^.
8AHPLINQ
• MIN. 40 •! N«
oc.
OURAIONO 01 « 3 OH K O >I m 1 U FILH
• */MI«. TO 17**C
ATT'M 10"11 * 11
Figure 2. Cartridge Sampling for Low PPB Levels of Halocarbons,
Aliphatics, and Aromatics in Air with Thermal Desorber Module
TUT SAMPLES WITH WIDE-RANQINQ VOLATILES
FOR CBYOTHAPPIMO. DESORPTIOM AND CAPUAMY OC ANALYSIS
SAMPLE CONCENTRATOR CDS S3O/GC
iri.ITl.Ilt C»HU.»«» OC
WITH CHYOFOCUIM*
Figure 3. Gasoline and Diesel Fuel Test Mixture Analyzed with
Cryotrapping, Thermal Desorption, and Capillary GC/FID System
303
-------
(a)
OIRECT COtUMH CUTOFOCUtlMQ
SflO -1 OUTSIDE Alfl
• ETCNTION TlMf 10
ISO ml LABOtUTOftT AMI ( b )
(Cj
SOOml PAINT SHOr A.IN
MIEHTKM TIME 10
Figure 4. Air Monitoring on Tenax Cartridge with Direct Colu»n
Cryofocusing, 500 ml Sample
(a) Outside Air, (b) Lab Air, (c) Paint Shop Kir
SAMPLE CONCENTRATOR - QC/M*1S SPECTROMETER
CARTRIDGE AEROSOL/PARTICULATE
(a)
RECOWTHUCTTD Kttt
CKftOHATOOAAM
(b)
Figure 5. (a) Reconstructed Ion Chromatogram of Cartridge
Aerosol/Particulate. Thermal Degradation
(260°C/5 min) GC-MS Analysis of Microcap fl
(b) Electron Impact MS of Scanset 1529
304
-------
SUPERCRITICAL FLUID EXTRACTION-CHROMATOGRAPHY
EXTRACTION
SIMULANT IN SOIC
1880,Hgm
CHROMATOGRAPHY
BIS-CETHYLHEXYL)
PHOSPHONATE
>
Figure 6. Supercritical Fluid Extraction-Chromatography (SFE-SFC/FID)
of Bis-(Ethylhexyl)phosphonate, 3000 psi CO2 Mobile Fluid
SUPERCRITICAL FLUID EXTRACTION-CHROMATOCRAPHY
(a)
SFE-SFC/ITD
(b)
SPECTRUM
11'I. IP A*"'.*?'!1 """."?'" w. .,*J
m in MI s
Figure 7. SFE-SFC Interfaced to Ion Trap Detector (ITD) of Polyolefin
Mixture with Naphthalene
(a) Reconstructed Chromatogram
(b) Mass Spectrum, m/z 128
305
-------
USING A FID-BASED ORGANIC VAPOR ANALYZER IN CONJUNCTION WITH
GC/MS SUMMA CANISTER ANALYSES TO ASSESS THE IMPACT OF LANDFILL
GASSES FROM A SUPERFUND SITE ON THE INDOOR AIR QUALITY OF AN
ADJACENT COMMERCIAL PROPERTY
Thomas H. Pritchett
U.S. Environmental Protection Agency
Edison, NJ
David Mickunas and Steven Schuetz
IT Corporation, REAC Contact
Edison, NJ
The ERT was tasked to access the degree that VOCs, which
may have been co-migrating with methane from a Superfund
site, were affecting the indoor air quality of a shopping mall.
Of particular concern to the Region was the fact that the mall
had actually been built on top of the site prior to its being
added to the NPL, The actual assessment used a combination
of both field screening methods and fixed laboratory meth-
ods to gather two separate sets of data: one set on the landfill
gases and the other set on the air inside the mall. OVA,
Explosivity, and HNU readings from all of the landfill vent
were used to select the vents from which the Summa canis-
ters would be taken for GC/MS and permanent gas analyses.
Concurrent with Summa sampling, the inside of the mall was
screened using an OVA - particularly at all of the likely
entry points for subsurface gases.
The analytical results were interpreted as follows: The
Summa results were used to determine the "worst case" ratio
of target compound to methane observed in the vent gases.
These values were then multiplied by the worst OVA read-
ings observed in the vicinity of a likely soil gas entry point
in order to predict the highest possible concentration of
VOCs that could have been present due to co-migration with
the methane from the landfill. These "worst case" predic-
tions clearly indicated that there was not an apparent long-
term health risk due to VOC migration from the landfill.
307
-------
FIELD ANALYTICAL SUPPORT PROJECT (FASP) USE TO PROVIDE DATA FOR
CHARACTERIZATION OF HAZARDOUS WASTE SITES FOR NOMINATION TO
THE NATIONAL PRIORITIES LIST (NPL):
ANALYSIS OF POLYCYCLIC AROMATIC HYDROCARBONS (PAHS)
AND PENTACHLOROPHENOL (PCP)
Lila Accra Transue, Andrew Hafferty, and Dr. Tracy Yerian
Ecology and Environment
101 Yesler Way, Suite 600
Seattle, Washington 98104
ABSTRACT
The path from initial discovery of a site
as potentially contaminated to its inclu-
sion on the National Priorities List (NPL)
requires numerous activities, most impor-
tantly the identification and quantitation
of hazardous wastes or contaminants asso-
ciated with the site and the surrounding
area. New guidance for NPL nomination
places greater emphasis on accurate deter-
•ination of the areal and volumetric extent
of contamination during the site assessment
phase of work. Under this guidance, exten-
sive sampling is a prerequisite for charac-
terization of a site. This places a heavy
burden on the United States Environmental
Protection Agency (EPA) regions' ability to
provide quality assurance oversight for
data generated by Contract Laboratory Pro-
gram (CLP) analysis of these samples, and
adds considerable costs and time to the
nomination process. If the contaminants of
concern have been identified previously, it
•ay be appropriate to characterize the site
using field analytical support. In Region
10, the Field Analytical Support Project
(FASP) program has been integrated into the
Screening Site Inspection (SSI) and Listing
Site Inspection (LSI) process to provide
cost savings and near real-time analytical
information about the site. FASP methods
are designed to meet the data quality ob-
jectives (DQOs) established for each site.
All FASP data used for site characteriza-
tion are confirmed by analyzing 10 percent
of the samples collected for full Target
Compound List (TCL) analysis through the
CLP. Gas chromatographic methodologies for
field analysis of selected PAHs and PCP
have been developed for FASP in response
to a regional need for site characteriza-
tion at wood treating facilities. FASP
methods are developed for small volumes,
rapid extraction and analysis, and minimum
labor intensity. Methods developed for
FASP will be presented, as well as the
results from two LSIs, including a compari-
son of .FASP data to CLP confirmation re-
sults at each site.
INTRODUCTION
The United States Environmental Protection
Agency (EPA), under the Superfund Amend-
ments and Reauthorization Act of 1986
(SARA), uses the National Hazardous Waste
Site Investigation program to identify
hazardous waste sites for inclusion on the
National Priorities List (NPL). Ecology
and Environment, Inc. (E & E) holds the
Zone 2 Field Investigation Team (FIT)
contract, under which potential hazardous
waste sites are investigated, and relative
risks and threats to human health and the
environment are evaluated. FIT assists the
EPA in its goal of identifying sites for
the NPL in three stages: 1) Preliminary
Assessments (PAs), 2) Screening Site
Inspections (SSIs), and 3) Listing Site
Inspections (LSIs). A potential hazardous
waste site would go through all three
phases before it could be listed on the
NPL.
In 1988, EPA released the proposed re-
visions to the Hazard Ranking System (HRS),
which is used to score potential hazardous
waste sites based on an assessment of rela-
309
-------
tive risks. Prior to the revised HRS
(rHRS), the extent of contamination at a
hazardous waste site was determined only
after the site actually was placed on the
NPL (during the remedial investigation
phase of site cleanup). The rHRS includes
nev guidance for nomination to the NPL, and
places greater emphasis on accurate deter-
mination of the areal and volumetric extent
of contamination during the site assessment
phase of work. Coupled with congressional
mandates aimed at streamlining the listing
process, the new guidance places a heavy
burden on the limited analytical resources
available in terms of the number of samples
required for accurate site characteriza-
tion, and rapid turnaround of analytical
data after sample collection.
The site assessment program obtains most of
its required data through the EPA CLP,
since the CLP provides cost-effective
analyses for a large number of
contaminants. Sometimes, however, it may
be impractical to utilize the CLP to
characterize a site if preliminary data are
already available that identify the target
analytes of concern. The costs and time
involved with a large-scale sampling plan
can be minimized by tailoring the type of
sample analyses performed to the specific
project needs. Also, information obtained
from the laboratory during the sampling
event may allow the field team to optimize
sample locations for proper identification
of site boundaries, while minimizing the
total number of sample analyses required.
These types of laboratory interaction and
sample location tailoring are currently
difficult to obtain through CLP Routine
Analytical Services (RAS).
In addition, RAS contract required quanti-
tation limits may not be adequate to deter-
mine the extent of on-site contamination at
sites where the NPL listing criteria estab-
lishes a need for the lowest obtainable
quantitation limits. It also is possible
that CLP methodology may be inappropriate
under specific matrix conditions present at
a site, potentially resulting in further
elevation of the quantitation limit above
required action levels. Determination of a
matrix interference in advance, through
real-time analysis, may allow for modifica-
tion of the CLP method as requested through
the Special Analytical Services (SAS) pro-
cess, to minimize the necessity of resamp-
ling.
This paper describes an alternative to the
exclusive use of full organics and in-
organics CLP RAS analysis of samples col-
lected during the SSI and LSI processes.
When compared to CLP RAS, this alternative
often results in cost and time savings
while providing analytical information that
satisfies the data quality objectives
(DQOs) for each site.
DQOs
DQOs are statements regarding the level of
uncertainty that a data user or decision-
maker is willing to accept in results de-
rived from environmental measurements. The
DQO process is designed to help the data
user match quality needs with the appro-
priate analytical laboratory and methods so
that the right type, quality, and amount of
data are collected (1).
When applied to hazardous waste site inves-
tigations, the DQO process provides a quan-
titative basis for designing rigorous, de-
fensible, and cost-effective investiga-
tions. The DQO planning process recognizes
that decision making is driven by regula-
tory requirements and by risks to public
health and that the uncertainty in deci-
sions will be affected by the type and
quality of data collected. DQOs provide a
qualitative and quantitative framework
around which data collection programs are
designed, and can serve as performance
criteria for assessing projects (2).
DQOs determine the level of analytical sup-
port necessary to provide decision-makers
with sufficient confidence upon which to
select options with known levels of uncer-
tainty. Choice of specific analytical op-
tions may be determined by:
o Health-based concerns,
o Sample analysis cost,
o Analytes of concern or target/indicator
analytes,
o Regulatory action levels that dictate
method quantitation limits,
o Sample matrices,
o Sample collection, handling, and storage
requirements, and
o Statistical uncertainty in the qualita-
tive identification of analytes and
errors associated with the quantitation.
310
-------
All of the above considerations must be
weighed to determine the appropriate analy-
tical needs for the project data. Rarely,
if ever will a single analytical program
provide the best technical information and
the most cost effective solution to address
all concerns at the site.
The "art" of field analytical support is to
natch analytical capability to the DQOs re-
quired for a specific site in a cost-effi-
cient manner. Once the acceptable level of
error in the result is determined, the
acceptable level of inherent error in the
Measurement system can be addressed.
FIELD ANALYTICAL SUPPORT PROJECT (FASP)
PROGRAM
Broadly defined, field analytical support
is the use of chemists in an analytical
laboratory at or near the site of a hazar-
dous waste investigation, removal, or re-
Bed ial action. Field analytical support is
•ore than a facility or vehicle stocked
with instrumentation, glassware, and
expendables; it is the interactive
management process by which decision-makers
and the personnel who provide the
analytical results integrate planning,
execution, and assessment of analytical
data collection into environmental studies.
These procedures form the basis of the FASP
program.
In the late 1970s and early 1980s, field
analytical support for determinations of
contaminants at hazardous waste sites was
almost exclusively restricted to health and
safety monitoring of on-site personnel.
Early site screening was limited primarily
to air monitoring for volatile organic com-
pounds with hand-held instruments such as
the HNu PI101 (photoionization detection)
and the Foxboro OVA (flame ionization
detection). Within the last decade, more
sophisticated analytical instrumentation,
such as portable (hand-carried) and trans-
portable (mobile laboratory supported) gas
chromatographs and light-weight, compact
X-Ray fluorescence and atomic absorption
analyzers, have begun to be employed rou-
;tinely in hazardous waste site investiga-
tions. These new instruments, coupled with
field-experienced chemists, have provided
near real-time organic and inorganic
analyses for contaminants in air, soil,
Vater, and other matrices (3).
Under E & E's Zone 2 FIT contract, a FASP
program was initiated in 1984. The main
purpose of FASP is to support the PA, SSI,
and LSI process by utilizing field analyti-
cal methods to provide useful information
about site contaminants on a real- or near
real-time basis. FASP can be a cost- and
time-effective alternative or supplement to
conventional laboratory sample analysis in
many situations. Turnaround time for
conventional laboratory analyses, such as
CLP RAS is 40 days after receipt of the
samples. CLP data for site assessment
activities must undergo data validation by
a FIT chemist which takes approximately two
weeks. By contrast, FASP data are
generally provided verbally within 24 hours
of sample receipt, and a final deliverable
is often available approximately 14 days
after the project is completed. FASP data
are evaluated during laboratory projects.
Additional data validation time is not
required.
The EPA recognizes that field analytical
methods such as FASP provides, are appro-
priate for many decisions made in Superfund
(American Environmental Laboratory, October
1990). The EPA encourages the use of these
field analytical methods for screening,
monitoring and other assessments requiring
rapid turnaround of data, and for decisions
where unconfirmed analyte identity and
estimated concentrations are appropriate.
FASP methods are currently included in
EPA's revised Field Analytical Methods
Catalogue. FASP data have been used to:
o Optimize sampling grids,
o Select groundwater well screen depths,
o Guide remedial disposal requirements,
o Provide guidance to cleanup contractors,
o Assist in spill response,
o Select well locations based on soil gas
monitoring,
o Provide enhanced site characterization,
o Identify the most appropriate samples
for CLP analysis,
o Estimate waste quantities,
o Determine extent of contamination migra-
tion, and
o Find "hot-spots".
FASP is not a replacement for or an equiva-
lent of the EPA CLP. FASP does provide
real-time data of known (legally admis-
sible) quality, which may be used in situa-
tions where data generated by a certified
laboratory and standard methodology is not
a requirement for decision making. All
FASP analytes are, by definition, tenta-
311
-------
lively identified, and all FASP quantita-
tive data are estimated concentrations be-
cause methods and quality control (QC) are
a subset or variants of standard CLP QC.
Although both qualitative and quantitative
accuracy and precision may nearly equal
CLP, no attempt is made to alter these
limitations. Therfore, to properly iden-
tify FASP data as tentatively identified
with estimated concentrations, all FASP
data in Region 10 are annotated with the
qualifier "F". This qualifier also indi-
cates that field methodologies were
employed to generate the data.
FASP often is used at sites where previous
sampling has been performed and target
analytes have been identified. When
analytes have been identified previously,
unambiguous identification (i.e., mass
spectral detection) may not be required.
FASP is used most efficiently in the
analysis of samples for a limited group of
analytes requiring only one or two analyti-
cal methodologies. FASP is not used rou-
tinely for analysis of samples for unknown
contaminants.
FASP STANDARD OPERATING GUIDELINES (SOGs)
The FASP program functions under SOGs that
provide guidance on general QC and analyte-
matrix-specific methodologies which have
been developed within the FASP program.
Methodologies are developed on an as-needed
basis, to accommodate the FIT program, or
any other program in which FASP is uti-
lized. FASP methods are designed to pro-
vide near real-time data to field person-
nel. To accomplish this goal, the methods
utilize simplified sample preparation tech-
niques (disposable glassware, smaller scale
extractions) based on more exhaustive con-
ventional laboratory methods, such as CLP
methods. As field analytical methodologies
and the associated QC are generated, they
are standardized, reviewed by FASP
chemists, and submitted to EPA for review
by the Analytical Operations Branch (AOB)
Field Methods Workgroup for final approval.
By the use of standardized and approved
SOGs, consistent data of known quality are
generated.
Like EPA or other standard methods, SOGs
prepared for field analytical support pro-
vide information on the approximate pre-
cision and accuracy that the methods may
provide for sample analysis. However, FASP
methods often are tailored to meet site-
specific requirements. This increases the
probability of obtaining useful data by
overcoming matrix problems, establishing
appropriate quantitation limits for the
project DQOs, or focusing on specific
target analytes.
QC
FASP QC is based on the needs of the FIT
program and may vary according to the
analytical method and/or specific project
needs. There are, however, some general
guidelines provided by SOGs which are
consistently employed.
Instrument Calibration
Gas chromatographic response to target
analytes for the external standard method
of quantitation is measured by determining
calibration factors (CFs), which are the
ratio of the response (peak area or height)
to the mass injected. An initial calibra-
tion designed to demonstrate the instru-
ment's linear response is generated for
each target analyte by analyzing a minimum
of three standard concentrations which
cover the working range of the instrument.
Using the calibration factors calculated
from the initial calibration, the percent
relative standard deviation (ZRSD) is cal-
culated for each analyte at each concentra-
tion level. The percent relative standard
deviation generally is required to be less
than or equal to 25 percent.
The mean initial calibration factor for
each analyte is verified by the continuing
calibration during each operational period
(daily) to ensure detector stability. Mid-
range standards are analyzed, and calibra-
tion factors are compared to the mean
initial calibration factor for each
analyte. The relative percent difference
generally is required to be less than or
equal to 25 percent. If the continuing
calibration criteria are not met for each
target analyte, a new initial calibration
is performed.
Final calibrations are performed at the end
of a project, or sampling effort to ensure
analytical instrument stability. The cali-
bration factor from the final calibration
is compared to the mean initial calibration
factor for each analyte. The relative per-
cent difference is required to be less than
or equal to 50 percent. If the relative
percent difference meets continuing cali-
312
-------
bration criteria, the final calibration
also may be used as a continuing calibra-
tion.
Analyte Identification and Quantitation
Qualitative identification of target
analytes is based on both detector selec-
tivity and relative retention time as com-
pared to known standards, using the
external standard method. Generally,
individual peak retention time windows
should be less than ±5 percent for packed
columns.
The concentration of an analyte in the
sample is calculated using the calibration
factor for that analyte calculated from the
continuing calibration. Reported results
are in micrograms per kilogram (ug/kg)
vithout correction for blank results, spike
recovery, or percent moisture.
Sample chromatograms may not match identi-
cally with those of analytical standards.
Vhen positive identification is question-
able, the chemist may calculate and report
a maximum possible concentration (flagged
as < the numerical value) which allows the
data user to determine if additional (e.g.,
CLP RAS or SAS) analysis is required or if
the reported concentration is below action
levels and project objectives and DQOs have
been met.
Similarly, when sample concentration ex-
ceeds the linear range, the analyst may
report a probable minimum level (flagged as
> the numerical value) which allows the
data user to determine if additional (e.g.,
CLP RAS or SAS) analysis is required or if
the reported concentration is above action
levels and project objectives and DQOs have
been met.
Blank Analysis
A method blank is performed with every set
of samples extracted; a minimum of one
•ethod blank per 20 samples is performed.
The method blank must contain less than the
project quantitation limit, the minimum
reportable value, for each target analyte.
Matrix Spike Analysis
Accuracy is defined as the closeness to
-------
analyte concentrations may be used as a
comparison of the two data sets.
FASP POLYCYCLIC AROMATIC HYDROCARBONS
(PAHs) ANALYTICAL METHODOLOGY
FASP PAH methodology provides identifica-
tion of a subset of the base/neutral acid
(BNA) compounds included on the CLP Target
Compound List (TCL). The method provides
tentative identification of the PAH com-
pounds listed below, at estimated concen-
trations:
Naphthalene
Acenaphthylene
Acenaphthene
Fluorene
Phenanthrene
Anthracene
Fluoranthene
Pyrene
Chrysene
Benzo(a)anthracene
Benzo(b)fluoran thene
Benzo(k)fluoranthene
Benzo(a)pyrene
Indeno(1,2,3-cd)pyrene
Dibenzo(a,h)anthracene
Benzo(g,h,i)perylene
For the soil matrix, a veil homogenized 2
or 3g sample is weighed into a disposable
culture tube with a Teflon-lined cap. The
sample is extracted with 6 mLs of methylene
chloride twice by vortexing for 2 minutes,
combining the extracts. The final extract
is dried with a small amount of sodium
sulfate and then solvent exchanged into
isooctane.
Isolation of the target analytes is accomp-
lished by a small-scale silica gel column
cleanup. A disposable glass 4 mL giant
pipette is filled with a plug of glass
wool, silica gel, and sodium sulfate. The
column is eluted first with methylene
chloride, then petroleum ether (10 mLs of
each). The sample, in isooctane, is then
introduced onto the column. After the
sample is introduced to the column, the
column is first eluted with petroleum ether
(6 mLs) in order to allow interfering
contaminants, such as hydrocarbons, to be
removed. The PAHs are then eluted with
methylene chloride (10 mLs), and the final
volume of the extract is reduced to 1.0
mL under a stream of nitrogen.
The sample is analyzed by gas chromato-
graphy, using a J&U 0.53 mm x 15 m DB-5
fused silica megabore column and employing
flame ionization detection. A temperature
program is utilized to optimize separation
of the analytes. The gas chromatographic
analysis time is approximately 30 minutes.
Samples are quantitated using the external
standard method. Standard mixes are pur-
chased from a commercial manufacturer and
diluted to appropriate concentrations for
instrument calibration. Calibration
factors are calculated for each analyte in
the initial and continuing calibrations.
The concentration of the analyte(s) in a
sample is calculated based on the analyte
calibration factors calculated from con-
tinuing calibrations.
The quantitation limits for the FASP PAH
methodology are 1,000 ug/kg, while CLP RAS
required quantitation limits are 330 ug/kg.
As the CLP samples do not undergo silica
gel cleanup, the final matrix potentially
contains a higher degree of interference
from petroleum hydrocarbons, which are
often present along with the PAHs. When
petroleum hydrocarbon interferences are
present, the sample often requires dilution
before an accurate analysis can occur.
This results in an elevation of the actual
contractual quantitation limits. Samples
analyzed by FASP methodology are relatively
free of these interferences, and generally
do not require dilution.
The total time for preparation and analysis
of 10 soil samples is 490 minutes. In a
10-hour day, the maximum capacity for a
field analytical laboratory equipped with
one gas chromatographic system is approxi-
mately 11 samples during the first day of
operation, and 20 samples each day there-
after. This projected capacity does not
take into account any dilutions which may
be required when high target analyte levels
are present.
This method employs only disposable glass-
ware, eliminating time required for clean-
ing glassware, and minimizing the potential
for cross contamination. Solvent volumes
are minimal, requiring a total of only 40
mLs per sample, compared to the CLP method
for BNAs which requires 300 mLs of solvent
per extraction.
314
-------
FASP PENTACHLOROPHENOL (PCP) ANALYTICAL
METHODOLOGY
For soil, a well homogenized 2 or 3g sample
is weighed into a disposable culture tube
with a Teflon-lined cap. The soil is dried
by adding a small amount of sodium sulfate.
The sample is then extracted with methanol
(10 mLs) by vortexing for 2 minutes. Five
mLs of the extract is transferred into a
clean culture tube.
The extract is derivitized with a solution
of pentafluorobenzyl bromide and hexacyclo-
octadecane (18-crown-6 ether) in 2-pro-
panol. One mL of the derivitization solu-
tion is added to the sample extract, along
vith 3 mg of potassium carbonate. The
culture tube is then capped, gently shaken,
and left in a hot water bath at 80°C for 4
hours. The culture tube is allowed to
cool, then the sample is extracted with 5
•Ls of hexane by vortexing for 1 minute.
Five mLs of carbon-free water are added to
the culture tube, and vortexed for an addi-
tional minute. The hexane layer, which
contains the derivitized PCP, is trans-
ferred to a clean culture tube and dried
Vith a small amount of sodium sulfate. The
extract is then ready for analysis.
The extract is analyzed by gas chromato-
graphy using a 1.0 m, glass column packed
with 1.5% SP-2250/1.95% SP-2401 and employ-
ing electron capture detection. The iso-
thermal column oven temperature is 275°C,
and gas chromatographic analysis time is
approximately 20 minutes.
Sanples are quantitated using the external
standard method. Standards, blanks, and
appropriate quality control samples are
prepared with each batch of samples de-
rivitized.
The quantitation limit for PCP using this
•ethodology is 50 ug/kg. The quantitation
Unit for PCP by CLP BNA methodology is
significantly higher (1,600 ug/kg). FASP
•ethodology allows for the lower quantita-
tion limit by isolating the PCP present in
the sample and removing matrix interfer-
ences, and then using a more sensitive in-
strumental technique (GC/ECD).
The total time for preparation and analysis
of 10 soil samples for PCP is 530 minutes.
In a 10-hour day, the maximum capacity for
afield analytical laboratory equipped with
one gas chromatographic system is approxi-
mately 10 samples during the first day of
operation, and 20 samples each day there-
after. This projected capacity does not
take into account any dilutions which may
be required due to high target analyte con-
centration in the sample.
This method, like the PAH method, employs
only disposable glassware, and consumes
only minimal solvent volumes (21 mLs total)
compared to CLP solvent volumes of 300 mLs
per sample extracted.
CASE STUDY 1
E & E was tasked to perform an LSI at an
active wood treating facility occupying 19
acres in Oregon. The facility operations
involve pressure treating wood products
using creosote (containing PAH compounds)
and PCP in a petroleum oil carrier. The
determination of the extent of on-site sur-
face contamination was defined as one of
the objectives of the LSI, requiring
analysis of 56 on-site grid surface soil
samples. Since the target analytes were
known, it was determined that site-specific
DQOs could be met by using FASP at a sub-
stantial cost and time savings compared to
a full CLP sample analysis scheme.
Sixty-two surface soil samples were col-
lected at the site for FASP analysis, in-
cluding six duplicate, or colocated
samples. The samples were shipped to the
FASP Seattle Base Laboratory for analysis,
as the project was not large enough to
justify mobilization. The sample analyses
were completed within 24 hours of receipt
of the last sample shipment.
A cost comparison was calculated for FASP
versus CLP RAS analysis of the samples.
The total FASP costs included the purchase
of required expendables, which totaled
approximately $4,546.00 and labor, which
totaled approximately $13,300 for 350 hours
of effort. If CLP had been utilized for
these analyses, the total cost would have
been $27,308, which accounts for laboratory
charges and data validation. This amounts
to a savings of $9,461 by utilization of
the FASP program. This comparison indi-
cates that full organics CLP RAS would not
be appropriate for these samples. Rather,
a focused analysis, such as CLP SAS or FASP
would be more appropriate. For near real-
time availability of sample data, FASP
would be the preferred alternative.
The confirmatory samples were analyzed for
315
-------
BNA compounds by a CLP laboratory at a fre-
quency of approximately 10 percent (8
samples). Sample quantltation limits vere
consistently higher for the CLP data set
due to the matrix interferences from the
oil present in the samples. For most
samples, quantitation limits were elevated
2 to 300 times above the contract-required
quantitation levels.
Correlation between the FASP and CLP data
sets was excellent. FASP identification of
PAHs and PCP vas confirmed, and relative
trends in concentrations generally agreed.
A statistical analysis of the data sets was
performed using correlation coefficients.
FASP and CLP data sets were compared for
analytes where four or more pairs of data
points were available (i.e., four or more
samples sent for confirmatory analysis had
results above method quantitation limits
for the analyte). The calculated correla-
tion coefficients are summarized in Table
1.
Table 1. CORRELATION COEFFICIENTS FOR FASP
AND CLP DATA: CASE STUDY 1
Analyte
Data
Pairs
Used
Correlat ion
Coefficient
(r)
Phenanthrene/
Anthracene 6 0.999
Fluoranthene 6 0.999
Pyrene 6 0.999
Chrysene/
Ben7o(a)anthracene 8 0.9997
Benzo(b)fluoranthene/
Benzo(k)fluoranthene 8 0.9775
Benzo(a)pyrene A 0.9703
Pentachlorophenol 6 0.9696
As a result of the FASP analysis and CLP
confirmation, the data generated by FASP
were determined to be acceptable for use in
determining the on-site hazardous waste
quantity. This allowed data users to
accurately measure the relative risks
resulting from on-site contamination.
CASE STUDY 2
An LSI was performed at an inactive pipe-
coating facility, which had generated coal
tar, coal tar epoxies, asphalt, and cement
mortar wastes over the 51 acres for
approximately 30 years. Several target
analyte groups had been identified pre-
viously, including volatile organic com-
pounds, PAHs, and polychlorinated biphenyls
(PCBs). The project objectives required
on-site surface soil contamination to be
characterized. An on-site grid sampling
pattern was used, resulting in collection
of 54 samples.
Previous site sampling events had identi-
fied the target analytes, allowing for FASP
analysis of the on-site surface soil
samples while maintaining the project DQOs.
The soil samples were analyzed for volatile
organic compounds, PAHs, and PCBs at the
FASP Seattle Base facility. It was more
cost-effective to analyze the samples at
the base facility due to the variety of
analyses required and the relatively small
size of the project.
The cost of FASP analysis of the 54 samples
and four field duplicate samples vas
$20,900 ($1,900 for supplies, $19,000 for
labor) compared to CLP analysis costs which
would have totaled $57,408. This amounted
to a total savings of $36,508 by utilizing
FASP. All sample analyses were completed
within 7 days of the last sample shipment
date.
Six samples (approximately 10 percent of
the total number of samples) were split and
sent to a CLP laboratory for confirmatory
volatile, BNA, and pesticide/PCB analysis.
Again, matrix interferences prevented CLP
BNA analysis without elevated quantitation
limits due to the presence of oil. FASP
methodology, involving sample cleanup for
specific analyses, removed much of the oil
interference.
Correlation between the two data sets was
excellent. FASP identification of volatile
compounds, PAHs, and PCBs was confirmed by
CLP data, and relative trends in analyte
concentrations agreed. Calculated correla-
tion coefficients were generated where four
or more data pairs were available. One
split sample contained extremely high
levels of PAHs. CLP results were signifi-
cantly and consistently higher than the
FASP results for all PAHs detected in this
sample. It is most likely that this
phenomenon was due to the non-homogeneous
nature of the soil matrix. Therefore, this
data pair was not included in the correla-
tion coefficient calculation. The correla-
tion coefficients are presented in Table 2.
316
-------
Table 2. CORRELATION COEFFICIENTS FOR FASP
AND CLP DATA: CASE STUDY 2
DataCorrelation
Pairs Coefficient
Analyte Used (r)
Fluoranthene 4 1.000
Pyrene 4 1.000
Chrysene/
Benzo(a)anthracene 4 1.000
Benzo(b)fluoranthene/
Benzo(k)fluoranthene 4 1.000
Benzo(a)pyrene 4 1.000
Indeno(l,2,3-cd)
pyrene/Dibenzo(a,h)
anthracene 4 0.999
Benzo(g,h,i)perylene 4 0.999
Aroclor 1254 5 0.945
A statistical analysis of matrix spike re-
covery data for eight samples collected at
both of the sites described above is pre-
sented in Table 3.
CONCLUSION
Recently, EPA has placed a greater emphasis
on the determination of extent of contami-
nation during site assessments. FASP was
initiated under E & E's Zone 2 FIT contract
in 1984, and is a viable alternative or
supplement available to address the
analytical demands for determining relative
risks at hazardous waste sites. FASP
provides data of known quality, using
standard methodologies and QC modified to
meet the project DQOs. FASP data can be
obtained at a substantial cost and time
savings when compared to conventional CLP
analysis, and has been used successfully
for characterization of sites with known
target analytes.
Table 3. AVERAGE MATRIX SPIKE RECOVERIES
FOR SOIL SAMPLES AT HAZARDOUS WASTE SITES
Analyte
Average
Percent
Recovery
Standard
Deviation
Naphthalene 70.0 36.8
Acenaphthylene 103 47.3
Acenaphthene 94.3 36.6
Fluorene 90.3 24.9
Phenanthrene/
Anthracene 93.4 38.9
Fluoranthene 118 53.5
Pyrene 123 53.8
Chrysene/Benzo(a)
anthracene 121 34.5
Benzo(b)
fluoranthene/
Benzo(k)
fluoranthene 107 26.6
Benzo(a)pyrene 112 24.0
Indeno(l,2,3-cd)
pyrene/
Dibenzo(a,h,)
anthracene 98.7 25.8
Benzo(g,h,i)
perylene 88.0 28.8
Pentachlorophenol 122 51.4
REFERENCES
1. Cram, S.P., American Environmental
Laboratory, September 1989, pp. 19.
2. Neptune, D., E.P. Brantly, M.J.
Messner, D.I. Michael, May-June 1990,
Hazardous Materials Control, Volume 3,
Number 3, pp. 19.
3. Hafferty, Andrew, September 1989, "A
Cost Summary of Field Screening Implementa-
tion in Region 10", Division of Environ-
mental Chemistry, Proceedings, American
Chemical Society National Meeting, Miami
Beach, Florida.
DISCUSSION
DOUG PEERV: You were talking about doing 20 samples in a ten-hour day with
30-minute run time. Does thai include your QA/QC or did you have another four
hours of work time to cover that?
LILA ACCRA-TRANSUE: We did ten samples or 20 sample analyses. So that
includes the QC samples that we need to run.
DOUG PEERY: So you're talking about your standards and your QC's within
that 20 number.
LILA ACCRA-TRANSUE: Right.
VICKI TAYLOR: How many split sample pairs did you take?
LILA ACCRA-TRANSUE: We take approximately 10%. For the first project
we'd taken eight and for the second project, six.
VICKI TAYLOR: So you were basically presenting a correlation coefficient for
all the split samples that you took?
LILA ACCRA-TRANSUE: Right. All of the comparable data pairs are reported
where they were hits in both samples.
317
-------
Thermal Desorption Gas Chromatography-Mass Spectrometry
Field Methods for the Detection of Organic Compounds
A. Robbat, Jr., T-Y Liu, B. Abraham, and C-J Liu,
Tufts University, Chemistry Department, Trace
Analytical Measurement Laboratory,
Medford, MA 02155
INTRODUCTION
The overwhelming amount of information required to characterize
purported hazardous waste sites, as well as to support Superfund
site cleanup and closure activities, have catalyzed the development
of field instrumentation capable of providing site managers with
immediate access to chemical and physical data. The demand for
field "practical" methods and instrumentation has been recognized
by the U.S. Environmental Protection Agency (1, 2).
Faster data turnaround times and ease of operation have been the
primary motivation for selecting field gas chromatographic (GC)
methods of analysis. Despite recent advancements in field GC
instrumentation, typical applications focus on the detection of EPA
listed volatile organic compounds (VOCs) in water, air, or soil
gas. The primary limitation of commonly employed field GC's is
the non-definitive signal response of the detectors (including
photoionization, flame ionization, thermal conductivity, and
electron capture) which are incapable of providing unambiguous
identification of the wide variety of organic compounds that may
be present in a highly contaminated sample. Generally, ten to
twenty percent of the samples analyzed on-site are "split" for
confirmation by GC with mass spectrometric (MS) detection.
Since most commercially available mass spectrometers have
traditionally been housed and operated in a clean air, temperature
controlled room and the notion that economies of scale require
highly trained MS operators to be based in multi-MS laboratories,
misapprehensions have arisen as to whether MS's can be operated
successfully (and profitably) in the field.
The limited availability of field GC-MS's is not a function of MS
operating requirements, but more, the perception that significant
sample cleanup and QA/QC procedures will be required to obtain
useful data as well as the apparent reluctance of instrument
manufacturers to enter the field marketplace. Until recently, these
misconceptions have perpetuated the myth that GC-MS's belong
solely in the laboratory.
Over the last several years, we have discussed field GC-MS
applications utilizing Bruker Instruments' mobile mass
spectrometer (2-6). The MS, initially designed for NATO as a
chemical warfare detector, was manufactured from the outset as a
field instrument. In our studies, the MS was transported from
site-to-site in a mid-sized truck and was battery operated for - 8
to 10-hr at ambient conditions. For example, samples have been
analyzed with outdoor conditions, where; temperatures have been
between 10 "F and 90 "F, rain, snow, and high humidity. Gas
cylinders were not necessary for GC operation since charcoal
filtered ambient air served as the carrier gas.
Simple field methods have been developed based on analyte
introduction by thermal desorption CTD) followed by fast GC
separation and MS detection. Screening level and more
quantitative TDGC-MS methods have been submitted to EPA's
EMSL-Las Vegas for VOCs in water, soil/sediment, soil gas, air
and polychlorinated biphenyls (PCBs) and polycyclic aromatic
hydrocarbons (PAHs) in soil/sediment for inclusion into the
compendium of field methods that will be published by EPA's
Analytical Operations Branch. The methods include a menu of
QA/QC procedures whose implementation depends upon a given
study's objectives. The goal is to provide a practical GC-MS tool
that can deliver the quality of data required for the study with
minimal sample cleanup. Presented in this paper are typical
examples of data quality and a comparison of field and laboratory
results one can expect from both the screening and more
quantitative field TDGC-MS methods for PCBs, PAHs, and
pesticides.
EXPERIMENTAL SECTION
A mobile mass spectrometer (Bruker Instruments, Billerica, MA)
was used in these studies. The TDGC-MS was powered by
battery or electrical supply from the site. The MS was transported
to Superfund sites in Westborough (Hocomonco Pond; PAHs) and
319
-------
North Dartmouth, MA (Resolve; PCBs) in a Chevrolet Blazer, In
addition to the instrument's internal data collection and monitoring
system, the MS was equipped with an external data system and
thermal desorption sampling probe. Sample introduction was
made by thermally desorbing (TD) the analyte directly from
soil/sediment or from an organic extract through the TD sampling
probe's (SP) short 3.5 m fused silica capillary column. For direct
TD soil/sediment experiments, 0.5 g of soil was placed on an
aluminum foil covered petri dish. An internal standard was
injected into the soil before the measurement was made. In
contrast, the more quantitative measurements required several
additional steps: 1) 0.5 g of soil was weighed and extracted with
2 ml of solvent; 2) prior to extraction, a known quantity of
surrogate (or target) compound(s) was added to the soil (or field
blank) to determine extraction efficiencies (note: this step was
required since a single 2 ml extraction yielded analyte recoveries
of less than 100%; 3) co-inject known aliquots of extract and
internal standard onto aluminum foil covered petri dish; 4)
thermally desorb analyte. Shown below are the TDGC-MS
operating and PCB, PAH, and pesticide experimental conditions:
Operating Conditions
Mass Spectrometer Bruker Instruments (Billerica, MA)
electron energy 70 volts (nominal)
mass range 45 to 400 amu
scan time 2 sec
MS tune autocalibrate (HjO,,,; FC-77); 18, 69,
119, 169,331 amu
mass resolution set to unity; ca. 10% valley
definition
ion detection 17 stage Cu-Be dynode electron
multiplier with self-scaling
integration amplifier (108
linearity)
Sampling Probe Head 260 °C
GC Column
dimensions
DBS (J & W Scientific, Folsom, CA)
3.5m x 0.32mm i.d.; 0.25/x film
thickness
ambient air purified through carbon
filters
3 to 4 ml/min
earner gas
flow rate
PCBs PAHs Pesticides
initial temp 140°C, 30 sec 70°C, 40 sec 120 °C
temp prog 120°C/min 35°C/min 17°C/min
final temp 200°C, 90 sec 233°C, 80 sec 233°C
Internal d,«-pyrene d,-naphthalene d,0-
Standards or d,0-pyrene phenanthrene
Data were acquired by using the internal monitor's selected ion
monitoring program. The data system reported the total ion
current as a logarithmic value. The antilog value is used in
conjunction with MS response factors and analyte recoveries to
calculate concentrations in the sample. Standards were purchased
commercially from the following companies: PCBs (Ultra
Scientific, Hope, RI); PAHs (Supelco, Inc., Bellefonte, PA);
Pesticides (Chem Service, West Chester, PA); internal standards
(Cambridge Isotope Laboratories, Woburn, MA). All standards
and soil recovery experiments were prepared with high purity
solvents (> 96 %) as received.
RESULTS and DISCUSSION
The objective of this study was to develop fast TDGC-MS
methods (< 20 min/sample including sample cleanup). Two
methods were developed. Analyte introduction for quantitative
measurements were made by co-injecting organic extracts (or
standard solutions) of PCBs, PAHs, or pesticides and internal
standard(s) onto an aluminum covered petri dish followed by
TDGC-MS and for screening measurements by direct thermal
desorption from soil/sediment.
The surface monitor program mode was employed in this study.
Target compounds (maximum number twelve) were detected by
selected ion monitoring (SIM) MS. The (logarithm) ion current
was recorded and displayed visually on the system's monitor.
Found in Figure 1 are typical PCB and pesticide outputs. Three
fragment ions representative of each compound(s) and an
impossible ion (see below for rationale) were selected for
detection. For example, in cell A the target ions and their relative
intensities for the three monochlorinated PCBs were 188 (100%),
190 (33.5%), 152 (31.1%), and 189 (0%). Similarly, cells B-H
in Figure la illustrate the SIM four ion current responses for
chlorination levels 2 - 8, respectively; cell I, d,0-pyrene (internal
standard); cells J - K, PAH surrogates; and cell L, hydrocarbon
signals indicative of matrix complexity. Detection was made, and
printed on screen, when the signals from the four ions relative to
each other agreed to within preset criteria over a predetermined
retention time window. In this mode, SIM response may be
considered analogous to selective GC detection. Note above, that
the last fragment ion for the monochlorinated PCBs had a relative
intensity of 0%. Inclusion of an impossible ion served to provide
selective detection. For example, an increase in fragment 1 ion
current relative to fragments 2-4 within the target compound's
retention window precluded compound identification. Thus, the
mathematical algorithm assisted in screening out interferants
present in the sample.
solvent C»H,4
extraction
CH,C12
C6HM
320
-------
EURFACE MONITOR
M PC8/DIOXINS/CL
8
7-
6-
5-
4-
3-
2-
Die-PHEJWTHR
D-12-CHRYSENE
HYDROCHRBONS
CL3-BIPHET1YL
CL4-BIPHEWL
CL5-BIPHEWL
CLS-BIPHEJ1YL
B CL2-BIPHO1YL
G CL7-BIPHENYL
F 4.8
F 2.9
F 5.3
F 4.9
F 5.0
F 3.3
F 3.4
fiBCDEF6HI3KL
Figure la
5SFKZ KiiTO?. 28.18.315:44
V FESTICIuE
ft fl,3/E/0-BHD F 4.S
5 Die-Ft£?MHF£ F £.7
S-j CHEPTACllOR F 4.3 153
7-
6-
3-
4-
3-
2-
>H
^'-DOD F3.3 2
D ftLDRI/1 F 4.5
F DIEffi!?! F 3.8
EffiT-EFCKIDE F 4.4 [p
B 4,4 ' -DDE
1 VTT
A6!
.4 ' -rjDT
I
1
U F5.2
F5.4
)
7
DEFSHim
184
H
The fast GC linear temperature programs and MS detection
provided sufficient separation to identify compound(s) as shown in
Table 1. Figure 2 is a typical instrument print out for the amount
(4-ion total current count, in log values, left vertical axis) vs. time
response curves (horizontal axis) for four of the chlorinated
pesticides shown in Figure Ib. In addition to the compound and
amount detected, other information visible on the display included
"real-time" monitoring of: logarithm of ion current, left vertical
axis; MS vacuum pressure, right vertical axis; and column
temperature, above right vertical axis.
ajon toinot 38.u.a IBS
V PESTICIDE fflOUHT VS Tfft
I ft^E/HHC F 4.8
8-
7-
S"
4-
3-
2-
1
IS?
2
5
'V^'U !
# 1
tSSs
SUJFflK tttHITOR 3fl.1B.58 18:45
V PETICIIE5 fraW VS THE
t i£FT.EPIKir£ F4.4
1-
t
3-
4'
3-
1
172
2
i
H
i
WWHWMH ^V-V ^WM) 7
183s
aim nonna 38.ia.aa u:c
D MEW F 4.C
7-
£•
t
4-
3-
2-
171
2
!
f
,V*, .v"' A :
•nren KwrV n •?
imii
tffil
ssFsumnai 2a.ta.3B u;«
v PEsriciio nmin vs ire
S4/4'-0£ F5.2
J-
7-
£•
5-
3-
2-
1
173
2
S
1
i
':•.;•, It
Ssi'V ?
188 s
Figure Ib
igure 1. Typical Field TDGC-MS SIM response of a standard
ution containing PCBs (la) and chlorinated pesticides (Ib).
Figure 2. Amount versus time curvfr for several chlorinated
pesticides shown in Figure 1.
321
-------
TDGC-MS experiments were performed between the concentration
range of 40 and 4000 ng/compound. Repetitive measurements at
each concentration yielded differences in the log value of + 0.13
producing ion current differences of less than 30%. Table 1 lists
typical response factors (RF) and percent relative standard
deviations (%RSD) calculated for PCBs, PAHs, and pesticides
thermally desorbed from an organic extract. Plots of signal versus
concentration were linear (r= 0.999) with the %RSD for the
average RF less than 30%, meeting initial and continuing
calibration criteria in the Contract Laboratory Program. Table 2
lists representative RF and RSDs for PCBs and PAHs thermally
desorbed directly from soil. Despite somewhat larger percent
RSDs for some PAHs, measurement precision at this level will
only be critical at site cleanup "action" levels. It should be
pointed out that thermal desorption extraction efficiencies differ
greatly for some PAHs (see Table 3 for minimum detectable
quantity. Note: RF in Table 2 calculated over linear range as
shown in Table 3). Minimum detection levels for most
compounds were ~ 1 ppm for soil/solvent extraction and slightly
higher for direct soil thermal desorption. Because TDGC-MS
experiments can be performed in 5 to 20 min depending on the
method employed (with known data quality), many more analyses
can be performed than currently practiced for site characterization,
stockpiling, and worker/community protection activities. The
frequency for performing continuing calibration checks may be
determined (on-site) by following surrogate compound RF values
(see below).
Research has shown that compound recoveries vary with soil-type.
For example, PCB/hexane (0.5 g/2 ml hexane, 2 min) extraction
recoveries were 69 + 5% for 50 ppm backyard (organic) soil, 80
± 2 %, for 25 ppm sandy material from the Resolve Superfund
site in North Dartmouth, MA, and 73 ± 5 % for an ERA, 35
ppm, soil. Therefore, appropriate surrogate compound(s) and/or
target standards must be added to samples as the soil-type varies.
Such experiments can be used to determine instrument
performance as well.
Tables 4-7 illustrate typical examples of data quality one can
expect from the field TDGC-MS methods. Split samples were
collected by EPA's Region 1 oversight contractor and analyzed in
the field (Tufts) and lab (Lockheed ESC, Las Vegas, NV). Table
4 compares field and lab GC-MS measurements for total PCB
present in several samples obtained from the Resolve site while
Table 5 delineates chlorination level comparisons for two of the
samples. The field and lab results are in excellent agreement.
Shown in Tables 6 and 7 are field and lab comparisons for four
PAH samples from the Hocomonco Pond (Creosote contaminated
Superfund) site. Note that the samples in Table 6 and the sample
labeled HP-SB5 in Table 7 were performed by SIM using the
system's internal monitor as described above. In contrast, the
sample labeled pond (Table 7) was analyzed by total ion current,
selected ion monitoring extraction. The advantage of this
detection method was that full mass spectral fragmentation data
and compound library matching was applied. On the other hand,
the disadvantage was that ion current from matrix components may
add to the SIM signal resulting in higher concentrations than what
might actually be present. This, however, is no different than
what can occur using traditional CLP, MS methods. Field and lab
comparisons for PAH samples also appear to be in good
agreement.
Additional data will be presented describing further application of
the field TDGC-MS methods. Illustrations will be given
documenting cost effectiveness. Results will show that GC-MSs
can be operated in the field, provide rapid access of data, and
allow project managers to make decisions on-site.
ACKNOWLEDGEMENTS
Partial financial support for this project was provided by the U.S.
Environmental Protection Agency, EMSL-LV; New Jersey
Institute of Technology's Northeast Hazardous Substance Research
Center; and Tufts University's Center for Environmental
Management. The authors wish to thank EPA's Region 1
Hazardous Waste Division for providing access to Superfund sites
and samples and to the oversight contractors for their cooperation.
REFERENCES
1) Williams, L.R., Editorial Article, American Environmental
Laboratory, October, 1990 (see additional articles by EPA).
2) U.S. Environmental Protection Agency, Sixth Annual Waste
Testing and Quality Assurance Symposium, July 16-20, 1990,
Washington, DC; Field Analytical Methods Workgroup sponsored
by Analytical Operations Branch. See Proceedings, Robbat, A.,
Xyrafas, G., Abraham, A., "A Fast Field Method for The
Identification of Organics in Soil", 1-350..
3) "Method Evaluation for Field Analysis of PCBs and VOCs
Using a Field Deployable GC-MS", Xyrafas, G., Ph.D. Thesis,
Tufts University, Chemistry Dept., Medford, MA. O2155.
3) "A fieldable GC-MS for the Detection and Quantitation of
Hazardous Compounds: Analytical Chemistry in the Field?"
Robbat, A., Jr., Xyrafas, G., 198th American Chemical Society
National Meeting, 411, 29(2), 1989, Miami Beach, Florida.
4) "Evaluation of a Field-Based, Mobile, Gas Chromatograph-
Mass Spectrometer for the Identification and Quantification of
Volatile Organic Compounds on EPA's Hazardous Substance
List", Robbat, A., Jr., Xyrafas, G., In Proceedings of the First
International Symposium on Field Screening Methods for
Hazardous Waste Site Investigations, October 11-13, 1988, Las
Vegas, Nevada, Pg. 343.
5) "On-Site Soil Gas Analysis of Gasoline Components Using a
Field-Designed Gas Chromatograph-Mass Spectrometer", Robbat,
A., Jr., Xyrafas, G., In Proceedings of the First International
Symposium on Field Screening Methods for Hazardous Waste Site
Investigations, October 11-13, 1988, Las Vegas, Nevada, Pg. 481.
322
-------
Table 1. Thermal Desorption Field GC-MS Response Factors and
Percent Relative Standard Deviations - from Extract (Quantitative
Method)
Polvchlorinated Biphenvls
Table 2. Thermal Desorption Field GC-MS Response Factors and
Percent Relative Standard Deviations - Direct from Soil
Polvchlorinated Biphenvls
Chlorination Level
CM
Cl-2
Cl-3
Cl-4
Cl-5
Cl-6
Cl-7
Cl-8
Ave RF(n=5)
0.47
0.26
0.27
0.16
0.15
0.10
0.06
0.03
%RSD
20
17
17
15
12
15
17
10
Chlorination Level
Cl-1
Cl-2
Cl-3
Cl-4
Cl-5
Cl-6
Cl-7
Cl-8
Polvcvclic Aromatic
Ave RF(n=5)
13.44
3.75
3.91
2.55
2.02
1.61
1.04
0.36
Hydrocarbons
Polvcvclic Aromatic Hydrocarbons
naphthalene
acenaphthylene
acenaphthene
fluorene
phenanthrene/anthracene
fluoranthene/pyrene
chrysene/benz(a)anthracene
Chlorinated
BHCs
Heptachlor
Aldrin
Heptachlorepoxide
Dieldrin
4,4'-DDE
4,4'-DDD
4,4'-DDT
1.37
8.63
0.82
0.58
4.59
9.52
0.90
Pesticides
0.10
0.02
0.07
0.03
0.02
0.32
0.16
0.12
123
123
24J8
123
123
95
133
96
235
163
109
25.4
163
186
19.7
naphthalene
acenaphthylene
acenaphthene
fluorene
phenanthrene/anthracene
fluoranthene/pyrene
chrysene/benz(a)anthracene
2.39
1.21
0.33
0.16
0.25
0.06
0.003
RSD
19
25
23
16
16
16
23
19
75
356
52.1
165
213
310
229
323
-------
Table 3. PAH Dynamic Range Directly Desorbed from (0.5 g)
Soil Matrix.
Concentration Signal Linearity
Compound(s') Cne) (11=5) (rt
Table 4. Comparison of Field and Lab GC-MS Results for Total
PCBs in Samples from the Resolve Superfund Site, North
Dartmouth, MA
Naphthalene
Acenaphthylene
Acenaphthene
Fluorene
Phenanthrene &
Anthracene
Fluoranthene &
Pyrene
4000
2000
1600
800
120
80
40
4000
2000
1600
800
80
40
4000
2000
1600
800
80
40
4000
2000
1600
800
120
80
40
8000
4000
3200
1600
240
160
80
8000
4000
3200
1600
240
160
510084 + 22.9% 0.999
255648 ± 22.9%
210541 ± 31.2%
110357 ± 12.9%
28371 ± 13.2%
5312 ± 23.4%
1995'± 22.4%
255648 ± 22.9% 0.999
129245 ± 26.1%
94858 ± 10.8%
51454 ± 26.1%
3575 ± 13.2%
794'+ 17.2%
94858 ± 10.8%' 0.999
23397 ± 12.7%
20307 ± 22.9%
9936 ± 34.5%
740 ± 12.7%
251"+ 16.8%
52714 ± 11.0% 0.999
24197 ± 32.8%
18585 ± 12.7%
8971 ± 13.2%
371 ± 12.8%
794a± 13.4%
251"+ 15.4%
42578 ± 35.4% 0.999
21245 + 12.2%
16271 ± 26.1%
8084 ± 22.9%
877 + 24.3%
3981'+ 18.6%
316" ± 20.2%
17498 ± 24.3% 0.999
8629 ± 13.8%
6854 + 13.8%
3221 ±40.1%
371 ± 12.8%
195'+ 15.3%
Quantitative Screening Level
TDGC-MS TDGC-MS Lab GC-MS
EPA ID# (ppm) {ppm) (ppm)
TUF-RS-SO-A26-2-4 368.3 309.4 298.6
TUF-RS-SO-AI-5-2 274.6 213.6 260.0
TUF-RS-SO-A42-6-8 23.1 7.2 15.9
TUF-RS-SO-A37-0-2 9.1 3.2 1.3
TUF-RS-SO-AI4-0-2 7.6 1.6 5.0
TUF-RS-SO-A5A-2-4 1.7 1.7 0.4
TUF-RS-SO-NH24-2-4 1.7
TUF-RS-SO-A14-6-8 1.3 - 3.0
TUF-RS-SO-A7-4-6 ND ND ND
ND, compound not detected
Sample comparison on an as collected basis (i.e., soils were not
dried)
Lab GC-MS performed by Lockheed ESC, Las Vegas, NV
Field GC-MS performed by Tufts University
Sample collected by EPA's Region 1 oversight contractor
-, Samples were not analyzed
Table 5. Comparison of Field and Lab GC-MS by Chlorination
Level (ppm), Resolve Superfund site, North Dartmouth, MA.
ID Sample » TUF-RS-SO-A 1 5-2TUF-RS-SO-A42-6-8
Cl-level Field Lab Field Lab
TDGC-MS GC-MS TDGC-MS GC-MS
Cl-1 12.5 ND 0.5 ND
Cl-2 7.6 10.8 1.5 1.0
Cl-3 60.3 56.5 4.5 4 1
Cl-4 121.4 122.8 5.1 5.3
Cl-5 59.5 53.6 6.3 4.3
Cl-6 20.9 15.9 3.0 1 2
Cl-7 1.7 0.4 0.3 ND
Cl-8 0.7 ND 1.9 ND
total PCS 274.6 260.0 23.1 15.9
These values were not included in the dynamic range.
ND, compound not detected
Sample comparison on an as collected basis (i.e., soils were not
dried)
Lab GC-MS performed by Lockheed ESC, Las Vegas, NV
Field GC-MS (Quantitative Method) performed by Tufts
University
Sample collected by EPA's Region 1 oversight contractor
324
-------
Table 6. Comparison of Field and Lab GC-MS Results for PAH's
From the Hocomonco Pond Superfund Site in Westborough, MA,
in ppm.
Table 7. Comparison of Field and Lab GC-MS Results for PAH's
From the Hocomonco Pond Superfund Site in Westborough, MA,
in ppm.
DSTB22(0'-2') DSTB22(2'-4')
Lab Field1 Lab Field1
Naphthalene 0.1 ND 2.2 ND
Acenaphthylene 0.1 0.1 ND 0.7
Acenaphthene 1.4 0.1 6.0 0.2
Fluorene 2.9 1.5 16.3 3.0
Anthracene & 8.3 40.3 81.8 72,7
Phenanthrene
Pyrene& 11.8 10.6 112.2 60.5
Fluoranthene
Chrysene& 6.0 6.2 37.2 37.2
Benz(a)anthracene
Benz(b)fluoranthene, 3.2 23.8 17.7 22.3
Benz(k)fluoranthene, &
Benz(a)pyrene
ND, compound not detected
Sample comparison on an as collected basis (i.e., soils were not
dried)
Lab GC-MS performed by Lockheed ESC, Las Vegas, NV
Field GC-MS performed by Tufts University (Thermal Desorption
of Methylene Chloride Extract)
'Data collected by Selected Ion Monitoring (Internal Data System)
POND HP-SB5
Lab Field2 Lab Field1
Naphthalene 1.3 1.9 54.8 32.0
Acenaphthylene 1.4 ND ND ND
Acenaphthene 0.7 ND ND 1.2
Fluorene 2.5 ND ND 0.8
Anthracene & 16.7 10.4 ND ND
Phenanthrene
Pyrene& 30.7 43.6 ND ND
Fluoranthene
Chrysene& 37.2 55.2 ND ND
Benz(a)anthracene
ND, compound not detected
Sample comparison on an as collected basis (i.e., soils were not
dried)
Lab GC-MS performed by Lockheed ESC Las Vegas NV
Field GC-MS performed by Tufts University (Thermal Desorption
of Methylene Chloride Extract)
'Data collected by Selected Ion Monitoring (Internal Data System)
'Data collected as Total Ion Current Chromatogram and quantified
by Selected Ion Monitoring Extraction (External Data System)
DISCUSSION
ALAN CROCKETT: I found your presentation and the results extremely
informative and the accuracy or the precision you were getting was fantastic. Did
you say that you were using two tenths of a gram sample or a two-milligram
sample?
AL ROBBAT: A half a gram.
ALAN CROCKETT: That's impressive just being able to sub-sample a jar of
soil as repetitively as you've been able to. What's your preparation procedure for
homogenization of soil that comes into your facility?
ALROBBAT: These samples were all homogenized by EPA Region I. We didn't
do anything more after we got them, except stir them up a little bit.
ALAN CROCKETT: How did they homogenize when you get them so
homogeneous?
ALROBBAT: Basically they screen them and then they collected them in a large
jar and simply just rotated them. We did not do any of the real homogenization
of the sample.
ALAN CROCKETT: What's the cost of the instrumentation by the way?
AL ROBBAT: I think it's about $ 180,000 but your best bet is to ask Bruckner
Instruments.
JON GABRY: What are your power requirements for the unit?
ALROBBAT: We use six 24-volt batteries. Six 24 volt batteries out in the site.
We also can power-up at the site if there's electrical supply. So again, if you're
interested in those types of details. 1 would suggest you visit the Bruckner
Instruments booth.
325
-------
RAPID DETERMINATION OF SEMIVOLATILE POLLUTANTS BY
THERMAL EXTRACTION/GAS CHROMATOGRAPHY/MASS
SPECTROMETRY
T. Junk, V. Shirley, C. B. Henry, T. R. Irvin, E. B. Overton
LSU Institute for Environmental Studies
42 Atkinson Hall, Baton Rouge, LA 70803
J. E. Zumberge, C. Sutton, R. D. Worden
Ruska Laboratories, Inc., 3601 Dunvale, Houston, TX 77063
Abstract
There is considerable interest in rapid,
field deployable analytical systems!
Conventional gas chromatography/mass
spectrometry analytical techniques
provide sensitivity and specificity but
require cumbersome solvent
extractions. Thermal extraction offers a
fast and safe alternative to classical
extraction procedures for a wide range
of semivolatile pollutants. In this
technique samples are loaded into
porous quartz crucibles with no
preparation other than weighing
required prior to analysis. Analytes are
volatized into the helium carrier gas
flow at controlled preprogrammable
temperature profiles and subsequently
cyrocondensed onto a conventional gas
chromatographic column. The method
was demonstrated by analyzing for a
representative group of organic
pollutants covering a wide range of
polarity/volatility contained in natural
soil matrices at concentrations as low as
0.5 ppm using a Pyran Thermal
Chromatograph. Analyses were
independently performed by three
different laboratories (Institute for
Environmental Studies, Louisiana State
University; Engineering Toxicology,
Texas A & M University, Ruska
Laboratories, Inc.) using an on-line
Finnigan Ion Trap Detector for
identification and quantification.
Average correlation coefficients for
calibration curves ranged from 0.938 to
0.997 for compounds less volatile than
naphthalene. Naphthalene and more
volatile compounds experienced
variable losses during open-air sample
loading. Dialkylphthalates underwent
partial decomposition during the
thermal extraction process. Recoveries
varied depending on soil types as well
as on the physical and chemical nature
of analytes, with generally the highest
thermal extraction yields for river silt
and the lowest yields for clay. Typical
recoveries were 10 to 30% for
polynuclear aromatic hydrocarbons, 60
to 70% for hexachlorobenzene, and
nearly 100% for chloronaphthalenes.
However, the pesticide aldrin showed
recoveries of at most 19%. A majority
of the analytical results are within an
accepted range for quantitative analysis.
The Pyran system can be adapted to be
327
-------
deployable. With sample turn-around
times of typically 30-60 minutes this
instrument should greatly facilitate
remediation and hazardous waste
cleanup efforts.
Introduction
Transportation, field deployable
analytical systems that provide
unambiguous data on the amount of
semivolatile organic pollutants can aid
in the rapid assessment and cleanup of
hazardous waste sites. By
complementing the Environmental
Protection Agency's Control Laboratory
Program through interactive field
management, the efficient remediation
of hazardous wastes sites can be
accomplished (R. J. Bath, personal
communication).
Mass spectometry provides the
specificity and sensitivity necessary for
the identification and quantification of
most environmental pollutants.
However, to introduce analytes into the
mass spectrometer, the pollutants must
first be extracted from the soils.
Normally, organic solvents are used for
this purpose, a cumbersome and labor
intensive approach. Thermal extraction,
in contrast, desorbs analytes from their
matrices (soils) by controlled heating
under conditions which avoids analyte
decomposition (as opposed to pyrolysis).
In this report, we describe results from
a study aimed at verifying the
suitability of thermal extraction as
alternative to conventional extraction
for a representative cross section of
semivolatile organic pollutants. We
establish the factors controlling analyte
recoveries from different types of
matrices. Three laboratories
participated in this study, using
identical instrumentation (Institute for
Environmental Studies, Louisiana State
University, Texas A & M University,
College Station, and Ruska Laboratories,
Inc., Houston).
Instrumentation
A Level 2 Thermal Chromatograph
(Ruska Laboratories, Houston, Texas)
was interfaced with a Finnigan Ion Trap
Detector. Samples were heated in a
quartz chamber using a linear
temperature program and semivolatile
analytes purged with helium gas. These
analytes were cyrocondensed onto a
fused silica chromatographic column
(Hewlett Packard HP-5, 12 m x 0.2 mm)
cooled with liquid carbon dioxide,
separated, and identified by mass
spectroscopy. Thermal extraction
efficiencies for specific toxicants were
also monitored by thermal extraction
under identical conditions in an
identical quartz chamber coupled to a
flame ionization detector (Level 1
Thermal Extractor). Schematic diagrams
of these instruments are shown in Fig. 1.
Other experimental parameters were
chosen as follows: 30 ml/min He carrier
flow during thermal extraction phase,
30:1 split ration between thermal
extraction chamber and GC column, 1
ml/min carrier flow through GC column.
Standards Preparation
Test soils were prepared by adding
stock solutions of 20 semivolatile
organic pollutants covering a wide range
of polarity/volatility to three different
organic-lean natural soil matrices:
kaolin clay, sandy river silt, and
subsurface terrestrial soil from
Livingston Parish, Louisiana containing
30% clay, 66% silt, and 4% sand with a
total organic content of 0.11%. stock
solutions of the 20 standards (see Table
1) were prepared by weighing pure
328
-------
compound standards (primarily from
Aldrich Chemical Co.) and diluting 4000
Hg/ml stock standard (PP-HC8, Chem
Service, Inc.; lot #25-121B) with
dichloromethane to 20 ng/|il per
component.
The three soils were crushed using a
mortar and pestle and sieved through a
850 |im sieve. The sieved soils were
slurried for 1 hour with the appropriate
amount of stock standards (pure
dichloromethane for controls), the
solvent then removed at room
temperature by evaporation under a
fume hood to produce two sets of test
soils with concentrations of 50 ppm and
0.5 ppm, respectively, per analyte. The
soil standards were then sent to the
three participating laboratories for
independent analyses in well-filled
teflon lined screw cap vials and stored
at 6° to avoid analyte losses.
Methods
Soil samples were weighed into the
porous fused silica crucibles, while
standard stock solutions (20 ng/^il) were
injected onto the porous fused silica lids
of the sample crucibles using a 10 u.1
syringe just prior to loading into the
thermal extraction chamber. All
samples were heated from 30° to 260°
at 30°/min and held isothermally at
260° for 10 min before cooling to 30°.
The "trap" and "splitter" regions (see Fig.
1) were held isothermally at 300° and
310°, respectively; interface and
transfer line temperatures to the MS
were held between 280° and 290°. The
column was held at 5° until the thermal
extraction process was complete, the
temperature programmed to 285° at
10%nin and kept isothermal for 5 min.
Total cycle time was 59 min. The ion
trap detector was scanned from 47 to
440 amu at 1 scan/sec, peak threshold
was set at 2, and a mass defect of 100
mmu/100 amu was used. Full scan
mass spectra of eluting compounds
standards were verified using the NBS
mass spectra library. Areas and
retention times of characteristic ion
masses were recorded after each run for
each of the 20 compounds and internal
standards. Calibration curves for each
of the 20 compounds in the stock
solution (20 ng/jil) were obtained by
injecting 2, 5, 10, 15 and 20 ill onto the
crucible lids (corresponding to 40, 100,
200, 300, and 400 ng/component,
respectively). Ten u 1 (200
ng/component) of the deuterated
internal standards (Table 2) were also
added to the lid prior to each of the
above five runs. This experiment was
done in triplicate at Ruska Laboratories,
using a Finnigan Ion Trap Detector for
two runs as described above and a
Hewlett Packard Mass Selective Detector
(MSD) once for comparison. Just prior to
each run of the standard soils (10.0 to
13.8 mg for the 50 ppm standards and
approx. 100 mg for the 0.5 ppm
standards), 10 ul (200 ng/component)
of the deuterated internal standards
were injected into the soil/sediment.
Response factors (RF) and percent
relative standard deviations (%RSD)
were calculated for each compound
based on EPA's "Test Methods for
Evaluating Solid Waste, Physical,
Chemical Methods", SW-846, Third
Edition, Method 8270 (GC-MS for
semivolatile organics, capillary column
technique). RF values are based upon
the results of the on-lid injections of the
stock solutions.
Soil/sediment samples were also
analyzed using the Level 1- FID
instrument (see Fig. 1) to further
329
-------
eludicate the thermal extraction process
in an independent study at Louisiana
State University. This set of
experiments seeks to identify factors
influencing analyte recoveries by
systematically varying operator-
controllable variables including gas flow
rates, additives to facilitate extraction,
extraction temperature and duration; as
well as to define limiting factors for
target analytes and matrices. Three
analyte solutions were prepared: n-
triacontane ("C-30"), pyrene, and
hexachlorobenzene ("HCB"). These
compounds were chosen for their
thermal stabilities and chemical
inertness. Two are structurally similar,
all three are neutral and devoid of
reactive functionalites. Ten u.1 of stock
solutions in dichloromethane (10 mg/ml
for pyrene, HCB; 2 mg/ml for C-30)
were spiked onto the soils immediately
prior to analysis. The resulting FID
signals were integrated to calculate
analyte recoveries (Table 3), with the
FID signal of the pure analytes (no
matrix) as reference.
Conclusions
Level 1 Thermal Extraction/FID
Thermal extraction efficiencies vary
considerably with the nature of analytes
as well as matrices (Table 3). While
conventional solvent extraction
procedures would be expected to
produce similarly high recoveries for n-
triacontane, pyrene, and
hexachlorobenzene, thermal extraction
produced markedly different results for
clay as matrix (Fig. 2a). HCB recoveries
were quantitative, while C-30 and
pyrene recoveries ranged at approx.
30%. Variation of the matrix had a less
pronounced effect on the recovery of
pyrene. These results cannot be
explained solely in terms of polarity or
volatility. Not surprisingly, percent
deviations of recovery decrease
dramatically in the presence of a soil
matrix (Fig. 2b). The increase of helium
flow during the thermal extraction
process from 40 to 100 ml/min did not
increase the extraction yields of C-30
significantly (see Table 3); however,
addition of polar additives to the soil
samples immediately before thermal
extraction, such as water or phosphoric
acid, improved the recovery of pyrene
from clay markedly (Fig. 3c). Figure 2d
illustrates blockage of reactive sites of
the soil matrices by repeated spiking of
the same river silt sample. Thermal
extraction efficiences increased from 25
to 65%. Simple physical obstruction of
the carrer gas flow is certainly one of
the factors contributing to reduced
recoveries. The soil samples "cake" and
block the desorption of analytes into the
carrer gas flow. Thus, recoveries sank
to 69% for pyrene and to 82% for C-30
when standards were spiked onto the
lids of crucibles filled with 100 mg clay
without direct contact between analyte
and matrix (Table 2). Repeated thermal
extraction, the increase of extraction
temperatures above 450<>, or an
extension of extraction times were not
promising, as illustrated by Fig. 4a-c.
These figures compare the thermal
desorption of identical amounts of
pyrene (100 ng) from a porous quartz
crucible (Fig. 4a) and spiked into a
kaolinite clay sample (Fig. Figure 4b)
using the temperature profile shown in
Fig. 4c under otherwise identical
conditions. Not only is the thermal
desorption of the standard from the
spiked clay considerable below 100%,
but it is also shifted to higher
temperatures. At 450°, no further
analyte was released upon prolonged
heating. The fate of the unextracted
330
-------
analytes is currently unknown and
subject to future investigations.
Level 2 Thermal Extraction/GC/MS
The results of analyses from all three
laboratories are summarized in Table 1.
The 20 organic compounds and
corresponding characteristic ion masses
are listed along with linear correlation
coefficients (r) derived from the five
point calibration curves of the on-lid
stock solution injections. Fig. 3 shows
examples of four calibration curves
from one laboratory; the more volatile
components (e.g. naphthalene)
experience variable rates of evaporation
after injection of the standard stock
solution onto the porous quartz crucible
lids prior to sample insertion into the
pyrocell (approx. 2 min from injection
onto the lid until sample loading).
Dioctyl phthalate signals were relatively
low except at high concentration levels
(300-400 ng); after it appears that
much of this compound degraded to
phthalic anhydride (which was always
detected) during the on-lid calibration
runs. Diethyl phthalate, in comparison,
showed good linearity and less
degradation. Pentachlorophenol
linearity was not as good as that of
other compounds in the same volatility
range. All other compounds showed
good linearity.
Also listed in Table 1 are the percent
relative standard deviations (%RSD) of
calculated response factors based on the
on-lid injections of 20 ng/u,l mix of 20
compounds plus the deuterated internal
standards listed in Table 2. Since %RSD
values are also a measure of the
precision for each compound, it is not
surprising that most volatile compounds
also show the highest deviations.
Although there is some variation
between the participating laboratories,
specific compounds tend to yield high
%RSD values while others showed
consistently good precision. The same
holds for deuterated standards. Again,
the more volatile naphthalene-d8 and
dichlorobenzene-d4 showed the most
variation, phenanthrene-dlO and
chrysene-d!2 the least.
From the obtained data set, recoveries
could be calculated either by the
external standard method using the
least square fits of the five point
calibration curves for all compounds or,
alternatively, by internal standard
quantitation based on the response
factors calculated for each compounds.
Table 1 lists results for both methods,
which do not reflect the expected
improved accuracy for the internal
standard method. Due to the
considerably different chemical and
physical environments the standards
experience while being partially
adsorbed by the soil samples and
partially by the porous crucibles, no
high degree of accuracy can be expected
by the internal standard method. The
implicit assumption made in
conventional chromatography, namely
that standards and analytes are
subjected to identical environments,
cannot easily be realized in thermal
extraction.
Percent recovery appears to be
dependent on a number of factors
including polarity, molecular weight,
and interactions with constituents of the
soil matrix, both organic and inorganic.
Not surprisingly, recovery was
significantly greater for many
compounds from the river silt than from
the clay or subsurface soils (e.g.,
phenanthrene: 11% from clay and 31%
from silt) while chloronaphthalene was
close to 100% for both clay and silt. The
331
-------
recovery of diphenylamine was equally
low (approx. 5%) for clay and silt. Since
the subsurface soil contains about 30%
clay, percent recovery is generally in
between those for clay and silt. It is
interesting to compare recoveries for
the structurally similar tricyclic
compounds dibenzothiophene, fluorene,
and carbazole. In all three soil types
the order of recovery efficiency was
dibenzothiophene>fluorene>carbazole,
which likely reflects increasing binding
to the soil matrix. At the 0.5 ppm
concentration levels, naphthalene,
chloronaph thalene, fluorene,
hexachlorobenzene, dibenzothiophene,
phenanthrene, aldrin, and pyrene were
all detected in the soil standards in at
least two of the three laboratories.
It is apparent from these results that
small aliquots of soils can be analyzed
by thermal extraction/GC/MS without
any prior sample preparation. While
the method is generally suited for
situations requiring high precision or
low detection limits, it performs well a
analyte concentrations>50 ppm, is
amenable to full automation and will
serve for rapid screening of soils
contaminated with thermally stable
organic semivolatiles, a class of
compounds that includes PNA's PCB's,
most petroleum products and pesticides
and is commonly encountered in
hazardous waste cleanup efforts.
References
Bath, R.F., personal communication
(1989).
Environmental Protection Agency, "Test
Methods for Evaluating Solid Waste,
Physical, Chemical Methods", SW-846,
Third Edition, Method 8270 (GC/MS for
semi-volatile organics: capillary column
technique).
Henry, C.B., Overton, E.B., and Sutton, C.,
"Applications of the Pyran Thermal
Extraction-GC/MS for the Rapid
Characterization and Monitoring of
Hazardous Waste Sites"; Proceedings of
the First International Symposium for
Hazardous Waste Site Investigations,
399-405 (1988).
Overton, E.B., Henry, C.B., and Martin,
S.J., "A Field Deployable Instrument for
the Analysis of Semi-volatile
Compounds in Hazardous Waste";
Pittsburgh Conference and Exposition on
Analytical Chemistry and Applied
Spectroscopy, New Orleans, LA.,
Abstract (1988).
Zumberge, J.E., Sutton, C., Martin, S.J.,
and Worden, R.D., "Determining Oil
General Kinetic Parameters by Using a
Fused Quartz Pyrolysis System"; Energy
and Fuels, 2, 264-266 (1988).
Junk, T., Irvin, T.R., Donnelly, K.C., and
Marek, D., "Quantification of Pesticides
on Soils by Thermal Extraction-GC/MS",
in preparation.
Acknowledgements
We thank Drs. R.J. Bath and D. Flory for
helpful comments and suggestions.
332
-------
u
U
AIR
Sample Crucible
LC02
LEVEL I-FID ANALYZER
Figure 1
sScale
10cm
MS
COI.UUN LXII
LLTT OR RIGHT
SIDE
PYROCEU
-------
MATRIX vs RECOVERY
Flow 40 ml/min
% Racovery
N°n« Clay
•I Pyrene E3 C-30
Subsoil Sill
Hexachlorobenzene
Fig 2a
ADDITIVES DURING EXTRACTION
Pyrene on Clay
so/
Nona
Phosphoric Acid
DEVIATION OF RECOVERY
Comparison of Different Matreces
% Sid Dev.
Fig 20
Nons Clay Subsoil Silt
^ Pyrene IZ2 C-30 [^ Hexachlorobenzene
REPEATED SPIKING OF SOILS
Pyrene on Sand
Extraction #
FJg 2d
-------
NAPHTHALENE
HEXACHLOROBENZENE
Area(m/e 128)(Thousands)
Aroa(nVe 284)(Thousands)
Fig 31
Fig. 3D
PHENANTHRENE
Area(nVe 178)fT"housands)
100 200 300 400 500
BENZO(A)PYRENE
100 200
400 500
Fig. 3c
Fig 3D
-------
'ABLE i. Precision ol semi -volatile standard calibration curves and percent recovory
COfvPCUvD
2-chlorophenol
4-methylphenol
2.4-dichlorophenol
naphthalene
4-chloro-3-melhylphenol
l-chloronaphlhalene
2.s-dinitrotoluene
lluorono
cJlo ihylphlhalu lo
diphcnylamine
hexachlorobenzene
dibenzothiophc-ne
peniachlorophenol
phenanlhrene
carbazole
nldrin
pyrcno
bis[2-ethylhexyl)ph!halati
benzo(a)pyrene
Based on 25 on-lid injeclio
2 Response (actors (RF) and
(area cua/i mass of dcul. st
3 Percent theoretical recovery
50 ppm per component)
•• Rased on deutcratcd intcrna
5 Average linear correlation
These compounds were also
QUAN
MASS (m/z)
128
107
152
128
107
162
165
166
1 45
169
284
184
266
178
167
66
202
149
252
252
TS; 4 different
sercent relative
. (ng compoun
based upon I
3 different in
standards sp
II
coefficient lor
II
detected from
II
I
(vCAN
r5
0.9152
0.9481
0.8962
0 7484
0.9708
0.9385
0.9848
0.0007
0.07 1 "/
0.9801
0.9901
0.9946
0.9537
0.9943
0.9932
0.9924
0 9971
0.9922
0.9931 |
inslrumen
standard c
d)
nearily of
Irumenls a
kcd diroc[l
the on-lid
!—
the 0.5 pp
MEAN1
%RSO
110
85
67
66
29
23
26
1 -1
34
20
10
44
8
10
1 1
7
38
37
s: 3 differ
eviation ("/
%RSD . 1
id injeclio
t 3 diflere
Y on soil c
live poin
H-
m slandar
%RSC
RL
147 73
96 76
78 50 78
88 35 59
27 26 27
29 6 22
18 41 16
10 9 24
14 GG 14
21 32 14
674
734
51 56 28
835
11 115
845
730
26 19 39
38 10 36
>nt laboratories.
»RSD) were calc
00 (SD/RF)
ns (slope and y
nl laboratories
nd corrcspondin
calibration fror
ds in at least 2
OF F
TAW
;>4
R4
68
43
1 9
??
?4
0
i a
8
22
1 5
28
1 9
1 3
29
1 4
53
38
38
40
ulaied
interc
n RF
~T
n lhr£
nf the
I
IF 2
LSU
64
103
46
35
31
22
GO
25
9
5
56
7
1 1
8
5
70
6'
o 400
I
based
cpl) ant
values-
I
e diffo
i
3 labor
n
Cl
RL TAM LSI
15 46
8 10 V
9 26 33
60 100 72
7119
69 77 141
10 6 4
4159
835
71 76 57
12 28 19
~8 13 12
434
234
4 12 5
10 1 1
9 1
:g per componenl
on EPA method 8
not on deuterale
only PNAr. nro cc
ent laboratories.
atones.
.AY
\ AVE
30
1 0
23
' 771
9
96'
7
9'
5
68'
20'
11 '
4
3
7
4
4
per in
>70 ol
d inter
mparo
% RECOVERY 2
50 PPM STANDARDS
SILT
RL TAM LSU AVE
19 25 - 22
8 6 6 77
23 23 15 20
93 128 138 120'
14 12 9 12
83 118 109 103'
11 12 7 10
20 31 17 25'
04 5 '6
71 67 54 64'
40 39 26 • 35'
36 33 23 31'
10 7 6 8
23 20 15 19'
20 24 13 19'
4779
14 5 6 8
eclion.
SW-846; i.e.: RF: (area qu<
nal standards spiked inlo
3 duo to tho chemical sim
SUBSOIL
RL TAM LSU AVE
2 o
9847
25 44 9 26
27 41 121 63'
9 15 4 9
>1 78 - 70'
10 16 4 10
12 27 10 19'
87 46
58 78 103 SO'
•2 53 66 50'
8 47 46 40
8 12 7 9
1 21 13 15
1 32 17 20
018 4 11
0 20 2 11
in mass compound) (nrj
soil. Average oJ 9 re
aritics with the slnndn
% RECOVERY 4
CLAY SILT SOIL
33 93 67
3.1 55 45
50 78 85
35 65 52
3 15 10
2 149
deut. internal standard)/
ilicates (about lOmg soi
ds.
-------
TABLE 2
Variations of Internal Standard Areas
Deuterated
Internal Standards
Quan. Mass
Mean %S Ruska Ruska Ruska TAMU
LSU
.1 ,4-dichlorobenzene-d4
naphthalene-d8
acenaphthene-dlO
phenanthrene-dlO
chryscnc-d!2
perylene-dl2
152
136
164
188
240
264
106
57
17
8
7
34
75
41
7
6
6
43
155
74
22
9
11
38
139
70
13
11
11
26
73
43
10
6
3
19
89
57
32
8
4
45
* Values are based on 25 on-lid injections recorded at 3 different laboratories; 200 ng/component,
injection.
-------
TABLE 3
Evaluation of Thermal Extraction on Pyran Level 1
Exp# Analyte Amt(ug) Matrix He ml/minNo.of Obs. Avg %Dev
1
2
3
4
5
6
7
8
9
10
pyrene
pyrene
pyrene
pyrene
C-30
C-30
C-30
C-30
HCB
HCB
100
100
100
100
20
20
20
20
100
100
none
clay
subsoil
r.silt
none
clay
none
clay
none
clay
40
40
40
40
40
40
100
100
40
40
3
3
3
3
3
3
3
3
3
3
2837.3
717.0
716.3
1063.0
1101.0
341.0
1241.3
396.7
613.7
667.7
3.7
1.8
18.3
29.7
1.4
18.5
2.0
2.4
8.4
2.1
%Rec
100.0
25.3
25.2
37.5
100.0
31.0
100.0
32.0
100.0
108.8
Addition of phosphoric acid, 85%, 0.1 ml
11 pyrene
Addition of water, 0.2 ml
100
r.silt
40
1246.5
18.1
43.9
12 pyrene 100 clay
Standards on lid of crucible filled with soil
40
868.0
30.6
13
14
pyrene
C-30
100
20
clay
clay
40
100
1984.0
1018.7
4.7
69.9
82.1
Repeated spiking of previously extracted soil
15 pyrene
16 pyrene
17 pyrene
C-30: n-triacontane
HCB: hexachlorobenzene
100
100
100
r.silt
r.silt
r.silt
40
40
40
Extr.#
1
1
1
DISCUSSION
768.0
1330.0
1860.0
27.1
46.9
65.6
AL ROBBAT: Have you tried looking at organic extracts? Can you place an
organic extract in the soil and look at the thermal desorption properties? In other
words, lake the soil, extract it with melhylene chloride, taken out of part of the
extraction, and run your experiment?
THOMAS JUNK: In other words, you introduce an organic extract that has been
extracted in a conventional procedure to see how that behaves in the instrument
itself? I'm not quite sure that I understand your question.
AL ROBBAT: We found the same thing. For example, for PAHs, if you take the
thermal desorption sample probe and place it directly over the soil, you get
between 7% and 15% extraction recoveries. What I'm suggesting is that if you
use the simple 2 mL extraction procedure that I described, take a half a gram of
soil, add 2 mLs of methylene chloride, extract it, add that extract to your cell, can
you perform that experiment? Can you use say, 1Ou]s or 20 uJs of extract in your
cell. Have you tried that experiment?
THOMAS JUNK: Yes, you can. You can use a conventional extract and absorb
it onto the porous quartz crucible. In order words you could go through the
addition of a small amount of solvent into the soils that would then by and large
produce a similar effect as the one you just mentioned.
AL ROBBAT: Right.
THOMAS JUNK: Yes, we've tried that. And in some cases it produces
satisfactory results. However we have not consistently found an improvement
over classic extraction procedures. I think I can cover that together with the
addition of various cold solvents that I just mentioned, such as with phosphoric
acid or water. And yes, we do see an increase, bu the increase for phosphoric acid
for example, was much more significant than that for other cold solvents.
STEPHEN BILLETS: I want to say from the standpoint of testing this
technology that Ruska Thermal Extraction System is in the EMSL-Las Vegas
Laboratory currently undergoing evaluation for possible use in a field
demonstration study. So we are conducting, in our laboratory, a complementary
effort to what Thomas described.
338
-------
THE APPLICATION OF A MOBILE ION TRAP MASS SPECTROMETER SYSTEM
TO ENVIRONMENTAL SCREENING AND MONITORING
William H. McClennen, Neil S. Arnold, Henk L.C. Meuzelaar,
Erich Ludwig* and JoAnn S. Lighty**
Center for Micro Analysis &
Reaction Chemistry, University of
Utah, Salt Lake City, UT 84112
*GSF Munchen Institut fur
Okologische Chemie, Ingolstadter
Landstrasse 1, 8042 Neuherberg,
Germany
**Chemical Engineering
Department, University of Utah,
Salt Lake City, UT 84112
ABSTRACT
This paper presents examples of the use of a mobile Ion
Trap Mass Spectrometer (ITMS, Finnigan MAT) for on-
site environmental screening and monitoring of vapors by
gas chromatography/mass spectrometry (GC/MS). The
instrument is built around a miniaturized ITMS system,
with a novel direct vapor sampling inlet and coupled to a
high speed transfer line GC column (short capillary
column with fixed pressure drop). The column is
temperature controlled inside the standard ion trap
transfer line housing. This provides for high speed
analyses at 10-60 s intervals using an automated sampling
system constructed with only inert materials in the
sample path.
Specific laboratory and field applications exemplify key
characteristics of the system including sensitivity,
specificity for a broad range of compounds, ruggedness
for field testing in harsh environments, and general speed
and versatility of the analytical technique. The system
has been calibrated for alkylbenzenes at concentrations as
low as 4 ppb in air and used to monitor these compounds
in an office space. Both the MINITMASS and a simpler
Ion Trap Detector (ITD) based system have been used to
monitor organic vapors from acetone through 5 ring
polycyclic aromatic hydrocarbons produced in laboratory
scale reactors for studying the thermal desorption and
incineration of hazardous wastes. The ruggedness of the
MINITMASS system has been demonstrated by vapor
sampling in the Utah summer desert and at a 600 MW
coal fired power plant. Finally, the analysis speed and
versatility are described for vapor monitoring of volatile
organic compounds at an EPA national priority list waste
site.
INTRODUCTION
Preliminary data obtained with a miniaturized Ion Trap
Mass Spectrometer (MINITMASS) system developed in
close collaboration with the manufacturer (Finnigan MAT
Corp.) were presented at the first International
Symposium on Field Screening Methods for Hazardous
Waste Site Investigations (1). The MINITMASS system
was shown to be capable of performing tandem MS
(MS") analyses in electron ionization (El) as well as
chemical ionization (CI) mode and featured a special air
sampling inlet in combination with so-called "transfer line
gas chromatography" capability (2,3). Due to its
relatively low weight (approx. 280 Ibs.), the
MINITMASS system was readily transported inside a
small mobile laboratory mounted on a regular 3/4 ton
pick-up truck (1). Some of the main shortcomings of the
MINITMASS system included: insufficient sensitivity
(high ppb/low ppm range), limited mobility for many
field screening applications (due to a shock sensitive
turbomolecular pump), untested performance with low
volatile (e.g., PAH type) compounds and lack of field test
data at actual hazardous waste sites.
Since October 1988 the MINITMASS system has been
tested under a variety of conditions at several outdoor as
well as indoor locations. Moreover, several hardware and
software improvements have markedly increased its
sensitivity (currently ~1 ppb for alkylbenzenes (3)) and
applicability to low volatile compounds (e.g., 3-5 ring
PAH's (4)) while enabling true mobility through the
installation of a more rugged vacuum pump. In addition,
a simplified mobile ITD (Ion Trap Detector) system was
constructed and tested for dedicated hazardous waste
combustion applications which do not require MS"
339
-------
capabilities (5,6). Hazardous waste related monitoring
applications of both systems will be described in the
following paragraphs. Some of these results have been
presented elsewhere (2-6) but are included with various
new data for completeness in this overview of the
instrument's current performance.
EXPERIMENTAL
General Parameters
The direct atmospheric vapor sampling inlet described in
detail elsewhere (2,5,6), consists of three concentric tubes
with appropriate flow control plumbing and electronics.
The inlet system is made from deactivated fused silica,
quartz and glass, or glass-lined metal tubing. The sample
path contains no moving parts. When sampling, the gas
is exposed to the column inlet for a controlled period of
time (0.3 to 2 s) while 30 to 200 ^L of sample is
admitted to the column. Helium carrier gas flow is then
restored for the rest of the sampling cycle and GC
separation of the sample takes place.
This inlet is coupled to a 1 m long fused silica capillary
column which enables nominal GC separation of
components and provides a pressure drop between the
near-ambient sampling environment and the high vacuum
of the mass spectrometer ion source. With the fixed
pressure drop, the chromatographic conditions are
controlled primarily by column length, radius and
temperature (2). The fused silica capillary column used
in this work was either a 0.15 mm ID x 1.2 urn methyl
silicone film (CP-Sil-5 CB, Chrompak) with 1.5-2 ml/min
He How or a 0.18 mm ID x 0.4 urn film DB-5 (J&W
Scientific) with ca. 4 ml/min He flow.
This transfer line GC inlet system was used with both a
regular Finnigan MAT ITD and the MINITMASS system
with axial modulation and tandem MS capabilities (1),
although this paper describes results from use in only the
single MS mode. In addition to these special capabilities,
the MINITMASS system permitted higher flow rates by
virtue of the axial modulation feature and was thus the
only system which could use the 0.18 mm ID column.
The combination of increased MS resolution and
increased flow rates resulted in higher sensitivity.
Field Testing
In field tests sponsored by the EPA, vapor standards were
diluted into a 5 m long x 2.5 cm ID glass and Teflon®
manifold with a 2.2 1 s'1 total flow. These standards
included both 50 ppm compressed gas mixtures and
equilibrium headspace vapors of pure compounds injected
into the manifold with a motor driven syringe pump. The
50 ppm gas standards were diluted to calibration mixtures
of 20 to 350 parts per billion (ppb) in air, while the
syringe pump produced mixtures ranging from 16 ppb to
over 10 ppm, depending on the compound vapor pressure,
syringe diameter and motor speed.
Vapor standard calibration data were obtained scanning
from m/z 45 or 50 to 200 at 4 scans s"1. The inlet and
transfer line were maintained at 25°C, while the ion trap
was maintained at 85°C. The temperature of the mixing
chamber was ambient and not controlled. The vapor inlet
drew up to 120 ml min"1 from the EPA vapor manifold.
A .5 s vapor sample pulse was frequently used, although
the EPA experiments involved varying the sample time
from 330 ms to 2.5 s with a routine pulse width of 715
ms.
Combustion Monitoring
Exhaust from a rotary kiln simulator was monitored for
gas phase hydrocarbons during the combustion of
polymeric medical supplies (6). The 11 g batch samples
were loaded into the kiln and incinerated at 600°C.
Rapid on-line analyses were obtained using the
unmodified ITD system. A sample flow of 25 to 50 ml
min"1 was drawn from the kiln exhaust gases in a
transition area preceding the afterburner. Vapor samples
were taken at 10 s intervals to monitor concentration
transients during sample combustion. The .15 mm ID
column was used in this system at a constant temperature
of 30°C or 82°C with the vapor inlet at 60 or 100°C, and
the mass spectrometer was scanned from m/z 35 to 120
or 50 to 148 at 4 scans s"1. Exhaust from a thermal bed
reactor for hazardous waste studies (7) was monitored
using the MINITMASS during thermal dcsorption of
polycyclic aromatic hydrocarbons (PAHs) from contam-
inated soils obtained at former coal gas plant sites (4).
The soils were heated to 400°C under a radiant heater
with a preheated nitrogen flow of .5 1 min'1 above the bed
of soil. The exhaust gas was sampled at 60 s intervals
with 2 s vapor sampling pulses. The inlet was operated
between 150 and 175°C and the transfer line was
maintained at 125°C. The separation was performed in
the .18 mm ID column specified above. The ion trap
manifold was 200°C and the MS was scanned from m/z
100-300 at 4 scans s-1.
RESULTS AND DISCUSSION
Sensitivity and Dynamic Range
The basic objective of vapor sampling short column gas
chromatography is to provide sufficient separation of the
340
-------
organic compounds of interest from the major
atmospheric constituents to allow optimum use of the
sensitivity, specificity and speed of the detector. The
sensitivity of the MINITMASS system utilizing this
principle is demonstrated by the analysis of toluene vapor
in air in Figure 1. This figure shows ion chromatograms
for six repetitive samples of toluene in air at a
concentration of 16 parts per billion (ppb, volume or
molar ratio). The sampling points are indicated in the
total ion chromatogram (TIC) of Figure la by the
baseline depressions at 25 second intervals. The short
pulses of air, ca. 70 jil in 0.7 s, suppress the baseline by
overloading the ion trap with air ions so that even back-
ground ions in the MS are not detected. However, the
well resolved toluene peaks elute from the short column
20 s later with excellent sensitivity and signal to noise
(s/n) as indicated by the selected ion chromatogram of
summed ion peaks at m/z 91 and 92 shown in Figure Ib.
In addition to this 16 ppb data, Table 1 presents a set of
calibration points for toluene showing the degree of
sample repeatability expressed as relative standard
deviation. For statistical reasons, a minimum of 5
consecutive vapor samples were taken at each
concentration. These concentrations were prepared via
the syringe pump method described above, and response
was measured via the peak area of the m/z 91 trace from
a .715 s vapor sample. A linear fit to the full set of data
points was obtained with the correlation coefficient R =
.998, indicating a linear dynamic range of 3 or more
orders of magnitude.
A practical application of alkylbcnzcne analysis is shown
in Figure 2 using the thin film column (0.4 urn rather
than 1.2 ^im used in Figure 1). Figure 2 presents two ion
chromatogram traces from indoor atmospheric sampling
in one of our office work areas. The m/z 91 trace clearly
shows peaks for toluene, ethylbenzene, m- and p-xylene
and o-xylenc. The estimated toluene concentration is 70
ppb, presumably derived from the glue used on the
recently installed ceiling tiles. The MINITMASS system,
with its axial modulation capability and the higher flow
of the .18 mm ID column has shown alkylbenzene
detection limits near 1 ppb with a s/n greater than 2. The
normal ITD-based system, limited to the 1.5 ml/min of
the .15 mm ID column, has shown detection limits of
approximately 20 ppb.
Speed and Selectivity
An example of the GC/MS vapor analysis of a 7
component standard gas mixture is shown in the partial
chromatograms of Figure 3. Figure 3a shows the total
ion chromatogram (TIC) while the concurrent selected ion
chromatogram profiles show the major ions from four of
the test compounds. Arrows in the TIC profile indicate
the beginning points for 3 subsequent 715 ms samples
with a 30 second sampling interval. Although the 1,1,1-
trichlorocthane (111TCA) and benzene peaks are not
completely resolved in the TIC, they are readily
quantitatcd based on the selected ions from their unique
mass spectra. Note that the small vinyl chloride peaks at
scan numbers 5, 125 and 245 are partially cut off by the
large air pulse baseline disturbances from which they are
incompletely separated. However, even this early eluting
compound had a reproducible, linear response curve over
the range of 20 to 350 ppb which we tested. In other
words, the limited resolution of the short GC column is
sufficient to greatly enhance the specificity and selectivity
of the mass spectrometer.
A major asset of the speed of short column GC/MS is the
ability to do on-line monitoring in a nearly real-time
mode. Figure 4 shows a set of chromatograms
monitoring the evolution of volatile organics from the
combustion of polypropylene materials in a rotary kiln
simulator (6). For these experiments the ITD system
with the .15 mm ID, thick film (1.2 ^im) column sampled
gases just prior to the afterburner. A 0.5 s sampling
every 10s with the column at 82 C was sufficient to
follow the transient concentrations of aromatics during
the 2 min experiment. Selected ion traces at m/z 78 and
91 show the specific benzene and toluene peaks in the
repetitive analyses. Figure 4b explicitly plots the
quantitatcd evolution curves obtained from peak areas of
selected ions for benzene, toluene, phenol and styrcne.
With Ihc column at 30°C, compounds as small as acetone
were separated from air with benzene eluting at 7s. The
peak concentrations of these hydrocarbons occurred
before and after the point at which the melting plastic
was totally engulfed in flame in the 600 C rotary kiln
simulator.
Figure 5 illustrates the high boiling range of compounds
which can benefit from the speed and selectivity of short
column GC/MS in an analysis of polycyclic aromatic
hydrocarbon (PAH) vapors during thermal treatment of a
contaminated soil (4). By elevating the temperature of
the thin film column to 125 C, compounds with boiling
points ranging from 218 to 340 C were readily analyzed
with a 60 s sampling internal. The numbered peaks in
the TIC represent 1) naphthalene, 2) methylnaphthalcnes,
3) C2 naphthalenes, 4) fluorene, 5) phenanlhrene and 6)
anthracene. Although the GC resolution was insufficient
to separate some of these isomers, the selected ion
chromatograms show the obvious benefits. In Figure 4b
the m/z 154 trace shows a prominent peak for biphenyl
with additional unresolved humps for fragments from
341
-------
larger compounds. The m/z 168 trace indicates that the
short column is able to resolve the methylbiphenyls (at
12 s) from the dominant dibenzofuran peak (at 14 s).
The m/z 184 trace shows separate peaks for the C4
naphthalenes at 22 s, dibenzothiophene at 37 s, and ions
associated with the intense phenanthrene (m/z 178) peak.
Instrument Ruggedness
Although the commercial versions of both the ITD and
ITMS instruments were designed principally for
stationary laboratory operation, the modified ITMS,
which was built around a normal ITD chassis, has held
up well under the rigors of harsh transportation and
environmental operating conditions. Two examples of
applications in particularly harsh environments include
the Utah summer desert and flue gas sampling from a
coal fired 600 MW power plant.
The desert testing was the maiden use of the system to
discover potential field problems. It involved operation
of the instrument in the mobile lab on the back of a 3/4
ton pickup truck lo sample various chemical vapors
released from a permanent dissemination line (1). The
instrument was severely bumped and jostled as we
maneuvered the truck to sample in the shifting winds on
the brush covered terrain. Operation was complicated at
the time by the necessity of venting and then restarting
the instrument before and after each move to avoid a
turbomolccular pump crash. Typical down time was
approximately one hour between data acquisitions,
including cool down and warm up. Despite the rigorous
handling, the instrument's only failure during the 2 weeks
of testing was an overrated power transistor which had
previously failed in normal lab operation on a different
ITMS. The turbomolecular pump has recently been
replaced with a more rugged model to allow true
"mobile" use in addition to "transportable" operation but
has not yet been rigorously tested in the "mobile" mode.
The flue gas sampling involved rolling the MINITMASS
instrument to a seventh floor site at a Utah Power and
Light Company 600 MW coal fired power plant. The
objective was to do on-line analysis for aromatic
hydrocarbons in the 350 C flue gas as it exited the main
boiler sections. No organics were observed in the 1.5%
excess oxygen combustion products although our on-site
detection limits were 4 and 10 ppb for alkyl benzenes and
alkylnaphthalenes respectively. However, simply
operating the instrument was a major accomplishment in
this harsh environment. Fly ash was continuously raining
in the ambient atmosphere from the overhead structures;
the whole work site was constantly rocking and rumbling;
and the ambient temperature ranged from ca. 4 to 35 C
(40 to 95 F). A makeshift plastic tent with a crude
window fan and filter was assembled over the instrument
to supply some measure of environmental control (in
addition to the chassis mounted fans and filters).
A recent example of the instrument's more "routine"
transportability was demonstrated in EPA testing in New
Jersey. The MINITMASS system was successfully
driven across the country in its mobile lab and brought
into operation with verified performance capability within
6 hours after arrival at the national priority list (NPL)
landfill site. This start up time included transferring the
instrument from the mobile lab to the EPA site trailer and
diagnosis and repair of a broken thermocouple.
Operation in the mobile lab itself had previously been
verified in less than 1 hour.
Versatility of Inlet Sampling
One of the unique features of our vapor sampling inlet is
the ability to readily vary the sample volume injected
onto the column by a simple change in the sample pulse
width. Figure 6 shows calibration data for toluene from a
gas standard sampled with a 495 ms pulse compared to
the typical 715 ms sampling. The linear regression lines
along with 95% confidence intervals are shown for each
of the two data sets. The shorter sampling pulse has very
similar repeatability with the smaller response slope
corresponding directly to the smaller sample and
reduction in pulse time within 4%. Figure 7 also
demonstrates the linearity of mean peak area response
versus sample pulse width for two different scries of gas
mixtures. The lower concentration data are from the
analysis of an "unknown" gas mixture (ca. 30 ppb of the
gas standard shown in Figure 3) which was run with 2.5
s sampling to maximize sensitivity for the later eluting
compounds. The high concentration data was a test of
increasing GC resolution for the overlapping peaks
dichloromethane (DCM), 1,2-dichloroethane (12DCE) and
1,1-dichloroelhane (11DCA) in a gas mixture standard by
decreasing the pulse width to 330 ms. The combined
data sets demonstrate excellent linearity over an eightfold
change in sample size.
Figure 8 illustrates the effect of sampling duration on
resolution and sensitivity. The same mixture of DCM,
12DCE, 11DCA and tetrachloromethane at 350 ppb each
is used in 8a with a sampling time of 330 ms and in 8b
with a time of 715 ms. Clearly the 330 ms time
improves the resolution of the early eluting compounds,
but comparison of the m/z 82 selected ion trace indicates
that the tetrachloromethane which elutes at 7 s has only
70% of the peak height obtained at 715 ms. For
detection of compounds at the lowest levels, peak height
342
-------
becomes the limiting factor as the ion counts must exceed
the noise threshold.
Another aspect of the vapor sampling performance of our
inlet is the ability to readily sample from atmospheres at
different ambient pressure and limited small total volume.
Most of the EPA testing involved sampling standard gas
mixtures from a high flow manifold operating at reduced
pressure conditions similar to those which might exist in
a system drawing samples from many separate remote
points on a fence line. However, at the end of this series
of experiments, several Tedlar bag samples were also
analyzed. In order to maximize the number of separate
samples from a single 1 liter Tedlar bag, the total sample
drawn into our inlet was reduced from 120 ml/mm to 25
ml/min. One of the gas mixture standards was then
diluted to 2.5 ppm in a Tedlar bag and our duplicate
analysis of it showed responses within 4% of perfect
linearity for the manifold calibration lines.
Figure 9 shows the total ion chromatograms for the
analysis of landfill wellhead vapors from the NPL site
which were sampled into Tedlar bags and diluted by 1/5
with clean air. The identified compounds from these two
analyses are listed in Table 2 with detected concentrations
for those previously calibrated.
CONCLUSIONS
The results presented here have demonstrated
MINITMASS performance in several key areas of
capability which might be expected from an on-site,
vapor sampling, short column GC/MS system based on
an ion trap mass spectrometer. These include
specifically: detection limits of less than 10 ppb for a
variety of volatile organic compounds; selective analysis
of 21 compounds or more in a single one minute
chromatogram with boiling point windows depending on
column type and temperature; repetitive sampling as
frequent as each 10 s for monitoring transient vapor
concentrations; and direct variation of sample size with
sample pulse time to readily optimize GC resolution
versus ultimate sensitivity. The examples of operation in
harsh environments and at remote sites further suggest
that the instrument is rugged enough for most field
screening and hazardous waste site investigations. These
specific capabilities also apply to a similarly equipped
standard ITD system except for the order of magnitude
sensitivity difference with the ITMS enhancement.
Although the MINITMASS has the additional advantages
of capabilities for MS" and truly mobile operation as
compared to the ITD described here, many applications
do not require tandem MS and the more rugged turbo
could be user installed. However, the main advantage of
the standard ITD for vapor sampling GC/MS is its high
sensitivity in a commercially available instrument.
Coupling the vapor inlet to bench top MS such as the
ITD or a Hewlett Packard MSD is also a much more
economical way to get transportable GC/MS into a field
screening, stack monitoring or even process control
application than the prototype MINITMASS.
This paper also suggests the need for further development
in associated areas of field testing instrumentation.
Foremost are the advantages in compound range and
analysis speed which could be gained by broader
temperature range operation and temperature
programming for the transfer line column. Second is the
capability for rapid on-line enrichment in case of more
dilute target compounds or less sensitive detectors. And
finally, there is the need for continuing development in
all aspects of miniaturization (size, weight, power
requirements) and ruggedization of fieldable GC/MS
systems.
ACKNOWLEDGEMENTS
This work was funded by the Advanced Combustion
Engineering Research Center (ACERC, which is
supported by the National Science Foundation, the state
of Utah, 23 industrial participants and the U.S.
Department of Energy), U.S. Environmental Protection
Agency (EPA), Finnigan MAT Corporation, the U.S.
Army Chemical Research Development and Engineering
Center (CRDEC) and Utah Power and Light Company
(UP&L).
The authors also wish to acknowledge the invaluable
assistance of Dale Urban for general field preparations,
David Mickunas (USEPA), A. Peter Snyder (CRDEC)
and Kenneth R. Thompson (UP&L) for specific project
arrangements and ACERC participants Sue Anne Sheya,
David Wagner, Eric Lindgren, and David Pershing for
support of rotary kiln and desorption studies.
REFERENCES
1. Mcuzelaar, Henk L.C.; McClennen, William H.;
Arnold, Neil S.; Reynolds, Tim K.; Maswadeh,
Wallace; Jones, Patrick R.; and Urban, Dale T.,
"Development of the MINITMASS, A Mobile
Tandem Mass Spectrometer for Monitoring
Vapors and Particulate Matter in Air," First
International Symposium on Field Screening
Methods for Hazardous Waste Site Investigations,
Las Vegas, Nevada, October 1988.
343
-------
2. Arnold, Neil S.; McClennen, William H.; and
Meuzelaar, Henk L.C.; "Vapor Sampling Device
for Direct Short Column GC/MS Analyses of
Atmospheric Vapors," Analytical Chemistry, 1991,
in press,
3. Arnold, Neil S.; McClennen, William H.; and
Meuzelaar, Henk L.C.; "On-site Transfer Line
GC/MS" Analysis of Environmental Vapors Using
a Modified Ion Trap Mass Spectrometer," ACS
Preprints, Div. of Environmental Chemistry, Vol.
29, No. 2, Miami Beach, Florida, September 1989.
4. McClennen, William H.; Arnold, Neil S.; Roberts,
Kenneth A.; and Meuzelaar, Henk L.C.; Lighty,
JoAnn S.; and Lindgren, Eric R.; "Fast, Repetitive
GC/MS Analysis of Thermally Desorbed
Polycyclic Aromatic Hydrocarbons (PAHs) from
Contaminated Soils," Combust. Sci. and Tech.
74(1-6), 1990, 297.
5. McClennen, William H.; Arnold, Neil S.; Shcya,
Sue Anne N.; Lighty, JoAnn S.; and Meuzelaar,
Henk L.C.; "Development of Novel, Mass
Spectrometric Combustion Monitoring
Techniques," ACS Preprints, Div. of Fuel Chcm.,
Washington, D.C., August 1990.
6. Lighty, JoAnn S.; Wagner, David; Deng, Xiao
Xuei Pershing, David W.; McClennen, William
H.; Sheya, Sue Anne N.; Arnold, Neil S.; and
Meuzelaar, Henk L.C.; "On-line GC/MS Sampling
of Exhaust Gas from a Rotary Kiln Simulator,"
Air and Waste Management Assoc. International
Specialty Conf. on Waste Combustion in Boilers
and Industrial Furnaces, Kansas City, Missouri,
April 1990.
7. Lighty, JoAnn S.; Pershing, David W.; Cundy,
V.A.; Linz, D.G.; "Characterization of Thermal
Dcsorption Phenomena for the Cleanup of
Contaminated Soil," Nuclear Chem. Waste
Management, 8, 1988, 225.
Table 1
Concentration Peak Area RSD
Table 2
Tedlar Bag Samples of NPL Site Wellhead Vapors
Scan
No.
8
9
21
26
36
46
48
55
60
68
72
91
102
165
179
210
217
396
430
460
490
Compound2
11DCE
DCM
12DCA
Benzene
Trichloroethene
Cyclooctane
Chlorocyclopentene
Cyclooctane
Cyclooctane
Toluene
Cyclooctane
Cyclooctane
Perchloroethene
Elhylbenzene
m,p-Xylene
Styrene
o-Xylene
C3-benzene
Cj-benzene
C3-benzene
C3-benzene
Concentration
1-4 (1/5) 1-6
85
480
94
340
30,000 :
430
1300
3900
800
(Ppb)
(1/5)3
~
42
210
15
--
-
t
--
1,500
t
t
—
82
270
75
—
—
«
—
1 Scan number corresponding to chromatograms
in Figure 9. This number divided by four
equals retention time in seconds.
2 Positive identification for standard compounds
which were used in EPA testing while others
are tentative, e.g. cyclooctanes could be
octenes.
1 Compounds in 1-4 but not seen in 1-6 indicated
by dashes while t indicates traces detected.
16
161
245
9450
227
1430
2540
86530
14
3.4
1.0
1.7
344
-------
100%
a. TIC (Total Ion Chromatogram)
12C
150
26 -
13 -
Q
C
b) m,
ll
30
'z 91 +
l
60
Tu
92
I
j
90 1
->e (s)
I !
JO
Figure 1. (a) Total ion chromatogram for 6 repetitive
samplings of a 16 ppb toluene vapor standard. The
points of injection can be identified by the suppression of
the baseline, (b) A selected ion trace of m/z 91 and 92
for the same 6 samples. The threshold setting of the MS
prohibits exact calculation of the signal to noise ratio, but
it appears to be 8 to 1 based on height, and >20 to 1
based on area. Adapted from reference 2.
100
=
0)
_o
o
m/z 91
2l 4
T i • i • r • i' i' i' i' i''' I''' i' i
0 20 40 60 80 100
Time (s)
Figure 2. Repetitive sampling of background levels of
alkylbenzenes in a room with a recently tiled ceiling.
Labelled compounds in (b) the selected ion trace for m/z
91 are: 1) toluene; 2) ethylbenzene, 3) m,p-xylcnes and 4)
o-xylene.
inyl chloride
' I ' I ' I I
^ 12DCE
33%
35<
Sea
Tin
7o
1 i ' i
n #
>e(s)
(1 benzene
i • i ' 1 ' i
i ' i ' | ' i
50
12
i ' i ' i • | • i •
Ji toluene
l ' i • i • | • i ' i •
100
25
« d. m/z 78 *
1 11
1 l ' ' ' ' ' ' ' ' ' ' ' '
\ e. m/z 91
150 200 250
37 50 62
Figure 3. Example analysis of a 7 component gas
mixture at 350 ppb showing 3 repetitive samplings at 30
s intervals in a) total ion chromatogram and selected ion
chromatograms for quanlitation of b) 12DCE, c) 111TCA,
d) benzene and e) toluene. Note presence of ions m/z 61
from vinyl chloride, 111TCA, and trichloroelhene and
m/z 97 from trichloroethene that do not interfere with
quanlitation because of chromatographic separation.
-------
Co
».
O)
100 -
50 -
a
0)
a
a
u
«
7.4 -
3.7 ~
0.0
3.4 q
1 .7 -
0.0
1300 -,
XI
p,
a
C
O
—>
C
U
«J
C
EJ
U
a
a
o
975 -
650 -
325 -
Total Ion Current
100%
m/z 78
m/z 91
>%
05
o
t->-p-, , ^|-,T-,-^|
0 10 20 30 40 50 60 70 80 90 100
Time (s)
2.5%
1
«
b)
— 4.4%
10 20 30 40 50 60 70 80 90 100
Time (s)
Figure 4. (a) Ion profiles for a sequence of vapor
samples taken during the combustion of 11 g of poly-
propylene materials in a laboratory scale rotary kiln
simulator. The ion m/z 78 is due to benzene, while m/z
91 indicates first toluene, and then partially resolved
ethyleenzcne and xylene isomers as in Fig. 2. (b) Concen-
tration profiles in parts per billion (ppb) for 4 compounds
obtained from the integrated peak areas of selected ions
78, 91, 104 and 94 for benzene, toluene, styrcnc and
phenol respectively. Adapted from reference 2.
0.2%
SOIL "A" UftPORS
a) TIC
5,6
• I i | I I I | I i | I
b) m/z 154
c) m/z 168
d) m/z 184
I I I | T T | I | 1 p
° 8 22 Time(s) 37 «
Figure 5. On-line vapor sampling GC/MS analysis
durinr a 400 C thermal desorption of a contaminated soil.
Chromatograms of the total and selected ion signals are
shown for the vapor sample taken 12 min into run.
PAHs labeled in the TIC trace: 1) naphthalene, 2)
methylnaphthalenes, 3) C2 naphthalenes, 4) fluorene, and
5,6) unresolved phenanthrene and anthracene. Additional
compounds indicated as prominent peaks in the selected
ion traces are: biphenyl, m/z 154; dibenzofuran, m/z 168;
and dibenzothiophene, m/z 184. Adapted from ref. 4.
-------
3000 -
o 2000 -
1000
Toluene (m/z 91)
O 22May, 715 ms
V 23May, 715 ms
24May, 715 ms
22May, 495 ms
23May. 495 ms
24May, 495 ms
2000
1500
o
4)
100 200
Concentration (ppb)
JOO
400
1000
Figure 6. Comparison of toluene calibration data from
715 ms and 495 ms pulse widlh sampling. These
standard concentrations were diluted from a 50 ppm in air
compressed gas mixture. Note that the spread of
individual data points comprises three separate days of
system operation.
O
o
c
o
500
Compound, Cone.
O 12DCE, 30ppb
V 111TCA, 30ppb
D Benzene, 30ppb
A Toluene, 30ppb
• DCM, 350ppb
T 12DCE, 350ppb
• 11 OCA, 350ppb
A CCI 350ppb
0 500 1000 1500 2000
Sample Pulse Width (ms)
Figure 7. Mean peak area response versus sample pulse
width data for one gas mixture at ca. 30 ppb and another
at 350 ppb.
2500
-------
50 H
c
I)
- 0
Total Ion Current
B o
c
o
Retention Time (s)
Total Ion Current
a)
0
"""*
u
>
^J
o
Bi
26 -
13 -
0 -
m/z 61
JYi m/z 65 x 3.3
M. m/z 82 x 2.5 ,
1 M ' \
/' i '' •
/' V ''
10
b)
/& -
13 :
0 -
m/z 61
fl\i ~ ~ m/z 65 x 3.3 >'\
// 1, m/z 82 x 2.5/ \
0 5
1 i
'0
Retention Time (s)
Figure 8. Ion profiles showing the effect of sampling
time on chromatographic resolution and sensitivity, (a)
was obtained from a 330 ms vapor sample and shows the
separation of dichlorornethane (the first peak) from the
complex of 1,2 dichloroethene (m/z 61) and 1,1
dichloroethane (m/z 65). Separating these latter two
compounds may be possible with even shorter sample
pulses. The ion profiles in (b) were obtained at 715 ms
and illustrate the relationship between sample duration
and sensitivity. The m/z 82 (tetrachloromethane) trace
has a 50% increase in peak height which correlates to
absolute sensitivity of the instrument. Adapted from
reference 2.
100%
xlO
TEDLAR BAG 1-4, 1/5 dilution
vA
Scan No.
Time (s)
100%
50
12
r i' i -1 -1' i -1
100 150
25 37
A
' | ' 1 ' 1 ' 1 ' 1 ' ' 1
200 250
50 62
TEDLAR BAG 1-6, 1/5 dilution
Scan No.
Time(s)
i ' i ' i ' i i | i i ' i
50 100 150
12 25 37
I • i • i • i M"' | i I i I i i i i i
200 250
50 62
Figure 9. Total ion chromatograms from analysis of 1/5
dilution Tedlar bag samples from NPL landfill site
injection wells 1-4 and 1-6.
-------
DISCUSSION
MAHADEVA SINHA: In all your slides when you lalk about sensitivity, 1
couldn't get a good feeling. How did you find your detection limit of sensitivity
of 1 ppb or 10 ppb?
WILLI AM McCLENNEN: Basically it is a matter of looking at peak areas that
we're doing quantitation on. Routinely we would compare the peak area of any
peak we could find at the retention time we were looking at to any background
peak in that vicinity or another area of the chromatogram. So the one picture I put
into the paper that will accompany the proceedings of this will show a 16 ppb
toluene peak with virtually no baseline. There is, I think, three little blips on the
baseline. We estimated that there are signal to noises, at least the factor of 20 or
more. So basically it's a matter of looking at any discernable peak, any disruption
of the baseline noise in the area of the retention time that we're identifying the
compound.
MAHADEVA SINHA: Toluene works very good because you have almost the
base peak for the entire mass spectrum there. Other peaks are pretty small. But
when you get the mass spectrum itself, do you just identify the peak itself, one
parent peak, or the complete mass spectrum of that?
WILLIAM McCLENNEN: The mass spectrum. Typically we take data with a
full scan, so we can look at all the ions that we would typically expect. And we
can compare them to library spectra or spectra that we know.
MAHADEVA SINHA: What if I change that compound and go to, let's say.
carbon tet or a dichloroethylene, what becomes your sensitivity at that point?
WILLIAM McCLENNEN: Again, that's where I've hedged a little bit. For
compounds that do not show one nice peak or one nice fragment ion or mo lecu lar
ion we have a slightly lower detection limit. All of our quantitation for the results
I've shown you has been on single ions, not trying to combine several ions. But
we still have detection limits for the chlorinated compounds that I showed. We
were still looking at standards that were less than 20 ppb and getting good spectra
for them.
349
-------
FIELD MEASUREMENT OF VOLATILE ORGANIC COMPOUNDS
BY ION TRAP MASS SPECTROMETRY
M.E. Cisper, J.E. Alarid, and P.H. Hemberger, Chemical and Laser Sciences Division
E.P. Vanderveer, Mechanical and Electronic Engineering Division
Los Alamos National Laboratory
Los Alamos, NM 87545
ABSTRACT
We have developed a second generation transportable
gas chromatograph/ion trap detector (GC/ITD) for the
in-situ characterization of chemical waste sites. This
instrument is extensively based on commercial
instrumentation and can be used for field analysis of
volatile organic compounds (VOCs) in soil and water.
A purge and trap GC is used for sampling and
separation of VOCs from the environmental matrix
before their introduction to the ion trap detector for
mass spectral analysis. A secondary microprocessor
controls the sampling and GC hardware in parallel
with the ion trap detector, which in turn is controlled
by the host PC. The analysis of water samples is
demonstrated by using surrogate samples spiked with
the 24 VOCs contained in the Supelco A & B
Purgeable Standards, acetone, methyl ethyl ketone, and
methyl isobutyl ketone. Our first transportable
GC/ITD was demonstrated at a chemical waste site for
the analysis of volatile organic compounds in soil [1].
The second generation instrument incorporates
significant improvements in several areas and is nearly
ready for field deployment. This instrument has been
extensively characterized in the laboratory. In these
tests, we have found anomalies in quantitation that
might arise during field use. Once these problems -
which may occur with any ion trap based field
instrument - are resolved, the second generation
GC/ITD will be tested and demonstrated in the field.
The second generation transportable GC/ITD will be
described in this manuscript. Some comparisons will
be made to the first generation instrument where
appropriate.
INTRODUCTION
We describe and compare two modular field-
deployable gas chromatograph/ion trap detector
systems for characterization of hazardous waste sites .
Extraction of the analyte from the matrix occurs in the
sampling module. Although the sampling module
could be readily adapted for extraction of volatile
organic compounds from air [2], the focus of this work
has been soil and water analysis. A separations
module, i.e., a gas chromatograph, provides separation
of the extracted VOCs. A Finnigan MAT ion trap
detector (ITD) provides a simple and reliable mass
analyzer on which to base field instrumentation [3].
The ITD serves as a universal detection module. A
turn-key operating system has been developed for this
instrument. This operating system, which incorporates
additional hardware and software, allows the
instrument to be operated by personnel with minimal
technical background. Because the instrument
operates under nearly full computer control, very little
operator interaction is required. The first generation
instrument (GC/TTD-1) was developed using a purge
and trap gas chromatograph built in our laboratory;
commercially available equipment was used wherever
possible in the second generation transportable
instrument (GC/ITD-2).
EXPERIMENTAL
Ion Trap Detector
The ion trap detector was used in conventional fashion
and without modification in GC/ITD-2. An SRI
Instruments 8610 gas chromatograph was connected
directly to the ITD transfer line. Electron impact
ionization was used with the Finnigan Version 4.10
software with automatic gain control (AGC) [4]. The
Finnigan Programmer's Option Package was used to
generate FORTH subroutines and keystroke sequences.
The ion trap was tuned using perfluorotributylamine
(FC-43) as a tuning and mass calibration standard with
the automated tuning procedures contained in the ITD
software.
A schematic diagram of GC/ITD-2 is shown in Figure
1. Automation software consists of the host interface
351
-------
software, the extended FORTH Finnigan ITD
software, and keystroke sequences (i.e., macros) added
to the ITD software. The SRI Model 8690 purge and
trap device and Model 8610 gas chromatograph are
controlled via the Microstar Laboratories DAP 1200/4
data acquisition processor. In this fashion, all
components required for the second-generation
transportable GC/ITD are readily and inexpensively
available.
Although no special considerations were given to
reducing the size, weight, and power requirements of
GC/ITD-2, it easily fits into the back of a 4X4 vehicle
and can be generator powered. GC/ITD-2 is self-
contained in a housing approximately 60 cm on each
side. This housing contains the gas chromatograph,
ion trap detector, sampling system, and a 6.8-1 cylinder
of ultrapure helium. The mechanical backing pump
for the ion trap vacuum system was mounted next to
the housing. The instrument weighs ca. 80 kg and
requires less than 1.5 KVA of power.
Modifications of the Finnigan ion trap were required
for use in GC/ITD-1. The conductance-limiting
interelectrode spacers and the open-split interface were
eliminated. A wide-bore capillary column was directly
coupled to the ion trap and the standard 50 1 s'1
turbomolecular pump was replaced with a 240 1 s'1
pump to maintain the ion trap at the proper pressure
with the higher gas load imposed by the direct coupled
column. A system based on the Hitachi
HD637BO5ZOF microprocessor was designed and
built to control the sampling and chromatographic
hardware.
Surrogates and Samples
Water standards and surrogates were prepared by
adding the Purgeable Mixture A (200 |ig ml'1 of each
of 13 VOCs) and Purgeable Mixture B (200 (Jg ml'1 of
each of 11 VOCs) (Supelco, catalog nos. 4-8851 and 4-
8852, respectively) to HPLC grade water (J.T. Baker).
Aqueous solutions of acetone, 2-butanone (methyl
ethyl ketone), and 4-methyl-2-pentanone (methyl
isobutyl ketone) were also prepared and added at the
appropriate concentration. Standards at 1 part-per-
million (ppm), 100 part-per-billion (ppb), 10 ppb, and
1 ppb concentrations were prepared by dilution of a
stock solution containing the A & B Purgeables and
three ketones at the 2 ppm level. Standard solutions
were refrigerated to prevent evaporation of the VOCs.
Fresh stock solutions and standards were prepared at
approximately 3 day intervals. Fluorobenzene was
used as an internal standard at the 32 ppb level in the
aqueous standards and surrogates. A 10 ppm standard
of fluorobenzene (Aldrich) in reagent grade methanol
was prepared and used as the spike solution. Water,
methanol, and the fluorobenzene/methanol solution
were periodically analyzed with GC/ITD-2 as a check
for impurities.
Purge and Trap Chromatography
For water analysis, a 5 ml aliquot of the standard or
surrogate was loaded into a Pyrex glass tube (12.5 cm
length by 16 mm i.d.). The sample was heated to 80°C
during purging with helium at a flow rate of 60 ml
min'l. Sample recovery under these purge conditions
is shown in Figure 2. Trap effluent during the purge
cycle is diverted away from the analytical column and
vented.
After the sample purge, the adsorbent trap (50/50
Graphpak GB and Chromosorb W) was ballistically
heated to 375°C from ambient temperature at a rate of
1800°C min'1 and flushed with helium at the carrier
flow rate of 3 ml min'1 (carrier gas velocity 21 cm s'1).
The analytical column is a "VOCOL" fused silica
capillary column (30 m length by 0.53 mm i.d.) with a
3.0 nm film (Supelco, cat no. 2-5320M). The oven
temperature for the chromatography column was held
at 30°C during the purge cycle. Following heating of
the adsorbent trap, the oven temperature was ramped at
20°C mhr1 to 40°C, held for 1 minute, then ramped at
4°C min to a final temperature of 150°C. The
adsorbent trap is continually flushed with helium
during the chromatographic run; the chromatographic
oven is heated to 250°C for system cleanout after data
acquisition.
RESULTS
A list of the volatile organic compounds used in these
studies (the A & B purgeables and ketones) and the
reconstructed ion chromatogram (RIC) from the purge
and trap analysis of 5 ml of water containing 100 ppb
of these 27 compounds is shown in Figure 3. Working
calibration curves from 1 ppb to 100 ppb of 1,1,2-
trichloroethane in 5 ml of water are shown in Figure 4.
The value of the average slope of these three lines is
0.82 +/- 0.02 and the correlation coefficient for these
data is 1.00. This calibration curve was obtained using
the Finnigan Automatic Gain Control (AGC) software.
The use of the automatic gain control does not in itself
ensure that the ion trap will not operate under space-
charge conditions. At higher sample concentrations
(microgram levels) the AGC should be manually tuned
to prevent mass spectral degradation due to space-
charging. Although the calibration curve for 1,1,2-
trichloroethane is quite good, calibration curves for
other compounds can exhibit non-linearity, especially
at lower concentration levels. For these compounds,
relative sensitivity factors will vary as a function of
concentration as shown in Figure 5. This non-linearity
can be partly explained by the occurrence of
ion/molecule reactions of analyte molecules with ions
from water (H2O+, H3O+) and methanol (CH3O+,
CH3OH+, and others). In the AGC mode, the Finnigan
ion trap detector is pre-programmed (in the
instrument's firmware) to store all ions above m/z 20
(ionization DAC value of 125). However, water ions
below m/z 20 can remain in the ion trap during and
after ionization and are able to react with analyte ions
352
-------
during the analysis sequence. A plot of relative
intensity of HaO* ions (m/z 19) as a function of the
mass storage level during ionization is shown in Figure
6. These data were obtained with a Finnigan Ion Trap
Mass Spectrometer (ITMS), which does allow the
operator to select the storage level (DAC value) during
ionization. With the Finnigan ITD, the storage level
during ionization is fixed in the AGC mode (i.e., it is
not user adjustable) and it is likely that the non-
linearities displayed in Figure 5 will be observed with
any purge and trap GC/ITD system. It is also possible
for analyte ions to react with neutral water and
methanol molecules. The creation of protonated water
molecules, HjO"1", via reactions of analyte ions (shown
in the upper cnromatogram) with water background in
the ion trap detector, is demonstrated in Figure 7. We
estimate water loss during our purge cycle to be about
1-2 mg mur1. Moisture may either adsorb or simply
condense on interior surfaces of the sampling system
and subsequently be introduced to the ion trap during
trap desorption. It is important to remember here that
nanogram levels of analyte are being measured and
that seemingly insignificant amounts of moisture in the
ion trap (for example, water introduced by venting the
instrument to transport it from one area to another) can
affect the reliability of quantitation. An effective
solution to this problem will require modification of
the ITD. We are presently developing methodologies
to reduce the ion/molecule chemistry that occurs in
GC/ITD-2. These new techniques should improve
response linearity for most of the VOCs in this study.
The chromatographic retention times obtained with
GC/ITD-2 are very reproducible. Table 1 shows the
precision of retention times for several compounds at
concentration levels from 1 pbb to 100 ppb. We have
found that changes in retention time are often
accompanied by changes in mass spectral peak
distribution and intensity. However, because of the
excellent retention time reproducibility shown here,
retention time measurement of the internal standard
(fluorobenzene) can be a readily observed metric of
performance during field analysis and provides a real-
time check on instrument performance. A retention
time control chart is shown in Figure 8. The three out-
of-control points occurred following scheduled power
outages in our laboratory.
Compounds are identified on the basis of their
chromatographic retention time and their
experimentally obtained mass spectrum via an
automated identification routine. Mass spectral library
matches are highly accurate on a "first-hit" basis for
nearly all compounds in the A & B Purgeable and
ketone mixture with GC/ITD-2. Exceptions are
toluene and those compounds that coelute from the
chromatograph (1,1-dichloroethylene and methylene
chloride; benzene and 1,2-dichloroethane). Coeluting
compounds can often be identified with some mass
spectral interpretation. At this time, this capability is
not programmed into our automated library search
routines and, in fact, such interpretation might be
better left to those reviewing the data after die field
work is completed. Table 2 shows the accuracy of
library matches with GC/ITD-2. The "ID Hit" is the
rank of the correct compound identification in a list of
the 10 most probable compounds selected by the
library search routine. The purity value provides an
indication of how closely related the sample mass
spectrum is to the library spectrum on a scale of 0 -
1000. A purity of 800 or higher implies a very good
match between the two. Even though some of die
identifications shown in Table 2 do not have high
purity values, the purity values obtained so far have
been quite reproducible. This reproducibility provides
a means to screen library search data for consistency.
For example, c/s-l,3-dichloropropene is identified on a
"second-hit" with a purity value of about 400. If an
analysis of an unknown mixture were to identify a
compound as cw-l.S-dichloropropene with a
significantly higher purity value, say 800, that
identification might be suspect even though the purity
value suggests otherwise. In general, we have found
that the number of "first-hit" identified compounds is
higher with GC/ITD-2 than with GC/ITD-1. This is
most likely due to better regulation of the helium
buffer gas partial pressure in GC/ITD-2 where the
conductance limiting spacers are used and the ion trap
is coupled to the chromatography column via the open-
split interface.
DISCUSSION
The design of this transportable instrument, coupled
with the flexible control system provided by the
Microstar Laboratories DAP1200 data acquisition
processor, is well suited to problems in field analysis.
The ability to address and control ancillary
instruments, such as sampling devices, via the host
computer provides great flexibility for different
analytical problems. Keystroke sequences can provide
customized data reduction procedures for different
applications.
The combination of the data acquisition processor,
Forth interface software, and Forth keystroke
sequences added to the ITD software allow the
instrument to be operated in a turnkey fashion. That is,
in the intended mode of operation for the transportable
GC/ITD, the operator only needs to select the
appropriate operational procedures from the menu.
Figure 9 shows the menu by which the instrument is
controlled. Once selected, each item proceeds
automatically. For example, one can choose to
calibrate the instrument ("Trap Setup") or to generate
an analytical report ("Quantitation"). If "Acquisition"
is selected from the menu, a sub-menu is displayed to
provide optional analysis procedures, e.g., to analyze
for a suite of VOCs in soil or to monitor a single
compound in groundwater.
At this time, an analysis with GC/ITD-2 requires about
60 minutes. We have recently purchased the SRI
353
-------
Model 8680 Purge and Trap Device; that device should
allow us to significantly reduce the analysis time
required by GC/TTD-2. Detection limits with GC/TTD-
2 have not been determined. These limits cannot be
reliably estimated until the quantitation anomalies
described above are resolved. The response of
GC/ITD-2 to many compounds at the 1 ppb level is
quite good and we expect that detection limits will be
below the part-per-billion level for laboratory
standards. However, it is quite likely that the ultimate
detection limits for the instrument will be limited by
the nature of the sample in the field.
The gas chromatograph/ion trap detector (GC/ITD)
configuration described here has other advantages over
commercially available transportable gas
chromatograph/quadrupole mass spectrometers [5].
Ion trap detectors are inherently simple. The ion trap
does not suffer from the complexity added by external
ion sources and ion lenses. The ion trap electrode
assembly is small and rugged although the radio-
frequency and power transformers, gas supplies, data
system, and other equipment can offset the size
advantage. When operated as an ion trap detector (that
is, as a simple single-stage mass analyzer), no direct
current (dc) potentials are applied to the trap
electrodes. Without dc potentials applied to the trap
electrodes, charging phenomena are minimized as the
electrodes become dirty. Another advantage derives
from the inherently high sensitivity of the ion trap
detector. The ITD is roughly 10 to 100 times more
sensitive than a conventional transmission quadrupole
[6]. This sensitivity advantage can be pushed even
further with the addition of axial modulation to the ion
trap [7]. Finally, the ion trap can provide mass
spectrometry/mass spectrometry capabilities (MS/MS)
[8] far more simply than tandem quadrupole
instruments, which require at least two independent
quadrupole mass filters [9]. The advantages added by
axial modulation and MS/MS capabilities do however
require more expensive versions of the Finnigan ion
trap than the Ion Trap Detector used in this work; these
are the Finnigan ITS-40 Ion Trap System and the Ion
Trap Mass Spectrometer (ITMS).
The configuration of GC/ITD-2 has advantages in
several areas over the configuration of GC/TTD-1. The
first advantage is a cost savings. Savings are realized
through the use of commercially available equipment
The time and costs associated with the development,
construction, and trouble-shooting of chromatographic
hardware and the custom interface were eliminated.
GC/ITD-1 also required an expensive turbomolecular
pump and mechanical modification of the Finnigan
vacuum manifold. The second advantage is reliability.
The gas load imposed by the removal of the
conductance limiting spacers and the high flow rates in
GC/ITD-1 resulted in reduced lifetimes for the
ionizing filaments and the electron multiplier. The
DAP 1200 processor also appears to be more
compatible with the Finnigan operating system than
the custom interface and computer "hang-ups" have
been eliminated. The third advantage is the ease with
which future instruments may be built. The
construction of GC/ITD-2 by other laboratories should
be quite straightforward due to the extensive use of
commercially available equipment. In general, copies
of instruments that use custom designed and built
hardware are more difficult to build with the same
performance of the original instrument vis-a-vis
instruments that are based on proven "off-the-shelf
hardware. Finally, the Hitachi processor used in
GC/TTD-1 is no longer available and therefore not
suitable for new designs.
CONCLUSION
Work is currently in progress to fully characterize the
second generation instrument in the area of
quantitation accuracy. Once quantitation methods are
developed and tested in the laboratory, detection limits
for a wide number of compounds of environmental
significance can be determined. Analysis time will be
reduced. Tests with soil samples will be performed.
We do not anticipate significant problems in the purge
and trap analysis of soils since the SRI Purge and Trap
devices used in this instrument have been developed
for both soil and water samples. New containment for
GC/ITD-2 is being designed; the present containment
is designed to minimize instrument footprint but it
does make minor maintenance operations cumbersome
to carry out.
The ion trap detector provides many advantages as a
mass analyzer in this application. It is simple to
maintain and operate. The high sensitivity of the ion
trap and the inherent universality of the modular mass
spectrometer system are perhaps the most important
features for a field analytical instrument. The
instrument provides high specificity for compound
identification due to the two-dimensional information
provided by chromatographic retention time and mass
spectral library identification. Mobile ion trap mass
spectrometers operating in the MS/MS mode have
been successfully applied for the direct, continuous or
near-continuous analysis of permanent gases and
condensable vapors [10]. Ion trap mass spectrometer
systems have also been developed for rapid screening
of volatile organics in environmental matrices by
MS/MS techniques [11]. Such instruments can
provide highly complementary information to that
obtained from a transportable GC/ITD system. Both
types of instruments will find widespread use in
environmental restoration activities. Other instruments
used in field applications, such as a gas chromatograph
[12] or Fourier-transform infrared spectrometer [13],
do not provide the combination of sensitivity,
specificity, and universality demonstrated by the
transportable ion trap instrument.
354
-------
REFERENCES
1. Hemberger, P.H., Aland, J.E., Cameron, D.,
Leibman, C.P., Cannon, T.M., Wolf, M.A., and
Kaiser, R.E., Int J. Mass Spectrom. Ion
Processes., in press.
2. Cameron, D., Alarid, I.E., Hemberger, P.H., and
Leibman, C.P., "Real-Time Analysis of
Organic Pollutants by Ion Trap Mass
Spectrometry," Ion Trapping in Mass
Spectrometry Conference, American Society for
Mass Spectrometry. Sanibel Island, FL,
January, 1990.
3. Stafford, G.C., Kelley, P.E., Syka, J.E.P.,
Reynolds, W.E.,and Tood, J.F.J., Int. J. Mass
Spectrom. Ion Processes, 60,1984,85.
4. Yost, R.A., McClennan, W., and Snyder, A.P.,
"Picogram to Microgram Analysis by Ion Trap
Mass Spectrometry," Proc. 35th Ann. Conf.
Mass Spectrom. Allied Topics, Denver, CO,
1987, p. 789.
5. Cannon, D.R., Ind. Chem. News, 7,1986,1.
Specific instruments include:
Mobile Environmental Monitor (MEM) Broker
Instruments, Manning Park, Billerica, MA.
PETRA, VG Gas Analysis Systems, Aston
Way, Chesire, England.
TAGA, Sciex, Inc., Thornhill, Ontario, Canada.
SpectraTrak, Viking Instrument Corp., Reston,
VA.
6. Nourse, B.D., and Cooks, R.G., Anal. Chim.
Acta,228, 1990,1.
7. Weber-Grabau, M., Kelley, P.E., Bradshaw,
S.C., and Hoekman, D.J., "Advances in MS/MS
Analysis with the Ion Trap Mass Spectrometer,"
Proc. 36th Ann. Conf. Mass Spectrom. Allied
Topics, San Francisco, CA, 1988, p. 1106
8. Louris, J.N., Cooks, R.G., Syka, J.E.P., Kelley,
P.E., Stafford, G.C., and Todd, J.F.J., "New
Advances in the Operation of the Ion Trap Mass
Spectrometer," Proc. 33rd Ann. Conf. Mass
Spectrom. Allied Topics, San Diego, CA, 1985,
p. 707.
9. Yost, R.A. and Enke, C.G., Anal. Chem., 51,
1979,69.
10. McClennan, W.H., Arnold, N.S., Sheya, S.A.,
Lighty, J.S., Meuzelaar, H.L.C., "Direct
Transfer Line GC/MS Analyses of Incomplete
Combustion Products from the Incineration of
Medical Wastes and the Thermal Treatment of
Contaminated Soils," Proc. 38th Ann. Conf.
Mass Spectrom. Allied Topics, Tucson, AZ,
1990, p. 611.
11. Wise, M.B., Buchanan, M.V., Guerin, M.R.,
"Rapid Environmental Analysis by Direct
Sampling Glow Discharge Mass Spectrometry
and Ion Trap Mass Spectrometry," Oak Ridge
National Laboratory TM-11538, Oak Ridge,
TN, 1990.
12. Blanchard, R.D. and Hardy, J.K., Anal. Chem.,
58,1986,1529.
13. Spartz, M.L., Witkowski, M.R., Fately, J.H.,
Jarvis, J.M., White, J.S., Paukstelis, J.V.,
Hammker, R.M., Fately, W.G., Carter, R.E.,
Thoma, M., Lane, D.D., Marotz, G.A., Fairless,
B.J., Holloway, T,. Hudson, J.L., and Gurka,
D.F., Amer. Environ. Lab., 1, 189, 15.
ACKNOWLEDGEMENTS
P.H.H. wishes to acknowledge the efforts and
contributions of C.P. Leibman in the development,
testing, and demonstration of the first generation
transportable GC/ITD. P.H.H. also thanks Dale Spall
for his invaluable technical contributions, support, and
encouragement throughout this project.
This work has been financially supported through the
Department of Energy - Office of Technology
Development.
355
-------
Compound
Methyl Ethyl Ketone
1,1,1 -Trichloroethane
Carbon Tetrachloride
Benzene
Fluorobenzene
Trichloroethylene
1,2-Dichloropropene
Bromodichlorome thane
Toluene
1,1,2-Trichloroethane
Tetrachloroethylene
Dibromochloromethane
Chlorobenzene
Table 1.
Reproducibility of Retention Times
tR*
(1 ppb)
23.3
25.0
25.7
26.4
27.1
28.2
28.9
29.6
32.4
33.6
34.4
35.2
37.2
(lo'ppb)
23.5
25.0
25.8
26.4
27.1
28.3
28.9
29.6
32.4
33.6
34.3
35.2
37.2
(lOlfppb)
23.4
25.0
25.7
26.4
27.1
28.2
28.9
29.6
32.3
33.5
34.4
35.2
37.2
Average %rsd
23.4
25.0
25.7
26.4
27.1
28.2
28.9
29.6
32.3
33.5
34.3
35.2
37.3
0.3%
0.07%
0.1%
0.1%
0.1%
0.2%
0.1%
0.1%
0.1%
0.1%
0.2%
0.2%
0.2%
* Retention times in minutes. The start of the purge cycle is time t = 0.
** Average of 9 runs from 1 ppb to 100 ppb.
Compound
Table 2.
Reproducibility of Library Matching
ID Hit Purity* %rsd
Methyl Ethyl Ketone 1 931
1,1,1-Trichloroethane 1 599
Benzene** 1 759
Trichloroethylene 1 801
Bromodichloromethane 1 861
c-l,3-Dichloropropene 2 413
Toluene
Tetrachloroethylene 1 657
Chlorobenzene 1 826
Bromoform 1 856
* Average of 3 analyses at 100 ppb.
** Coelutes with 1,2-dichloroethane.
5%
7%
2%
2%
5%
356
-------
Figure 1. A Schematic Diagram of the Second Generation
Transportable Gas Chromatograph/Ion Trap Detector.
SRI 8690 or 8680
Purge & Trap Device
SRI 8610 Gas
Chromatograph
Finnigan
Ion Trap Detector
Helium Supply |
Microstar
Data Acquisition
Processor
-------
8!
CD
o
<
CO
6
100 A
Figure 2.
Sample Recovery for 100 ppb of
VOC / Ketone Mixture
O Methyl Ethyl Ketone
• Methyl Isobutyl Ketone
A A & B Purgeables
Analysis Number
Los Alamos
CLS-91-630
-------
Figure 3. Reconstructed Ion Chromatogram from the Purge and Trap
Analysis of 5 ml of Water Containing 100 ppb each of 28 Volatile
Organic Compounds.
11,12
T01-
2 3-4
- A.
I
fl
13:21
1280
I
1609
26:41 33:21
Scan Number
Retention Time (min)
2400
40:01
1. Acetone (not observed)
2. Trichlorofluoromethane
3.1,1-Dichloroethylene
4. Methylene Chloride
5. t-12-Dichloroethylene
6. 1,1-Dichloroethane
7. Methyl Ethyl Ketone
8. Chloroform
9.1,1,1-Trichloroethane
10. Carbon Tetrachloride
11. Benzene
12.1,2-Dichloroethane
13. Fluorobenzene
14. Trichloroethylene
15.1,2-Dichloropropene
16. Bromodichloromethane
17.2-Chloroethyl Vinyl Ether
18. Methyl Isobutyl Ketone
19. c-13-Dichloropropene
20. Toluene
21. r-1,2-Dichloropropene
22.1,1,2-Trichloroethane
23. Tetrachloroethylene
24. Dibromochloromethane
25. Chlorobenzene
26. Ethyl Benzene
27. Bromoform
28.1,1,2,2-Tetrachloroethane
-------
Figure 4.
Calibration Curve for 1,1, 2-Trichloroethane
O Dec 90
D Dec 90
A Dec 90
1E-2
1 10 100
Concentration (parts-per-billion) in a 5 ml Water Sample
Los Alamos
CLS-91-578
-------
Figure 5.
Sensitivity Factor Relative to Fluorpbenzene
as a Function of Concentration
32 ppb Fluorobenzene in 5 ml VOC / Water Standard
10.0
o
W
c
o
(0
o
I
-------
Figure 6.
V)
4)
DC
Relative Intensity of M/Z 19 as Function
of lonization DAC Value
100
90
80
70
60
50
40
30
20
10
0
12
2(
I
ft
Puncorr =
A
^
r
•
>
•
>
•
I
1
^
_..
8x10'8torr ~
d
«
—
—
—
—
—
~
'5 135 145 155 165 175 185 195 205 215 225
3 23 26 29 33 36
DAC Value
Mass
LOS ALAMOS
CLS-91-632
-------
Figure 7.
RECONSTRUCTED ION CHROMATOGRAMS FOR
A & B PURGEABLES AND KETONES
100%-
IT)
ID
OJ
i
Methyl Ethyl
Ketone
2-Chloroethyl
Vinyl Ether
Methyl Isobulyl Ketone
1.1,2-Trichloroethane
15% -,
O)
N
E
fromXH+%H2O-> X'+H3Of
I
1400 1600
I
1800
2000
2200
2400
SCAN NUMBER
LOS ALAMOS
CLS-91-628
-------
Figure 8.
Retention Time Control Chart for
Dibromochloromethane
VSL.
+99%
+95%
^f
-95%
-99%
I I I I I I I I I I I
0
o o o o
0
J 0^ -0 f—
— — — — — — — — — — — — —
— — — — — — — — — — — — —
ocan
2116
2114
2110
2106
2104
1234 56789 10 11
Run
Los Alamos
CLS-91-580
-------
Figure 9.
Instrument Control Menu
SELECTION MENU
1) Trap Setup
2) Bakeout
3) Acquisition
4) Trap cool down
5) DAP reset
6) Quantitation
(SETUP)
(BAKEOUT)
(ACQMENU)
(TRAPCOOL)
(DAPCLR)
(QUANT)
9) Quit Menu
ENTER SELECTION NUMBER:
DISCUSSION
RUSSELL SLOBODA: Once you finished prototype work, what was the
average length of time when you turned off the motor and the vehicle driving the
trailer off the site and when your calibration could actually be done?
CHRISTOPHER LEIBMAN: Start again you mean? Actually we have a fairly
good size turbo pump, the 240 liter/second turbo pump. That repre-
sents one of the modifications. It pumps out fairly quickly and we're back in
business within a day. We drive up to the site at night, and let the unit pump down;
that's been our practice. Clearly that's something we're going to have to look at
very hard.
RUSSELL SLOBODA: Do you need power throughout the night for one day's
events?
CHRISTOPHER LEIBMAN: The way it is currently configured, yes.
RUSSELL SLOBODA: Can you peel off the back of the system and then shut
things off and then turn it on the next day?
CHRISTOPHER LEIBMAN: We've considered that. But the power require-
ments when the instrument is in the standby mode just aren't considerable. The
GC's off and the turbo's up to speed, so there's very little load. The one thing I
would like to also add is, I have this demonstration on a diskette. I suspect it will
run a little more smoothly on your instrument than in this form. But if you would
like to see any data after this presentation, write and let me know.
BRIAN ECKENRODE: I was wondering about your success on the library
searching. For example, I noticed you had one spectrum, tetrachloroethylene,
that had peaks beyond the molecular ion. How is that affecting your ability to get
a hit in the library?
CHRISTOPHER LEIBMAN: We've done quite well in terms of the library
searching, the library matching. We're looking both based on retention time and
the mass spectra. So while it may see those other ions in the mass spectrum, it's
only looking for the presence of certain target ions. Operated in that mode, the
coelution does not pose a problem.
PHIL HEMBERGER: You had mentioned going to a hydrogen generator. Is
that also to supply buffer gas to the ion trap?
CHRISTOPHER LEIBMAN: Yes. And in fact some interesting work has been
done in that area. Scott McClucky and Gary Clish and O'Krish have looked at
the presence of hydrogen in the ion trap, and what they've observed is an
enhancement in sensitivity by a factor of two.
PHIL HEMBERGER: With hydrogen you would expect much different
collision of cooling of the ions then would be provided by helium. The
fragmentation patterns should also change. Do you anticipate that you're going
to have to build your own library?
CHRISTOPHER LEIBMAN: I think you need to look at the search algorithm
that we're using. The library we're using to conduct the search is one we create.
So when we go through the chromatogram and identify each of the peaks, we
essentially then identify that fragmentation pattern. So if it changes with different
collision gases so be it. It's taken into account in the way we set upour calibration
files. But, a very good point.
365
-------
TRANSPORTABLE GC/ION TRAP MASS SPECTROMETRY FOR TRACE FIELD
ANALYSIS OF ORGANIC COMPOUNDS
Chris P. Leibman. David Dogruel, Health and Environmental Chemistry Group, HSE-9,
Eric P. Vanderveer, Instrumentation Group, MEE-3 Los Alamos National Laboratory,
M/S K-484, Los Alamos, New Mexico, 87545
Abstract
A transportable purge and trap/GC/MS based on the Finnigan
Ion Trap Detector (ITD) has been developed at Los Alamos
National Laboratory for the identification and quantification
of volatile organic compounds present at chemical waste sites.
This instrumentation is being evaluated for use to support
environmental surveillance and the characterization/clean-up
of hazardous waste sites. A custom purge and trap/GC
sampling system was integrated with a modified ITD to
achieve instrument operation consistent with field activities.
The sampling system is controlled by an ancillary
microprocessor designed at Los Alamos National Laboratory.
The instrument is extensively automated and can be operated
with minimal training. Instrument operation transparent to
the field user has been achieved by integrating sampling
system control software with the operating software of the
ITD.
The instrumentation and associated methods parallel those
outlined in method 8260, SW-846. Qualitative and
quantitative analysis for the 68 target compounds and the
associated internal standards and surrogates is completed in an
automated sequence that is executed every 25 minutes. Sample
purging, analysis, data reduction, and preliminary report
generation proceeds automatically. The instrument can be
operated in a continuous mode, pausing only for sample
loading and data file specification. All data are archived on
floppy disk for subsequent review by a skilled analyst.
Part-per-trillion detection limits can be attained for many
compounds from either 5 gram soil or 5 milliliter water
samples.
Introduction
The development and use of field transportable analytical
instrumentation can significantly reduce the cost associated
with environmental surveillance and restoration activities.
Field analytical support minimizes the analytical data
turnaround time, which can expedite site characterization and
provide analytical data to field personnel for guidance of
ongoing work. Clean-up personnel can be used more
efficiently since these teams will not have to be released and
reassembled weeks later after receipt of analytical results from
a remote laboratory.
Field analytical support can directly impact the expense of
environmental clean-up by reducing the cost-per-analysis.
Cost for sample packaging, shipment, receiving and
management are eliminated if analyses are performed on site.
Field analytical support improves the chances that schedules
and monetary constraints associated with remedial activities
are met.
Performing analyses on-site can enhance the quality of
analytical data generated. Field analyses reduce the possibility
that samples will be compromised from transport and
handling. Reduced sample handling and the analysis of
samples within minutes of collection minimizes the potential
loss of volatile components. Near real-time data can also be
used to direct subsequent sampling efforts. Additionally,
initial site characterization can help delineate the sampling
grid used for collection of samples to be sent to a remote
laboratory.
A transportable purge and trap/GC/MS has been developed at
Los Alamos National Laboratory to provide field analytical
support for environmental restoration activities. The
instrument is based on a Finnigan Ion Trap Detector (ITD)1,
a rugged and simple mass spectrometer. This transportable
GC/ITD has been designed specifically to support field
operations and to provide analytical data of sufficient quality
to meet higher level data quality objectives. Our focus has
been to attempt to meet the quality control criteria outlined in
chapter 1 of SW-846 and to use procedures which parallel
method 8260, SW-846.
Experimental
Purge and Trap/Gas Chromatograph
A custom purge and trap/GC was fabricated for sampling
volatile organics in water or soil samples. The purge and
trap/GC has two sampling loops, each loop consisting of a
needle sparger and an adsorbent resin trap. A schematic of
this sampling system is shown in Figure 1. In position A,
simultaneous with the purging and concentration of one
sample onto trap B the contents of trap A are desorbed onto
the capillary column. Subsequent to the analysis of trap A, the
ten port valve (Valco Instruments Co.) is rotated to position B
367
-------
and the contents of trap B are desorbed onto the capillary
while purging/concentration occurs on the other sampling
loop. Backpressure regulation via a split maintains column
carrier gas flow. Splitless injection is performed for 20
seconds during adsorbent trap desorption. This is sufficient
for quantitative transfer of trapped target analytes while
serving to minimize water transfer to the analytical system.
Additionally, the capillary column is maintained at 10 °C
during desorption. This serves to cryofocus target analytes
onto the head of the column while allowing any water to pass
unretained.
The adsorbent traps are packed with equal amounts of 2,6-
diphenylene polymer and silica gel. Traps are heated to 200°C
at a rate of 500°C/min. Heater jackets are provided for the 5
milliliter sparger tubes and maintain a purge temperature of 35
°C. The temperature programmable GC oven can be
programmed with up to 35 multiple ramps and is capable of
sub-ambient operation. All sample transfer lines are
deactivated fused silica and are heated to 85 °C. All valves,
heaters and the GC oven associated with this sampling system
are controlled by a dedicated microprocessor. A 30m x
0.32mm i.d. DB-624 (J&W Scientific) fused silica capillary
column with 1 um film thickness was used. The capillary
column was directly coupled to the ITD via a heated transfer
line.
In addition to soil or water sampling, soil gas analysis can be
accomplished by replacing the needle spargers with an
adsorbent trap thermal desorption unit. Soil gas or air samples
are collected on an adsorbent trap using an air sampling pump.
The air sampling tubes are then transported to the instrument
and placed in a heater assembly whereby trap contents are
thermally desorbed onto the primary adsorbent traps shown in
Figure 1. Conversion from soil/water analysis to air analysis
can be accomplished within 5 minutes. Instrument operation
is modified via the computer to accommodate the air sampling
trap desorption/analysis.
Mass Spectrometer
A Finnigan MAT Ion Trap Detector1 was used with the
following modification. The supplied transfer line and open
split interface were eliminated. The 50 L/sec turbomolecular
pump was replaced with a 240 L/sec turbomolecular pump.
The larger pump was required to handle the increased gas load
realized from the direct coupling of the capillary column to
the ITD. The larger turbomolecular pump also reduces pump
down time following system venting.
Electron impact ionization was used; the ionization period was
regulated using Finnigan supplied automatic gain control
software.
Data System/Automation
A Zenith Supersport 286 laptop computer was used for data
acquisition and instrument control. All aspects of mass
spectrometer and sampling system operation were controlled
through the dedicated laptop computer. Finnigan supplied
ITD control software (version 4.10) with the programmer's
option served as the platform for system automation. FORTH
subroutines and keystroke sequences were incorporated with
the Finnigan supplied software to automate ITD data
acquisition, quantitation, and report generation.
Communication to the sampling system microprocessor was
through the serial port of the laptop computer. Sampling
system control was achieved using assembly language
programs. Parameters for sampling system event sequencing
and heater or GC oven temperatures were written into the
Finnigan ITD software using the programming option.
Physical Requirements
The total instrument dimensions are 17.5" x 23.5" x 26" ( H x
W x D ) exclusive of the laptop computer. The instrument can
be deployed in a vehicle equipped with compressed gas supply
and a small liquid nitrogen dewar if cryogenic operation is
required. A portable generator or line power is required.
Power consumption is less than 1.5 kW. For field test to date,
the instrument has been deployed in a 12 foot trailer.
Methods/Operational Sequence
The methods used with the transportable purge and
trap/GC/ITD parallel those outlined in method 8260, SW-8462.
Following instrument pump down and warm up of all heated
zones, filament continuity and water concentration in the ITD
are checked. Mass calibration is verified using
perfluorotributylamine (PFTBA). Depending on the end use
of field generated data, different levels of quality control can
be used. An ITD tuning check can be performed using 4-
bromofluorobenzene (BFB) to ensure that abundance criteria
recently specified in method 524.2 is met. Figure 2 shows ion
chromatogram derived from the molecular ion of 4-
bromofluorbenzene obtained by purging a solution containing
50 ng 4-bromofluorobenzene. The mass spectrum obtained at
the scan indicated by the cursor shown in Figure 2, is shown
in Figure 3. This mass spectrum meets the abundance criteria
specified in method 8260, SW-846. Following tune
verification, a calibration curve can be established or
continued adherence to the calibration curve can be checked
using a midpoint standard. The midpoint standard check can
also be used to update retention times on a daily basis.
Currently our target list comprises the 68 target compounds
shown (with their corresponding internal standards and
surrogates) in Table 1. Following analysis of a blank, sample
analysis is performed in a continuous mode, pausing only for
sample loading and data file specification. Data acquisition is
followed automatically by data reduction. Target compounds
are identified by 1) elution of sample component at the
appropriate elution window and 2) comparison of the sample
mass spectrum with the standard reference mass spectrum.
Standard reference mass spectra are obtained from the analysis
of calibration mixtures. If any targeted compounds are
detected, a hardcopy preliminary report is generated
immediately. All data are archived on machine readable
media for subsequent review by a skilled analyst.
Results/Discussion
We have successfully deployed the transportable GC/ITD at
waste sites at Los Alamos National Laboratory. To date field
trials have been completed without significant instrument
failures. The qualitative and quantitative analysis for the 68
target compounds and the associated internal standards and
368
-------
surrogates is accomplished in the field in an automated
sequence executed every 25 minutes. A portion of the total
ion chromatogram obtained in the field from 5 mis of a 50 ppb
water standard is shown in Figure 4. Retention times for the
target compounds reflected in Figure 4 are given in Table 1.
Chromatographic development is completed within 16 minutes.
The need to obtain the Chromatographic resolution displayed
in Figure 4 is dependent on the site specific data quality
objectives established. If only screening is required or the
target list is more limited, Chromatographic development can
be reduced by using a steeper GC oven temperature ramp for
faster elution. An example of instrument operation in the fast
screening mode is shown in Figure 5, representing a
chromatogram obtained from the internal standards/surrogates
spiking mixture used in this work.
During field trials, co-located samples were taken for
comparisons between the transportable GC/ITD and a
laboratory based GC/quadrupole mass spectrometer. Table 2.
shows the comparison between an analysis performed at a
waste site with the GC/ITD and results obtained from the
remote laboratory. The results shown in Table 2. reflect a 1 to
100 dilution (high level-methanol extraction method)2.
Difference between the field results to those obtained at the
remote laboratory may reflect the loss of volatile components
during sample transport.
Low part-per-trillion detection limits have been achieved for
some compounds from 5g soil or 5 mis water samples with this
instrumentation. An extracted ion current profile of m/z 98,
the quantitation ion of toluene d8 (a surrogate), obtained from
a 20 ppt solution of the compounds listed in Table 1. (100
picograms/component in 5mls) is shown in Figure 6. The
signal to noise ratio for m/z 98 in this chromatogram is
approximately 10:1. The complete background subtracted
mass spectrum for toluene d8 is shown in figure 7. No toluene
d8 was detected in the blank which preceded the ion
chromatogram shown in figure 6. Low part-per-trillion
detection limits cannot be routinely achieved in the field.
However, the high sensitivity of the instrument increases the
degree of confidence in the automated mass spectral
identifications performed in the field for a higher working
concentration range. Our targeted working concentration
range for field studies is from 100 ppt to 100 ppb for most
compounds.
ensure data quality. A high degree of specificity for
compound identification is achieved with retention time and
mass spectral information. The instrument is fast, analysis for
the 68 target compounds outlined in method 8260, SW-8462
can be achieved in 25 minutes. For screening, the
Chromatographic performance can be reduced to reduce
analysis time. Additionally, part-per-trillion detection limits
have been demonstrated for many compounds with this
instrument.
References
1) Stafford, G.C., Kelley, P.E., Syka, J.E.P., Reynolds, W.E.,
and Todd, J.F.J. "Recent Improvements in and Analytical
Applications of Advanced Ion Trap Technology " Int. J. Mass
Spectrom. Ion Processes, 60, Special Issue, Sept., 1984, 85.
2) "Test Method for Evaluating Solid Waste Physical/Chemical
Methods, SW-846, Third Ed., Update I, Method 8260. Office
of Solid Waste and Emergency Response, U.S. Environmental
Protection Agency, Washington, D.C.
Acknowledgements
C.P.L. wishes to acknowledge the contributions of P.H.
Hemberger, T.M. Cannon, and M.A. Wolf during the initial
development of this instrumentation.
This work has been financially supported by the Department
of Energy Office of Environmental Restoration and Waste
Management, Los Alamos National Laboratory Environmental
Restoration Program, Applied Instrumentation Technology.
Conclusion
A transportable GC/MS for the qualitative and quantitative
analysis of volatile organic compounds present at chemical
waste sites has been developed at Los Alamos National
Laboratory. System components have been integrated to
produce an instrument which is extensively automated and can
be operated with minimal training. Protocols for field
technicians with subsequent data review by a skilled analyst
369
-------
TABLE 1.
VOLATILE INTERNAL STANDARDS WITH CORRESPONDING
ANALYTES ASSIGNED FOR QUANTITATION
Pentafluorobenzene
Dichlorodifluoromethane
Chloromethane
Vinyl Chloride
Trichlorofluoromethane
l,l-Dich!oroethene
Trichlorotrifluoroethane
lodomethane
Carbon Disulfide
Acetone
Methylene Chloride
Acrylonitrile
trans- 1,2-Dichlorethene
1,1-Dichloroethane
Vinyl Acetate
2,2-Dichloropropane
cis-1,2-Dichloroethene
2-Butanone
Bromochloromethane
Chloroform
1,1,1 -Trichloroethane
Carbon Tetrachloride
1,1 -Dichloropropene
Benzene
1,2-Dichloroethane-d4 (surrogate)
1,2-Dichloroethane
1.4-Difluorobenzene
Trichloroethene
1,2-Dichloropropane
Dibromomethane
Bromodichloromethane
2-Chlorovinylether
trans-1,3-Dichloropropene
Toluene d8 (surrogate)
Ret. Time
(min: sec)
5:45
:04
:18
:28
1:18
2:00
2:03
2:10
2:09
2:08
2:43
3:13
3:09
3:59
4:18
4:59
5:02
5:09
5:28
5:28
5:34
5:45
5:46
5:58
.5:56
6:01
6:24
6:36
6:48
6:54
7:04
7:20
7:26
7:38
Chlorobenzene d5
4-Methyl-2-pentanone
Toluene
cis-1,3-Dichloropropene
1,1,2-Trichloroethane
Tetrachloroethene
1,3-Dichloropropane
Chlorodibromomethane
2-Hexanone
1,2-Dibromomethane
Chlorobenzene
1,1,1,2-Trichloroethane
Ethylbenzene
m,p-Xylene
o-Xylene
Styrene
Broinoform
Isopropyl Benzene
1.4-Dichloroethane-d4
4-Bromofluorobenzene (surrogate)
Bromobenzene
1,2,3-Trichloropropane
1,1,2,2-Tetrachloroethane
n-Propylbenzene
2-ChIorotoluene
4-Chlorotoluene
1,3,5-Trimethylbenzene
t-Butylbenzene
1,2,4-Trimethylbenzene
S-Butylbenzene
1,3-DichIorobenzene
1,4-Dichlorobenzene
p-Isopropyltoluene
1,2-Dichlorobenzene
n-Butylbenzene
1,2-Dibromo-3-Chloropropane
1,2,4-Trichlorobenzene
Napthalene
Hexachlorobutadiene
1,2,3-Trichlorobenzene
Ret. Timg
(min: sec)
8:56
7:36
7:42
7:55
8:05
8:10
8:13
8:25
8:19
8:31
8:58
9:04
9:07
9:15
9:39
9:41
9:52
10:04
11:36
10:15
10:24
10:29
10:26
10:33
10:39
10:46
10:46
11:09
11:12
11:24
11:31
11:38
11:34
12:03
12:04
13:03
14:11
14:21
14:27
14:54
370
-------
TABLE 2.
COMPARISON OF FIELD ANALYSIS TO LABORARORY BASED ANALYSIS*
Compound Field Laboratory
1,1,1-Trichloroethane 1900 ppb 1340 ppb
Tetrachloroethene 1800 ppb 1500 ppb
2-Butanone 140 ppb 40 ppb
Trichlorotrifluoroethane 2950 ppb 810 ppb
*100x Dilution Required Prior to Anaylsis
371
-------
POSITION A
ANALYSIS • TRAP A (-
SAMPLE COLLECTION - TRAP B (
H« PURGE
GAS
POSITION B
ANALYSIS • TRAP B (
SAMPLE COLLECTION - TRAP A (-
Figure 1. Schematic of Transportable GC/ITD Sampling System.
372
-------
o
(X
U
c
o
TJ
174-
569
9:21
9:41 18:61
620
10:21
10:41
Scan Number
Retention Time (min:sec)
Figure 2. Extracted Ion Current Profile Derived from Molecular Ion of 4-Bromofluorobenzene
(50ng).
95
I INT
75
6168
81
174
104113 128 141
i i i
120 140 160
m/z
Figure 3. Mass Spectrum of 4-Bromofluorobenzene Which Meets Tune Criteria.
373
-------
r;
U
*
2 to
s275
c
c
o
u
a:
<-- xlO
50 180 150 206 250 300 358 408 450 500 550 608 650 706 750 800 850 906
0:51 1:41 2:31 3:21 4:11 5:01 5:51 6:41 7:31 8:21 9:11 18:01 10:51 11:41 12:31 13:21 14:11 15:01
Scan Number
Retention Time (minisec)
Figure 4. Total Ion Chromatogram Obtained in the Field from 50 ppb Standard.
-------
3
u
I 58
"2 tO
§275
c
o
fyrr
50 100 150 200 250 300 350
0:51 1:41 2:31 3:21 4:11 5:01 5:51
Figure 5. Total Ion Chromatogram of Internal Standards Obtained in Fast Screening Operational
Mode.
100%
o
lH
ft,
a
U
-o
01
X
W
7:41
8:01
500
8:21
520
8:41
540
9:01
Figure 6. Extracted Ion Current Profile of Toluene d8 Molecular Ion from 20 ppt Standard Solution
(100 picograms/5 mis water).
375
-------
1062
3
ea
•i—'
<
CO
>
55
52
49
57
73
76
82
87
96
, IB 113 120
i ,i i
i
108
120
m/z
Figure 7. Background Subtracted Mass Spectrum Obtained from 20 ppt Standard Solution.
376
-------
The Use of Field Gas Chromatography to Protect
Groundwater Supplies
Thomas M. Spittler
Director, USEPA Regional Lab
60 Westview St., Lexington, MA 02173
Abstract
The Use of field instrumentation to
detect the presence of volatile chem-
icals in the environment has undergone
rapid and dramatic change in the past
fifteen years. This paper will give a
brief overview of this development and
indicate some of the promising uses to
which this equipment can be put in the
service of groundwater protection.
The use of non-specific detectors to de-
termine the presence of volatile organ-
ics in the environment is really a pre-
liminary phase of field gas chromato-
graphy. Such instruments can determine
low ppm levels of most volatiles and are
now equipped with on-board data-logging
capability.
Portable gas chromatographs are obvi-
ously more useful for identification and
guantitation of mixtures of volatile or-
ganic contaminants. While there are
many different instruments on the mar-
ket, only a limited number meet the
qualifications of true portability, rug-
gedness and high sensitivity that are
frequently required in field studies.
The capabilities and limitations of
several field instruments will be de-
scribed in some detail.
The remarkable sensitivity of some field
gas chromatographs enables a field
chemist to detect very low levels in
ambient or vadose zone soil gas surveys
(low ppb levels on a wt/wt basis).
Also, the use of headspace analysis in
the field provides ppt sensitivity for
volatile organics in water. These tech-
niques provide a field analyst with an
ability to detect contamination in
potable water even when levels are well
below any need for health concerns.
The above techniques and equipment pro-
vide the basis for a truly preventive
strategy to protect groundwater sup-
plies. Some discussion of the various
stages useful in developing and imple-
menting a groundwater protection stra-
tegy will be discussed.
Introduction
Contamination of groundwater in the
1970's was primarily a matter of con-
cern for bacterial, odorous or visible
constituents deemed undesireable for
potability. With the advent of sensi-
tive detection equipment, attention
began to focus on the presence of or-
ganic contaminants as well. It did not
take long to realize that in most areas
of the country groundwater contamination
was principally caused by volatile, low-
solubility solvents and hydrocarbons.
Now it is widely acknowledged that the
greatest threat to groundwater is from
fuel leaks and other solvent losses to
the ground and water table.
Total Organic Analysers
Field measurement of volatile organics
at waste sites began historically with
the use of portable total organic vapor
detection equipment.(1) Typical exam-
ples of this equipment were the HNu PM
101, Century Systems OVA and the AID PID
377
-------
analyser. Using either photoionization
or flame ionization as the detection
principle, these instruments were able
to detect most volatile organics at
about the 1-5 ppm range. Despite their
obvious lack of specificity, such equip-
ment is still widely used to detect the
presence of volatile organics from fuel
leaks, spills and improper disposal of
solvents in pits, ponds and lagoons.(2-
5)
Depending on the nature of the detector,
some instruments were much more sensi-
tive to certain classes of volatile
organics. For example, the PID detector
can detect about 1 ppm of benzene and
chlorinated ethylenes, but only about 40
ppm of the chlorinated alkanes. In
fact, some PID detectors of an early de-
sign (HNu and AID) had almost no sensi-
tivity to chloroalkane molecules. This
was based on a paper on photoionization
theory as first proposed by Driscoll of
HNu.(6) When Photovac introduced their
total PID detector it was apparent that
this detector, while still operating
with lamps at 10.2 or 10.6 ev, could
readily detect compounds with ionization
potentials above the rated energy of the
lamp. While no theoretical explanation
of this vastly heightened sensitivity
has been published, users have known for
years that 111 TCA can be detected with
the Photovac gas chromatograph.
About three years ago a significant ad-
vance on the total analysers appeared on
the market in the form of data-logging
capability. First came the "Smart Port-
able" from Thermo Environmental Instru-
ments (formerly AID) which was followed
closely by the Photovac "Microtip". Now
field data could be stored in computer
memory and dumped via an interface cable
to a computer for later display in
either tabular or graphic format.
?ield Gas Chromatographs
The limitations of non-specificity soon
led investigators to taking small field-
designed gas chromatographs with them
for field studies.(1,7) In the 1970's
this usually meant either the Century
OVA equipped with an ambient temperature
gas chromatographic column or the
isothermal unit by AID.(8) Both
instruments had the advantage of careful
attention to field needs: they were
ruggedly designed, had a good track
record for field usefulness and
contained unique features that made for
versatility in addressing a wide range
of field problems. For Example, the
AID unit could accept five different
detectors in the same gas handling and
electronics package. The Century unit
contained features that allowed a field
chemist to rapidly screen samples using
a total analysis mode and at any time
switch into the gas chromatographic mode
when the detector showed the presence of
volatile organics in the sample stream.
(1)
Headspace Analysis
To demonstrate the usefulness of field
chromatography, consider screening water
samples by analysing headspace above
collected drinking water samples. These
samples can be rapidly screened by sim-
ply injecting 200 ul of headspace gas
into the GC septum and observing the
total response of the detector in the
backflush configuration. When the back-
flush peak exceeds some low limit, the
presence of ppb levels of dissolved vol-
atiles is indicated.(1) At this point a
sample can be injected into the septum
using the GC configuration and rapid
analysis with rather good resolution can
be achieved for the typical list of vol-
atiles found in contaminated ground
water. Identification is performed by
comparing peak retention time to known
standard mixtures in the field. Quan-
titation is achieved by comparing the
unknown peaks to known standards with
identical retention times.(9) Where re-
tention times are ambiguous, it is a
simple matter to change the column and
repeat the sample and standard to de-
termine retention times again on a
different packed column.
Soil Gas Analysis
In addition to field headspace analysis,
a second powerful tool to aid in field
investigations was the technique known
as "Soil Gas Analysis". It was
principally the remarkable sensitivity
of the Photovac PID detector that made
this method of vapor detection so
widespread. Now it was possible to
detect typical aromatics or chlorinated
alkenes at the ppb (wt/wt) level in
air.(10,11) With this increased
sensitivity, many investigators began to
determine the presence of volatiles from
spills or underground tank leaks by the
simple expedient of measuring their con-
centration in the vadose zone. (12-14)
Soil gas analysis, coupled with
headspace analysis provides the field
378
-------
investigator with tools to locate and
track under-ground plumes or tank leaks
and determine rapidly their impact on
local ground-water. There are certain
features of the Photovac GC which enable
an experienced field analyst to do this
work more effectively. First, the
instruments are typically supplied with
two columns so that the field analyst
can not only perform reliable identi-
fications using the two-column technique
but he can also use a short column in
one position to speed up screening of
samples.(15) Second, the use of
inexpensive zero-grade compressed air
eliminates the typical problem of
sample matrix interference with the
early part of the chromatogram since the
carrier gas is now identical to the
usual sample matrix, ambient air or
headspace above soil or aqueous samples.
(12) Third, the use of an isothermal
oven containing a wide-bore capillary
column has greatly enhanced the reso-
lution of field gas chromatography.
When discussing the necessity for high
resolution in field work, this observa-
tion has been made frequently by exper-
ienced field investigators. Most field
contamination incidents do not involve a
large number of volatile compounds. In
fact, it is more common to find two to
five volatiles in the typical field
study. Even in this situation, only one
or two compounds predominate at a site
and are the principal reason for the
investigation and remediation.(16)
Where a larger number of gas chromato-
graphic peaks are found the experienced
field chemist will immediately suspect
the presence of some type of hydrocarbon
fuel. In these cases it is much easier
to continue the investigation by merely
observing the pattern and relating it to
a type of fuel (e.g. gasoline, diesel,
jet fuel etc.). Even number two fuel
oil has enough of a volatile fraction to
be rather readily detected in under-
ground tank leaks and spills.(17,18)
Groundwater Protection Strategy
Using some of the above observations and
equipment, it is possible now to devise
a practical strategy for groundwater
protection which has much more of the
"prevention" aspect than the "reactive"
component so often found in environ-
mental contamination incidents.(19)
The first consideration in a real pro-
tective strategy would be the unusual
sensitivity of today's field gas chro-
matographs. For example, the Photovac
has a sensitivity to volatiles in air in
the range of parts per billion (pg/cc).
(11) Consider a volatile like benzene
dissolved in groundwater. The benzene
will partition into the headspace of a
closed vial with a distribution coef-
ficient of about 1 at room temperature.
(20) This means that if benzene was
present in the aqueous phase at the ppb
level, it is present in the vapor phase
at the ppm level. But we have said that
the Photovac PID is capable of detecting
one ppb of benzene in air. It follows
that when water concentrations are in
the ppt range, it is still possible to
detect them using the headspace
technique.
Consider now the fact that many if not
most public water supplies drawing on
ground water are surrounded by test bor-
ings or other experimental wells often
drilled when the original supply was
under consideration for exploitation.
Where this is not the case, it is a
relatively simple matter to place such
test wells in strategic locations up-
gradient of the supply wells so that
they can be used as an early warning
monitoring field. Instead of regular
testing of production wells, these test
wells can and should be regularly tested
for possible signs of early incursion of
contaminant plumes into the production
well field. When subsequent tests show
the presence of increasing levels (even
at the ppt level) of volatiles like
aromatics from gasoline leaks or chlor-
inated hydrocarbons from other sources,
it is time to investigate potential
sources. The mere fact of increasing
levels, low though they may be at the
present time, is clear evidence that the
water supply is under threat of future
higher contamination which might even-
tually render it unfit for consumption.
In the early investigation stage of this
process, soil gas analysis often can and
will play a key role. Establishing soil
gas sampling profiles will often show
clearly whence the contaminant plume or-
iginates. Because of the low levels of
contamination present in the perimeter
of the production well field, there is
still adequate time to lay out a plan to
find and stop the source of contamina-
tion. In this exercise great care must
be taken that not only the source be
removed, but also the contaminated soil
beneath the source be remediated as soon
and as efficiently as possible to stop
any further discharge of contamination
379
-------
to the aquifer. For this purpose there
are now many alternatives. Among the
most practical is vacuum extraction.(21,
22)
Vacuum Extraction Technique
Vacuum extractions is a method of choice
for removal of volatiles from soil for
several reasons. First, it addresses
the problem at the point where volaltile
organic contamination directly enters
the aquifer. Second, vacuum extraction
can rapidly and effectively remove pre-
cisely that fraction of the organic
contamination which is most volatile and
water soluble. In fact, volatility and
water solubility of hydrocarbons go hand
in hand. Hexane is about 13 ppm water
soluble, octane is 0.6 ppm soluble and
n-dodecane is only 4 ppb water soluble.
Thus, spills involving fuels less vola-
tile than number 2 fuel oil have very
little tendency to dissolve in rain
water and reach the aquifer. Even a
product layer of these heavier hydro-
carbon fuels will only contribute very
low levels of organic contamination to
the underlying groundwater. On the
other hand, the very soluble aromatic
fraction of gasoline, diesel and fuel
oil has sufficient volatility that it is
rapidly removed from the soil by vacuum
extraction.
It has been the experience of many who
use this removal technique of vacuum
extraction that once the bulk of the
more volatile constituents are removed
from the spill site, normal bacterial
action is enhanced and rapidly consumes
the higher boiling constituents which
cannot be removed from soil by vacuum
extraction.(23) The combination of
enhanced aeration and reduced volatile
content in the soil is precisely the set
of conditions which most favor natural
degradation of the residual hydrocarbons
from any fuel contamination situation.
Interceptor Well Installation
The last step in a prevention oriented
clean-up strategy should be to install
interceptor wells in the path of the
groundwater contamination plume. For
this purpose, it is imperative that
careful depth profiling of contamination
in the aquifer precede any attempt to
install interceptor or barrier wells.
The literature is replete with studies
which indicate that groundwater plumes
are usually confined in their vertical
and horizontal dimensions by the natural
geological features and the wide range
of permeabilities of aquifer materials.
(24) By careful placement of screens in
the interceptor wells a contamination
plume can be cost-effectively removed
from the aquifer while permitting clean
water from other parts of the aquifer to
continue to supply the production well.
Field GC and Groundwater Protection
Routine field monitoring of the test
wells and withdrawn and aerated water by
a field chromatograph will insure that
the water supply remains in a potable
condition. At the same time vacuum ex-
tracted volatiles can be monitored to
prevent air pollution. The use of real-
time monitoring will prevent inadvertant
environmental damage during the cleanup.
Spray Aeration Technique
Regarding techniques to aid in restoring
the quality of water removed from the
plume zone, I wish to discuss a method
developed several years ago in Miami by
Paul Wood.(25) Wood's technique was
spray aeration as an alternative to the
more conventional stripping towers. In
his system, water is sprayed upward in a
box the dimensions of which are not cri-
tical as long as about 14 ft in height
is used. At the base of the box is a
tank to receive the sprayed water. A
second pump and spray head is placed in
series in a second box to form a two-
stage system of spray aeration. Wood
measured an average 90% efficiency per
aeration stage. It is then a simple
matter to place enough stages in series
to remove dissolved volatile organic
contamination down to whatever level is
satisfactory for recycling back into the
aquifer.
Contrasting Wood's spray aeration with
stripping towers, the following five
points should be made. 1) Spray
aeration is considerably less expensive
to build and maintain. 2) Biofouling of
the aeration system is almost non exist-
ent. Compare this to the continual
buildup of large bacterial colonies in
stripping towers. 3) Buildup of oxide
films on the stripping tower coils is
absent in the spray aeration box. 4)
Concentration of removed volatiles is
much higher in the natural draft exhaust
than in the large volume-high flow rate
coming from a stripping tower. 5) The
highly concentrated exhaust stream of
the aeration box is more easily captured
and can be more cost-effectively
380
-------
prevented from becoming an air pollu-
tion problem.
Field GC and Spray Aeration
Once more, the use of a portable gas
chromatograph to monitor the performance
of the vapor handling system is obvious.
Data can be obtained in real time to
monitor the efficiency of spray aeration
using headspace analysis of inlet and
finished water. Vapor concentration of
the volatiles from the aeration tower
discharge can also be monitored in real
time to assist in designing and oper-
ating a vapor recovery system. Where
vapor recovery is necessary, systems
such as the CRS technique (26) will also
benefit from ongoing field monitoring
for optimization of performance.
Conclusions
In conclusion, we have seen that the
availability of inexpensive, sensitive
and rugged field gas chromatographs can
make a substantial contribution towards
a practical groundwater protecticr. stra-
tegy. There are now many examples in
the northeast where these ideas have
been applied at the local municipal and
county 1evel.(17,19) With the limited
resources of federal programs and the
dwindling funding of state programs,
this :r.ove towards local self-help is not
only welcome but long overdue. Such a
strategy, under the guidance and tech-
nical overview of state and federal
programs promises to be a cost-effective
way to insure the future purity of
groundwater resources. There is also a
growing awareness that such inexpensive
techniques as field gas chromatography
are sorely needed in many other
countries where emerging environmental
awareness cannot keep pace with limited
budgets, but where a high level of
dependence on groundwater is a practical
necessity.
Bibliography
1. Spittler, T.M., "Use of Portable
Organic Vapor Detectors for Hazardous
Waste Site Investigations", HMCRI Con-
ference, Washington, DC, Oct. 15-17,
1980.
2. Fitzgerald, J . J . , "On-Site Analytical
Screening of Gasoline Contaminated Media
Using a Jar Headspace Procedure",
Petroleum Contaminated Soils, V. II,
Lewis Publishers, Inc., Chelsea, MI,
1989, pp. 119-136.
3. Robbins, G.A., Bristol, R.D. & Roe,
V.D., "A Field Screening Method for
Gasoline Contamination Using a Poly-
ethylene Bag Sampling System", GWMR,
Fall, 1989, pp. 87-97.
4. Robbins, G.A., Deyo, E.G. et al,
"Soil Gas Surveying for Subsurface
Gasoline Contamination Using Total
Organic Vapor Detectors - Ft. I,
Theory", GWMR, Summer, 1990, pp. 122-31.
5. Robbins, G.A., Deyo, E.G. et al,
"Soil Gas Surveying for Subsurface
Gasoline Contamination Using Total
Organic Vapor Detectors - Pt. II, Field
Experimentation", GWMR in press.
6. Driscoll, J.N. & Spaziani, F-F;,'
"Trace Gas Analysis by Photoionization",
Analytical Instrumentation Division of
ISA, King of Prussia, PA, May, 1975.
7. Clay, P.P. & Spittler, T.M., "The Use
of Portable Instruments in Hazardous
Waste Site Investigations", HMCRI Con-
ference, Washington, DC, Nov. 29-Dec. 1,
1982, pp. 40-44.
8. Mehran, M. , Michael, J. S. Sirota, E. ,
"Delineation of Underground Hydrocarbon
Leaks by Organic Vapor Detection", Pro-
ceedings of HMCRI Conference, Washington
DC, Oct. 31, 1983, pp 94-97.
9. Clark, A.E., LatailU, M.M. &
Taylor, E.L., "The Use of a Portable GC
for Rapid Screening of Samples for
Purgeable Organic Compounds in the Field
and in the Laboratory", USEPA Region I
Lab SOP, Lexington, MA, 1982.
10. Barker, N.J. & Leveson, R.C., "A
Portable Photoionization GC for Direct
Air Analysis", American Laboratory, Dec.
1980, pp. 76-83.
11. Leveson, R.C. & Barker, N.J., "A
Portable Multi-Component Air Purity
Analyser Having Sub-ppb Capabililty
without Sample Preconcentration", 27th
ISA Conference Proceedings, St. Louis,
MO, Spring, 1981, pp. 7-12.
12. Spittler, T.M., Clifford, W.S. &
Fitch, L.G., "A New Method for Detection
of Organic Vapors in the Vadose Zone",
AWWA Conference Proceedings, Denver, CO,
Nov. 19-21, 1985, pp. 236-246.
13. Marrin, D.L. & Thompson, G.M.,
"Remote Detection of Volatile Organic
Contaminants in Groundwater via Shallow
Soil Gas Sampling", NWWA Conference
Proceedings, Petroleum Hydrocarbons and
Organic Chemicals in Groundwater, 1984.
14. Nadeau, R.J., Stone, T.S. & Klinger,
G.S., "Sampling Soil Vapors to Detect
Subsurface Contamination: A Technique
and Case Study", NWWA Conference
Proceedings, Denver, CO, Nov. 19-21,
1985, pp. 215-226.
15. Spittler, T.M., Lataille, M.M. &
Parks, P.A., "Correlation between Field
GC Measurement of Volatile Organics and
381
-------
Laboratory Confirmation of Collected
Field Samples Using a GC/MS", APCA
Specialty Conference, Chicago, IL, Mar.
1983.
16. Doty, C.B. & Travis, C.C., "The
Superfund Remedial Action Decision
Process: A Review of Fifty Records of
Decision", JAPCA, v. 39, # 12, 1989, pp.
1535-43.
17. Stiefel, C.L. & Heufelder, G.R.,
"Regulation and Testing of Residential
Underground Fuel Storage Tanks" Journal
NEWWA, Dec., 1988, pp. 256-266.
18. Wilhelm, R.W. & Bouchard, R.J.,
"Assessment and Remediation of Residen-
tial Properties Contaminated with Home
Heating Oil", Petroleum Contaminated
Soils, v. II, ch. 27, Lewis Publishers,
Inc, Chelsea, MI, 1989, pp. 329-346.
19. Anon., "A Groundwater Protection
Strategy for Local Municipalities",
Water Connection, NEIWPCC Newsletter, v
3, # 2, 1986.
20. Mackay, D. & Shiu, W.Y., "A Critical
Review of Henry's Law Constants for
Chemicals of Environmental Interest", J
Phys. Chem., v. 10, # 4, 1981, pp. 1175-
21. Malot, J.J., "Cleanup of a Gasoline
Contaminated Site Using Vacuum Extrac-
tion Technology", Petroleum Contaminated
Soils, v. II, Lewis Publishers, Inc.,
Chelsea, MI, 1988, pp. 283-301.
22.Bowman, R.S., "Manipulation of the
Vadose Zone to Enhance Toxic Organic
Chemical Removal", 2nd. Int'l. Workshop:
Behaviour.of Pollutants in Porous Media,
Bet Dagan, Israel, June, 1987.
23. Connor, J.R., "A Case Study of In-
Situ Soil Venting in Conjunction with
Field Recharged Carbon Adsorption for
Gasoline Removal", Unpublished, Private
Communication.
24. Cherry, Prof. J.A., University of
Waterloo, Waterloo, Ont. Canada, Private
Communication.
25. Malot, J. J. & Wood, P. R., "Low
Cost, Site Specific Total Apoproach to
Decontamination", Soils Contaminated by
Petroleum, Wiley and Sons, NY, 1988, pp
331-354.
26. Patterson, J.H., "Case History:
Soil Venting as a Construction Safety-
Remediation Method for Development of
Contaminated Property", Proceedings of
Seminar for Ct. State Civil Engineers,
Session IV, Berlin, CT, Nov. 2-3, 1989.
382
-------
FIELD SCREENING PROCEDURES FOR DETERMINING THE,PRESENCE
OF VOLATILE ORGANIC COMPOUNDS IN SOIL
Alan B. Crockett and Mark S. DeHaan
Idaho National Engineering Laboratory, EG&G Idaho, Inc.
Idaho Falls, ID 83415
ABSTRACT
Many field screening procedures have been used
to detect the presence of volatile organic
compounds (VOC) in soils but almost none have
been documented and verified. Users of these
procedures have not really known whether their
objectives in screening were met. A reliable
VOC screening procedure could significantly
reduce the number of samples currently being
submitted to laboratories, thereby reducing
costs and improving site characterization. The
Environmental Protection Agency's Environmental
Monitoring Systems Laboratory in Las Vegas
(EMSL-LV) has therefore sponsored a research
effort to evaluate and improve headspace methods
for screening soils for VOC in the field. The
research involved comparing several extraction
procedures using soils from actual waste sites,
and determining the agitation and mixing
necessary to achieve equilibrium. Headspace was
analyzed using a relatively simple portable gas
chromatograph with a short column. The results
were variable and show that several procedures
should be attempted and the results evaluated
before selecting a screening procedure.
INTRODUCTION
A recent study by the Office of Technology
Assessment and the National Academy of
Sciences has indicated that the U.S.
Environmental Protection Agency (EPA) should be
collecting 10 times as many samples as is the
Work performed for the U.S. Environmental
Protection Agency under Agreement No. DE-SA07-
90ID012989, through the U.S. Department of
Energy, Contract No. DE-AC07-76ID01570.
current practice under the Resource Conservation
and Recovery Act (RCRA) and Superfund.
Considering that the approximately 80
laboratories in EPA's Contract Laboratory
Program (CLP) are already operating at full
capacity, that there will be no extension of
clean-up deadlines and no increase in funding,
there will have to be major changes in the
programs to increase efficiency. One way of
increasing efficiency is to reduce the number of
samples being analyzed under CLP protocols that
show no or only very low contamination levels.
At present, 80% and 90% of the samples submitted
to CLP laboratories for analysis of volatile and
semivolatile organics, respectively, fall in
this category (personal communication, Dave
Bottrell, EMSL-LV). One means of reducing the
numbers of such samples is to screen samples
prior to submission for CLP analyses. In
theory, 80% of the volatile samples being
submitted to CLP laboratories could be
eliminated or the CLP productive capacity could
be expanded by a factor of 5 if adequate
screening methods for VOC were available.
Specific examples would be the Department of
Energy's (DOE) Hanford Site and Savannah River
Plant, where about 126 soil or sediment samples
were collected and analyzed for volatile
organics as part of the DOE Environmental Survey
(1, 2). Of the samples collected, approximately
59% equaled or exceeded the Contract Required
Detection Limit (CRDL) for one or more volatile
compounds. Many of these samples were flagged
with a B, indicating blank contamination.
Only 29% of the samples had positive detections
above the CRDL and no B flag. Therefore, in
theory, somewhere between 41% and 71% of the
samples could have been rejected if screened in
383
-------
the field to CRDL limits. If it costs DOE
$250-5500 for each VOC analysis, the savings
could have been $13,000-$44,000 for analytical
services, which does not include costs for data
management and report preparation.
This research was designed to evaluate several
headspace methods for screening soil samples in
the field for the presence of VOC. The
objective was to determine whether or not to
send a sample to a CLP laboratory for gas
chromatography/mass spectrometry (GC/MS)
analysis.
BACKGROUND
The term "volatile organic compounds" refers to
a group of chemicals that readily pass from a
solid or liquid form to the gaseous phase
(volatile) and are composed of carbon-based
molecules (organic). Many VOC, toxins,
carcinogens, or mutagens, are hazardous to the
health of human and nonhuman organisms and are
common environmental contaminants. Volatile
organics (Table 1 for EPA list of VOC) are
particularly significant because they constitute
15 of the 25 most frequently identified
substances at 546 superfund sites (3).
A standard operating procedure for field
screening of VOC could decrease the demand for
CLP analyses, and at the same time result in
improved characterization of hazardous waste
sites. More samples could be collected and
screened, thus increasing the size of the area
characterized at a site, the intensity of that
characterization, and maximizing the usefulness
of those samples sent to the laboratory for
analysis. Near real time data would permit the
field sampler to redesign the sampling effort
while still in the field to characterize "hot
spots" and plumes. Other potential uses include
preliminary input for risk-assessment studies,
monitoring for efficacy of clean-up actions, and
research on the transport of VOC in soils (4).
Laboratory Analysis of Soils. EPA Method 8240
(5) for VOC in soils uses an inert gas to purge
(11 min.) VOC from a mixture of 5 g of soil and
5 mL of water into a Tenax trap. The VOC is
thermally desorbed and swept into the GC/MS for
analysis. The "high-level method" involves
extraction of 1 g of soil with 10 mL of methanol
(including spiking solution) by hand shaking for
2 Mention of specific products and/or
manufacturers in this document implies neither
endorsement or preference, nor disapproval by
the U.S. Government, any of its agencies, or
EG&G Idaho, Inc., of the use of a specific
product for any purpose.
2 min., transfer of an aliquot of the extract to
a purge and trap device, and analysis by GC/MS.
Laboratory Screening of Soils by Headspace
Analysis. EPA Method 3810 (5) allows rapid
screening of large numbers of samples. Ten g of
soil are placed in two 125-mL septum-seal glass
vials and one is spiked with calibration
standards. The two vials plus a third
containing only the standards are allowed to
equilibrate in a 90°C water bath for 1 hour.
Then, 2 mL of headspace is withdrawn and
injected into a GC. Detection limits for this
method may vary widely among samples because of
the large variability and complicated matrices
of waste samples. The sensitivity of the method
depends upon the equilibria of VOC between the
vapor and dissolved phases.
Field Screening. The most commonly used
procedures for field screening soil samples
involves analyzing headspace with an organic
vapor analyzer such as a Photovac TIP, HNU
PI-101 or Century Systems OVA-108. These
instruments respond to flame or photo ionizable
materials in air. They are very portable, easy
to use but do require relatively large sample
flow (0.25-1 L/min.) and have detection limits
in the lower ppm range. Very little data have
been published on their effectiveness as field
screening devices for determining the presence
of VOC in soils.
To improve detection limits and reduce sample
size, some field personnel have used portable
gas chromatographs which are up to 3 orders of
magnitude more sensitive and only require
headspace samples of 1-2 mL or even low jzL
quantities. Much more has been published on the
use of such devices for screening and analysis
of soil headspace in the field. Most screening
methods for soil samples are based on headspace
methods used for water samples.
Cheatham (6) effectively used a close support
laboratory to provide rapid assessment of
presence or absence of organic contamination.
The study used two portable HNU 301 GCs equipped
with a photo ionization detector (PID) connected
in series with a flame ionization detector, or
an electron capture detector. The method gave
unacceptable resolution of the indicator
compounds when no column packings were used.
So, direct headspace and purge and trap
techniques were studied using packed columns.
Samples were prepared by sealing 10 g of soil in
tared 100 ml serum bottles, which were placed in
a 90° water bath and allowed to equilibrate
approximately 1 hour, after which time the
headspace gas was analyzed. Samples were
submitted to a CLP laboratory for confirmation.
The results were of sufficient quality to
increase the accuracy of the site
384
-------
TABLE 1. CLP ANALYTICAL DATA FOR TEST SOILS
TARGET COMPOUNDS
Acetone
Benzene
Bromodichloromethane
Bromoform
Bromomethane
2-Butanone
Carbon dtsulfide
Carbon tetrachlorlde
Chlorobenzene
Chloroethane
Chloroform
Chloromethane
c 1 s-1 . 3-d1ch loropropene
Dlbromochloromethane
1 , 1-Oichloroethane
l,2-D1chloroethane
l,l-D1chloroethene
1,2-Dichloroethene
(total)
1 , 2-Dichloropropane
Ethyl benzene
2-Hexanone
Hethylene chloride
4-Nethyl-2-pentanone
Styrene
1,1,2,2-
Tetrachloroethane
Tetrachloroethene
Toluene
trans-l,2-d1chloroethene
trans-1,3-
dlch loropropene
1,1, 1-Tr Ichloroethane
1,1,2-Trlchloroethane
Trlchloroethene
Vinyl acetate
Vinyl chloride
Xylenes (total)
m-xylene
o-xylene
SAMPLE
KC804203
(Ug/kg)
0
66
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
24
0
170
0
0
0
0
0
0
84
0
0
0
0
0
0
31
870
0
0
KC804214
OigAg)
0
45
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
26
0
160
0
0
0
0
0
0
62
0
0
0
0
0
0
0
860
0
0
LA61201
(|ig/kg)
95
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
62
0
0
0
0
0
0
0
LA82301
(ligAg)
180
0
0
0
0
120
0
0
0
450
59
0
0
0
510
0
140
6200
0
3
0
190
0
0
0
0
47
0
0
160
9
33
0
24
0
0
0
LA82302
(Ug/kg)
260
0
0
0
0
440
0
0
0
300
320
0
0
0
1100
0
210
5800
0
0
0
160
0
0
0
0
72
0
0
1200
32
55
0
26
0
0
0
SR52702
(Ug/kg)
1200
220
0
0
0
220
17
0
0
0
0
0
0
0
0
0
0
10
0
200
0
65
0
0
0
40
1300
0
0
0
0
16
0
1100
380
0
0
Batch 1
(Ug/kg)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
0
0
0
37
0
0
0
0
0
0
0
6-4
(Ug/kg)
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
47
2
1
0
0
2
0
4
0
0
0
0
0
PARK LOT
(|lg/kg)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
138
36
0
0
0
3
0
30
0
0
0
0
680
characterization and improve plume mapping. The
quick turnaround time (-2 hours) allowed the
field staff to better understand the site and
select sampling locations using known data
rather than "best guess", thus optimizing the
limited project budget.
Clark, et al. (7) used a method developed by
Spittler in which 10 mL of soil was added to a
tared 40-mL VOA vial containing 20 mL of water
and 20 /*L of 2% mercuric chloride. In the lab,
the vials were warmed to room temperature,
shaken for 1 minute, and the headspace analyzed
using a Photovac model 10A10 GC. Headspace
sample volumes varied from 10 /iL to 1 mL. The
method was used to screen samples prior to GC/MS
analysis to avoid overloading the GC/MS and to
provide an indication of the presence or absence
385
-------
of organics. Use of the method has reduced
expensive GC/MS time and greatly reduced lost
analysis time in the laboratory. When screening
showed no organics, Clark thought it safe to
assume that no priority pollutants were present
at the ppb level, but that there might be
exceptions.
Griffith (8) and Spittler et al. (9) spiked soil
with three known VOC and placed aliquots of that
soil in water. The soil/water mixture,
contained in a glass vial, was placed in a water
bath and allowed to equilibrate. The air above
the water was then analyzed for VOC using a
Century Systems Model 128 GC portable organic
vapor analyzer. Results showed good recovery of
the compounds introduced to the soil specimens.
The method was tested under field conditions
with duplicate samples sent to an independent
laboratory. Comparison of field and laboratory
results showed good correlation for the aromatic
compounds under study.
OBJECTIVES
The objective of this research was to evaluate
eight headspace procedures for determining the
presence or absence of volatile organic
compounds (CLP list) in soil at less than ppm
detection limits. The intent was to be able to
screen soils for the presence of VOC and to
decide whether the samples need further analysis
by CLP methods or could be considered clean,
(i.e., contain insignificant levels of VOC.)
To achieve this objective, a variety of
information was required. First, an extraction
technique was needed that would maximize the
concentration of volatile compounds in the
headspace (more important for instruments
without low detection limits), in a reasonable
period of time, using practical field
procedures. This extraction procedure should be
fast (minutes), efficient (90% of equilibrium),
and easily accomplished under field conditions.
Secondly, a suitable analytical device was
required for detection of VOC in headspace. A
Photovac 10A10 field portable GC with a PID was
selected although other instrumentation could be
substituted.
APPROACH
This study was designed to build upon previous
studies. While a quantitative approach is
reasonable for water, many uncontrollable
factors make soil headspace analysis much more
complicated. This study was not designed to
investigate theory but to develop empirical
evidence on the utility and limitations of soil
headspace analysis using "naturally"
contaminated soils.
The first step was to compare a variety of soil
extraction methods to maximize headspace
concentration of volatiles when used on
"naturally" (as opposed to spiked) contaminated
soils. The second step was to determine the
"best" method of achieving 90% of headspace
equilibrium using the extraction method
identified in Step 1. The third step (not
conducted) would have been to use the screening
procedure in the field on samples that would
also be analyzed by GC/MS. The methods and
procedures are compared to each other, as well
as to a modified Method 3810 (5) in which the
10A10 was substituted for the prescribed
analytical instrumentation.
TABLE 2. EXPERIMENTAL TREATMENTS
1
2
3
4-6
7
8
1 g of soil in 29.5 mL of water
in a 40 mL VOA vial
5 g of soil in 27.5 mL of water
in a 40 mL VOA vial
20 g of soil in 20.0 mL of water
in a 40 mL VOA vial
same as above with a saturated
NaCl solution
10 g of soil in a 125 mL septum
cap bottle heated to 90*C in a
water bath for 1 h.
5 g of soil in 5 mL of methanol
followed of 0.6 mL of methanol to
29.5 mL of water in a VOA vial
STEP 1: COMPARISON OF SIX EXTRACTION METHODS TO
ACHIEVE MAXIMUM HEADSPACE CONCENTRATION. The
first tests were to determine maximum headspace
concentration after vigorous agitation using the
treatments listed in Table 2. The first
treatment was suggested by Spittler and has been
investigated by Griffith (8) and Spittler et al.
(9). A salt solution was selected for testing
since it is well known that adding salts to
water samples can increase headspace
concentration (salt also can be used as a
nonhazardous preservative). Treatments 1-6 and
8 were planned around 10 mL of headspace in
volatile organic analysis (VOA) vials.
Treatments 7 and 8 were included for comparison
to a standard laboratory screening procedure
(Method 3810) (5) and a solvent extraction
method (Method 8240, "high level" method which
uses methanol as the extracting solvent) (5).
386
-------
The importance of achieving maximum
concentrations in the headspace was to improve
the detection limit. While a GC/MS may not be
as sensitive as the Photovac 10A10 to some
compounds, a much larger sample can be actually
analyzed by using the Tenax sorbent in Method
8240. Essentially all the VOC contained in the
5 g sample are placed in the instrument using
Method 8240 while with headspace techniques,
only a small fraction of the VOC are injected.
Thus, to approach CLP detection limits for as
many compounds as possible, the objective was to
maximize static headspace concentration.
Obtaining fresh soil samples with good
analytical data was a problem, so archived
samples, known to have contained volatile
contaminants, were used. Although the samples
were quite old they still had detectable
concentrations of VOC, so their age did not
matter. The interest was in how efficient the
extraction procedures were and how long it took
to reach equilibrium (Step 2, below).
The headspace was sampled with standard syringes
(2-40 nt injections) and analyzed using the
Photovac 10A10 which had a 10.6 eV PID. Ultra
zero air was used as the carrier gas. While it
was not possible to specifically identify or
quantify the VOC present, it was the relative
concentration among the treatments that was
important.
The sources of the soil samples are given below
and the contaminants originally reported present
are listed in Table 1. Some of the VOC have
high ionization potentials and thus are
difficult to detect with the 10.6 eV PID.
Batch 1 provided by EPA and collected from
the Times Oil Superfund site in
Tacoma, Washington.
G-4 laboratory column sample provided by
EPA.
KC804203 subsurface soil sample collected
from under a pond at the DOE's
Kansas City Plant.
KC804214 very similar to the above.
LA61201 soil collected from a canyon wall at
DOE's Los Alamos
2LA61202 National Laboratory.
LA82301 sludge samples from an inactive
septic tank at DOE's Los
LA82302 Alamos National Laboratory.
Park Lot EPA provided sample from Tacoma,
Washington.
SR52702 subsurface soil sample collected
from an oil basin at DOE's Savannah
River Plant.
These bottles of soil were to be homogenized and
cored to obtain the required aliquot which would
be added to a tared vial containing the
extractant. This proved impractical and a
laboratory scoop was used to accomplish the
transfer as quickly as possible. Vials were
weighed again to obtain the actual amount of
soil. Tap water was used for the extractions
since headspace blanks showed no major
interference. Extractant volume was measured
using a graduated cylinder (most practical for
field work).
If it is assumed that the VOC are totally
desorbed from the soil in the extraction tests,
the 20 mL soil test should result in a 29.5
times higher concentration in water than 1 mL
soil (20 times more organic in about 1/3 less
water). While salts in an aqueous medium will
reduce the solubility of VOC in water thus
increasing headspace concentration in a two
phase system, their effect on headspace
concentration in a soil-water-air equilibrium is
not readily predictable.
To rapidly achieve static equilibrium, a Spex
Mill was used to violently agitate the soil
water mixture for 5 min. Headspace was analyzed
using a 6 in. SE-30 column at ambient
temperature. Because the treatment's
effectiveness was expected to vary greatly for
different soils and VOCs, it was deemed
preferable to test several soils once rather
than perform repeated tests on the same sample.
The limited volume of each sample was also a
factor. It was expected that violent shaking
for 5 min. would approximate equilibrium (or at
least be as rigorous as any practical field
extraction technique).
STEP 2: TIME TO REACH EQUILIBRIUM. The second
step was to minimize the time it takes for soil,
water, and headspace to reach near equilibrium
(goal was to obtain at least 90% of equilibrium
conditions in headspace) using practical field
procedures. "Naturally" contaminated soils were
again used instead of spiked soils because a
common criticism of spiking is "easy on, easy
off".
The four extraction procedures listed in Table 3
were compared with 5 min. of agitation on the
Spex Mill. The null hypothesis was that there
is no difference in extraction rate among the
treatments.
387
-------
TABLE 3. EXTRACTION PROCEDURES
1
2
3
4
Violent hand shaking for 1 min.
Agitation in a sonic bath for 5
min.
Agitation using a vibrator for 1
min.
Combinations of the above and
repeated analysis over time
STEP 3: COMPARISON TO CLP PROCEDURES. The
final evaluation of the selected extraction
procedure identified in Steps 1 and 2 was to be
a field trial of the recommended procedures.
This would involve making a decision as to
whether volatiles were present or not and then
evaluating those decisions based on laboratory
analytical data (ideally using an improved
Method 8240 where 5 g samples were sealed in the
field and never opened in the laboratory). This
step still needs to be conducted.
Water and NaCl blanks were also prepared and
treated as samples. A benzene in water standard
was run as a retention time and instrument
response check. No significant attempts were
made to identify contaminants or quantify
concentrations. The number of samples analyzed
varied, but ideally, duplicate samples with
duplicate injections were used. Peak height
data were measured and recorded by hand. The
raw data were normalized before statistical
analysis to eliminate differences in actual
versus planned soil weight, attenuation setting,
and volume of sample injected into the GC. An
example of a normalized data set is presented in
Table 4 which shows: sample number, treatment
(D- duplicate sample), nominal amounts of soil
and extractant used, and normalized peak heights
with replicate injection data.
TABLE 4. NORMALIZED PEAK HEIGHT DATA FOR ANALYZED SAMPLES
TREATMENT.
NOMINAL AMOUNT
OF SOIL (g).
and
EXTRACT (mLl
Treatment
Ambient
5.0 0
5.0 0 0
Treatment H20
5.0 27.5
5.0 27.50
1.0 29.5
1.0 29.50
Treatment
Heated Bot
10.0 0
10.0 0 0
Treatment
Heated Wet
10.0 10
10.0 10 D
Treatment
Met ha no 1
5.0 5.0
5.0 5,0 D
Treatment NaCl
5.0 27.5
5.0 27.50
1.0 29.5
1.0 29. 50
NORMALIZED HEIGHT. PARK LOT SAMPLE
(cm)
PEAK 1
0.78
0.85
0.69
0.77
0.17
0.21
20.16
20.73
14.59
9.70
1.23
0.53
0.91
0.79
0.21
0.29
PEAK 1R
0.91
0.70
0.00
0.25
21.09
26.97
18.16
14.71
1.07
0.00
0.61
l.OZ
0.00
0.00
PEAK 2
0.35
0.39
0.25
0.25
0.35
0.18
7.44
8.23
11.02
3.44
0.61
0.42
0.40
0.31
0.18
0.41
PEAK 2R
0.69
0.25
0.00
0.69
12.40
12.21
13.62
7.51
0.61
0.00
0.30
0.40
0.00
0.00
PEAK 3
1.02
0.89
1.23
1.99
0.42
0.46
21.71
23.57
20.75
12.52
4.70
5.56
1.28
1.67
0.48
0.53
PEAK 3R
2.26
1.75
0.00
0.71
23.26
29.81
27.24
18.78
4.09
0.00
0.99
2.01
0.00
0.00
PEAK 4
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.61
0.53
0.00
0.00
0.00
0.00
PEAK 4R
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.51
0.00
0.00
0.00
0.00
0.00
PEAK 5
2.04
2.09
1.79
1.65
1.83
1.49
10.85
12.49
3.89
3.76
1.84
2.12
1.78
0.94
1.66
1.57
PEAK 5R
1.86
1.44
0.00
1.59
10.85
15.05
4.86
5.01
1.58
0.00
0.99
1.11
0.00
0.00
388
-------
RESULTS
Initial work showed major problems with
conducting the standard laboratory procedures
(treatments 7 and 8, heated headspace and
methanol extraction). After some early eluting
peaks, a massive tailing peak was produced which
masked all other data. It was thought that the
level of methanol in the headspace was
interfering with the analysis and that condensed
water may have produced a similar effect.
Therefore, early experiments eliminated these
treatments. Later, the procedures were tried
again with better results in some cases.
Statistical Methods
The normalized data were analyzed with analysis
of variance (ANOVA) techniques to test for
statistically significant treatment effects (a
treatment is a specific extracting method and
soil amount combination). In many cases the
effect of the amount of soil extracted was
acceptably linear, but in some cases the
relationship did not appear linear. Therefore
the results were also analyzed with the soil
amount log-transformed to better linearize the
response. It was found that the
log-transformation did not change the results.
Because the treatments might have different
effects when used on the different soil samples,
the analyses were run separately for each
sample.
For each experimental run with a given
treatment, from four to six peaks were measured
and used. There were high correlations among
the peak heights for a given treatment, i.e. if
a treatment affected peak 1 height, it similarly
affected all the other peaks. To take advantage
of this correlation, a Multivariate Analysis of
Variance (MANOVA) was used to analyze the
treatment effects. The soil amount was treated
as a covariate in the analysis. The analyses
were done using the GLM procedure of the SAS*
(10) statistical software package.
For the experiments using soil types KC804203,
KC804214, and LA61201, there were no true
replicates analyzed, therefore it was impossible
to directly estimate the experimental error or
natural variation. Instead, subsampling or
measurement error was used as a lower bound
estimate of experimental error, and all
analytical results for these samples should be
viewed conservatively.
Statistical Results of Extraction Tests
Sample SR52702
Sample SR52702 was a subsurface soil sample
collected from an oil basin at a depth of 6-12
in. Table 1 shows that the sample contained 12
VOC when analyzed by GC/MS for the DOE
Environmental Survey. In appearance, the soil
was heterogeneous with black specks and rocks.
Because of concern for sampling error, the
contents of the sample jar were dumped into a
beaker, the rocks removed and the aggregate
broken up as much as possible to improve
homogeneity prior to subsampling. Some VOC were
certainly lost in this process but there were
enough left to use for testing.
The MANOVA involved comparison of treatments 1-6
(treatments 3 and 6 used 10 g instead of the
planned 20 g of soil due to limited amount of
soil available) and showed that the NaCl and
water treatments did not result in significantly
different peak heights (Wilks' Lambda F=2.2,
p=0.23) . However, the amount of soil (1, 5,
and 10 g) had a statistically significant effect
on the peak heights, with larger soil amounts
resulting in higher peaks (Wilks' Lambda F=12.8,
p=0.01). The height of the 10 g peaks were from
1.6 to 3.4 times higher than the 1-g water
extract peaks.
Sample Batch 1
The Batch 1 sample was collected by EMSL-LV
personnel from the Times Oil superfund site in
mid March, 1989. The sample was very dense,
black, and contained small pebbles. The
chromatograms showed 2 large peaks of very early
eluting compounds (possibly including vinyl
chloride). Because they were so large and
early, they could not be separated and measured
accurately with the 6-in. column and only the
later 4 peaks were analyzed.
The analysis of treatments 1-6 showed that the
NaCl and water extractions did not result in
significantly different peak heights (Wilks'
Lambda F«3.2, p-0.14). The amount of soil used
had a statistically significant effect on peak
heights, with larger soil amounts resulting in
higher measured peaks (Wilks' Lambda F=80.6,
p=0.0004). The 20-g samples produced peaks 15
to 48 times larger than the 1-g water extract
samples. It is interesting to note that the
effect the amount of soil had depended on the
extractant used.
3 The Wilks' Lambda statistic tests
whether the treatments have significantly
different effects on the measured peak heights.
The p-values given are the probability that the
observed results are simply due to chance, and
not due to treatment effect.
389
-------
Sample KC804203
The experiment included all eight treatments
(data for the 20-g NaCl treatment was lost) but
the methanol extract was diluted 100:1 instead
of 50:1. The analysis (5 peaks) showed that the
different extraction methods resulted in
significantly different peak heights (Milks'
Lambda F=5.6, p=0.003). The water extract gave
consistently higher peaks, with NaCl second
highest. The methanol extracting and the heated
dry treatment gave generally lower peaks than
the other two treatments.
The amount of soil used also had a statistically
significant effect on the peak heights (Wilks'
Lambda F=157, p^O.OOOl). For all the peaks, the
greater the amount of soil used, the higher the
measured peak, i.e. there was a positive
correlation between amount of soil used and the
peak height. For the water extraction, the 20-g
extract increased peak heights from 2 to 6 times
compared with the 1-g extracts. For the five
peaks measured, methanol produced from 0.2 to
0.5 times the response compared to the 1-g water
extract. Methanol appears to be relatively more
efficient in recovering the late eluting peaks
(higher boiling compounds). For the heated dry
bottle, response varied from 0.5 to 0.8 times
the comparable peaks height for the 1-g water
sample.
Sample KC8Q4214
This experiment included all 8 treatments but
there were no replicate injections or duplicate
samples. Therefore the statistical tests were
not very powerful or accurate. Additionally,
the results for the 5- and 20-g water extraction
appear to be outliers (peak heights for both
treatments were very similar, contrary to the
NaCl data). The analysis (6 peaks) using all
the data showed that the extraction methods did
not result in significantly different peak
heights (p > 0.3). Also, the amount of soil
used did not seem to affect peak heights (p >
0.1). However, visual interpretation of the
data, excluding the outliers, showed results
very similar to sample KC804203.
Sample LA6120I
The experiment included all 8 treatments but
about 20 g of soil was used for the heated
bottle (Treatment 7) instead of the planned 10 g
(the normalizing program therefore cut peak
heights in half for data analysis). The
analysis of all 6 peaks showed a fairly
significant extraction treatment effect (Wilks'
Lambda F=11.9, p=0.03). The results were mixed
though, with water giving very much higher peak
2 readings, and NaCl giving somewhat higher peak
6 measurements. Overall, water and NaCl
extracts produced comparable peak heights for
the same soil amount, methanol produced smaller
early eluting peaks than the 1-g water or NaCl
extractions, but the last peak was 2.9 times
larger. The greatest peak heights were achieved
using the heated dry bottle (11-66 times than
the 1-g water extract).
The amount of soil used also had a very
significant effect on the peak heights (Milks'
Lambda F=2678, p=0.0001). For all the peaks,
the greater the amount of soil used, the higher
the measured peak. Peak heights were from 4-12
times higher using 20 g of soil than 1 g.
Sample LA82301
The experiment involved only treatments 1,2,4,
and 5 (1- and 5-g extractions using water and
NaCl) because of the small soil volume
available. The analysis of 7 peaks showed that
the different extraction methods resulted in
significantly different peak heights (Wilks'
Lambda F=240.5, p=0.05). The NaCl extraction
gave consistently higher peaks than the water
(up to about 3 times greater peak height).
The amount of soil used also had a very
significant effect on the peak heights (Milks'
Lambda F=3304, p=0.01). The 5-g treatments
generally produced 3-5 times higher peaks than
the 1-g water treatment.
Sample G-4
The analysis of data (4 peaks) from treatments
1-6 showed no significant difference in peak
heights for the water and NaCl extractions
(Wilks' Lambda F=2.1, p=0.2). The amount of
soil used had a statistically significant effect
on peak heights for the water treatments, with
larger soil amounts resulting in higher measured
peaks (Wilks' Lambda F=31.7, p=0.001). The 20-g
water extraction peaks were 2-6 times larger
than the 1-g peaks. It was interesting that
this soil effect was not apparent with the NaCl
extractant. These data, however, are suspect
since they had to be reanalyzed several days
after initial extraction due to analytical
problems.
Sample PARK LOT
The experiment included treatments 1, 2, 4, 5
7, and 8 (the 20-g water and NaCl treatments
were omitted due to limited soil volume
available) plus a modification of the heated
bottle (Treatment 7) technique (10 ml of water
added to bottle). Additionally, a 5-g sample
was placed in a VOA vial at ambient room
temperature. The analysis of data (Table 4)
from treatments 1-6 showed no statistically
significant differences between the water and
390
-------
NaCl extractions (Wilks' Lambda F=0.95, p=0.6).
However, the heated bottle (treatment 7, listed
as "Heated Bot" in Table 4) produced the
greatest peak heights followed by the same
treatment with 10 ml of added water ("Heated
Wet" in Table 4), and methanol. Interestingly,
the VGA vial with 5 g of dry soil at ambient
temperature ("Ambient B" in Table 4) provided
greater peak heights than the 1-g water and NaCl
extractions.
The amount of soil (1 versus 5 g) did somewhat
affect the measured peak heights, with larger
soil amounts resulting in higher peaks (Milks'
Lambda F-90, p=0.002).
Statistical Results of Equilibrum Tests
Sample 2LA61201
Step 2 experiments for this soil included the
treatments in Table 3. Additionally, since Step
1 data on treatments 7 and 8 were limited
(heated bottle and methanol) those treatments
were included to supplement Step 1 data.
Sample 2LA61201 was extracted using hand
shaking, the Spex Mill, vibration and sonication
as well as the same treatments over a period of
2+ hours. Statistical analysis of the data
showed there were some differences in the
treatment effects (Wilks' Lambda F=4.2,
p<0.001), but only for 2 of the 5 peaks
analyzed. To help compare treatment effects,
both Duncan's multiple range test and Scheffe's
multiple comparison were calculated at the 0.05
significance level. While there were some
differences, no treatment clearly emerged as
superior. Although not tested statistically, it
appears that initial extraction using the Spex
Mill is more efficient since peak heights
increased less over time that with the other
treatments. The Spex Mill initially provided
the highest average peak height for all 5 peaks;
up to 2-3 times higher in some cases. After a
couple of hours, all treatments show very
similar results.
The heated dry bottle extraction produced much
greater peak heights than the reference method
of 1 g of soil extracted by the Spex Mill. The
response was about 8-34 times greater for the 5
peaks analyzed. A methanol extraction treatment
was also conducted but the results were unusable
due to interferences from the methanol.
Sample LA82302
The Step 2 treatments for this soil included
hand shaking for 1 min., Spex Mill shaking for
5 min., sonication for 5 min. after hand shaking
for 1 min., and vibration treatment for 1 min.
after hand shaking for 1 min. The samples from
the hand shaking and sonication treatments were
shaken again by hand after 3 hours and
reanalysed. The MANOVA results indicated that
there were some significant differences (Wilks'
Lambda F«2.9, p=0.04).
The individual ANOVA analysis for 3 of the 4
peaks analyzed showed no statistically
significant differences in treatment effects in
spite of some peaks being twice the size of
those in other treatments. For one peak, the
initial Spex Mill treatment was superior all
other treatments (p=0.006). For all peaks, the
vibration treatment produced the lowest peaks.
No other obvious patterns appeared that would
suggest one treatment was better than the
others.
Both a heated bottle and methanol treatment were
run at the same time to supplement the Step 1
data. The methanol results showed gross
interferences from the methanol even though a
fresh bottle of HPLC grade methanol was used.
The heated bottle treatment produced
significantly greater peak heights for all peaks
than any of the treatments discussed above
(p<.01). Compared to the initial hand and Spex
Mill extractions, the heated bottle method
produced 5-13 times higher peaks than the
initial 1-g soil sample extracted with water
using the Spex Mill.
CONCLUSIONS
The comparison of six treatment combinations of
soil amount and extracting solution (water or
NaCl), showed variable results. While NaCl
extractions produced significantly larger peaks
for one test sample, that was the only data
demonstrating clear superiority. Even then, the
differences were only a factor of about 3. The
conclusion is that water is generally a superior
extractant to the saturated NaCl solution for
soil headspace analysis.
The tests of the effect of extracted soil
amounts, clearly showed that larger quantities
of soil extracted into the same volume of
headspace produces higher headspace
concentrations. For the 5- and 20-g soil
samples, one would expect a 5 and 29 fold
increase in headspace concentration over a 1-g
sample if all soil contaminants were transferred
to the water (ideal but not possible). Table 5
shows the effect of sample size on VOC headspace
concentration, by sample. The 20 g sample
produced from 3.6 to 24 times greater response
than the 1 g sample extracted with water using
the Spex Mill. Overall, the increase was about
a factor of 6.5 improvement calculated using a
geometric mean. The 5 g soil samples provided
between 1.8 and 7.8 times greater response with
a geometric mean of 3.3. While increased soil
391
-------
amounts do increase headspace concentration, the
increase is usually not as great as
theoretically possible.
The methanol extraction data shown in Table 6
shows that methanol is sometimes superior
(factor of 4.2) and sometimes inferior (factor
of 0.3) to the water extraction using 1 g of
soil and the Spex Mill. Overall, there was
little reason to select methanol over water
(geometric mean = 1.1). The disadvantages of
methanol extraction for field screening are that
it involves the transport, use, and disposal of
a hazardous chemical, and requires the
additional field steps of quantitatively
transferring an aliquot to a VOA vial after
settling. Methanol may also interfere with
analysis on some instruments and it may be
difficult to obtain clear supernatant from some
samples. The advantages are that the extraction
step is quick, the sample in methanol should be
relatively stable, composite samples can be
collected, the same sample used for field
screening can be sent to a laboratory for
analysis, and the extraction method is based on
a standard EPA analytical procedure.
Dry heated head space analysis method using 10 g
of soil was sometimes far superior to the 1-g
soil/Spex Mill treatment for maximizing peak
heights. As shown in Table 6 the relative
headspace concentration for the heated bottle
treatment compared to the 1 g water extraction
varied from 0.7 to 48 times greater, with a
geometric mean of 6.8. The disadvantages of the
heated bottle approach are the time requirement
of heating for 1 hour, the need for a water bath
(requiring electrical power) and possible
analytical problems related to condensed water
in the analytical device. Some GC columns do
not work well with saturated vapor samples.
For extracting samples in the field, violent
shaking such as provided by the Spex Mill is
efficient but not possible without a power
supply. Hand shaking for 1 min. seems slightly
inferior initially, but with time, headspace
concentration becomes the same as with Spex Mill
extraction.
The overall recommendation is that several
procedures should be evaluated and compared
using site specific contaminated soils. A
standard operating procedure that helps a user
select the best screening procedure for the
intended use should be developed. Also,
documentation on the effectiveness of screening
procedures versus standard quantitative methods
is needed so that screening effectiveness can be
evaluated including a rough estimation of
detection limits.
TABLE 5. RELATIVE HEADSPACE CONCENTRATION OF VOC
for 20, 5, and 1 g soil samples extracted with water
Sample
Size
20 g
5 g
i g
Batch 1
24
7.8
1
G-4
3.6
3.5
1
KC804203
3.6
1.8
1
KC804214
5.1
5.6
1
LA61201
7.2
2.9
1
LA82301
3.8
1
Park Lot
__
2.3
1
SR52702
2.4*
1.8
1
Arithmetic
Mean
8.7
3.7
1
Geometric
Mean
6.5
3.3
1
*Sample size was 10 g., not included in means.
TABLE 6. RELATIVE HEADSPACE CONCENTRATION OF VOC for Heated Bottle and
Methanol Treatment Compared to 16 Soil/Sample Extracted Water
SAMPLE
Heated
Bottle
Methanol
1 9
soil/H20
KC804203
0.7
0.3
1
KC804214
2.0
1.6
1
LA61201
28
0.8
1
2LA61201
26*
—
1
LA82302
7.6
--
1
PARK LOT
48
4.2
1
Arithmetic
Mean
17
1.7
1
Geometric
Mean
6.8
1.1
1
*Not included in means.
392
-------
REFERENCES 7.
1. EG&G Idaho. Hanford Site Sampling and
Analysis Data Document, Vol. 1-4,
EGG-ES-7953, Rough Draft, 1988.
2. EG&G Idaho. Savannah River Sampling and
Analysis Data Document, Vol. 1-4, 8.
EG&G-ES-8042, Draft, 1989.
3. McCoy & Assoc., Hazardous Waste
Consultant, Vol. 3, No. 2, 1985.
4. Barth, Del. Literature Review on 9.
Preparation of Soil Samples for OC
Analysis, Environmental Research Center,
University of Nevada, Las Vegas, 1988.
5. U.S. EPA. Test Methods for Evaluating
Solid Wastes, SW-846, Vol IB, Office of
Solid Waste and Emergency Response,
Washington, D. C., 1986. 10.
6. Cheatham, Richard, Jeffrey Benson, Jeralyn
Guthrie, William Berning, and Roger Olsen,
"Rapid, Cost-Effective GC Screening for
Chlorinated Pesticides and Volatile
Organics at CERCLA Sites" Monograph
Series Screening Techniques, HMCRI 9300
Columbia Boulevard, Silver Spring, MD, p.
86, Undated.
Clark, Arthur, Moira Lataille, and Edward
Taylor, "The use of a Portable PID Gas
Chromatograph for Rapid Screening of
Samples for Purgeable Organic Compounds in
the Field and in the Lab." US EPA Region
I Laboratory, 1983, unpublished.
Griffith, J. Tyler. A New Method for
Field Analysis of Soils Contaminated with
Aromatic Hydrocarbon Compounds, M.S.
Thesis, University of Connecticut, Stores,
1988.
Spittler, Thomas, Mary Cuzzupe, and J.
Tyler Griffith, "A Field Method for
Determination of Volatile Organics in Soil
Samples", First International Symposium,
Field Screening Methods for Hazardous
Waste Site Investigations, Oct 11-13,
1988.
SAS Institute, SAS/STAT User's Guide,
Release 6.03 Edition, SAS Institute, Cary,
North Carolina, 1988.
DISCUSSION
RANDY GOLDING: What is the correlation between what might be perceived
as standard accepted analytical practices and the results you get from these
screening methods. What's their predictive value and which one is the best?
ALAN CROCKETT: I don't real ly have a good answer for you, I'm afraid. How
predictive are they? That was to be step three which was never completed. I don't
know right now how well the CLP data would correlate with the field screening
data. That's why I'm asking people right now, who have the interest in following
these procedures, to analyze their samples two different ways and publish the
results.
THOMAS SPITTLER: Justone brief response to that, because we had used this
•water extraction for volatiles in soil for quite a few years in our own region, and
I've been in touch with other people who have been doing it, particularly some
of the people doing research at the University of Connecticut. They have found
extremely good reproducibility and very high sensitivity in extracting volatiles
out of soil. The only thing is most of this was done with synthetic soil samples.
There's a group at the Cold Regions Research Lab in Hanover, New Hampshire,
who have also done spiking of soil samples, and had done some very nice work
on extracting volatiles from soil samples. Tom Jenkins is one of the two. I think
he's got some very interesting insights on this problem of volatiles in soil
analysis. It's a major issue because so many people are out there digging up tanks,
trying to comply with state regulations on how much is too much contamination
in soil. An incredible amount of lousy data is coming out of samples collected,
shipped off to laboratories, and reported back months later only to find that what
was obviously there when the sample was collected is no longer there when the
sample is analyzed. That's a lesson in biodegradation and vapor loss, among
other things.
393
-------
COMPARISON OF FIELD HEADSPACE VERSUS FIELD SOIL GAS ANALYSIS VERSUS STANDARD
METHOD ANALYSIS OF VOLATILE PETROLEUM HYDROCARBONS IN WATER AND SOIL
RANDY D. GOLDING MARTY FAVERO GLEP>
TRACER RESEARCH CORPORATION
THOMPSON
ABSTRACT
Twelve sites in the state of Iowa were evaluated for hydrocarbon
contamination associated with the use of existing underground
storage facilities. Using driven probe technology, soil gas,
groundwater and/or soil samples were taken from each sampling
location. Each soil gas, groundwater and soil sample was
analyzed on location using a field mobilized gas chromatograph.
Each groundwater and soil sample was also analyzed by a
contracted laboratory using the appropriate standard method.
Correlations between the various analytical approaches were
examined.
The correlation between field analytical results and the laboratory
analytical results was 0.87 over four orders of magnitude for
twenty-five samples. The correlation for toluene in soil gas
samples versus toluene in soil samples was 0.81 over five orders
of magnitude.
INTRODUCTION
In order to examine the utility of soil gas investigations and field
analyses in evaluating the contamination at underground storage
tank (UST) sites, temporary approval was granted to Tracer
Research Corporation (TRC) to apply soil gas methods at several
UST sites in Iowa. The results of the soil gas investigations were
to be compared to analyses of soil and water samples collected at
the same time. The results of this comparative study were to be
used to establish appropriate action criteria for soil gas
investigations used at UST site audits for insurance purposes.
BACKGROUND
The background section consists of two parts. First a brief
description or definition of soil gas methodology and second a
discussion of how the instrument that makes the total petroleum
measurement (TPHC) works. This background is very important
in understanding the first part of the results section.
Shallow Soil Gas Investigation - Methodology
Shallow soil gas investigation refers to a method developed by
TRC for investigating underground contamination from volatile
organic chemicals (VOCs) such as industrial solvents, cleaning
fluids and petroleum products by looking for their vapors in the
shallow soil gas. The method involves pumping a small amount
of soil gas out of the ground through a hollow probe driven into
the ground and analyzing the gas for the presence of volatile
contaminants. The presence of VOCs in shallow soil gas indicates
the observed compounds may either be in the ground near the
probe or in groundwater below the probe. The soil gas technology
is most effective in mapping low molecular weight halogenated
solvent chemicals and petroleum hydrocarbons possessing high
vapor pressures and low aqueous solubilities. These compounds
readily partition out of the groundwater and into die soil gas as a
result of their high gas/liquid partitioning coefficients. Once in
the soil gas, VOCs diffuse vertically and horizontally through die
soil to die ground surface where they dissipate into die
atmosphere. The contamination acts as a source and the above
ground atmosphere acts as a sink, and typically a concentration
gradient develops between the two. The concentration gradient
in soil gas between the source and ground surface may be locally
distorted by hydrologic and geologic anomalies; however, soil gas
mapping generally remains effective because distribution of the
contamination is usually broader in area! extent than die local
geologic barriers and is defined using a large data base. The
presence of geologic obstructions on a small scale tends to create
anomalies in the soil gas-groundwater correlation, but generally
does not obscure the broader area! picture of the contaminant
distribution.
Soil gas contaminant mapping helps to reduce the time and cost
required to delineate underground contamination by volatile
contaminants. The soil gas investigation does this by outlining the
general area! extent of contamination.
How the Hydrocarbon Measurement is Made
To illustrate some of the advantages of soil gas hydrocarbon
measurements, representative chromatograms produced during
the comparative study at selected Iowa UST sites are presented
with a brief explanation.
A chromatogram is a graph of the analytical signal output by the
chromatograph. When a sample is analyzed using a gas
chromatograph, it is injected into a tube through which a gas is
flowing towards a detector. The sample is carried toward the
detector by the flowing gas stream. Between the injection point
and the detector is a long tube called a column that contains a
powder or fluid that absorbs substances like gasoline vapors.
Gasoline is a mixture of many different substances that are very
similar. The column is most absorptive to substances with high
boiling points such as xylenes (280 degrees F) and less absorptive
to substances with low boiling points such as methane (-260
(below zero) degrees F). Thus low boiling or very volatile
substances like methane flow rapidly through die column and high
boiling or not so volatile substances like xylenes are retarded by
the column and flow through die column more slowly.
When a substance exits from the column and is carried by the gas
stream into the detector, it is burned in the flame (flame ionization
detector) and an electrode senses an increase in combustible
395
-------
substances in the (lame and the result is an increase in voltage at
the signal output.
The chromatogram is a plot of this voltage versus time. When
nothing combustible is entering the detector the recorder draws
a straight line (baseline) along the left side of the page. When a
combustible substance such as methane enters the detector, the
increase in voltage from the chromatograph causes the recorder
pen to move until the substance is completely burned. The
recorder pen then returns and stays at the baseline until the next
combustible substance enters the detector. This triangular shaped
deflection is called a peak. The point in time that the peak occurs
indicates what kind of substance it is and the time is printed on
the chromatogram next to the peak. The distance that the pen
moves or how tall the peak is indicates how much of the substance
entered the detector.
INJECT 86/98/9813:86:13 STORED TO BIN » 85
/ ^
•Wane
Benzene1
"^ XyleneA
Figure 1. Standard mixture for calibration.
Figure 1 is a chromatogram of a mixture of methane, benzene,
toluene, and xylene, all substances found in gasoline and most
other petroleum fuels. Note the exit lime for toluene is 1.01
minutes.
CtWKl A INJECT 86/88/98 H:3B:89 REPUWtJ mi BIN « 9<
n i
. 1.56
Figure 2. Gasoline vapors from a gasoline tank. NOTE; Most of the
components of gasoline exit before toluene.
Figure 2 is a chromatogram of gasoline vapors taken from a
gasoline tank.
INJECT 86/86V»» H:I4:33 REPUWED FICM BIN I 93
— 5.32
Figure 3 is a chromatogram of kerosene vapors taken from a
kerosene tank.
•JECT 06X08/90 13:35:47 REPLAYED FROM
AZ 1
1 .8
.54
15 .89
19 .01
Figure J. Kerosene vapors from a kerosene tank. NOTE: Most of the
components exil after toluene. Kerosene is not as volatile as gasoline.
Figure •«. Diesel vapors from a dlesel tank. NOTE: Almost all
components exil after toluene. However, there are some volatile
substances in dieseL
Figure 4 is a chromatogram of diesel vapors taken from a diesel
tank. Only a small portion of the constituents of the diesel vapors
exit before toluene.
Kerosene is more volatile than diese 1 and gasoline is more volatile
than kerosene. These figures illustrate that the more volatile
substances exit the column first and the less volatile substances
exit the column later. They also serve to illustrate that diesel
contains a substantial amount of relatively volatile substances.
PROCEDURES
The following describes in a general way the procedures used to
acquire the data in this study.
Equipment
Tracer Research Corporation utilized a one-ton analytical field
van that is equipped with one gas chromatograph and two
computing integrators. In addition, the van has two built-in
gasoline powered generators that provide the electrical power
(110 volts AC) to operate all of the gas chromatographic
instruments and field equipment. A specialized hydraulic
mechanism consisting of two cylinders and a set of jaws was used
to drive and withdraw the sampling probes. A hydraulic hammer
was used to assist in driving probes past cobbles and through
unusually hard soil.
396
-------
Soil Gas Sampling Procedures
Sampling probes consist of 7-14 foot lengths of 3/4 inch diameter
hollow steel pipe that are fitted with detachable drive tips. Once
inserted to the desired depth, the above-ground end of the
sampling probes were fitted with an aluminum reducer and a
length of polyethylene tubing leading to a vacuum pump. Gas flow
is monitored by a vacuum gauge to insure that an adequate flow
is obtained.
To adequately purge the volume of air widiin the probe, 2 to 5
liters of gas is evacuated with a vacuum pump. During the soil gas
evacuation, samples are collected in a glass syringe by inserting a
syringe needle through a silicone rubber segment in toe
evacuation line and down into the steel probe. Ten milliliters of
gas are collected for immediate analysis in the TRC analytical
Geld van. Soil gas is sub-sampled in volumes ranging from 1 uL
to 2 mL, depending on the VOC concentration at any particular
location.
Soil Sampling Procedures
Soil samples were collected by pushing the soil gas probes into
the ground without the detachable drive point, thus allowing soil
to accumulate in the probe. The soil was removed by inserting a
one-half inch diameter pipe through the probe to push the soil
out. Approximately 10 grams of soil and 10 mL of water were
placed in a 40 mL teflon sealed VO A bottle leaving approximately
20 mL of headspace. Each VO A was then shaken vigorously for
30 seconds before the headspace was analyzed. This allows for
the desorption of volatile compounds into the headspace of the
vial. Headspace vapor is sub-sampled in volumes ranging from 1
uLto2mL.
Groundwater Sampling Procedures
Groundwater samples were collected by driving the hollow
probes with detachable drive points below the water table. Once
at the desired depth the probe was withdrawn several inches to
permit water inflow into the resulting hole. Once inserted into
the ground, the above-ground end of the sampling probes were
fitted with a vacuum adaptor (metal reducer) and a length of
polyethylene tubing leading to a vacuum pump. A vacuum of up
to 24 inches of mercury was applied to the interior of the probe
and open hole for 1 to 15 minutes or until the water was drawn up
the probe. The water thus accumulated was then removed by
drawing a vacuum on a 1/4 inch polyethylene tube inserted down
the probe to the bottom of the open hole. Loss of volatile
compounds by evaporation is minimized when water is induced
to flow into the very narrow hole, because it can be sampled with
little exposure to air. The polyethylene tubing was used once and
then discarded to avoid cross-contamination.
Groundwater samples were collected in 40 mL VOC vials that are
filled to exclude any air and then capped with Teflon-lined septum
caps. Groundwater samples were analyzed by injecting
headspace in the sample container created by decanting off
approximately half of the liquid in die bottle. Headspace analysis
is the preferred technique when a large number of water samples
are to be performed daily. The method is more time efficient for
the measurement of volatile organics than direct injection.
Depending upon the partitioning coefficient of a given
compound, the headspace analysis technique can also yield
greater sensitivity than the direct injection technique.
Field Analytical Procedures
A Varian 3300 gas chromatograph, equipped with a flame
ionization detector (FID), was used for the soil gas, soil, and
groundwater analyses. Compounds were separated by a 6" by 178"
OD packed column widi O v-101 as die stationary phase at 100°F
in a temperature controlled oven. Nitrogen was used as the
carrier gas.
Hydrocarbon compounds detected in soil gas, soil, and
groundwater were identified by chromatographic retention time.
Quantification of compounds was achieved by comparison of the
detector response of the sample with the response measured for
calibration standards (external standardization). Instrument
calibration checks were run periodically throughout the day and
system blanks were run at the beginning of the day to check for
contamination in the soil gas sampling equipment. Air samples
were also routinely analyzed to check for background levels in the
atmosphere.
The GC was calibrated for soil and groundwater headspace
analysis by decanting 10 to 20 mL off of the known aqueous
standard so as to leave approximately die same amount of
headspace dial is in die field samples. The bottle is then resealed
and shaken vigorously for 30 seconds. An analysis of the
headspace in the vial determines the Response Factor (RF) which
is then used to estimate soil or groundwater concentrations.
Detection limits for the compounds of interest are a function of
the injection volume as well as the detector sensitivity for
individual compounds. Thus, the detection limit varies with die
sample size. Generally, the larger the injection size the greater
the sensitivity. However, peaks for compounds of interest must
be kept within die linear range of the analytical equipment. If any
compound has a high concentration, it is necessary to use small
injections, and in some cases to dilute the sample to keep it within
linear range. This may cause decreased detection limits for other
compounds in die analyses.
The detection limits for the selected compounds vary depending
on the conditions of the measurement, in particular, the sample
size. If any component being analyzed is not detected, the
detection limit for that compound in that analysis is given as a less
than" value (e.g. ug/L). Detection limits obtained from GC
analyses are calculated from the current response factor, the
sample size, and the estimated minimum peak size (area) that
would have been visible under the conditions of the measurement.
Laboratory Analytical Procedures
Groundwater samples were analyzed using analytical protocols
outlined hi EPA methods 5030 and 8015. A purge and trap step
is used to strip the hydrocarbons out of die water.
Soil samples were analyzed by a method stipulated by the Iowa
Department of Natural Resources referred to as OA-1. The
method is substantially derived from EPA methods 5030 and
8015. Methanol is used to extract hydrocarbons from the soil.
The Methanol extract is then diluted at least 25 to 1 in reagent
water. The water is then analyzed in essentially the same manner
as die groundwater samples.
Quality Assurance/Quality Control Procedures
Tracer Research Corporation's normal quality assurance
procedures were followed in order to prevent any
cross-contamination of soil gas, soil, and groundwater samples.
• Steel probes were used only once during the day and then
washed with high pressure soap and hot water spray or
steam-cleaned to eliminate the possibility of
cross-contamination. Enough probes were carried on each
van to avoid the need to reuse any during the day.
. Probe adaptors (TRC's special design) were used to connect
the sample probe to the vacuum pump. The adaptor was
designed to eliminate the possibility of exposing the soil gas
stream to any part of the adaptor. Associated tubing
connecting the adaptor to the vacuum pump was replaced
periodical as needed during the job to insure cleanliness
and good fit. At the end of each day die adaptor was cleaned
with soap and water.
• Silicone tubing (which acts as a septum for the syringe
needle) was replaced as needed to insure proper sealing
397
-------
around the syringe needle. The tubing does not directly
contact soil gas samples.
» Glass syringes were used for one sample only per day and
were washed and baked out at night.
. Injector port septa through which samples were injected into
the chromatograph were replaced on a daily basis to prevent
possible gas leaks from the chromatographic column.
. Analytical instruments were calibrated each day by analytical
standards from Chem Service, Inc. Calibration checks were
also run after approximately every five sampling locations.
. Sub-sampling syringes were checked for contamination prior
to sampling each day by injecting nitrogen into the gas
chromatograph.
. Prior to sampling each day, system blanks were run to check
the sampling apparatus (probe, adaptor, and 10 cc syringe)
for contamination by drawing ambient air from above ground
through the system and comparing the analysis to a
concurrently sampled ambient air analysis.
• All sampling and sub-sampling syringes were
decontaminated each day and no such equipment was reused
before being decontaminated. Microliter size sub-sampling
syringes were reused only after a nitrogen blank was run to
insure it was not contaminated by the previous sample.
• Soil gas pumping was monitored by a vacuum gauge to insure
that an adequate gas Dow from the vadose zone was
maintained. A reliable gas sample can be obtained if the
negative pressure reading on the vacuum gauge was at least
2 inches Hg less than the maximum pressure of the pump.
RESULTS
Twelve UST sites were evaluated by comparative mediods
previous to May 29, 1990. In all cases the condition of the site
determined by soil gas or Geld analytical measurements agreed
with the results obtained from soil and water samples, if the soil
gas action levels recommended by TRC were used to interpret
the soil gas data and the current Iowa UST Board action levels
were used to interpret soil or water data. When the data from
soils or water indicated that the site was contaminated, the data
from soil gas samples also indicated that the site was
contaminated. In Iowa, a site is considered contaminated if the
level of TPHC in soil is greater than 100,000 ug/Kg or the level of
benzene in the groundwater is greater than 5 ug/L.
There was no existing standard or action level for soil gas or for
field analytical methods. The action levels proposed for TPHC
levels were 1000 ug/L, 10,000 ug/Kg, and 500 ug/L for soil gas,
soils, and water, respectively.
In addition, the results of soil and water analyses using TRC field
analytical methods were compared to results obtained using
standard laboratory methods. The correlation between field
measurements and laboratory measurements was good for water
samples. The correlation between Geld measurements and
laboratory measurements was not as good for soil samples. This
result is not surprising, however, since water samples can be
homogeneous and soil samples are not. In many cases the soil
samples being compared were taken from the same bore hole but
were taken from different cores. It is very likely that discrepancies
between field and lab results for soils represent real differences
between samples as much as disagreements between analytical
methods.
Duplicate samples were also sent to two laboratories to check the
inter-laboratory reproducibility. The agreement between TRC
field methods and the samples sent to the two laboratories is as
good as the agreement between the two laboratories.
Soil Gas Sampling Versus Soil Sampling: Sample Integrity
INJECT B5/22/W 16:54; 12 STORED TO BIN I 12
ri.25
Figure 5. Soil gas sample #1 taken at a depth of 6 feet. NOTE: '
peaks are much larger than the later peaks, like fresh gasoline.
The early
A INJECT 05/22/9017:39:31 STORED TO BIN « 22
733
=5
1.25
Figure 6. Soil sample #1 taken at a depth of 6 feet NOTE: The early
peaks are smaller tnan the later peaks. Most of the volatile substances
were lost while handling the soU
Figure 5 is a chromatogram of a soil gas sample taken at a farmer's
coop in north-central Iowa. Notice that the early peaks are much
larger than the later peaks. Also notice that the chromatogram
looks very similar to the chromatogram for fresh gasoline vapors
(see Figure 2). Compare the chromatogram in Figure 5 to Figure
6 which is a chromatogram from the analysis of a soil sample taken
from the same location within a few inches. Notice that the early
peaks in Figure 6 are smaller than the later peaks. This is because
most of the volatile compounds have been lost during the handling
of the soil sample. Also die addition of water and agitation to the
soil sample prior to headspace analysis increases the signal from
lower volatility compounds.
T(l BIN * 59
Figure 7. SoU gas sample #8 taken at a depth of 3 feet NOTE: The early
peaks are larger than the later peaks, like fresh gasoline.
IN.IFC.T M/MM U:3ft:47 STORED TO BIN » 48
Figures. Soil sample #8 taken at a depth of3 feeL NOTE: The early
peaks are smaller than the later peaks. Most of the more volatile
substances were lost while handling the soiL
A comparison of Figure 7 and Figure 8 illustrates the same
principles. The chromatogram in Figure 7 is the result of the
analysis of a soil gas sample from Sioux City, LA, and the analysis
398
-------
in Figure 8 is the result of the analysis of a soil sample taken from
the same location. Once again the analysis of TPHC in soil gas
sample is more similar to that of fresh gasoline than that of
petroleum product left in the soil sample after Handling and
exposure of the soil to ambient air.
Site By Site Comparison
The results from four of the twelve skes are presented here.
These examples are intended to be representative of what
occurred in the study.
• SIOUX CITY, IOWA
This is a relatively new site where the UST's were recently
installed. The site is fairly clean, except for isolated
contamination near the elbows in the piping where the turn
is made toward the pump islands (see Table 1 and Figure 9
at sampling location 8). A leak in the pipe is indicated. There
is excellent agreement between the soil gas measurement
and the analysis of soil samples at this site. Both would have
located the problem. However, having the analytical
laboratory on site allowed additional samples to be taken at
locations 12, 13 and 14, which verified that location 8 is
within a few feet of the release point of the hydrocarbon
contamination.
TABLE 1
Sioux City, Iowa
05-23-90
CONDENSED DATA
SAMPLE
\-V
2-8'
W
4-8'
5-8-
t-vr
7-3-
•w
w
10-3'
11-3'
Shaded valu
TRC
TPHC
SG
ug/L
<4
<4
<4
a?
<4
<4
<4
5800
<4
2
SO
TRC
TPHC
SOIL
ug/l
nil
6
4
<2
<2
-------
• OLWEIN, IOWA
See Table 2 and Figure 10 for a summary of the results of
this site investigation. The results from the soil gas
investigation, the field analyses of soil samples, and the
laboratory analyses of soil samples all indicate that
contamination is not general throughout the site but is
localized around the eastern end of the tank pit, the piping
trench between the building and the pump islands and
around the southern pump island. On the map in Figure 10,
the largest indications of contamination are at sampling
locations 4, 6, 7, and 9. Analysis of soil samples indicated
problems at two locations where soil gas measurements did
not, but analysis of soil gas samples indicated contamination
problems in two locations where analysis of soil samples
indicated no problem. At this site, the same conclusions are
reached by using either investigative method as long as
multiple sample locations are examined in reaching those
conclusions.
TABLE Z
Olwein. Iowa
05-25.90
CONDENSED DATA
SAMPLE
1-9'
2.9-
3-W
4-0'
5-9'
6-91
7-3'
i-y
9-31
10- J1
ll-J'
TRC, TPHC. SG. ug/1
<1
<4
2700
58
as
1500
24000
5
5300
<4
3300
TRC, TPHC, SOIL.
ut/l
<8
<6
2200
5400
<13
16000
%000
too
21000
^4
310
KEYSTONE TPHC
SOIL uj/1
< 5000
-------
• GAS STA TION, DBS MOINES, IOWA
Because the surface of this tank pit was covered with grass,
which makes the tank pit a groundwater recharge area, and
the backfill material is native clayey soil, the soil gas survey
was not relied upon to survey this site. Soil samples were
collected at depths equal to or greater than the tank
bottoms. The samples were analyzed by the TRC field
method for soils and by two different independent
Laboratories. The agreement between the TRC results and
the Laboratory results was as good as the agreement
between the two laboratories (see Table 4 and Figure 12).
This illustrates the variability of soil samples.
The result of the investigation is that the site is contaminated
throughout the tank pit at the depth of the tank bottoms.
TABLE*
Gas Sution/D«s Moinc*. Iowa
06-11-90
CONDENSED DATA
SAMPLE
1
TRC,
TPHC
GW-HS
ug/L
600
NET
TPHC
GW
ug/L
140
PACE
TPHC
GW
ug/L
190
SAMPLE
2
3
4
5
6
7
TRC
TPHC
SOIL
"g/Kg
<0.5
1,200,000
62.000
26J
290,350
200
NET
TPHC
SOIL
Ug/Kg
< 15.000
860.000
43.000
< 15.000
26,000
< 15.000
PACE
TPHC
SOIL
ugrtCg
clO
40.000
HOW
<10
984.000
<10
Shaded value* represent concentraiions above action leveti.
TRC umple* analyzed by Tracer Research Corporation in mobile lab.
PACE tamplet analyzed by Pace Laboratories, Coralville, LA.
NET samples analyzed by NET Midwest Laboratories, Cedar Falls, LA.
TPHC signifies Toul Petroleum HydroCarbons.
SOIL signifies a «>iJ sample.
GW signifies a groundwater sample,
GW-HS signifies a groundwater mole analyzed using a bead space method
EXPLANATION
• 1 Prob* Location
. Pipeline location
o, Fill
Pump Island
10
'11
'12
3A
3B
N
FIGURE 12. GAS STATION,
DES MOINES, IA.
Figure 12. Facilities and sampling locations al a gas station In DCS
Moines, Iowa. Set Table 4 for analytical results.
401
-------
I CONVENIENCE STORE 2, DES MOINES, IOWA
High levels of contamination were discovered at all sample
locations at this site (see Table 3 and Figure 11). When
contamination was discovered early in the investigation, the
objective of the investigation changed to an effort to
determine the extent of the contamination. It was
discovered that the contamination extends mainly eastward
from the north end of the tank pit and does not diminish at
least to the border of the property. The extent of this
contamination indicates that petroleum is being released
underground currently and has been for some time. The
very high levels of hydrocarbons in the soil gas (100,000 ug/L
or greater) over a widespread area is typical of a significant
ongoing contamination problem. This is also the ideal
condition for the best correlation between different
investigation approaches. The results at this site were unlike
the results at almost all the other sites where isolated pockets
of minor contamination were indicated.
TABLES
Convenience Store ?2/Des Moines, Iowa
06-10.00
CONDENSED DATA
SAMPLE
TRC.TPHC,
SG.ug/l
110,000
10,000
25.000
23,000
53.000
S2.000
100,000
TRC.TPHC,
GW-HS, aifl
TRC.TPHC,
SOIL, ug/kt
60.000
PACE,
TPHC,
GW-HS. ug/l
PACE,
TPHC, SOU,
LSO.OOO
gK«^«4 values represent concentration* above action level
TRC samples analyzed by Tracer Research Corporation in mobile lab.
PACE samples analyzed by Pace Laboratories, CoraMIe, IA.
SG signifies toil gas sample.
SOIL signifies a soil sample.
TPHC signifies Total Petroleum HydroCarbons.
GW signifies a groundwaler sample.
GW-HS signifies a groundwater sample analyzed using a bead space method.
EXPLANATION
.pj Prob* Location
pipallna Location
o. HII
PI 3
•P14
PIS
P12
P17
•P7
P11
-a
P16
N
20
FIGURE!I. CONVENIENCE
STORE, DES MOINES, IA.
Figure 11. Faculties and sampling locations at a convenience store In
DCS Monies, Iowa. See Table 3 for analytical results.
402
-------
SUMMARY RESULTS
The following discussion is concerned with the general
conclusions that can be drawn from the data discussed above.
Comparison of Field and Laboratory Analytical Results.
Figure 13 is a plot of the field analytical results versus the
Laboratory analytical results for total petroleum hydrocarbons in
water samples collected at the same sampling locations. The data
are presented in log-log scaled plots because of the wide ranges
of data values. Table 5 is a summary of a regression analysis of
the data The regression results are reported as logarithms. As
can be seen in Figure 13, the agreement between the two methods
is good. The correlation coefficient yielded by the data is 0.87.
TRACER VCTSJS LABORATORr
Figure L3.Comparlson of field analytical results using a head space
method with laboratory results using a standard method for hydrocarbons
In water. See Table 5 for a list of linear coefficients and statistical results.
TABLES
LINEAR COEFFICIENTS AND STATISTICAL RESULTS OF A
COMPARISON OF FIELD HEADSPACE ANALYSIS AND STANDARD
METHOD ANALYZING OP HYDROCARBONS IN WATER
Slope (theoretically - 1)
Intercept (theoretical • 0)
Correlation Coefficient (r)
Thmbold value for
Mauttkal ligniTiciiKe of r
(at 99% confidence levd)
REGRESSION
RESULTS
a«s
ai4
.87
0.2B
STD.
ERROR
an
aw
.10
Also the theory line, which is generated by the assumption that
the field results should be equal to the laboratory result (slope =
1, intercept = 0), is just as good a representation of the data as is
the line generated by a linear regression analysis.
Some differences between field and laboratory water analyses and
between field and laboratory soil analyses are parallel. In both
cases the Geld analysis uses what is called a head space method,
and also in both cases the laboratory analysis uses what is called
a purge and trap method. For these reasons the TRC field
analysis for soils should compare favorably with standard
laboratory methods for soils if identical samples are analyzed.
The correlation between field and laboratory analyses of soil
samples should not be expected to correlate as well as water
samples. Soil samples are not typically homogeneous and,
therefore, should not be rigorously considered as split samples.
Very different samples can be collected form nearly the same
location. After examining the data in Table 4, it can be seen that
duplicate soil samples collected in this study are not reliably
similar. Also, the laboratory method uses methanol to extract
hydrocarbons from the soil The field headspace analysis relies
on water to wet the soil particles and displace the absorbed
hydrocarbons. A methanol extraction of the hydrocarbons
should be more efficient than a water displacement because the
hydrocarbons are more soluble in methanol while the methanol
is able to strongly wet the soil particle.
Figure 14 compares the field and laboratory analytical results.
While there is more scatter than in the comparison of
groundwater analyses the correlation is still strong (0.69 See
Table 6). However, the line generated by a regression analysis of
the data does not agree well with the theory une which assumes
both results should be equal. The log mean value of the ratio of
the laboratory analytical result divided by the field analytical
result is 13, which indicates a ratio of 20. This could be related
to the better extraction efficiency of the methanol.
If, as is the case for low level analyses, the soil sample is not
extracted with methanol but is simply mixed with water and placed
in the purge vessel, the field headspace method would yield results
that are roughly equivalent to the standard laboratory method.
TKAOEJt VERSUS LABCMATCAY
Figure 14. Comparison of Held analytical results using a headspace
method with laboratory results, using a standard melbanol for
hydrocarbons in soils. See Table 6 for a list of linear coefficients and
statistical results.
TABLE 6
LINEAR COEFFICIENTS AND STATISTICAL RESULTS OP A
COMPARISON OF HELD HEADSPACE ANALYSIS WITH STANDARD
LABORATORY ANALYSIS OF SIMILAR SOIL SAMPLES FOR
HYDROCARBONS
Slope (theoretically - 1)
Intercept (theoretically » 0)
, , field null
(Aver.,, of to, ^^
Correlation Coeflkienl
Tbreahold value for lUtiatiol lipiifionce
of r (at the W* confidence lever)
REGRESSION
RESULT
12
-IT
U"
O.M
0.24
STD. ERROR
0.16
U
aw
• See Equation 3
" See Equation 4
Comparison of Soil Gas Measurements with Soils Analyses
If the correlation of hydrocarbons in sol gas is represented by Cg
and the level of hydrocarbons absorbed to the soil and dissolved
in the water, which is adsorbed to the soil is represented by Cg,
the ratio of tie two values could be represented by K.
Equation 1:
Cs
K or,
K is not an equilibrium constant, but if the system were at
equilibrium, K would be proportional to the equilibrium constant.
Note also that K contains many variable factors such as the surface
area of the soil, the water content of the soil, the soil porosity and
the soil temperature. Since the data is plotted in the log-log scale,
the logarithm of Equation 1 becomes Equation 2:
403
-------
log -£ = log K or log Cg =
Cs
Which is equivilent to Equation 3:
logCg=logCs
A set of hydrocarbon levels governed by a single ratio, K, plotted
on a log-log scale would fallon a line with a slope of 1.0 and an
intercept equal to log K. In this way the intercept from a
regression analysis might be related to log K. See Tables 7-9.
An alternative method of obtaining log K would be to average the
logarithms of the ratios of the soil gas and the soil hydrocarbon
levels. Equation 4:
Average log K = \ log
CS(ri)
Figure 15 is a plot of soil gas levels of total petroleum
hydrocarbons with measurements of total petroleum
hydrocarbons in soils taken from nearly the same locations. The
correlation coefficient for the data set is 0.73, which is highly
significant. Also, the slope of 0.81 is more than the std. error
different from 1.0, but the theoretical slope of 1.0 is within the
95% confidence interval of the regression calculated slope.
There are many reasons why the correlation is not perfect. The
ratio of the amount of petroleum hydrocarbons in the soil gas
versus the amount adsorbed to the soil changes depending upon
the soil type and water content. The ratio also depends on the type
of fuel, length of time in the ground, or the distance between the
sampling point and the original source of contamination.
Hydrocarbons can be detected in soil gas at greater distances
from the source than in soil samples. Also, the condition, species,
and concentration of microbes in the soil have an effect. Finally,
the amount of volatiles lost from soil samples during handling
varies a great deal with soil types, water content, sampling
operator, and analyst.
Figure 15. Comparison of soil gas levels of total petroleum hydrocarbons
(TPH) with TPH levels In soil samples taken from nearly the same
location. Both samples were analyzed In the field. See Table 7 for a list of
linear coefficients and statistical results.
TABLE 7
LINEAR COEFFICIENTS AND STATISTICAL RESULTS OF A COMPARISON
OF FIELD SOIL GAS ANALYSIS AND SOIL ANALYSIS BY A FIELD METHOD
Slope (theoretically » 1)
Intercept (log K)'
Average of log K**
Correlation Coefficient (r)
Threshold value for statistical significance
of r (at the 99% confidence levd)
REGRESSION
RESULT
(177
-.117
..74
.73
.27
STD.
ERROR
0,10
\-2
.10
•See Equation 3
••See Equation 4
To illustrate the effect of some of these factors, consider Figure
16. This is a plot of the level of toluene in soil gas versus the level
of toluene in soil samples. (Note that in this analysis toluene is
not completely separated from the other hydrocarbons in the
sample.) Toluene is a common component of gasoline that is less
volatile than most of the components in gasoline (see Figure 2).
It is therefore less susceptible to loss during sampling. Therefore,
as expected, the correlation coefficient (0.81) for toluene in soil
gas versus toluene in soils is better than for total hydrocarbons in
these two kinds of samples (0.73).
Also, the slope of 1.0 calculated by the regression analysis is in
excellent agreement with theory. The exact agreement is
probably only circumstantial. The scatter in the data sets is best
represented by the standard error of the intercept. The standard
error of the intercept for the plot comparing toluene levels (0.82)
is also reduced from 1.2, which is the standard error of the
intercept for the comparison of total hydrocarbon measurements.
Figure 17 is a plot of total petroleum hydrocarbons in soil gas
versus total petroleum hydrocarbons in soils as determined by
keystone laboratories. The correlation coefficient (0.63) for the
data set is highly significant and the slope of the regression line is
in excellent agreement with expectation.
Although the scatter in each of these plots that compare soil gas
levels to levek in soils is large, as great as 12 (remember that this
is the error in log K) it is easily accounted for by allowing for the
possible variation in soil surface area alone. In other words,
clayey soils would tend to give rise to data skewed towards the X
axis and sandy soils would tend to give rise to data skewed towards
the Y axis.
Figure 16. Comparison of soil gas levels of Toluene with Toluene levels in
sou samples taken from nearly the same location. Both samples were
analyzed In the Held. See Table 7 for a list of linear coefficients and
statistical results.
TABLES
UNEAR COEFFICIENTS AND STATISTICAL RESULTS OF A
COMPARISON OF SOIL GAS LEVELS OP TOLUENE WITH TOLUENE
LEVELS IN SOIL SAMPLES
Slope (theoretically - LO)
Intercept (log K)*
Anri|e log K"
Correlation Coefficient
Threshold value for italistical tignificance
of r (it 99% confidence level)
REGRESSION
RESULTS
LO
-.079
•0,79
0,81
033
STD.
ERROR
O.U
O.E
0.12
•See equation J
" See equation 4
404
-------
Appropriate Action Levels for Soil Gas
The cleanup action level for TPHC in soils in Iowa, and some
other states, is 100,000 ug/Kg or 100 mg/Kg.
If a vertical line is drawn through the graph (Figure 17) at the value
of 100,000 ug/Kg, it divides the data into two groups, those above
the action level, called positives and those below the action level
called negatives. From a total of 51 samples, 15 are positives.
These soil samples are classified as contaminated above the action
level.
If a horizontal line is drawn through the intersection of the vertical
line and either the regression line or the theory line, it will intersect
the Y axis of the graph in Figure 17 at a value of approximately
200 ug/L. This horozontal fine divides the data into two sets.
Those levels above the line are called soil gas positives and those
below the line are called soil gas negatives.
When the analytical result for a sampling location falls above the
action level for soils and for soil gas, both methods are in
agreement. When the measured levels of TPHC falls below the
action level for soils and for soil gas, once again both methods are
in agreement. The frequency of agreement between field soil gas
and laboratory soil measurements by this approach is 0.8.
If, however, the intent is to use the soil gas survey as a screening
method, and the occurence of one or more contaminated samples
causes a site to recieve a closer look, the discrepancies that cause
the greatest concern are those in which soil gas analysis gives a
negative results when soils analysis would have yielded a positive
one. This mighr be called a false negative. An estimate of the
frequency or probability of false negatives from the data in Figure
17 is 0.08. Finally, it should be noted that soil gas samples are less
costly to collect and analyze than are soil samples. If therefore,
multiple soil gas samples are analyzed, the chances of a continued
false negative becomes (0.08)", in which n is the number of soil
gas samples. Very quickly, the chances of obtaining a repeated
false negative becomes vanishingly small ((0.08)2 = 0.006, [0.008]3
= 0.0005).
After this evaluation the action level for TPHC in soil gas was set
at 1000 ug/L as a compromise between false negatives and false
positives and to compensate for the fact that soil gas samples
would be collected closer to the contaminated sources. The use
of 1000 ug/L as the soil gas action level raises the frequency of
false negatives for soil gas to 10% in the data set in Figure 17. The
frequency of agreement of 0.8 is not affected.
TABLES
LINEAR COEFFICIENTS AND STATISTICAL RESULTS OF A
COMPARISON OF FIELD SOIL GAS LEVELS OF TOTAL PETROLEUM
H YDROCARBONS (TPHC) WITH TPHC LEVELS IN SOIL SAMPLES BY A
STANDARD METHOD"'
Slox (theoretically " 1)
Intercept flog K)*
Average log K"
Correlation coefficient (r)
TtimtioM value for statistical significance
of rlalW. confidence level)
REGRESSION
RESULTS
o.«
-13
•2,6
0.63
0.32
STD.
ERROR
0.17
L:
.012
*See equauon 3
" See equation 4
•"Metliod a OA-1 which is an lowi modification of EPA 8023
CONCLUSIONS
After a review of this data set it can be concluded that soil gas
investigations are a useful complement to soil and water sampling
approaches to site evaluations. It is important, though, to use
appropriate action levels for soil gas measurements.
Reasonable and practical guidelines can be written to ensure that
soil gas investigations be used at locations for which it is
appropriate. For those locations where soil gas measurements are
inappropriate, soil or water samples can be collected. It has been
shown that the Geld headspace analysis of volatile petroleum
hydrocarbons yields results that correlate very well with results
from standard purge and trap methods. The correlation between
the measurement of hydrocarbon levels' in soils by the Geld
headspace method and by OA-1 was not as good. Whether the
differences arose from heterogeneous samples or differences in
extraction efficiency was not determined.
It should be remembered that these results were obtained using
laboratory grade analytical equipment which was mobilized for
field use. Some soil gas investigations are conducted using hand
held instruments or portable gas chromatographs with little or no
temperature control of the sample stream. Since these devices
are not as reliable, caution should be used in applying these results
to those approaches.
This study has shown that total hydrocarbons in soil and water
samples can be reliably assessed using Geld analytical methods.
Figure 17. Comparison of soil gas levels of total petroleum hydrocarbons
(TPH) with TPII levels in soil samples taken from nearly the same
location. The soil samples were analyzed by a contract laboratory using
OA-1 which is an Iowa modification of EPA method 8015. See Table 9 Tor
a list of linear coefficients and statistical results.
405
-------
DISCUSSION
STEVE KNOLLMEYER: You seem to assume that the soil gas emanates from
the same place you collect the soil sample, rather than from the water table or soil
contamination deeper in the ground. Is that something you always see, or did you
study that at all? And secondly, did you correct any of the soil gas readings for
methane that may be there naturally?
RANDY GOLDING: The chromatograph was able to separate the natural
methane and it wasn't included in any of these numbers. We didn't assume that
the source for the soil gas vapors were in the same region of space as the sample
collected. But, since we were evaluating this method as a screening tool, we
simply collected the soil gas sample from the same region of space that any other
contractor would have collected the soil sample. We were comparing whether or
not the answers would agree.
DAVID CLIFT: What did you use for the standards?
RANDY GOLDING: This will surprise you, but we simply used a mixture of
benzene, toluene, and xylene, and we averaged the response factor for those to
calculate the total hydrocarbon number. That should over-estimate the hydrocarbon
number. We didn't use gasoline samples as our standard.
DAVID CLIFT: That was just gasoline that you're looking at, right?
RANDYGOLDING: Well, wedidn'tknow what we were looking at necessarily.
since the tank pits all contain multiple products.
DAVID CLIFT: You couldn't determine if they were aliphatic or aromatic then?
RANDY GOLDING: Not from the FID detector. Only from retention time. We
were limited by resolution problems because we were using this as a screening
method, therefore speed was also a factor. If we had a water sample in later
studies, the action level for water was defined by benzene levels. The action level
for soil was defined by total hydrocarbons. And so at that point we would take
the gas chromatograph, cool the temperature down, and try to separate benzene
out from the other products. And we easily attained the 5 ppb action level for
benzene. It was easy toobtain using a headspace method. We achieved a 0.02 ppb
or 20 ppt detection limit for benzene, if we did a good separation. Then we tried
to look to see whether or not benzene was the problem if we had groundwater in
the sample.
DOUG PEERY: What was the spectral difference between your soil gas and
your soil samples? How much apart were they?
RANDY GOLDING: It varied because being able to collect the soil sample in
that probe wasn't as reliable as one would like. But typically I would say that it
was within two feet. Sometimes we had to go down the same hole repeatedly
because we didn't get enough sample. Remember we would try to collect split
samples, one for the lab, one for us. And so sometimes, not always, the sample
was collected by two different excursions into the same hole, and so it's possible
that shavings from different depths were included in different samples. That's a
disadvantage and a weakness of the study.
DOUG PEERY: If I understand correctly, you did your soil gas and sometimes
you went back into the same hole and collected your soil samples?
RANDY GOLDING: Well, we would collect soil gas after the probe arrived.
And then before removing the probe we would usually collect the soil sample.
DOUG PEERY: Would that not bias your results in your soil samples because
you did your soil gas first, because the volatiles would be removed?
RANDY GOLDING: The partitioning ratio that you would get from the data
was always at least as large as you would predict just by doing a batch study in
the laboratory in a very controlled environment. There isn't any evidence in the
results that this occurred. And I agree that it's a concern that should have been
raised. We didn't evacuate large volumes of ground. In order to flush the probe
adequately, you only had to evacuate soil gas from a sphere that had a radius of
a few inches, perhaps four inches. But then we would often push well beyond that
point 10 or 12 inches to collect the soil sample.
DOUG PEERY: Okay. 1 know in some procedures for soil gas there is a purging
of the probe, and then we have found, (Dr. Spittler said of his previous work) that
over a period of time there is an equilibrium that is reached.
RANDY GOLDING: Then it would recover. There is a steady state that you
reach if you keep pumping, and then if you stop pumping it will recover.
DOUG PEERY: Right. But I take it that you did not go to the steady state
position?
RANDY GOLDING: No, we only Hushed the probe. We only tried to get a grab
sample of the soil gas.
406
-------
FIELD SCREENING OF BTEX IN GASOLINE-CONTAMINATED GROUNDWATER
AND SOIL SAMPLES BY A MANUAL, STATIC HEADSPACE GC METHOD
James _D_1_St_uart:, Suya Wang, Dept. of Chemistry, U-60, Gary A. Robblns, Dept. of Geology and Geophysics,
U-45, The University of Connecticut, Storrs, CT 06269-3060 and
Clayton Wood, HNU Systems, Inc., 160 Charlemont St., Newton Highlands, MA 02161-9987.
ABSTRACT
A manual, static headspace GC method has
been developed and used in the field for the
screening of gasoline-contaminated groundwater
and soil samples. This developed, static
headspace method has focused primarily on the
analyses of benzene (B), toluene (T),
ethylbenzene (E), and the three xylene isomers
(X) (often collectively abbreviated as BTEX).
However, this method also allows for the
determination of methyl-t-butyl ether (MTBE),
trichloroethylene (TCE) and tetrachloroethylene
(PCE) in the headspace above the aqueous layer
as detected by a photoionization detector (PID)
of a field-portable gas chromatograph. The
headspace method is performed in the same 40-mL
VOA vial in which the sample is collected, hence
reducing the possibility of sample loss due to
volatilization. Good agreement was found
between the field, static headspace method, a
laboratory-based manual, static method and a
laboratory-based, purge-and-trap method. The
results of field screening for BTEX, MTBE, and
PCE at several sites in the New England area
will be presented.
INTRODUCTION
Simple field methods associated with the
use of portable instruments have been reported
to give dependable data while saving time and
money (1,2). These methods can provide for the
rapid screening of large numbers of samples in
the field, thus providing for more effective and
timely site assessment and evaluation of
on-going remediation efforts. In addition,
sample loss due to volatilization and/or
bacterial alteration of the targeted compounds
compounds can be effectively avoided.
Reports by Spittler, et_al^ (3-5) and
Grob, et_al^ (6-7) have shown that the static
headspace method can be used as a rapid and
effective method for the analysis of various
volatile organic pollutants in groundwater and
soil samples. Wylie found that using optimized
conditions and the same analytical instrumention
that the static headspace method can be as
sensitive and as reproducible as the dynamic,
purge-and-trap method. He noted that the static
method is obviously more portable and better
able to be used on a variety of environmental
samples, such as soils and sludges (8).
Recently, we published a brief technical note
that described using a manual, static headspace
method for the analyses of BTEX in gasoline-
contaminated groundwater and soil samples (9).
Since the time of that publication, we have
employed the manual, static headspace method in
the field at four sites in Connecticut that have
experienced contamination due to leaking
underground storage tanks (LUST). A report of
our findings will be described in this paper.
EXPERIMENTAL
Instrumentation. The field separations were
performed on a portable gas chromatograph (HNU
Systems, Model 311). A splitless injection was
employed onto a narrow bore, 0.32 mm i.d., 25 m
in length, capillary column having a 1.0 micron
film thickness of dimethyl polysiloxane
(Nordibond NB-30, HNU Systems). A column flow
rate of 5.0 mL/min was used. The column's
eluent was passed to a photoionization detector
(PID) equipped with a 10.2-eV lamp whose output
was to a built-in integrator on the Model 311.
The column was set isothermally at 60°C, while
the injector's and detector's temperature were
set at 90°C. Manual injections of the head-
407
-------
space vapors were accomplished using 50- or
100-microliter, gas-right, fixed needle
microsyringes (Scientific Glass Engineering).
For comparison work, a laboratory-based manual
static headspace method was performed using the
splitless injection mode onto a capillary
column, gas chromatograph (Hewlett-Packard Model
5890A). A megabore capillary column, 0.55 mm
i.d., 30 m in length with a 3.0 micron film of
dimethyl polysiloxane (DB-1, J&W Scientific) was
used. The column's eluent, at a flow rate of
8.0 mL/min, was passed through a PID (HNU
Systems, Model 52-02A) equipped with a 10.2-eV
lamp, followed by a flame ionization detector
(FID). The output of each detector was
displayed on an integrator (Hewlett-Packard,
Model 3396A). The following column oven
temperature program was used: initial
temperature, 40° C, initial time 1 min,
temperature program rate, 8 °C/min, final
temperature, 190° C. For the laboratory-based,
megabore columns, a 200-nncroliter portion of
the headspace was injected using a
250-microliter, gas-night microsyringe.
For the laboratory based purge-and-trap
comparisons, an equivalent procedure to the one
described for Method 52A.2, "Measurement of
Purgeable Organic Compounds in Water by
Capillary Column GC/MS", was employed with the
following equipment: a dynamic headspace
concentrator (Tekmar, Model 2000) equipped with
sixteen port, 5-mL glass sparge chambers on an
automat*.]c liquid sampler (Teckmar, Model 2016).
A cryogenic focusing interface (Tekmar Capillary
Column Interface) was used to attach the purge-
and-trap system to the splitless injection
systems of a Hewlett-Packard 5890A gas
chromatograph. The detector system was a mass
selective detector (Hewlett-Packard 5970) with
an associated Hewlett-Packard 5895 Chem Station.
A narrow bore, 0.32 mm i.d. capillary column, 30
m in length with a 1.8 micron film thickness of
DB-624 (J&W Scientific) was used. The column's
flow rate was adjusted to 2.0 mL/min.
In subsequent studies, a second purge-and-trap
unit: was used which consisted of a dynamic
headspace concentrator (Teckmar, Model 4000)
with ten port, 5-mL glass sparge chambers
(Teckmar Model ALS) that: was connected to a
packed column, gas chromatograph (Perkin-Elmer,
Model 3920B) equipped with two detectors, a PID
(HNU Model 52) with a 10.2 ev lamp and a FID.
The output of both detectors were sent to a
2-pen recorder (Perkin-Elmer Model 023) and to
two integrators (Hewlett-Packard Model 3390A).
A packed column, 8 ft. long, 0.125 in. o.d.,
0.085 in. i.d. packed with 1% SP-1000 on
Carbopak B 60/80 Mesh (Supelco, Inc.) was
employed; a column flow rate of 40 mL/min was
used. The following column oven temperature
was used: a 4 min. hold at an initial column
temperature of 15°C, followed by a 8&C/min.
temperature program to 220° C, with a variable
final temperature hold. With these chromato-
graphic conditions, the peak for MTBE eluted at
about 16 min and excellent: resolution for MTBE
and the BTEX compounds were obtained.
Vials for the Static Headspace Method. The
gasoline-polluted groundwater samples were
direcly collected in 40-mL glass vials (Supelco,
Part No. 2-3278), with hole caps (Supelco, Part
No. 2-3283) and Teflon®-faced septa (Supelco,
Part No. 2-3281). Prior to the field sampling,
100 microliters of a 24,000 mg/L aqueous
solution of mercuric chloride were added to each
vial. A final concentration of 60 mg/L of
mercuric chloride in groundwater samples was
proven to be an effective method of BTEX
preservation against microbial degradation (9).
Immediately after sampling, the capped vials
were inverted to reduce the loss of volatile
organics and placed in a 25.0°C water bath if
analysis were to be performed in the field or
packed on ice and returned to the laboratory
where they were kept refrigerated at 4dC.
Analyses of Groundwater Samples. The VGA vials
containing the 40 mL sample of the groundwater
were placed in a 25.0°C water bath in order to
reach thermal equilibrium. Then a 1.5 In. long,
22-gauge needle was inserted through the septum
to allow air to enter. Next, a similar needle
attached to a 10-mL Luer-Lock syringe was used
to remove 10.0 mL of the aqueous phase. The
vial was kept in an inverted position and shaken
thoroughly for 2 min. The vial, with the 10.0
mL of headspace, was again placed In the 25.0°C
water bath and allowed to reach thermal and
phase equilibrium. At the time of analysis,
normally 50 microliters for the portable GC and
200 microliters for the laboratory GC were
withdrawn with a gas-tight syringe and Injected
into the gas chromatograph.
Analyses of Soil Samples. First, an
identification label needs to be placed on each
clean, empty vial equipped with its individual
holed-cap and septa. Then the weight of the
empty vial is measured to within ± 0.010 g.
Thereupon 25.0 mL of distilled water is
carefully pipetted into each vial and 100
microliters of the 24,000 rag/L mercuric chloride
added as a preservative. The vial with its cap
and septa is then reweighed. During the field
sampling, the soil sample with a range from 5 to
10 g. is carefully added to the vial which is
then quickly capped. The vial and its contents
are then thoroughly shaken for 2.0 min and the
entire contents reweighed. The weight gain
corresponded to the weight of the soil sample
408
-------
taken for analysis. Depending upon whether the
sample is to be analyzed in the field or in the
laboratory, the vial is either placed in the
25.0°C water bath or on ice for transportation
back to the laboratory.
RESULTS AND DISCUSSION
Figure la shows the separation obtained on the
HNU-311 portable gas chromatograph for an eight
component aqueous standard. The concentrations
for the BTEX components in the aqueous phase
were at: the 880 ppb levels, while MTBE was 1820
ppb, TCE 3008 ppb and PCE 1747 ppb. Referring
to Fig. la, it may be seen that almost complete
return to baseline occured between the peak due
to ethylbenzene (peak 6) and the peak due to the
co-elution of m- and o-xylene (peak 7). However,
it should be noted that only a very small peak
(peak 1) is obtained by the static headspace
method for MTBE, even at a significant
concentration of 1820 ppb in the aqueous phase.
This is because the Henry's Law constant for
MTBE is very small. A preliminary estimate of
<0.01 (in unitless terms) has been obtained in
our work. This means that MTBE tends to remain
in the aqueous phase and does not significantly
partition into the headspace. Table 1 summarizes
average Henry's Law constants for the compounds
used in this paper.
On the laboratory-based, HP-5890 gas chromato-
graph, a series of monthly calibrations had
established that for MTBE and the BTEX compounds
that there were linear increases in peak areas
with increases in concentration over 3- to
4-orders of magnitude for both the PID and FID
detectors . Table 2 gives values for the method
detection limits obtained for the static
headspace method using the portable gas
chromatograph (HNU Systems, Model 311) with its
associated detector and integrators settings
commonly employed in the BTEX analyses of
groundwater samples.
Table 3 presents data that compares the results
of analyses of the same gasoline contaminated
groundwater sample performed in the laboratory
by the manual, static headspace method to an
automated, purge-and-trap GC/MS method,
equivalent to EPA Method 524.2. It may be seen
that for most of the comparison of the results .
by the two very different analysis methods that
there is in general very good agreement. As
expected the purge-and-trap method reported
concentrations in the lower ppb range that were
not detected using the headspace method. For
the more contaminated samples, the headspace
method tended to give higher concentrations.
Upon examining the chromatograms, it appeared
that more peaks coeluted with the peaks due to
MTBE, benzene and/or toluene in the short, about
20 min. analysis time, in comparison to the
longer, about 40 min. elution time employed by
the purge-and-trap. Also, significant column
overload of the narrow-bore, capillary column of
the portable GC was observed. As may be seen
in Fig. Ib, the integrator's plot of the
portable GC remained above scale for a
significant portion of the chromatogram, and
reported values of 0.00 ppb for both MTBE and
benzene, whereas the laboratory-based GC
reported values of 54,600 and 1260 ppm levels,
respectively for the same groundwater sample
(not listed in Table 3).
In July of 1990, the opportunity arose to
perform field analyses on groundwater samples
from nineteen monitoring wells at a State Lust
site in Westbrook, CT. The site is located in
the middle of the small town at a busy
intersection of coastal Route //I (Boston Post
Rd.). This is a complex site where it was
believed that gasoline-contamination may have
been caused from two leaking underground storage
tank (LUST) locations. In January of 1989, in
response to a report of gasoline fumes in an
nearby commercial building, the State Dept. of
Environmental Protection authorized a private
engineering firm to perform investigatory and
remedial action. After conducting geographical
studies and historical record searching, soil
gas probings and volatile organic analyses of
groundwater and soil samples were performed. It
was surmised that a plume of underground
gasoline-contamination eminated from at least
one of the underground storage tanks (UST) and
travelled in a general NNWest direction towards
and around the northside of the commercial
building, in a line generally delineated by
monitoring wells (MW1, MW2 and MW7, Table 4). A
total of six, underground storage tanks were
removed from the site during the spring of 1989.
Holes were found in several tanks, and a film of
free product noted on the water in the
excavation pits. At approximately the location
where the underground tanks were removed, a
recovery shed housing groundwater pumps and a
stripping tower were installed. Table 4
summarizes the analyses for MTBE and BTEX
performed in the field on the HNU-311 at the
Westbrook, CT site.
In the two days of intensive sampling at the
Westbrook Site, the HNU-311 portable GC was able
to rapidly and effectively screen for MTBE and
the BTEX compounds in the many groundwater
samples. It should be noted that analyses of
MW 4,5,9 and 17 located at a distance and to the
southwest of the expected plume were found to
have only significant levels of MTBE. Fig. Ic
shows the chromatogram of the headspace for MW5,
showing only a single peak for MTBE. In an
interesting article, Garrett, et al., have
suggested that MTBE provides an excellent
indicator for the outer limits of a gasoline
409
-------
plume because It spreads further and faster than
gasoline (10). Also, It is expected that MTBE
is not readily degraded by the subsurface
bacteria.
In August of 1990, the opportunity presented
itself to use the HNU-311 portable GC at a site
known to be contaminated with tetrachloro-
ethylene (PCE). In a small shopping center in
Weston, CT, a dry cleaning shop had been in
operation for a number of year. In recent
months, levels of PCE in the 10-100 ppb range
had been found in drinking water wells of homes
downgradient from the shopping center. In
one day of field work, in conjunction with State
regulatory officers, samples from about five
home drinking water wells, groundwater
monitoring wells and soil samples were all
analyzed on-site. A shallow hole was manually
dug almost at the back-door of the dry cleaning
shop. Soil samples were taken at various
depths. The results of the field analyses on
these soil samples by the manual, static
headspace method using both the HNU-311 and a
Phot ovac 10S50 (operated by a State regulatory
personnel) and later in the laboratory by
purge-and-trap, packed column GC with PID and
FID detectors are presented in Table 5. These
results indicated that, the manual, static
headspace method was an excellent, field
screening method for the determination of PCE
and other such unsaturated, branched chlorinated
solvents in soil and water samples. It was
found that less loss of TCE occured if the
samples were analyzed in the field.
CONCLUSION
A manual static headspace method has been used
in the field on groundwater and soil samples at
several sites found to have a wide variation in
organic contamination. The method works well
for BTEX, TCE and PCE as they are readily
detected in the low ppb levels by the PID
detector. The method has been found to be
especially valuable in that it is portable and
may be relatively easily performed. But, above
all, the analytical results are available almost
immediately to aid in evaluating any on-going
site characterization and/or remediation
efforts.
References
1. "Field Measurements. Dependable Data When
You Need It", EPA/530/UST-90/003, Office of
Underground Storage Tanks, Washington, DC
20460, July 1990, pp. 1-92.
2. Kerfoot, H.B., Amick, E.N., Pierett, S.L. ,
Lewis, T.E., Bottrell, D.W., "Field Eval-
uation of Portable Gas Chromatographs",
EPA/600/X-89/030, USEPA Environmental
Monitoring Systems Lab, Las Vegas, NV, Feb.
1989, pp. 1-76, Appen. A-G.
3. Spittler, T.M., "Field Measurement of PCBs in
Soil and Sediment Using a Portable Gas
Chromatograph", Proc. of the Nat. Conf. on
Management of Uncontrolled Hazardous Waste
Sites, Hazardous Materials Control Rsch.
Inst., Silver Springs, MD, 1983, pp. 105-107.
4. Clay, P.F. and Spittler, T.M., "The Use of
Portable Instruments in Hazardous Waste
Site Characterization", Proc. of the Nat.
Conf. on Management of Uncontrolled
Hazardous Waste Sites, Hazardous Materials
Control Rsch. Inst., Silver Springs, MD,
1983, pp. 40-44.
5. Spittler, T.M., Siscanaw, R.J. and Lataille,
M.M., "Correlation Between Field GC
Measurement of Volatile Organics and
Laboratory Confirmation of Collected Field
Samples Using the GC/MS", Proc. of the Nat.
Conf. on Management of Uncontrolled
Hazardous Waste Sites, Hazardous Materials
Control Rsch. Inst., Silver Springs, MD,
1982, p. 57.
6. Umbreit, G.R. and Grob, R.L., "Experimental
Application of Gas Chromatographic
Headspace Analysis to Priority Pollutants",
•L._!!iy.ll°!-l:.J>ci- _Health, A15 (5), 1980,
429-466. ~ •-••••- -
7. Kiang, P.H. and Grob, R.L., "A Headspace
Technique for the Determination of Volatile
Compounds in Soil", J^ JEn^rijron L_Sci^Health,
A21 U), 1986, 71-100. - -
8. Wylie, P.L., "Comparing Headspace with Purge
and Trap for Analysis of Volatile
Priority Pollutants", Ji_Am^_Wa_ter_Works
Assoc^, 80, 1988, 65-72.
9. Roe, V.D.7 Lacy, M.J., Stuart, J.D. and
Robbins, G.A., "Manual Headspace Method to
Analyze for the Volatile Aromatics of
Gasoline in Groundwater and Soil Samples",
A-nali_c_heJ5i> II> 1989> 2584-2585.
lO.Garrett, P., Moreau, M. and Lowry, J.D.,
"MTBE as a Ground Water Contaminant",
Proc. of Conf. of Petroleum Hydrocarbons and
Organic Chemicals in Ground Water, 1986.
DISCUSSION
MATT STINCHFIELD: Have you looked at using other disinfectants such as
an organomercury compound, which is much more toxic and might allow you to
use much lower concentrations.
JAMES STUART: No
THOMAS SPITTLER: We have, with things like acids. And I believe in our
Cincinnati lab they did quite a bit in biodegradation studies. Herb Brass in
particular. He'll tell you a lot of the different things that he tried. Mercuric
chloride just seemed to be the simplest. It didn't change the pH, was easy to use
and it worked.
410
-------
Table 1. Certain aqueous properties of the compounds studied, at 25 °C
Compound Henry's Law Constant1'2 Solubility1
(dimensionless) (mg/1)
benzene 0.22 1800
toluene 0.27 510
ethylbenzene 0.35 160
m-xylene 0.29 160
p_-xylene 0.29 180
o-xylene 0.20 190
1,1,1-trichloroethylene (TCE) 0.42 1000
tetrachloroethylene (PCE) 0.70 400
MTBE3'4 <0.01 43000
1 Calculated from the data of Mackay, D., Shiu, W.Y., "Critical review of Henry's Law
constants for chemicals of environmental interest", J. Phys. Chem. Ref. Data 10
No. 4, 1981, 1175-1199. "
2 Calculated from the data of Ashworth, R.A., Howe, G.B., Mullins, M.E. and Rogers, T.N.,
"Air-water partitioning coefficients of organics in dilute aqueous solutions",
^_M2!rdous_Materj.als, !§» W&&, 25-36.
3 Approximate Henry's Law constant for
MTBE from our studies.
Garrett, P., et al. (10).
Table 2. Method detection limitsl in ppb for the manual, static headspace method,
obtained on the portable gas chromatograph (HNU Systems, Model 311).
On the headspace of a 8.8-30 ppb aqueous standard using the photoionization detector.
Benzene Toluene Ethylbenzene m-and p_-xylene o-xylene
3.2 3.7 7.3 7.9 15.3
MTBE TCE PCE
5.0 6.9 7.4
1 Method Detection Limits measured according to: Appendix B, Part 136, Federal Register,
40 CFR Ch.l (7-1-88 Edition), pp. 510-512.
411
-------
Table 3. Comparison of results, In ppb, between the static headspace method
automated purge-and-trap GC/MS method.
to an
Sample Method MTBE Benzene Toluene Ethylbenz. m-and p_-xylene o-xylene
TMW-1
TMW-1
TMW-6
TMW-6
MMW-1
MMW-1
MMW-4
MMW-4
MMW-6
MMW-6
MMW-7
MMW-7
MMW-9
MMW-9
MMW-10
MMW-10
MMW-1 1
MMW-1 1
heads pace
purge&trap
heads pace
purge&trap
headspace
purge&trap
headspace
purge&trap
headspace
purge&trap
headspace
purge&trap
headspace
purge&trap
headspace
purge&trap
headspace
purge&trap
ND
<2
ND
8.7
72
130
483
390
ND
11
co-elut
165
ND
<50
ND
1.3
ND
2.2
ND
ND
ND
ND
ND
ND
17.3
12.7
1144
709
496
330
84.5
ND
10.5
8.1
21.2
18.0
ND
ND
ND
ND
ND
ND
61.8
40.6
4320
2170
3180
2800
ND
ND
ND
ND
ND
ND
ND
ND
13
ND
861
411
28.5
15.3
1500
1420
489
379
365
361
23.3
13.5
ND
ND
ND
ND
ND
ND
1830
1220
690
315
6940
4170
4260
4430
664
944
74.8
70.2
ND
1.4
ND
ND
ND
ND
426
276
399
322
2030
1980
2080
2120
33.5
ND
ND
ND
ND
ND
412
-------
Table 4. Results of field analyses of MTBE and BTEX at the Westbrook Site, July 1990,
In ppb.
MW Location MTBE Benzene Toluene Ethylbenx. m-and g.-xylene o-xylene
MW1
MW2
MW3
MW4
MW5
MW6
MW7
MW8
MW9
MW10
MW11
MW12
MW15
MW16
MW17
MW18
MW19
MW20
B-l
B-2
B-3
B-4
B-5
B-6
ND =
-MW14
Not accessible used for groundwater air stripping
ID 5950 27100 2180 6220 4690
ND ND ND
ND ND ND
ND ND ND
1490 4530 4400
ND ND ND
ND ND ND
ND ND ND
ND ND ND
791 778 832
1.1 1.6 1.9
2.6 3.0 3.8
ND ND ND
9.5 10.2 8.8
ND ND ND
ND
5950
27100
Abandoned
117
453
ND
52100
ND
1570
5.8
ND
Used for
1730
Abandoned
469
ND
0.4
ND
ND
ND
ND
ND
NI
ND
16.
0.
ND
soil
818
ND
ND
ND
1.
ND
0.3
ND
ND
12400
ND
2 ND
6 ND
ND
vapor extractions
790
3.8
1.6
ND
1 4.1
ND
nv nu nu
not sampled, had septic leachate
4210 262 512
ND ND ND
ND ND ND
813000* 33500* 123000*
94 375 146
ND ND ND
ND ND ND
5610 17500* 10800*
Not Detected, NI • Not Integrated, * = Integration count exceed limit
Table 5. Comparison of the analyses for PCE in soil samples, Weston, Ct. during
August of 1990, expressed as mg of PCE per kg of soil.
Soil
Depth
Method & Instrument
PCE Cone.
1 ft. headspace, HNU-311 0.395
headspace, Photovac 10S50 0.19
purge&trap, Perkin-Elmer^ 0.135
Soil
Depth
2 ft.
Method & Instrument
PCE Cone.
headspace, HNU-311 36.8
headspace, Photovac 10S50 12.82
purge&trap, Parkin-Elmer^ 1.52
3 ft. headspace, HNU-311 3.71
headspace, Photovac 10S50 3.70
purge&trap, Perkln-Elmer^-- 0.17
3.5ft. headspace, HNU-311 1.08
headspace, Photovac 10S50 0.61
purge&trap, Perkin-Elmer-^ ND
Lab analyses performed 7 and 8 days after sampling.
in polyethylene bags with a significant headspace.
Integrator reported over-ranged.
Soil samples improperly stored
413
-------
la.
03 OQ QQ (D GQ OQ OQ
a. CL. a. a. o- a. o.
0. Q- 0. tt- 1- Q- fl-
Ib.
1C.
U
z a. a. o m OQ m
ue a. o, a. a. a. a.
•-»• a. a. ft* A.
C C J3 * *
;sxix
- c r X X
a * — i i
>- m LU i o
£"'
Fig. la. Separation of an eight component mixture by injecting the headspace of an aqueous standard
solution onto the HNU-311 portable gas chromatograph.
Ib. Chromatogram showing significant column overload of a highly gasoline-contaminated ground-
water sample onto the narrow-bore, capillary column of the HNU-311 portable gas chromatograph.
Ic. Headspace sample of a groundwater sample found to contain only methyl-t-butyl ether (MTBE).
414
-------
Comparison of
Aqueous Headspace Air Standard vs STJMMA Canister Air Standard
for Volatile Organic Compound Field Screening
H. Wang
Roy F. Weston, Inc.,ESAT Project
Landmark One, One Van de Graff Drive
Burlington, MA 01803
W. S. Clifford
United States Environmental Protection Agency
Region I, Environmental Service Division
60 Westview Street
Lexington, MA 02173
Abstract
Introduction
This paper describes the application of
SUMMA canister and aqueous headspace air
standards for ambient air volatile organic
compound(VOC) field screening to perform quick
on-site analysis using a portable gas
chromatograph(GC). Studies were conducted
comparing aqueous headspace standards to
SUMMA canister standards using a portable gas
chromatograph. A comparison of SUMMA canister
analytical results from the portable GC versus
GC/MS (gas chromatograph/mass spectrometer)
was provided. Research on time dependent stability
and temperature dependency of SUMMA canister
standards was also conducted. A Photovac 10A10
portable gas chromatograph(GC), an HP 5890/5970
gas chromatograph/mass select detector(GC/MSD)
and a Tekmar 5000 thermal desorber modified for
canister analysis were employed.
Toxic volatile organic air pollution is a growing
concern because of its widespread presence in the
atmosphere, adversely affecting public health. There
has been much interest in monitoring ambient air
for these toxic compounds. The United States
Environmental Protection Agency(U.S. EPA) has
developed several methods for measuring toxic
organic compounds in ambient air. These include
collection on solid adsorbents, such as Tenax GC
and spherocarb traps, as well as the collection of
whole air in suitable canisters[l,2]. With the
increasing interest in air analysis, field screening for
ambient air is becoming more important. When
performing on-site ambient air analysis for volatile
organic compounds (VOCs) using portable gas
chromatography, it is important to have a suitable
standard to be able to identify and quantitate
compounds of interest. The headspace above a
415
-------
Headspace
10 //g/L aqueous standard kept at a constant
temperature can be used as a VOC field screening
standard to perform quick on-site air analysis using
a portable gas chromatograph. The headspace
standard is a very simple and inexpensive technique
for standard preparation. VOCs, with their
relatively high vapor pressures, have a natural
tendency to migrate from water into air. In a
closed VOA(volatile organic analysis) vial filled
three-quarters full with an aqueous VOC standard
(Figure 1), VOCs
will move from the
water into the air
above the water
(headspace) until an
equilibrium is
reached. This air
above the water is
a perfect medium for
an air VOC standard
since it consists of
air and the migrating
VOCs from the
water.
the mole fraction or the volume-concentration
(ppb/v) of a component i in is:
Aqueous Solution
Figure 1. Aqueous Headspace
By the ideal gas equation of state:
PV=nRT
P: pressure V: volume
R: gas constant n: moles
T: absolute temperature (K)
for a single component i in a gas mixture:
PiV=niRT
Pi: Partial pressure of component i
n;.- Moles of component i
(1)
Mi=ni/2nj=piRT/2pjRT=pi/PT
(2)
MJ: Mole fraction or volume concentration of
component i
PT: The total pressure of the gas mixture(=2pj)
Because the aqueous solution is very diluted(10
fig/L), it can be treated as a ideal solution.
According to Henry's law:
pi=ki(T)Xi
(3)
kj(T): Henry's law constant;
Xj: Mole fraction of solute i in the water
solution.
Therefore, the concentration in the headspace is
shown as:
Mi=Pi/P=ki(T)Xi/PT
Henry's law constant k; varies with temperature(T),
therefore, the concentration in the headspace is a
function of temperature(T), the total pressure above
the solution(P-i) and the mole fraction in the water
solution(Xj).
Mi=/(Xi,T,P)
Whether an aqueous headspace standard can be
used for an air analysis standard depends upon the
ability to control the concentration of VOCs in the
aqueous solution, the pressure of the headspace and
the temperature of the standard. The first variable
Xj which reflects the concentration of a VOC in the
416
-------
water solution can be simply controlled by
preparing a solution with a known concentration of
the VOC. The temperature(T) of the standard can
be controlled and kept constant by placing the
VGA vial in an ice-water bath, keeping the
aqueous standard solution in the vial at
approximately CF - l°C. The total pressure(Pr)
above the solution is equal to atmospheric pressure.
Relative changes in atmosphere pressure are
negligible. Therefore, for this screening application,
the total pressure (Pj) can be considered a
constant. With the ability to control the variables
above, the aqueous headspace can be used for
ambient air field screening analysis as an external
standard.
SUMMA canister based sampling systems have
gained wide acceptance for the collection of
integrated whole ambient air samples containing
volatile organic compounds. Utilization of this
sample collection method has increased
significantly. Some recent research has used
SUMMA canisters as VOA standards[4]. As an
application, SUMMA canisters are able to be used
as field screening standards as well. Canister
standards present the true concentration of VOCs
within the can and the VOCs stored in a canister
exhibit relatively long term stability. In addition, the
transportation of a canister standard is easy;
therefore, the development of this method can be
a very effective approach of SUMMA canister
methodology and VOC field screening for ambient
air.
The following work is on the method studies of
SUMMA canister application and aqueous
headspace as ambient air standards for field
screening of VOCs. The comparison of aqueous
headspace standards to SUMMA canister standards
using a portable gas chromatograph and a
comparison of SUMMA canister analytical results
from the portable GC versus GC/MS were
performed. Research on time dependent stability
and temperature dependency of SUMMA canister
standards was also conducted.
Experimental
(1) Evaluation of aqueous headspace standards:
Experimentation was performed to determine
the actual concentration of selected VOCs in the
headspace above a 10 ^g/L aqueous standard
contained in a closed 40ml VOA vial filled with
30ml of aqueous standard kept at a temperature of
0°-lt. Analysis was performed on a Photovac
10A10 portable gas chromatograph equipped with a
4' 1/8" SE-30 column and a photoionization
detector, calibrated with a Research Triangle
Institute(RTI) certified mixture of VOCs traceable
to NBS primary gas standards. (Table 1.) The
Table 1. RTI Certified Concentration for IS Component Mixture
Containing Volatile Toxic Organic Compound
Traceable to NBS Primary Gas Standards
Cylinder No. ALL 21378
Compound
Vinyl Chloride
Bromomethane
Trichlorofluoromethane
(Freon 11)
Methylene chloride
Chloroform
1,2-Dichloroethane
1,1,1-Trichloroethanc
Benzene
Carbon tetrachloridc
1,2-Dichloropropanc
Trichlorocthylene
Toluene
1,2-Dibromocthane
Tetrachloroet hylene
Chlorobenzenc
Ethyl benzene
o-Xylene
Concentration (ppb/v)
5.19
5.68
5.15
4.48
4.86
5.02
5.22
5.15
5.02
0.5
0.6
0.7
OS
OS
OS
0.4
0.3
OS
5.15 ± 0.3
5.11 ± 0.3
5.19 ± 0.3
4.83 ± 0.5
5.24 + 03
5.27 ± 03
4.85 ± 0.3
5.12 t 0.5
417
-------
Photovac portable GC was calibrated by running a
syringe blank and a single point of the RTT
certified cylinder standard. Concentrations of VOCs
in the cylinder were approximately 5ppb. An
aqueous standard (working standard) was prepared
using 5.0 fA of commercial (Supelco) and EPA
repository standard mix solution (200 ^g/ml for
each component) diluted to 100 ml with VOC-free
water giving a final concentration of 10 ng/L for
each component. Table 2 shows the component and
concentration of the Supelco and EPA repository
stock standards. The working standards were stored
in 40ml VOA vials with zero headspace at 0° - It.
Before the analysis, 10 ml of water solution was
Table 2. Components and the
and of the Canister
of the Stock Standard Solution (Supelco Purgeable A and B, and EPA Repository)
Compounds
Concentration
Stock Solution Canister Standards
(^g/ml) 0/g/m ^ (ppb)
Trichlorofluoromethane
1,1-Dichlorocthylcne
Methylene Chloride
t-l,2-Dichloroethylene
1,1-Dichloroethane
Chloroform
Bromochloromethane
1,1,1-Trichloroethane
Carbon Tctrachloride
Benzene
1,2-Dichloroelhane
Trichloroethylene
1,2-Dichloropropanc
Bromodichlorome thane
1,3-Dichloropropene
Toluene
1,1,2-Trichloroethanc
Tetrachloroethylene
Chlorobenzene
Ethyl Benzene
Bromoform
m-Xylene
o-Xylene
1,1,2,2-Tetrachloroethane
1,3-Dichlorobenzene
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
200
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
55.6
9.9
14.0
16.1
14.0
13.8
11.5
10.6
10.2
12.0
17.5
13.8
10.4
12.1
8.3
12.3
14.8
10.2
8.2
12.1
12.9
5.4
12.9
12.9
8.1
9.3
withdrawn from the vial, leaving 30 ml of the
standard water solution and 10 ml headspace in the
vial. The vial was then placed into an ice-water
bath to equilibrate for about 30 minutes at 0" -
fC. After equilibration, the headspace was
analyzed on a portable GC. Four trials of analysis
were conducted separately. The data in Table 3
presents the analytical results from the portable GC
of selected VOC concentrations in the headspace.
(2). Preparation and GC/MS Calibration of
SUMMA canister standards:
Duplicate standards were prepared in 6L pre-
vacuumed Anderson made SUMMA passivated
canisters. The canisters were cleaned by vacuum
and heating. Canisters were vacuumed to < 5 mmHg
at about lOOt for 4 hours. 5.0 n\ of the stock
standard solution (Supelco purgeable A and B, and
EPA VOC repository) was injected into the each
canister. The canisters were then pressurized to 30
418
-------
Tible 3. Coaccntntiaa of Selected VOCs in Aqucouc HeadspaccC)
Compound
Benzene
Trichloroethylene
Toluene
Tetrachloroethylene
Ethyl benzene
o-Xylene
Concentration
Trial 1
146
163
166
212
128
98
(ppb/v)
Trial 2
161
129
153
229
154
122
Trial 3
161
141
191
177
134
95
Trial 4
135
134
127
187
115
47
Average
1S1
142
159
201
133
91
RSD%
7.3
9.2
14.5
10.2
10.6
30.1
(•): Above a 10 /ig/L water solution at 0°-1°C
psi with 25% relative humidity. The concentrations
of each VOC in the canister was 55.6 ng/1. Table
2 shows the VOC concentration of the canister
standards. Following a 24 hours equilibration period
at room temperature, the canisters were analyzed
on a Hewlett-Packard 5890 gas chromatograph
equipped with a 60 m megabore capillary column
and 5970 Mass Selective Detector. A Tekmar 5000
thermal desorber modified for canister analysis was
used for desorbing. The calibration results are
shown in Table 4. The canister cleaning and
analysis were performed according to EPA Method
Table 4. GC/MS Certified Concentration of Standard Canisters
Compound
Trichlorofluoromethane
(Freon 11)
Methylene chloride
Chloroform
1,2-Dichloroelhane
1,1,1-Trichloroethane
Benzene
Carbon tetrachloride
1 ,2-Dichloropropane
Trichloroethylene
Toluene
Tetrachloroethylene
Chlorobenzene
Ethyl benzene
o-Xylene
Calc.O
9.9
16.0
11.4
13.7
10.2
17.4
11.9
12.0
10.3
14.8
8.2
12.1
12.8
118
Concentration
10.7
12.2
9.7
12.0
9.3
15.7
10.0
11.3
9.1
13.6
7.3
10.7
12.5
11.4
(ppb/v)
Can A
± 1.3
± 2.0
± 0.6
± 0.9
± 1.1
± 2.6
± 1.0
± 1.2
± 0.9
0.7
0.6
0.7
1.6
0.9
GC/MS
Can B
11.4
13.6
10.9
13.0
10.7
17.4
11.2
12.2
10.4
14.8
8.1
11.5
13.2
11.7
±
*
±
±
±
+
±
±
t
+
+
±
*
+
2.6
2.9
1.3
1.7
1.7
2.6
19
1.9
1.3
1.6
1.0
1.6
13
15
(•): Calculated according to dilution.
TO-14 and EPA Region I draft SOP for ambient
air VOC analysis.
(3). Portable GC analytical results of canister
standards:
The GC/MS certified canister standards were
analyzed on a Photovac 10A10 portable GC which
was calibrated with the aqueous headspace working
standard of VOCs. Table 5 shows the analytical
results of benzene, trichloroethylene, toluene,
tetrachloroethylene, ethyl benzene and o-xylene.
419
-------
Table S. Comparison of GC/MS results versus portable GC on Canister Standards
Concentration (ppb/v)
Compound
Benzene
Trichloroethylene
Toluene
Tetrachloroethylene
Chlorobenzene
Ethyl benzene
o-Xylenc
CALC.
17.4
10.3
14.8
8.2
12.1
12.8
12.8
Can A
GC/MS
15.7
9.1
13.6
73
10.7
12.5
11.4
PGC
12.0
8.7
14.0
9.9
10.5
12.0
10.0
Can B
GC/MS
17.4
10.4
14.8
8.1
11J
13.2
11.7
PGC
11.6
8.7
14.1
10.1
9.0
12.0
13.0
(4).Comparison of aqueous headspace standards
versus canister standards:
The certified canister standards and aqueous
headspace standards were used for calibrating the
portable GC to analyze prepared air samples. Four
air samples of different concentrations were
prepared in SUMMA canisters and were certified
on GC/MS. The samples were then analyzed on the
portable GC. Table 6 shows the analytical results.
Based on different calibration standards, two groups
Table 6. Comparison of Aqueous Headspace Standards vc Canister Standards on Portable GC Analysis
Compound
Concentration (ppb/v)
Sample: #1 #2 #3
(MS)' AH Can (MS) AH CAN (MS) AH
#4
CAN (MS) AH CAN
Benzene
Trichloroethylene
Toluene
Tetrachloroethylene
Ethyl benzene
o-Xylcne
(27)
(16)
(23)
(13)
(20)
(20)
22
24
32
17
18
22
32
29
34
14
16
20
(14)
(8.1)
(12)
(6.4)
(10)
(10)
12
8.1
13
7.6
9.0
13
18
9.7
14
6.0
8.0
12
(6.8)
(4.1)
(5.8)
(3.2)
(5.0)
(5.0)
6.1
35
6.5
5.0
63
4.5
8.8
4.2
63
4.2
55
4.2
(2.7)
(1.6)
(2.3)
(1.3)
(2.0)
(2.0)
3.3
1.6
2.4
1.3
ND"
ND
5.0
2.1
2.4
1.6
ND
ND
(•): Results of GC/MS analysis
(*•): Non-detected
of data were obtained. The data in the columns
under "AH"(Aqueous Headspace) are the results
from the portable GC calibrated by the aqueous
headspace standard and those in the columns under
"CAN"(CANister) are the results from the portable
GC using the canister calibration. Comparisons
between the two data groups on benzene, toluene,
trichloroethylene, tetra-chloroethylene, ethyl
benzene and o-xylene were performed.
(5). Stability of canister standards and temperature
dependency:
The study of time dependent stability of
different manufactures' SUMMA canisters has been
reported[3]. The study in this paper focused on the
time dependent stability of benzene, toluene,
trichloroethylene, tetrachloroethylene, ethyl benzene
and xylenes. Two canister standards were analyzed
on a periodic basis using HP 5890/5970 GC/MSD.
With seven months of analytical results, there were
no significant variation of VOC concentrations in
the canisters. Figure 2 shows the time dependent
420
-------
Cone. (ng/L)
so
100 150
Time (day)
20O
260
Chloroform ~*~ Bvnzon*
Chlorob«nz«n«—*- o-Xyl«rM
TCE -B- Tolu
T*trachloro*thyl«na
Figure 2. lime dependent Stability of Canister Standards
stability for several selected compounds.
As a standard for field screening, the
temperature variation is an important factor for the
canisters. The temperature dependency study was
performed at various temperatures ranging from
-20C to 45°C. A canister standard and a duplicate
were stored at each temperature environment for
at least three hours, then analyzed on GC/MSD
Cone, (ppb/v)
14
12
10
8
6
4 -
2 -
-30 -20 -10 0 10 20 30 40 60
Temperature (Dg. C)
-•- TCE -•- Tolu.n.
-°- o-Xyl«n» -«- T«tr«chloro«thyl«n«
(a). Portable CC Analysis
Cone. (ng/L)
60
20
-30 -20 -10 0 10 SO 30
Temperature (Dg. C)
40 60
Chloroform -*- Blnan*
Chlorobinunt~»- o-Xylint
TCE -o- Tolinn.
Tttraohloro«lhyl«nt
(b). GC/NS Analysis
Figure 3. Temperature Dependency of Canister Standards
and on the portable GC. Figure 3 shows the
temperature dependency result of canisters. The
data shows those standards were very stable over
the temperature range of -20°C to 4Q°C.
Conclusions
According to both the theoretical and
experimental results, the aqueous headspace
standard is a suitable VOC standard for ambient
air field screening analysis. The field screening
headspace standard is easy to prepare with materials
that are readily available in any environmental
laboratory. It takes very little time to prepare and
the cost to prepare this type standard is minimal.
Canister standards possess high accuracy for
most of the VOCs and reflect real concentrations of
the VOCs inside. Canisters are easy to store and
transport, and are reusable. The temperature
dependency study on canister standards showed that
VOC concentration in canisters remain stable over
the normal field condition ambient temperature
range of -20^ to 4Q°C. The time dependent canister
stability tests showed that canister VOC standards
have long term stability. Over a seven month time
period, VOC concentrations in canister standards
remained stable.
Compared with aqueous headspace standards,
canister standards are relatively expensive. Because
of all the necessary accessary equipment needed for
standard preparation, this method is only
recommended to those laboratories which have a
canister analysis set up.
421
-------
Acknowledgment
The authors wish to thank the Regional
Laboratory of the U. S. EPA Region I, in which all
of the experiments were conducted.
Reference
[1].W. A. McClenny and K. D. Oliver.'Toxic
Monitoring with Canister-Based Systems",
APCA 80th Annual Meeting, 1987, New York,
New York, June 21-26, 1987.
[2]. A. R. Gholson, etc., "Evaluation of Canister for
Measuring Emissions of Volatile Organic Air
Pollution from Hazardous Waste Incineration",
JAPCA 39 1210-1217, (1989)
[3]. R. E. Berkley and K. Kronmiller, etc.,
"Performance Optimization of Photovac 10S70
Portable Gas Chromatograph" p849-854,
Proceedings of the 1990 EPA/AWMA
International Symposium, Raleigh, North
Carolina, May 1-4, 1990.
[4]. R. W. Harrell, W. J. Mitchell, etc., "Humidified
Canister Stability of Selected VOCs" p726-
730, Proceedings of the 1990 EPA/AWMA
International Symposium, Raleigh, North
Carolina, May 1-4, 1990.
DISCUSSION
THOMAS SPITTLER: I have to say that I saw this data just about two or three
days before I left for this meeting, and I was literally astounded at how well that
the simple, inexpensive, standard preparation correlated with the incredibly
expensive, complicated GC/MS canister technology in our lab. We didn't show
you half the slides he had of all the equipment required to prepare and analyze
the canister standard.
RALPH SULLIVAN: When you remove about ten cc's of liquid, did you back
fill with air, or make any provision to clean up that air as it went in, if you did that?
HUI WANG: No, that's just the same procedure as we prepared the headspace
standard for soil and water screening. We just pull the air out and just use it. We
have a mobile lab that's relatively clean. Is your question about the cross-
contamination from the air getting into the headspace?
RALPH SULLIVAN: Yes.
HUI WANG: The mobile lab is relatively clean. It has probably very, very low
levels of those target compounds. It's never going to affect our standard.
THOMAS SPITTLER: We've actually looked at the lab air in our building and
it'sascIeanastheoutsideair.Andthatmeansabout 1 ppb of benzene and toluene
and nothing else. We've never seen much need to put a big charcoal scrubber in
for that make-up air.
RALPH SULLIVAN: But you do put another hypodermic syringe in there to
dissolve the air?
THOMAS SPITTLER: Yes, the needle is just stuck in there. Room air just goes
in to replace the drawn water.
RALPH SULLIVAN: The next question has to do with the canisters. I saw
nothing in the diagram that indicated that you put any water into the canister. Did
you put water into the canister with a syringe?
HUI WANG: Yes, in many canisters there's about 25% relative humidity. I
calculated the amount of water we need and injected more water to the canister
directly.
RALPH SULLIVAN: Do you have mixtures of standards or single component
standards in the vials and in the canisters?
HUI WANG: Yes, composed of multi-components not only single components.
RALPH SULLIVAN: So, you added various quantities of say, neat compounds
into the vials and also into the canisters?
HUI WANG: Yes. Actually, I used the same stock solution. We use the same
stock solution for canister and headspace standards.
422
-------
Quantitative Soil Gas Sampler Implant for Monitoring
Dump Site Subsurface Hazardous Fluids
Kenneth T. Lang
Douglas T. Scarborough
U.S. Army Toxic and Hazardous
Materials Agency
Technical Support Division
Aberdeen Proving Ground,
Maryland 21010-5401
Mark Glover
D.P. Lucero
IIT Research Institute
Maryland Technology Center
4600 Forbes Blvd.
Lanham, Maryland 20706
ABSTRACT
In conjunction with a
triservice (Army, Navy, and
Air Force) program to develop
a cone penetrometer with
associated sensors and
detectors, a prototype soil
gas sampling system has been
fabricated and functionally
tested. The system, referred
to as TerraTrog,
quantitatively samples
hazardous soil gases and
vapors. TerraTrog can be
deployed by a cone
penetrometer to depths of 100
ft far less expensively than
drilling monitoring wells.
The device may be permanently
implanted or may be retrieved
and deployed at multiple
locations using the cone
penetrometer.
TerraTrog comprises two
modules: an implant of small
dimensions containing a
gas-permeable membrane of high
diffusion impedance (located
at subsurface levels) and a
sampling and calibration
interface with a pneumatic
manifold (located at ground
level). Unlike conventional
non-quantitative soil gas
sampling techniques requiring
vacuum to operate, TerraTrog
relies only on soil gas
diffusion for subsurface soil
gas collection and a carrier
gas stream flowing at a slight
positive pressure for lifting
the sample to the surface.
Because the sampling is
diffusion-limited by a
membrane of known impedance,
the sampling rate and sample
size are independent of soil
permeability. Sampling does
not deplete the local soil gas
or vapor, guaranteeing the
accuracy of measurements made
with the device even after
long periods of continuous
sampling. The system has a 15
rain. maximum
time-rate-of-response.
Functional and
performance testing has been
performed with
trichloroethylene in soil,
water, and air, using a
Photovac 10S70 portable gas
chromatograph. The implant
has been demonstrated to
operate as designed, i.e., is
diffusion-limited with implant
response directly proportional
to external soil gas partial
pressure.
INTRODUCTION
A major problem in the
cleanup process or assessment
of sites contaminated by
hazardous waste and toxic
chemicals stems from the
paucity of information
regarding site subsurface
characteristics, composition,
and aerial and volumetric
extent. Performing a general
prospecting or screening
423
-------
survey of the site hazardous
fluids and their mobility or
stability is of significant
value in developing
preliminary overall
containment and treatment
plans (1). A network of
relatively low-cost implanted
soil gas samplers deployed
throughout the site vadose and
peripheral zones as well as
adjacent aquifers and high
permeability strata can be
utilized effectively for site
prospecting and
characterization. The notion
of an implanted sampler
network is a viable concept
only if waste characterization
data can be provided quickly
and inexpensively and if the
sampler can provide samples of
all hazardous soil fluids and
contaminants and can interface
at the dump site with a
variety of analyzers or
monitors and secondary
samplers.
This soil gas sampler
system, called TerraTrog for
easy reference, is described
below and addresses the above
requirements satisfactorily,
offering features that promote
simple, low cost sampler
deployment; minimal soil
disturbance from deployment;
minimal sample extraction
during each sampling episode,
providing a correspondingly
more representative sample of
soil gases; minimal hardware;
and small dimensions. The
TerraTrog implant has a 1-in.
lateral dimension and can be
deployed by cone penetrometers
available commercially (2).
In addition, sampler operation
is independent of the soil
permeability over a range of
0.1 to 1000 mD, and therefore,
quantitative data are obtained
for sandy as well as clay soil
types. These operational
features also render the
sample obtained independent of
sampling chamber volume, line
length, sampling pump head,
and corresponding pressure
losses.
An important consequence
of using implants in the
initial prospecting process
and then progressing to the
characterization and
monitoring phases is that the
network of permanent implants
deployed initially can be used
for the life of the dump site
cleanup and monitoring tasks.
Thus, an implanted sampler is
a very attractive concept for
long-term site monitoring
requirements.
THE SAMPLER IMPLANT SYSTEM
TerraTrog comprises two
modules: the subsurface
implant and the surface
control interface. Figure 1
illustrates the system. Soil
gases enter the implant at
flow rates proportional to the
individual gas partial
pressures and the partial and
vapor pressures of dissolved
and pure liquids,
respectively, regardless of
the soil permeability. The
soil gases are lifted to the
surface by the carrier gas
stream, which enters the
surface module and flows at a
controlled and measured flow
rate to and through the
implant and returning to the
surface as shown. Soil gas
analysis and monitoring is
accomplished by the analyzer
or monitor attached to the
carrier gas stream return line
at the interface. The
SUBSURFACE
lUPUNT
FIGURE 1 - TEHRATROG SOL GAS SAMPLER SYSTEM
424
-------
analyzer/monitor and carrier
gas used are compatible with
all aspects of TerraTrog and
the data quality requirements
of the application. In situ
calibration of the implant is
performed with an innocuous or
surrogate gas, which is
carried to the implant by the
calibration gas line. During
one calibration episode, it is
anticipated that less than
lOOO^il of calibration gas is
injected directly into the
soil surrounding the implant.
In addition, a secondary
sampling device (grab bag,
bubbler, etc.) may be attached
to the interface, and soil gas
may be collected in batches
for subsequent laboratory
analysis. With a sufficiently
large carrier gas stream flow
rate, one or more
analyzers/monitors and/or one
or more secondary samplers can
be attached to the carrier gas
outlet of the interface and
can be operated concurrently.
Implant
The cross-sectional
illustration of the implant
(Figure 2) depicts a
cylindrical array of eight
metal rods approximately 6 in.
long, contained within a 1-in.
diameter envelope. These are
surrounded externally by a
1-in. diameter, 0.002-in.
thick Teflon tubular membrane.
The rods provide mechanical
support for the tubular
membrane.
Both ends of the tube are
sealed from the surrounding
soil by 0-rings and a top and
bottom header. A sealing cap
compresses the tube and 0-ring
into a groove on the bottom
header. A nut and the cone
tip maintain the sealing cap
in place as shown. Carrier
gas is introduced to the tube
interior and returned to the
surface through the top
header, which also serves as a
gas manifold. There is no
pneumatic communication
between the calibration gas
and the implant interior or
the carrier gas. Calibration
gas enters the header and
flows directly to the
periphery of the calibration
gas diffuser cap where it is
injected into the region
external and adjacent to the
tubular membrane. The
calibration gas diffuser cap
also serves as the sealing cap
in an identical fashion as the
bottom sealing cap. It is
maintained in place by a metal
gasket and nut. The thread
sealing gasket ensures that
calibration gas does not leak
through the threads shown.
Surface Control Interface
The surface control
interface module comprises a
panel attached to a metal
stake embedded in the soil. A
pneumatic gas control network
is mounted to the panel back
side. Gas connections are
made through fittings that
lead to the carrier and
calibration gas supplies and
the respective pneumatic lines
to the implant. The carrier
gas return line connects to a
manifold for an
analyzer/monitor and/or for
secondary sampling devices.
All gas connections are
made at the panel face. All
pneumatic lines contain inline
filters. The carrier gas flow
rate is controlled by a
precision pressure regulator
and flow adjustment valve. A
0 to 60 psig pressure gauge
measures the regulated
pressure, and a rotameter
measures the carrier gas flow
rate. Calibration gas flow
rate is controlled by the gas
supply pressure regulator and
a flow-limiting orifice in the
surface module network. Each
gas line contains inline
pressure relief and shutoff
valves downstream to prevent
overpressurization of the
implant and to assist in the
startup and checkout process.
When not in use, the
sample ports are capped or
plugged. Note that no
electrical power is required
to either maintain or operate
the interface or the implant
as described. All power
requirements are associated
425
-------
with the analyzer or monitor.
PRINCIPLE OF OPERATION
Implant operation is
based on a flow of soil gases
by diffusion through the
semi-permeable tubular
membrane of Figure 2 (3-7).
In addition, the soil gas flow
rate is diffusion limited by
the membrane and consequently
independent of the soil
permeability. As carrier gas
flows through the implant, the
concentration of the soil gas
species at the surface is a
ratio of the two gas flow
rates:
= (QS/QC)
10-
(1)
where
[G] = soil gas species
concentration in the
carrier gas stream at the
interface module, parts
per billion (ppb, v/v);
QS - soil gas species
Gamer Gas Out
Carrier Gas In
Calibration Gas n
Top Haider and
Gas Manifold
Cafibratjon Gas
DiffuSK- Can
CaBbration Gas
Cut
Membrane Tubs
Gamer Gas TuCe
/ Bottom Header
flow rate into the
implant, std ml/min; and
QC = carrier gas
flow rate, std ml/min.
The carrier gas flow rate
is measured at the surface
interface module. The soil
gas species flow rate is the
product of the soil gas
species membrane conductance
and partial pressure in the
surrounding soil. By lumping
the membrane and carrier gas
parameters into the term, y,
the soil gas partial pressure
is related to [G] as follows
(6,7):
(2)
where
sg
7 [G]
Rgire 2 - Sampler Implant
PSO = soil gas species
partialypressure in the soil,
torr.
The system response time
is the sum of the time to
saturate the tubular membrane
with soil gas to an
equilibrium concentration, the
carrier gas lag time in the
pneumatic lines, and the time
required from startup to
establish carrier flow through
the interior gas volume of the
implant to the condition where
equilibrium concentration is
established.
For an implant with a
1.000-in. Teflon tubular
membrane, 0.002-in. thick, and
a soil gas diffusion
coefficient of 10 cm /s (8),
the time to saturate the
tubular membrane is
approximately 128 s. The lag
time will depend on the inside
diameter of the carrier gas
pneumatic lines, the depth of
the implant, and the carrier
gas flow rate. For TerraTrog
operating with a 50-std ml/min
carrier gas flow rate stream,
0.0625-in. inside diameter
pneumatic lines, and an
implant 50 ft. below the
surface, the lag time of the
system is 73 s. Approximately
7.7 min are required to
exchange five implant gas
426
-------
volumes at a 50-ml/min carrier
gas flow rate. It is certain
that the tubular membrane will
saturate shortly after
deployment of the implant and
long before carrier gas is
flowing and thus contributes
insignificantly to the system
response. Thus, the time
required to obtain an
equilibrium reading at the
ground surface from startup is
approximately 10 to 11 min for
the conditions listed above.
Equations 1 and 2
describe the soil gas species
concentration at the surface
interface for TerraTrog
operating in the dynamic
sampling mode, i.e., the
operating mode in which the
carrier gas flows continuously
through the implant. The
implant can also be used in
the static sampling mode,
i.e., the operating mode in
which the carrier gas does not
flow (Q = 0) for a prescribed
period of time preceding
dynamic sampling but flows
only after the equilibrium
condition described below is
attained. Note that soil gas
flow into the implant will
continue, regardless, until
the soil gas partial pressure
difference across the tubular
membrane is 0. At this point,
the net flow of soil gas into
the implant is 0, and an
equilibrium soil gas
concentration internally and
externally of the tubular
membrane is obtained. After
this equilibrium is attained,
the carrier is used to lift
the soil gas accumulated in
the implant.
For the initial
condition, where the soil gas
partial pressure, P ,
measured at the surface
interface is 0 (P-_» the
sampler implant intergral soil
gas partial pressure is 0),
the time required to obtain
the static equilibrium
condition with the soil gas
pressure, i.e., P = P is
7.22 days (6,7) f§? theig
implant dimensions listed
above and a soil gas
permeability coefficient of 2
x 10~ std ml/min-cm -torr/cm
(9).
P. will be less than P
at equilibrium because of g
mixing and subsequent dilution
of the soil gas accumulated in
the implant by the carrier gas
stream. It is estimated that
five implant chamber volume
exchanges with carrier gas
will be required to remove the
soil gas accumulated in the
implant and transport it to
the surface. Assuming
homogeneous mixing, the
average soil gas concentration
or partial pressure measured
at the surface will be
one-fifth the soil gas partial
pressure during the five
volume gas exchanges; at a
carrier gas flow rate of 50
std ml/min/ the time period is
approximately 7.7 min.
In situ calibration of
the implant is performed with
calibration gas supplied to
the external surface of the
tubular membrane. A
calibration gas stream enters
the calibration gas inlet,
flows through the top header
and gas manifold to the
calibration gas diffuser cap
and through the holes in the
cap to the external surface of
the tubular membrane. The
implant operates on the
calibration gas as it does on
soil gas.
DESIGN AND OPERATIONAL
CONSIDERATIONS
Aside from fundamental
system analytical and
monitoring performance
requirements, the system
design constraints are
established by reliability and
service life requirements and
deployment flexibility.
TerraTrog reliability
corresponds generally and most
importantly with the
exigencies of maintaining the
relationship of soil gas
species partial pressure, P ,
and the measured soil gas sg
species concentration, [G],
described by equation 2.
Adherence to this relationship
is predicated on the design
427
-------
and operational integrity of
the tubular membrane and the
pneumatic lines leading to the
surface. It is essential that
the soil gas flow into the
implant by a diffusion process
only, and therefore, the
tubular membrane must be free
of tears, punctures, and pin
holes and other pneumatic
leaks. Thus, pre- and
post-assembly inspection of
the tubular membrane as well
as an implant leak check is
required. The tubular
membrane must not be damaged
during the deployment and
operational processes.
Relatively inert implant
fabrication materials, e.g.,
stainless steel and Teflon,
are used, because after
deployment, every external and
perhaps some internal
component or surface will be
exposed to chemical and
physical attack. It is of
prime importance that the
tubular membrane material be
chemically and physically
inactive with the soil and
with benign as well as
hazardous soil fluids to
ensure that the tubular
membrane material diffusion
conductance (3-5) remains
unchanged during the life of
the implant. Teflon of any
form is regarded as the most
suitable material for the
tubular membrane. The
membranes currently in use are
fabricated of 2 mil (0.002
in.) thick Teflon film.
The maximum typical soil
gas sample flow rate into the
implant is approximately 0.01
std uL/min for arbitrary but
realistic conditions. In a
relative sense, it is a very
small sample, yet large enough
to produce a [G] for many soil
gas species within the
response range of many gas
phase analyzers/monitors that
may be attached to the
interface.
There are three important
aspects to the relatively
small sample size or flow
rate: First, the disturbance
to the soil is minimized;
consequently, a more
representative sample is
obtained independent of soil
fluid conditions. Second, for
soil strata in and around dump
sites, the soil gas flow rate
into the implant is diffusion
limited by the tubular
membrane and is independent of
the gas permeability of the
surrounding soil. Thirdly and
most importantly, these
conditions lead to a
quantitative measurement of
the soil gas partial pressure.
The soil gas flow rate is
proportional to the soil gas
species pressure only, without
regard to the form of the
sample, i.e., gas phase,
liquid phase, or dissolved
gas/liquid phase. For
example, the implant can
obtain information regarding
dissolved trichloroethylene
(TCE) in water or TCE
saturated in water, and
insoluble gases contained in
the water. Furthermore, the
implant also functions as
described immersed completely
in an aquifer or other body of
water or liquid.
SOIL GAS ANALYSIS AND
MONITORING
In a relative sense, the
actual soil gas monitoring and
analysis of the transport gas
output stream from the
interface panel is the most
simple and direct procedure of
the entire system. A variety
of analyzers, monitors, and
secondary sampling devices can
be used singularly or
simultaneously. The user,
however, must establish
preliminary requirements for
the target species and the
lower detection limits of the
analytical devices
contemplated, i.e., it is
essential to consider the
analyzer/monitor performance
specifications to specify and
adjust the operating
conditions of TerraTrog.
TERRATROG PERFORMANCE
TerraTrog response to
trichloroethylene (TCE) in
soil, water and air has been
characterized in the
428
-------
laboratory. The test
equipment was arranged in a
configuration identical to
Figure 1, with the implant
suspended in a specially built
test vessel. In separate
tests, soil, water and air
with measured concentrations
of TCE were contained in the
vessel to simulate implant
field deployment conditions.
The vessel was pneumatically
sealed to prevent loss of TCE
vapors except by diffusion
through the implant gas
permeable membrane and
subsequent removal on the
implant carrier gas stream.
The response
characteristics of the implant
were determined by monitoring
the concentration of TCE
within the test vessel
external to the implant and by
monitoring the carrier gas
entering and exiting the
implant in dynamic sampling
mode, and by direct analysis
of the implant contents in
static sampling mode. In all
tests, it was verified that
the concentration and
therefore partial pressure of
TCE in the soil, water or air
external to the implant
remained stable and constant
during the period of the test.
All concentration measurements
were made on a Photovac 10S70
portable gas chromatograph
with a
10-m CPSIL5CB capillary column
and a photoionization
detector. The gas
chromatograph was calibrated
using air standards prepared
from aqueous solutions of TCE
of known concentration (9).
Representative data for
the TerraTrog
time-rate-of-response in
static sampling mode to
dissolved phase TCE in water
is shown in Figure 3. The
internal concentration reaches
equilibrium with the external
TCE concentration in 7 days
(168 hr.). This is in
excellent agreement with the
calculated equilibrium
response time, 7.22 days
(6,7).
Representative dynamic
sampling mode data for the
I©
SMpLi
An«lyt»r;
004-2021
TCE 20 ppb- (20 u»/l)
IB W«t«r. 22 C
Static
Fbot«vm« 10370
Figure 3 Implant Tim*-R4t*-of-Reiponse
to TCE In W.tar
implant response, [G], to TCE
as a function of inverse
carrier flow rate, 1/QC/ is
shown in Figure 4. A
multipoint calibration curve,
Figure 5, shows the
relationship between implant
response, [05] / and external
dissolved phase TCE
concentration in water.
Figure 4 show the implant
response is linear regardless
of whether the implant is
deployed in soil, water, or
air. Additionally, the flow
rate of TCE into the implant,
Q , is constant when the
implant is sampled in dynamic
mode in an environment of
constant external TCE
concentration. The sample
flow rate, Q , is relatively_5
small, ranging from 9.2 x 10~
std./Jtl/min for dissolved
phase TCE at 100 ppbm external
concentration in water, to 3.7
xlO std. ul/min for gas
phase TCE at 227 ppmv external
concentration in air. Figure
5 shows that the implant
response, and hence Q , varies
linearly with the extlrnal TCE
concentration and therefore
with the external TCE partial
429
-------
pressure. The data
demonstrate that Q is
dependent only on the
permeability, P , of the
implant gas permeable
membrane, and the external TCE
partial pressure, or
concentration (6,7).
Therefore, the implant
operation is diffusion-limited
by the implant gas permeable
membrane and the implant
response is directly
proportional to the external
TCE partial pressure in soil,
water, or air, exactly as
T.rrtTrcu StH: 003-2001
EnviranMnt: TCZ 21 ppH
Sail rhea.
S«*plln| Hod.: DrnMuc
Carrier CM: Air
Analyser: Photov.i
Figure 4 In-plane Response Co TCE In Soil
TunTiai S/tt: 003-2001
Envlrotwrnt: TCE In H«t«t. 22 C
Con««ntr*tion •• Net
SMpUn« Hod*: Dynmic
Carrlvr 0«f: Air
Cfttift OM Flow UK*: 40 iU/.un
An«lyt«r: PhetevH 10S70
' Inplwit C«libration Fttctor K:
O W.t*E TCE Co«c4>ntr«tlon
. / lopltint R»ipon.», |C]
*2.92 i
300 030 600
VlattiT TCE CtMfttMtrctlM., ppb*>
described by equations 1 and
2.
The multipoint
calibration curve, Figure 5,
can be used with the implant
to directly measure the TCE
concentration in contaminated
groundwater in the field. For
example, a user would deploy
the implant to the desired
depth in a monitoring well or
other body of water and
establish a carrier flow rate
of 40 ml/min. The implant
response would be measured
using a conventional gas-phase
TCE analyzer, such as the
Photovac 10S70. The implant
response would then be located
on the vertical axis of the
calibration curve, and the
corresponding TCE
concentration in the
groundwater read off the
horizontal axis of the curve.
Since the implant
response is shown to be linear
in the multipoint calibration
curve of Figure 5, it may be
replaced with a single point
calibration which yields a
linear calibration factor, K:
[Ce]/[G]
where
Figure 5 calibration Curve for Inplanc Response
Co TCE In Uaccr versus TCC Concencration
K = implant linear
calibration factor, ppbm/ppbv
[C 1 = external TCE
concentration, ppbm
[G] = implant
response, ppbv
The linear calibration
factor may be used exactly as
the multipoint calibration
curve to make direct field
measurements.
TEST PLANS
Laboratory and Controlled
Field Testing
Further laboratory and
field testing is in progress
at the National Institute for
Petroleum and Energy Research
(NIPER). Laboratory testing
is planned to demonstrate the
quantitative TerraTrog
sampling characteristics over
a range of controlled soil
430
-------
permeability from 0.1 to 1500
mD.
TerraTrog will be
deployed in soil by a cone
penetrometer to depths
approaching 50 ft to develop
deployment procedures and
techniques, optimize the
pneumatic line dimensions and
configuration, and determine
the effects of the surrounding
subsoil mass on the implant
operational integrity.
Optimum deployment procedures
will be developed regarding
the mechanical aspects, the
pneumatic tubing, and grouting
and sealing the bore hole. In
addition, the in situ
calibration scheme described
above will be implemented to
develop optimum calibration
procedures. This work will
determine the utility and
validity of the in situ
calibration. The TerraTrog
performance after calibration
will be assessed with
calibration gas injections.
CONCLUSION
The development of the
TerraTrog is viewed as having
real potential for future use
in the evaluation of hazardous
waste sites. The potential
utility of the device includes
not only initial site
assessment, but possibly of
more importance, its use in
the routine monitoring that is
essential to the long term
assessment of a site before,
during and after remedial
activities are accomplished.
Although initially designed to
be used in a cone
penetrometer, the utility of
the device for routine
groundwater monitoring is also
recognized due to its small
diameter and ability to
descend down standard well
casings.
ACKNOWLEDGEMENTS
The authors thank Susan
Hendrickson of IIT Research
Institute for valuable
technical critique and editing
of this manuscript.
Funding for the initial
phase of this work was
provided by the Program
Manager for Rocky Mountain
Arsenal, Gregory Mohrman,
Project Monitor.
REFERENCES
(1) S.C. Golian, E.E. Dodge, and
B. Bixler, " Conducting
Remedial Investigations on
Feasibility Studies Under
CERCLA," Superfund '88
Proceedings of the 9th National
Conference. Hazardous
Materials Control Research
Institute, Silver Spring, MD,
p. 1, 1988.
(2) L.R. Taylor and N. Berzins,
"Subsurface Contamination
Screening by Combined Soil
Gas/Ground water Survey
Procedure," Superfund '88
Proceedings of the 9th National
Conference. Hazardous
Materials Control Research
Institute, Silver Spring, MD,
p. 158, 1988.
(3) R.M. Barrer, Diffusion In and
Through Solids. Cambridge
University Press, London,
Chapter 1, 1981.
(4) D.P. Lucero, "Performance
Characteristics of Permeation
Tubes," ApfllyHnal Chemistry
43. p. 1744, 1971.
(5) D.P. Lucero, "Ultra Low-level
Calibration Gas Generation by
MuM-Stage Dilution
Techniques," Ca1i^raH"n in AIT
Monitoring, ASTM STP 598,
American Society for Testing
Materials, Philadelphia, PA,
pp. 301-319, 1976.
(6) D.P. Lucero, "Soil Gas Sampler
Implant for Monitoring Dump
Sltp Subsurface Hazardous
Fluids," Proceedings of the 6th
National Conference on
Hazardous
Wastes
and
Hazardous
Materials.
Orleans, Louisiana,
12-14, 1989.
New
April
431
-------
(7) D.P. Lucero, "A Soil Gas
Sampler Implant for Monitoring
Dump gjt-g Subsurface
Hazardous Fluids," Hazardous
Mafr-pTi'als Control _3, 5, p. 36,
1990.
(8) D.P. Lucero, "Performance of
Membrane-Covered Polar! graphic
Detectors," Analy
Chemistry 40. p. 707, 1968.
(9) T. Spittler, unpublished notes
of December 6, U.S.
Environmental Protection
Agency Regional Laboratory, 60
Westview Street, Lexington,
MA, 1989.
DISCUSSION
ROBERT LUGAR: With the TerraTrog probe, have you considered how to
avoid cross-contamination between holes orcarrying it toa lower depth if you're
doing a depth profile?
KEN LANG: We've considered that a potential problem, but since we don't
have any experience with actually pushing it into the ground yet, we're not sure
whether or not it is going to be a problem, and as such we're not going to try to
engineer that problem out until we see whether or not it does happen. I really don't
think it's going to be a problem because, as was presented in the Navy's work,
on some sites that have fairly high contamination, the device seemed to be rather
self-cleansing as it was pushed down in through the soil. They made that
determination based on the sapphire window, and the fact that their readings of
contamination dropped off rather rapidly. And we're hopeful that that's going to
happen with this device. The membrane does not come in direct contact with the
soil, so I'm hoping that that will not turn out to be a problem.
SKIP WEISBERG: When you inject the carrier gas in to drive the soi 1 gas out,
do you then feed that directly into the GC, or do you add an additional
supplemental carrier?
KEN LANG: The testing that we've done so far involves feeding that carrier gas
directly into the gas chromatograph. We may find out later on, especially if our
aim is toward reducing detection limits, that there may be other things that we
have to do. Calculations suggest that in order to cleanse the chamber, the interior
of the membrane, you would want to get at least five volume changes. I think we
calculated that the internal volume of that is somewhere around 80 cc's. So, we
already believe that for some instrumentation we're going to have to preconcentrate
that material before it's actually introduced into the analytical instrument.
SKIP WEISBERG: I was wondering if you would suffer a dilution in loss of
sensitivity due to the injection of a carrier gas down into the soil?
STEVE KNOLLMEYER: I was wondering if the tenon membrane acts to
cause certain molecules to diffuse more through the membrane and retain some
of them outside just because of their molecular size?
DANIEL LUCERO: Yes, the conductance of the membrane varies with the
analyte. For example, benzene has a higher permeability coefficient than does
TCE. And of course what that means is that the system has to be calibrated if you
want accurate results. The variations are not great for organic molecules. But if
you want more accurate results, then you do have to calibrate. And if you implant
the device and start seeing a gas that you haven't calibrated against, then that was
the reason for the inside calibration ports that you have in there. There are no
holes in the membrane; it's not a porous membrane. It works by solution/
dissolution. And as such there is a difference in impedance from analyte to
analyte.
THOMAS SPITTLER: Have you ever tried to push the cone penetrometer
through glacial till very far?
KEN LANG: No, we haven't pushed the penetrometer at all. That's the next step.
THOMAS SPITTLER: I don't like to be rough about it, but I think you're in
for a shock.
KEN LANG: We have not pushed the penetrometer with the TerraTrog on it. We
have made lots of pushes of the penetrometer with other devices on it, but not in
glacial till. There are obviously some limitations. One that would come to mind
right off would be once you hit bedrock that's as far as you go. Chert is also a
problem, and we have tried to push through chert materials and it doesn't work
there, either. So, there are some limitations of the cone penetrometer that we're
already aware of.
432
-------
TUNABLE CO2 LASER-BASED PHOTO-OPTICAL SYSTEMS FOR
SURVEILLANCE OF INDOOR WORKPLACE POLLUTANTS
Harley V. Piltingsrud
National Institute for Occupational
Safety and Health, 4676 Columbia Parkway
Cincinnati, Ohio 45226
Disclaimer; Mention of company names or
products does not constitute endorsement
by the National Institute for
Occupational Safety and Health (NIOSH).
Abstract
This paper describes research work
recently conducted at the National
Institute for Occupational Safety and
Health (NIOSH) on the use of CO2 laser
lidar for surveillance of indoor
workplace pollutants. Long-term goals of
this program included an ability to
real-time map the spatial distributions
of a variety of pollutants in a workplace
atmosphere. In pursuit of this goal, a
rapidly-tunable, Q-switched, low pressure
laser was developed for time-of-flight
aerosol backscatter differential
absorption measurements. This enabled
effective atmospheric freezing as well as
the evaluation of possible methods for
suppressing the severe multi-reflection
scatter problems common in work
environments. Experiments were performed
to determine the practicality of
utilizing aerosol backscatter methods in
the desired lidar system. The results of
the experiments indicated that present
practical technology did not support such
a methodology. Other special hardware
requirements of a field-deployable lidar
system were explored, including
high-speed, high-sensitivity detector
systems, and miniature detector
cryocoolers.
Introduction
During discussions held at the National
Institute for Occupational Safety and
Health (NIOSH) on research needs for gas
and vapor monitoring, industrial
hygienists and instrumentation
specialists alike voiced a need for
portable surveillance equipment to map
concentration distributions of selected
vapor phase chemical pollutants in the
workplace atmosphere, on a real time
basis. It was suggested that this should
be some type of optical instrument,
requiring only line of sight from it to
the areas to be monitored, investigating
possible approaches to this task, a
number of recommendations regarding the
desirable characteristics of a gas and
vapor surveillance system were gathered.
These recommendations indicated that an
ideal system should: (a) be portable
(movable into various workplace settings
for surveys etc.), (b) be single-ended
(no use of retro-reflector arrays or
transmitters and receivers separated by
large distances), (c) give real-time
mapping of the spatial distribution,
concentration and identity of a variety
of airborne pollutants commonly found in
the workplace, (d), not present any
hazard to the people in the workplace
(from laser radiation, chemicals, liquid
nitrogen, etc.), (e) require minimal
maintenance (some routine service perhaps
every several months), (f) have a
sensitivity on the order of 10-100 parts
per billion (over sampling distances of
five meters) for some pollutants, (g)
have a range up to a few hundred meters,
(h) be automated and programmable for
unattended data gathering, (i) have a
spatial resolution of at least five
meters full-width-half-
maximum (using a point source), and (j)
433
-------
survey the workplace at least every five
minutes. In reviewing previous research
in this field, two principle
methodologies were identified: Fourier
transform infrared (FTIR) spectroscopy,
and laser-based light detection and
ranging (lidar) systems.
Fourier-Transformed Infra-red
(FTIR) Systems
Open-beam, long-path FTIR systems appear
to be gaining much attention, and showing
at least a limited potential for
qualitative assessments of atmospheric
gas and vapor contents for toxic waste
site monitoring, as well as fence-line
monitoring. Such a system has been used
by Herget for various outdoor studies'1'2'
as well as an workroom _study in an
aluminum refinery potroom.
(3)
Other investigators have used FTIR
equipment similar to that used in
analytical laboratories, for a "real
time" analysis of gaseous pollutants in
the workplace by use of a multi-port
sampling apparatus with sample lines
running from ,work locations to a central
FTIR system.
(4.5)
FTIR has several advantages over
alternate methodologies including: (a)
high sensitivity, (b) high specificity,
(c) a large variety of chemicals
detectable and identifiable, (d)
technology is well tested, based on
similar laboratory use, (e) proven field
ruggedness and dependability, (f) no
support requirements other than
electricity (if detector cooling is not
used), (g) commercial availability, (h)
an inherent self-calibration of certain
parameters of the equipment, (i) "real
time" analysis capability, and (j)
portability of the apparatus, making its
use for temporary applications feasible.
Possible problems associated with this
approach include: (a) a lack of an
adequate gas phase high resolution
spectral library when peak identification
methods are used to identify unknown
compounds, (b) likely problems with
chemical interferences in the analysis,
(c) a lack of adequate spatial resolution
for some applications, such as measuring
breathing zone chemical concentrations,
(d) a requirement for either a two-ended
system or one using retro-reflectors, (e)
a lack of adequate scan speeds to achieve
"atmospheric freezing" and, (f) a
reduction in the systems sensitivity due
to atmospheric water vapor.
Currently, a Nicolet Instrument Co.,
Madison, Wisconsin, custom portable open
beam path FTIR system is being evaluated
for industrial hygiene applications by
the University of Michigan School of
Public Health (S. Levine) and the
University of California, Berkeley (R.
Spear). This work is being supported
by a NIOSH grant #1-\R01-OH02666-01. The
Bomem Inc., Quebec, Canada, Model DA2
open beam path FTIR system is currently
being used for a variety of atmospheric
pollution studies by M.^Spartz at the
Kansas State University.'"
(7)
Laser Remote Detection Systems
Laser systems have been quite useful in
making remote measurements of chemical
contents of the atmosphere for a variety
of purposes and using several methods.
Most remote photo-optical pollution
detection work has been accomplished
using light detection and ranging (lidar)
methods where laser produced radiation
was transmitted into the atmosphere to be
measured and the return radiation (either
aerosol scattered or terrain or
retroreflector reflected) is analyzed to
obtain the desired information about the
chemical contents of the atmosphere.
This has been accomplished by a variety
of methods'8' including:
a.
The analysis of Raman scattered
radiation: In this method, Raman
scattered short pulses of laser
radiation (usually at ultraviolet
(UV) and blue wavelengths) are
analyzed for their photon intensity
vs. wavelength distributions as a
function of the time from the
emission of the radiation pulse from
the laser. This results in the
identification and quantification of
the pollutant of interest as a
434
-------
b.
function of distance from the
transmitter. The return signal is
relatively weak due to the small
Raman scattering coefficient. This
limits the range of this technique
to a few hundred meters, even using
very powerful lasers. Its
time-of-flight (the time between a
transmitted laser pulse and the
return signal pulse, indicating by
the time differential what the
distance was between the transmitter
and a particular scattering or
reflecting location) position
resolution (ranqinq) can be on the
order of 10 meters. This approach
has been used in many
investigations, usually to measure
pollutants in exhaust plumes from
factories or power plants. Raman
lidar has the advantage of being
applicable to a large variety of
pollutants. However, the spectral
resolution and bandwidth of the
Raman signal is relatively broad
leading to significant interferences
in complex mixtures of pollutants.
Interference often results from
sunlight or other bright lighting.
Principle limitations are a severe
lack in sensitivity of the technique
(a very weak Raman return signal)
requiring the use of very powerful
laser systems which are typically
not "eye safe" and would be very
difficult to make "eye safe" while
preserving the system performance.
Also, spatial resolution has been
poor due to the poor data statistics
of the return signal, and the near
field interferences in time of
flight measurements.
Resonance Absorption Methods: These
methods depend on transmitting a
laser beam having a wavelength very
close to that of a resonance
absorption line of a chemical
species being measured, and
measuring a return signal reflected
off of an object or from aerosols in
the laser beam's path. It is
generally a much more sensitive
method then that of Raman scattering
due to the much higher scattering
coefficient for Mie scattering.
This increased scattering
coefficient could result in an
extended range, the use of lower
powered lasers, the use of less
sensitive detectors, better range
resolution, and faster measurements
than Raman lidar. It has the
disadvantage of being able to
measure only one pollutant at a time
when only one laser wavelength is
used. Unresolved interferences can
result when only one transmitted
wavelength is used. One version of
such a system uses differential
absorption lidar (DIAL) whereby the
laser radiation wavelength is
shifted between two (or more)
wavelengths, one of which is very
near to the resonance absorption
line of the pollutant being measured
and another line which is not. The
ratio of these measurements is used
to normalize the measurement system
calibration with regard to the
scattering characteristics of the
atmosphere being studied. By
measuring the return signal
intensity for successive locations
along the beam path, subtraction
yields the signal loss due to the
incremental decrease in the observed
signal associated with a range
increment AR, and arising from the
attenuation of the specific
molecular constituent for which the
laser is tuned. This is described
by the following equation:
(8)
4E0. - [BU0.R) -
AR) ] - tE(Xu,R) -
AR) ]
where:
AE0« = the incremental decrease in
the differential signal E(Ao,R) =
the return signal energy at range R
and wavelength X0, representing a
maximum absorption by the chemical
of interest.
E(XH,R) = the return signal energy
at range R and wavelength A.u,
representing an absorption off the
resonance line of the chemical of
interest.
435
-------
Using conventional timing methods,
DIAL systems can have ranges of
several km, and spatial resolutions
of as small as 10 m.
hanging Methods
Principal methods used for position
determination (ranging), include: (a)
time-of-flight, (b) triangulation ranging
(such as methods used by the U.S. Bureau
of Mines for measurement of methane gas
concentrations at coal seams in mines<9)),
and (c) the use of variable focal
distance optical systems. Variable focal
distance optical systems have been used
successfully in laser doppler wind
velocity measuring apparatus using CO2
lasers and heterodyned detection methods,
and have produced spatial resolutions of
approximately 10 m at a range of
approximately 100 m, with maximum useful
ranges of up to 1 km. However, this
method requires a coherent detection
system.<10> Practical problems have been
encountered with these systems due to
their instabilities and engineering
difficulties, including frequency
instability, harmonic generation, phase
shifted echoes and loss of wavefront
parallelism. In addition, there are
significant problems associated with the
application of coherent detection systems
with rapid wavelength shifting, when the
parallel retuning of both the
transmitting laser and a local oscillator
laser must occur or when optoacoustic
wavelength shifting of part of the
transmitter signals is used as the local
oscillator. In addition, coherent
detection is probably not practical for
short distance time-of-flight
measurements (believed to be necessary
for our application) due the heterodyne
frequencies required to achieve a 30 ns
timing resolution. This is especially
significant when considering optoacoustic
wavelength shifting.
Laser Types
Many laser systems exist for lidar
applications, from extremely
short-pulse-length, high-energy/pulse
systems to continuous-emission, low-power
devices covering wavelengths from the far
ultraviolet to the far infrared. If a
very rugged, low maintenance, compact
system is needed, tunable over a range of
wavelengths, the number of available
lasers becomes quite small. Possible
candidates include dye lasers, solid
state diode lasers and C02 lasers. The
dye lasers suffer from rather narrow
tuning ranges for a specific dye,
moderate to high maintenance requirements
and moderate to large sizes. They
operate mostly in the near ultraviolet to
near infrared region. Operation at those
wavelengths may make design provisions
for personnel eye protection more
difficult due to lower allowable limits
of radiation exposure.<11>
Tunable diode lasers are available at
less than 1 W outputs covering wavelength
ranges from approximately 2 - 30 /zm.
Significant problems associated with
these devices are: (a) a very
narrow tuning range, requiring an array
of such devices in order to cover a wide
wavelength range, (b) high cost, (c) low
power output, and (d) large size and
weight when a cooling apparatus is
included.
Many gases and vapors of interest for
monitoring purposes have rich absorption
spectra in the near to far infrared,
associated with their
rotational-vibrational molecular
transitions. The CO2 laser emits
radiation in any of about 80 lines in the
region 9-11 j«n. The number of available
wavelengths can be expanded through the
use of isotopes of carbon or oxygen in
the CO2. One of these lines can often be
closely matched to a rotation-vibration
absorption line of a pollutant of
interest so that it is feasible to
monitor a large variety of compounds
(though not necessarily simultaneously).
A wide variety of chemicals have been
identified as being detectable using a
C02 laser based lidar system (see Table I
for a partial list of gases which can be
detected using CO2 lines(12>) . Using rapid
tuning, a single laser could presumably
be used for differential absorption
monitoring of more than a single
substance, and under many circumstances,
without interference by other vapors or
aerosols in the atmosphere. Humidity
would be expected to produce little
interference in monitoring pollutants in
the 9-11 /jm region from direct absorption
alone. However, in cases where the two
differential absorption lines are
separated by more than 2 x 10*Z jra,
moisture effects on the scattering and
absorption properties of certain aerosols
may become a concern.03' Other advantages
of the CO2 laser include the potential for
a durable, compact and modest costing
laser, lacking extensive utility and
436
-------
maintenance requirements, a more liberal
allowable irradiance level for laser
radiation at long wavelengths vs the UV
and visible01' (this may prove to be an
advantage, depending on other system
performance factors), and that
wavelengths produced by the CO2 laser are
transmitted well through normal
atmosphere.
Tabl* I. Partial liat of C03 la««r d«t«ctabl* ga«««)) (the parameter
characterizing the backscatter
efficiency) pertaining specifically to
infrared radiation and workplace
aerosols. Since the aerosol environment
varies considerably from one workplace to
another, the backscatter coefficient also
will vary.(U-15)
Past Uses of Lidar Systems
for Workplace Monitoring
There have been a few attempts in the
past to use laser based photo-optical
technology for workplace monitoring, in
1981, Britain's Imperial Chemical
Industries and G.P. Elliot Electronic
Systems, Ltd. made a brief report on a
system they were working on using a CO2
laser-based lidar system for scanning a
workplace.06> TO our knowledge this
device never was implemented. MDA
Scientific, Norcross, Georgia, (formerly
Tecan Remote and Environmental Laser
Systems), claims to have obtained an
exclusive license to use that technology
for their applications.
In 1983, Egan of Bethlehem Steel's Homer
Research Laboratories reported on the
trial of an Er:YAG laser based
differential absorption aerosol
backscatter lidar for methane detection
in mines.<17) In 1985, Litton of the U.S.
Bureau of Mines reported on their work in
developing a methane monitor using a
laser diode operating at 3.3 jum,, and a
triangulation system for ranging. This
resulted from problems with sensitivity
and explosion proofing of the equipment,
as well as a desire for tunability.
Results are not yet available/9'
In 1985, Persson, at Chalmers University,
Sweden reported on a dual CO2 laser
differential absorption detection system
for use in a workplace.<18) It used a
continuous wave (CW) laser with
retro-reflectors placed at the end of the
laser beam path. The device measured
total column content of the pollutant
chemical, thus yielding an average
concentration along the path.
Tecan Remote produced a commercial
differential absorption CO2 laser based
system for workplace pollutant
monitoring.09' They used a total column
content method and retro-reflectors.
They have installed systems at some major
chemical manufacturers facilities for
monitoring around process equipment of
special concern. Wavelength changing to
monitor different chemicals monitored was
possible by a manual retuning of both
lasers.
Other photo-optical methods were
considered including non-coherent pulsed
ultraviolet, visible and infrared sources
with sensitive spectroradiometric
437
-------
detection systems. The method seems to
be limited by the need for both very
intense and very short light pulses when
applied to aerosol scattering methods.
Generally, adequate high intensity
broadband sources having pulse lengths
under 10 /is are not yet commercially
available. This tends to force the use
of long-path measurement methods making
use of retro-reflectors, and the
acceptance of much poorer spatial
resolution. The system's spectral
resolution could cause problems in the
presence of chemical interferences.
Problems in Applying CO2 Laser
DIAL Lidar Technology to
Workplace Monitoring
Based on the above information, it
appeared that a C02laser DIAL system had
significant potential for leading to a
working system that would satisfy most of
the system characteristics stated
earlier. FTIR methods were also
considered promising; however, at least
two other research groups were
investigating that methodology. Many
uncertainties remained in the application
of current technology to producing a
lidar system satisfying the recommended
objectives. These included:
(a) problems in producing a system that
can scan a workplace atmosphere
while keeping transmitter and
receiver beam path alignments
precisely co-ordinated,
(b) problems in achieving overall system
sensitivities using components that
did not require frequent servicing
or supplies of materials such as
liquid nitrogen, cooling water,
etc. ,
(c) developing methods for unfolding the
identities and quantities of unknown
contaminants in complex mixtures,
using multiple wavelength
measurements,
(d) developing techniques for dealing
with data errors due to the effects
of rapidly changing aerosol
concentrations with time and
position. These could contribute to
differences in sequential
measurements on and off resonance
lines,
(e) designing a system that was field
portable (that was moveable into a
workplace as one or several modules
that can be handled by two people).
This could place severe restrictions
on the equipment selected for use,
(f) designing a system that could
operate continuously in an
industrial environment without
contamination of optical components
or without typical temperature and
humidity extremes posing a problem,
(g) a significant probability that
non-aerosol scattering (scattering
from objects and walls in the
workplace) could produce a sizeable
interfering signal in the detection
system, leading to erroneous results
or a greatly reduced system
sensitivity.
There were significant problems
associated with large magnitude
non-aerosol scattered return signals from
objects in the workplace causing
erroneous responses in the receiver.
Scatter rejection could be the most
severe technical difficulty to overcome.
This effect could be reduced by using a
coincidence of time-of-flight ranging and
triangulation ranging, as well a possible
use of both transmitter and receiver
polarization for the rejection of
multiple scattered return signals (see
Figure 1) . The triangulation ranging
uses a stationary alignment between the
transmitted beam and a linear array of
receiver detector elements. This fixed
transmitter-receiver relationship could
also help avoid the difficulty in
achieving adequate tracking between a
scanning receiver and a stationary or
scanning laser beam. In addition, this
approach could help to eliminate another
potential problem associated with short-
range lidar signals, which is a lack of
adequate dynamic range in the receiver
for return signals arriving from both
near and distant scatter sources. By
having given detectors look only at a
narrow distance range of return signals,
acceptable dynamic ranges should result.
It was anticipated that combining the
narrow field-of-view with time-of-flight
ranging would allow a 5m FWHM spatial
resolution. It was felt that HgCdTe
detectors cooled to liquid nitrogen
temperatures might be required here due
to the low allowable laser power levels
for eye safe conditions, as well as many
other demands of the system design.
Calculations for a specific system were
performed, using values for variables
438
-------
suited to the desired system performance.
The choice of values for these variables
resulted from a review of the
specifications of available components
and subassemblies (including a
consideration of their costs). An
example can be seen in Appendix I. The
calculations show that we would need a
single pulse energy of 2.75 x 10'3 J for
the system to function under the
conditions defined. The stability of the
system's electronics and optical
equipment may limit reliable differential
measurements to 1 to 2 percent ;<1Z'ZO)
consequently, signal to noise ratio (SNR)
values greater then 500 to 1000 may not
be useful.
The results of these calculations
indicated that if the assumptions used
were correct, it would be possible to
produce a lidar system having the
required sensitivity, beam irradience,
pulse energy, etc. However, these
calculations: (a) assumed a hypothetical
value for the volume backscatter
coefficient (/3(A0,R<>)) which may not
represent workplace conditions well, (b)
did not address large scattering signals
from surrounding objects, (c) assumed
perfect performance of optical and
electronic components, (d) did not
address interfering chemical species, (e)
did not address pulsed electrical noise
from the laser and Q-switch being
introduced into the detector signal,
degrading the systems SNR, and (f)
assumed that the limited information on
detector D* values were valid for the
fast signals necessary for the
time-of-flight measurements required.
Equipment and Methods
Laboratory Test System
Based on the review conducted, and the
potential benefit of a workplace
pollutant monitoring system based on a
DIAL system, a laboratory evaluation was
conducted of certain aspects of a CO2
laser workplace DIAL system using time-
of -flight and triangulation ranging. In
prder to make measurements of certain un-
evaluated parameters relative to the
performance of the lidar system, to
verify the practicability of certain
design concepts/ and to provide for
experimental development of a working
system, a laboratory test system was
assembled. The goals in designing this
system were to construct a laboratory
apparatus having a maximum flexibility to
evaluate various methods of assembling a
workable lidar system. It also needed to
be capable of measuring many of the
system parameters necessary for tne
development of future designs of a field
useable system.
The laboratory test system consisted of
the following subassemblies:
Laser
System performance goals necessitated the
use of a recently developed low pressure
pulsed CO2 laser in combination with a
high performance intercavity Q-switch for
producing near-Gaussian (no tailing as is
associated with TEA CO2 lasers), short,
high-intensity pulses. In differential
absorption lidar systems, it is important
that differential absorption measurements
in an atmosphere take place on and off of
the resonance line of the chemical
compound of interest, with a very short
time interval between the two
measurements. In the past, this has been
accomplished by using multiple lasers
tuned to different wavelengths, fired
sequentially with a short time interval
between pulses<20). It was considered
desirable that the two pulses of the pair
be produced by the same laser, lowering
equipment size, weight, .costs and
alignment problems.
Rapid wavelength changing of a single
laser could allow for a rapid change of
chemicals monitored, thus allowing a more
frequent monitoring of pollutants of
interest. This has not proved to be
practical in the TEA lasers typically
used, due to both power supply
constraints as well as the mechanical
behavior of the lasing medium. It was
also desirable that the laser be compact,
require a minimum of cooling, have an
extremely long lifetime (greater that 5 x
107 shots), and use low radio frequency
interference (RFI) components.
A laser was constructed using a modified
Pulse Systems, Los Alamos, NM, model LP30
low pressure CO2 amplifier section, due
to the long upper lasinq level lifetime
(greater than 60 ms) of low-pressure CO2
lasers. This lifetime permitted the
Q-switching of two output pulses from a
single laser amplifier electrical
transverse discharge pulse, while
allowing several microseconds for
wavelength changing between pulses. An
intracavity beam telescope was employed
to use the amplifier discharge cavity
439
-------
cross-section efficiently with the small
CdTe Q-switch crystals available. A 1200
Hz oscillating grating with a
high-resolution grating position sensor
was used to change and reprogram
wavelengths rapidly. Programming of
wavelengths was accomplished by selecting
appropriate delay times from the grating
position reference signal for triggering
the laser amplifier and Q-switch (see
Figure 1) . The resulting laser was
relatively compact and had a low mass,
and low power consumption. It had output
pulses of approximately Gaussian shape
with full-width-half-maximum_ values
adjustable between 50 and 100 ns; an
ability to produce "pulse pairs" having
interpulse spacings of 5-50 /is, each
pulse of the pair being independently
wavelength selectable over most CO2 laser
emission lines; a repetition rate for the
"pulse pair" up to 10 Hz;, a pulse energy
of approximately 5-10 mJ; and pulse pair
wavelengths reprogrammable between pulse
pairs. The output of the laser was
emitted through a beam expanding
telescope such that the irradience was
well below maximum permissible occular
exposure limits<1X> for the pulse widths
and repetition rates used. Most of the
basic performance goals of the device
were achieved in the laboratory
prototype. A full description of this
laser is presented in a paper soon to be
published.<21>
FIGURE 1. Q-SWITCHED DOUBLE-PULSE CO2 LASER
•v
Beam Power/Energy Monitor
Provision was made for extracting a fixed
percentage of the output beam to a beam
power and energy monitor. The beam
power/energy monitor was used for
continuously monitoring the beam pulse-
to-pulse energy differences to normalize
the results of differential absorption
measurements to constant beam pulse
energy conditions. This monitor
consisted of a fast (<1 ns rise and fall
time) room-temperature HgCdTe detector
(Boston Electronics Corp., Boston, MA,
model R004-0) and a Comlinear Corp.,
Loveland, CO, model CLC100 low noise
amplifier. The monitor had a short term
(1 h) reproducibility better then +1.0
percent and an absolute long term
accuracy better then +5 percent (2a) .
This detector monitored the laser output
radiation intensity and waveform via
radiation reflected from the tuning
grating.
Receiving Telescope
The receiving telescope was a Newtonian
type having a mirror diameter of 8 inches
and an effective aperture of
approximately f/4.5. It was used for
gathering the return signal laser
radiation and projecting a line image of
the aerosol scattered laser beam onto
HgCdTe detectors.
Detectors
Several types of detectors were available
for sensing radiation in the 9-11 fn&
wavelength band; however, only the HgCdTe
detectors had sufficiently high
detectivities (as indicated by D*) for
use in this application. HgCdTe
detectors are manufactured in a variety
of forms, varying in surface area,
wavelength sensitivity, frequency
response, etc., depending on their
applications. Unfortunately, the
manufacturers' data on their products
were often sketchy, and their testing
methods often did not include actual
measurements of fast pulses of radiation.
Thus, one was often left to speculate on
their actual performance in a particular
application. Competing performance
parameters, frequency response and
detectivity, were both critical to the
function of the lidar concepts to be
evaluated. The detectors evaluated in
the receiver system were cooled HgCdTe
detectors of photovoltaic (77 K) and
photoconductive (77 K and 200 K) types,
and were selected as representative of
devices commercially available at the
time. The photoconductive detectors were
specified to have approximately 10 degree
fields of view, 1.3 x 1.3 mm size, D*
values of approximately 1 x 1011 cm Hz1/2
W"1 at 10 kHz, and a high frequency
roll-off at approximately 10 MHz. The
elements were useable as single elements
or as a linear array with 5 elements.
Element six was a photovoltaic detector l
440
-------
mm diameter, had a 10° field-of-view, a D*
rating of 2 x 1010 cm Hz1/2 W1 at 100 kHz,
and a high frequency roll-off at
approximately 50 MHz. The detector array
was manufactured by InfraRed
Associates,Inc., Cranbury, NJ, as their
model #89-251R. For very high-speed
measurements (<10 ns rise and fall times)
a Judsen model J15TE4:10-MC31G-S01M
thermoelectric cooled (200K) HgCdTe
photoconductive detector with a model
TC-4 controller (both manufactured by
EG&G Judson, Montgomery, PA) was used.
It had a 1 mm diameter detector element,
a D* rating of 3 x 10* cm Hzvz W1 at 10
kHz, and a high frequency roll-off at
approximately 100 MHz. Generally, the
HgCdTe detectors designed for higher
operating temperatures exhibited much
better frequency responses, but this is
accompanied by a significantly lower D*
value. Due to the photovoltaic detectors
large capacitances and consequential long
time constants, a Comlinear model AJP401
transimpedance preamplifier was used to
help reduce the effective time constant.
Comlinear Model CLC100 video voltage
preamplifiers were used with the
photoconductive detectors.
Detector Cooling
The detectors used for detecting return
signals were cooled by two methods.
The linear array detector was kept at its
operating temperature (45-77 K) by a
closed cycle miniature refrigeration
system (a Philips/Magnavox Model MX 7043,
Magnavox Electro-optical Systems, Mahwah,
NJ). This consisted of a split Stirling
cycle, linear motor device using no
bearings or lubricants, and having
clearance seals. The overall device was
hermetically sealed. Mean times between
failures for this device are guaranteed
to exceed 2500 operating hours, with test
data implying lifetimes >10,000 h. Heat
power capacity at 77 K was approximately
1 W, allowing reasonably large assemblies
of detectors to be cooled. Total power
consumption at those operating conditions
was approximately 50 W. The vibration
produced by the mechanical refrigerators
can cause unwanted motion in the
detector, which may degrade system
performance by modulating the detector's
position relative to the focused photon
beam to be detected. The satisfactory
performance of such a closed-cycle cooler
could provide a considerable advantage to
the performance of a field-deployable
workplace lidar, allowing the utilization
of high sensitivity detector systems
without the difficulty of supplying
liquid nitrogen for it. The single
element photoconductive detector (see
identified in "Detector" section) was
cooled to 200 K by a four-stage
thermoelectric cooler. Such coolers had
extremely long expected lifetimes,
produced no vibration and required little
power (8 W for the one used). The lower
temperature limit of such coolers
resulted in a less than optimum D* value
for the detector, when detection
sensitivity at very high frequencies was
not critical.
Transient Waveform Analyzer
A high speed waveform analyzer was needed
both for general system diagnostics as
well as for analyzing lidar return signal
time of flight information. This
consisted of two LeCroy Corp., Spring
Valley, NY, Model TR8828 Transient
Waveform Recorders, capable of capturing
two fast waveforms simultaneously and
recording them in temporary memory.
These recorders digitized the waveform
information in 5 ns increments, with an
8-bit accuracy, and stored the
information in file lengths up to 32k.
These recorders were interfaced via a
CAMAC IEEE-488 interface to a data
processing system for display and data
reduction. ASYST waveform analysis
software (Asyst Software Technologies,
inc., Rochester, NY) was used to display
and treat lidar return signal data. A
specialized • fast-Fourier-transform
filtration method utilizing a Blackman
attenuation function*22' was used to remove
unwanted high frequency components of the
data, allowing a better extraction of
return signal information. After using a
wide range of cutoff frequencies with the
data, a 30 ns cutoff was selected which,
considering the slope of Blackman filter
function, appeared to have an effective
frequency cutoff of approximately 100
MHz. Some data were recorded using a
model 54021A oscilloscope manufactured by
Hewlett Packard Co., Palo Alto, CA.
Data Logger
A data logger consisting of at least 12
parallel channels of fast sample and hold
amplifiers coupled to 10 bit ADC's, was
interfaced to the data processing system
via a CAMAC IEEE-488 interface. The
sample and hold amplifiers had gate times
as small as 10 ns. This enabled the
setting of individual time of flight
range windows for individual detectors in
441
-------
a linear array, and a rapid shift of
signals from one set of input channels to
another between short-interval
pulse-pairs.
With the assembled laboratory test system
(see Figure 2), a series of experiments
was performed to help to understand
better the potential for using aerosol
backscatter as a signal source for a
RGURE 2. LABORATORY TEST SYSTEM.
workplace DIAL system. Measurements were
carried out in a corridor having
dimensions of 1.85 x 2.75 x 79 m. Major
components of the laboratory test system
were mounted on a movable bench
positioned at one end of the corridor.
The transmitted laser beam was then
projected along the major axes of the
corridor under several conditions (see
Figure 3) including: (a) co-axial with
the corridor and intercepted by the end
of the corridor;, (b) as in (a), except
with an object (a light source mounted on
a tripod), and separately, a generated
aerosol (wood dust, and bleached flour)
in the beam path located 28m from the
source; (c) with the laser beam projected
to reflect off the walls of the corridor,
as well as the end of the corridor.
Return signal measurements were made
using both the thermoelectric-cooled and
Figure 3. Corridor Measurements Using
Laboratory Test System.
Stirling-cycle-cooled detector systems,
located at the focal point of the
receiver telescope for particular
distance intervals along the transmitted
laser beams path. The profile of the
transmitted laser beam was mapped at a
position approximately 28 m from the
transmitter, using the
thermoelectric-cooled HgCdTe detector
operated above its normal operating
temperature of 200 K, to reduce its
efficiency. The detector was scanned in
2.5 cm intervals from -15 to +15 cm along
both the x and y axes on a plane
orthoganal to the beam axes.
Measurements were made at 10.6 nm
wavelength (10P20 line), with a laser
output of approximately 2 mJ/pulse.
Measurement of the overlap of the
receiver acceptance angle and solid angle
of the transmitted beam at a given focal
distance were made by scanning the
detector element across the focal plane
for a point reflector backscatter source
located at the extremes of the beam cross
section. This was accomplished at
distances from the transmitter/receiver
of 15 and 28 m.
Data and Discussion
Table II shows the transmitted beam cross
section relative irradiance profile.
Tabl* II. Happing of tha R«lat
at 30 n.
Relative Irradlanca (H CM'3)
-12.5 -10 -7.5
y-ax«s
(cm)
-12. 5
-10
-7.5
-5.0
-2.5
0
2.5
5.0
7.5
10
112.5
0 15
15 60
17S 270
100 200
50 25
100 70
From the slight bipolar shape, it appears
that there are probably two cavity
resonance modes present. Figure 4 shows
the output from the laser as measured by
the 200 K detector. The approximately 6
/.s interval between two sequential pulses
on the 10P16 and 10P20 lines respectively
can be seen. Figures 5a-5c show the
return pulse (diffusely reflected from an
object at 10 m) as monitored by a
photoconductive detector element (77 K)
a photovoltaic detector element (77 K)
and a photoconductive (200 K)
thermoelectrically cooled detector
respectively. It should be noted that
the polarity of the pulses shown in
Figure 5 varies with the detector and
amplifier used. The output of the 77 K
442
-------
Figure 4. Dual-pulse Laser Output.
,,,, ,
•;•••:
Pulse I (IOPI6)
/
Pulse 2 (IOP20)
Time (1000 ns/dlv)
Figure 5. Frequency Response
Characteristics of Three Types of
HgCdTe Detectors.
Photoconductlve • 77 K
Time (200 ns/ulv)
Distance (30 m/dlv)
/ \
Photovoltaic • 77 K
Photoconductlve • 200 K
*—**-A A ^-~-^
Time (100 ns/dlv)
Distance (15 m/dlv)
photoconductor and photovoltaic detectors
is positive-going, and the 200 K
photoconductive detector is negative-
going. It is clear that the frequency
response of the photoresistive detector
is quite poor, with a 1/e time constant
of approximately 500 ns. The 77 K
photovoltaic detector frequency response
was much better, with a 1/e time constant
of approximately 80 ns. The frequency
response of both of these detectors was
somewhat poorer than anticipated.
Manufacturer D* specifications for their
detectors are often based on low
frequency measurements (1-10 kHz) and
extrapolated to higher frequencies, with
possibilities for substantial errors.
Pulse shape distortions by the detectors
for fast rise and fall time pulses were
often not specified by the manufacturer.
The time constant of the 200 K
photoconductive detector was small enough
(specified as <10 ns by the manufacturer)
that the laser pulse shape detected by it
was not easily distinguishable from that
displayed from the high-speed
room-temperature HgCdTe detector.
Examining the time resolution required
for a 10 m time-of-flight (round-trip for
the desired 5 m spatial resolution,
approximately 33 ns) , it seems apparent
that, based on frequency response alone,
the 77 K photovoltaic detector would be
quite marginal, since tailing in its
output pulse excessively extended over
neighboring spaces, and only the 200K and
room temperature photoconductive
detectors would be fully adequate.
Considering that 60 ns FWHM is probably a
lower limit for the transmitted laser
pulse width, any further degradation of
that pulse width would be unacceptable.
It should be noted that it would be
difficult to achieve a 5 m spatial
resolution from a continuous aerosol
backscatter source using time-of-flight
methods and a 60 ns FWHM transmitted
pulse. We anticipated that the
combination of triangulation and time-of-
flight could make this possible).
Examining the D* ratings of these
detectors, it is clear that substantial
system design compromises would be
necessary to utilize either one, with the
room-temperature device being
particularly poor. Considering the above
findings, it appears that the earlier use
of a D* value of 1 x 1011 cm HzV2W"1 in the
example calculations was overly
optimistic, and that in practice a value
of 108 cm Hz1/2 W"1 may be more realistic.
This is a result of both the pulse shape
distortion resulting from some detectors,
as well as an inaccurate extrapolation of
D* values to higher frequencies. It is
possible that this could be improved by
the use of lower capacitance photovoltaic
detectors, detectors with a much smaller
surface area, or the use of coherent
detection methods. The use of much
smaller surface area detectors would
demand a much more sophisticated optical
assembly to achieve a sufficiently stable
focus of the return photons on the
appropriate detector element. Some
practical improvements could probably be
achieved with this; however, their
magnitude would be difficult to estimate.
Additional system sensitivity
improvements could be achieved by
increasing the receiving telescope
443
-------
diameter (a 30 cm diameter would increase
AO by a factor of 2), and by increasing
laser power (narrowing safety margins to
400% should allow an increase by
approximately a factor of 10) to
approximately 25 mJ/pulse. Both the
closed-cycle Stirling cooler and the
thermoelectric cooler worked well during
the several hundred hours of operation
they were used.
The curves in Figure 6 show the effect of
filtration on a single return pulse.
Figure 7 shows the overlap of an
end-of-corridor return signal with a
space separated from it by 20 meters,
using the high-speed 200 K HgCdTe
detector, and an approximately 90 ns FWHM
transmitted pulse. The ratio of the
end-of-corridor signal to the adjacent
space along the beam path (assumed to be
an insignificant signal for the aerosol
backscatter component) is approximately
300. Sources of this unwanted signal may
include stray light, aberrations in the
receiver optics, inadequate collimation
of the detector, and the shape and width
of the transmitted pulse.
Figure 6. Example of Filtration
of Single-pulse Waveform.
Figure 7. Effect of Large Signal
Source on Detection of
Neighboring Small Signal Source.
Time <80ns/dlv)
Distance (I 2 m/dlv)
If the ratio of the aerosol scattered
signal to that from a diffusely
reflective surface is approximately 10" ,
it appears that the return signal from a
space adjacent to such a strong signal
would have to be isolated from that
signal source by a ratio of at least 10 ,
in order for aerosol backscatter from the
space to be detectable and useful.
Depending on the aerosol concentration
20 m
Time (300ns/dlv)
Distance (45 m/dlv)
and its reflective properties the aerosol
scattered signal could vary over several
orders of magnitude. It appears from the
data in Figure 7 that the "tails" of the
very strong specular or diffuse
reflections from walls or other solid
objects cause overlaps of signal in
neighboring spaces such that the
resulting SNR ratings for the adjacent
space along the beam path woulxJ be
inappropriate for detection of aerosol
backscatter signals.
Figure 8 shows an example of an aerosol
backscatter signal produced by a fine
mahogany wood dust aerosol along an
approximately 3 m pathlength of the laser
beam, at a distance of approximately 28 m
from the transmitter/receiver. The
waveform was the result of the
subtraction of a return signal without
added aerosol from one with the added
aerosol. The limiting background noise
in the signal appeared to be the result
of electrical noise from the Q-switch
induced into the detector signal path,
rather than amplifier and detector noise.
This aerosol was poorly characterized;
however, it very likely represented wood
dust levels in excess of OSHA allowable
limits (a visible cloud). This
experiment was used to produce an example
of an upper practical limit of the amount
of backscatter signal that could be
obtained. Normal aerosol concentrations
in the air-conditioned laboratory space
were insufficient to observe aerosol
backscatter with the receiver efficiency
and transmitted laser power of the
present system.
444
-------
An examination of the difficulties in
using aerosol backscatter as a signal
source for a workplace DIAL system
prompted an examination of alternatives.
The above data suggest that the scatter
from workplace objects and walls could
provide a strong signal source for a
"column content" system. Time-of-flight
return signals, such as in Figure 9,
could be cross-correlated with the
transmitted signal to enhance the
separation of return pulses having
differing time-of-flight values, thus
enabling a determination if only one
significant scattering of the transmitted
Figure 8. Backscatter From Wood
Dust Aerosol.
Time (100 ns/dlv)
Dlntince (IS m/dlv)
Figure 9. Multiple-scattered
Return Signal.
Time (200 ns/dlv)
signal had taken place, as well as to
determine the round-trip path length of
the transmitted pulse to the receiver
(see Figure 10) . The continuous radial
scanning of a workplace utilizing this
methodology may provide a useful means of
generating an angular mapping of average
beam-path concentrations in the
workplace. The use of more than one
transmitter/receiver system, having
overlapping monitoring fields, could
allow construction of a workplace
pollutant concentration map. It is also
possible that an aerosol backscatter
signal could be used if the reflective
surfaces the beam could intercept were
treated to reduce their refectivities
(perhaps by a few orders of magnitude),
or in outdoor workplaces where no
significant non-aerosol backscatter
sources were in the field-of-view.
Outdoor settings should also provide a
substantially larger backscatter
coefficient.
Figure 10 Radial-scanning Workplace
Monitor.
Conclusions
This study identified several substantial
limits to the use of an
aerosol-backscatter DIAL system for
workplace monitoring. These were (a) the
limitations of currently available
detectors, when applied to a high spatial
resolution time-of-flight lidar, (b) the
difficulty in providing very high optical
isolation ratios for adjacent spaces
along the laser beam, (c) the probable
low aerosol concentrations for air
conditioned workplaces (compared to
outdoor concentrations),
-------
non-aerosol backscatter sources in the
field-of-view could be a suitable setting
for using aerosol backscatter methods.
As an alternative, it appears that the
use of backscatter from workplace objects
may provide a useful means of generating
an Angular mapping of average beam-path
concentrations in workplace. Further
work is needed to indicate the viability
of this approach.
Acknowledgment
The author wishes to acknowledge the
efforts of Gregory J. Deye, physicist,
for his work in the difficult task of
adapting the Asyst software to the
requirements of the data output of the
laboratory test system, and David
Hartley, research physicist, for his
initiation of the concept for this
project. Both are employees of the
Division of Physical Sciences and
Engineering, the National Institute for
Occupational Safety and Health,
Cincinnati. Ohio.
References
1. Herget, W.F. and J.D. Brasher:
"Remote Fourier Transform Infrared Air
Pollution Studies," Optical Engineering
19(4):508 (July/August) 1980.
2. Herget, W.F.: "Analysis of gaseous
air pollutants using a mobile FTIR
system," American Laboratory (December)
1982.
3. Herget, W.F.: "Long-Path Infrared
Measurements of Gaseous Emissions from
Aluminum Refinery Potrooms," U.S. EPA
Internal Document (May) 1980.
4. Sawyer, R.R. and J. Coppola:
"Automatic Monitoring Systems for
Determining Time Weighted Average
Workplace Levels," Symposium Proceedings
" Control Technology in the Plastics and
Resins Industry, DHHS (NIOSH) Pub. No.
81-107. U.S. Govt. Printing Office,
Washington, D.C. (January) 1981.
5. Watson, W.M.: "Continuous
Environmental Monitoring of Nickel
Carbonyl by Fourier Transform Infrared
spectrometry and Plasma Chromatography,"
Environmenta1 Science and Technology,
Vol. 13, (October) 1979.
6. Ying, L., S.P. Levine, C.R. Strang
and W.F. Herget: "Fourier Transform
Infrared (FTIR) Spectroscopy for'
Monitoring Airborne Gases and Vapors of
Industrial Hygiene Concern," Am. Ind.
Hyg. Assoc. J. 50(7):354-359 (1989).
7. Spartz, M.L., M.R. Wilkowski, J.M.
Fateley, J.s. Jarvin, et aJL.: "Evaluation
of a Mobile FT-IR System for Rapid VOC
Determinations," American Environmental
Laboratory (December) 1989.
8. Measures, R.M.: Laser Remote
Sensing, John Wiley and Sons, New York
(1984).
9. Litton, C.D.: "Remote Measurements
of Methane in Underground Mines",
Proceedings of Topical Meeting on Optical
Remote Sensing of the Atmosphere, Lake
Tahoe, Nevada (January) 1985.
10. Schwiesow, R.L., Cupp, R.E., Post,
M.J. and Calfee, R.: "Coherent
Differential Doppler Measurements of
Transverse Velocity at a Remote Point "
Applied Optics. Vol.16, #5, (May) 1977.
11. ANSI Z136.1, 1986, American
National Standard for the Safe Use of
Lasers, American National Standards
Institute, NY, NY (1986).
12. Grant, W.B. and R.T. Menzies: "A
Survey of Laser and Selected Optical
Systems for Remote Measurement of
Pollutant Gas Concentrations," j. Of the
Air Poll. Control Assoc. i3:188 (1983).
13. Petheram, J.C.: "Differential
backscatter from the atmospheric aerosol-
the implications for IR differential
absorption lidar," Applied Optics.
(November) 1981. —
14. Post, M.J.: "Aerosol Backscatterina
Profiles at CO2 Wavelengths: the NOAA
Data Base," APPl. Opt. 21:2507 (1984).
15. Schwiesow,R.L., Cupp, R.E., Derr
V.E., Barrett, E.W., and Pueschel, R.F'
"Aerosol Backscatter Coefficient Profiles
Measured at 10.6 mm", j^ App
Met eroro logy, Vol. 20, #2, (February)"
1981»
16. Mannon, J.H.: "Infrared Laser is
Key to British Gas Detector," Chemical
Engineering 47 (December) 1981.
17. DeFreez, R.: "Remote DIAL
Measurements of Methane in Coal Mines,"
Proceedings of Topical Meeting on Optical
Remote Sensing of the Atmosphere, Lake
Tahoe, Nevada (January) 1985.
446
-------
18. Persson, U. S., Lundquist, B.
Marthinsson and S.T. Eng:
"Computer-Automated COj-Laser Long-Path
Absorption System for Air Quality
Monitoring in the Working Environment,"
Appl. Opt. 2^:998 (1984).
19. Simpson, 0.: "Remote Sensing
Technologies for Hazardous Gas
Detection," Sensors (July) 1987.
20. Staehr, H., Lahmann, W., and
.Weitkamp, C.: "Range-resolved
differential absorption lidar:
optimization of range and sensitivity",
Applied Optics, Vol. 24, #13, (July)
1985.
21. Piltingsrud, H.V. : "A CO2 Laser for
Lidar Applications Producing Two
Narrowly-spaced Independently
Wavelength-selectable Q-switched Output
Pulses", Submitted to Applied Optics
(1990) .
22. Blackman, R.B., and Tukey, J.W.:
"The Measurement of Power Spectra",
Dover, 1958.
APPENDIX I
From Measures*8'
R
2/0 k(A0,R)dR
A0 {(X0) AR D'
where :
Emin = minimum laser energy pulse required
to observe a return signal at
range R (J)
R = range to AR being sensed (cm)
AR = range interval being sensed (cm)
SNRmin = signal to noise ratio
/9(X0 R) = volume backscatter
coefficient cm"1 sr"1
£(R) = overlap factor of laser and
receiver beam (geometric form
factor)
A0 = wavelength of resonance line at
which laser is operating (cm)
A0 = area of objective lens (cm2)
5 (X0) = receiver spectral transmission
factor
rd = detection time interval (s)
D* = specific detectivity (cm Hz1/2 W"1)
k(A.0,R) = normalized attenuation
coefficient for pollutant in
atmosphere (STP) (ppm cm)
-i
C = concentration of pollutant (STP)
(ppm)
B = detection bandwidth (sec"1)
(=l/2Td)
Ad = detector area (cm2) .
Using the following values for variables
in the previous equation:
R = 10,000 cm
AR = 500 cm
SNR =1.5
0(A0,R) = 10~8 cm"1 sr"1 (this value may
be much higher in an industrial
atmosphere or lower an air
conditioned atmosphere)
X0 = 9.639 /ira
A0 = 314 cm2
£(A.) = 1.0
rd = 5 x 10~8 s
D* = 1012 cm Hz1/2 W"1 (10° acceptance
angle) (practical values for this
variable using non-coherent
detection are probably on the
order of 1011)
k(X0) = 2.4 x 10~6 ppm"1 cm"1 (Benzene
at Ippm cnT1) (other compounds
may have values up to 102 times
that for benzene)
C = 1 ppm
Ad = 1.7 x 10~2 cm2 (0.05" on side).
Then,
E™1" = 4.13 X 10~6 J
L
If 10 sequential shots are accumulated,
then:
447
-------
5 x 10
~7
and
Emin = 1<3 x 1Q-6 j per
N
2 "3 .T
of 2.75 x 10"
DISCUSSION
CHUCK FLVNN: I'm curious about the decision, or finding, that it was not
useful because of the near scattering effects. If you could do away with this
scatter, would it be desirable?
HARLEY PILTINGSRUD: One of the things I mentioned in the paper, and
didn't have time to mention here, was that there possibly are some options in
some workplace situations where you could attenuate the backscatter from
objects in the workplace by some treatment. But it would be a little difficult
because you'd have to reduce it by a couple orders of magnitude and that's not
real easy to do.
CHUCK FLYNN: And if you took away the desire todo your ranging, could you
then do it easier?
HARLEY PILTINGSRUD: As I mentioned the second approach is one where
you lose some ranging. You know the direction the beam is pointed but the
measurement is a total column content one, so you don't know how the
concentration varies along the beam path. By using two such systems at different
angles of view, you could achieve some two-dimcntionsal spatial resolution.
448
-------
Immuno-based Personal Exposure Monitors
Arbor Drinkwine and Stan Spurlin
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110
Jeanette Van Emon
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada 89109
Viorica Lopez-Avila
Mid-Pacific Environmental Laboratory, Inc.
625-B Clyde Avenue
Mountain View, California 94043
Abstract
The feasibility of direct air monitoring using
immuno-based collection systems is being
investigated with the goal of developing personal
exposure monitors that take advantage of the high
specificity and sensitivity of immunochemical
systems. A system is under development which
will lead to compact, diffusion-based personal
exposure monitors for specific target analytes.
The interface problem of the aqueous based
antibody system and the air medium has been
overcome using semipermeable membrane tubes
with a very high surface to volume ratio. An
immuno-based collection and analysis system using
a monoclonal antibody developed specifically for
pentachlorophenol has been investigated. Similar
systems are being developed for aldicarb and
various nitroaromatics.
Introduction
The goal of this research is to develop personal
exposure monitors (PEMs) that use either a
polyclonal or a monoclonal antibody immobilized
onto a silica support for collection and detection
of a specific target analyte or compound class.
Selectivity is based on the inherent characteristics
of the antibody system used in the PEM device.
The device should be applicable to short- and
long-term monitoring and should allow the analysis
to be performed immediately after sampling and
in the field.
PEMs, both dynamic and passive, have long been
used for assessing occupational exposure to
hazardous materials. Recently, passive (diffusive)
sample collection systems have become popular as
PEMs, especially those that use sorbents. The
well characterized diffusion rates, high sample
capacity, and compact size of many of these
sorbent-based diffusion samplers make them ideal
for non-intrusive monitoring. Sorbents such as
charcoal or Tenax which allow the PEM to collect
an array of organic compounds are most
commonly used. These compounds are then
extracted and analyzed by gas chromatography or
gas chromatography/mass spectrometry in a
laboratory setting. This is an excellent approach
when several components are of interest and the
vapor is not characterized. However, this method
is not particularly cost effective when only one or
two compounds are to be monitored during a
specific exposure scenario, such as a pesticide
application or bag and drum operation. Only a
limited number of compound-specific PEMs are
currently available for this type of monitoring (i.e.,
formaldehyde and NOX).
Immuno-based collection and detection systems
have many attributes which enhance their appeal
as PEMs. The antibodies are generally selective
for a single compound or closely related class of
compounds. This selectivity is an advantage for
the isolation of target compounds. Also,
449
-------
antibodies generally exhibit high binding constants
for the target analytes, which means that their
collection efficiencies are high. Finally, the
antibodies can be easily regenerated to the
appropriate labeled or active form and reused.
The antibody-based detection systems available
exhibit good sensitivity, with detection limits often
in the 10 pg to 100 pg range. Recent applications
of amplification methods such as enzyme linked
immunosorbent assays (ELISA) have led to these
increases in sensitivity. The radioimmunoassay
(RIA) procedures, previously used in many
immunoassays, have been replaced with
colorimetric methods which allow for rapid
analysis of multiple samples using inexpensive
colorimetric readers (or even visual comparison to
standards in some cases). Most immunoassay
formats are also relatively simple to use and are
readily adaptable to field laboratories.
The limitation in using immunochemical
techniques for air sampling is that antibodies are
designed to work in aqueous systems. The few
attempts to use "dry" antibodies have not been
very successful. Studies have been carried out
using antibodies immobilized to substrates such as
polyethylene, Tenax, and indium [1,2]. The
immobilized antibodies were then exposed directly
to vapors without a wetting solution. Success was
limited to those systems which developed a
response in an aqueous solution after exposure.
This limited success was most likely the result of
the target analyte binding non-specifically to the
protein covered surface and then binding
specifically with the antibody after wetting. We
have overcome this limitation by using a vapor
permeable membrane as an interface.
The system we are investigating consists of a
vapor-permeable membrane and a cavity that
encapsulates the aqueous medium and the
immobilized antibody. The membrane acts as the
air-to-liquid interface, allowing vaporous analytes
to diffuse into the aqueous medium. The ideal
membrane should allow the molecules to pass
freely and should not interact with the analyte and
cause losses by nonselectively retaining the
analyte. It should also provide a high surface-
area-to-interstitial-volume ratio to keep the PEM
small and the mass transport process rapid.
Figure 1 shows a diagram of the PEM device.
The analyte in vapor form diffuses through the
porous membrane of a microdialysis tube. Once
inside the tube, the analyte is captured by the
antibody immobilized to the packing inside the
tube. The capture results in the release of a
labeled compound or an enzyme product
(depending on the type of detection system) to the
aqueous medium. At the end of the sampling
period, the tube fittings are attached to a syringe
or pump, the packing rinsed with solvent to
remove the labeled compound or the enzyme
product, and the rinsate analyzed to determine the
concentration of the target analyte that diffused
into the microdialysis tubing.
Information on the diffusion rates of selected
compounds across the membranes is presented in
this paper. Preliminary data on an antibody-based
collection system for pentachlorophenol (PCP) are
also presented.
Experimental
The membranes under evaluation are regenerated
cellulose microdialysis tubing (Spectrum Medical).
Typical tube dimensions are 50 /im in external
diameter, 35 nm internal diameter, and 20 cm in
length. Tubes are collected in bundles of 22 or 88
with pore sizes rated at 6,000 or 9,000 molecular
weight cutoff (MWCO).
The vapor chamber (Figure 2) consists of a 10 L
stainless steel vessel with air sampling ports and a
stirrer to ensure even vapor distribution within the
chamber. Previous evaluations have indicated that
the distribution is even throughout the chamber
using this propeller mixer. The saturated vapor is
created by placing the material (e.g., radiolabeled
PCP) in the bottom of the chamber and allowing
it to equilibrate until saturation is reached as
determined by collecting periodic air samples on
solid sorbents and analyzing them by gas
chromatography (GC) or liquid scintillation
counting (LSC).
PCP, 2,4-dinitrotoluene (DNT), and 2,4,6-
trinitrotoluene (TNT) were obtained from Aldrich
Scientific at 98% purity or better. The C14
radiolabeled materials were obtained from New
England Nuclear. The radiochemical purity of
each material was greater than 95%. The
monoclonal PCP antibody was developed by
WBAS, Inc. of Rockville, Maryland, and obtained
through the U.S. Environmental Protection
450
-------
Agency, Environmental Monitoring Systems
Laboratory - Las Vegas. All other reagents were
obtained from Sigma Chemical.
GC analysis of DNT and TNT was performed on
a Varian 3700 equipped with an electron capture
detector and a DB-5 30m, 0.25mm ID fused-silica
open-tubular column. LSC was performed using a
Packard 4170 Scintillation Counter.
The procedure used for the PCP vapor/liquid
diffusion study is as follows. A 9-inch, 22-fiber
bundle was masked off with Teflon tape to allow
only 1.5 in. of fiber to be exposed to the
atmosphere. The bundle was rinsed with ethanol
and phosphate-buffered saline (PBS) to wash out
the plasticizer according to the manufacturer's
instructions. The bundle was filled with PBS,
fitted into the cell holder, and placed into the 14C-
PCP vapor chamber. At 5-min intervals for 1 hr,
2 mL of fresh PBS were drawn through the
bundle. The 2 mL of fluid, which contained the
PCP samples during the interval, were transferred
to a LSC vial. At the end of the 1-hr exposure
time, the bundle was removed from the chamber
and hooked up to a peristaltic pump and fraction
collector. PBS was pumped through the bundle at
a rate of 0.25 mL/min for 5 hr and the effluent
collected in LSC vials. This post exposure study
was performed to help identify and quantitate any
hysteresis effect from PCP adsorbed to the tubing
but not yet migrated into the filling solution.
Quadruplicate vapor samples were taken from the
exposure chamber while the bundles were being
exposed and analyzed for PCP.
Preliminary experiments for nitroaromatics used
water-filled dialysis tubing as the analyte collector
in the chamber. These evaluations were
performed in deionized water because it was
unknown at this time what buffer system would be
used. This allowed the determination of the
analyte diffusion rates into the internal filling
solution of the tubing. The bundles were not
masked with Teflon tape as they were for PCP.
This results in a larger sampling surface area for
these evaluations than for the subsequent PCP
evaluations. The microtubes were filled with
deionized water after the manufacturer's extensive
solvent rinsing procedure was performed. The
filled tubes were placed into the chamber
containing 14C labeled DNT or TNT for 0, 5, 10,
20 and 60 minute periods, then removed, and the
labeled analyte amounts contained in both the
filling solution and the tubing material was
determined by LSC.
The procedure for the diffusion studies using an
antibody filling solution is as follows. The PCP
antibody was suspended in a PBS-Tween 20
solution and injected into the microtubes. The
tubes were then suspended for 15 minutes in the
chamber containing the radiolabeled PCP vapor at
3 ng/L. Control tubes were also placed into the
chamber, including one filled with PBS-Tween and
another with Bovine Serum Albumin (BSA)
protein suspended in PBS-Tween. The antibody-
filled tubes were then dialyzed against deionized
water for 4 hours to remove any unbound PCP
and then analyzed.
Results
Table 1 lists the approximate equilibrium vapor
concentration determined for the three analytes
using solid sorbents to sample the vapor chamber.
Repetitive samples were collected until the
chamber concentration reached equilibrium.
These values are not corrected for recovery
efficiencies from the collection sorbents because
previous work at MRI has indicated recoveries of
better than 90% for these particular analytes.
Table 1. Chamber Vapor Concentrations
Estimated Vapor
Concentration (Mg/L)
from Solid Sorbent Confirmation
Analyte Collection Method
DNT
TNT
PCP
3
100
3
GC
GC
LSC
Figure 3 contains the results from an evaluation of
the diffusion of DNT and TNT through the 6000
MWCO and 9000 MWCO dialysis microtubes.
The results for TNT and DNT are similar. The
lower apparent diffusion rates for TNT are
expected because TNT has a much lower
equilibrium vapor concentration than DNT. The
451
-------
plotted values are the average of two
determinations at each time interval and have
been corrected for surface area differences.
Future work will couple this collection system with
antibody-based detection systems.
The larger pore tubing (9000 MWCO) does
exhibit faster diffusion than the smaller pore
tubing (6000 MWCO). However, the magnitude
of the difference is not large enough to justify the
use of the larger pore tubing, which is more likely
to lose water by evaporation over extended
sampling periods.
The results from the diffusion rate studies of PCP
are shown in Tables 2, 3, and Figure 4. Duplicate
assays of each bundle type were performed. The
results of assaying the PBS effluent while the
bundles were in the exposure chamber are
presented in Table 2, and representative plots of
each bundle type are presented in Figure 4.
Linear regression analysis was performed on the
cumulative PCP passed through the membrane
and into the PBS buffer as a function of time.
The data indicate an average mass sampling rate
of 14.3 ng/min for the 6000 MWCO bundle and
19.2 ng/min for the 9000 MWCO bundle. The
effective sampling rate, calculated by division of
the mass sampling rate by the PCP vapor
concentration, averages 0.27 L/min for the 6000
MWCO bundle and 0.30 L/min for the 9000
MWCO bundle.
The correlation coefficients (Table 2) indicate
good linearity for both bundle types with the
MWCO 6000 bundles being greater than 0.970
and the 9000 being greater than 0.995. The x-
intercept indicates the approximate delay time in
which the vapor permeates into the bundles and
into the PBS. These times are 6.7 min for the
MWCO 6000 bundle and 2.9 min for the MWCO
9000 bundle.
The total amount of PCP that permeated the
membrane during and after exposure is presented
in Table 3. In the 5 hr after the bundles were
removed from the chamber, 1049 and 709 ng of
PCP continued to migrate through the MWCO
6000 bundle. Likewise, these values are 507 ng
and 438 ng for the 9000 MWCO bundle. These
results indicate a large fraction (46 percent for the
6000 bundle and 30 percent for the 9000 bundle)
of the PCP takes a considerable amount of time to
permeate through the membrane and into the
buffer. However, this should not affect the usage
since a waiting time of 1 hour can be inserted into
the analysis scheme.
Representative plots of the postexposure
permeation results for each bundle type are
presented in Figure 5. These results indicate a
very rapid PCP passthrough during the first 30
min, followed by a very slow accumulation during
the next 4.5 hr. The slow permeation phase may
be reaching an asymptotic limit, but it is
impossible to estimate the limit from these data.
A reasonable explanation for this phenomenon is
that the fast permeation phase is the PCP
diffusing through the aqueous portion of the open
pores of the membranes, while the slow phase is
the PCP diffusing through the regenerated
cellulose portion of the membrane. The
membrane may also become saturated with PCP
at this vapor concentration.
To provide a mass balance, another type of
exposure experiment was conducted in which the
bundle was exposed while being completely static.
A 9000 MWCO bundle was filled with PBS buffer
and placed in the PCP exposure chamber. After 1
hr, the bundle was removed from the chamber and
10 mL of PBS buffer immediately washed through
the bundle and assayed. Next, the buffer was
pumped through the bundle and collected for 5 hr.
At the end of the experiment, the bundle was
sacrificed and assayed for PCP. This experiment
detected 1400 ng PCP in the first of 10 mL of
effluent, 376 ng in the postexposure effluent, and
61.5 ng in the bundles. The fraction of PCP
permeating through the bundle was 20 percent of
the total PCP sampled by the membrane. Only 3
percent of the PCP remained in the membrane
indicating a low degree of permanent nonspecific
adsorption. Therefore, while diffusion may be
slow, very little of the analyte becomes
permanently affixed to the tubing.
Table 4 contains the results from the preliminary
evaluations of PCP-antibody loaded tubes. The
tubes were exposed to the radiolabeled PCP vapor
at 3 Mg/L for 15 minutes in the static exposure
chamber previously described. Tubes filled with
the PBS-Tween solution and PBS-Tween/BSA
were suspended in the chamber as controls. The
purpose of this experiment was to demonstrate
that the antibody-based collection system would
irreversibly bind the target analyte for later
452
-------
analysis. Such binding is important because
analyte exposure may occur in an episodic manner
and the retention of the analyte is key to
accurately determining the exposure.
In assessing this limited data set, it is clear that
the PCP-antibody is binding the PCP diffusing into
the tube. This can be inferred, in a non-
quantitative way, by comparing the amount of
PCP retained in the PCP-antibody loaded tubes
versus that retained in the control tubes after
dialysis as determined by LSC. More definitive
experiments are underway to quantify the relative
retentions.
Conclusions
The above data indicate a high probability of
success for the application of antibody-based
PEMs. Monitoring limits are of course
constrained by the detection capability of the
antibody-based system. However, based on the
reported limit for the PCP assay (1 ng) [3] and the
diffusion measurements reported in this paper, the
limit of detection for the PEM device should be
from 1-5 ng of PCP. Based on a vapor
concentration of 5 ppb (arbitrarily chosen as a
representative air concentration for PCP), this
would convert to a minimum exposure time
around 20 minutes for the analyte to reach a
detectable quantity within the PEM device.
The data, even though preliminary, demonstrate
the viability of using such PEMs. Studies carried
out as part of this program have also indicated
that the 6000 MWCO dialysis microtubing exhibits
sufficient collection efficiency for other
targetanalytes. Several antibody systems are under
evaluation for use in PEM devices, and new
systems will be evaluated as they become
available. These early studies indicate that it will
indeed be possible to apply antibodies to direct air
monitoring systems through the use of
microdialysis tubing as a semipermeable barrier
which allows vapor diffusion without significant
moisture loss.
References
1. Lukens, H.R., "Solid Substrate Immunological
Assay for Monitoring Organic Environmental
Contaminants," EPA Report No. 600/1-77-018,
Environmental Health Effects Research Series,
1977.
2. Giaever, I., "The Antibody-Antigen Reaction -
A Visual Observation," J. Immunology, 110, No 5
(1973) 144.
3. EPA Internal Report, "Evaluation of
Westinghouse Bioanalytic Systems PCP
Immunoassay," EPA/600/X-90/146, July 1990.
NOTICE: Although the research described in this
paper has been supported by the United States
Environmental Protection Agency, it has not been
subjected to Agency review and therefore does not
necessarily reflect the views of the Agency, and no
official endorsement should be inferred. Mention
of trade names or commercial products does not
constitute endorsement or recommendation for
use.
TABLE 2. RESULTS OF UTILIZING MICRODIALYSIS TUBING FOR
SAMPLING PCP VAPOR
Linear Regression Analysis
Bundle
type
Correlation
coefficient
(r2)
x-Intercept
(min)
Slope
(ng/min)
PCP vapor
concentration
(ng/L)
Effective
sampling rate
(L/min)
MWCO 6000
MWCO 9000
0.970
0.972
0.996
0.995
6.81
6.69
3.08
2.86
19.5
14.8
20.5
17.9
71.8
55.4
58.3
52.8
0.272
0.267
0.351
0.340
453
-------
TABLE 3. DIFFUSION OF PCP THROUGH THE HOLLOW FIBER
BUNDLES
Amount of PCP diffused through membrane
Bundle Type
MWCO6000
MWCO9000
With bundle in
(ng)
1135
857
1162
1049
chamber
(%)
52.0
54.7
69.6
70.5
With bundle
(ng)
1049
709
507
438
out of chamber
(%)
48.0
45.3
30.4
29.4
TABLE 4. COLLECTION OF PCP BY ANTIBODY SUSPENDED IN
MICRODIALYSIS TUBES (15 min exposure time)
Filling Solution
PBS-Tween/PCP Antibody
PBS-Tween
PBS-Tween/BSA
PCP Amount (ng)'
184
162
18
29
39
57
Duplicate determinations by LSC; Exposure time of 15 minutes
454
-------
FIGURE 1
Personal Exposure Monitor
Detach Tubing
from Holder
Badge Assembly
Attach Leur-loc
fittings and syringe-
empty into analysis vial,
455
-------
FIGURE 2
Exposure Chamber Used for Exposure Studies
Internal
Drive Shaft
Plugs
and Cover
Friction Fit
Removable
Aluminum Lid
Seamless
Aluminum Can
Guide Housing
(Threaded to Lid)1
10 Blade
Fan
Shaft to Stirring
Motor, Variable Speed
Control
Thermocouple
Probe
Sample coupons (4 x 1.25cm) suspended
from wire hooks
Aluminum
Plugs
Nickel
Support
Wire
Quartz Wool
Sample Tube
(Two at
Right Angles)
Swagelock
Connectors
To Digital
Readout and
Recorder
456
-------
FIGURE 3
Uptake of DNT and TNT by
Water Filled Dialysis Tubes
270
CM
.CJ
0)
T3
0)
"o
O
Q
**—
O
4-*
c
o
E
52.5
-45.0
-37.5
-30.0
n Tubing-DNT
O Tubing-TNT
-15.0
CM
I
CD
-------
FIGURE 4
PCP Vapor Diffusion,
MWCO 6000, and MWCO 9000 Bundle
1200
00
in
-
Exposure Time (min)
-------
FIGURE 5
PCP Vapor Diffusion,
MWCO 6000, and MWCO 9000
Bundle Removed From PCP Chamber
1050-n
O)
0
120
160
200
240
280
320
Exposure Time (min)
-------
A REMOTE SENSING INFRARED AIR MONITORING SYSTEM FOR GASES AND
VAPORS
Levine, S.P. (A,*), Xiao, H.K. (A), Herget, W. (B), Spear, R.
(C) and Pritchett, T.(D)
(A) University of Michigan, School of Public Health, Ann Arbor,
Michigan 48109-2029
(B) Nicolet Analytical, 5225-1 Verona, Madison, WI 53711-0508.
(C)University of California, School of Public Health, Berkeley,
CA 94720.
D) U.S. EPA-ERT, 2890 Woodbridge, Edison, NJ 08837-3679.
Abstract
A prototype transportable remote
sensing instrument has been built
that is capable of performing real
time quantitative analysis of gas and
vapor contaminants in air.
Introduction
In the early 1970's, papers appeared
in the literature on the
investigation of "remote sensing of
emissions" (ROSE) for air monitoring
(1-5). In all of those papers, the
instruments were large, and the data
analysis was off-line and depended on
an experienced spectroscopist
interpretation.
In the mid-1980's the use of FTIR
without remote sensing, but instead
with the use of a closed gas cell
into which the sample was pumped, and
using the classical least squares fit
(LSF) data analysis methods of
Haaland (8,9), was explored for air
monitoring applications (6-14).
Recently, several groups have begun
experimentation with ROSE-FTIR (15)
and differential absorption laser
systems for remote sensing of
pollutants at hazardous waste sites,
and potentially for fenceline
emergency chemical-release
monitoring. The first such study at a
hazardous waste site, a joint effort
in 1987 between the EPA-ERT and The
University of Michigan, resulted in
the definition of questions of
instrument stability under field
conditions, and aiming problems over
long (km) distances (16). Other
studies have been summarized (17,18).
All reported systems require manual
interpretation of data.
In this paper, we report on some of
the design and operational bases
behind this small, transportable
ROSE-FTIR system.
Experimental
The light source/optical bench weighs
16 kg, and measures approximately 20
cm (h) X 48 cm (w) X 36 cm (1). The
infra-red (IR) light source is an air
cooled Globar operating at 1300 K.
The optical bench contains a "porch-
swing" interferometer capable of up
to 2 cm""1" resolution at scan speeds
as high as 2 scans/second.
The receiver/detector module consists
of an 8 inch (20 cm) diameter
Cassagrain telescope equipped with
first-surface aluminum coated
mirrors, and a 3 inch (7.6 cm)
diameter convex spherical secondary
mirror. The IR light detector is
liquid nitrogen cooled HgCdTe (MCT),
2
with an image area of 1 mm
461
-------
The receiver/detector is mounted on
an aluminum plate, and also weighs 16
kg. Dimensions of this module are 28
cm (h) X 38 cm (w) X 36 cm (1), plus
a 30 cm (1) X 25 cm (d) extension on
the telescope cover. The electrical
requirement for the complete
instrument is <10 amps of 115 V.
electrical service.
Mirrors may be used to direct the IR
beam around the monitoring site.
These mirrors are 1 foot square (929
cm2) first surface aluminum coated
mirrors. All modules, including
mirrors, but not including the
computer, may be mounted on tripods.
The computer is a Dell 310 20 Mhz
80386 system with an 80387 co-
processor, 150 Mb hard drive, 4 Mb of
RAM, a Dell VGA-Plus color card and
NEC Multisync II color monitor, a
Nicolet Fourier transform co-
processor board, and a Nicolet A/D
controller board. Software is Nicolet
PC/IR, equipped with special systems
to aid in ROSE operation.
Discussion
The objective of this work was to
design, build and evaluate a small,
transportable remote sensing (ROSE)
Fourier transform infrared (FTIR)
spectrophotometer system designed
specifically for use as a gas and
vapor air monitor for the workplace
or in emergency response situations.
The system has a maximum viewing
distance of 40 meters, and can yield
one or more analyses every minute.
The IR beam can be placed linearly
or, using mirrors, around monitoring
stations that are not linear. In
order to aid in the accurate aiming
of the beam, the He-Ne laser beam
that emerges from the instrument is
co-axial with the IR light beam. The
laser beam has an intensity of 260
microwatts/cm2 at 1 meter distance
from the instrument, so protective
glasses are not needed.
In theory, the beam could be moved
around the workplace using a digital
stepper motor-controlled aiming
mirror. Thus, the most important
paths within an entire workplace or
emergency response site could be
traversed in a few minutes. The
optimal choice of the beam path is a
question being explored using a large
exposure chamber at the University of
California-Berkeley.
Evaluation of the system has shown
that beam path length and detector
response, under conditions of
constant and uniform concentration,
are directly related for paths tested
up to 12 meters. Path lengths beyond
that have not yet been tested.
Most important appears to be the
presence of non-analyte contaminants
in the "clean" background air
spectrum. These non-analytes cause
baseline non-linearities in the
spectral regions in which analytes
must be determined.
The key advance that has made the use
of ROSE-FTIR and closed cell-FTIR
methods possible for air monitoring
applications has been the use of LSF
analysis of the data. Since LSF
techniques make assumptions with
regard to the linear behavior of the
baseline, "poorly behaved" baselines
(with non-linear regions) degrade the
performance of the LSF software.
This, in turn, results in higher
limits of detection (LOD), poorer
linearity, and degraded accuracy and
precision.
The effects of temperature (10-35 °C)
and relative humidity (20-85%) have
been evaluated and found to be
minimal. However, when the instrument
is moved between monitoring sites and
the telescope optics are realigned,
minor variations in the baseline may
be significant with respect to the
performance of the LSF method.
The solution of this problem is
therefore central to the use of ROSE-
FTIR methods. In initial tests in the
462
-------
laboratory, this problem has been
solved through the use of negative
and positive least squares fitting.
With an analyte vapor mixture of five
components and a seven component non-
analyte mixture spectrum, both at 1
ppm concentration in ambient air per
component, the mean recovery of
analyte was 103% with a standard
deviation of 10%. Without this
method, the mean recovery degraded
significantly, as did the precision
of the results.
The use of an iterative classical
least squares fit (ILSF) approach has
also been evaluated for the
identification of unknown compounds
in the spectra of mixtures of vapors
in air (19). This method appears, in
preliminary tests, to be capable of
identifying unknown substances in
mixtures. However, field testing of
this method has not yet been
performed.
Conclusions
1. A transportable remote sensing
ROSE-FTIR instrument has been
designed, built and tested under
controlled conditions.
2. Positive and negative LSF methods
can be used, under controlled
conditions, to compensate for the
presence of non-analytes in the
background spectrum.
3. Iterative LSF methods can be used,
under controlled conditions, to
identify unknown components of the
spectra of mixtures.
4. Further testing is needed.
Acknowledgements
We thank the Environmental Response
Team (EPA-ERT) (contract 68-03-3255,
and related contracts) and the
Centers for Disease Control (CDC-
NIOSH) (research grants 1-R01-02404
and 1-R01-02666) for their generous
support. In addition, we would like
to acknowledge the support of the
Office of the Vice President for
Research at The University of
Michigan.
References
1. Analytical Methods Applied to Air
Pollution Measurements, R.K. Stevens
and W.F. Herget, Eds., Ann Arbor
Science, Ann Arbor, MI, 1974.
2. McClenney, W.A., W.F. Herget and
R.K. Stevens, A Comparative Review of
Open-Path Spectroscopic Absorption
Methods for Ambient Air Pollutants,
Chap. 6 in Ibid.
3. Barnes, H.M., Jr., Herget, W.F.
and Rollins, R., Remote Sensing of
SO2 in Power Plant Plumes Using
Ultraviolet Absorption and Infrared
Emission Spectroscopy, Chap. 12 in
Ibid.
4. Herget, W.F. and J.D. Brasher:
Remote Optical Sensing of Emission.
Appl. Opt. 18: 3404 (1979).
5. Herget, W.F.: Analysis of Gaseous
Air Pollutants Using a Mobile FTIR
System. Am. Labs. 72 (1982).
6. Herget, W.F., J. Staab, H.
Klingenberg, and W.J. Riedel:
Progress in the Prototype Development
of the New Multicomponent Exhaust Gas
Sampling and Analyzing System. Soc.
Automot. Engin. Conference, Feb.,
1984, paper no. 840470.
7. Herget, W.F. and S.P. Levine:
Fourier Transform Infrared (FTIR)
Spectroscopy for Monitoring
Semiconductor Process Gas Emissions.
Appl. Indus. Hyg. 1: 110 (1986).
8. Haaland, D.M. and R.G. Easterling:
Improved Sensitivity of Infrared
Spectroscopy by the Applications of
Least Squares Methods. Appl.
Spectrosc. 34: 59 (1980).
9. Haaland, D.M., R.G. Easterling and
D.A. Vopicka: Multivariate Least
Squares Methods Applied to the
Quantitative Spectral Analysis of
Multicomponent Samples. Appl.
Spectrosc. 39: 73-84 (1985).
10. Strang, C.R., S.P. Levine and
W.F. Herget: Evaluation of the
Fourier Transform Infrared (FTIR)
Spectrometer as a Quantitative Air
Monitor for Semiconductor
Manufacturing Process Emissions.
Amer. Ind. Hyg. Assoc. J. 50: 70-78
(1989).
463
-------
11. Strang, C.R. and S.P. Levine: The
Limits of Detection for the
Monitoring of semiconductor
Manufacturing Gas and Vapor Emissions
by Fourier Transform Infrared
Spectroscopy. Amer. Ind. Hyg. Assoc.
J. 50: 79-84 (1989).
12. Ying, L.S., S.P. Levine, C.R.
Strang and W.F. Herget: Fourier
Transform Infrared Spectroscopy for
Monitoring Airborne Gases and Vapors
of Industrial Hygiene Concern. Amer.
Ind. Hyg. Assoc. J., 50: 354-359
(1989).
13. Ying, L. S. and S. P. Levine, The
Use of Fourier Transform Infrared
Least-Squares Methods for the
Quantitative Analysis of Multi-
Component Mixtures of Airborne Vapors
of Industrial Hygiene Concern. Anal.
Chem., 61: 677-683 (1989).
14. Small, G.W., R.T. Kroutil, J.T.
Ditillo, and W.R. Loerop. Detection
of Atmospheric Pollutants by Direct
Analysis of Passive Fourier Transform
Infrared Interferograms. Anal. Chem.
60:264-269 (1988).
15. Spartz, M.L., et al, Evaluation
of a Mobile FT-IR System for Rapid
VOC Determination, Amer. Environ.
Lab., 15-30 (Nov., 1989).
16. Levine, S.P., L.S. Ying, C.R.
Strang, and H.K. Xiao, Advantages and
disadvantages in the use of FTIR
spectrometers for monitoring airborne
gases and vapors of industrial
hygiene concern. Appl. Indus. Hyg.,
Appl. Ind. Hyg., 4: 180-187 (1989).
17. Proceedings of the Symposium on
Measurement of Toxic and Related
Pollutants, EPA/AWMA, Research
Triangle Park, NC, 1989, 1990.
18. Minnich, T.R., R.L. Scotto, R.H.
Kagann, O.A. Simpson, Optical Remote
Sensors Ready to Tackle superfund,
RCRA Emissions Monitoring Tasks,
Hazmat World, 42-59 (May, 1990).
19. Xiao, H.K., Levine, S.P., D'Arcy,
J.B., Iterative Least Squares Fit
Procedures for the Identification of
Organic Vapor Mixtures by FTIR
Spectrophotometry. Anal. Chem., 61:
2708-2714 (1989).
DISCUSSION
MAURY FEE: In your expert system, did you have to reduce the interference
by successive subtraction, like a water vapor or CO2, in identifying your species?
STEVE LEVINE: The iterative least squares, which is the algorithmic heart of
the expert system, does that essentially by doing a least squares fit and getting an
optimal fit through several successive iterations. Instead of doing a classical
library search, we'll take the spectra we think are there one at a lime, and look
at the residual. What it does is attempts, through the windows that have been
chosen in the spectra, to do that with least squares fitting. So the answer is yes,
but not in the way that people used to do it.
JUDD POSNER: Have you ever considered the use of neural networks to make
it an artificial spectroscopist?
STEVE LEVINE: I did not pay him to ask that question! In fact, we had just
published a paper on neural nets in Analytical Chemistry. We did indeed try it.
The difficulty with neural nets is that you have to train them. To train them you
have to be able to see the peaks. When you're looking at the FTIR spectrum of
environmental or industrial gases or vapor mixtures you can't see a lot of the
peaks. And so the neural net can't be easily trained. The net failed at any
concentration below about 50 ppm for anything above about three component
mixtures that we tried. It was an interesting idea. We were hoping to use it as a
pre-screening tool for the iterative least squares to speed up the process and
reduce the algorithmic load on the system. But aside from producing an
interesting publication, it did not work.
BRIAN PIERCE: My question concerns the localization of leaks throughout
the plant. Could you distribute an array or retroreflectors and then direct your
source at these — to enable the ability to localize such a leak?
STEVE LEVINE: That of course is what we're ultimately hoping to do. At this
point we're hoping to make everything work with a single beam, manually
operated. Again you have to look at the things that MDA has done with their
DIAL laser system and their digitally controlled mirror for moving the beam
around, and what the Army has done with their seven position moving mirror that
surveys the battlefield in their XM21 FTIR. So this is something that others have
done, and we want to be able to do it in the workplace. We haven't done it yet.
DON LAVERY: On the qualitative part of your study, you're doing something
rather similar, I think, to what folk sat the Atomic Energy Commission of Canada
have done in the UV. In their library search they look first for the candidate
spectrum that will explain the largest amount of variation. And then proceed
through the secondary and tertiary candidates so that they pay the most attention
to the most important first. Have you done anything like that?
STEVE LEVINE: We've tried three strategies. One is to take the entire spectral
library in one array and see which combination of fits matches the known peaks
that we have best. That turned out to be the least satisfactory and the slowest. We
then tried the set building method where we go through the library one compound
at a time and look for the best matches that way. And the third is the set reduction
method where we start with 16 compounds in the library at a time, and match the
best for each 16, and attempt to squeeze the positives down to the fewest number.
The set building method which, in a way, parallels what you say the Canadians
have done on the U V system, is the method that is the fastest and has given us the
best results. The difficulty with me giving you a definitive answer is, again, it's
not robust. We have one publication on it. It needs more work.
464
-------
ADRIAMYCIN EXPOSURE STUDY AMONG HOSPITAL PERSONNEL
R L Stechenson C I H., Thomson Consumer Electronics Inc., Marion. IN.
CHRice ?hD. C.I.H., andJ.Dimos, M.S.. University of Clnc.no.t!,
Department of Environmental Health, Kettering Labs, Cmti., OH.
ABSTRACT
Using antineoplastic drugs is one of the routine treatment regimes employed in
combatting cancer. Nearly 250,000 cancer patients are treated annually with
chemotherapeutic agents usually in either hospital outpatient or inpatient
settings, however, some patients receive these drugs in physicians' offices or
outpatient clinics.1 The number and types of health care professionals who
are potentially exposed to antineoplastic drugs includes but is not limited to
4,000 medical oncologists, 10,000 oncology nurses, 30,000 pharmacists, and
even greater numbers of staff nurses and support personnel.1*2
A project was undertaken to utilize a visible light fluorescent method
previously developed at the University of Cincinnati, to document areas of
adriamycin contamination (skin and work surfaces). Specific aims of the study
were to compare the degree of contamination among pharmacists or pharmacy
technicians, intravenous (I.V.) therapy administration personnel, oncology
nurses, patient care nurses, maintenance workers, and hospital laundry
workers, and to describe the population at risk of dermal exposure to this
antineoplastic agent and recommend measures to prevent exposure.
Sixty-four observations/monitoring sessions for adriamycin exposures in a
hospital setting were conducted from June - August 1988, for dermal contact
with adriamycin. No dermal exposures to adriamycin among the hospital staff
monitored were found. In addition, no adriamycin contamination was documented
on any work surfaces. There were however, several noteworthy findings. The
ability to detect fluorescence of adriamycin (2 mg/ml to 0.002 mg/ml) applied
to some of the disposable latex gloves, disposable gowns/smocks, toweling,
front covers of the infusion pumps, and several other work surfaces typically
found in the hospitals surveyed, varied according to the material/work surface
and the concentration applied. Understanding of the sensitivity of the
methods and potential interferences will greatly aid in the interpretation of
positive results.
465
-------
INTRODUCTION
Background
Antineoplastic drugs such as adriamycin, cyclophosphamide, methotrexate,
mitomycin, dacarbazine, and cisplatin, are used in cancer treatment. Several
of these drugs produce mutagenic, carcinogenic, and teratogenic effects in
some cancer patients.3>4>5 Clinical research has been concerned with
patients and their health status following drug administration. Relatively
little attention has been given to persons exposed to antineoplastic drugs
occupationally, during preparation and administration of the drugs, or
following treatment of the patient. However, there has been a growing number
of studies in this area, some of which have shown mutagens or thioethers in
the urine and sister chromatid exchanges among personnel regularly handling
cytotoxic drugs.6«7>8
Some 25 antineoplastic drugs are commonly used in cancer therapy.5
Thirty-two agents are commercially available for treatment, and another 80 are
in clinical development.1 Many of these modern antineoplastic drugs are
highly toxic. Health care personnel who formulate, administer, and clean
surfaces that have contacted these drugs may be at risk of developing a number
of adverse effects, including cancer and fetal loss. The population presently
estimated to be occupationally exposed to antineoplastic agents includes, but
is not limited to, thousands of employees in the pharmaceutical manufacturing
plants, 30,000 hospital pharmacists, 4000 medical oncologists, 10,000 oncology
nurses, and even greater numbers of general staff nurses and support personnel
in hospital laundry maintenance and housekeeping. Effects of chronic exposure
to these drugs at very low dosages, as would be expected occupationally,
cannot be predicted with presently available data. Only two studies have been
published reporting any environmental sampling for antineoplastics in the
hospital setting. Both of these papers reported airborne levels of
antineoplastics, though neither study developed acceptable methods for
generalized use in airborne exposure assessment.9t10 |n addition, no
studies are available to confirm the effectiveness of engineering controls,
protective apparel, or work practices, along with proper handling and disposal
techniques for controlling the risk of contact with these drugs. Few if any
reports have been published which examined drug contamination of hospital
surfaces, staff clothing, or soiled bedclothes or methods to remove residual
drugs.
Antineoplastic drugs include alkylatlng agents, antimetabolites, antimitotic
agents, antibiotics and other drugs. The main therapeutic purpose is to
destroy cancer cells by blocking various biochemical pathways. The specific
site of action varies, depending on the particular class of agent. The
general mechanism of action is either through direct interaction with DNA or
inhibition of nucleic acid synthesis.11 Many of these drugs have been shown
to be carcinogenic, mutagenic and teratogenic in experimental systems, and
therapeutic doses of antineoplastic drugs have been associated with the
466
-------
development of secondary tumors in patients receiving chemotherapy. Aside
from their actions on tumor cells, antineoplastic agents can interfere with
normal body cells resulting in damage and, in some cases, cell death.11
Surgery, radiation, and chemotherapy are three types of medical treatment
commonly employed to combat cancer. Chemotherapeutic agents such as
adriamycin, are used because of their cytotoxicity. Adriamycin is one of the
most widely used of all the anticancer agents. It is frequently used to treat
tumors characteristic of leukemias, lymphomas, Hodgkins disease, and
carcinomas of the breast, ovaries, bladder, stomach, lung, thyroid, and
bronchus. Most of the antineoplastic agents currently in use today are
supplied as powders in vials or as liquid solutions requiring reconstitution
or dilution prior to administration by intravenous or parenteral injection.
Often pharmacists wearing gloves and protective smocks handle these drugs in
biological safety cabinets. The concentration is usually 2 mg/ral adriamycin
hydrochloride in sterile saline solution and the volume is adjusted for each
patient using additional sterile saline for dilution. The recommended dosage
schedule (adriamycin) for adult patients is 60-75 milligrams per square meter
(mg/m2) of body surface as a single I.V. infusion. The drugs are
administered in various schedules, such as once every three weeks, or on three
successive days every four weeks, until a total dose of 550 mg/n^
(adriamycin) has been given.12 The exact regimen depends on the drugs used,
type of cancer and the health status and responsiveness of the patient.
Patients may receive drug therapy in a variety of settings; hospital
inpatient, hospital outpatient, in the physician's office or in the home.
A variety of personnel are potentially exposed to the antineoplastic drugs
including nurses, doctors, and pharmacists who prepare and administer the
drugs, and maintenance and housekeeping staff, who repair, clean, and/or
dispose of equipment following administration of the drugs or work in the
rooms or offices where the drugs were administered. Adrianycin, like most
other drugs, is often not fully utilized by the patient (the dose administered
is not fully absorbed: some of the drug is excreted as is and some is excreted
in metabolized forms). Therefore, vomitus and excreta may contain the drug
and/or its metabolites. Housekeeping and custodial staff nay be exposed
during routine operations. Patient care personnel must handle bed linens
contaminated with vomitus and excreta which may contain drugs. Unprotected
laundry workers may unknowingly transfer drugs froa the linens to their
hands. In general, contaminated waste, bedlinens, vomitus and excreta may be
handled by a nunsber of persons involved in either treatment, patient care or
facility maintenance and the extent to which the personnel contaminate their
skin as a result of contact with drugs, waste or soiled linens has not been
documented. Other ways that the antineoplastic drugs can be released into the
work environment include contaminated packaging (broken vials damaged during
shipping), powders and liquid sprays (aerosols) released during preparation,
administration and routine cleanup operations, spills or leakage froa
syringes, I.V. bags, residual contamination on used syringes, gloves, linens,
vials, I.V. bags, and tubing. Although volatilization is not a property of
the currently availablp cytotoxic agents, aerosolization of the drugs can
occur during preparation and administration. Routes of entry into the body
467
-------
are through skin absorption (dermal), inhalation of aerosolized drug,
accidental self-innoculation and iagestion. Ingestion can occur during
mouth-breathing, smoking, eating, drinking, or other hand-to-mouth
contamination. Direct skin contact and inhalation of aerosolized drug are
often the greatest sources of exposure.
Adriamvcin
Adriamycin, also known as doxorubicin, is a red crystalline solid that is
soluble in water, aqueovs alcohols and methanol. This cytotoxic antibiotic is
isolated from cultures of Streptomyces peucntius. It is produced by three
companies; one in Japan, one in Italy and by one domestic manufacturer.
Spectrofluorometric methods have been used for identification and estimation
of the drug in biological fluids and tissues.12
By knowing the excitation and emission wavelengths characteristic of a
compound, one can use the fluorescence phenomenon to identify and quantitate
such compounds. One of the physical characteristics of adriamycin is that it
fluoresces when activated by certain wavelengths of visible and short wave
ultraviolet light. In prior research studies conducted by Rice, Van Raalte,
and Dimes et. al.13 at the University of Cincinnati, a spectrophotometer was
used to characterize the abs'orption spectrum of adriamycin hydrochloride in
saline solution with lactose, as it is constituted for patient
administration. They found that absorption in the visible range took place,
with a peak at 470nm. Using a spectrofluorometer they examined the
excitation/emission spectrum for adriamycin hydrochloride in saline solution
and found a maximum intensity occurring at 580 nm. The examination of
fluorescence excitation/emission was confined to the visible region since
ultraviolet illumination was not considered as an option for the project.
468
-------
EVALUATION METHODS
To insure the easy availability of the equipment used for this project, only
readily accessible materials were considered for the various components shown
in Figure I. A Kodak model AF-1 Ektagraphic slide projector was used as a
light source to stimulate fluorescence; the optical system of the projector
was equipped with a condensing lens and an infrared filter. The projector was
equipped with a 300 watt tungsten-halogen projection lamp and a gliss filter
(BG-12 4084 Filter) which selectively passed short wave (blue) visible light
was placed into the slide projection compartment. A 35mm single lens reflex
camera with a Vivitar 55mm 1:2.8 macrolens and a Kodak Wratten number 21
gelatin filter (75mm x 75mm) was used to photograph the fluorescent emission
from adriamycin. The Wratten filter absorbed the stimulating blue light
emitted by the light source, allowing only the orange-red fluorescent glow of
the adriamycin to be photographed. Sunglasses were worn during visual
observations to filter out the interfering light emitted by the stimulating
light source; ultraviolet and blue filtering sunglasses manufactured by Sun
Tiger (Pasadena, CA) block transmission of light below 550 nm and were used in
this research. To insure constant intensity and maximize sensitivity,
attempts were made to maintain the background light levels at a minimum. All
photographs were taken with the stimulating light source (projector) and
camera held at 20-25cm from the fluorescent materials, and the angle between
the light source and camera held to less than 45 degrees. With background
light levels under 10 lux, using Ektachrome 160 tungsten film, exposure times
between 1/4 and 1 second, and a maximum aperture setting of 2.8, the presence
of adriamycin fluorescence on test materials was demonstrated with
concentrations ranging from 2.0 to 0.002 mg/ml placed on various materials
including but not limited to stainless steel, benchtop absorbent padding, a
cotton lab coat cloth and latex glove material. These materials were felt to
be typical of the types of materials on which antineoplastics might spill or
leak in the clinical setting. Orange-red fluorescence was observed on all
test materials at all concentrations except for the most dilute which was not
observed to fluoresce on stainless steel or latex. Monitoring was conducted
on the worksurfaces, protective clothing and exposed skin both prior to and
after handling adriamycin itself or materials possibly contaminated with
adriamycin.
469
-------
RESULTS
Sixty-four separate monitoring sessions for adriamycin exposures in hospital
environments were conducted from June - August 1988, for dermal contact with
adriamycin. The various jobs monitored for adriamycin exposure included
full-time pharmacists, pharmacy interns, technicians, physician assistants,
nurses, laundry workers and maintenance workers. The areas and work surfaces
monitored for adriamycin contamination included chemotherapy preparation
areas, outpatient departments, filters in biological safety cabinets and HVAC
systems, hospital laundry areas, and chemotherapy infusion equipment. No
dermal exposures to adriamycin among the hospital staff monitored were found.
Many of the pharmacists monitored were double gloved and wore protective
smocks when they mixed adriamycin. Furthermore, all the hospital personnel
surveyed followed good work practices when handling antineoplastic drugs. In
addition, no adriamycin contamination was documented on any work surfaces.
There were however, several noteworthy findings.
The ability to detect fluorescence of adriamycin (2 mg/ml to 0.002 mg/ml)
applied to some of the disposable gloves, disposable gowns/smocks, toweling,
front covers of the infusion pumps, and several other work surfaces typically
found in the hospitals surveyed, varied according to the material/work surface
and the concentration applied. For example, the ability to detect
fluorescence on some types of latex gloves and especially orange-red colored
gloves, and on stainless steel, was reduced with the more dilute
concentrations of adriamycin. Allowing the eyes time for dark adaptation may
play a role in increasing one's ability to detect fainter fluorescence over
smaller areas. The increased illuminance of background light levels in the
survey area which were more than the optimum range of 10 to 35 lux, provides
for an additional interference problem. The excitation BG-12 4084 filter
mounted on the projector was chosen due to its availability and because it has
a transmission peak at AOOnm; it passes relatively little of the fluorescence
stimulating energy of 480 nm wavelength. While the filter was adequate,
fluorescence intensity would likely increase with the use of a filter with
peak transmission at 480nm. Lastly, some difficulties arose in conducting the
field evaluation using the bulky equipment. No doubt, miniaturization of the
detection system would provide greater acceptance for its use throughout the
health care environment.
CONCLUSIONS
This research demonstrates that a unique fluorescent detection system can be
used to reduce present uncertainties involved in occupational exposure to
antineoplastic drugs. Fluorescence detection provides a simple means of
measuring the contacted area. This method is quite useful in assessing the
adequacy of cleanup after the drug has been spilled. The method is sensitive,
minimal equipment is required; and very little training is needed to enable
personnel to monitor their own work areas and skin. For less than $100.00,
along with a slide projector, any work area can be monitored for adriamycin
contamination on a continuing basis by visual observations. Further research
is needed to define the limit of detection of visible light stimulated
fluorescence detection of adriamycin and optimize the method for use in the
field. The use of visible light to stimulate fluorescence may have broader
applications in industrial hygiene and dermal exposure and surface
contamination studies. Stimulating light sources equipped with several
interchangeable filters could allow for rapid detection of a number of
compounds.
470
-------
REFERENCES
1. DeVita, V. ed., Cancer-Principles and Practice of Oncology, 2nd Ed.,
Lippincott, 1985.
2. United States Department of Labor, Bureau of Labor Statistics, "Employment
by industry and occupation...hospitals." p.1675, 1983.
3. Sorsa, M. et al. "Occupational exposure to anticancer drugs-potential and
real hazards." Mutation Research, 154:135-149, 1985.
4. Solimando, D. and Wilson, J. "Demostration of skin fluorescence following
exposure to doxorubicin." Cancer Nursing, 6:313-315, August 1983.
5. Vaughn, M. and Christensen, W. "Occupational exposure to cancer
chemotherapeutic drugs: a literature review." AIHC Jour., 46:B8-B16, June
1985.
6. Falck, K.G., Grohn, P., et al., "Mutagenicity in urine of nurses handling
cytostatic drugs." Lancet, 1979.;1:1250-1.
7. Rogers, B. (1987). "Work practices of nurses who handle antineoplastic
agents." AAOHN Jour. #35, pp.24-31.
8. Norppa, Sorsa, Vainio, et al., "Increased sister chromatid exchange
frequencies in lymphocytes of nurses handling cytostatic drugs." Scand.
Jour. Work and Environ. Hlth., //6, pp. 299-301. 1980.
9. Kleinberg, M. and Quinn, M. "Airborne drug levels in a laminar-flow hood."
Am. J. Hosp. Phann., 38:1301-1313, 1981.
10. Neal, A., et al. "Exposure of hospital workers to airborne antineoplastic
agents." Am. J. Hosp. Phar., 40:597-601, 1983.
11. Vanderpool, H. "The ethics of clinical experimentation with anticancer
drugs." in Cancer Treatment and Research in a Humanistic Perspective,
Gross and Garb, eds. 1985.
12. World Health Organization, International Agency for Research on Cancer,
IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to
Humans, Vol. #10, IARC Lyon, France.
13. Van Raalte, J. "Detection of Adriamycin by Visible Light Stimulated
Fluorescence." Master's Thesis, University of Cincinnati, June 13, 1986.
BARRIER FILTER
Figure £_, Schematic of fluorescence detection equipment.
471
-------
DISCUSSION
PHIL GREENBAUM: I wondered if you had checked to see if any studies had
been done as far as birth defects related to this drug?
RICH STEPHENSON: Perhaps there's some literature that has been published
on that. I can't recall them off the lop of my head, though.
JUDD POSNER: It seems to me that this red drug had a capacity for being
determined with probably the simplest of the spectrophotometers that involve the
eye, and how much more sensitive in general was the UV than just looking
around for red spots?
RICH STEPHENSON: We didn't use the UV detector because of the hazards
associated with the UV light. We went with something that posed less of a hazard,
the visible light source.
JLDD POSNER: How much more sensitive was the visible light measurement
than the eye could see? I mean,do you have some ideaabout what kind of increase
in sensitivity that gave you?
RICH STEPHENSON: That wasn't pan of my thesis. It perhaps would be an
interesting topic for additional work.
HARRY SALEM: You slated that there was no dermal contamination, yet I
observed from your slide the pharmacist was wearing double gloves and short
sleeves. Was the potential dermal contamination tested on the bare arms or under
the gloves or protective clothing?
RICH STEPHENSON: Before the pharmacist or intern mixed or applied the
drug, we looked at the hands and the arms, and any exposed skin surface. And
definitely before they donned any gloves. And then we looked at it afterwards.
In prior studies done by Rice and VanRaulty, they did find some contamination.
But I think just knowing that we were present in the hospital environment and
telling the participants what we were looking for and why we were there, there
was a learning curve that happened right on the spot. So they took extreme
caution to fol low good work practices and not spi 11 any on their hands or clothing.
472
-------
REAL-TIME PERSONAL MONITORING IN THE WORKPLACE
USING RADIO TELEMETRY
Ronald J. Kovein
National Institute for Occupational
Safety and Health
Paul A. Hentz
National Institute for Occupational
Safety and Health
Disclaimer
Mention of a company name or product does
not constitute endorsement by the
National Institute for Occupational
Safety and Health.
Abstract
A system used to radio transmit data from
remote locations within a workplace to a
personal computer for immediate
interpretation has been developed by
NIOSH researchers. The system consists
of several radio transmitters and a base
receiver that is capable of multi-channel
reception. Exposure data obtained from
any direct-reading instrument with a
recorder output signal can be displayed
and stored at the computer. The worker
being monitored carries the instrument in
a backpack, along with a radio
transmitter. Using telemetry, the
concentration of the airborne contaminant
under study can then be plotted on a
video monitor for immediate assessment.
If there is more than one worker under
study at a time, multiple exposure curves
can be displayed on the screen.
The base receiver is a commercial
(frequency) scanner that has been
modified to accept RS-232 serial
communication. The manual keypad has
been removed so that channel selection is
accomplished through system software.
The radio transmitter is similar to a
wireless telephone; there are no wires to
entangle. The radio telemetry system
allows a worker unrestricted movement
within its effective range. It also can
be used to monitor up to five individual
workers per program execution.
A case study involving a furniture
refinisher's exposure to methylene
chloride is described here to demonstrate
the utility of radio telemetry. The
worker carried a radio transmitter that
was attached to a photoionization air
analyzer. Qualitative methylene chloride
exposures were remotely monitored on a
video monitor throughout the day.
Increases in exposure levels, due to job
tasks, work practices, and emission
sources were immediately identified so
that corrective action could be taken at
that time.
It is important for researchers who are
developing real-time monitoring
techniques to consider the procedures
discussed in this report. Also,
researchers conducting field studies
should be cognizant of the variety of
real-time monitoring techniques and use
them to their advantage in evaluating
worker exposure.
473
-------
Introduction
In recent years, researchers at the
National Institute for Occupational
Safety and Health (NIOSH) have used
microcomputers, data loggers or coaxial
cables, and video-taping techniques
during hazard control studies to help
acquire real-time exposure data.*1'2'3'
The data source was a portable
direct-reading instrument that measures a
worker's exposure to a hazardous vapor,
gas or dust. Such instruments measure
the concentration of airborne pollutants
through detection by flame ionization,
photoionization, electrochemical
reactions, infrared and ultraviolet
radiation, and chemiluminescence. *4) Most
of these devices provide a continuous
analog signal through an output connector
that is proportional to the concentration
detected. The signal can be routed to a
strip chart recorder for a continuous
record of exposure levels, or it can be
routed to an electro-mechanical device
(e.g., a solenoid valve) for process
control. Or, the analog signal can be
converted into an audio signal and
transmitted over a radio channel to a
distant receiver for immediate processing
by a microcomputer.
The radio telemetry system, discussed
herein, was developed to improve this
real-time exposure monitoring by
allowing: (1) the acquisition and
analysis of exposure data at locations
remote from non-stationary sources and
(2) immediate access to exposure
information that is not possible with
data loggers. The radio telemetry system
overcomes the distinct disadvantage of
delay by continuously supplying data to a
personal computer throughout a monitoring
session.'A* An interface modification of
a commercial (frequency) scanner allows
the computer to selectively tune up to
five separate frequencies through its
RS-232 serial port (i.e. the system would
permit monitoring of five individual
workers per program execution). The
-------
for transmission over the radio channel.
Using frequency-shift keying (FSK)
modulation, the advantages include an
error rate that is essentially
independent of signal amplitude, equal
per-digit error probabilities for a mark
and space, and simple noncoherent
detection without need to process the
carrier.
The base receiver, a modified Regency Z60
Programmable Scanner, contains the FSK
demodulator needed to reconvert a
transmitter's audio-frequency signals
(FSK tones) back to binary signals. The
Z60's normal keyboard programming has
been replaced with an RS-232 interface
connection to the IBM AT* or compatible
computer. Programs in BASIC were written
during system development to evaluate
performance; C-language programs are used
in the field for data collection.
The input section of the remote
transmitter consists of an 8-bit
successive approximation
analog-to-digital (A/D) converter. The
A/D converter operates continuously in a
free-running mode. Each frequency-
synthesized transmitter can be programmed
to operate on any frequency in the FM
broadcast band.*6' Selection of the
carrier frequency is important since
reception can vary from one broadcast
region to the next. Engineering firms
that specialize in radio frequency (RF)
design can build a transmitter to Part 15
(CFR 47) specifications. Independent RF
testing laboratories can provide the
certification required by the Federal
Communications Commission (FCC).
Field Demonstration
The radio telemetry system's first field
use was in a furniture refinishing
facility where the substance under study
was methylene chloride (NIOSH recommends
that worker exposure to methylene
chloride be controlled to the lowest
feasible limit).'7' A photoionization
air analyzer (Photovac TIP II") was
strapped to a rack which, in turn, was
attached to a tubular-framed backpack
(Figure 2). The TIP II™ comes with a
receptacle that allows an electrical
connection to a portable chart recorder.
Instead of the chart recorder, the analog
signal was split between a data logger
(Rustrak* Ranger) and a radio
transmitter. Two air sampling pumps also
were attached to the backpack. (The data
logger and pumps shown in Figure 2 were
unrelated to the demonstration of the
telemetry system, and will not be further
discussed.) Although data collected
through radio telemetry could be used to
determine if exposure limits were
exceeded, the data presented here are the
result of a qualitative approach.
Instead of parts per million (ppm), DC
voltage (an analog output of the
direct-reading instrument) was used to
identify and minimize peak exposures to
methylene chloride. For example, 0,5 VDC
may represent a concentration of 50 ppm,
but it is approximately one-half the
exposure represented by 1.0 VDC.
During the field demonstration, the
subject worker and equipment backpack
provided qualitative exposure data from
three separate work areas within the
facility.(8) Figures 3 and 4 show where
furniture finishes were removed
(stripped) and rinsed off, respectively.
In the absence of chemical leaks or
spills, the worker was not subjected to a
vapor buildup in the third work area.
A cart provided mobility for the
computer, video monitor, and base
receiver. The system demonstrated during
test runs that data transmissions could
be accurately received over 100-ft.
distances within the facility.
Figure 5 shows the worker stripping the
finish from a wooden chair, while wearing
an air-purifying respirator. (It should
be noted that NIOSH recommends the use of
either a supplied-air respirator, or a
self-contained breathing apparatus with a
full facepiece and operated in a
pressure-demand mode for any detectable
concentration of methylene chloride.) <9>
Each time that he completed a piece of
furniture, a table or chair, the worker
had to carry it into the rinse room to
spray off the chemical residue. After
the rinse, the furniture was usually
placed outside the double doors (in the
area of the cart) to air dry. The worker
also sometimes left the stripping room to
replenish his supply of solution before
475
-------
starting another piece. Regardless of
the worker's activity or location, it was
important to maintain accurate records
during monitoring.(8)
While the video monitor displayed the
worker's exposure to methylene chloride
vapors, the computer simultaneously
stored the data to disk. Figures 6
through 8 are reproductions from selected
data that were stored on the disk. The
reproductions are similar to the original
graphs that were temporarily viewed at
the work site as the exposures occurred.
With few exceptions, the emission sources
and work practices which led to increased
exposure were easily identified.
It should be noted that researchers
should consider instrument response time
when viewing work processes. To pinpoint
the sources and practices that contribute
most to the overall level of exposure,
the researcher should avoid choosing
instruments with a response time of more
than 10 seconds (5 seconds in some
applications). A delayed response may
create difficulties in interpreting the
relationship between instantaneous
exposure levels and the work process
itself.<10> If the duration of a delayed
response is known, data that have been
saved on disk can be offset to reflect
the true time for each exposure level.
Figure 6 illustrates the worker's
relative exposure to methylene chloride
during the removal of a table's finish.
When the worker was observed leaning
directly over the table, the particular
exposure level that was seen provided
evidence that worker participation in the
elimination, or reduction, of personal
exposure is as important as engineering
controls. The relationship between the
exposure of a worker to solvents and the
work method used is often a comparison of
actions and consequences.
Dramatic improvements can be obtained in
the job environment by relatively simple
changes in work practices. A method
called Picture Mix Exposure (PIMEX) was
developed at the National Institute of
Occupational Health in Solna, Sweden that
specifically addresses the problem of
employee awareness.<:I1^ The method
assumes that exposure depends to some
extent on the way the individual employee
works at his/her workplace. Good
examples of this are a spray painter's
exposure to solvents, a welder's exposure
to welding fumes, and many other cases in
which an employee handles the source of
the contaminant. Swedish field studies
employing video techniques make it
possible to identify problems on
videotape. Through the use of a video
mixer, measurements obtained by direct-
reading instruments are superimposed at
the edge of a monitor screen in a form
similar to a bar chart. The height of
the bar is proportional at all times with
the level of the signal. Since the PIMEX
method uses radio transmitters to route
the measurements to the mixer, video
mixing is performed on site so that
superimposed videotapes can be available
for the company debriefing. While NIOSH
researchers use a videotaping system,
present video techniques exclude video
mixing on location.
-------
Besides registering a sizeable peak in
exposure as a result of the initial water
rinse, the contamination is slow, in
terms of breath cycles, to dissipate.
The initial peak, followed by a gradual
reduction in the relative vapor
concentration, was characteristic for
both types of furniture. Figure 8 shows
the water rinse of another piece of
furniture. The series of smaller peaks
following the initial one is a response
to the jet of water striking fresh
solvent on unrinsed areas of the
furniture.
There appeared to be a potential
ventilation problem in the rinse booth
that required further investigation.
NIOSH researchers were alerted by the
exposure levels that were being displayed
on the video monitor (Figure 8). Once
alerted, a smoke tube was discharged in
the booth and its trail was followed. It
was found that the temperature of the
chemical solution was outside the
chemical manufacturer's recommended
range, resulting in a breakdown in a
paraffin-based vapor barrier which
created unnecessary fuming and product
loss.*8) Through the use of radio
telemetry, immediate action was taken to
correct the problem.
Conclusion
The usefulness of radio telemetry, a
system that instantly transmits exposure
data from direct-reading instruments to a
microcomputer, has been demonstrated on a
field survey at a furniture-stripping
operation. The system offers advantages
over other methods of data collection.
Using telemetry, workers under study can
enjoy more natural movement than they can
by being tied with coaxial cables.
Unlike data loggers, radio signals
produce instantaneous results. The
researcher can view exposure information
received from several workers and/or
processes on a moment-by-moment basis,
thus preserving the primary advantage
(instant feedback) of direct-reading
instrumentation. Immediate exposure
(B> Temperature Range: 60° to 85°F per
product label instructions for paint
remover #2105 manufactured by Kwick Kleen
Industrial Solvents, Inc., Vincennes, IN.
determinations are needed in the
workplace to prevent employee injury and
to advise management. Such
determinations allow the swift
elimination of emission sources and the
timely development of exposure scenarios,
proper work procedures and training aids.
Video techniques used by Swedish
researchers have recognized the
expediency of radio telemetry. Although
information that has been saved on disk
can be later mixed with videotape to
produce training aids, there is an
advantage in video mixing at the job
site. By combining the results as they
occur, the need to mix hours of data
files and videotape at another time and
place is eliminated.
Finally, the application of radio
telemetry provides more flexibility and
personal involvement while monitoring in
the workplace. Personnel involved in a
field study are generally more
productive. There is more interaction,
observation, and discussion. The focal
point of this activity centers around the
video monitor. As exposure information
is updated on the screen, comments can be
made and notes taken. Researchers have
the opportunity to take additional
measures in response to
higher-than-expected exposure readings.
Follow-up investigations could become
unnecessary when corrective actions are
proven effective before leaving the work
site.
It is important for researchers who are
developing real-time monitoring
techniques to consider the procedures
discussed in this report. Also,
researchers conducting field studies
should be cognizant of the variety of
real-time monitoring techniques and use
them to their advantage in evaluating
worker exposure.
Acknowledgment
We would like to thank Martin Abell of
NIOSH and William Metzger, a NIOSH summer
employee, for their assistance in the
development of the radio telemetry system
software.
477
-------
References
1. Sheehy, John, et al., "Methodolgy,"
Control of Asbestos Exposure During
Brake Drum Service, DHHS (NIOSH) Pub.
No. 89-121, U.S. Dept. of Health and
Human Services, Cincinnati, Ohio,
1989, pp. 12-13.
2. Gressel, Michael, et al. , "Advantages
of Real-time Data Acquisition for
Exposure Assessment," Applied
Industrial Hygiene, Vol. 3, No. 11,
November 1988, p. 316.
3. Caplan, Paul, et al., "Study Results
and Discussion," Survey Report:
Control Technology of Solid Material
Handling, NTIS Pub. No. PB-86-191285,
National Technical Information
Service, Springfield, Virginia, 1985,
pp. 22-24.
4. First, Melvin, "Sampling and Analysis
of Air Contaminants: An Overview,"
Applied Industrial Hygiene, Vol. 3,
No. 12, December 1988, p. F-20.
5. Beasley, Amy and Fairfield, Cheryl,
Survey Report: The Control of
Methylene Chloride During Furniture
Stripping at the Association for
Retarded Citizens, ECTB Rpt. No. 170-
18, U.S. Dept. of Health and Human
Services, Cincinnati, Ohio, 1990.
6. Code of Federal Regulations. Federal
Communications Commission. 47 CFR
15.239, rev. October 1, 1989.
7. Current Intelligence Bulletin #46.
Methylene Chloride, DHHS (NIOSH) Pub.
No. 86-114, U.S. Dept. of Health and
Human Services, Cincinnati, Ohio,
1986, pp. 2-3.
8. Kovein, Ronald and Hentz, Paul,
"Real-Time Personal Monitoring in the
Workplace Using Radio Telemetry,"
Applied Occupational and
Environmental Hygiene (in press).
9. NIOSH Pocket Guide to Chemical
Hazards, DHHS (NIOSH) Pub. No. 90-
117, U.S. Government Printing Office,
Washington, D.C. 1990, pp. 150-151.
10. Carlsson, Jan-Ake, "PIMEX: A New
Method for Worksite Environment
Control," Journal of the oil and
Colour Chemists' Association, Vol.
72, No. 6, June 1989, p. 211.
11. Rosen, Gunnar, "Video Filming and
Pollution Measurement as a Teaching
Aid in Reducing Exposure to Airborne
Pollutants," Visualization Methods
and Emission Studies as Aids in the
Control of Exposure to Airborne
Contaminants, Vol. l, Arbete & Halsa,
Solna, Sweden, 1989, pp. inn -
111:10.
">*IO!I IRHNSMimns
V
J9L
y -
_d d
'x *? IM «3 IM *4
88 ' MH* 90 5 MHZ I 01.1 MHZ
BflSt STRTIDN
11.7 MHZ
• S-IO* MH2
Figure l. Radio Telemetry
Block Diagram
478
-------
Figure 2. Equipment Backpack
Figure 4. Rinse Room
Figure 3. Workstation
479
-------
Figure 5. Stripping Furniture
200 400 600 800 1000 1200 1400
ELAPSED TIME (SECONDS)
Figure 7. Chemical Stripping/Water
Rinsing
oo
25
50 75 100 125 150 175
ELAPSED TIME (SECONDS)
200
WORKER LEANING OVER TABLE
WHILE STRIPPING
Figure 8,
Peak Exposures in
the Rinse Room
50 75 100 125 150 175
ELAPSED TIME (SECONDS)
Figure 6. Personal Exposure as a
Result of Poor Work
Practices
200
480
-------
DISCUSSION
MATT STINCHFIELD: Did you attempt to calibrate your relative reading to
actual air concentrations, and if so, how did you do it?
RONALD KOVEIN: That's something we took for granted. We were not out
to determine the exact concentration in that room. This is like a background sniff
to identify the sources, practices, so on. In other words, along the y axis there may
or may not be 50 ppm. But in a relative fashion we know the air quality of that
room, that rinse booth, was far less superior than. say. close proximity to the
ventilation system.
MATT STINCHFIELD: Yes. the work procedure and the corresponding spike
that we saw was apparent. Have you considered how you might go about
calibrating your field instrument?
RONALD KOVEIN: Yes, it is highly doable at this stage. In my mind it is even
secondary to others because it is so doable. We've tested the hardware. For
instance, the A to D in the front end of that transmitter is highly linear. We can
improve system software. It's still growing. All these things can help make the
system more user friendly and to allow for actual concentration readings. It is
highly doable.
MATT STINCHFIELD: One final pan to that. The software you were using for
receiving the information through your RS-232- is that a commercially available
type of data acquisition software, or was that something custom?
RONALD KOVEIN: Yes. it is. But once it's on the floppy disk it's part of the
data storage and you can use just about anything you want, similar to a data
logger. It is custom in that everything was built from scratch. We're not aware of
any thing out there commercially that does quite what we do. And in fact, we think
it is quite unique, though it may be considered very basic. You'll notice it is
simple. It is asynchronous, one-way communicalions. We have nothing sophis-
ticated or as expensive as cyclic redundancy checking, for instance. 1 don't feel
we need it. We have improvements down the way. and it is going to get better in
range. There is so much left to do. The commercial scanner by the way. is rough.
It did a great job for what we wanted it to do. but we can optimize it. We can give
it belter filtration making it much more sensitive and selective. And at the same
time we have improvements to do on the prototype transmitters. We can increase
field intensity for instance. Now with the revision of CFR 47 fora Part 15 device.
it can go from about 50 microvolts per meter up to 250. So as I say, things are
starting for our system in both performance and quality.
481
-------
IMPROVEMENTS IN THE MONITORING
OF PPM LEVEL ORGANIC VAPORS
WITH FIELD PORTABLE INSTRUMENTS
Gerald Moore
GMD Systems, Inc.
An Overview of the Scope for Monitoring
In the whole field of gas detection and measurement, organic
compounds represent by far the largest group of interest to users
due to the wide variety of these substances that are used in
different applications. Measurements are made for reasons of
process control, flammability hazard or toxicity and sometimes
for all of these reasons with the same substance.
Before focusing on the specific area of toxic level monitoring,
it is interesting to review the number of organic compounds and
the needs for monitoring.
One of the most widely used reference books, the CRC Handbook of
Chemistry and Phvsics*T lists a grand total of 17,746 chemical
compounds, both organic and inorganic. This listing breaks down
into 4,126 inorganic and 13,620 organic compounds; i.e.,
approximately two-thirds of all the listed compounds are
organ!cs.
However, when one looks for a more specific listing of compounds
likely to be of interest for monitoring in industry, there is a
vast difference in the numbers. One of the most comprehensive
list of organic vapors in common use is that published by the
National Fire Protection Association (NFPA) in the U.S.A.2 This
listing covers the fire hazard properties of liquids, gases and
volatile solids and has been progressively updated over the
years. After eliminating synonyms and inorganics, a total of
approximately 1,250 substances remains, suggesting that this may
be a reliable estimate of the number of organic compounds in
common use for which organic vapor monitoring may be needed.
Clearly this number of compounds far exceeds any other group of
compounds that require monitoring in air, but represents a small
fraction (less than 10%) of the total organics listing.
When one thinks of toxic hazards, TLV guidelines come to mind as
the accepted numerical classification. The American Conference
of Governmental Industrial Hygienists (ACGIH) recently published
a guide for TLVs3 in various countries . In this listing there
are 972 compounds of which 355 are organic vapors having a
defined and listed TLV. The other 6'17 are dusts inorganics and
organics having no listed TLV (carcinogens, etc). Since this
paper is concerned with monitoring toxic organic vapors, the
analysis will concentrate on the 355 listed TLVs in an attempt to
put an overall context on to organic vapor monitoring needs.
483
-------
An Overview of the Scope for Monitoring (continued)
Figure 1 shows the results of this numerical breakdown. From a
starting point of 1,250 substances in common use, those
substances clearly of interest only for their flammabi1ity were
next identified. This produced a surprisingly small total of 44
substances, typified by methane, ethylene and substances such as
corn oil, soybean oil, etc. Taking out this group, together with
the 355 listed by the ACGIH, leaves a total of 851 (68% of the
total number) which the author believes to be toxic at some level
even though they have no formal listing. Obvious examples are
members of the same chemical family or series such as ketones or
aldehydes, where one member of the family has a TLV listing and
the others do not. It could be argued that only certain members
of the series are in fact toxic, but the author believe that a
more reasonable explanation Is that the medical and
epidemiological data has not yet been produced and formalized.
From a starting point of 1,250 organic vapors, one can only,
therefore, subtract around 4% as being purely of Interest from a
flammabllity point of view. All the rest should be considered as
having either listed or unlisted toxic properties and might,
therefore, be of interest for monitoring in the atmosphere. This
is quite a challenge, requiring monitors for over 1,200 different
compounds at toxic levels.
The Need for Sensitivity In Monitorlng
A breakdown of the ACGIH listed TLVs for organic vapors is shown
in Figure 2. It will be seen that the numbers of TLV listings
below 100 ppb or above 1000 ppm are very small and that the
largest single groups are in the ranges 1-10 ppm and 10-100 ppm.
A flexible direct reading instrument having a range from
approximately 0.1 — >1000 ppm would therefore be of great utility
in measuring around 90% of the toxic organic vapors and 70% of
the listings could be covered with a range of 0.1 — >100 ppm.
Detection Principles in Common Use
The user has available a wide variety of detection techniques to
bring to bear on the problem of organic vapor monitoring.
Figure 3 provides a simplified overview of these techniques in
relation to one another and to the spectrum of requirements.
Eliminating those techniques that are highly specific in nature,
e.g., colorImetric paper tapes and those that are primarily
intended for use as flammabi1ity monitors, leaves the following
list of methods as suitable for portable direct reading monitors:
484
-------
connoNLv LISTED ORGANIC VAPORS
1258
LISTED ILU
28X €355)
NON-TOXIC
3.5X (44)
HOT LISTED
TOXICS
C051)
BENZENE
ACRVLOMITRILE
METHYL EIHVL KETONE
METHANE
ETHVLENE
1,4-DICHLOROBUTEHE
2>4-DICHLOROPHEHOL
96X LISTED AND NON-LISTED TOXICS
4X FLftHNftBLE OHLV
Figure I
485
-------
IZBr
00
o>
NUHBEB 68
B
11
1-9.9
TOXIC OBGANIC UAPOBS
DISTORTION OF LISTED TLU-TUft'S
19
53
63
15
1B-99PPB 188-999 PPB 1-9.9 PPH 18-99 Pffl 188-999 PPB 1888 PPB+
TLH-TW LISTING (flCGIH GUIDE 1996)
Figure 2
-------
SENSITIVITY RfiNGES FDR ORGflNIC VftPDR MONITORING INSTRUMENTS
£
-4
UNITS
PRINCIPflL
CONCERN
METHODS
flVftlLfiBLE
1
1
1
V
PPRTS PER BILLION (10~9) 1 PflRTS PER MILLION
1
0.1 1 10 100 1 10 100
1
<10~6> 1 PERCENT GflS (X)
\
1000 1 10 10ft
1
t
t
1
1
1
1
COLORIMETRIC PflPER TfiPES
1
1
1
1
t
__ __ )
1
1
LftSER / COMERft / FTIR
1
DXIDflTION (£)
(PPM RftNGE)
1 OXIDATION <1>
1 (LFL RftNBE)
1
t
- > 1
t
1
1 OXYGEN
DEFICIENCY
< THE RMflL >
CONDUCTIVITY
< — ELECTROCHEM— >
Og CELLS
Figure 3
-------
Detection Principles in Common Use (continued)
* Closed Path Infrared Absorption (IR)
* Flame lonlzatlon (FID)
* Gas Chromatography with FID (FID/GC)
* Catalytic Oxidation (Catalytic)
* Photoionization (PID)
* Gas Chromatography with PID (PID/GC)
The last two categories have seen many improvements in the last
five years, resulting in a new generation of lightweight direct-
reading instruments for monitoring organic vapors. The need for
specific compound identification in complex mixtures has
stimulated the development of portable gas chromatographs of high
sensitivity, suitable for field use.
Since the other techniques mentioned have been described
frequently in other published material, this paper focuses on the
particular properties of the photoionization detector and the
ongoing development of truly portable instruments based on the
use of this detector in conjunction with gas chromatographic
t echnique.
General Description of PID Based Instruments
Photoionization is becoming a very popular technique for the
detection of organic vapors at low ppm levels. This detection
principle, especially when combined with gas chromatography can
provide a powerful tool in Identification and quantitation. A
photoionization detector is similar to an FID except that the
ionization energy is provided by an ultra-violet emitting lamp.
A photon of UV radiation is absorbed by a molecule of the organic
vapor causing Ionization of the molecule. An ion flux is
generated between two high voltage electrodes and a detectable
current results. The detector usually consists of a sealed
interchangeable UV lamp that emits photons of specific energy.
If this energy level is high enough to ionize the organic
molecule, this will be detected in the ionization chamber as an
electrical signal which is electronically converted to a measured
concentration. If the proper lamp Is chosen, sensitivities can
be as low as 0.1 ppm, but there is a trade-off between high
energy emissions and lamp life. The high cost of replacement
lamp (>U.S.$500) makes this trade-off of importance. These
direct reading instruments are simple to use, highly portable and
reasonably cheap. However, the range of relative responses is
very wide as will be discussed later. Examples are the Photovac
"TIP"4 and MSA "Photon"5 instruments.
488
-------
General Description of PIP Based Instruments (continued)
As mentioned earlier, if this type of detector is used In
conjunction with gas chromatography, specificity improves and
limits of detection can be lowered. Several instruments having
this configuration have been designed and brought on to the
market in recent years. A good example is the PHOTOVAC Model
10S708. One of the little used advantages of a PID based GC is
the possibility of using air as the carrier gas, thus eliminating
all consumable supplies and the need for compressed gas
cylinders. This feature will be discussed in more detail in the
next section of this paper.
Special Features of PID Based Instruments
The versatility of a PID based direct reading portable instrument
and its ease of use has stimulated the development of a number of
commercially available instruments and encouraged their
widespread use in the field. However, the particular advantages
and limitations of the technique are not always well understood
and this can lead to improper application.
As stated earlier, the extent to which an organic vapor is
ionized depends on the energy level of the photons emitted by the
UV lamp, relative to the level required to ionize the vapor
(lonizatlon Potential). There are various published tables of
ionization potentials and some of these also give the relative
responses of these detectors to a variety of organic
vapors. »'>1" UV lamps are available having energy levels
ranging from 8.5 --> 11.7 eV with the higher energy lamps
typically having shorter lifetimes.
Figure 4 provides a breakdown of the number of organic vapors
that can be measured with each of the three most commonly used
lamp types, assuming the same total number of 1,250 as used
earlier. It will be seen that the use of a 9.5 eV lamp would
drastically curtail the usefulness of this technique and,
thereby, possibly give many "false negatives" in field use. A
10.2 eV lamp gives a far more comprehensive general purpose
response, allowing around 85% of all commonly used organic vapors
to be measured. It should be particularly noted, however, that
many of the compounds of most interest from a toxicity point of
view are halogenated organics which require higher energy lamps
to achieve ionization. Therefore, the improvement in the numbers
detectable with an 11.7 eV lamp is not as superficial as might at
first appear, since the halogenated compounds are heavily
represented in this difference number. Care should, therefore,
be taken in choosing a PID that is appropriate to the part icular
compounds in a given application.
489
-------
1298-
CD
O
DETECTABLE 888
COHPOUNDS
(1258 TOTAL) 688
B
275 (2ZZ)
TOXIC OBGANIC UAPOBS
PHOTOIONIZATION DETECTOB
NUHBEB DETECTABLE UEBSUS HUP TVPE
H ui i u fim M •«> •»» i»t iui( »)
9.5eU
IB.ZeU
WKPTffE
1213 (97Z)
11.7GU
Figure 4
-------
Special Features of PIP Based Instruments (continued)
The main advantage of a PID based portable monitor is its
universality of response. Subject to the careful choice of lamp
as described earlier, the monitor will give reading for an
extremely wide range of organic vapors. Typically such an
instrument is calibrated on the substance of choice, or
alternatively on a so-called "typical" substance. For other
substances, there will obviously be a range of relative responses
both above and below the nominal calibration.
Unfortunately, some suppliers use an aromatic compound such as
benzene for their "nominal" calibration. This is undesirable
since aromatics, in general, are the most easily ionized and,
therefore, respond with most types of UV lamps. In addition,ff
they tend to give high responses, meaning that most other
substances will have lower relative responses, in some cases as
little as l/100th that of benzene. Clearly, a more general
purpose instrument would be calibrated on a "median response"
substance so that relative responses would be distributed more
evenly above and below the nominal calibration (this assumes that
an individual calibration for a particular substance is not
avallable).
This approach was researched using the available references, and
the result is shown in Figure 5. By listing all available
relative responses and scaling them to the median value, a
compound could be selected (methyl propyl ketone) and the
distribution of relative responses of such a median-calibrated
instrument is shown. A useful number of compounds (263 or 21% of
the total) fall within relative response range of ±25% which
could be considered sufficiently accurate for many applications.
However, it will also be noted that almost half of the total
number are outside the range of ±80% so some care is still
necessary. Clearly, the PID technique is wide-ranging in its
uses but some care is necessary if general purpose uses are being
considered.
PID GC Improvements
The addition of capillary column gas chromatography to a PID-
based monitor results in an extremely flexible and portable
organic vapor monitor. The authors have been involved with the
development of such a system during the last few years and
believe that it represents a significant advance in technique.
This technique was originally developed at the Swedish National
Defense Research Institute (FOA) and was licensed to GMD Systems,
Inc., who will be marketing it under the trade name "Autograph".
The unit is presently at the advanced prototype stage and will be
available on the market during 19916.
491
-------
78--
2 OF
TOTAL 58
(1258)
PHOTOIONIZftTIQN
1EMIIUE RESPONSES TO TOXIC
mm
mm COMPOUND: RETHYL PBOPVL KETONE
532(663)
342(425)
.WA'.V.V.ViVt'.V.'tVtVi
tU**tfH»H*J**HH*
111.11111111 »"l MI At CM
HtiiiHWy«Yntn'H
IIHIIIIIIIIllllllllll
*/-lH2 */-25Z */-582 +/-8B2
BELfiTIHE BESPONSE BfUGE (COHPABO TO
472(588)
iVfujVn
-------
PIP GC Improvements (continued)
The Autograph is a fully portable PID based GC system which can
be operated either in direct reading "detector" mode with the
sample flowing directly to the PID; or alternatively in "GC" mode
permitting selectivity and identification of compounds in mixed
atmospheres.
Figure 6 shows the block diagram of the system, emphasizing the
extreme simplicity of the flow path and minimum of functional
components.
It should be noted that no external supply of carrier gas is
needed, since outside air from the environment is used as the
carrier and cleaned of organics by passing through the sorbent
trap shown, which consists of a small tube containing one or more
synthetic sorbent materials in granular form.
The sorbent trap thus, simultaneously serves to provide a clean
air "carrier" stream and also acts as a preconcentrator for the
pollutants of interest.
The sorbent tube is tightly wound with an electrical heater
winding; and after a preset sampling period, the tube is
electrically heated to around 250° C for a few seconds, thereby
desorbing the organics previously adsorbed. The desorbed
substances pass on to the heated capillary column (typically 10
meter 0.3 nm I.D.) in which they are separated from each other.
Each component of the mixture emerges from the column after a
fixed retention time and is detected and quantified by the PID.
Since retention times are dependent on flowrate and column
temperature, these two parameters are tightly controlled via the
microprocessor and appropriate sensors.
This technique of preconcentrating the organic pollutants via a
sorbent with subsequent rapid thermal desorption, improves the
lower detection limit by about 50 - 100 times as compared to a
PID alone.
The preconcentrat or sorbent tube is also removable and can be
used alone as a diffusion-operated personal sampler. By
distributing a number of these samplers and analyzing them later,
all the functions of a normal analytical laboratory can be
performed In the field with a single Instrument which can also be
used for survey work.
The microprocessor calculates concentration data, identifies
compounds against a known substance library and displays all
parameters on the instrument's LCD graphics display panel. All
data Is also stored for each sample period.
493
-------
DIRECT INLET
D
£
S
0
R
B
POUER
T
E
n
p
s
i
G
N
A
L
MICRO
PROCESSOR
R
A
I
E
KEYBOARD
DISPUW
PRINT/PEG!
ALARM
Figure 6
PID BASED GAS CH10HATOGIAPH
GHD "AUTOGRAPH"
494
-------
PIP GC Improvements (continued)
Operating parameters such as column temperature, length of
sampling period, etc., can be programmed by the operator either
via an Integral keypad or via a serial data link from an external
laptop or other personal computer, which also serves to download
and analyze data which is stored in the Instrument's memory
during field use.
Typical analysis cycle times are around 4 minutes, and it will be
appreciated that a new sample is being collected during this
cycle, since the sorbent continues to act as a filter to the
incoming air. However, longer sampling periods can be used,
giving increased sensitivity of detection, since the mass of
sample on the sorbent is increased. Sensitivities down to
0.01 ppb have been achieved in this way.
The versatility of the technique is enhanced by providing a
sample bypass valve which diverts the incoming air directly to
the PID. This permits rapid qualitative screening of the test
atmosphere with the concentration displayed in analog bar-graph
format on the graphic display. If organics are detected, the
Instrument can then be put into "GC" mode for Identification if
requi red.
The complete instrument is compact in size (approximately 4-1/2"
high x 10" wide x 12.0" deep) and weighs approximately 10 pounds.
It is battery operated and runs for up to 8 hours on a single
charge.
The Autograph demonstrates the degree to which it is possible to
simplify and miniaturize these techniques and shows the potential
for carrying out many analytical functions with a field-portable
inst rument.
Conclus1ons
It has been shown that there are around 1,250 organic compounds
In general use requiring organic vapor monitoring. Approximately
96% should be considered as listed or potentially toxic and,
therefore, there may be a need for toxic level monitoring. This
group of compounds represents by far the most numerous family of
substances for which specific monitoring is required.
The range of toxic concentrations is mainly from 0.1 --> 1000 ppm
with only very small numbers outside this range. Specific
detectors are often available for particular organics at ppb
levels, but the remainder require a good general purpose
monitoring technique.
495
-------
Conclus i ons (continued)
Improvement in PID-based instruments have made them more
available and easy to use in recent years. High sensitivity
broad-ranging response are the advantages, while cost of the
detector and non-uniformity of relative responses are the main
d i sadvantages.
and
PID-based portable GC systems are now widely available. This
combination of sensitivity, selectivity and portability is
unique. However, not all instruments take full advantage of the
technique as regards true portability, elimination of carrier
gases, etc. The development of the GMD "Autograph" has been
described as an example of good exploitation of this potential.
REFERENCES
1. CRC Handbook of Chemistry and Physics, 60th Edition,
April, 1979
2. NFPA 325M, Firo Hazard Properties of Flammable Liquids, Gases
and Volatile Solids, 19S4 Edition
3. Guide to Occupational Exposure Vulues- 1990, American
Conferenci? of Governmental Industrial Hygienists,
Cincinnati, OH
4. Photovac International, Inc., Huntington, NY 11743
5. MSA Corporation, Pittsburgh, PA 15230
6. GMD Systems, Inc., Bendersonvi11e, PA 15339
7. Listing of Molecular lonization Potentials, unknown
textbook source, GMD files
8. Technical Bulletin No. 11, Photovac International, Inc.,
Huntington, NY 11743
9. Rittfcldt, Lars, The National Defense Research Institute of
Sweden, personal communication of PID relative response
fact ors
10. Langhorst, M.L., "Phot oi orii zat i on Detector Sensitivity of
Organic Compounds", Journal of Chromatographic Science,
Vol. 19, February 1981
DISCUSSION
JUDD POSNER: My question relates to using air as a carrier gas. My
recollection way back when I was using PID was that there was a negative beat
for oxygen and that might be a problem with using air as acarrier gas. Does it not,
in fact, lower your sensitivity somewhat?
The second thing is, having spent so much time arguing with people that tubes
are not really a very good design for passive monitoring, I'm appalled that you
should come up here and tell me that you can use that little external tube as a
passive monitor.
GERALD MOORE: You got me with a tube, I have to admit. It's kind of a neat
feature of particular techniques, so why not use it. If I could have made it look
like a passive monitor, I really would have done it out of principle.
As to the more serious point about the oxygen, we haven't actually seen any
interfering peaks that interfere with the identification. I think the more serious
problem that's been raised by several people we've talked to is possible oxidation
of the thermally unstable compounds with an air carrier either in the system, on
the column, or wherever. Not having the advantage of a nitrogen carrier, I think
that is going to be a problem for some compounds. We don't have enough
experience to know how many. We hope that the advantages of the technique
considering the wide field we're dealing with here will be greater than the
disadvantages.
496
-------
RAPID ASSESSMENT OF SUPERFUND SITES FOR HAZARDOUS MATERIALS
WITH X-RAY FLUORESCENCE SPECTROMETRY
H.H. Cole III, R.E. Enwall, G.A. Raab, C.A. Kuharic
Lockheed Engineering & Sciences Co., Las Vegas, NV
W.H. Engelmann, L.A. Eccles
U.S. Environmental Protection Agency, Las Vegas, NV
Abstract
Field-portable X-ray fluorescence (FPXRF)
is nationally recognized as an excellent
screening tool for inorganic contaminants
on hazardous waste sites. However, FPXRF
is more than a screening tool when used
correctly. Properly calibrated and
monitored, FPXRF produces quantitative
data of known quality. Albeit, the data
are often of lower quality than intensive
laboratory analytical methods, but one
must consider the end-use of the data.
In many situations, definition of the
spatial distribution of the contaminants
can be accomplished most cost effectively
by taking numerous FPXRF measurements
rather than a limited number of
laboratory analyses of higher precision
and accuracy.
INTRODUCTION
Sampling procedures traditionally
employed for site characterization
revolve around the analytical laboratory.
Physical samples are collected onsite,
packaged, decontaminated if necessary,
and shipped to a laboratory for analysis
under chain-of-custody restrictions (see
Figure 1). Laboratory residence time is
typically 20 to 40 days, with analytical
costs of $150 or more per sample using
contract laboratory program (CLP)
procedures. Analytical results are
compiled into data packs and shipped to
site personnel for review. Clearly, this
approach is costly, time consuming, and
affords many opportunities for the
incurrence of errors and data loss. It
discourages multistage sampling and, in
some cases, has led to development of
inadequate sampling plans in order to
avoid the inherent expense.
The disadvantages of the foregoing
approach have prompted new emphasis on
the development of rapid, inexpensive
field analytical methods. In the case of
inorganic contaminants, X-ray
fluorescence (XRF) is particularly
applicable, especially for heavier
elements such as the transition metals.
XRF spectrometers capable of being
transported to field sites and operated
thereon are commercially available. Some
require operation in a field laboratory,
whereas others can be transported
manually and operated directly on the
site surface. Availability of the latter
has led to development of a field
portable X-ray fluorescence (FPXRF)
approach by Lockheed Engineering &
Sciences Co. (LESC) as part of the U.S.
Environmental Protection Agency (EPA)
effort to evaluate and develop field
screening and analytical methods.
The FPXRF approach consists of performing
in situ analyses with a field portable
instrument (see Figure 2). Maximum
flexibility and minimum data turnaround
can be achieved when the portable
instrument is supported with a field
laboratory containing equipment for
sample preparation, data analysis and
display, and a laboratory grade XRF
instrument for analyzing calibration
standards and confirmatory sample. If
desired, data and a report can be
delivered to site personnel prior to
demobilization.
497
-------
X-MET 880 FIELD SPECTROMETER
Field implementation employed in LESC
studies to date is limited to the X-Met
880. This instrument is a field-
portable, energy-dispersive spectrometer
commercially distributed through
Outokumpu Electronics, Inc., Langhorne,
PA. It is self-contained, battery
powered, and weighs 8.5 kg. These
characteristics, and the fact that it is
hermetically sealed and can therefore be
decontaminated, allow operation directly
onsite. X-ray fluorescence is induced by
a low intensity 244Cm or 241Am gamma-ray
source housed, along with a gas
proportional detector, in the sampling
probe. Operational safety is maintained
by a shutter approved by the Nuclear
Regulatory Commission.
Analysis with the X-Met 880 consists
simply of placing the probe in direct
contact with the sampling medium and
opening the shutter with a trigger.
Fluorescent x-ray photons are counted
over a user-specified period of time by a
counting circuit and classified into
discrete energy levels by a multichannel
analyzer to produce a spectrum
characteristic of the elements in the
sampling medium. Net intensities for
each target element are calculated by
software deconvolution of the
characteristic spectrum and converted to
concentration values by means of a
calibration model. This model is derived
empirically by measuring the net
intensities of the target elements in a
set of calibration standards, and fitting
a linear function that relates net
intensity to concentration by a multiple
regression procedure.
As is the case with all XRF systems, the
relationship between net intensity and
concentration varies with the
characteristics of the sample matrix. In
the case of solid, inhomogeneous
particulate media such as soils or
sludges, the concentration-intensity
relationship is particularly influenced
by variability in the grain size
distribution, bulk density, and the
geometric relationships between discrete
grains containing the target element(s)
and the detector. The geometry problem
is exacerbated by the very small volume
(roughly 0.04 cubic inch) measured by the
probe. Net intensities can be
artificially enhanced or absorbed by
certain non-target elements that may be
present. Moisture has also been reported
to affect intensities(1) . Data quality
can be significantly influenced by any or
all of these matrix effects which must
therefore be taken into account in the
calibration procedure, a subject that is
discussed in greater detail in the next
section.
ROUTINE FIELD PROCEDURES
Calibration
The X-Met 880 has no fundamental
parameter capabilities which would allow
for standardless calibration. It uses
calibration curves based on matrices
similar to those of the routine samples.
The instrument has 32 calibration models,
each of which can contain up to 6
calibration curves. Therefore, each
calibration model can simultaneously
quantify up to six analytes. The number
of calibration standards required for
each calibration model depends on the
number of analytes of interest; generally
eight to ten standards per analyte.
Site Typical
A site typical calibration curve is based
on samples similar in composition, but
not necessarily matrix matched. Extreme
caution should be exercised when using a
site-typical calibration curve. The
authors have encountered the situation
where increased iron levels in mine
tailings relative to the calibration
standards resulted in anomalously high
chromium results (in excess of several
wt% Cr!). Corroboratory analyses found
chromium in the zero to 40 mg/kg range.
Site Specific
To minimize enhancement/absorption and
spectral interference errors, calibration
standards should be collected from the
specific site in question. These Site
Specific Calibration (SSC) standards must
closely emulate the physical and chemical
matrix of the routine samples. The SSC
standards are prepared as loose soils
(screened through 2 mm but unpulverized)
so that the particle size bias of the
routine samples is included in the
instrument calibration.
Characterization of the SSC standards
must be done using a total digestion
procedure rather than a partial
extraction (i.e., CLP "Total Metals'"21),
because XRF is a total analyte method
regardless of phase or speciation.
498
-------
Co-calibration
If more than one X-Met is used on a site,
the two instruments should be co-
calibrated with the same SSC standards.
Even with co-calibrated instruments,
there can be inter-instrument bias. This
should be tracked and quantified by using
splits of the same check samples
(discussed below) to monitor the
instruments.
The authors were involved in a site
screening which was restrained to using 2
X-Mets that were not co-calibrated. The
instruments produced two very different
populations of data for chromium, an
order of magnitude apart. The higher
concentration set exceed the state
regulatory levels. Despite the low
chromium values in the laboratory
corroboratory samples, the high FPXRF
values are still an issue with the state
health agency involved (approximately 1.5
years later).
Sampling
In situ analysis does not require that a
physical sample be removed from the
ground. The FPXRF probe is placed on the
ground and the analysis mode is activated
by pulling the trigger. Acquisition time
can be preset at any desired length; 30
to 120 seconds is the most common range.
FPXRF in situ analyses are very
beneficial during remediation. FPXRF can
be employed in iterative passes following
contaminated soil removal efforts and
quickly produce the results of each
remediation attempt.
In situ analyses are the quickest way to
obtain soil chemistry data but these data
also contain the highest degree of
variability (error). A large source of
error is the extreme heterogeneity of
most soils. The radioactive source in
the probe is exciting a very small cross
section of the soil (20 mm diameter by 2
mm depth) . When the probe is moved
laterally 5 cm, the detector is 'seeing'
a very different sample. Another source
of error is the extremely wide range of
particle sizes in the sample. This
produces a known negative bias(3) that can
be somewhat compensated for by using
loose, unpulverized SSC standards. A
third source of error is surface
microvariability which is caused by such
physical factors as wind or running
water, vehicles or footsteps, and by
chemical alteration of the surface. An
excellent example of a rapid chemical
change in surface phenomena was
encountered by the authors at a mine
tailings site in New Mexico. A white
precipitate of ZnSO4 (?) formed surface
crusts in the late morning as the
tailings piles dried out. Every
afternoon, thunder storms washed the
precipitate back into the soil. No
precipitate was apparent the next morning
until the sun began to dry the soil.
Different daily spatial patterns could
radically alter concentration maps.
Surface microvariability can be mitigated
by in situ homogenization or _ by
collecting intrusive samples, i.e.
samples that are physically removed from
the sampling media'41.
DATA QUALITY
The procedures in this section address
only in situ analyses. Intrusive samples
have several sources of variability that
do not occur with in situ sampling such
as collection, handling, and preparation
errors.
Quality Control Procedures
Replicate Analyses
All FPXRF routine samples are analyzed in
triplicate and the means are the reported
values. The three in situ measurements
are made in a 6 by 6 by 6 inch triangular
pattern around the sample location
marker. After the third measurement of
every fifth sample location, the probe is
left in place and analysis is repeated
two more times (stationary probe
triplicate).
Before the first in situ sample location
and after every tenth sample location
low- and mid-calibration range quality
control (QC) check samples are analyzed
in triplicate. QC check samples are
loose soils in a 31 mm diameter by 2.5 cm
cup, approximately one half full. Each
QC check sample is analyzed in
triplicate, by removing the cup from the
detector between each analysis, shaking
the soil in the cup, then lightly tapping
the cup on a smooth surface before
replacing it on the detector. Analyzing
these samples periodically will warn the
technician of gain change or other
instrument problems. The authors have
found this particularly helpful in
detecting a low battery before the X-Met
software gives the "low battery" warning.
499
-------
Confirmatory Samples
The number of samples for laboratory
confirmation of the FPXRF instrument
results are based on the overall number
of FPXRF samples points and the budget of
the onsite coordinator. One confirmatory
in 40 routine samples is adequate on a
site with approximately 300 sample
locations. Confirmatory sample frequency
can decrease with increased sample
locations.
Quality Assurance Parameters
Precision
Different levels of precision can be
determined. Minimum instrument
variability in the field is measured from
the stationary probe triplicate. A more
comprehensive assessment of precision is
measured from the performance of the QC
check samples. Physical agitation of the
sample between measurements yields some
component of the soil microvariability.
The QC samples are analyzed over the
entire time span of analyses and,
consequently, yield an overall FPXRF
system precision.
Accuracy
Accuracy is also determined from the low-
and mid-calibration range QC check
samples. QC values are plotted in
control charts (one for the low range
sample and one for the high range
samples) with concentration on the
ordinate and successive measurements on
the abscissa. Accuracy and the
instrument bias from the "true" values
can be quickly determined from the
control charts. Table 1 compares FPXRF
accuracy and precision performance to
that of the CLP Data Quality Objectives
(DQOs)<2>.
Detection Limits
Detection limits are defined as three
times the standard deviation (SD) of the
low-calibration range QC check sample.
CLP contract required detection limits*2*
are compared to FPXRF detection limits in
Table 2.
Data Quality Objectives
Tables 1 and 2 give some indication of
levels for precision and accuracy that
can be achieved under conditions deemed
to be typical for the types of waste
sites investigated. These values should
be viewed with some caution, however,
recalling that the matrix-specific nature
of the intensity-concentration
relationship dictates that achievable QA
levels are also site-specific.
Precision, accuracy, and detection limits
may vary significantly from matrix to
matrix, in contrast with "wet chemical"
procedures in which physical matrix
effects are eliminated by taking samples
into solution prior to analysis. QA
levels displayed in Tables 1 and 2
probably represent neither the best or
worst cases. Evaluating the applicability
of FPXRF to meet DQOs for a given site
therefore requires careful evaluation of
matrix character and variability when
using empirically calibrated instruments
such as the X-Met 880.
In the authors' experience, there seems
to be some reluctance to consider FPXRF
as anything more than a screening tool.
Part of this problem stems from the
failure of site personnel to consider the
end-use of the data when defining DQOs
and evaluating potential analytical
methods. Examples come from site
managers who have asked if specific
detection limits can be achieved for
certain target elements when those limits
are one or more orders of magnitude lower
than potential action levels. Perhaps an
even more widespread problem concerns
spatial applications. When determination
of the spatial distribution of target
elements constitutes the data end-use,
DQO definition almost invariably focusses
on errors relating to sampling and
analysis and not on errors relating to
spatial interpolation. This problem is
so significant in terms of the
application of FPXRF that it warrants
some detailed consideration.
The inferential link between samples and
the spatially distributed population they
are intended to represent is established
through the process of spatial
interpolation. This process consists of
estimating concentration values at
unsampled points, usually located at the
nodes of a regular grid, by applying an
appropriate algorithm to sample values.
It results in a spatial model of target
element concentration, and serves as the
basis of such graphical decision-making
tools as isometric diagrams and contour
plots. The reliability of decisions
based upon a spatial model ultimately
depend on estimation errors incurred at
the grid nodes.
500
-------
Geostatistical theory shows that spatial
estimation errors consist of several
components related to sample collection,
preparation, and analysis, and of a
component related to spatial extension of
the sample value to an unsampled
location'5'. Extension error is a
function of the spatial variability of
concentration and of the distances
between samples and grid nodes, and is
usually much larger than sample-related
errors1*'. Thus, the largest source of
error can be reduced most effectively by
increasing sampling density rather than
by improving the precision and accuracy
of the analytical method. It follows
that DQO definition for spatial
applications should focus on spatial
estimation errors rather than on
analytical errors. It is possible to
meet or exceed high DQOs with an
inexpensive method such as FPXRF, even
though its precision and accuracy may be
less than those of CLP wet-chemical
methods. This is achieved by employing
higher sampling densities, at less cost,
thereby reducing the largest source of
error in the spatial model that
represents the instrument of decision.
The proviso, of course, is that the
concentrations of interest exceed the
detection limit of the FPXRF for the
target element under the prevailing
matrix conditions.
To demonstrate the foregoing points,
Figure 3 shows a variogram model
describing the spatial variability of
lead concentrations obtained with the
FPXRF method in contaminated soil at a
Superfund site. Intersection of the
model at a high value along the GAMMA
axis indicates high analytical error. A
second variogram in Figure 3 represents a
model with no analytical error, clearly
an ideal but unachievable situation.
Using these models, the spatial errors
incurred in estimating the average lead
concentration of a 50' x 50' x 1'
remediation block were calculated for
different densities of samples taken on a
square grid pattern within the block (see
Table 3). Direct analytical cost of $150
per sample was assumed for the ideal
analytical case, and $13 per sample
location was assumed for the FPXRF case.
The latter value is based on CLP
laboratory analysis of 25 calibration
samples at $150 per sample, and a
sampling campaign totaling 300 hundred
sample locations. Applying these values
to the number of samples for each
sampling density results in the total
costs shown in Table 3. Spatial
estimation error, expressed as relative
standard deviation, are plotted against
sampling costs for each case in Figure 4.
These plots show that FPXRF can achieve
levels of spatial estimation error
similar to those of the best laboratory
methods, and that a given level of error
can be achieved by FPXRF at significantly
less cost than laboratory methods. This
means that FPXRF should not be restricted
to screening, but can also be employed
for site characterization and remedial
evaluation sampling in many situations.
As before, these conclusions are based on
the assumption that concentrations of
interest are above the FPXRF detection
limits for the given situation.
CONCLUSIONS
Based on results achieved thus far, FPXRF
has demonstrated the capability of
providing data necessary for screening
and characterizing many inorganic
contaminants, both rapidly and
inexpensively. The authors believe that
it will play an important and perhaps
central role in site remediation in times
to come. Like any other analytical
method, FPXRF has limitations as well as
strengths. Many of these will be
improved by recent and future
improvements in technology. Sensitivity
and detection limits, for instance, can
now be materially improved with high
resolution detector systems designed for
field portable instruments. Calibration
constraints can be relaxed with
incorporation of fundamental parameter
techniques and better software systems.
However, several points need to be
emphasized for attainment of optimal
results when field sampling with the X-
Met 880 and similar instruments:
f Site specific calibration standards
are absolutely necessary to obtain
defensible quantitative data.
»• Proper sampling protocols must be
designed to allow quality assessment
of the data.
* DQOs must be correctly defined.
»• Geostatistical procedures are
essential for proper definition of
DQOs and for QA evaluation of
spatially distributed FPXRF data.
If these points are followed, the authors
believe that FPXRF can meet or exceed
501
-------
traditional CLP procedures more cost-
effectively in many situations.
REFERENCES
(1) Glanzman, R.K. 1990. personal
communication. Senior
Geohydrologist/Geochemist. CH2M
HILL, Denver, CO.
(2) U.S. EPA, "Contractor Laboratory
Program Statement of Work for
Inorganic Analyses, SOW No. 788,"
Attachment A, U.S. Environmental
Protection Agency, Washington, D.C.,
1989.
(3) Wheeler, B.D, "Accuracy in X-ray
Spectrochemical Analysis as Related
to Sample Preparation,"
Spectroscopy 3., 1988, 24-33.
(4) U.S. EPA, "Data Quality Objectives
for Remedial Response Activities:
Development Process," EPA/540/G-
87/004, U.S. Environmental
Protection Agency, Washington, D.C.,
1987, 154pp.
(5) Journal, A.G. and Huijbregts, C.J.,
"Mining Geostatistics," Academic
Press, Inc., New York, 1978.
(6) U.S. EPA, "Soil Sampling Quality
Assurance Guide, Second Edition,"
EPA/600/8-89/046, U.S. Environmental
Protection Agency, Las Vegas, 1989
pp. 26-27.
NOTICE: Although the research described in this article has been
supported by the United States Environmental Protection Agency through
Contract No. 68-CO-0049 to Lockheed Engineering & Sciences Company, it
has not been subjected to Agency review and therefore does not necessarily
reflect the views of the Agency and no official endorsement should be
inferred.
Table 1. Comparison of CLP DQOs to FPXRF Performance
for soil samples.
Analytical Method
CLP Spectroscopic (AA/ICP)
FPXRF
Accuracy
+/- 25 %
+/- 50 %
Precision
+/- 35 %
+/- 35 %
Table 2. Comparison of Detection
Limits for CLP and FPXRF.
(all values in mg/kg)
Analytes
Cu
Zn
As
Pb
CLP*
5
4
2
0.6
FPXRF
410 - 470
105 - 200
100 - 250
120 - 513
* Assumed soil weight of 1 g and end volume of 200 mL.
DISCUSSION
JIM PASMORE: I have a question on the data you showed for precision. What
was your measurement time on that, or the precision measurements, table 2.
BILL COLE: Thirty seconds.
STEPHEN KNOLLMEYER: I was just wondering why you didn't have any
detection limits for the cadmium? Didn't you take it?
BOB ENWALL: That was just a slight mistake on our part. We haven't really
had any experience with cadmium that would give us a number.
502
-------
Traditional Pathway for Sampling and Analysis
PACKING & SHIPPING
FIGURE 1. Traditional Approach to Site Characterization.
503
-------
Table 3 . Comparison of Costs for Estimating the Average
Concentration of a 50' x 50' x 1' Remediation
Block Using Different Sampling Densities.
SAMPLE DENSITY
(# SAMPLES)
1
4
9
16
25
36
49
ESTIMATION
ERROR
FPXRF
1.8298
0.8447
0.5510
0.4094
0.3268
0.2709
0.2318
ESTIMATION
ERROR
IDEAL ANAL.
0.9211
0.2977
0.1604
0.1067
0.0825
0.0629
0.0522
COST
$
FPXRF
13
52
117
208
325
468
637
COST
$
IDEAL ANAL.
150
600
1350
2400
3750
5400
7350
Screening Data
to Computer
Develop Semivariogram
from Geostatistics
Develop Kriging
Concentration Isopleth
Maps, Strategy and
Sampling Plans from
Geostatistics
3-
2—
10 20 30 tti tO 60 70
Separation Oixtanc*
FIGURE 2. FPXRF Approach Employing In Situ Measurements.
504
-------
6.0-
4&-
4.0-
&6-
9.0-
H
2.0-
0.6-
0.0-
.A.
A A
ZT
I I I I I I I I I [ I I I—I | I I I I |—I I I I | I I I I |
10 20 80 40 SO M
i I i i i i I i i i i I i i i i I i i i i
70 M
I ' ' ' ' I
80 100
DISTANCE (to*)
aouo LINE: mar MODEL DASHED UNE OEM. ANALYTICAL MODEL
RGURE 3. RELATIVE VAR10GRAM8 FOR FPXRF AND IDEAL ANALYTICAL CASES.
2.0-f
1.5-
; 14)-
0.6-
ao-i
1 ' ' M
HX>
woo
COST (DOLLARS)
SOUOUNE FPXRF CA9E DASHED UMB DEAL ANAIYIICAL CASE
RQURE 4. ESTIMATION OF A 50' X 50' X1' REMEDIATION BLOCK. ESTIMATION
ERROR VS. COST FOR FPXRF AND IDEAL ANALYTICAL CASES.
505
-------
A HIGH RESOLUTION PORTABLE XRF HgI2 SPECTROMETER FOR FIELD
SCREENING OF HAZARDOUS WASTES.
J. B. Ashe
Ashe Analytics
Austin,
Texas 78746
P. F. Berry and G. R. Voots
TN Technologies, Inc.
Round Rock, Texas 78664
M. Bernick
Roy F. Weston, Inc.
Edison, NJ 08837
G. Prince
U.S.E.P.A.
Environmental
Response Team
Edison, NJ 08837
ABSTRACT
A field portable XRF spectrometer based on
a mercuric Iodide (Hgl,) semiconductor x-ray
detector is described. Its multi-element
capabilities will be illustrated with mea-
surements on chemically-analyzed samples
representing materials collected from sever-
al hazardous waste sites containing differ-
ent metallic pollutants in a variety of soil
matrices.
The range of the analyzer extends from Ca to
U, and a typical configuration provides for
about 20 elements which are simultaneously
reported together with the analytical preci-
sion. Minimum detection limits for most
elements are in the range of 50 -200 mg/kg
with a 200 second analysis time.
The solid state detector is operated near
ambient temperature and affords an energy
resolution of better than 300 eV for the Mn
K x-rays. Intrinsic detector efficiency
exceeds 60% for energies up to 100 keV.
Dual radioisotopic source excitation is
provided from the list: Fe-55, Cd-109, Cm-
244, Am-241. A 6.5 kg battery-powered Data
Processing Unit features menu-driven opera-
tion and on-board dual 2000 channel MCA
spectrum display capability. Internal
storage provides for the retention of 30
spectra and 100 multi-element analytical
reports.
A "fundamental parameters" based analysis
algorithm is used to compute elemental
concentrations. This computational ap-
proach, together with the comprehensive
element coverage, permits "standardless"
measurements over a wide range of material
compositions.
INTRODUCTION
Soil contamination by hazardous metallic
waste is present at the level of concern on
more than 50% of the sites on the National
Priorities List. Complete evaluation of the
degree of hazard and measurement of the
spatial extent of those hazards involves the
analysis of literally millions of samples.
Although procedures have been approved by
the Environmental Protection Agency for use
under the Contract Laboratory Program for
performing enforcement quality analyses on
environmental samples, a need still exists
for rapid, reliable, and cost-effective
assays to expedite the characterization and
remediation of sites. Ideally such assays
could be accomplished by direct measurement
in the field.
Several investigators (1-15) have reported
on the application of energy dispersive
x-ray fluorescence (XRF) to the assay of
metal-contaminated soil. Savings in time and
analysis cost over the standard EPA-approved
chemical methods are significant (1,5,11).
Essentially a nondestructive technique
suited to the measurement of almost any kind
of material in powder, liquid or solid form,
XRF is further distinguished by its ability
to analyze for many elements, including the
unexpected, in a truly simultaneous fashion.
Thus it offers low cost and rapid turnaround
time per analysis.
By means of XRF, two environmental applica-
tion methodologies - in situ assay with
field portable instruments and intrusive
sample analysis with laboratory grade equip-
ment - are currently being pursued and are
reported to yield good definition of the
magnitude and extent of contamination -
507
-------
particularly when coupled with geostatisti-
cal sampling methods and data analysis
(12,14). Laboratory-grade instruments gener-
ally offer better precision, higher accuracy
and greater sensitivity than portable equip-
ment since, with no weight and power re-
strictions, x-ray tube excitation (as op-
posed to radioisotopes) and high resolution
spectrometers of the cryogenically-cooled
semiconductor Si(Li) detector type can be
used. Portable instrument designs have so
far employed gas proportional detectors and
their capability for multielement applica-
tion has been somewhat impaired by the
limited energy resolution of that type of
detector. The in-situ measurement capabili-
ty of a portable instrument, albeit at some
trade-off in analytical performance, is of
advantage in reducing the time delays and
data-integrity risks associated with sample
handling procedures. Other benefits are its
low cost per analysis, its utility in delin-
eating hot spots as an aid in the collection
of samples for enforcement quality assay or
to guide the work of site remediation.
Quantitative XRF application always requires
an appropriate calibration, usually by
measurement on a representative suite of
chemically known standards. Alternatively,
as is now the option on most laboratory-
grade instruments using Fundamental Parame-
ter (FP) methods of analysis, only pure
elements or a few standards ( which need not
be of site-specific composition) are re-
quired. Currently available portable instru-
ments with their lower x-ray resolving power
are more restricted to the use of an empiri-
cally structured analysis algorithm of
limited element coverage. They, therefore,
require a multi-sample calibration on site-
specific material (11). This is a major
drawback to their general use, and the
quality of the resultant analyses is opera-
tor sensitive and highly dependent on the
validity of the calibration samples.
The development of a high-resolution non-
cryogenic semiconductor x-ray detector has
made possible a new field-portable instru-
ment design which can provide for the appli-
cation of FP analysis of soils using a site-
independent calibration based on pure ele-
ments standards. Our paper will present some
of the results obtained with this instru-
ment.
INSTRUMENTATION
The instrument used to evaluate the applica-
tion of an FP-based XRF method for field
analysis of metal-contaminated soil is shown
in Figure 1. Similar instruments are now
used industry-wide for on-site verification
of alloy materials (16). The system operates
CD CD COB
Figure 1
Field-Portable XRF Spectrometer
off either AC-power or rechargeable NiCd
batteries and weighs approx. 17 Ib (8kg).
Compared to the alloy-analysis design, the
soil application unit was modified only in
regard to the type of isotopic excitation
sources contained in the hand-held probe,
and minor revisions to the PROM-based oper-
ating software in the data processor module.
The sample measurements we report were
performed with an isotope combination of Cd-
109 and Am-241; each of an effective 3mCi
source activity. The sources are separately
shielded in a motorized turret and are
positioned for measurement under program
control. An on-screen set-up menu allows the
individual source exposure time to be se-
lected from 1 to 999 sees. All reported
measurements used a 200 sec. selection.
Probe accessories such as a detachable base,
shown in the figure, and clip-on front-end
attachments, facilitate the measurement of
contained samples, but the main utility of
the probe is afforded by its compact, hand-
size, design for direct application to in-
situ material. Measurement is initiated by
momentary push-button action, either at the
probe or on the instrument panel, so free-
standing long period assay and, as required,
operation within an environmentally sealed
plastic enclosure are quite practical. A
tough replaceable x-ray entrance window also
seals the probe face over a measurement
aperture of 0.5 x 0.75 inch (1.25 x 2.0 cm).
508
-------
Spectrometer-Analysis Operations
The x-ray analytical capability of the
instrument is established mainly by its use
of a new-technology, high resolution energy
dispersive x-ray detection device based on
semiconductor HgI2. This detector is con-
tained in a capsule within the probe and
operates at a controlled, less-than-ambient
temperature by low-power thermo-electric
cooling. The energy resolution, expressed as
FWHM, is of the order of 300 eV for the Mn
K x-ray line.
The data processor performs all of the
necessary analog/digital electronic func-
tions to translate the detected x-ray infor-
mation into quantitative analytical results.
Pulse-amplitude records, for instance,
representing the sample x-ray fluorescence
spectra are generated and stored for each
excitation source in a 2x2000-channel memo-
ry. These spectra, if desired, can be pre-
sented on the instrument's LCD panel accom-
panied with the usual control features of a
multichannel analyzer. Peak identifiers,
regions of interest, and element-line mark-
ers, for example, are operative in that mode
and are displayed in calibrated x-ray energy
units. Non-volatile RAM storage allows up to
30 spectra to be retained together with the
analytical results for more than 100 field
measurements. An RS-232 serial port is
provided for printer and computer communica-
tion.
All operations are prompted by on-screen
menus which indicate the available options
and how to proceed. An example of the "turn-
on" menu is as follows:
MAIN MENU
Enter Choice of Operation
1 SOILS ANALYSIS
2 RECALL STORED RESULTS
3 REVIEW/CHANGE SET-UP
4 STORE/RECALL SPECTRUM
5 OTHER FUNCTIONS
Routine operation proceeds by option #1,
which leads to a set up of the data acquisi-
tion times per source and initialization of
the probe controls for measurement. Measure-
ment concludes with an audible signal fol-
lowed by an on-screen report of the analyzed
elements. Results are labelled by element
symbol and, as later described, include both
element concentrations and an indication of
the computed standard deviations. Options,
such as the storage and printout of results,
follow the on-screen report if pre-selected
from the Main Menu. Spectrum operations,
available as shown, are normally by-passed
in the routine measurement sequence which
returns directly to a "ready" status after
the analysis report.
Element concentrations are computed using
Fundamental Parameter derived coefficients
in an algorithm of the form:
CONC = R x S x (1 + SUM{anxCn})
Where, "R" is the measured analyte x-ray
intensity relative to the pure element: "S"
is a calculated sensitivity coefficient: and
the quantity SUM{) is a summation of "n"-
element absorption-enhancement terms con-
taining calculated alpha-coefficients (17)
and iteratively computed element concentra-
tions. Preparation of the instrument for the
measurements reported in the next sections
entailed only a normalization to the pure
element response. No other calibration was
performed. X-ray intensities are processed
for more than 20 elements but only those
determined to be in excess of three-times
the standard deviation are presented in the
analysis report. All element x-ray intensi-
ties, however, can be viewed on the screen.
SAMPLE MEASUREMENTS
Although the instrument is capable of per-
forming in-situ measurements, the effective-
ness of an FP-analytical approach was evalu-
ated by analyzing intrusive samples so that
comparative analyses could be obtained. A
total of 55 samples were measured represent-
ing material from four NPL sites character-
ized by complex metallic contaminations. The
samples were air dried, disagglomerated but
not ground, and passed through a 20 mesh
screen (.84 mm hole). The oversize was
discarded. The undersize was split so that
replicate samples were available for the
chemical and x-ray assays. The powdered
samples, ranging in mass from 2 to 8 gms,
were placed in 30mm sample cups and covered
with a 0.005 mm polypropylene x-ray window
for measurement in the upright probe geome-
try. Sample thicknesses ranged from 8.5 to
16 mm and bulk densities of the loose pow-
ders ranged from 0.26 to 1.2 gm/cc. The
chemical assays of the sample splits were
performed by a commercial laboratory (18)
using flame Atomic Absorption (AA) analysis
for the elements of interest.
Two of the sampled sites (noted as Sites 1
and 3 in the results) are inactive metal
plating locations. The samples were collect-
ed from settling lagoons and consisted of
plating sludges mixed with the local soils.
Site 2 had been the location of a smelter;
a metal working facility; and, in recent
time, a Ni-Cd battery manufacturing opera-
tion. It is an estuary location, submerged
except during low tide, and the soil would
509
-------
be best described as contaminated bay sedi-
ment. Site 4 was a scrap metal storage-
segregation facility, involving a wide
variety of metals and metallic compounds in
open areas throughout the site.
An example of the instrument-generated pulse
height spectrum (excited by Cm-244) for one
of the soil samples from a plating lagoon is
shown in Figure 2, overlaid with a spectrum
for the same sample obtained on a gas-filled
proportional-detector instrument. The supe-
rior x-ray resolving power of the HgI2
detector is obvious and is seen to provide
well for quantitative analysis of minor
elements in the presence of adjacent atomic-
number elements at high concentration.
o
rr
x
6 8 10 12 14
X-RAY ENERGY (keV>
16
Figure 2. Example XRF Spectra for a HgI2
and a Gas Proportional Detector on the Same
Soil Sample (Cm-244 Excitation).
The instrument-reported XRF assay results of
a Cd-109/Am-241 excitation measurement on
the soil sample are given in Figure 3.
Where available, the results of prior analy-
sis by Atomic Absorption are tabulated.
DATE:
TIME:
MODE:
07/10/90
10:15:55
STANDARD
COMPOSITION:
C;i
Cr
Fe
Co
Ml
Cu
Zn
Cd
Sn
f'b
CONC.
(PCNT)
5.035
0.513
8.453
.118
1.570
1.661
11.865
.683
.453
.047
ST. DEV.
(PCNT)
0.705
.055
.064
.024
.020
.015
.053
.005
.005
.002
X-RAT
(mg/kg>
50,035
5,130
84,453
1,180
15,700
16,610
118,865
6,830
4,530
470
ATOMIC ABSORP.
(mg/kg)
*
6,600
*
*
16,000
*
110,000
7,900
*
430
The format of the XRF results shown in the
left-hand columns of Figure 3 is that of the
resident operating software which reports
both the element concentrations and the
computed statistical uncertainties in units
of weight-per-cent-element. For comparison
with the AA mg/kg assays in the figure, and
for the results discussed in the next sec-
tion, a factor of 104 has been applied to
the x-ray data. In this single-sample
comparison, the XRF and AA results are seen
to be in good agreement for the elements
measured by both techniques. However, a
number of contaminants not known to be at
the site were detected at a statistically
significant level by XRF.
COMPARATIVE RESULTS
Graphical comparisons of the XRF results
with AA analyses for the 55 split samples
from the four sites are presented in Figures
4-9. Because of suspected intra-sample
heterogeneity, a statistical comparison of
the results of the two measurements on a
simple point-by-point basis was not consid-
ered appropriate. Under such circumstances,
global data comparisons are more valid. Two
global methods were used: The XRF results
were regressed against the AA results to
reveal any relative biases in the data sets;
and the relative percent difference (RPD)
between the two determinations were calcu-
lated to indicate the average disparity
between the analyses. Any disparity would
reflect the imprecision of each determina-
tion and real chemical differences in the
sample splits. Where the measurements span
a range of values well above the detection
threshold, the average RPD should be a good
indicator of the intra-sample heterogeneity
uninfluenced by the analytical precision of
either technique. The average RPD's from
samples with contamination levels greater
than 10 times the standard deviation were
computed for each analyte and are shown in
Table 1 along with the main results of the
regression analysis:
Table 1 Some XRF vs. AA Correlation Data
* = Not Analyzed
Figure 3. X-Ray Analysis Report and AA
Assay Data for the Soil Sample of Figure 2.
ANALYTE
Cr
Fe
Ni
Zn
Cd*
Cd"
Pb
0
0
0
0
0
0
RANGE
(mg/kg)
- 28,000
- 350,000
- 20,000
- 150,000
- 10,000
- 580,000
REGRESSION DATA
SLOPE
1.02
1.36
1.11
0.99
0.84
1.10
1.06
a sip.
.09
.05
.05
.02
.07
.06
.09
K2
0.84
0.98
0.96
0.9S
0.90
0.95
0.85
RELATIVE
% OIFF (RPO)
38
23
30
23
33
17
30
Cd uncorrected for bulk density
Cd corrected for bulk density (see text)
510
-------
XR
30000
24000
18000
12000
RfYW
0
F mg/kg CHROMIUM
"*u #
•
ff
*
D
D/"
D
7
/
/
n
Sltel
Site 3
Site 4
0 6000 12000 18000 24000 30000
Cr, Atomic Absorption, mg/kg
XRI
20000
15000
10000
lyyyv
Q.
F mg/kg NICKEL
•D /
"/*
D
/
/
/
D
SHe1
B
Site 2
•
Site 4
0 5000 10000 15000 20000
Ni, Atomic Absorption, mg/kg
Figure 4
Comparative Assay Results for Chromium
Figure 6
Comparative Assay Results for Nickel
XRF
500000
400000
200000
100000
QJ
mg/kg IRON
/
/
m
/
/
Site 2
Site 4
0 100000 200000 300000 400000 50OOOC
Fe, Atomic Absorption, mg/kg
XRF
80000
60000
40000
OfWYV
Q|
mg/kg ZINC
a ^
£
a
ef'
<'a
a
a
/'
/
' u '
a
Sltel
•
Site 2
Site 4
0 20000 40000 60000 80000
Zn, Atomic Absorption, mg/kg
Figure 5
Comparative Assay Results for Iron
Figure 1
Comparative Assay Results for Zinc
511
-------
XRF mg/kg CADMIUM
15000
12000
9000
6000
3000-
0
<
Dens. Corrected
Uncorrected
. . •
J^
/
"
a QD
•
X
] a
/
3000 6000 9000 12000 15000
Cd, Atomic Absorption, mg/kg
XRF
600000
450000
300000
150000'
0
(
mg/kg LEAD
•
...••'
..-•'
v
I
/
/
/
.-'
/
/
n
Srtel
Site 2
Site 4
) 150000 300000 450000 600000
Pb, Atomic Absorption, mg/kg
Figure 8
Comparative Assay Results for Cadmium
(Illustrating Bulk Density Corrections)
Figure 9b
Comparative Assay Results for Lead
(High Range)
XF
800
600
400
200-
F mg/kg LEAD
K W
:
_
/
n /
[-3 "
D i
•
n
j ..'-''
__
a
Sitel
Site 2
Site 4
—
0 200 400 600 800
Pb, Atomic Absorption, mg/kg
Figure 9a
Comparative Assay Results for Lead
(Low Range)
Concerning the density correction on the Cd
assays (noted in Table .1), it is generally
recognized that for XRF analysis of loosely
packed powder material there is always some
influence of bulk density on the measured x-
ray intensity. The magnitude of the effect
depends on the excitation and fluorescent x-
ray energies and was investigated by mea-
surement on a specially prepared suite of
"doped" samples covering a density range of
0.2 to 2.0 gm/cc. A functional dependence
of the x-ray intensity, relative to a "high
density" sample, was derived to be of the
form;
(mu*rho)/(mu*rho+l)
where "rho" is the bulk density and "mu" is
a calculated absorption coefficient for the
analyte x-rays. The value of "mu" decreases
with increasing x-ray energy and, in low
bulk density material, becomes a factor in
the XRF analysis of elements such as Cd, Sn,
Sb, etc., excited by the relatively high
energy emission of Am-241. Over the density
range of the measured samples (0.26 to 1.2
gm/cc) the values of the calculated x-ray
intensity correction were in the range of
1.2 to 1.9. The results are illustrated in
the data of Figure 8.
The XRF precision data reported alongside
the assay values (as previously noted in
Table 1) are based on the calculated statis-
tical errors associated with each analyte x-
512
-------
ray intensity and include the compounding
effect of the iterative solution of the
multielement analysis algorithm. The data,
therefore, can convey a realistic indication
of analytical detection limit, consistent
with all of the conditions of measurement.
An evaluation of the precision was conducted
by repeated independent analysis on the same
undisturbed sample over a three-day period
and the results are shown in Table 2.
Table 2 Precision Evaluation Data
ANALYTE
Fe
Ni
Cu
Zn
Cd
Sn
Pb
Bi
ANALYTE
CONCENTRATION
-------
expected that similar accuracy and precision
values would be achieved for other analytes
in the same part of the periodic table as
those reported here.
The MDL's reported in Table 3 were deter-
mined under the available but less-than-opt-
imum instrumental parameters. Count rates
with the sources employed were substantially
lower than the practical limit of the sys-
tem. The detector head, shown in Figure 1,
is designed for optimum presentation of
small metallic samples commonly encountered
in alloy-sorting applications. A sample
presentation geometry of more appropriate
design for soil assay should yield a signif-
icant improvement in the measurement effi-
ciency. Most data were acquired with the
source pair of Cd-109 and Am-241. Chromium,
and perhaps other analytes, could be more
sensitively assayed with other isotopic
sources. Improvements in these areas of the
instrument design are expected to lower the
MDL of all analytes by at least a factor of
two, and should lower the MDL of Cr into the
150 - 200 mg/kg range.
CONCLUSIONS
The results obtained from this study clearly
show the potential of a Fundamental Parame-
ter approach to the analysis of soil contam-
ination with a high-resolution energy-dispe-
rsive XRF analyzer of field portable design.
The operational convenience of a calibration
based only on measurement of pure element
standards is well demonstrated. Considering
the diversity of soil types tested and their
wide variation in the level of contamina-
tion, the overall accuracy is good and
certainly adequate for screening tests where
operational requirements are generally set
at +/- 50% accuracy and +/- 1°% precision
(20). The results reported here easily
satisfy those requirements.
FOOTNOTES and REFERENCES
1. Furst, C.A.; Spittler, T. and Titlinghast, V. "Screening
for Metals at Hazardous Waste Sites: A Rapid Cost-Effec-
tive Technique Using X-Ray Fluorescence" Proc.of 6th
National Conference on Management of Uncontrolled
Hazardous Waste Sites. Hazardous Materials Control
Research Institute, Silver Springs, MO. (1985)
2. Grupp, D.J.; Everitt, D.A.; Bath, R.J. and Spear, R.
"Use of a Transportable XRF Spectrometer for On-site
Analysis of Hg in Soils" American Environmental Labora-
tories, pp 32-40, Nov, 1989
3. Grupp, D.; Fendler.K.; Mathers, V.; Bath, R. and Spear,
R. "On-Site Multi-Element Analysis of Hazardous Waste
Site Soil Samples Using X-Ray Fluorescence" Conf. Proc.
HAZTECH 90, Houston, TX. pp 8B-887 - 897 (1990)
4. Jacobus, N. "Screening of Hazardous Waste With an Energy
Dispersive X-Ray Fluorescence Spectrometer" Advances in
X-Ray Technology, Vol. 33, p655-663. C.S. Barrett et al,
ed. Plenum Press, New York (1990)
5. Meiri.D.; Bradfield, D.G. and Downs, D.M. "Delineation
of Heavy Metals in Surface Soil by Portable X-ray
Fluorescence Analysis Screening" Fourth National
Outdoor Action Conference on Aquifer Restoration, Ground
Water Monitoring and Geophysical Methods, pp 1067-1079.
Las Vegas (1990)
6. Mernitz, S. and Staible, T. "Use of a Portable X-Ray
Analyzer and Geostatistical Methods to Detect and
Evaluate Hazardous Metals in Mine/Milt Tailings" Proc.of
6th National Conference on Management of Uncontrolled
Hazardous Waste Sites. Hazardous Materials Control
Research Institute, Silver Springs, MD. (1985)
7. Perils, R. and Chapin, M. "Lou Level XRF Screening
Analysis" Fifth International Symposium on Field
Screening Methods for Hazardous Waste Site Investiga-
tions, pp 81-94. (1988)
8. Piorek, S. and Rhodes, J.R. "Hazardous Waste Screening
Using a Portable X-Ray Analyzer" Symposium on Waste
Minimization and Environmental Programs within D.O.D.,
Long Beach, California, April 1987
9. Piorek, S. "XRF Technique as a Method of Choice for
On-site Analysis of Soil Contaminants and Waste Materi-
al" Advances in X-Ray Technology, Vol. 33, p629-637.
C.S. Barrett et al, ed. Plenum Press, New York (1990)
10. Raab, G.A.; Faber, H.L. and Simon, S.J. "Evaluation of
a Prototype Field-Portable X-Ray Fluorescence System for
Hazardous Waste Screening" EPA/600/4-87/021. U.S.
Environmental Protection Agency, Las Vegas, Nevada.(1987)
11. Raab, G.A.; Kuharic, C.A.; Cole III, W.H.; Enuall, R.E.
and Duggan, J.S. "The Use of Field-Portable X-Ray
Fluorescence Technology in the Hazardous Waste Industry"
Advances in X-Ray Technology, Vol. 33, p629-637. C.S.
Barrett et al, ed. Plenum Press, Neu York (1990)
12. Raab, G.A.; EnwalL, R.E.; Cole III, W.H.; Faber, M.L. and
Eccles, L.A. "X-Ray Fluorescence Field Method for
Screening of Inorganic Contaminants at Hazardous Waste
Sites" Hazardous Waste Measurements, M.Simmons, ed. Lewis
Publishers Chelsea, MI (in press,1990)
13. Sackman, A.R.; Perlis, R. and Chapman, M. "Applications
of X-Ray Fluorescence Spectroscopy for Site Screening"
14. Walsh, J.P.; Harding, T and Aulenbach, S. "Using X-Ray
Fluorescence Spectrometry and Geostatistics for Soil-Me-
tal Contamination" Proceedings of Conference on Plan-
ning, Rehabilitation and Treatment of Disturbed Lands, pp
307-317. Billings, MT (1990)
15. Watson, W.; Walsh, J.P. and Glynn, B. "On-site X-ray
Fluorescence Spectrometry Mapping of Metal Contaminants
in Soils at Superfund Sites" American Laboratory, pp
60-68. July, 1989.
16. Berry, P.F. and Voots, G.R. "On-site Verification of
Alloy Materials with a Neu Field-Portable XRF Analyzer
Based on a High-Resolution HgI2 Semiconductor X-ray
Detector" Proc. 12th World Conf. on Non-Destructive
Testing (Amsterdam). Vol.1. p737. 1989. Elsevier Science
Pub.
17. de Jongh W. K. "X-ray Fluorescence Analysis Applying
Theoretical Matrix Corrections" X-ray Spectrometry 2.
p151, 1973
18. Atomic Absorption analyses were performed by Jayanti
Patel and Pradeep Gupta of Roy E. Weston's REAC Laborato-
ry.
19. American Chemical Society Committee on Environmental
Quality "Principals of Environmental Analysis." Anal.
Chem 55 (1983)
20. Raab, G.A.; Faber, M.L. and Simon, S.J. "Development of
a Field-Portable X-Ray Ftuorsescence System for On-Site
Hazardous Waste Screening" Lockheed Engineering and
Management Services Co, Report for Contract No. 68-03-3-
249.
514
-------
DISCUSSION
JOHN MORRIS: The fundamental parameters technique assumes that it's PETER BERRY: The value of the fundamental parameters is only really
dealing with all the components of the system. This question is twofold. Do you achieved if you measure all of the elements. And so naturally we do expose all
routinely switch over radionuclide sources to span spectrum for all samples to do of the sources to obtain the data to apply that model. The light elements are not
this? And do you make assumptions about the light component, or do you practically measured in the field, so everything is expressed relative to a lighter
calculate that with the Raley-Compton ratio? element like silica. Elements like calcium can be determined and the coefficients
derived, but we cannot measure silica and quantify that. So that one is assumed
to be the balance of the material.
515
-------
LOW CONCENTRATION SOIL CONTAMINANT
CHARACTERIZATION USING EDXRF ANALYSIS
A.R. Harding
Spectrace Instruments, Inc.
345 E. Middlefield Rd.
Mountain View, CA 94043
INTRODUCTION
Effective assessment and remediation of
hazardous waste sites dictates that
analytical methodologies be developed which
assist in the evaluation of site
contamination and simultaneously make
efficient use of sampling time and resources
(1). Optimally, a technique would provide
on-site personnel with immediate and
accurate information concerning the identity
and concentration of inorganic soil
contaminants (2).
Inorganic pollutants can be readily
determined in contaminated soils with energy
dispersive X-ray fluorescence spectrometry
(EDXRF) using a thermoelectrically cooled
Si(Li) detector (3). A field mobile
laboratory van or trailer can accommodate
the EDXRF system because the electrically
cooled detector, which provides high
resolution EDXRF spectra, does not require
cryogenic cooling. Soil sample preparation
for EDXRF analysis is minimal, therefore,
short turnaround times are realized between
sampling and reporting results.
This report will describe an EDXRF method
developed to determine four inorganic soil
contaminants: lead, arsenic, zinc, and
cadmium at four sampling depths, the EDXRF
results for approximately one hundred eighty
soil samples will be compared to results
obtained for sample splits submitted for
analysis at an independent laboratory.
Evaluation of low concentration arsenic
detectability with elevated lead
concentrations in these samples will be
discussed. Accuracy and precision of the
EDXRF method will also be compared to the
independent methods using a standard
reference material and soil samples
submitted in triplicate to both
laboratories.
EXPERIMENTAL
The field mobile EDXRF spectrometer used in
this work was a Spectrace 6000 (Spectrace
Instruments, Inc., Mountain View, CA). The
EDXRF system consists of three modules: the
spectrometer, the control/pulse processing
electronics, and the data analysis computer.
The compact size and weight (90 Ibs.) of the
modules permits installation of the system
in a laboratory trailer or van.
The bench top spectrometer module, which can
accommodate a single soil sample, is powered
by 110 V line or generator feed. The
excitation source used is a low powered Rh
anode X-ray tube (50 kV, 0.35 mA (17 W)
maximum output) positioned at a 45° incident
angle to the sample. Three primary
radiation filters permit optimum spectral
acquisition conditions to be computer
selected.
The thermoelectrically cooled Si(Li) X-ray
detector is mounted at a 45° take-off angle
in an inverted geometry with respect to the
sample. The 20 mm2 Si(Li) crystal, which is
protected by a 0.5 mil Be window, is cooled
to -90°C for operation using a multi-stage
thermoelectric (Peltier effect) cooler. The
300 watts produced at the detector heat sink
are dissipated by forced ambient air.
Thermoelectrically cooled detectors provide
typical resolutions of 185 eV (Mn Ko).
517
-------
A card cage module is interfaced between the
spectrometer and an IBM PS/2 or PC/AT series
computer. The card cage components include
the detector high voltage supply, the pulse
processing electronics, and the control
circuit board for the EDXRF spectrometer.
The data analysis software executed on the
PC is capable of either a fundamental
parameters or empirical data treatment
scheme using a combination of standard
reference materials and/or site specific
standards.
Sampling of the suspected waste site was
performed using EPA approved protocols in a
9500'x 3500* rectangular area. Forty three
(43) cores were collected and partitioned
into four depth levels: surface to 2"; 2" to
6"; 6" to 12"; and 12" to 18", and
designated levels 1 through 4, respectively.
At the site, samples were first homogenized
and then split into two fractions. One was
submitted for EDXRF analysis and the other
sent to an independent lab for analysis.
The independent laboratory used EPA SW 846
(methods 3050 and 6010) methodology to
determine Cd, Pb, and Zn concentrations in
the soil sample splits. Arsenic was
determined in those splits using SW 846
method 3050 and EPA method 206.4
(spectrophotometric).
Sample preparation for EDXRF analysis
consisted of drying the sample for 4 minutes
in a microwave oven followed by sieving the
dried sample. Material passing through the
2 mm sieve was collected as sample and was
free of large foreign objects such as
pebbles and sticks. Drying the sample was
required due to the variable moisture
content in the submitted soils; some surface
samples had the consistency of mud. The
sieved soil was then ground in a Spex
shatterbox grinder (Spex Ind., Edison, NJ)
using tungsten carbide cups for 2 minutes.
Grinding cups were subsequently cleaned
using soap and tap water. The cleaned cups
were rinsed with distilled/deionized water
followed by isopropanol. Approximately 5
grams of prepared sample were poured into a
disposable 32 mm X-ray sample cup and
covered with a 6.3 ^m polypropylene film.
Five grams of dried sample gave the
equivalent of a 15 mm sample depth in the
cup. Approximately twenty five samples were
prepared and analyzed per day.
STANDARDIZATION METHOD
Two sets of excitation conditions were
employed to determine seven elements in the
soil samples, four of which are of specific
environmental concern: Zn, As, Pb, and Cd.
Table 1 lists the two sets of spectral
acquisition conditions and which conditions
were used to determine each analyte. Figure
1 is a mid Z spectrum of a soil sample that
was found to contain 125 ppm As, 1100 ppm
Pb, and 729 ppm Zn. A multiple linear least
squares peak fitting routine was used for
deconvolution of overlapped peaks.
The soil characterization method was
standardized using four standard reference
materials (SRM): NBS 1648 (urban
particulate); NBS 2704 (river sediment); SO-
1 and SO-3, two soil standards available
from the Canada Centre for Mineral and
Energy Technology. Standards labeled NBS
are available from the National Institute
for Standards and Technology (NIST). These
SRMs have certified concentrations of Fe,
Mn, Cu, Zn, Pb, and Cd.
A fundamental parameters (FP) method (5) was
employed as the data treatment scheme and
used certified concentrations of Fe, Mn, Cu,
Zn, Pb, and Cd in the four standard
materials. To compute instrumental
sensitivity (emission peak counts per second
per ppm), the balance of the standard was
assumed to be comprised of Si05 to account
for the contribution of the matrix on the
measured analyte X-ray intensity. The
balance component Si05 was selected to mimic
the concentration of Si and 0 in typical
soils, approximately 24% Si and 45% 0.
Since none of the selected SRM's contain
arsenic, As sensitivity (cps/ppm) was
determined using a fundamental parameters
theoretical calculation based on the
computed Zn sensitivity. Table 2 lists the
analyte sensitivities computed by the FP
method.
There are some advantages to using an FP
method for standardization compared to site
specific soil standards. The FP method can
use readily available, well characterized
SRMs to measure analyte sensitivities. Site
specific soil standards, by contrast, are
usually collected with a separate sampling
mobilization. The FP method standardized
with SRM's can provide accurate analyte
518
-------
concentrations to be determined in samples
with fairly wide matrix variations without
restandardization, unlike methods
incorporating site specific standards.
RESULTS
Table 2 lists the lower limits of detection
determined using the two sets of spectral
acquisition conditions (Table 1). The
pertinent equation is: LLD (ppm)=
3/Ib/m(/T), where Ib is the background
intensity (cps), m is the analyte
sensitivity (cps/ppm), and T is the
acquisition livetime in seconds (6).
Calculated LLD values are dependent upon
spectrum acquisition times, sample matrix,
and excitation conditions. The conditions
in Table 1 were selected to optimize the Pb
and Cd spectral regions. Improved LLD's are
possible with EDXRF using longer spectrum
acquisition livetimes and optimized
excitation conditions for selected spectral
regions.
Results for the determination of four
analytes by EDXRF in 180 samples (43 cores
at 4 levels, two SRM's, three samples in
triplicate) were compared to independent
analysis results in order to evaluate the
level of agreement between the two methods.
Table 3 lists the correlation plot data for
the analytes in terms of actual slope,
intercept, errors, and the correlation
coefficient of the fit. Each analyte
correlation plot included approximately 150
data points.
As shown in Table 3, slopes of the plots for
Pb, Cd, Zn, and As are within 8% of 1.00 and
all correlation coefficients are greater
than 0.92. The calculated slope near 1.00
and correlation coefficients greater than
0.90 indicates agreement between the two
analytical techniques. Figure 2 is a plot
of 94 data points in the range of 0 to 300
ppm Pb. Figure 3 is a plot of 110 EDXRF and
ICP analyzed samples in the range of 0 to
100 ppm Cd and also indicates agreement
between the results of the two methods.
EFFECT OF LEAD ON EDXRF ARSENIC
DETECTABILITY
Figure 1 illustrates the spectral
interference between the emission lines of
lead and arsenic in the EDXRF measurement.
The As Ka (10.53 keV) and the Pb La (10.55
keV) peaks are directly superimposed. Peak
deconvolution software must, therefore, rely
on the relatively low intensity As KB (11.73
keV) peak for unobstructed arsenic peak
shape data. However, to the low energy side
of the As KB is the Pb Ln (11.35 keV) and to
the high energy side of the As KB are the Pb
Ln6 (12.14 keV), the Pb Ln4 (12.31 keV), and
the Pb LB! (12.61 keV) peaks which appear as
a single peak shape in the spectrum. As the
Pb emission lines increase in intensity with
increasing lead concentration the arsenic KB
peak becomes indistinguishable.
The nature of the arsenic/lead interference
in the EDXRF spectrum has a detrimental
effect on the arsenic lower detection limit
(LLD) in soils containing high Pb
concentrations. The magnitude of the
interference effect is directly related to
the resolution provided by the EDXRF
spectrometer. EDXRF spectrometers with
improved resolutions would exhibit reduced
As/Pb spectral interference. Reduced
spectral interference thereby reduces the
detrimental effect of elevated Pb
concentrations on the As LLD.
To quantify the effect of As/Pb spectral
interference on the EDXRF arsenic LLD, 148
samples in this study were evaluated. Of
the 148 samples, 43 samples were reported as
not detected for arsenic by EDXRF. Of those
43 samples, 31 were reported by the
independent lab as containing 12 ppm As or
higher. Arsenic non-detects, reported by
EDXRF, were evaluated with respect to the As
and Pb concentrations reported by the
independent lab for the same sample.
The overall findings of the 148 samples can
be illustrated using analysis results of
four samples (Table 4) as examples. Two of
the samples have non-detected As reported by
EDXRF and two had detected EDXRF arsenic
concentrations. In sample A, the
independent lab reported an As concentration
of 12 ppm while EDXRF reported a non-
detected (NO) arsenic concentration. Note
that 12 ppm is the EDXRF arsenic detection
limit. Calculating the ratio of As to Pb
concentrations, as determined by the
independent lab, a value of 0.083 was
obtained. Nearly the same ratio was found
for sample B, again where EDXRF reported a
ND while the independent lab determined 17
ppm As.
The largest absolute As concentration found
by the independent lab that was reported as
ND by EDXRF was 67 ppm As. That sample
contains 1310 ppm Pb (1217 ppm Pb determined
by EDXRF) which is an As/Pb ratio of 0.051.
EDXRF reported a non-detected As result for
all samples containing an As/Pb
concentration ratio below 0.046.
519
-------
Table 4 also lists two examples of EDXRF
successfully analyzing low concentrations of
arsenic in the presence of lead. The As/Pb
concentration ratios for samples C and D
were 0.046 and 0.053, respectively.
From the data in Table 4 and the correlation
data shown in Table 3, three findings
emerge. First, the EDXRF spectrometer used
here is unable to determine arsenic in
samples containing an As/Pb concentration
ratio of less than 0.046. Secondly, arsenic
determination by EDXRF is unreliable for
samples containing As/Pb concentration
ratios in the range of 0.046 and 0.083.
This is due, in large part, to the errors in
counting statistics for EDXRF measurements
near the arsenic LLD. Lastly, EDXRF results
show excellent correlation with the
independent lab results for samples
containing As/Pb concentration ratios above
0.083.
ACCURACY AND PRECISION
To evaluate the accuracy provided by the
EDXRF method two SRMs were submitted as
unknowns for EDXRF analysis as well as being
submitted to the independent lab for
analysis. Table 5 lists the results for SRM
SO-2. EDXRF analysis of SO-2 provides
results that are in good agreement with
certified values. The independent ICP
analysis of zinc in SO-2, however, is biased
low by a factor of one-half.
Precision was evaluated by submitting three
samples a total of three times for
independent and EDXRF analysis. Table 6
shows the results for the two methods along
with the calculated standard deviation (in
ppm) of the three replicate analyses. Note
that Cd in sample C was only reported by
EDXRF to the nearest 1 ppm and three values
of 9 ppm Cd were determined, hence the zero
standard deviation for the three replicates.
EDXRF precision is better than 10% relative
standard deviation in all but one case (As
in sample C) and compares well with that
provided by the independent lab.
CONCLUSION
Field mobile EDXRF analysis of soils
suspected of being contaminated provides
information concerning the nature, extent,
and magnitude of the contamination. Due to
the minimal sample preparation necessary for
EDXRF analysis, sampling to result
turnaround time is relatively short so the
most effective use of sampling resources is
realized. EDXRF detection limits below 20
ppm were obtained for the elements of
environmental concern. The effect of
increasing lead concentration on arsenic
detectability was quantified. Using the
EDXRF method described here, reliable As
results were found for those samples
containing As/Pb concentration ratios above
0.083. Accuracy and precision for the
analytes of interest using the EDXRF method
was shown to be comparable to results
obtained by independent analysis.
Comparable results for Cd, As, Pb, and Zn
between independent and EDXRF methods
validates the use of EDXRF analysis for
hazardous waste site investigation and
remediation.
ACKNOWLEDGMENT
The author would like to acknowledge James
P. Walsh and Associates for site sampling
and providing the independent analysis data.
REFERENCES
(1) Vincent, H.; Field Screening Methods for
Hazardous Waste Site Investigations
Symposium Proceedings. (1988), 61.
(2) Perl is, R.; Chapin, M.; "Low Level XRF
Screening Analysis of Hazardous Waste
Sites", Field Screening Methods for
Hazardous Waste Site Investigations
Symposium Proceedings. (1988), 81.
(3) Madden, N.W.; Hanepen, G.H.; Clark,
B.C.; "A Low Power High Resolution
Thermoelectrically Cooled Si(Li)
Spectrometer", IEEE Trans. Nuc. Sci.. 33,
(1), (1986), 303.
(4) Data Quality Objectives For Remedial
Response Activities. U.S. EPA, EPA 540/G-
87,003, Appendix A, (1987).
(5) Criss, J.W.; Birks, L.S.; "Calculation
Methods for Fluorescent X-ray Spectrometry-
Empirical Coefficients vs. Fundamental
Parameters", Analytical Chemistry. 40,
(1968), 1080.
(6) Currie, L.A.; "Limits for Qualitative
Detection and Quantitative Determination",
Analytical Chemistry. 40, (1968), 586.
520
-------
Table 1. Spectral acquisition conditions for
the EDXRF analysis of soils.
SPECTRAL REGION
MID Z
HIGH Z
CONDITIONS
35 kV, 0.35 mA, 0.13mm
Rh filter, 200 s livetime
50 kV, 0.35 mA, 0.63mm
Cu filter, 200 s livetime
ANALYTES
Mn, Fe, Cu
Zn, Pb, As
Cd
Table 2. Sensitivity and lower limits of detection for the
analytes of interest.
ANALYTE
Mn
Fe
Cu
Zn
Pb
As
Cd
SENSITIVITY (cps/ppm)
0.010
0.015
0.046
0.067
0.084
0.132
0.107
LLD (ppm)
21
19
26
19
7
12
4
Table 3. Correlation plot data for the four analytes
of environmental interest.
ANALYTE
Pb
As
Cd
Zn
SLOPE
1.01+0.03
1.08+0.05
1.02+0.03
1.02+0.02
INTERCEPT
10.0+13.8
0.98+3.54
3.09+2.19
63.0+13.6
CORRELATION
COEFFICIENT
0.96
0.92
0.94
0.98
Table 4. Examples of four samples illustrating the
effect of lead concentration on the arsenic
lower limit of detection. All concentration
values are in ppm.
SAMPLE
A
B
C
D
As XRF
ND
ND
28
16
Pb XRF
153
200
381
217
As AA
12
17
16
11
Pb AA
144
209
349
209
As/Pb
0.083
0.081
0.046
0.053
521
-------
Table 5. Results of the analysis of SRM SO-2 by ICP
and EDXRF methods. All values in ppm.
SAMPLE
SO-2
ANALYTE
Pb
Zn
ICP
19
55
EDXRF
17
123
CERTIFIED
21
124
Table 6. EDXRF and independent lab results for three soil
samples each analyzed in triplicate. All values in
ppm.
SAMPLE
A
B
C
ELEMENT
As
Cd
Pb
Zn
As
Cd
Pb
Zn
As
Cd
Pb
Zn
I NO. LAB
45 + 4
20 + 2
286 + 28
185 ± 15
17 + 3
80 + 6
141 + 15
556 ± 39
17 + 1
10.0 + 0.9
117 + 8
173 + 26
EDXRF
41 + 3
31 + 3
312 + 12
134 + 10
14 + 1
58 + 4
158 + 3
529 ± 46
19 + 4
9 + 0
142 + 14
128 + 3
522
-------
7
KBV
r i r r i ri r i i i i i i i
? 10 11 12
i i i i i i i i i i
13 14
13 16
Figure 1. Mid Z spectrum of a soil sample containing 1100 ppm Pb,
729 ppm Zn, and 125 ppm As. Full scale on the y-axis is
2,000 counts.
SO 100 190 200 250 300
PB (ppm) BY EDXRF
8
20 40 BO 80
CO (ppm) BY EDXRF
Figure 2. Pb correlation plot
for 94 samples.
Figure 3. Cd correlation plot
for 110 samples.
523
-------
DISCUSSION
ROGER JENKINS: Would you care to speculate on the effect of the microwave
heating on any mercury that might be in the sample. Do you think that would
drive it off or not?
ANTHONY HARDING: I don't have any experience with mercury in soils and
heating. I'm sorry.
ROGER JENKINS: Secondly, the drawings, sieving and grinding adds a
considerable amount of time to the total sample analysis time. How much do you
think that buys you in terms of increased precision or accuracy? In other words,
if we eliminated that, and just sort of mainly chopped up the sample and stuck it
in the x-ray system, what could we get?
ANTHONY HARDING: In terms of precision degradation, it would probably
be a factor of two or three. In other words, your coarsely graded sample would
be about a factor of three more or imprecise relative to the method that we took
here. And of course, if your precision has degraded, your accuracy is likely to
have degraded as well.
MARTY HARSHBARGER-KELLY: I'm familiar with medical x-ray
generating devices, and they typically use a tungsten or molybdenum target for
the mammographic units. What is the target material in your x-ray source?
ANTHONY HARDING: Rhodium.
MARTY HARSHBARGER-KELLY: And you use a rhodium filter to attenuate
the beam too, for the mid-range Z?
ANTHONY HARDING: Yes, we modify the spectral distribution from the x-
ray lube lo minimize the background and produce improved excitation efficiency
for a spectral region.
MARTY HARSHBARGER-KELLY: And the rhodium is used for both the
mid-range and the high Z metals?
ANTHONY HARDING: Rhodium is selected because it's a very good general
purpose anode material. Typically unless there is a specific excitation advantage
to going to a tungsten anode, molybdenum anode, silver anode, most applications
are done quite adequately with the general rhodium anode tube.
RUDOLF GREULICH: I'd like to ask you to comment on the background
labels you might expect in your samples. You are talking about low soil
concentrations. Those you have been showing are rather high and rather low.
Some of them might be influenced by the background. Do you know anything
about that?
ANTHONY HARDING: The detection limits were determined as interference
free detection limits in the soil matrix.
RUDOLF GREULICH: You don't know what the background levels might be?
ANTHONY HARDING: The background levels are going to be varying site to
site, in different regions.
MARK BERNICK: I'd like to comment on the, what appeared to be, 50%
recoveries of the AA lab and the zinc. And what that means in terms of the actual
accuracy of the digestive and analytical method.
ANTHONY HARDING: I wish I could comment on that particular effect.
Because it does have some ramifications in terms of our correlation data counted
later, or earlier, whatever. I really don't have a particular reason. All I know is we
submitted that sample lo two different laboratories. They were both using ICP for
thai particular soil, and they both got 55 ppm. So, unfortunately, I can't explain
it.
JOHN MORRIS: On your arsenic lead peak stripping, most of your arsenic data
were fairly low. I was wondering if you had tried MBS SRM 1645. It's the older
river sediment. It's no longer commercially available, but you ought to be able
to find some somewhere in some labs. Arsenic was 66. And the lead was 715,1
believe.
ANTHONY HARDING: That's about 10%.
JOHN MORRIS: Yes. I was just wondering whether you could do it with the
higher levels.
ANTHONY HARDING: We were able lo do it from up to 125 ppm arsenic. But
that correlation plot data was up to 125.
JOHN MORRIS: Your actual points that you showed were lower.
ANTHONY HARDING: Well, yes. I didn't show the arsenic correlation.
JACK HERNDON: I was wondering how long it takes for the detector to
stabilize after, say, a period of 24 hours or longer if the unit is turned off for that
period. How long does it lake to stabilize before you can start taking readings?
ANTHONY HARDING: From room temperature the detector takes about 45
minutes to an hour to cool down to operational temperature. I'd give it another
30 minutes for temperature stabilization an the pulse processing electronics.
We're pretty insensitive to temperature variations that are normally obtained in
a laboratory van or trailer.
JACK HERNDON: What is the range of metals that your unit can detect? How
light?
ANTHONY HARDING: We can delect sodium through uranium, atomic
numbers 11 through 92.
JACK HERNDON: Do you have a vacuum system for lower ranges?
ANTHONY HARDING: Yes.That particularchamberthat I showed isevacuable.
524
-------
DATA QUALITY ASSURANCE/QUALITY CONTROL FOR FIELD
X-RAY FLUORESCENCE SPECTROMETRY
Clark D. Carlson and John R. Alexander
U.S.E.P.A ESAT Region 10
The Bionetics Corporation
Port Orchard, Washington
ABSTRACT
Because of the nature of field screening
with portable x-ray fluorescence (XRF) spectrome-
try, the majority of quality assurance/quality con-
trol is performed prior to and following the field
activity. Prior to any field screening activity, the
calibration of the instrument is the most vital area
for QA/QC in the analysis. It is recommended that
a suite of site specific calibration samples be
prepared with soil which is representative of the
site to be investigated. This soil should be col-
lected at the site to be investigated and should
include a set of both clean samples and contami-
nated samples, if possible. The concentrations of
these samples should be verified by laboratory
techniques. These samples are then mixed and
spiked with the appropriate analytes to give the
suite of calibration samples. From this suite of
calibration samples is selected one or more
samples to be used for initial and continuing
calibration verification (ICV and CCV). During a
sampling activity, the QA/QC measures are the
assurance of representativeness of the sample,
particle size , the CCV and duplicate analysis.
Samples must also be collected periodically,
approximately 10% of samples analyzed, for
laboratory verification of the result. The results of
two field screening activities are presented with
the protocols for quality assurance of the data.
INTRODUCTION
The use of x-ray fluorescence(XRF) spectrometry
as a method of screening hazardous waste sites
for inorganic contaminants has become a viable
option due to the commercial availability of field
portable instrumentation (1). The usefulness of a
method as a screening tool depends on portability
of the instrument, the speed of the analysis, and
the precision and accuracy of the data. The size
of the instruments makes them highly portable
and the results from the analyses can be present-
ed immediately for interpretation. The question of
precision and accuracy of that data is then of the
highest priority. In studies at mines and hazard-
ous waste sites, the usefulness of data obtained
from field portable XRF units has been reported
(2).
As with any analytical method, the quality of the
data must be maintained to protect the ultimate
usefulness of any data obtained. The USEPA
defines the parameters of data quality as the
precision, accuracy, representativeness, com-
pleteness, and comparability (3). The methods for
meeting the requirements of data quality are, for
the most part, universal for all types of chemical
analysis and field screening techniques, however,
there are some special situations in XRF spec-
trometry that require unique forms of quality
control. These special situations include particle
size affects(4), and preparation of calibration
samples(S). The other, standard, QA methods are
calibration of the instrumentation(linear correlation
of the calibration), calibration verifications, dupli-
cate analyses, laboratory verification, and sample
representativeness. This paper will describe the
quality control protocols used in field screening of
various hazardous waste sites.
PROCEDURE
The field portable XRF unit used in acquiring all
data was an X-MET 880 manufactured by
Outokumpu Electronics. The X-MET 880 is an
525
-------
energy dispersive x-ray fluorescence unit with two
radioactive sources (244Cm and 241Am). For the
examples which are presented in this paper, only
the low energy source (244Cm) is used. The
calibration curves are calculated with the instru-
ments linear regression software utilizing all of the
calibration sample.
Prior to explaining the quality assurance/quality
control associated with the field use of the instru-
ment, a rudimentary knowledge of the procedures
used will assist in the evaluation of the QA/QC
protocols. Below is the general outline used
when sampling with the X-MET 880 X-ray fluores-
cence spectrometer.
1. Calibration of the instrument.
a. only necessary if elements not
previously calibrated for are re-
quired, or if a close match be-
tween the matrices of the cali-
bration standard and the sam-
ples is desired.
b. calibration can be performed
with spiked samples or previ-
ously analyzed samples. A
minimum of four samples per
element (up to 20 samples)
should be used for the calibra-
tion.
2. Check the calibration of each model to be
used (A model is a calibration including up
to six elements). This is to be done by
measuring a check sample (a control stan-
dard). If the values of the elements in the
sample are outside of one sigma (1 a), the
STA command on the X-MET 880 can be
used to reslope the calibration curve.
3. Prepare the instrument for field use. Check
the charge on the batteries (each battery is
good for 8 hours of use), inspect the instru-
ment to see that all cords are in place and
in good condition , and cover the probe
window with the polypropylene film to keep
it clean during use. Bring extra polypropyl-
ene film for replacement if necessary.
4. Upon arrival at the site, the instrument is to
be turned on and allowed to equilibrate for
at least 30 minutes. The instrument also
needs to be allowed to gain control for 5
minutes after each 20-25 minutes of use. If
the surrounding temperature is changing
rapidly, the gain control should be per-
formed at shorter time intervals.
5. Prepare the site for sampling
a. determine the frequency of
sampling
b. make a map of where sampling
is to occur (these two steps
may be performed prior to ar-
rival)
c. determine the number of meas-
urements per point on the map
which will give a representative
value for the points. This num-
ber will be between four and
seven (seven giving a confi-
dence interval of >90%)
6. Preparation of each point for measurement
Immediately prior to taking the measure-
ments at a point on the map, a representa-
tive sample to be measured must be ex-
posed.
a. for surface studies, organic
matter (stick, grass, bugs, etc.)
and any large unrepresentative
objects should be removed.
The surface may be scraped
with a rake or a shovel.
b. for subsurface studies, the
proper amount of surface mate-
rial must be removed. This can
be done using a shovel or oth-
er digging apparatus and get-
ting to the level of interest, or
526
-------
by obtaining a core sample
(with a coring tool) and mea-
suring at the proper depth.
c. the samples (four to seven) are
homogenized by the quartering
technique and sieved to 9mm.
7. After preparing each point, the measure-
ments can be obtained. One measurement
is required for each model used the analy-
sis. A counting time of 50 seconds should
be used for each measurement. This value
can be adjusted by the operator if the matrix
characteristics are such that longer or short-
er counting times are indicated. Factors
influencing this decision include particle
size, moisture content and homogeneity of
the matrix.
The data to be recorded for each measure-
ment are:
a. Concentration response for
each element
b. counting statistic for each ele-
ment (gives the standard devia-
tion due to the measuring time
used)
c. intensity data for each element
8. For quality assurance, the control sample
should be measured before the measure-
ments begin, after each 10 measurements,
and after the final measurement. As indicat-
ed earlier, the control sample should fall
within ten percent (10 %) of the actual value
or the model should be restandardized
using the standardization function on the
instrument. A duplicate analysis should
also be performed for precision analysis.
9. Repack the instrument for travel back to the
lab, inspecting for any problems.
It should be noted that the quality control referred
to in section 8 is that associated with only the field
analysis and not the pre-screening QC.
The QA/QC involved in the process of field
screening for metals with a field portable XRF unit
can be divided into two main categories, the pre-
screening QA/QC followed by the field (screening)
and laboratory (post-screening) QA/QC. These
two main categories can then be further separated
into the individual elements that make up the
protocols involved in the assurance of the quality
of the data.
PRE-SCREENING QA/QC
Prior to any field screening activity, a number of
procedures can be followed to assist in the quality
control of the final product. The first of these is in
the calibration of the instrumentation. There are
two primary methods for obtaining a calibration
curve for a field screening application which will
produce data of acceptable quality; a site specific
calibration and a generic calibration. For the site
specific calibration, a number of samples with
varying concentrations of interest must be ob-
tained for each site which will be screened. For a
generic calibration, a suite of samples prepared
by the spiking of a generic, or common, soil with
various levels of the analytes of concern. The
former of these methods will be more time con-
suming and therefore have a greater cost associ-
ated with the analysis, while the latter may not
take into consideration matrix affects from the soil
on the site. Of the two methods the one that
appears to give the best results is the site specific
calibration.
The best way of preparing site specific calibration
samples is to have a series of analyzed samples
obtained from the site of interest which contain the
proper ranges of analytes of interest. This method
would give a reliable calibration but would defeat
the purpose of having an instrument to screen a
site for possible contamination since the site
would already be well characterized. An alterna-
tive involves the spiking of samples obtained at
the site to be screened. In some cases it is not
possible to obtain samples prior to a screening
activity and in these cases a generic calibration
will be the only option.
527
-------
When a request for screening at a site is made, a
minimum of two samples from the site are re-
quired to be used in creating the calibration sam-
ples. One of the samples must be from a part of
the site which is considered to be uncontami-
nated, or "clean", while the subsequent samples
should be obtained from what is expected to be
highly contaminated areas of the site. These
samples are dried, sieved through a 9mm pore
size sieve, homogenized and analyzed in the
laboratory. The method for analyzing the samples
to be used in the calibration will depend on the
type of results are desired. The XRF can emulate
whatever method is called for in the project plan.
If the analysis is to emulate a total contaminant
digestion, then the calibration samples should be
analyzed using SW846 Method 3051 (6) followed
by ICP-AES. The data can also emulate a TCLP
type digestion or a total digestion (i.e. hydrogen
fluoride). In addition to using this data in prepar-
ing the calibration samples, the analysis may also
give some idea as to what unexpected contami-
nants may be present at the site. The clean soil
is used as the blank, the soil to be spiked, and
the diluent for the contaminated soil. The contam-
inated soil is used as the limit for the calibration
curve, unless a greater range of concentrations
than this will account for is requested. In this
case, the clean soil can be spiked at higher levels
or spikes can be added to the contaminated soils.
From the samples obtained at the site, five calibra-
tion samples are prepared. These five consist of
the clean soil, the contaminated soil, a 25/75
mixture of the two samples, a 50/50 mixture of the
two samples, and 75/25 mixture of the two sam-
ples. The next 15 to 25 samples are prepared by
spiking these five mixtures with the analytes of
interest, giving 20 to 30 calibration samples to
create the calibration curve. The spiking of the
samples is preformed with the oxides and nitrates
of the analytes of interest in varying ratios as is
seen in the table below:
Table 1 Site 1 Calibration Samples
#
1
2
3
4
5
6
7
8
9
10
As
<3.0
<3.0
<3.0
<3.0
10000
6000
5100
2000
1000
5000
Cr
12
136
110
90
6012
4012
1112
10012
2012
5012
Pb
201
2590
2110
1630
4590
12590
8590
3590
6590
7590
Zn
200
6870
5540
4200
7870
9070
16870
10870
12870
11870
Fe
12200
47800
40680
33560
12200
12200
12200
12200
12200
12200
all concentrations in mg/Kg
The use of the oxides and nitrates is due mainly
to their availability and ease of handling. The
above table is a partial list of the values used in
the calibration for the screening of a superfund
site in region 10 (see results section). The prepa-
ration of the calibration samples normally requires
three to four labor hours. Twenty gram samples
of the dried and sieved clean soil are measured
out, one for each element to be analyzed. To
these samples is added a nitrate or an oxide of
the analyte of concern in a proportion to give a
sample concentration of ten weight percent.
These samples are then homogenized and used
in the proper ratios with the clean soil to give the
calibration samples. Three of these samples are
analyzed using Total digestion and ICP-AES to
verify the concentrations. One of the three sam-
ples is chosen as the control sample for the ICV
and CCVs. For this site the sample chosen as the
control sample was # 10. Normally one of the
samples which has not been spiked is used as
the control sample but since there was little or no
arsenic or chromium in the sample, it was decided
to use a sample which contained all of the anal-
ytes of interest.
When preparing the calibration curves, the corre-
lation coefficient shows the linearity of the calibra-
528
-------
tion. Since there is a direct correlation between
intensity and concentration in XRF spectrometry,
the quality of the data will be dependant on the
linear correlation. The acceptance limit for the
linear correlation used in this study was 0.990. If
the correlation falls below the limit, the intensities
for the calibration samples are recollected and if
the correlation is still low, the samples with low
values (far off the calibration line) are re-prepared
and reanalyzed.
FIELD AND LABORATORY QA/QC
In the field portion of the analysis, there are a
number of areas where the quality of the data
must be documented. As with any instrumental
technique, there are QC requirements in XRF
spectrometry including the initial and continuing
calibration verification and duplicate analysis. An
aspect of XRF spectrometry which can cause
unique problems is that particle size can affect the
results and so must be controlled. Finally, the
field analysis itself gives rise to possible sources
of error, such as how representative the sample
collected is and laboratory verification of the
results. All of these factors must have associated
QC/QA to document the quality of the results.
The normal QC which is followed with any instru-
mental technique include the initial and continuing
calibration verification (ICV and CCV respectively).
The ICV is performed prior to any sampling and
the CCVs are performed after every ten sampling
sites. If any of the verifications are out of the
control limits, the control limits being ± 20%, then
the calibration needs to be restandardized. The
X-MET 880 software has a restandardization
function built in so that all that is required is a
remeasurement of a standard sample. Since there
are no moving parts within the instrument, there is
rarely a need for a restandardization during a field
screening activity. A duplicate analysis is run to
give an idea as to the precision of the analysis. A
sample is chosen at random for the duplicate
analysis.
Particle size affects can cause discrepancies in
XRF data and so it is necessary to minimize these
affects. One way of proceeding to this end is to
match the field sample particles size to that used
in the calibration samples as closely as possible.
By avoiding a difference in particle size, the affect
of particle size should be a minimum. To accom-
plish this, all samples are sieved with a standard
sieve to less than 9mm. This size of sieve was
used so that all particles that can be considered
soil are included in the analysis. To get an even
closer match in particle size would require some
sort of particle size reduction which would greatly
increase the time required for the field screening
procedure.
The problem of obtaining representative samples
will be of concern when the site is large and the
sampling areas are spaced some distance apart.
There is a need to make sure that the readings
obtained are a reasonable reflection of the con-
tamination at the sampling site. To obtain a
statistically representative sample, seven unique
portions are obtained from the site and then
homogenized using the quartering technique.
The homogenized sample is then sieved (see
above) and analyzed with the XRF spectrometer.
The final QC requirement for the field analysis is
the collection of samples to be used for laboratory
verification. The samples are to be collected from
the sieved material at a frequency of approximate-
ly 10% of the total samples. This gives a range of
sample concentrations to help in the interpretation
of the field data. These samples will then be
analyzed by the USEPA approved method at a
laboratory to determine the accepted concen-
trations of contaminants at the site.
RESULTS
This section will give the result obtained at two
hazardous waste sites using the X-MET 880 to
analyze for a variety of elements. Both of the sites
were analyzed using a site specific calibration. In
both cases, the calibration samples were prepared
using two samples from the site, one contaminat-
ed and the other uncontaminated, with spiking of
the soils.
529
-------
Site 1 is a \unkyard which at one time contained
transformers and lead-acid batteries in addition to
other types of scrap metal. This is an eleven acre
site with many type of soil including fill brought in
from other places. The element of most concern
at this site was lead. The three figures below
(Figures 1, 2, and 3) show the correlation between
the laboratory results and the field XRF spectros-
copy results. In the preparation of the calibration
curve, the analysis on the preliminary samples
was performed using SW846 method 3051 (6) for
the digestion and ICP-AES for the analysis (see
Table 1). The laboratory analyses on the verifica-
tion samples used the same procedures.
Figure 1 Site 1 correlation for lead results
Figure 2 Site 1 correlation for zinc results.
Zinc
^
a.
a.
<2 <"
3 "g
W
"- 3
sj:
o
.B
Jj
10
9
8
7
6
5
4
3
2
1
0
.
'm
m
" •"
5 10
Thousands
XRF results (ppm)
F'gure 3 Site 1 correlation for iron results.
35
30
Lead
E
c.
3 "§20
05
u
O
10
5
0
E
c.
70
60
50
Iron
"3 c 40
5 10 15 20 25 30 35
Thousands
XRF results (ppm)
X3
3
20
10
0
0 10 20 30 40 50 60 70 80 90
Thousands
XRF results (ppm)
530
-------
Table 2 Duplicate XRF Analysis for Lead
at Site 1
Sample #
1
2
3
4
5
6
7
8
9
10
Analysis #1
8600
28630
12600
12580
16480
5430
14110
8540
7910
10270
Analysis #2
7270
9370
6970
15200
9500
5870
14370
7250
7290
13100
From the Figures and Tables above, it is apparent
that the analysis simulated the laboratory data
fairly well. Other elements were analyzed for (As,
Cr, and Cu) but they were all present below the
detection limit of the calibration. The discrepanc-
ies in the duplicate analysis can be explained as
a problem in the variability of the contamination at
the points and the difficulty in obtaining a large
amount during the duplicate analysis. The large
variability may be due to one small nugget in the
first sample which contained a large amount of
lead.
There were no problems encountered in the
continuing calibration verification. All of the
values for the control sample were within 10% of
that found in the laboratory analysis so restand-
ardiation was not required during the analysis.
Site 2 is a former Oil filtering operation for the
reclamation of used oil. The filtercake material
was buried and the covered over with gravel to
make a parking lot. The samples were obtained
using a drilling rig. This was a very small site with
the sampling occurring at 12 boring holes. The
main element of concern was lead. Table 2
shows the comparison of the field readings and
the laboratory values for the two samples used in
the verification. The ultimate purpose of the XRF
screening was to determine, on the site, the hot
spot for lead contamination. In the preparation of
the calibration curve, the analysis on the prelimi-
nary samples was performed using SW846 meth-
od 3051 for the digestion and ICP-AES for the
analysis. The laboratory analyses on the verifica-
tion samples used the same procedure.
Table 3 Results from Site 2
Hot Spot
Representative
Pb
As
Cr
Lab
(ppm)
15384
18.11
41.57
XRF
(ppm)
15500
<180
<100
Lab
(ppm)
4608
24.99
90.39
XRF
(ppm)
5050
<180
110
The results show that the X-MET 880 XRF spec-
trometer gave results that were close to that found
during the laboratory verification.
No problems encountered in the continuing
calibration verification. All of the values for the
control sample were within 10% of that found in
the laboratory analysis so restandardiation was not
required during the analysis.
SUMMARY
This study has shown the quality objectives for
utilizing XRF as a screening tool for metals at
hazardous waste sites. The results from the field
screening appear to emulate the data obtained
from the laboratory verification. A major factor in
the quality of the results would appear to be the
site specific calibration. The use of the site
specific calibration appears to give good quality
data without adding a great amount of time to the
pre-screening process. The data presented is the
product of two sites which had very different soil
types which could create difficulties when using a
531
-------
generic calibration.
The other methods of quality control are also
responsible for the quality of the data received
from the screening process. These methods
include the calibration verification, duplicate
analysis, assurance of the representativeness of
the sample, particle size, and laboratory verifica-
tion. These steps are already accepted methods
in the collection and analysis of any environmental
samples.
When the protocols listed above are followed, the
data obtained from the screening of hazardous
waste sites for inorganic contaminants by x-ray
fluorescence spectroscopy correlates well with
confirmatory results and require minimal reanaly-
sis in the field. The use of XRF spectroscopy as
a screening tool will meet the criteria establish for
these tools, those being speed of analysis, accu-
racy of the method, cost effectiveness and quality
of the data.
REFERENCES
(1) Chudyk, W.; "Field Screening of Hazardous
Wastes"; Environ. Sci. Technol., Vol 23, No.
5, 1989, pg. 504
(2) Chappell, R.W.; Davis, A.O.; Olsen, R.L;-
"Portable X-Ray Fluorescence as a Screen-
ing Tool for Analysis of Heavy Metals in
Soils and Mine Wastes", Proceedings of the
Conference on Management of Uncontrolled
Hazardous Waste Sites, U.S. Environmental
Protection Agency, U.S. Government Print-
ing Office, Washington D.C., 1987, EPA-
540/G-87/004, pg. 115
(3) USEPA, "Data Quality Objectives for Reme-
dial Response Activities: Development
Process" U.S. Environmental Protection
Agency, U.S. Government Printing Office,
Washington D.C., 1987, EPA-540/G-87/003,
P9-4
(4) Grant, C.L;Pelton, P.A.; "Influence of Sam-
pling on the Quality of Analysis with Empha-
sis on Powders" Advances in X-ray Analysis,
Vol. 17, 1974, pg.44
(5) Piorek, S.;Rhodes, J.R.; "A New Calibration
Technique for X-ray Analyzers Used in
Hazardous Waste Screening" Proceedings
of the 5th National RCRA/Superfund Confer-
ence, U.S. Environmental Protection Agen-
cy, Environmental Monitoring and Support
Laboratory: Las Vegas, NV, 1986, pg. 428
(6) USEPA, "Test Methods for Evaluating Solid
Waste" Fourth Edition, U.S. Environmental
Protection Agency, U.S. Government Print-
ing Office, Washington D.C., 1988, SW846,
Vol.16, pg. 3051-1
ACKNOWLEDGEMENTS
The authors would like to thank the U.S. Environ-
mental Protection Agency for funding this work.
We would also like to express our thanks to Mr.
Roy R. Jones for his many useful discussions and
assistance throughout this project.
532
-------
DISCUSSION
RUSS SLOBODA: My question involves the site specific calibration. Most soil
studies usually involve a variety of matrices at a specific site and it would seem
to be rather naive to say that one calibration is specific to the whole site. What
type of protocol do you have to help you decide? Is it a geologist helping you to
decide if these are different mineral matrices? Or do you test in your base lab
every single thing that might be a different matrix before you decide on how may
site-specific calibrations are necessary for the types of matrices you're
encountering.
CLARK CARLSON: What we like to do, of course, is to have enough samples
from the site. We require at least two. But depending on the size of the sites
(because most of the sites that I 've done haven't been that large), the assumption
that just a few samples will get us a fairly good correlation as to the matrix over
the whole site, I feel, is a good one. But when you're talking about very large sites,
you may run into some problems with the matrix, and you may have to do more
than one site-specific calibration.
JOHN BARICH: When specifying a job, what is your rule of thumb as to what
percentage of your budgeted dollars and what percentage of your performance
period should be devoted to the QA program?
CLARK CARLSON: As far as the time allotted, it usually takes on the average,
about four or five hours to make our calibration samples. And depending on the
size of the site, of course, that's going to affect the percentage of time that you're
going to use for that particular portion of the QA/QC. But other than that, since
we're doing the duplicate samples and the sample verification, the rule of thumb
that I've been using is roughly between 20% and 25% of the time.
533
-------
A Study of the Calibration of a Portable Energy Dispersive
X-ray Fluorescence Spectrometer
C.A. Ramsey, D.J. Smith, and E.L. BourO *
United States Environmental Protection Agency
National Enforcement Investigations Center
Box 25227, Denver, Colorado 80225
ABSTRACT
Generation of reliable concentration
information from portable energy
dispersive X-ray fluorescence
necessitates the development and use of
appropriate calibration methods. The
present study considers some of the
difficulties encountered with the use of
empirical calibration, which is obtained
from the measurement of standards with a
similar matrix to that of the samples
being investigated. The effectiveness
of two sets of empirical calibration
standards have been investigated, namely
a set of site specific samples analyzed
using a referee method, and a set of
artificially prepared calibration
samples produced by spiking an
uncontaminated soil matrix. It was
found that sample matrix variation
unaccounted for in the calibration leads
to uncontrolled bias in the analytical
results.
INTRODUCTION
Of major concern in the characterization
of many hazardous waste sites is the
rapid and accurate determination of
metal concentrations in solid and liquid
samples. Generally, these
determinations are carried out in an
off-site laboratory using techniques
such as atomic absorption (AA)
spectrophotometry or inductively coupled
plasma (ICP) atomic emission
spectrometry. These methods, while
relatively accurate, are time consuming
and expensive; they also cannot
practically provide on-site, real-time
results. Thus, there is a need for a
supplementary technique that can provide
adequate information about a site in a
timely and less expensive manner.
Portable energy dispersive X-ray
fluorescence (FPXRF) spectrometry is a
promising method to meet this need.
Instrumentation is commercially
available which utilizes radioisotope
sources and gas-filled proportional
counters.
Problems exist, however, which are
inherent to X-ray fluorescence (XRF)
methods that must be addressed to
provide useful concentration data for
unknown samples. Some of these
problems, such as particle size and
mineralogical effects, are common to all
types of XRF and can be minimized by
preparation of all samples and standards
to a uniform state. Other well-known
problems consist of interelement
effects, which include spectral overlap,
primary and secondary absorption, and
enhancement. Absorption and enhancement
effects are commonly addressed in
laboratory XRF using computational
procedures such as the "Fundamental
Parameters" approach (Jenkins, 1974).
With contemporary laboratory wavelength-
and energy-dispersive instrumentation,
spectral interferences present little
difficulty in all but a few cases.
Thus, in laboratory implementations of
XRF, results comparable to AA and ICP
are routinely achievable.
To date, the aforementioned problems
have not been adequately addressed in
applications of FPXRF to environmental
monitoring; of major concern is the
relatively poor spectral resolution of
the proportional counter detectors,
which require larger overlap correction
factors for adjacent elements than solid
535
-------
state dectectors. In addition,
approaches such as "Fundamental
Parameters" are not practical to
implement in FPXRF because of the
presence of many unresolved spectral
interferences, computation requirements,
and the inability to measure all
components in the sample. Thus,
empirical calibration approaches have
been used in FPXRF; these relate the
measured spectra of a "training set"
(calibration set) to the concentrations
of elements present in the training set
samples, typically using an approach
such as multiple linear regression
(MLR). Two types of training sets are
common, namely real samples analyzed by
referee methods (site-specific models),
and synthetically prepared standards
(generic models). For both types of
training sets, the analyst attempts to
match the matrix of the training set to
the anticipated unknown samples.
Obviously, selection of a proper
training set matrix and the
determination of its applicability to
unknown samples are critical problems.
This study involved the application of
empirical calibration-based FPXRF to an
environmental monitoring situation,
namely, the determination of the
elements chromium, zinc, and lead in a
soil-like waste material. Samples from
a site contaminated with heavy metals
were collected and analyzed by ICP.
These samples were used to evaluate the
performance of site-specific and generic
training sets.
XRF THEORY
The first step of any X-ray analysis is
the removal of inner shell electrons
from the atom. A vacancy is then
created which is immediately filled by
an electron from a higher energy shell.
The resulting free energy is emitted as
radiation that is characteristic of the
excited element. The second stage is
the selection of an emission line from
the element of interest by means of a
wavelength or energy dispersive
spectrometer. Next is the detection and
integration of the characteristic
emission line of interest, and finally,
the conversion of intensity to
concentration by the use of some
calibration procedure.
Until the late 1960s nearly all X-ray
spectrometers were the wavelength
dispersive type in which wavelengths are
separated by Bragg diffraction from a
single crystal. Although wavelength
dispersive X-ray fluorescence (WDXRF)
spectrometers have high spectral
resolution, they are bulky, expensive,
and require a high power x-ray tube as
an excitation source.
More suited for field work are energy
dispersive X-ray fluorescence (EDXRF)
spectrometers. EDXRF spectrometers,
being more efficient than wavelength
dispersive spectrometers, can utilize
small radioactive sources for excitation
instead of large X-ray tubes. In
addition, the separation of emission
lines does not require the use of a
large crystal chamber and goniometer as
does a WDXRF spectrometer. There are
several types of detectors that are
employed in FPXRF spectrometers:
scintillation counters, solid state
detectors, and gas-filled proportional
counters. The scintillation counter has
very poor resolution and requires the
use of balanced filters to discriminate
between lines. The need for filters
increases the mechanical complexity and
limits the flexibility of an instrument.
Solid state detectors use a crystal of
lithium drifted silicon. Silicon
detectors have very good resolution
(-0.16 KeV full width at half maximum
for the 5.89 KeV Mn-Ka) but require the
use of liquid nitrogen or thermoelectric
cooling to minimize electronic noise.
Gas filled proportional counters have an
intermediate resolution (-0.77 KeV for
Mn-Kg) between the scintillation counter
and the solid state detector but do not
require cooling. Thus, the combination
of radioisotopic source for excitation
and gas-filled proportional counters for
detection are more suited for field
portable instruments.
EXPERIMENT
Samples and Standards. Contaminated
soil samples were collected from a site
bearing potentially hazardous waste
material. Initially, 25 samples were
taken from soil that was contaminated
with heavy metals from a metal recycling
plant. Fifteen of these 25 samples were
used as a site-specific training
(calibration) set, referred to hereafter
as TRAIN1. The TRAIN1 samples were
selected to provide the maximum range of
concentrations for both the analytes and
major elements. The remaining 10
536
-------
samples were used as an "unknown"
testset; these are referred to as TEST1.
An additional testset (TEST2) of eleven
samples, taken from a different area on
the same site, consisted of material
from the same process produced at an
earlier time. ICP analysis showed that
the levels of iron, calcium, and silicon
were higher in the TEST2 samples than in
the soil from the TRAIN1 group (refer to
Table 1).
The generic training set (TRAIN2)
consisted of synthetically prepared
standards obtained commercially. These
standards had been produced by spiking
an uncontaminated sandy soil matrix with
chromium, copper, zinc, arsenic,
cadmium, and lead.
Sample Preparation. Approximately one
kilogram of each soil sample was air-
dried to constant weight and sieved
through a two millimeter nylon sieve. A
50 gram subsample was selected by
randomly taking approximately 50 one
gram aliquots from the primary sample.
This subsample was then ground in a
tungsten carbide rotary ring mill to
minus 200 mesh particle size. From the
ground subsample, an analytical sample
was selected for FPXRF analysis, which
consisted of approximately 30-40
increments of about 0.25 grams each,
which were then placed in a
polypropylene cup for the analysis.
Analysis of replicates by ICP verified
the homogeneity of the ground samples
(less than five percent relative
standard deviation between 0.25 gram
aliquots). These measures minimized
subsampling and particle size effects.
Samples were prepared for ICP by fusing
0.25 grams of ground material with 2.0
grams potassium hydroxide (KOH),
followed by dissolution of the melt in
hydrochloric and nitric acid (HCl-HNOo).
Fusion was selected over commonly used
acid digestions such as EPA Method 3050
because of greater accuracy for critical
elements such as chromium and iron;
procedures such as Method 3050 tend to
yield low results for many elements
because of poor attack of silica-based
minerals. This fusion procedure
produced reliable analytical results for
thirty elements of interest, as
evidenced by acceptable results for
spiked samples, replicate samples, and
reference materials. Accuracy of these
results was further verified by
determination of the analytes using a
laboratory-based EDXRF.
FPXRF Measurements. An Outokumpu X-HET
880 FPXRF, containing a 30 mCi 244Cm
excitation source and an argon
proportional counter, was used in these
studies. Data was acquired using a 200
second measurement time, and a
consistent sample presentation geometry.
Chromium, zinc, lead, and backscatter
(BS), as well as several potential
interfering elements were measured in
the training set. The Ka emission lines
for chromium and zinc were selected and
the La line for lead was used in Trainl
because no arsenic was present in the
samples. The 1^ line for lead was used
in Train2 due to the presence of arsenic
in the standards. Spectral
interferences were treated using a
Gaussian elimination algorithm
(subtracting the portion of signal due
to the interferant after measuring
interferant intensity and using a pre-
established correction coefficient)
provided for in the X-MET software
(Outokumpu Oy). Stepwise multiple
linear regression was used to develop
models accounting for absorption and
enhancement effects; interference-
corrected spectral intensities were used
as dependent variables; concentrations
of the elements were used as independent
variables. The significances of
independent variable effects were
determined using t-tests. Some
standards were omitted from the model if
significant improvement in r was
achieved by doing so. A summary of the
models generated is presented in Table
2.
RESULTS AND DISCUSSION
Examination of Table 3 reveals that bias
for all elements generally increases in
the order (TRAIN1 ; TEST1)
< (TRAIN1 ; TEST2) « (TRAIN2 ; TEST1 or
TEST2). Chromium could not be
effectively quantitated in any samples
by the TRAIN2 (generic) model. Indeed,
chromium was not detected for most
samples even though 400-600 mg/kg Cr was
actually present. The bias problem was
not due to lack of measurement precision
(see Table 4). A possible explanation
is as follows: The Cr-Ka, being of low
energy, is particularly susceptible to
sample matrix effects, such as the
presence of iron at varying levels. The
analogous effect is not as pronounced
537
-------
for the higher energy Zn-Ka and Pb-La.
Additionally, the Cr-Ka line is subject
to spectral interference from the Fe-Ka
line, which is problematic when the
concentration of the interferant (Fe) is
high compared to the analyte (Cr).
A difficulty with the TRAIN2 (generic)
calibration model is the lack of
definition of the sample matrix effect
(differing absorption/enhancement
correction coefficients). Elemental
analysis of the TRAIN2 standards by lab-
based EDXRF revealed very little range
in the concentrations of major elements
such as iron, silicon, and calcium.
Empirical training models containing no
variance in influential parameters (e.g.
matrix, concentration of interferants)
cannot be expected to produce models
which are robust with respect to these
variations. For TRAIN2, t-values for
iron and backscatter were statistically
insignificant in this set of standards,
indicating that no variances exist for
major elements in the TRAIN2 materials.
Differences in matrices are thus a main
pitfall in FPXRF. It is unrealistic to
believe that, for field applications,
the matrix will be identical for all
samples encountered. It is also
difficult to determine, in real time,
the applicability of a specific training
set to a particular sample. A specific
training set may be inadequate due to
the presence of unanticipated spectral
interferants, or large matrix
differences, or both problems. In this
study, changes of less than a factor of
two in the concentrations of iron,
calcium, and silicon between TEST1 and
TEST2 strongly influence the biases of
the resulting data.
FPXRF calibration models are frequently
evaluated based upon the value of the
regression coefficient, r . It is
essential to note that good correlation
alone does not ensure accurate results.
The regression coefficients for all
analytes using TRAIN1 and TRAIN2 were
close to unity, yet many predictions
were highly biased. While a model with
a low r2 value lacks any predictive
capability, a high r alone does not
guarantee its predictive capability for
a specific test sample.
CONCLUSIONS
FPXRF is an analytical technique which
is not highly robust with respect to
spectral interferences and sample matrix
effects. Even with site specific
empirical calibration, the applicability
of a particular model to a specific
sample cannot be ensured. In some cases
several models might have to be employed
to cover the entire range of analytes,
inteferents, and matrices. Quantitative
analyses of varying, unknown hazardous
waste streams present a challenge of the
highest order to FPXRF application;
substantial possibility exists for false
positive and false negative readings, as
well as highly biased quantitation. Low
atomic number elements and lower
concentrations appear to be more
susceptible to quantitation problems.
However, FPXRF is presently very useful
as a field analytical device for
problems such as contamination
delimination and segregation of waste
streams. Qualitative, semi-
quantitative, and quantitative
analytical results are all potentially
achievable on a case-by-case basis.
Assuming present instrumental hardware,
additional research in FPXRF should be
directed towards development of more
robust calibration methods, for example
chemometric calibration.
REFERENCES
Jenkins, Ron, An Introduction to X-ray
Spectrometry, Heyden, London, 1974.
Outokumpu Oy, - "Operation Instructions
X-MET 880 Analyzer Ver. 1", Outokumpu
Oy, Espoo, Finland.
538
-------
Table 1
Matrix and Analyte Concentration Ranges (rag/Kg)
Element TRAIN1 TEST2
Fe 83900-127000 87500-196000
Ca 17900-30700 27600-49100
Si 64400-146000 121000-220000
Cr 497-766 307-567
Pb 4740-8230 3210-8560
Zn 9280-20100 5840-26800
Table 2
Summary of Calibration Data
TRAIN1 (Site-Specific) Calibration Model
Interfering Points Regression
Analytes Elements Omitted Coefficient
Cr Cr, Ti, Fe, BS 2 .990
Pb Pb, Fe, BS 3 .990
Zn Zn, Fe, BS 1 .991
TRAIN2 (Generic) Calibration Model
Interfering Points Correlation
Analytes Elements Omitted Coefficient
Cr Cr, Mn, Fe, BS 0 .998
Pb Pb, BS 0 .996
Zn Zn, BS, Pb 0 .998
539
-------
Table 3
Bias Data for Samples
Absolute Percent Relative Error
TRAIN1 (Site-Specific) Calibration Model
Chromium Lead zinc
TRAIN1 7.0 9.3 5.6
TEST1 8.9 5.3 3.5
TEST2 100.0 23.3 13.1
TRAIN2 (Generic) Calibration Model
Chromium Lead zinc
TRAIN2 12.0 15.1 8.9
TEST1 97.9 33.8 39.8
TEST2 86.0 38.4 49.6
NOTES: The training set rows refer to the mean magnitude of
the relative errors produced by re-measurement of the
training set using the developed model. The testset rows
refer to the mean magnitude of the relative errors produced
by measurement of the samples using the developed model.
Table 4
Measurement Precision
Percent Relative standard Deviation
TRAIN1 (Site-specific) Calibration Model
Samples Chromium Lead zinc
TRAIN1 4.2 2.1 1.3
TEST1 8.4 1.7 0.7
TEST2 ND 1.9 1.9
TRAIN2 (Generic) Calibration Model
Samples Chromium Lead zinc
TRAIN2 1.3 4.7 0.5
TEST1 ND 2.7 1.8
TEST2 ND 2.9 0.8
NOTE: "ND" means only values of zero were measured;
precision was therefore not determined.
DISCUSSION
STEVE KNOLLMEYER: I have two questions. One, does the X-MET 880 DONALD SMITH: As far as I know there's no internal calibration available
haveaninternalmeasurementcapability?Thatis,needingnoexternalcalibration. with that. So, no we didn't.
And two, if so, do you compare errors using that internal calibration method
versus your matrix specific calibration standards?
540
-------
Use of Long-path FTIR Spectrometry in Conjunction with
Scintillometry to Measure Gas Fluxes.
Douglas I. Moore,
Clifford N. Dahm,
James R. Gosz,
Biology Department, University of New Mexico,
Albuquerque, NM, 87131
and
Reginald J. Hill,
NOAA/ERLAVave Propagation Laboratory
325 Broadway, Boulder, CO, 80303
ABSTRACT
Fourier Transform Infrared (FTIR) spectrometry is
rapidly becoming a technique of choice for analyzing
volatile hazardous waste emissions (1). We have
developed a field portable system that is capable of
measuring gas concentrations at up to a 1 Km pathlength.
The advantage of such a system is that it can analyze
samples virtually in real time for myriad compounds
simultaneously without introduction of any artifacts from
sample collection. Because detection sensitivity increases
with path length, analysis of compounds can often be made
down to ppb levels. While we have concentrated on
measuring gas emissions from biologic sources, we are
also capable of monitoring hazardous gas emissions that
have characteristic infrared absorbance peaks in regions
of the IR spectrum that are not dominated by water or CC>2.
Numerous volatile organic compounds such as chlorinated
hydrocarbons, aromatic hydrocarbons, alcohols, ketones,
esters, ethers and aldehydes fall into this category.
While gas concentrations are of interest, emission rates
are needed to accurately evaluate waste sites. Obtaining
such flux rates has become the focal point of our research.
Recently, the Wave Propagation Lab (WPL) at NOAA
(Boulder CO) has demonstrated that an optical-
scintillation instrument can measure path-averaged
momentum and heat fluxes. Development of
scintillometers by WPL, which is currently in progress,
will allow long-path measurement of water flux as well.
Combining these long-path flux measurements with
measurements of gradients of gas concentrations using the
FTIR has the potential to provide an estimate of flux rates
for numerous gases simultaneously. This technique will
then have application in natural, agricultural and human
impacted areas such as landfills and hazardous waste sites.
INTRODUCTION
Fourier Transform Infrared (FTIR) spectrometry is
rapidly becoming a technique of choice for analyzing
volatile hazardous waste emissions. A particular advantage
of FTIR spectrometry is that it can be configured for long-
path analysis. Large areas from which potential emissions
are occurring can be analyzed at one time. We have
developed a field portable system that is capable of
measuring a number of infrared absorbing trace gases over
long paths of up to a kilometer (Fig. 1). Many of these
gases have either natural and/or anthropogenic sources
and/or sinks. Increased path-length provides several
advantages such as increased detection level (ppb levels)
and long-path averaging of the high variability in gas
emissions that may often be the case on a smaller scale
(2).
Figure 1. Schematic of an infrared radiation source and
Fourier transform infrared (FTIR) spectrometer equipped
with telescopes to allow long-path analyses. The flat
mirror allows rapid manipulation of optical pathlength and
analyses over undisturbed terrain.
541
-------
The detection of the presence or absence of a particular IR
absorbing contaminant is diagnostic of what volatile
constituents are being released from a site. An even more
valuable piece of information would be an estimate of the
total flux of that contaminant through the site into the
atmosphere over time. Such an estimate requires
knowledge of the micrometeorological conditions of the site
during the period when long-path FTIR spectrometry
measurements are being made. Ideally, the micro-
meteorological measurements would also be path-averaged
over a comparable path being sampled by the FTIR
spectrometer. Scintillometry techniques, presently under
development by atmospheric scientists at the NOAA wave
propagation laboratory in Boulder, Colorado, provide a
potential method to link path averaged FTIR spectroscopic
profiling and micrometeorological characterization of the
near surface atmosphere. Such a combined data base may
make estimates of large scale contaminant fluxes possible.
The purpose of this article is to describe our long-path
FTIR spectrometer instrument, to present some analyses of
various IR absorbing trace gases measured using a long-
path configuration, and to discuss means of combining
micrometeorological data with long-path trace gas
concentrations to obtain estimates of fluxes. In particular,
combining of long-path scintillometry with long-path
FTIR gas concentration data shows the promise of providing
a better tool for quantifying gas fluxes from both natural
and polluted sources.
MATERIALS AND METHODS
FTIR Spectrometer
Our FTIR instrument is a Nicolet 740 optical bench
interfaced with a Nicolet 660 work station having a storage
module capable of storing 344 Mbytes of data. The 740
unit is capable of 0.3 cm-1 wavenumber resolution (3,4).
The optical bench is mounted on a wheeled cart in
conjunction with a 60 cm diameter Cassegrain telescope.
The infrared source (1000-W halogen quartz lamp or
globar) is mounted at the focal point of a second 60 cm
diameter Cassegrain telescope. The infrared beam is
transmitted to a 60 cm square flat mirror and returned to
the receiving telescope which focuses it on an adjustable
aperture (Fig. 1). The beam is then recollimated and
transmitted into the optical bench and subsequently
through the interferometer to the detector. The 740 bench
is equipped with an Hg-Cd-Te (MCT-A) detector. All
sample collection and processing is handled using Nicolet
software. Each collected sample can be either a single scan
or a composite of a series of scans. Sampling rates are a
function of scans per sample and the pathlength of the
moveable mirror. Most samples discussed in this paper
were collected at 0.5 cm-1 wavenumber resolution. Each
scan takes about a second at this resolution. Generally 16
scans per sample were collected although in some cases a
1-minute sampling rate was desired which resulted in 60
scans per sample. Each sample yields an interferogram
(Fig. 2) which must then be processed to give an
absorbance spectrum (Fig. 3) from which gas
concentrations can be calculated. This quantification is
performed using a multivariate least squares fit (LSF)
program developed by Haaland and Easterling (5). This
technique allows quantifying of multiple components whose
lines overlap in a given spectral region with better
precision and accuracy than can be obtained from a single
peak analysis.
7' 50 1250 1350 IHSO
DflTfl POINTS
Figure 2. Example interferogram from the FTIR. The
energy source was a quartz-halogen lamp, pathlength 407
m, atmospheric pressure 724 torr and instrument
resolution 0.3 crrH.
WflVENUMBER
Figure 3. The raw data of the interferogram is processed
to give an absorbance spectrum that allows calculation of
gas concentrations. This interferogram was collected at
Maricopa, Arizona over a cotton field, 10:26 hrs, 12 June
1988, pathlength 407 m.
542
-------
Optical Scintillometer
Scintillometers are devices that sense the scintillation
(intensity variations) in propagating electromagnetic
(EM) waves. The refractive index structure parameter
Cn2 characterizes the magnitude of this scintillation. EM
scintillation results from turbulent fluctuations in the
atmospheric refractive index which in turn result from
fluctuations in atmospheric temperature and humidity
fields. The scintillometer currently being used is an
optical scintillation inner-scale meter. In the experiment
reported here, it is deployed on a horizontal path of 150 m
at a height of 4 m. This instrument measures the variance
of the log intensity of diverging Laser light detected
through a 1mm-diameter hole as well as the variance of
the logarithm of aperture-averaged intensity from a
4.4-cm diameter, phase incoherent uniformly-
illuminated source. The ratio of these variances gives the
inner scale of turbulence I0- The refractive index
structure parameter Cn2 is also determined from this
instrument's data. An approximate correction for
saturation of scintillation of the Laser variances is
computed to extend the range of validity of the instrument.
The heat and momentum fluxes are deduced from Cn2 and I0
using Monin-Obukhov similarity relationships. This is
the "inertial dissipation" method of determining these
fluxes. In this configuration the instrument can be
considered a fluxes scintillometer.
For comparison purposes, a three-axis sonic anemometer
having platinum resistance-wire thermometer near its
center is also deployed at a height of 4 m on a tower near
the optical scintillation instrument. The sonic anemometer
measures all three fluctuating components of velocity at a
25 Hz data rate. The correlation of the vertical component
of velocity with the streamwise horizontal component gives
the momentum flux (divided by air density). The friction
velocity U- is the square root of the negative coefficient of
this correlation. The correlation of the vertical component
of velocity with the temperature from the resistance-wire
thermometer gives the temperature flux (heat flux is
temperature flux multiplied by the air's heat capacity.)
RESULTS AND DISCUSSION
FTIR
Much of our initial work has involved testing the utility of
the long-path FTIR for monitoring gases from natural as
well as anthropogenic origins. Gases that we most
routinely analyze include H20, COa, CH4l N20, and CO but
many other gases can also be quantified simultaneously or
at some later time. This points up one of the primary
advantages of FTIR. Spectra collected can be reprocessed at
a future time to look for oases that mav not have been of
primary interest at the time that the original analysis was
carried out. Below are examples of some uses of our
long-path FTIR.
Gas Emissions
The FTIR was set up to measure a 100 m path over a small
shallow lake on the Isleta Indian Reservation near
Albuquerque, N.M on June 26, 1989. A series of samples
was taken to establish ambient gas concentrations. The
bottom sediments were then disturbed by a person walking
around to force degassing of the sediments. A second series
of samples was collected during this period (Fig. 4).
Emission of CH4 is readily apparent. These emissions were
in fact point source emissions and the measured
concentrations denote a mean concentration for the entire
100 m path, most of which was not undergoing degassing.
This artificially induced gas emission points up the ability
of the FTIR to quantify such emissions but the experiment
was designed to test our ability to measure changes in path-
averaged atmospheric concentrations and was not meant to
estimate gas fluxes.
i,"d
•S 2.7-
e
1 "^
i"1
« 2.1-
5
1.9-
1.7-
U
Undisturbed
Disturbed
30
0 10 20
Time (min.)
Figure 4. Field demonstration of the ability of long-path
FTIR to detect CH4 emissions from a shallow lake in New
Mexico. Measurements were made over the lake on June
26, 1989 before and after bottom sediments were
disturbed along a portion of the pathlength.
Gradient Profiling
A primary focus of our study is to develop the capability to
measure gas fluxes over a long path which should average
out the large spatial variability encountered using small
scale techniques such as chambers. Micrometeorological
techniques such as Eddy correlation and Bowen Ratio
techniques have been shown to be capable of measuring heat
and water fluxes under certain ideal conditions. We have
explored the possibility of using the diffusivity constants
obtained from Bowen stations and eddy correlation stations
to plug into a flux gradient equation:
543
-------
F= k - dc/dz
d)
where F is the mixing ratio flux, dc is the difference in a
gas concentration (in ppm) measured at two heights and dz
is the difference in those two heights. We experimented
with this technique over a tall grass prairie site near the
Konza Prairie Long Term Ecological Research (LTER) site
near Manhattan, KA in July of 1989. The FTIR source and
spectrometer were mounted on the lift gate of a truck so
that the instrument could be raised from near-ground
level at 1 m to a 2 m level. Samples were collected at one
height for 10 minutes then moved to the other height for a
10 minute collection period. This procedure was repeated
over a 3 day period (July 25, 26, 27, 1989 - Julian days
206, 207. 208).
Our initial results were encouraging as a significant
gradient could be detected for most gases. In some cases,
such as for water vapor, this gradient agreed reasonably
well with that measured by a Bowen Ratio station in the
same vicinity (Fig. 5). During some of our measurements
the gradient was opposite to that expected. Presumably,
this was the result of the rapid changes in surface layer
conditions that were missed due to the time required to
move the instrument from one height to another and
realign it.
- 200
^ 10°
3 0
2, -100
1 -200
fc -300
Water Flux - Konza Prairie - July 27,1989
-400
Bowen
FTIR
8
14
16
10 12
Time (Hours)
Figure 5. Water fluxes as quantified by long-path FTIR
and a Bowen Ratio station at the Konza prairie research
area on July 27, 1989. The values are based on the
concentration gradient at 2 m and 1 m heights above the
ground. Negative values indicate that the direction of the
flux is away from the soil and vegetation.
Figure 6 shows the concentration gradients that were
monitored for the 3 days of sampling. There are several
noteworthy items: 1. Water gradients were always
negative as should be the case under stable surface layer
conditions (concentration at 2 m < concentration at 1 m).
2. Daytime COa gradients always showed a positive gradient
presumably reflecting the active COj absorption by plants.
3. Of even greater significance was the positive gradient
for CH4 coupled with the significant daily decreases In CH4
concentrations. A rain event on July 23 (Julian day 204)
was followed by a drying period for the remainder of the
sampling days. The sequence is likely due to a high rate of
oxidation of CH4 by methane oxidizers in the local soils
coupled with a decrease in atmospheric concentrations as
the general area dried. The regional source of the CH4 Is
not known. 4. CO concentrations also demonstrated a
constant drop through the sampling period. A shift in the
wind away from the direction of an interstate that ran just
north of the site to a more southeasterly direction is a
likely explanation for this change as anthropogenic sources
far exceed biological sources.
6a
I6.O-
15.5-
15.O
14.5-
14.0
H2O
m 2m
01m
206 207 208
Julian Day (1989)
6b
i
§
a
6c
C02
H 2m
Dim
206 207 208
Julian Day (1989)
2m
El 1m
206 207 208
Julian Day (1989)
544
-------
6d
S
I
g
a
ion
206
107
Julian Day (1989)
Figure 6. Concentration gradients for H2O (6a), and trace
gases- CO2 (6b), CH4 (6c) and CO (6d) measured by long-
path FTIR on the Konza Prairie research area during the
period July 25-27,1989.
Air Quality Monitoring
In an effort to explore the possibility of using the FTIR to
monitor air quality in the metropolitan area of
Albuquerque, N.M., we set up the instrument to analyze
over a path of about 500 m. A portion of this path was
just over the level of automobile traffic on one of the main
traffic arteries in Albuquerque. The sampling period
extended from rush hour through the period of peak
fireplace smoke emission. A primary purpose of the
experiment was to determine if methyl chloride (an
indicator of wood smoke source) could be detected.
Concentrations of this gas were found to be too low to be
detectable even at the 500 m pathlength due to both low
concentrations and relatively poor absorption by this
molecule. However, examination of the spectra showed that
a number of gases which are not normally detectable under
clean air conditions were present in elevated levels.
Ethylene and methanol were easily detected while ammonia
and formaldehyde were also above detectable
concentrations.
These tests were carried out on two nights during the
winter of 1989-90. The first was on Dec. 8, 1989. This
had been declared a no-burn night by the City's
Environmental Health Department meaning that no wood
burning in fireplaces or stoves was permitted. Gas
sampling began at about 5:40 PM and continued until
11:00 PM. Figure 7 shows the level of gas concentrations
for N2O,CO,CO2, CH4t H2OandC2H4 Values are means of
10 - one minute samples. All of the gases, with the
exception of N2O, increased until about 7:30 PM at which
time they began a steady decline. The second sampling date
was Jan. 12, 1990. This night was not declared a no-burn
night by the city although it was expected to be marginal
with respect to weather conditions which could assist with
the dispersal of gases emitted by wood burning or
automobiles. Indeed, surface wind conditions for both
nights were quite similar. On Jan. 12, most of the data for
the period 7:40 to 9:15 were lost due to problems with the
FTIR instrument (Fig. 8). As with the Dec. 8 date, all
gases tracked each other with the exception of N2O which
showed an inverse pattern. However, unlike the December
collection, the gases showed a steady increase from 5:00
PM until about 10:00 PM when they began to decline.
While the time of peak levels was different, concentration
of peak levels for all gases were generally comparable.
Table 1 gives the maximum, minimum, and mean
concentrations for all of the gases measured for the two
sampling dates. The sampling intervals for the two dates
were not the same although total sample numbers were
similar.
2
^-
i
o
U
O
550-
450-
350-
250-
150
No-Burn
-+- N20(ppb)
*• C0(ppb/10)
•*• C02(ppm)
-*• CH4(ppb/5)
•*• C2H4(ppb*5)
11 12
5 6 7 8 910
Time (PM) - Dec. 8,1989
Figure 7. Trace gas concentrations measured by long-path
FTIR during a no-burn (no wood burning allowed) night in
Albuquerque, New Mexico. Values are concentrations from
the means of 10 - 1 minute, samples on the night of
December 8 1989. Pathlength was 500 m.
550-
NZO(ppb)
C0(ppb/10)
C02 (ppm)
CH4 (ppb/5)
C2H4 (ppb*5)
Burn
_c
*•*
2
5
o
o
U
i/i
3
O
5 6 7 8 91011
Time (PM) • Jan. 12,1990
Figure 8. Trace gas concentrations measured by long-path
FTIR during a night when wood burning was allowed (Jan.
12, 1990). Pathlength was 500 m.
545
-------
Table 1. Mean, maximum and minimum concentrations of
trace gases measued with long-path FTIR on Dec. 8, 1989
(No-Burn) and Jan. 12, 1990 (Burn) in Albuquerque,
NM.
Gas Units
N2O
CO
CO2
CH4
H2O
CH3OH
C2H4
NH3
ppb
ppb
ppm
ppb
ppm
ppb
ppb
ppb
Mean
No-Burn
328
3316
426
1659
2037
7.7
61.6
16.7
Maximum
Burn
314
3438
405
1678
2204
15.2
59.2
19.7
No-Burn
337
4980
472
1790
2181
31
120
36
Burn
324
5274
463
1946
2300
30
110
26
Minimum
No-Burn
317
1518
386
1522
1926
0
23
9
Burn
297
1763
353
1494
2127
3
30
14
While the small sample size precludes any definitive
conclusions as to the relative contribution of
wood-burning and automobiles to the elevated levels of the
sampled gases, some things are suggested by the data.
1. The timing of the peak CO level on the burn night
suggests that wood-burning contributes a considerable
quantity of CO to the city air when permitted.
2. Based on mean and maximum CO levels for the two
nights, firewood burning contributes proportionately more
CH4, CHaOH and NHa than vehicle emissions while
automobile exhaust shows a greater contribution of C02,
N20andC2H4.
3. On the burn night, the CO concentrations were
extremely erratic during the time of maximum traffic.
This indicates a nearby source due to heavy vehicle traffic.
Later in the evening, the concentrations were less erratic
suggesting a less proximate source which would likely be
the case for wood smoke.
0.30 -,
0.20
o.to
-o.oo -
-0.10
«
t
f
,6'
4
J>*
•au
Fovorobta Wind Variability
Acceptable Wind Variability
OumlianabM Wind Variability
Eilreme Wind Variability
Poor Wind Direction
Unfavorable Cloud Cove
"oja
-010 -0.00 0.10 0.20
temp flux-eddy correlation
Figure 9. Temperature flux measured by the flux
scintillometer compared with temperature flux from eddy
correlation. Units are °C m s-1.
Scintillometry
Comparison of the temperature flux and friction velocity
determined from the fluxes scintillometer with that
obtained from the point sensors are shown in figures 9 and
10. These data show that the fluxes scintillometer gives
good values even in nonideal atmospheric conditions. Even
during periods of intermittent cloud cover and unfavorable
wind conditions, when the validity of the Monin-Obukhov
similarity is limited or unknown, measurements obtained
from the fluxes scintillometer compared well with the
point sensor measurements.
0.70 -
0.60 -
0.50
0.40
0.30
0.20 -
0.10
0.00
• rovoroble Wind Variability
« Acceptable Wind VoriabiMy
• OumtionoMe Wind Variability
• extreme wind Variability
* Poor Wind Direction
Unfavorable Cloud Cover
0.50 0.60 0.70
ni | I i I i | i i i i | i i i i I i i i i r* '
0.00 O.tO 0.20 0.30 0.40
U. - wind speed
Figure 10. Friction velocity measured by the flux
scintillometer compared with friction velocity deduced
from wind speed, surface roughness, and eddy-correlation
temperature flux. Units are m s-1
Integration of FTIR and Scintillometry
Our next step in the integration of the FTIR system with
the long-path measurement of water and heat fluxes
involves two steps. The first is the development of 2
scintillometers by the WPL One uses a 10.6 p.m
wavelength Laser source; the other uses a 3 mm
wavelength Gunn diode source. The combination of thes'
two scintillometers with the fluxes scintillometer allows
long-path measurement of the fluxes of latent heal
sensible heat and momentum. Modification of the FTIi
system to allow it to measure gas concentrations at twi
heights almost simultaneously is the second proposed
modification. This will be accomplished through the use of
a periscope system that will eliminate the lengthy delays
in realignment that are necessary when the entire
instrument must be moved vertically as was the case in thi
Konza Prairie experiment.
Two related methods will be used to derive gas fluxes from
FTIR and scintillometry measurements. These are outlined
546
-------
by Andreas (6). The first uses the Obukhov length L from
measurement of vertical momentum, u- and the
temperature and water vapor fluxes. These, combined with
the gas concentrations, GI and GZ measured at the two
heights zi and zg by the FTIR, will give an estimate of the
corresponding flux scale g- from the relation
AG = G2. G,=(g./K)[ln
• ug (z2/L) - ug(Z! /L)]
(2)
as suggested by Hicks and Liss (7). The gas flux is then
given by Fg= -u-g-. The second method will use a modified
Bowen ratio method which is based on the belief that ug
should be the same for all conservative scalars. In
particular, the difference in water-vapor mixing ratio,
AQ, also satisfies equation 2
AQ = QT . Q2 =(q-/K) [In (23/2,) - uq (z2/L) -
(3)
where q- is the water vapor flux scale. Since -u-g-= Fg
is the gas flux we are seeking and -u-q- = Fq is the water
flux, and since we assume ug=uq, Equations 2 and 3 yield:
Fg = Fq (AG/AQ)
(4)
To use equation 4, we obtain AG and AQ from the FTIR and
Fq from the scintillometers.An important assumption of the
application of gradient profiling is that the various gases
for which fluxes are being derived behave similarly to heat
and water vapor in the atmosphere. The scintillometry
system measures path averaged fluxes of heat, momentum
and humidity with scintillometers measuring at 1 mm,
10|im, and 1u,m. Gaseous flux estimates are then based on
these measurements of atmospheric conditions. All gases
might not have identical flux profile relationships that lead
to equation 2; that is it may be that ug*uq (8). It will be
critical in future work to determine the flux-profile
relationships for trace gases.
CONCLUSIONS
Our research to date with long-path FTIR spectroscopy and
path-averaged scintillometry has resulted in the following
conclusions.
1) Long-path FTIR has the analytical sensitivity to
measure numerous atmospheric trace gases both
anthropogenically and naturally derived at ambient
concentrations.
2) Increased pathlength commonly adds to the sensitivity
with which we can measure atmospheric trace gases.
3} Optical scintillometry is a potential tool for measuring
path-averaged fluxes of heat, momentum and humidity in
the atmosphere over the same path in which gas
concentrations are being determined by long-path FTIR.
4} Gradient profiling using long-path, path-averaged
FTIR spectroscopy has been used to show distinct vertical
structure of H2O, CO2 and CH4 in the atmosphere over a
prairie.
5) Coupling gradient profiling of atmospheric trace gases
with the long-path FTIR and path averaged scintillometry
is a promising means to begin to estimate gas fluxes at
larger spatial scales from various landscapes.
Applications are seen for both field screening of hazardous
waste and toxic chemical emissions to the atmosphere and
for many global greenhouse gases.
ACKNOWLEDGEMENTS
This work was supported by a grant from the National
Science Foundation (BSR 8618487). We also wish to
acknowledge Bill Herget of Nicolet Instruments for his
technical assistance in getting the FTIR instrument set up
and operational, Tom Garrison and Greg Shore for help
with collection and processing of FTIR data and Yorgos
Marinakis for help in collecting and interpreting
micrometeorological data. We would also thank Shashi
Verma for sharing his micrometeorological data from the
1989 FIFE experiment at Konza Prairie.
REFERENCES
1. Spartz, M.L. M.R. Witkowski, J.H. Fateley, J.M.
Jarvis, J.S. White, J.V. Paukstelis, R.M. Hammaker, W.G.
Fateley, R.E. Carter, Jr., M. Thomas, D.D. Lane, G.A.
Marotz, B.J. Fairless, T. Holloway, J.L. Hudson, D.F.
Gurka. Evaluation of a mobile FT-IR system for rapid
volatile organic compound determination, Part I:
Preliminary qualitative and quantitative calibration
results. Amer. Environ. Lab. 1 (2) Nov. 1989, pp. 15-
30.
2. Robertson, G.P., M.A. Huston, F.C. Evans and J.M.
Tiedje. Spatial variability in a successional plant
community: patterns of nitrogen availability. Ecology 6 9
(5) 1988, pp. 1517-1524.
3. Gosz, J.R., C.N. Dahm and P.G. Risser. Long-path FTIR
measurement of atmospheric trace gas concentrations.
Ecology 69(5) 1988, pp. 1326-1330.
547
-------
4. Gosz, J.R., C.N. Dahm, D. I. Moore and S. Hofstadler.
Field testing Long-path Fourier Transform Infrared
(FTIR) Spectroscopy for measurements of atmospheric gas
concentrations. Remote Sens. Environ. 32 1990, pp. 103-
110.
5. Haaland, D.M. and R.G. Easterling. Application of new
least squares methods for the quantitative infrared
analysis of multicomponent samples. Applied Spectroscopy
36(6) 1982, pp. 665-673.
6. Andreas,E.L Can long-path FTIR Spectroscopy yield gas
flux measurements through a variance technique? Atmos.
Environ, (submitted), 1990.
7. Hicks. B.B. and P.S. Liss. Transfer of SC>2 and other
reactive gases across the air-sea interface. Tellus 46,
1976, pp. 348-354.
8. Hill, R.J. Implications of Monin-Obukhov similarity
theory for scalar quantities. J. Atmos. Sci. 46, 1989, pp.
2236-2244.
DISCUSSION
JOHN EVANS: I was just curious how you calibrate the instrument and
secondarily, what sort of precision and accuracy of the path length average
concentration you get for something, like methane, for example.
DOUG MOORE: We calibrated FTIR using white cell, .25 meter white cell in
the laboratory. And for things like methane we get detection limits of about 30
ppb, plus or minus.
JOHN EVANS: What sort of precision and accuracy can you get on normal
measurements say, in the atmosphere?
DOUG MOORE: Well, yes, that's going to be path dependent. I'm not sure I
understand. What numbers do you want it in?
JOHN EVANS: Well, we see a graph up there with some numbers up and down.
Are they 1%, 10%, 50% accuracy precision?
DOUG MOORE: We're better than 5% accuracy. Probably better than that on
the long path.
TOM PRITCHETT: When you were actually measuring this flux, do you have
to shine the beam directly over the source, or do you shine the beam downwind
of the source in looking at any downwind transport?
CLIFF DAHM: Our appl ications are quite a bit d ifferent than many other people
who are looking at point source. We're not looking at point source emission.
We're looking at something that's broadly distributed across the environment.
So, what we're looking at is something where we really need to know something
about fetch length from which the sources are generated. But we're not looking
at a point source. If we are looking at a point source, we would go into point
source analysis, we would have to be downwind and perpendicular to that point
source, or over that area of point source. What we're looking at, though, is
landscape emissions of things that tend to be distributed rather, at least reasonably,
uniformly over the environmenl.
TOM PRITCHETT: So, basically you're looking at the flux as coming directly
underneath your beam, essentially?
CLIFF DAHM: That's a very difficult question as to exactly where those gases
are coming. They're coming from downwind. It's very dependent, of course, on
wind field conditions at the time you're making the measurements. But you can
be generating input terms to your vertical structure of the atmosphere that can be
anywhere up to 100 times the height you are above the ground, 100 times upwind
of that direction. Again, it depends very much on meteorological conditions at the
time of the emission.
548
-------
PATTERN RECOGNITION METHODS FOR FTIR REMOTE SENSING
Gary W. Small*
Department of Chemistry
University of Iowa
Iowa City, IA 52242
Robert T. Krout11
U. S. Army Chemical Research,
Development, and Engineering Center
Aberdeen Proving Ground, MD 21010
ABSTRACT
Digital filtering and pattern recognition meth-
ods are described that Implement an automated
detection algorithm for passive Fourier trans-
form Infrared (FTIR) remote sensing data. The
detection 1s performed with only a 76-po1nt
segment of the FTIR Interferogram, thereby
enabling a "short-scan" Interferometer to be
used. Two novel pattern recognition methods are
Introduced that provide for the Intelligent
selection of training set patterns and for the
calculation of collectively optimized plecewlse
linear discriminants. This methodology Is
evaluated with a large quantity of passive FTIR
remote sensing data and 1s shown to perform 1n a
highly effective manner.
INTRODUCTION
Fourier transform Infrared (FTIR) remote sensors
are environmental monitoring devices that employ
an Interferometer-based optical system to col-
lect Infrared spectral data 1n the outdoor
environment. The resulting data can be analyzed
for the presence of characteristic spectral
bands corresponding to target analytes.
Infrared remote sensors can be operated 1n two
ways. The spectrometer can be used with an
external blackbody Infrared source, or the
sensor can be employed 1n a passive mode simply
to collect whatever ambient Infrared background
radiation 1s present 1n the field of view. The
passive technique 1s the more flexible of the
two Implementations, as the sensor consists of a
single unit.
Specific application environments for passive
'Corresponding author.
FTIR sensors Include monitoring at hazardous
waste sites, leak detection at chemical plants,
and regulatory monitoring of smokestack emis-
sions. In these applications, the spectrometer
can be positioned 1n a stationary configuration
or mounted 1n a ground or airborne vehicle.
Two fundamental problems limit the applicability
of passive FTIR sensors In demanding monitoring
applications. First, the sensor must be rugged
enough to operate under the conditions required
for the application. For example, 1n an air-
borne Implementation, the spectrometer must be
stable enough to allow data collection under
conditions of moderate vibrations or varying G-
forces. Second, 1n the passive FTIR experiment,
no stable Infrared spectral background exists
for use 1n processing the collected data.
Standard laboratory spectral processing tech-
niques that employ a background or reference
spectrum cannot be used.
The most fragile component of a typical FTIR
remote sensor Is the Interferometer drive mecha-
nism, which must allow the collection of a
stable Interferogram of 1024 or 2048 points.
The required Interferogram length 1s dictated by
the spectral resolution required to detect the
target analyte(s) of Interest. This relation-
ship between Interferogram length and spectral
resolution derives from an Inherent characteris-
tic of the Fast Fourier Transform (FFT), the
data processing tool used to extract Infrared
spectra from the collected Interferograms. The
FFT assumes that the Interferogram 1s an Infi-
nitely long waveform that contains zeros for all
points not explicitly collected. This has the
effect of adding s1n(x)/x components to the
computed spectrum, resulting 1n spectral broad-
ening. As the number of collected Interferogram
points Is Increased, the s1n(x)/x contribution
549
-------
1s decreased. Correspondingly, spectra computed
from very short Interferograms are severely
distorted due to these effects.
One approach to Increasing the potential rugged-
ness of an FTIR remote sensor Is to adopt a
simplified "short-scan" Interferometer design.
The drive mechanism for such a system would
allow only the collection of a 100-200 point
Interferogram segment. Conceptually, this
system would be much more rugged than a conven-
tional design, as the moving mirror of the
Interferometer would need to maintain optical
alignment for only a very short distance. The
drawback to such a system 1s that a conventional
spectral-based analysis cannot be performed, due
to the characteristics of the FFT noted above.
Recently, we have Introduced an alternative FTIR
data processing strategy that 1s compatible with
short Interferograms (1). The approach used 1s
based on the application of bandpass digital
filters directly to short Interferogram seg-
ments. If the filter bandpass 1s chosen to
coincide with the frequencies of the spectral
band(s) of a target analyte, the application of
the filter has the effect of extracting specific
spectral Information directly from the Interfer-
ogram segment. This approach overcomes the
limitations of the FFT, but still allows the
data analysis to be based on selected Infrared
frequencies. Additionally, judicious choice of
the Interferogram segment allows the analysis to
be performed without the use of data describing
the Infrared background.
The principal drawback to this scheme 1s the
difficulty 1n Interpreting the filtered Inter-
ferogram data. Virtually every application of
FTIR remote sensing requires that the collected
data be Interpreted automatically and a decision
made as to the presence or absence of the target
analyte(s). In application scenarios such as
leak detection, a positive decision regarding
the presence of the analyte 1s used to trigger
an alarm. Clearly, 1n such cases, the decision-
making aspect of the analysis 1s critical.
In the work presented here, pattern recognition
techniques are described that allow the Imple-
mentation of an effective decision-making algo-
rithm for use 1n analyzing filtered Interfero-
gram segments. The utility of this methodology
1s demonstrated through the use of a large
quantity of passive FTIR remote sensing data.
EXPERIMENTAL
The FTIR remote sensing data used for this
research were collected with a passive FTIR
sensor built by M1dac Corp. (Costa Mesa, CA) to
the specifications of the U.S. Army Chemical
Research, Development, and Engineering Center,
Edgewood, MD. The spectrometer design 1s based
on a linear-drive Mlchelson Interferometer
coupled with a 11qu1d-n1trogen-cooled Hg:Cd:Te
detector that responds 1n the range of 8-12 urn.
The collected data consisted of I024-po1nt
Interferograms, with a corresponding spectral
resolution of approximately 4 cm~l. The data
collection was performed with the Instrument
placed on a tripod. Under a variety of Infrared
background conditions, a test analyte, SF§
(Matheson Gas Products, Secaucus, NJ), was
released In the field of view of the spectrome-
ter. SF5 was selected as a target because of
Its use as a standard test compound 1n pollution
monitoring. It has a single strong absorption
at 940 crrfl. Due to the great variety of Infra-
red backgrounds observed, the collected data
contained both SFs absorption and emission
bands.
The data analysis described here was performed
by use of software written 1n FORTRAN-77 and
assembly language. The design of digital fil-
ters and the selection of the pattern recogni-
tion training set were performed on a Prime 9955
computer system operating 1n the Gerard P. Weeg
Computing Center at the University of Iowa. The
pattern recognition analysis was performed on a
Hewlett-Packard Vectra RS/20c, a 20-MHz 80386
IBM PC-compatible microcomputer with 4 Mb RAM
(Hewlett Packard, Inc., Sunnyvale, CA). The MS-
DOS 3.3 operating system was used. The compil-
ers, assembler, and operating system used with
the Hewlett-Packard system were manufactured by
Microsoft, Inc. (Redmond, WA). This software
was executed under the Desqv1ew-386 multi-task-
ing environment (Quarterdeck Office Systems,
Santa Monica, CA).
OVERVIEW OF INTERFEROGRAM-BASED ANALYSIS
Figure 1 displays the action of a bandpass
digital filter 1n the spectral domain. A sin-
gle-beam spectrum Is displayed with an absorp-
tion band at 940 cm~l from the analyte, SF5.
The Interferogram producing this spectrum was
collected with the remote sensor positioned on
top of a building looking down at a ground
source of SF5, approximately 180 feet away.
Superimposed on the spectrum 1s a Gaussian-
shaped frequency response function of a digital
filter. The frequency response has a width at
half maximum of 33.0 cm"*-, and 1s centered on
the SFs absorption band. This filter can be
applied 1n the spectral domain by multiplying
the frequency response function by the single-
beam spectrum. The resulting filtered spectrum
Is zeroed outside of the filter bandpass, and
the SFs absorption 1s superimposed on the filter
bandpass function.
The same filtering procedure can be performed 1n
the Interferogram domain. Here, the correspond-
550
-------
Ing operation 1s the convolution of the Inter-
ferogram and the time-domain representation of
the frequency response function.
100 1000 noo
FUEQUENCr (em")
Figure 1. Single-beam spectrum (solid line)
with filter frequency response (dashed line)
superimposed.
In mathematical terms,
H(f)X(f) <-> /h(k)x(t-k) dk
(1)
where H(f)X(f) 1s the product of the frequency
response function, H, and the single beam spec-
trum, X. The Fourier transform pair of H(f)X(f)
Is the convolution of the raw Interferogram, x,
and the 1nterferogram-doma1n representation of
the filter bandpass, h (termed the Impulse
response of the filter). H and X are functions
of frequency, f, while h and x are functions of
the time variables, t and k.
In the Interferogram, the filtering operation
suppresses those sinusoidal signals whose fre-
quencies^ He outside of the filter bandpass.
The flittered 1nterferogranv1s thereby reduced to
two features: (1) the Interferogram representa-
tion of the Gaussian frequency response function
and (2) the corresponding representation of the
analyte band. As the Gaussian feature 1s wider
than the absorption band. Its Interferogram
representation damps at a faster rate. Thus,
beyond the point 1n the filtered Interferogram
where the representation of the Gaussian feature
has damped to zero, the dominant Information 1s
a sinusoidal signal whose amplitude 1s related
to the height of the analyte absorption band.
Figures 2 and 3 Illustrate these concepts.
Figure 2 depicts points 160-235 (relative to the
centerburst) 1n two unfUtered Interferograms.
The lower Interferogram corresponds to the
single-beam spectrum In Figure 1. The upper
Interferogram was collected during the same
experiment, but SFg was not present 1n the field
200 210
INTERFEROGRAM POINT
Figure 2. Segments (points 160-235) of two
Interferograms collected by the remote sen-
sor. SF5 was present when the lower Inter-
ferogram was collected.
INTERFEROGRAU POINT
Figure 3. Interferogram segments from Figure
2 after application of the bandpass filter.
The SFs Information 1s now clearly seen 1n
the lower plot.
551
-------
of view of the spectrometer. No discernible
difference can be seen 1n the two Interferograms
to Indicate the presence of SF5 Information In
the lower plot. Figure 3 displays the same
Interferogram segments after application of an
1nterferogram-doma1n digital filter developed to
approximate the frequency response 1n Figure 1.
By suppressing frequencies other than those
associated with the targeted spectral band, the
filtering operation produces a signal that can
be used to detect the presence of the analyte.
The 76-po1nt sinusoidal signals 1n Figure 3 form
the test data used In this work 1n the develop-
ment of an automated detection scheme for pas-
sive FTIR remote sensors.
DIGITAL FILTER DESIGN STRATEGIES
The design of a practical digital filter for use
1n the manner described above requires that the
h(k) values 1n eq. 1 be generated such that the
convolution Integral can be approximated accu-
rately. Additionally, the approximation must be
truncated to a finite number of terms. The most
common approach to this approximation takes the
form of
*'t
hOxt + hlxt-l
nkxt-k
(2)
where x'^ 1s point t 1n the filtered Interfero-
gram, the h(< are as defined above, and the xt_^
are points In the raw (I.e. unflltered) Inter-
ferogram. Since the Impulse response function
has been truncated to a finite number of terms,
filters of this type are termed finite Impulse
response (FIR) digital filters.
The most widely used approach to the generation
of FIR filter coefficients was developed by
McClellan and Parks (2). In this method, the
Remez-Exchange algorithm 1s used to generate a
series approximation to the frequency response
function of the filter. The h|< are then comput-
ed directly from this series approximation.
Recently, we have Introduced a design strategy
for FIR filters based on regression analysis
(3). This approach 1s based on two considera-
tions. First, some of the terms 1n eq. 2 are
undoubtedly more significant than others 1n
obtaining a good approximation to the convolu-
tion Integral. Therefore, 1t may be possible to
delete some terms without a significant loss In
filter performance. Given that eq. 2 1s a
linear model, standard regression analysis
techniques can be used to assess the signifi-
cance of the Individual terms. In this computa-
tion, a set of Interferograms collected by the
remote sensor Is used to build the regression
models.
Second, a better approximation to the convolu-
tion Integral may be obtained by utilizing a
different set of filter coefficients for each
Interferogram point. For example, Interferogram
points 160 and 161 would be filtered with dif-
ferent filter coefficients. Analogously, the
filter for use 1n application to a I00-po1nt
Interferogram segment would contain 100 sets of
filter coefficients.
We have termed these filters FIR matrix (FIRM)
filters, as the filter format defines a two-
dimensional matrix of filter coefficients. By
tailoring each filter to an Individual point,
smaller sets of filter coefficients can be used
at each point, thereby saving computation time.
In testing, FIRM filters outperformed conven-
tional FIR filters with twice the number of
coefficients.
For the work reported here, a FIRM filter was
generated for points 160-235 based on a set of
2429 Interferograms collected by the FTIR remote
sensor. Both SFs-conta1n1ng (429) and non-SF5
Interferograms (2000) were present. The filter
was generated to approximate the frequency
response shown 1n Figure 1. The region of
points 160-235 was selected as the point at
which the Information due to the Gaussian fre-
quency response function has effectively damped
to zero. Across this point range, the Gaussian
signal decreases from 0.2/8 to 0.000556 of Its
maximum value.
In the filter calculation, a stepwlse multiple
linear regression procedure was used to select
statistically significant terms from the region
of t-0 to t-100 1n eq. 2. To be Included 1n the
final model, terms had to meet a significance
level of 99.99%, based on the £ distribution.
The resulting filter averaged 37 coefficients
per point, while the average value of R2 for the
regression calculations was 93,2%. This filter
was used 1n the generation of Figure 3.
INTERFEROGRAM ANALYSIS BY PATTERN RECOGNITION
The filtered Interferogram segments 1n Figure 3
are easily differentiated. However, as the
analyte band decreases 1n Intensity to the limit
of detection, the corresponding filtered Inter-
ferogram segments are Indistinguishable from
those arising solely from background noise. If
a decision 1s to be made regarding the presence
of the analyte, a procedure must be devised for
distinguishing Interferogram segments exhibiting
weak analyte signals from those exhibiting only
noise.
Pattern recognition techniques are numerical
algorithms for use 1n classifying data objects
("patterns") Into categories or classes. These
methods have been used 1n a variety of applica-
tions In chemistry (4-6). In the present exam-
ple, Interferograms belong to one of two d1s-
552
-------
tlnct categories: (1) SF6-act1ve or (2) SFs-
1nact1ve. The patterns in this case are the 76-
polnt filtered Interferogram segments. These
Interferogram segments can be considered as
points 1n a 76-d1mens1onal vector space. If the
points corresponding to the given categories
cluster together In the data space, pattern
recognition techniques can be used to assign
unclassified points to the appropriate catego-
ries.
Two Issues are paramount 1n developing a suc-
cessful pattern recognition analysis scheme.
First, a representative set of example data must
be obtained for use 1n developing the data
classification algorithm. This "training set"
of data must» to the degree possible, encompass
the range of patterns to which the analysis will
be exposed. Second, the appropriate pattern
recognition algorithm must be selected. A
knowledge of the data space must be gained 1n
order to make the proper selection. Both of
these Issues are addressed below for the case of
the filtered Interferogram data.
OPTIMAL SELECTION OF TRAINING SET MEMBERS
When Initially forming a training set, 1t 1s
desirable to select a variety of patterns from a
pool of candidate patterns that 1s as large as
possible. A training set comprised of a large
number of patterns 1s not necessarily the same
as one comprised of a large variety of patterns,
however. Many of the available patterns may
effectively be duplicates. The Intelligent
selection of training set members Involves
maximizing pattern diversity while minimizing
the Inclusion of duplicate patterns.
This type of optimized selection of training set
members becomes Increasingly complicated as the
number of candidate patterns Increases and as
the dimensionality of the patterns themselves
Increases. In the present example, 31 different
data sets were collected, consisting of approxi-
mately 14,700 interferograms. Standard tech-
niques for deducing pattern similarity such as
the calculation of all palrwlse distances be-
tween patterns are computationally cumbersome
with data sets of this size. To address this
problem, we have developed an algorithm which
provides an automated way to select optimal
training sets which have the same characteris-
tics as the starting pool of candidate patterns.
The intelligent selection of patterns for use 1n
the training set requires some type of distance
calculation to quantify the relationships among
the data points. Performing this calculation
with the 76-dimens1onal data 1s undesirable,
however, due to the high computational cost.
The selection process can be greatly simplified
by reducing the dimensionality of the patterns.
This can be accomplished through the use of
principal components analysis (7,8). For n-
dlmenslonal patterns, an optimal ^.-dimensional
representation can be formed simply from the
projections of the n-d1mens1onal patterns onto
the first £ principal components. To Insure the
accuracy of any subsequent interpoint distance
calculations, £ 1s typically chosen to span a
large fraction (e.g. 9556) of the data variance.
Patterns for the training sets were chosen by
use of an algorithm which divides a £-dimension-
al principal components space Into smaller .p_-
dlmenslonal volumes. One pattern 1s then se-
lected from each of the smaller volumes, thus
providing equitable, global sampling of all
patterns 1n the principal components space. For
ease of conceptualization, this global sampling
strategy 1s Illustrated in three dimensions 1n
Figure 4. The total number of smaller volumes,
blocks 1n this case* as well as the shape of the
blocks, 1s determined by the number of specified
divisions along each principal component. The
number of divisions 1s termed the mesh size.
Each of the smaller volumes may contain several
patterns, as shown in the expanded view of one
of the blocks In Figure 4. To insure that the
pattern selected from that block 1s the most
different from patterns selected from neighbor-
Ing blocks, the pattern closest to the center of
each block is chosen. The center of the block
1n the expanded view 1s Indicated by a solid
dot. Applying this procedure to all blocks
results In a smaller set of selected patterns
Figure 4. Conceptual depiction of the divi-
sion of a three-dimensional principal compo-
nents space Into smaller three-dimensional
volumes. Patterns are selected from each of
the smaller volumes.
553
-------
which preserves the overall distribution of
patterns 1n the full data set. For the filtered
Interferogram data, the spread of patterns for
the two data classes 1s different. For this
reason, the principal components analysis and
pattern selection were performed Individually
for the two data classes.
After the number of blocks and their dimensions
are defined, each block 1s assigned a unique
number from 1 to the total number of blocks, N^.
M is defined as
(3)
where tru 1s the selected mesh size for the Jtn
principal component and a Is the number of
principal components being used. The relative
location of each pattern Is thus defined by
determining Its block number. Since the block
numbers are positive Integers, the real princi-
pal components coordinate values of each pattern
are also converted to positive whole numbers.
Coordinates along the J^h principal component
are transformed as
mJC(cj -
- Cj,m1n)] (4)
where cj' 1s the transformed coordinate, cj 1s
the original coordinate, mj 1s the mesh size
defined above, and cj,m-|n and Cj>max are the
minimum and maximum coordinates along the jtn
principal component. The computed Cj1 values
are then rounded to the nearest Integers, there-
by creating a new set of coordinates for each
pattern. The transformed Integer coordinates
are designated as Cj". The block number, B, of
any pattern can then be computed directly as
B =
C(C1" ~
1=2
1-1
IT
j=l
(5)
where c\n 1s the Integer coordinate value of the
pattern along the first principal component, cj"
1s the Integer coordinate value along the 1^n
principal component, and $. and mj are as defined
above. After computing the block number for a
given pattern, the center of the block and the
distance to the pattern are computed using the
cj1. This distance 1s used later to select the
pattern which is closest to the center of the
block. The patterns are then sorted by Increas-
ing block number, and the total number of occu-
pied blocks 1s calculated. The final set of
optimum patterns 1s selected by sampling each of
the occupied blocks.
This procedure was used to reduce the set of
14,700 Interferograms to 4000. Of the 4000
filtered Interferogram segments chosen to form
the training set, 2000 contained SF5 signals,
while 2000 contained no SFs Information. Six
principal components were used 1n the selection
of the non-SF6 patterns, while three principal
components were used In the selection of the
SFg-act1ve patterns. This training set was used
1n the development of a pattern recognition
scheme for the automated detection of SFg sig-
nals 1n the filtered Interferogram data.
PIECEWISE LINEAR DISCRIMINANT TECHNIQUES FOR THE
AUTOMATED DETECTION OF SF6
The selection of an appropriate pattern recogni-
tion technique for the SFs detection problem 1s
keyed by an Investigation of the manner 1n which
the SF5 and non-SF5 data classes cluster 1n the
76-d1mens1onal space. Principal components
analysis can be used to explore these relation-
ships visually. Figure 5 1s a plot of the
projections of the 4000 training set patterns
onto the first three principal components of the
data. SFs-actlve Interferogram segments are
Indicated by open circles, while non-SF5 Inter-
ferograms are Indicated by solid triangles. All
of the non-SFs points are clustered at the
center of the plot. To provide a better view of
the Interface between the data classes, Figure 6
1s an expanded view of the boxed region 1n
Figure 5. It 1s clear from an Inspection of
Figure 6 that the data classes merge at the
limit of detection of
PC2
PCI
Figure 5. Principal components plot depict-
ing the relationships among the 4000 patterns
1n the training set. SF5~act1ve patterns are
depicted as open circles. The non-SFs pat-
terns (solid triangles) are all located 1n
the boxed region.
554
-------
Figure 6. Expanded view of the boxed region
1n Figure 5. The SFg-act1ve patterns are
again Indicated as open circles, while the
non-SF5 patterns are Indicated as solid
triangles.
Consideration of the class distributions 1n
Figures 5 and 6 suggests that plecewlse linear
discriminant analysis (9,10) 1s the pattern
recognition method of choice for the filtered
Interferogram data. This technique 1s based on
the construction of boundaries or separating
surfaces between the data classes. The separat-
ing surfaces are termed discriminants, as they
define boundaries 1n the data space that allow
the classes to be discriminated. The plecewlse
linear discriminant consists of multiple linear
surfaces which collectively form a plecewlse
approximation of a nonlinear separating surface.
The need for a nonlinear discriminant Is clearly
motivated 1n Figures 5 and 6 by the circular
distribution of the SF5~act1ve points around the
points.
Each linear surface comprising the plecewlse
linear discriminant 1s defined by the locus of
points orthogonal to an n-d1mens1onal vector
termed a weight vector or discriminant, where n
1s the dimensionality of the pattern data (76 in
the present example). Each weight vector, w, is
calculated such that
wx
(6)
(7)
where xa represents a SFs-act1ve pattern, and xn
represents a non-SFg pattern. The dot products
1n eqs. 6 and 7 are termed discriminant scores.
The Individual weight vectors comprising the
plecewlse linear discriminant are calculated
sequentially, with each discriminant separating
a portion of the patterns 1n the training set.
The algorithm used for this work calculates
discriminants that have a pure-class subset on
one side of the discriminant and a mixture of
the two classes on the other side. A discrimi-
nant of this type is termed "single-sided".
After a single-sided discriminant has been
calculated, those patterns on the pure-class or
single-side of the discriminant are removed from
the calculation, and another discriminant 1s
computed In the same manner. The result 1s a
set of discriminants 1n which each discriminant
separates a different pure-class subset. Col-
lectively, the set of discriminants defines a
separating surface.
To classify an unknown pattern, each discrimi-
nant 1s applied to the pattern, producing one of
two possible results: (1) the pattern falls on
the pure-class side of the discriminant; or (2)
the pattern falls on the mixed-class side of the
discriminant. The first discriminant which
classifies the pattern onto the pure-class side
determines the class of the pattern. The last
discriminant determines the class If the unknown
pattern is never classified on the pure-class
side.
A multi-step procedure was devised to calculate
and optimize the set of discriminants comprising
the plecewlse linear separating surface. The
keys to this algorithm are. the use of Simplex
optimization techniques (11) to position each
discriminant and a novel discriminant recalcula-
tion procedure to perform a collective optimiza-
tion of the discriminants.
The Simplex algorithm computes a new weight
vector w by moving the previous w In an optimal
direction in the data space. The algorithm
consists of a set of rules which governs this
movement based on a numerical response function
that reflects the performance of the weight
vector. For the current work, many variations
of response functions were Implemented and
evaluated. To be effective, the response func-
tion must encode several characteristics related
to the performance of each weight vector, In-
cluding the number of patterns separated, and
whether the discriminant Is single-sided. In
addition, the response function should define a
continuous surface along which the optimization
can travel. The response function, R, used 1n
this work 1s defined as
(8)
(9)
R = Cl.O - S/f] S
S = (Ns/Nt)a Ns
555
-------
f = 2 [1og(Ns)+l]
(10)
tors.
S 1s termed the single-sided response, where Ns
1s the number of SFg-act1ve patterns separated,
Nt 1s the total number of patterns placed on the
single-side of the discriminant, and a 1s an
exponent that penalizes discriminants that are
not single-sided. Appropriate values of a have
been determined empirically to be 1n the range
of 10-200, depending on the magnitude of Ns.
For a single-sided discriminant, S 1s equal to
the number of SF5~act1ve patterns separated by
the discriminant (I.e. (Ns/Nt)a = 1.0). R 1s
made a continuous function by the use of s., the
standard deviation of the discriminant scores
for the non-SFs patterns. A smaller standard
deviation value produces a larger (I.e. more
optimum) value of the response function. It was
hypothesized that by minimizing the variation of
the non-SFg discriminant scores, the resultant
discriminant would be more nearly aligned with
the Interface between the two data classes. For
the data used here, the value of the standard
deviation 1s typically on the order of 10~5, and
consequently must be scaled to reduce Its Influ-
ence on R. The scaling factor, f, 1s used for
this purpose. Thus, the value of R can be
Interpreted as the value of S (I.e. number of
SF6-act1ve patterns) that has been penalized
based on the degree of variation among the
discriminant scores for the non-SF5 patterns.
The Simplex optimization described above 1s an
effective technique for optimizing each weight
vector. However, optimizing each weight vector
Individually may not produce the optimum piece-
wise linear discriminant, since the discriminant
consists of a set of weight vectors. To address
this problem, a collective optimization algo-
rithm was developed for this study. The proce-
dure used here 1s motivated by considering that
the calculation of the Initial set of weight
vectors 1s hierarchical 1n nature. The calcula-
tion of each weight vector 1s Influenced by the
performance of weight vectors that have been
previously computed. Each of these vectors 1s
computed such that 1t separates as many of the
remaining patterns as possible. In order to
effect a collective optimization, a method must
be developed to allow subsequent weight vectors
to Influence the calculation of previous weight
vectors.
The recalculation 1s performed Identically to
the single weight vector calculation described
above, but the data set of patterns 1s altered
to reflect the presence of other vectors 1n the
set. Prior to recalculating a given weight
vector, those patterns classified by later
weight vectors are removed from the data set.
This simple procedure allows the earlier weight
vectors to be reposltloned based on the classi-
fication performance of the later weight vec-
Employlng the 4000-member training set, a piece-
wise linear discriminant was computed consisting
of eight weight vectors. The Simplex optimiza-
tion algorithm was used to optimize each of the
vectors, and the recalculation procedure was
applied to optimize the set of weight vectors
collectively. The recalculated discriminant
classified 3945 of the 4000 patterns correctly
(98.655).
To evaluate the prediction performance of the
discriminants, two data sets were employed that
were not represented among the 4000 Interfero-
grams 1n the original data set. The two data
sets each contained 1000 Interferograms. The
application of a plecewlse linear discriminant
to a set of unknown patterns Is performed by
computing the discriminant score for each pat-
tern. In a graphical representation, the re-
sults can be displayed as a plot of the discrim-
inant scores vs. pattern number. Since multiple
weight vectors are used, there are multiple
discriminant scores that could be plotted. For
the purposes of this analysis, the largest
discriminant scores obtained by applying all
weight vectors to each pattern were used. For
SF6~act1ve patterns, the signal 1s then maxi-
mized, and for non-SFs patterns, the plotted
values then reflect the distance from the pat-
tern to the nonlinear separating surface.
Figures 7 and 8 show the resulting plots of
discriminant scores for the two prediction data
sets. The discriminant scores greater than zero
In the plots correspond to detections of SF6.
An Inspection of the figures and plots of trans-
formed spectra Indicate that the detections are
highly accurate. The rate of false alarms 1s
less than 1/6. These results suggest that the
combination of an Intelligent training set
selection algorithm along with the calculation
of an optimized plecewlse linear discriminant
produces a sensitive, effective detection scheme
for passive FTIR data.
CONCLUSION
The results presented here confirm that a short
Interferogram segment can be used for the reli-
able detection of target analytes from passive
FTIR data. The combination of digital filtering
and pattern recognition techniques allows this
detection algorithm to be Implemented. This
achievement makes possible the design of a new
generation of passive FTIR sensors based on the
"short-scan" Interferometer concept.
These results also Introduce two new general-
purpose algorithms for use 1n pattern recogni-
tion analyses. The training set selection
algorithm described here can be used to select
556
-------
training sets for use with any pattern recogni-
tion method. Results obtained 1n testing this
algorithm Indicate clearly that the method
outperforms pattern selection strategies based
on random sampling of a pool of candidate pat-
terns.
W^*^I|M4
100 200 300 400 500 600 700 100 000 1000
INTERFEROGRAM NUMBER
Figure 7. Plot of discriminant scores for the
first prediction data set. None of these 1nter-
ferograms were Included 1n the calculation of
the plecewlse linear discriminant.
100 200 JDO 400 500 (.00 700 BOO BOO 1000
INTERFEROGRAM NUMBER
Figure 8. Plot of discriminant scores for the
second prediction data set. None of these
Interferograms were Included 1n the calculation
of the plecewlse linear discriminant.
The multi-step procedure described above for
optimizing the placement of plecewlse linear
discriminants Is also a general approach that is
not limited to the remote sensing application
used here. The techniques developed 1n this
work are applicable to any pattern recognition
problem 1n which the Interface between the data
classes 1s complex. The optimized discriminants
are particularly suited to problems 1n which 1t
1s Important that the discriminants define the
limit of detection of a species.
Work 1s continuing 1n our laboratory on the
overall problem of collective optimization of
the weight vectors comprising a plecewlse linear
discriminant. We are currently exploring the
possibility of operating the simplex optimiza-
tion with a response function based on the
performance of all weight vectors simultaneous-
ly.
REFERENCES
(1) Small, G. W., Kroutil, R. T., Ditillo, J.
T., Loerop, W. R., "Detection of Atmospher-
ic Pollutants by Direct Analysis of Passive
Fourier Transform Infrared Interferograms",
Analytical Chemistry. 6Q» 1988, 264.
(2) McClellan, J. H., Parks, T. W., "A Unified
Approach to the Design of Optimum FIR
Linear-Phase Digital Filters", IEEE Trans-
actions no Circuit Theory. CT-20, 1973,
697.
(3) Small, G. W., Harms, A. C., Kroutll, R. T.,
Ditlllo, J. T., Loerop, W. R.,"Design of
Optimized Finite Impulse Response Digital
Filters for Use with Passive Fourier Trans-
form Infrared Interferograms", Analytical
Chemistry. £2, 1990, 1768.
(4) Jurs, P. C., "Pattern Recognition Used to
Investigate Multivarlate Data 1n Analytical
Chemistry", Science. 232, 1986, 1219.
(5) Derde, M. P., Massart, D. L., "Supervised
Pattern Recognition: the Ideal Method?",
Analytica Chimica Ac±a., 121, 1986, 1.
(6) Varmuza, K., "Pattern Recognition 1n Ana-
lytical Chemistry", Analytica Chimica .Ada.,
122, 1980, 227.
(7) Retelling, H.,"Analysis of a Complex of
Statistical Variables into Principal Compo-
nents", Journal M Educational Psychology,
2A, 1933, 417.
(8) Martens, H., Naes, T., "Methods for Cali-
bration", Multivariate Calibration. Wiley,
New York, 1989, p. 111.
557
-------
(9) Lee, T., Richards, J. A., "P1ecew1se Linear
Classification Using Seniority Logic Com-
mittee Methods, with Application to Remote
Sensing", Pattern Recognition. 1Z, 1984,
453.
(10) Duda, R. 0. and Possum, H., "Pattern Clas-
sification by Itenatively Determined Linear
and P1ecew1se Linear Discriminant Func-
tions", IEEE Transactions on Electronic
Computers. 15., 1966, 220.
(11) Routh, M. W.; Swartz, P. A.; Denton, M. B.,
"Performance of the Super Modified
Simplex1', Analytical Chemistry. 42, 1977,
1422.
DISCUSSION
DONALD GURKA: Can you visualize a digital filter analog to the Hadamard,
to eliminate the multi channel disadvantage on the transparent spectral component?
For example, use a series of digital filters which would only let through the
channels of information that you want to transform.
GARY SMALL: It's a nice idea. The problem would come in the construction
of the filters that would have many, very narrow individual band passes. The
problem that you get in the design of filters is that the narrower you want the band
pass, the more difficult it is to actually implement the filter that will work in the
time domain. The problem that you would come into would be having to have
either many individual filters, or to have a very complex filter that would have
multiple band passes. So, I think the key question would really be going back to
the electrical engineering techniques that one uses in designing filters to see
whether that would be viable. Our experience is that if you really want very
narrow band passes, it's a difficult problem in filter design. That might be too
tough, actually.
DONALD GURKA: So, the answer is yes, but it won't be easy?
GARY SMALL: The answer is it's conceptually a nice idea. I think implementing
it would be difficult.
558
-------
REMOTE VAPOR SENSING USING A MOBILE FTIR SENSOR
R.T. Kroutil, J.T. Ditillo, R.L.
Gross, R.J. Combs, W.R. Loerop; U.S.
Army Chemical Research, Development
and Engineering Center, Aberdeen
Proving Ground, Maryland 21010
G.W. Small; Department of chemistry,
University of Iowa, Iowa City, Iowa
52242
(A) Introduction
The remote-passive detection of
hazardous chemical vapors is an
important application for both the
military and civilian communities
interested in environmental issues.
Remote Fourier Transform Infrared
[FTIR] Spectrometers are capable of
detecting absorptions and emissions
of low-concentration chemical vapor
clouds using an ambient temperature
atmospheric background. For many
pollution-monitoring problems FTIR
spectroscopy represents the only
viable approach for the detection of
many environmental pollutants.
Remote FTIR technology,
developed to detect chemical warfare
agents, is directly applicable for
compliance assurance for many of the
chemicals listed in the U.S. Clean
Air Act. FTIR spectrometers have
the potential to monitor stack
emissions, hazardous components in
wood smoke, auto emissions, and
industrial releases. By using a
FTIR one can detect vapor
concentrations from chemical leaks
or spills. With advanced warning
provided by an FTIR, residents
located in a surrounding area might
be given enough time to safely
evacuate. In these applications an
FTIR would be mounted on an
emergency response team helicopter
to give an identification of a
particular chemical species.
An infrared remote chemical sensor
consists of a sensor and signal
processor that operate in parallel to
give an indication of the presence of a
pollutant. The sensor detects the
signatures of all chemical vapors and
backgrounds, while the signal processing
algorithms discriminate between the
spectral features associated with the
pollutant and background emissions. The
typical instrumentation required for
remote infrared chemical vapor sensing
consists of a two wave number resolution
interferometer with a specialized set of
collimating optics. The signal
processor detects a chemical cloud in a
fixed-site application by measuring the
background radiant emission profile as a
function of time. When a target cloud
moves into the field of an
interferometer a specific change in the
background radiant emission profile is
detected.
Recently, a number of new
applications using a remote chemical
sensor have been developed using
interferometers operating from
helicopters, aircraft, and earth-
orbiting satellites. The remote sensing
problem for these cases is severely
complicated because the background
emission profile changes rapidly with
respect to time. When operating in a
rapidly moving scenario, the change in
emission profile versus time cannot be
559
-------
used for detection of the presence
of a vapor cloud. This change in
the radiant background can be an
order of magnitude greater than the
spectral emission profile of the
vapor cloud. In order to remove the
changing background spectral
features a signal processing
technique is needed.
The collection of data from
mobile FTIR applications is further
complicated by the extreme _data-
processing requirements in which an
interferometer may collect up to 30
two wave number scans per second.
For many pollution monitoring
applications, size, power, cost, and
weight limitations require that a
low-powered signal-board computer be
used for real-time data analysis. A
recent development in the signal
processing hardware area is the
advent of the Digital Signal
Processing (DSP) chip. Current DSPs
are capable of processing up to 33
million floating point instructions
per second. Remote sensors operating
in mobile environments can benefit
from the high computational
throughput of single-board computers
using DSP technology.
(B) THE BACKGROUND OF THE PROBLEM
One can consider the radiance
incident on a remote sensor as
combinations of energy from the
background, the target vapor cloud
of interest, and the intervening
atmospheric gases. One can describe
an integral equation consisting of
infinitesimal layers of atmosphere.
In this case, a radiance emission
source will be absorbed in the layer
by both the target cloud and the
intervening atmospheric
constituents. The radiance measured
at the detector is given by the
following equation, where
(1)
s:
- KT(X)
- KN'(X) }dx
where kT and k. are the extinction
coefficients of the target gas and
the atmosphere
of a blackbody at the temperature on the
infinitesimal layer. N1 is the radiance
incident on the infinitesimal layer
traveling to the sensor, and x is the
length variable that is integrated for
the length of the target cloud and
intervening atmosphere.
Assuming homogeneous atmospheric
and target cloud species, the
integration of equation (1) gives the
power incident on a passive sensor as
shown in reference [1],
(2)
[TATTNBQ
- TATT) NT
where TA is the atmospheric
transmittance, TT is the target cloud
transmittance, NB is the radiance of the
background, NT is the radiance of a
blackbody at the ambient temperature, A
is the collector area, and SL is the
solid angle of acceptance of the sensor.
The atmospheric cloud transmittance is
(3) T. = e -KAR
where R is the distance of the
intervening atmosphere. The target
cloud transmittance is
(4)
= e-l
-------
1 to 20 ppm-m. The exact detection
level is dependent on the
absorptivity of each compound.
(C) SIGNAL PROCESSING OF REMOTE
SENSING DATA FOR DETECTION AND
ALARM
(1) BACKGROUND
Signal processing of remote
sensing data is required to extract
background spectral features from
those of a vapor target cloud. It
has been shown that digital
filtering used in either the
frequency or the interferogram space
can be used to extract spectral
background features.[2,3,4] Signal
processing algorithms used in the
interferogram space have advantages
with application to detection and
alarm algorithms for remote sensors.
First, the signal processing
algorithm does not require a
conversion into the frequency domain
by a Fast Fourier Transformation
(FFT). This reduces the number of
computations required for a
detection algorithm. It also
reduces some resolution degradation
caused by the apodization function
of the FFT in the transformed
spectrum. Second, the broad band
spectral background features are
somewhat separated by point number
in the interferogram space. In the
time domain, the central fringe of
an interferogram is at the zero
retardation of the moving mirror of
the interferometer and contains
information from all spectral
wavelengths. For remote sensing
data the central fringe contains a
disproportionate contribution of the
broad-band blackbody radiation
curve.[5,6] As one moves further
out into the ends of the
interferogram, the broad spectral
components constructively interfere
with each other more than the narrow
band spectral features. The effect
is a severe damping of the
contribution of the broad spectral
components. Signal processing in
the interferogram domain can take
advantage of this effect for remote
sensing by processing only short
interferogram segments located
adjacent to the center-burst of the
interferogram. Digital filters used
away from the center-burst are not
required to operate over the entire
16-bit dynamic range of the data.
Because of these reasons, digital
filters can be used more efficiently for
signal detection algorithms in the
interferogram space than in the
frequency domain.
(2) FINITE IMPULSE RESPONSE FILTERS
The most commonly used digital
filter for signal processing
applications is known as the finite-
impulse response (FIR) filter. The
basic form of the equation is shown in
the following equation.
N
(3)
U
1=1
b
where Y is the filtered data point
resulting from the application of the
filter, X is the input raw data values,
and b, are the filter coefficients.
The purpose of an FIR filter used
for signal detection in a remote sensing
interferogram is to generate narrow
bandpass responses to eliminate the
background information. One of the most
widely used techniques for generating
coefficients for an FIR filter is the
process known as the Remez Exchange
procedure. In this method, the filter
coefficients, b, are generated through
the use of the attenuation approximation
theorem. In this theorem, if the
approximation error of the frequency
response outside of a passband response
is uniformly distributed, the resulting
narrow bandpass filter response error
will be minimized.
The Remez Exchange procedure must
satisfy the condition where the response
is defined as P, where,
M
(4)
b cos (wn)
n=o
A weighted error function is defined
as the difference to the true response
from that of P where,
(5) E (e'w) = W(e'H) [ D - P(6JW) ]
561
-------
In this equation W is the
passband to stopband weighing
function, and D is the actual real
frequency response of the function.
The alternation theorem states
that for any selected set of
extremal frequencies, w, the
alternation condition must be
satisfied where,
(6)
E(e'JH) = -I
The Remez procedure is based on
the fact that an iterative solution
can be developed in a convergence
procedure between equations (5) and
(6) . When the correct values of b
are generated, then value of E(e"j")
in the two equations will converge.
Generating narrow bandpass
digital filters to extract out the
signal of interest has been applied
to the analysis of collected remote
sensing interferograms that contain
a spectral absorptions of the
simulant Sulfur Hexafluoride (SF6) .
In this example a narrow bandpass
digital filter using 40 coefficients
was developed that had a center
frequency corresponding to 940 wave
numbers. At the modulation frequency
of the interferometer this spectral
frequency corresponded to a bandpass
of 50 Hz with a center frequency of
2450 Hz. Figure 1 shows two short
segment interferograms collected
from an interferometer mounted on a
UH-1 Army helicopter travelling at
120 knots and 1000 feet altitude.
The bottom interferogram was
collected when the interferometer
was travelling past the target cloud
of SF6. Figure 2 shows the results
after filtering the segment with the
40-term digital filter. The bottom
interferogram segment shows the
fundamental frequency in which the
SF6 was present. It should be noted
that the digital filter strategy can
detect either the absorption or
emission case. The result of a case
of the target cloud being warmer
than the background is that the
resulting filter output will be 180
degrees out-of-phase from the
absorption case.
The result of collecting
successive interferogram segments
while moving is shown in figure 3.
In this figure a magnitude response
of the output for the 40-term digital
filter is shown as a function of time.
During this run an interferometer was
mounted on a UH-1 helicopter and flown
around a source of SF6. As the
helicopter passed by the target cloud
(three times) the instrument alarmed to
make a detection. The x-axis in this
figure corresponds to approximately
three minutes of collected data. The
target cloud was released at the
beginning of the data run. During each
helicopter pass the response of the
digital filter became weaker due to the
fact that the target cloud was
dispersing in a 10 mile per hour cross
wind.
(3) INFINITE IMPULSE RESPONSE
FILTER
Infinite impulse response (IIR)
digital filters are generally feedback
loop filters in which additional filter
coefficients are used. The basic form
of the equation is shown in the
following equation.
N
C
M
The coefficients, b, are identical
to the FIR case. The only difference is
in the feedback response which is a
weighted sum of the past output values.
The weighted sum of present and past
input values are added to the feedback
response. The major advantage in using
a narrow band IIR filter for an
interferogram is that the number of
coefficients is reduced making it a
highly efficient filter. This effect
can be illustrated in figure 4. The
attenuation response in this figure
shows that the IIR case has a better
attenuation than for the FIR case. The
two filters compared are on a
logarithmic scale and have roughly
equivalent numbers of computations for
an interferogram segment. The major
disadvantage of IIR filters are that
they can be unstable over large dynamic
ranges and can have phase non-
linearities. A practical problem in the
implementation of narrow-bandpass IIR
digital filters is the required
interferogram segment length for
feedback response to stabilize and the
output result to become constant. This
562
-------
requirement is currently being
studied in order to develop
alternate methods for implementation
of IIR filters for analyzing short
interferogram segments.
(D) INSTRUMENTATION FOR REMOTE
SENSING
Instrumentation currently being
used for the mobile-remote detection
of chemical vapors consists of an
interferometer, an infrared
detector, an analog signal module,
and a digital signal processing
module. The interferometer and
detector collect the infrared
background spectral radiance and
convert it into an analog signal.
The analog signal processing module
filters and amplifies the detector
signal. The analog module has a 16-
bit analog-to-digital converter to
convert the signal to digital form.
The digital signal processing module
can analyze the data using a wide
variety of signal processing
techniques.
The interferometer constructed
for the U.S. Army CRDEC by the Midac
Corporation, Costa Mesa, California
occupies 0.3 cu feet, weighs
approximaely 15 pounds, and uses
only 28 watts at 12 volts. The
interferometer consists of a linear
drive mechanical mechanism capable
of collecting two wave number
spectra at speeds of up to 11 scans
per second. The interferometer has
a Helium-Neon 10 milliwatt laser to
provide a reference signal for the
analog-to-digital converter. The
mechanical mechanism is controlled
by two small electronic servo cards
in which one card contains the
analog electronics and the second
card contains the digital
electronics. The infrared detector
used in the interferometer was
purchased from Judson Electronics,
Costa Mesa, California, and is. a 2
mm square Mercury Cadmium Teluride
(MCT) infrared detector. A narrow
band detector is used in this
application since the atmospheric
transparent spectral window for
remote sensing is only from 8 to 12
microns. A Zinc Selenide (ZeSe)
beamsplitter allows the instrument
to give a measured noise equivalent
spectral radiance (NESR) of
approximately 1.5 x 10"8 Watts/cm2 * sr
* cm'1.
(E) DIGITAL SIGNAL PROCESSING HARDWARE
FOR REMOTE SENSING
The requirement to perform real-
time data analysis for the mobile
chemical sensor can easily surpass that
of today's conventional microprocessors.
This is particularly true when
applications demand several of the time
series analytical methods. Conventional
processors are designed to perform a
wide variety of functions, resulting in
lackluster performance during
multiplication and summing operations.
To overcome this shortcoming, many
microcomputer users purchase an optional
numerics coprocessor.
The coprocessors are microcomputers
that have been optimized to perform a
variety of mathematical functions.
These functions include integer and
floating point arithmetic as well as
some algebraic functions. Coprocessors
can increase the performance of a
microcomputer dramatically; however,
even the slowest of the DSP chips can
outperform the processor-coprocessor
combination by a factor of ten for
common functions required for signal
processing applications.
DSP processors are much faster than
conventional processors because of
differences in the chip design
architecture. To perform the multitude
of functions required by the desktop
microcomputer, the internal architecture
of the general-purpose processor is not
tailored to any particular application.
Most microprocessors use a single-bus
architecture in which both program
instructions and data flow across the
same set of data lines. This
architecture, known as von Neumann
architecture, can result in a data
bottleneck caused by the path of flow on
the data bus.
The size of the general purpose
registers can also have a serious effect
on computational performance. Intel's
8088 and 80286 processors have only 16-
bit-wide registers, while the 80386
possesses 32-bit registers. To perform
math operations, the microprocessor
breaks the numbers into manageable
portions and performs a series of
software operations to obtain the
desired result. This process requires
563
-------
many machine cycles to complete.
Numeric processors reduce the number
of required cycles by employing
larger registers. The Intel 8087
numeric processor has 80-bit-wide
internal registers; however, it
still requires multiple cycles to
perform even the simplest
mathematical computation.
DSP chips are distinguished
from the general-purpose processor-
coprocessor by their ability to
perform instructions in a single
cycle. Internal architecture of a
DSP is optimized to perform single-
cycle computations that allow faster
performance of the sum-of-product
calculations required by many
digital signal-processing
algorithms. The performance is
obtained through the use of hardware
multipliers-accumulators, Harvard
architecture, or pipelining.
Hardware multipliers and adders
of the DSP eliminate the software
overhead required by conventional
processors in mathematical
operations. These units allow the
DSP to perform operations in a
single cycle and insure sufficient
register width for accurate results.
These multipliers and accumulators
are arranged to optimize the
multiplication followed by addition
type operations.
To take advantage of the high
speed advantage of the multiplier,
the DSP must insure a steady flow of
data into it. To achieve this, many
different techniques are employed;
however, most manufacturers use some
variation of the Harvard
architecture. Unlike the von
Neumann approach, the Harvard
architecture uses separate program
and data memories, each having its
own bus or buses. In a digital
filtering operation, this
architecture allows the data and a
corresponding coefficient to be
fetched from memory along separate
buses and loaded into the multiplier
simultaneously while an instruction
is fetched on the program bus.
Pipelining is another scheme
used to insure an adequate flow of
data to the multiplier-accumulator.
In a pipelined architecture, each
instruction is composed of several
steps such as FETCH, DECODE, MULTIPLY,
and ADD. Each subsequent instruction is
likewise divided; however, it always
follows one step behind the previous
instruction in the pipeline. In other
words, while the first instruction is
decoding the instruction fetched one
cycle before, the second instruction is
being fetched from memory. While
requiring multiple cycles to complete an
entire instruction, once the pipeline is
filled, a result is obtained every
cycle.
The general-purpose DSP is a
single-chip integrated circuit designed
to allow the greatest flexibility as
well as provide good overall throughput.
These integrated circuits range from 16-
bit to 32-bit floating point
architectures and are capable of real-
time signal processing on signals of up
to 200 KHz. Commonly offered features
include zero overhead looping, bit
reversed addressing, and external
interfaces to serial and parallel
devices. Numerous combinations of on-
board and off-board memory are also
available from several manufacturers.
The general-purpose DSP architectures
make them ideal for a wide range of
applications.
The general-purpose DSP can be
programmed to perform any of the digital
signal processing algorithms much like a
conventional microprocessor. They are
generally programmed in their native
assembly code; however, many of the DSP
chips have a high level language
compiler available. The use of a high
level language makes
software conversion from the
microprocessor-based systems much
easier.
Popular DSP chips include Texas
Instrument's 320 family (32010, 32020,
320C25, and the 32030), the AT & T
DSP32C, the Motorola 56001, and the
Analog Device ADSP2100. Development
boards for the popular IBM
PC/XT/AT/compatibles are often available
from either the manufacturer or a third-
party source. These boards are either
used for the development of stand-alone
design or as a high speed digital signal
processing coprocessor for a host
computer. Assemblers, compilers,
simulators, and debuggers are often
included with many of the development
boards.
564
-------
The Remote Sensing Group at
U.S. Army Chemical Research,
Development and Engineering Center
(CRDEC) has selected the AT & T
DSP32C as the target processor. The
DSP32C is a CMOS 32-bit floating-
point processor based on a piplined,
von Neumann architecture. This 25-
MIPS (million instructions per
second) device contains twenty-one
16-bit fixed-point registers for use
in control, address, and logic
functions and, in addition, four 40-
bit accumulators to perform 32-bit
floating-point mathematical
operations. On-chip memory includes
2 K of read-only memory and 4 K of
random access memory. The DSP has
an off-chip memory capability of 16
MB. The DSP32C also supports serial
I/O and a parallel I/O channel
designed for easy interfacing to
either an 8-bit or a 16-bit
microprocessor.
(F) CONCLUSIONS
Infrared interferometer
hardware, signal processing computer
hardware, and the application of new
mathematical algorithms have rapidly
advanced the remote sensing
technology during the last four
years. Lightweight, small
interferometers exist that can
withstand severe mechanical
vibrations while operating on
rapidly moving helicopter platforms.
Signal processing algorithms are
available which can extract infrared
background information in order to
give an automatic alarm indication
for the presence of a particular
chemical vapor species. Finally,
digital signal processing hardware
has been constructed which allows
infrared remote sensors to process
data in real-time. This advance
eliminates the need to collect data
for later analysis in a laboratory.
York (1990)
W Small, A.C.
"- j "
Chan.. 62
Harms, R.T.
"-"- Loerop'
, Chemical Analysis Monograph
Wiley, New York (1975) .
E.G. Codding, G. Horlick,
^ ,27,85 (1973).
REFERENCES:
[1] S. Chandrasekhar, Radiative
Transfer, Dover, New York (1960).
[2] G.W. Small, R.T. Kroutil, J.T.
Ditillo, W.R. Loerop, Anal. Chem..
60, 264-269 (1988) .
[3] R.T. Kroutil, J.T. Ditillo, G.W.
Small, Computer-Enhanced Analytical
565
-------
50
25--
0
-25-
50
150
200
250
300
150
200
250
300
Figure 1
-------
150
200
250
300
s
150
200
250
300
Figure 2
-------
MAGNITUDE RESPONSE OF FIR FILTER
en
O>
oo
W.U 1 W
uj 0.008-
o
a.
ft 0.006-
UJ
a
JD 0.004-
<
^ 0.002-
i
I
p
*r*u, tf^rm^^f^
Iff
i
\ •
M
R
it
Jw\
l*K>*rv^WW '" Wv^u-w-^V/^^SrV^^TT
0
100 200 300 400 500
COLLECTED INTERFEROGRAM NUMBER
Figure 3
-------
ATTENUATION OF FREQUENCY RESPONSE
100 COEFFICIENTS
FIR
12 ORDER IIR
-140
750 800 850 900 950 1000 1050 1100 1150
WAVENUMBERS
Figure h
-------
DISCUSSION
DONALD GURKA: Can you tell us something about the range and payload of
your drone?
JOHN DiTILLO: Right now, the payload is about 25 pounds, and that includes
everything, the video cameras, the interferometer, the gas, the whole bit. Just the
electronic portion or the sensor portion of it has to be limited to about 25 pounds.
It just happens that the specification for the aircraft the contractor had. You can
build them as large as you want, and the military drones that are under
development have very large payload capacities. This just happened to be the one
that we fell upon. A little bit about the aircraft: it flies at about 100 knots with the
gas on board, it can fly for about an hour, and it has autopilot capabilities so you
can send it out on a pre-planned mission and have it fly lazy eights or whatever
over a specific area. Some of the efforts we have this year are to tie a global
positioning system into that, so you can not only get video information back,
which isn't very realistic from a military standpoint, particularly if you look at
the scenario we have now in Iraq where the ground features aren't very distinct.
Video is not going to tell you a whole lot. So, we think it would be a much better
idea if you actually had grid coordinates as well as a response out of the
algorithm. So, there's some more effort that's going into the aircraft itself as far
as its capabilities, but that just happened to be what was available at the time.
TOM PRITCHETT: I'm familiar with calibrating the active FTIR units. How
do you calibrate a passive unit?
JOHN DiTILLO: I don't know that much about the optical end of things. The
XM-21 had an internal black body calibration that it went through on start up.
Beyond that I don't know that much about it.
CHIP MILLER: Seems like from the days of show pair, low pair days, I
remember problems with confusing silicates and absorption of silicates with the
phosphoryl absorption of the organophosphonates that you're interested in. Is
that still considered a problem, especially with regard with the desert scenario
and the silica?
JOHN DiTILLO: That's the montmorillinite and kaolin problem. A lot of time
and effort was spent on that problem. That was a serious problem early on. As you
can imagine in a military scenario, false alarms can be devastating. And through
years and years of testing and refinement of the algorithm, the XM-21 is virtually
fool proof. The instrument has been trained to eliminate a lot of those early
problems with dust or compounds that are similar to nerve agents and pesticides.
Just from a military standpoint, if a unit were to get a false alarm out of an XM-
21, the first thing they would do is go into mob gear. As soon as a unit does that,
their fighting efficiency goes down to about 10%. So, you can imagine if your
enemy knows that it hasn't hit you with agent, and it's looking across the field
and you're in mob gear, you can imagine what kind of ramifications that's going
to have. So, the army goes through great pains to eliminate any kind of false
alarms due to dusts and dirts and burning tires, and that type of stuff. And the
instrument has been trained to eliminate those problems.
570
-------
USE OF WIND DATA TO COMPARE POINT-SAMPLE AMBIENT AIR VOC
CONCENTRATIONS WITH THOSE OBTAINED BY OPEN-PATH FT-IR
Ray E. Carter, Jr., Dennis D. Lane, and
Glen A. Marotz
Department of Civil Engineering
4002 Learned Hall, University of Kansas
Lawrence, KS 66045
Mark J. Thomas and Jody L. Hudson
U.S. EPA, Region VII
25 Funston Rd., Kansas City, KS 66115
ABSTRACT
The technique of open-path FT-IR spectrometry
is being used increasingly to measure VOCs in
ambient air. Since the FT-IR technique
produces a path-integrated concentration and
most other techniques produce point
concentrations, some method of interconversion
is often desirable. In the case of a plume
generated by a single point source, a solution
to the interconversion problem can be found
through the use of wind data. A quantitative
relationship was developed between wind
direction frequency and concentration. This
relationship was used to predict path-
integrated concentrations, given point
concentrations. The same principle was used to
predict point concentrations, given path-
integrated concentrations.
The interconversion technique involved the use
of one-minute means of wind direction as inputs
to a Gaussian dispersion model. The one-minute
concentration-to-emission rate (C/Q) ratios
produced by the model were integrated over the
sampling period of the test to yield a C/Q
ratio that was based on the wind directional
frequency distribution, rather than on the
overall mean wind direction. These integrated
C/Q ratios for selected points were used to
develop predictive methods for both point
concentrations and path-integrated
concentrations; this process is described in
detail within the body of the paper.
The interconversion technique was tested, using
data from simulated field tests during which
VOC releases were made. The VOC plumes
generated were monitored along a line normal to
the projected plume centerline, using the FT-IR
technique and also by collecting whole-air
samples in evacuated stainless steel samplers
for subsequent GC/FID analysis. During the
final test set, no FT-IR measurements were
made; an observed path-integrated concentration
was produced by using the mean of
concentrations from point samples collected
five meters apart along the path, thus
providing an evaluation of the technique free
of any bias that might exist between the two
analytical methods.
Correlations between the observed
concentrations for the point samples and the
corresponding integrated C/Q ratios were
assessed for each test and found to be
significant at the 0.1 level in most cases.
Although a bias between the two analytical
methods was seen, the predicted path-integrated
concentrations were strongly correlated with
the observed values. For the test set in which
only point samples were collected, excellent
agreement between the predicted and observed
path-integrated concentrations was seen. The
predictive method for point concentrations did
a good job of predicting both the location and
the magnitude of the highest concentrations
from each test, and reflected the general shape
of the concentration-versus-crosswind curve
well.
INTRODUCTION
The technique of open-path Fourier transform
infrared (FT-IR) spectrometry is being used
increasingly to measure volatile organic
compounds (VOCs) in ambient air. Depending on
the circumstances (nature of the source,
receptors affected, etc.) this technique may
complement, or in some cases replace, the
collection of multiple whole-air samples with
subsequent laboratory analysis.
571
-------
The FT-IR technique produces a concentration
integrated over the length of the path from
source to detector for a given compound,
whereas the whole-air method produces a
concentration for only those points sampled.
In order to use the data produced by the two
techniques optimally, some method of
interconversion is desirable.
Solution of the interconversion problem would
be difficult in many cases. However, in the
case of a plume generated by a single point
source with relatively constant emission
characteristics, a solution can be found
through the use of wind data. The location and
movement of the plume centerline (and therefore
the maximum concentrations) are determined by
the wind direction. Thus, the concentration of
plume-derived compounds found in point samples
should be primarily a function of the amount of
time that the wind blew from the source toward
those points during the sampling period.
If a quantitative relationship between wind
direction frequency and concentration can be
derived, that relationship can be used to
calculate the concentration at any point along
the path from the IR source to the detector.
The mean of those values for points closely and
evenly spaced along the path should then
provide an approximate path-integrated
concentration. The same principle can be
applied to the conversion of path-integrated
concentrations to point concentrations.
APPROACH
Sampling and AnaJLys.is_Framework_Overvi_ew
Plumes consisting of both single compounds and
mixtures were generated from a stack two meters
above the ground. Point samples were collected
downwind from the source along a line normal to
the projected plume centerline for subsequent
analysis. During the first three of the four
test sets performed, the plume was also
monitored along the same path using an open-
path FT-IR spectrometric method developed at
Kansas State University (1).
During the fourth test set, point samples were
collected at five-meter intervals along the
path; no FT-IR measurements were made. This
closely-spaced network of point samplers
provided a more detailed characterization of
the concentration-versus-crosswind distance
curve. It also provided an observed path-
integrated concentration based on the mean of
concentrations from those samples, allowing
further evaluation of the interconversion
technique free of any bias that may exist
between the two analytical methods.
Wind data collected during the sampling period
were then used in conjunction with a Gaussian
dispersion model to determine a calculated
concentration-to-emission rate ratio (C/Q) for
selected points along the path for each minute
of the test. The C/Q values for each point
were then summed over the sampling period of
the test to produce an integrated C/Q (or C/Q*)
for each point.
The above data and calculated values were then
used in the following ways, to test the
validity and usefulness of the conversion
technique:
1) Correlations between the observed
concentrations for the point samples and
the corresponding C/Q* values were
performed for each test.
2) The observed point concentrations and the
calculated C/Q* values were used to
predict a path-integrated concentration.
This value was then compared to the
observed path-integrated concentration for
the same time period.
3) Using the observed path-i n t e gr ated
concentration and the wind data, an
equation was developed to predict
concentrations at selected points. These
predicted values were then compared to
observed concentrations present in samples
collected at those points.
Did ivi dual _T es t _Runs
Four sets of field tests were performed, one on
each of the following dates: February 24,
1989; October 20, 1989; April 25, 1990; and
October 22, 1990. VOC releases were made from
a stack two meters above the ground in an open
field near Lawrence, Kansas. During the first
three test sets, the general spatial
relationship of the VOC source and the sampling
devices was as shown in Figure 1. Specific
parameters (distances, number of point samples
collected, etc.), in addition to the field
deployment used in the fourth test set, are
described in the succeeding paragraphs.
Figure 1
On February 2U, 1989, a 20-minute release of
toluene was made. The release was monitored
along a line normal to the projected centerline
of the VOC plume 50 meters from the source,
using the open-path FT-IR methodology developed
at Kansas State (1) and also by collecting
whole-air samples in evacuated stainless steel
samplers for subsequent GC analysis. Sampling
and analysis of whole-air samples followed
protocols developed at the University of Kansas
572
-------
S = VOC Source
O = Point Sampler
C = Centerline Sampler
FD = FT-IR Detector
Fs = FT-IR Source
Figure 1. Sampling Network
(2). Nine whole-air samplers were deployed at
ground level along the path, with a = 30x. The
FT-IR path length was 120 meters.
On October 20, 1989, six 30-minute releases
were made. Only 1 ,1 ,1-trichloroethane was
released during the first three of these tests.
During the final three tests, various mixtures
of the following six compounds were released:
n-pentane, methylene chloride, methyl ethyl
ketone, tert-butyl alcohol, 1,1,1-
trichloroethane, and toluene. The releases
were again monitored along a line normal to the
projected plume centerline 50 meters from the
source, using both the FT-IR and whole-air
techniques. Five whole-air samples were
collected at ground level during each test,
with a = 30x. The FT-IR path length was 100
meters. During the final two tests of this
set, the wind had shifted approximately 30
degrees, making it questionable whether the
plume centerline was within the sampling
network. Data from those tests are not
reported.
On April 25, 1990, ten 12-minute releases were
made. Mixtures of the following five compounds
were released during the first nine tests:
methylene chloride, methyl ethyl ketone, tert-
butyl alcohol, 1,1,1-trichloroethane, and
toluene. Only 1 , 1 ,1 -trichloroethane was
released during the final test. The plumes were
again monitored along a line normal to the
projected plume centerline, this time 40 meters
from the source, using both the FT-IR and
whole-air techniques. A minimum of five whole-
air samplers were deployed along the path
during each test at one meter above the ground,
with a = 20x during Tests 1-3 and a = 30x
during Tests 1-10. The IR path length was 50
meters during Tests 1-3, 200 meters during
Tests 4-6, and 100 meters during Tests 7-10.
On October 22, 1990, three 12-minute releases
were made and again monitored along a line
normal to the projected plume centerline 40
meters from the source. Fifteen whole-air
samplers were collected at ground level along
the path at five-meter intervals in each of the
the three tests, as shown in Figure 2. No FT-
IR measurements were made. Only toluene and
1,1,1-trichloroethane were released.
The network of samplers was shifted five meters
to the left for Test 3 to account for what was
perceived to be a slight wind shift. Samplers
were thus arrayed from 40 meters left to 30
meters right of the projected centerline during
this test.
S = VOC Source
O = Point Sampler
C = Centerline Sampler
Figure 2. Sampling Network, 10/22/90
573
-------
The following meteorological data were
collected during all releases: one-minute
means of temperature and relative humidity, and
one-minute means and standard deviations of
wind direction and wind speed.
EXPLANATION OF METHOD
Determination_of_C/Q_VaUi es_
An algorithm was written to calculate
integrated concentration-to-emission rate
ratios (C/Q») for selected points. As it is
now written, the algorithm can produce a C/Q*
value for up to ten points per execution. The
following inputs are required:
- wind direction means for each minute of
the test
- wind speed means for each minute of the
test
- overall standard deviation of wind
direction for the test
- location of the centerline point
(distance and direction from the source)
- location of the remaining points
selected (distance left or right of the
centerline point)
Inputs were used in conjunction with a
dispersion model to produce C/Q* ratios for
selected points, in the following manner:
1) For each minute of the test, x- and y-
2)
Observations
S = VOC source
C = centerline point
D = designated point
SC = 50 meters
DC = 25 meters
network centerline direction
180C
meanwind direction (•»)
B - 180° - 160° = 20°
160C
coordinates were determined (with the
source as the origin and the mean wind
direction for that minute providing the
direction of the x-axis) for each of the
selected points. Each of the points was
assumed to be on a line perpendicular to
the line from the source to the centerline
point and the appropriate trigonometry was
then performed. An example for one such
point is shown in Figure 3.
The standard deviation of the wind
direction was then used to determine the
stability class. Based on that class, and
on the x-coordinate for a given point, a
value for o and o was determined for
each point (3)- An example of this
process is shown below, using the point
discussed in part 1 and assuming a
directional standard deviation of 10
degrees and a wind speed of 5 m/sec.
- direction standard deviation of 10°
indicates stability class D
- for that stability class,
o = 0.08 x (1 + O.OOOIx)
0.08(55.5m)(1
14.43m
0.06 x (1 +• 0.0015x)
0.0001(55.5m))
-1/2
3.20m
Trigonometry
DA - DC - AC = DC - SC tanB
SA = SC/cose
AB = DA sing = (DC - SC tanB) sinB
x = SB = SA + SB
= SC/cosB + (DC - SC tanB) sinB
= 55.5 meters
DB
DA cosa - 6.4 meters
Figure 3. Determination of x- and y- coordinates
574
-------
3) These values for o and o are then used
in the dispersion equation to calculate
C/Q, as shown below, also for the point
discussed in parts 1 and 2.
C/Q = — ~
1
-(6.4)2/2(4.43)2
4)
2Tr(4.43~)(3.22)(5)'
-4
= 7.907 x 10
The C/Q values for each point were summed
over the sampling period for that test,
which yielded an integrated C/Q (or C/Q*)
for each point. C/Q* values for the seven
points at which point samples were
collected during Test 10 on April 25, 1990
are as follows:
Location
30°L
20°L
10°L
centerline
10°R
20°R
30°R
C/Q*
3.206 x 10
2.020 x 10
4.
x 10
4.112 x 10
4.016 x 10
8.565 x 10
2.840 x10
-4
-2
-2
-2
-3
-6
Correlation of Observed PointConcentrations
Correlations between observed concentrations
from point samples and corresponding C/Q*
values were assessed. This operation produced
a correlation coefficient (r) and an associated
probability (p). The latter value represents
the probability that a correlation coefficient
of this magnitude could occur if there were no
relationship between wind directional frequency
and concentration. For Test 10 on April 25,
1990, the following concentrations of 1,1,1-
tr i chloroet hane were found in the point
samples':
Location
30°L
20°L
10°L
centerline
10°R
20°R
30°R
Concentration (ppb)
215
1056
1163
1188
175
46
0.0
These concentrations produce a correlation
coefficient of 0.968 with the corresponding
C/Q* values for the same test (shown in part 4
of the preceding section), with a probability
of <0.001.
Prediction_gf__Path-I_ntegrated Concentrations
The observed point concentrations and the
calculated C/QX ratios were used to produce a
predicted path-integrated concentration for
each test, in the following manner:
1) A linear regression was performed, with
the model-derived C/Q* values used as the
independent variable and the observed
concentrations in point samples the
dependent variable, yielding a functional
relationship between wind direction
frequency and concentration. A regression
equation of this form (y - ax + b)
normally contains a non-zero intercept
value, which can produce relatively high
predicted concentrations corresponding to
extremely low C/Q* values (an unrealistic
situation), or even negative
concentrations if the intercept is
negative.
Two solutions to this problem were
attempted. The first was to constrain the
regression line to pass through the origin
(intercept = 0). This solution eliminates
the problems caused by the intercept,
although it does not produce as good a fit
between predicted and observed values.
The second solution was to perform a log-
log regression. The equation produced
generally yielded a slightly better fit
than the zero-intercept method, while
still allowing predicted concentrations to
asymptotically approach zero for
decreasing values of C/Q*. Shown below
are regression equations produced from the
4/25/90 (Test 10) data, using both methods
described in this paragraph. Also shown
are F-ratios, indicating the significance
of the regression, and predicted versus
observed values.
Zero-Intercept Method
C = 33804 (C/Q*)
F = 88.0 (p <0.001)
Location Observed C Predicted C
30°L
20°L
10°L
centerline
10°R
20°R
30°R
215
1056
1163
1188
175
U6
0.0
11
683
1652
1390
136
0.
0.0
575
-------
Log-Log Method
In C - 0.392 (ln(C/Q*)+8.28
F = 280 (p <0.001)
Location Observed_C Predicted_C
30°L 215 168
20°L 1056 852
10°L 1163 1204
centerline 1188 1124
10°R 175 452
20°R 46 41
30°R 0.0 0.3
2) Once the regression equation was derived,
it was used to predict concentrations at
evenly spaced points along the path.
Predicted concentrations (based on a log-
log regression) for the 4/25/90 test are
as follows:
Location Concentration Location Concentration
2)
50m L
45m L
40m L
35m L
30m L
25m L
20m L
15m L
10m L
5m L
centerline
0.0
0.0
0.2
2
17
97
359
818
1146
1236
1129
3) The mean of the
50m R
45m R
40m R
35m R
30m R
25m R
20m R
15m R
10m R
5m R
predicted
0.0
0.0
0.0
0.0
0.0
0.1
2
34
219
658
concen
for evenly spaced points produced a
predicted path-integrated concentration.
The predicted and observed path-integrated
concentrations are shown below for the
4/25/90 test.
observed path-integrated concentration =
185 ppb
predicted path-integrated concentration =
272 ppb
Prediction ^_Polnt_Concentrations
The observed path-integrated concentration and
wind data collected for the same time period
were used to produce predicted concentrations
for selected points along the path in the
following manner:
1) Wind data and a dispersion model were used
to produce C/Q* values for evenly spaced
points along the path, as outlined
earlier.
3)
A conversion factor was determined, based
on the observed path-integrated
concentration and the mean of C/Q* values
for evenly spaced points along the path,
using the following equations:
PIC
n
F x (l
i
(C/Q*)i)/n
and
(Cpt). = F x (C/Q»)1
and
where F = conversion factor
PIC = path-integrated concentration
(C/Q*) = integrated C/Q for the ith point
n = total number of evenly spaced
points
(C ). = predicted concentration at the
ith point.
Based on the data collected during Test 10
on 4/25/90,
F = 185/7.955 x 10~3
F = 23255.8
The conversion factor was used in
conjunction with C/Q* values to predict
concentrations for selected points, which
were then compared with observed
concentrations from samples collected at
those points. For the centerline point on
4/25/90 (Test 10), the predicted 1,1,1-
trichloroethane concentration is as
follows:
~2
C t = 23255.8 x 4.112 x 10
C - 956 ppb
Shown below are 1 , 1 , 1-trichloroethane
concentrations predicted for all points at
which samples were collected during the
4/25/90 test. Also shown are the observed
concentrations for those points.
Location
30°L
20°L
10°L
centerline
10°R
20°R
30 °R
Concentration (ppb)
Predicted Observed
8
470
1137
956
93
0.
0.0
215
1056
1163
1188
175
46
0.0
576
-------
RESULTS, DISCUSSION
Correlation of Observed Point Concentrations
with C/Q* Values "
Table 1 shows correlations (and associated
probabilities) between observed point
concentrations and corresponding C/Q* values
for each test. For releases that consisted of
mixtures of two or more compounds, only 1,1,1-
tr ichloroethane (1,1,1-TCA) and toluene
concentrations were used in the analysis.
In all but three tests, the correlation
coefficients produced were associated with
probabilities of less than 0.1. In at least
one of those three cases, Test 3 on 10/20/89,
it appears that additional valid data points
would improve the correlation. These
statistics would support the belief that there
is a strong relationship between wind direction
frequency and concentration.
f££^i£t?S_Y.ers!i3 Obse_r_ved_P£th-Integr a ted
Concentrations
Table 2 shows observed pa t h-i nt egr a t ed
concentrations for all field tests. Again,
only results for 1,1,1-trichloroethane and
toluene are reported. Also shown are the
Table 1. Correlation of Concentrations with C/Q* Values
corresponding predicted path-integrated
concentrations, based on observed point
concentrations and wind data. The values
reported are those produced using log-log
regression of concentration against C/Q*. At
the time of this writing, FT-IR data was not
available for 2/21/89.
It should be noted that observed path-
integrated concentrations reported for the test
set of 10/22/90 were produced by using the mean
of concentrations from point samples collected
five meters apart along the path. Values of
predicted path-integrated concentration for
this data set were produced using log-log
regression of concentration against C/Q* for
the following five of the fifteen points at
which samples were collected: the projected
centerline point, points 10 meters left and
right of that point, and points 20 meters left
and right of that point.
Excellent agreement is seen in the observed and
predicted values from the 10/22/90 test set,
with the largest percent difference being 2.5%.
Values shown for the 10/20/89 and 4/25/90 test
sets indicate the possibility of a bias between
the two analytical methods, especially in the
analysis of 1,1 ,1-trichloroethane, making an
evaluation of the prediction technique more
difficult in this case. In order to gain some
Date
2/24/89
Stability Number of Valid Correlation Coef^ and Probabi_lUy_
Test _C_lass_ Data Points 1 ,1,1rfCA"vs C/Q* fofuene'vs ~c7Q*~
10/20/89
10/20/89
10/20/89
10/20/89
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
1
2
3
4
1
2
3
4
5
6
7
8
9
10
B
B
B
B
D
D
C
C
C
D
D
C
C
D
10/22/90 1 B
10/22/90 2 B
10/22/90 3 C
5
4
4
5
5
5
5
5
5
5
5
5
5
1
15
15
15
NR
r - 0.889 p<0.01
r = 0.900
r - 0.970
r = 0.885
r = 0.868
0.946
0.943
0.968
0.853
0.878
0.960
0.898
0.663
0.739
r = 0.968
r = 0.942
r - 0.968
r - 0.918
r
r
r
r
r
r
r
r
r
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
p<0
.05
.05
.2
.1
.02
.02
.01
.1
.1
.01
.05
.4
.2
.001
.001
.001
.001
NR
NR
NR
r -
r =
r =
r «=
r -
r -
r =
r =
r -
r •
r -
r =
r »
0
0
0
0
0
0
0
0
0
0
0
0
0
.858
.940
.944
.982
.895
.887
.916
.910
.675
.759
NR
.942
.969
.954
p<0.1
p<0.02
p<0.02
p<0.01
p<0.05
p<0.05
p<0.05
p<0.05
p<0.4
p<0.2
p<0.001
p<0.001
p<0.001
NR - Compound not released during this test
577
-------
Table 2. Predicted vs Observed Path-Integrated Concentrations
1,1,1-TCA Cone, (ppb) Toluene Cone, (ppb)
Date Test Observed Predicted Observed Predicted
2/24/89
NR
10/20/89
10/20/89
10/20/89
10/20/89
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
4/25/90
10/22/90
10/22/90
10/22/90
1
2
3
4
1
2
3
4
5
6
7
8
9
10
1
2
3
185
181
199
78
56
138
38
16
34
7
36
75
18
185
334
277
282
284
297
307
106
85
197
44
37
62
10
71
107
25
272
340
280
278
NA
NA
35
85
233
20
30
42
18
116
146
333
284
288
NR
NR
NR
NR
235
64
31
142
204
12
45
55
28
76
118
341
279
282
NR - Compound not released during this test
NA - FT-IR data not available
insight into the performance of the technique,
correlations between observed and predicted
values were assessed for the 4/25/90 test set.
For 1 , 1 , 1-trichloroethane, the correlation
coefficient was 0.994, with an associated
probability of much less than 0.001. For
toluene, the correlation coefficient produced
was 0.915, with an associated probability of
less than 0.001. These results indicate that
the technique presented for predicting path-
integrated concentrations is potentially a
sound one and warrants further study.
Predicted vs. Observed Point Concentrations
Table 3 (on following pages) shows observed
concentrations for all point samples collected
in the four test sets.
Also shown are the corresponding predicted
concentrations for the points, based on
observed path-integrated concentrations and
wind data. At the time of this writing, FT-IR
data was not available for 2/24/89.
As seen in Table 3, the technique presented for
predicting point concentrations does a
reasonably good job of predicting the general
shape of the concentration-versus-crosswind
distance curve, although concentrations 15 or
more degrees from the centerline are generally
predicted less accurately than are those nearer
the centerline. Given the bias present between
the two analytical methods, both the location
and the magnitude of the highest concentrations
from each test are predicted quite accurately.
CONCLUSIONS
The following conclusions are warranted, based
on the data presented:
1) The use of one-minute means of wind
direction as inputs to a Gaussian
dispersion model produce concentration-to-
emission rate ratios that are strongly
correlated with observed concentrations.
2) The predicted p a t h - i n t e g r a t e d
concentrations show good agreement with
the observed values, given the bias seen
between the two analytical methods.
3) The predicted point concentrations reflect
the general shape of the concentration-
versus-crosswind distance curve well. The
location and the relative magnitude of the
highest concentrations from each test are
predicted accurately.
4) The principles underlying the
interconversion methods are sound, and the
methods themselves warrant further testing
and development.
578
-------
Table 3- Predicted vs Observed Point Concentrations
Sample 1,1,1-TCA
Location Obs.
Cone.
Pred.
Toluene
Obs.
Cone.
Pred.
2/24/89 - Test 1
30° L
20° L
10° L
5° L
centerline NR
5° R
10° R
20° R
30° R
10/20/89 - Test 1
30° L 67
15° L 71 1
centerline 966
15° R 749
30° R 65
10/20/89 - Test 2
30° L 223
15° L 884
centerline X
15° R 283
30° R 47
10/20/89 - Test 3
30° "L "x
15° L 925
centerline 1043
15° R 241
30° R 43
10/20/89 - Test 4
30° L" ~ 48
15° L 272
centerline 278
15° R 297
30° R 20
4/25/90 - Test 1
20° L 33
10° L 182
centerline 295
10° R 134
20° R 44
4/25/90 - Test 2
20° L 78
10° L 525
centerline 737
10° R 277
20° R 28
206
336
405
304
34
181
376
386
232
88
236
372
410
294
68
65
98
141
176
57
6
170
191
63
0.8
8
286
715
73
0.1
214
891
1009
718
364
351
333
114
28
27
149
148
158
11
12
65
106
50
19
68
357
502
193
23
NA
NR
NR
NR
NA
4
106
119
39
0.5
5
176
440
45
0.1
Sample 1,1,1-TCA
Location Obs.
4/25/90 - Test 3
20° L 29
10° L 88
centerline 123
10° R 83
20° R 16
4/25/90 - Test 4
30° L 130
15° L 150
centerline 346
15° R 90
30° R 17
4/25/90 - Test 5
307 L 179
15° L 439
centerline 381
15° R 130
30° R 4
4/25/90 - Test 6
30 °T " 20
15° L 66
centerline 116
15° R 18
30° R 2
4/25/90 - Test 7
30° L 37
15° L 310
centerline 329
15° R 22
30° R 6
4/25/90 - Test 8
30° L 39
15° L 746
centerline 308
15° R 111
30° R 12
4/25/90 - Test 9
30° L 96
15° L 95
centerline 49
15° R 5
30° R 4
Cone.
Pred.
8
71
135
63
14
3
34
179
83
0.3
21
220
325
82
0.2
0.0
27
93
7
0.0
0.8
67
161
2
0.0
5
243
385
88
0.9
24
106
37
0.6
0.0
Toluene
Obs.
117
374
613
396
79
26
57
11 1
34
8
149
293
284
100
13
102
402
561
94
16
22
118
130
12
7
31
494
211
73
3
463
478
228
23
21
Cone.
Pred.
48
438
830
387
84
4
42
224
103
0.4
19
194
286
73
0.2
0.0
161
559
41
0.0
0.4
34
81
1
0.0
8
376
595
136
1
195
855
301
5
0.0
579
-------
Table 3. Predicted vs Observed Point Concentrations (cont'd)
Sample 1
Location
,1 ,1-TCA Cone.
Obs. Pred.
Toluene
Obs.
Cone.
Pred.
Sample
Location
1,1,1-TCA
Cone.
Obs. Pred.
Toluene
Obs.
Cone.
Pred.
4/25/90 - Test 10
30° L
20° L
10° L
centerline
10° R
20° R
30° R
10/22/90 -
35m L
30m L
25m L
20m L
15m L
10m L
5m L
centerline
5m R
10m R
15m R
20m R
25m R
30m R
35m R
215
1056
1163
1188
175
46
0.0
Test 1
6.1
21
61
1 44
479
909
789
691
607
574
482
232
8.0
2.8
1.8
8
470
1137
956
93
0.2
0.0
5.7
21
68
194
458
818
1027
911
650
429
249
116
43
13
3.4
5.7
22
63
148
486
920
785
686
591
575
470
224
9.9
2.8
4.5
NR
5.7
21
68
193
457
816
1024
909
648
428
248
116
43
13
3.4
centerline
5m R
10m R
15m R
20m R
25m R
30m R
35m R
10/22/90 -
40m L
35m L
30m L
25m L
20m L
15m L
10m L
5m L
centerline
5m R
10m R
15m R
20m R
25m R
30m R
610
479
553
333
229
80
25
42
Test 3
0.6
1.7
1.5
1.5
33
59
189
395
740
826
778
621
391
122
64
520
456
416
335
217
113
49
18
0.0
0.0
0.0
0.1
2.0
32
200
527
900
1127
777
404
188
56
9.8
592
506
561
344
237
83
27
44
0.5
0.6
0.7
1 .0
34
61
198
420
768
846
787
627
379
123
66
533
468
427
343
222
116
50
18
0.0
0.0
0.0
0.0
2.0
33
204
538
919
1151
793
412
192
57
10
10/22/90 - Test 2
35m
30m
25m
20m
15m
10m
5m
1.2
17
24
89
342
641
690
5.5
24
85
228
448
620
619
1.3
19
28
90
358
643
725
5.6
25
87
234
459
636
635
NR - Compound not released during this test
NA - FT-IR data not available
X - Whole-air sampling error, no data
REFERENCES
(1) Spartz, M.L., M.R. Witkowski, J.H.
Fateley, J.M. Jarvis, J.S. White, J.V.
Paukstelis, R.M. Hammaker, W.G. Fateley,
R.E. Carter, Jr., M. Thomas, D.D. Lane,
G.A. Marotz, B.J. Fairless, T. Holloway,
J.L. Hudson, D.F. Gurka. "Evaluation of a
mobile FT-IR system for rapid volatile
organic compound determination, Part I:
Preliminary qualitative and quantitative
calibration results," American
Environmental Laboratory, v. 1, no. 2,
November, 1989, PP 15-30
(2) Marotz, G.A., D.D. Lane, R.E. Carter, Jr.,
R. Tripp, J. Helvig. "Preliminary results
from a rapid deployment field study of
heavy gas detection and dispersion using a
whole-air technique," 87-103-7, EPA/APCA
Symposium on Measurement of Toxic Air
Pollutants, Raleigh, NC, 1987.
(3) Hanna, S.R., G.A. Briggs, R.P. Hosker,
Jr., "Gaussion plume model for continuous
sources , " H_£rid_book_on_A tmospheric
Diffusion , Technical" Information ~Center7
U.S. Department of Energy, 1982, pp 25-31!
580
-------
DISCUSSION
DONALD GURKA: How certain are we that there is no physical gradient across
the plume? That is, we're not looking at part gas and part aerosol?
RAY CARTER: We have found that, based on the studies that we've done over
a five-year period, the plume is well dispersed as it comes out of the stack. It is
in fine liquid droplets as it comes out of the stack, but the droplets are so dispersed
that they vaporize almost immediately. We have data from other tests that would
support this.
DONALD GURKA: I guess my point is that the physical gradianl might have
a differential effect on the open path concentration approach versus the canister
concentration approach.
RAY CARTER: I think I see your point, and there did seem to be a bias between
the two methods. And that is one good possibility if it is not all in vapor state. We
believe that it is, but since there is a bias it's still worth looking into.
TOM PRITCHETT: Since in essence you're doing a controlled release of
essentially a known emission rale, and in the case of your long path, open path
monitoring, you're calculating emission rate versus concentration ratio. Have
you used the open path monitoring to calculate ancxperimental emission rate and
compared that against the generation rate of vapors? And also, use, let's say, the
transect method of data reduction on the canister data, to also calculate an
emission rate and see which one of those methods wasgiving you an experimental
emission rate that was closest to your generation rate?
RAY CARTER: We did not have real good control on the FT1R methodology.
It was primarily done by the people from Kansas State, and so we haven't really
done anything with their data. I think your point is a good one, though. We could
use our data and the method that you suggest to see if we accurately predict the
emission rate that we did measure.
TOM PRITCHETT: Just as a follow up to that, if you looked at both the canister
data and the open-path monitoring data, since you have a consistent bias, you
might be able to resolve who's causing the bias by checking the two experimental
emission rates versus your generation rate.
RAY CARTER: Yes, I would agree with that. And if we can obtain all of data
from Kansas State that would be a good test to undertake.
ERNIE TUAZON: Would il be better to use a dye compound, like something
that is more evenly distributed in the atmosphere which can be measured by the
FTIR while it's doing its measurements, and also being sampled by your
canister? Like a compound as simple as methane in the air—it's almost relatively
constant. Even nitrous oxide or carbon dioxide. This will be in the data already.
You will have sampled it, and the FTIR will have measured it already. So, if
you're looking for that bias then, part of your answer may lie there.
RAY CARTER: You're suggesting using compounds that do naturally exist?
Then you would not have the gradient across the path?
ERNIE TUAZON: That's correct, this is not a test of your model, but it tells you
which one is producing a lower reading in that general direction.
RAY CARTER: Yes, since our main purpose was to test this model, then we
preferred to actually have a gradient across the path, but I see your point. That
would be another good method of determining which is the biased method. That
was not really our intent. When we discovered that there was a bias, the reason
we did the fourth test set was to Jry to test the method independent of the bias.
ERNIE TUAZON: I am aware of that. Also, one thing that might also affect your
comparison is the way you sample with the FTIR. There is a dead time between
FTIR measurements if you're calculating rightafteryou collect the interferograms.
In other words, there will be segments of interferograms being collected and then
dead time while you're calculating, and then you again collect interferograms.
In the canister, you're continuously sampling while that's occurring, aren't you?
RAY CARTER: I'm not sure I understand your question, but if I do, I believe
that was taken care of by the sequencing of the sampling. We attempted to
cooperate as much as possible with the people doing the FTIR measurements,
merely adjusted our sampling to fit whatever schedule they preferred.
DONALD GURKA: Yes, it seems to me that the RTPGroup with, I think that's
Bill McClenny, also saw this negative bias, but it seems to me that the bias was
within the combined experimental error. Is that correct?
BILL McCLENNY: The tests that we did were in the Delaware Site Program.
And the tests there were done with a plume that was originating from a nearby
industrial plant, and consisted of two primary emissions, paradichlorobenzene
and chlorobenzene. All tests that we did were by moving a canister along the path
next to the path of the augus — the line of sight for the beam. We had the system
set up with a source receiver at one end and a retroreflector at the other end. We
were carrying the canisters back and forth between those two locations. By
moving over a period of one-half hour, we gel an integrated canister sample, or
a sample that's integrated over time. And then the spectra from the Fourier
Transform System, were co-added over the same period of time. The two were
compared. The comparison was based on a common standard, i.e., the GC/MS
standard that was used to look at the paradichlorobenzene and the chlorobenzene,
also used for the FTIR System. By using a common standard we had a common
basis on which to compare, and we compared those measurements directly. And
those measurements were very close with paradichlorobenzene, even though we
were depending on a plume that was dispersing from a point source over which
we had no control. But, in our case we had to locate it at the right position, which
was an inconvenience. So, this type of comparison in which you have a control
source has advantages. For us, we were in the field. We had to locate downwind
of the source, and so our efficiency of taking these comparisons was reduced
because we had to wait for the right experimental conditions.
DONALD GURKA: But, the negative bias was within the combined experimental
error for the chemistry of these compounds?
BILL McCLENNY: We didn't have any bias that was discernable for the
paradichlorobenzene. But, we were dealing with concentrations that varied from
150 ppb down to about 11 ppb. And over that range, because paradichloroben-
zene has a very high absorption coefficient, we can see it easily with the FTIR
system, and therefore we had, I think, a good comparison.
581
-------
REMOTE DETECTION OF ORGANICS USING FOURIER TRANSFORM INFRARED SPECTROSCOPY*
Jack C. Demirgian and Sandra M. Spurgash
Analytical Chemistry Laboratory
Chemical Technology Division
Argonne National Laboratory
9700 South Cass Avenue
Argonne, IL 60439
(708)972-6807
ABSTRACT
Fourier transform infrared (FTIR) spectroscopy
is an ideal technique for remote detection of
organic emissions. There is an atmospheric
window in the 1200 to 800 cm"1 region, which
corresponds to the "fingerprint" region for
organic molecules. Virtually all organic
molecules have a unique absorption/emission
pattern in the fingerprint region. A remote-
passive FTIR relies on ambient emission of
infrared energy from organics to obtain
spectra. The instrumentation consists of inlet
optics, an interferometer, a mercury cadmium
telluride (MCT) detector, and an on-board
computer. The transportable unit measures 40
cm by 50 cm and has been used to collect data
while mounted on a helicopter or ground
vehicle. Through the use of this FTIR combined
with least squares software, it is possible to
analyze qualitatively and quantitatively for
organic vapors from either the air or ground.
The data presented will include quantitative
releases of common organics present in
incinerator stacks, hazardous wastes, and
illegal laboratories. Data will be presented
for pure compounds, mixtures, and target
analytes in the presence of interfering
compounds. The sensitivity, reproducibility,
and the potential of the technique will be
discussed.
INTRODUCTION
The emission of organic vapors is a concern for
environmental, health, safety, and regulatory
reasons. Sources of organic vapors include
industrial leaks, incinerator emissions, motor
vehicle exhausts, evaporation from contaminated
areas, leaking storage tanks, petroleum
refineries, and even illegal drug laboratories.
The EPA-certified procedures for organic
emissions consist of sampling a fixed amount of
air through a sampling apparatus such as a
volatile organic sampling train (VOST), which
traps the organics. The sample is then
transported for gas chromatographic (GC) or gas
chromatographic/mass spectroscopic (GC/MS)
laboratory analysis. Results are available in
a period of weeks or months. The entire
procedure is costly and time consuming. Also,
the GC or GC/MS analysis is a one-time
procedure. If the concentration of organics is
outside the acceptable instrumentation limits,
the sample cannot be reanalyzed.
Remote-passive FTIR offers the potential to
detect, identify, and monitor emissions in real
time in the field. FTIR spectroscopy is ideal
for remote detection. There is an atmospheric
window in the 1200 to 800 cnr1 region, which
corresponds to the "fingerprint" region for
organic molecules. Virtually all organic
molecules have a unique absorption/emission
pattern in the fingerprint region. Through the
use of FTIR combined with classical least
squares (CLS) or partial least squares (PLS)
software, it is possible to analyze
qualitatively and quantitatively for organic
emissions from either the air or ground.
The FTIR consists of infrared optics,
interferometer, MCT detector, on-board
computer, and external data collection system.
The remote FTIR used to collect the data
presented in this report was designed by the
U.S. Army and is designated as the XM21. It is
extremely rugged and can collect data mounted
on a helicopter or tank or ground mounted. It
can be programmed for target analytes and will
set off an alarm as soon as they are detected.
583
-------
The instrumentation emits no energy but detects
the natural emissions of organics due to the
difference in temperature of the organic
molecule from the background. For example,
organics emitted from an incinerator will be
thermally warm relative to the sky or ground
near the stack. A solvent exiting from an open
window or exhaust pipe will be warmer or colder
than the building from which it is emitted. On
a sunny day when the background is warmer than
the solvent vapor, a standard transmittance
spectrum is obtained. During the evening when
the background is cooler than the solvent
vapor, an emission spectrum is obtained.
This paper will focus on two potential
applications for remote FTIR spectroscopy: (1)
the monitoring of incinerator emissions and (2)
the detection of solvents emitted from a
building, such as an illegal drug laboratory or
production facility in which solvents are used.
RESULTS AND DISCUSSION
Data are presented for three pure liquids,
methanol (MEOH), chloroform (CHC13), and carbon
tetrachloride (CC14), to determine their
detection levels under laboratory conditions.
The liquids were released in front of a
blackbody that was maintained at 40°C. The
liquids represent common laboratory solvents,
two of which (CHC13 and CC14) are also
principal organic hazardous components (POHCs)
monitored in incinerator emissions.
1. Analysis of Pure Components
Methanol data were collected at flows
corresponding to concentrations of 3.7, 7.6,
13.9, 27.2, and 34 ppm-m. The IR spectra for
the data are presented in Fig. 1. The
quantitative results are presented in Table 1.
EXPERIMENTAL
1. Method
In a typical experiment, the distances from the
target (brick wall, blackbody, field release)
to the FTIR were accurately measured. A series
of releases of single components and mixtures
was performed using a vaporizer designed at
this facility. The vaporizer was capable of
converting a liquid flow to vapor, which was
released in front of the appropriate
background. The flow from the vaporizer was
determined using a hot wire anemometer. The
design of the vaporizer is described
elsewhere.x
2. Concentration Units
Data are reported in concentration-pathlength
Units of ppm-m. For example, a concentration-
pathlength release of 1 ppm-m is equivalent to
a release of 1 ppm over a width of 1 m. Our
release width was 10 cm, the width of the
vaporizer. Hence, when we release 10 ppm over
a width of 10 cm (0.1 m), the pathlength
concentration is 1 ppm-m. A 1 ppm-m release is
equivalent to a concentration of 1 ppm emitted
from a stack 1 m in diameter while the remote
FTIR is collecting data across the plume. The
parts per million level for the data presented
herein was determined by converting the liquid
flow to cubic centimeters of vapor and dividing
by the cubic meters of air released.
Table 1. Quantitative Data for MEOH Using a
Blackbody Background. Concentration-
pathlength units are ppm-m.
Cone.
34.0
27.2
13.9
7.6
3.7
Detected
2 std
STD
27.7
15.7
9.9
STD
Detected
3 std
STD
27.6
STD
8.8
STD
Agreement is excellent for the 27.2 ppm-m
sample, with deviation less than 2%. The
deviation was less than 13% for the 13.9 ppm-m
sample. However, when only two standards were
used, the error for the 7.6 ppm-m sample was
30%. This error decreased to 16% when a third
calibration standard was used.
Both CHC13 and CC14 are of special interest
because they are monitored during an
incinerator trial burn. Monitoring them in
real time enables the determination of on-
stream destruction removal efficiency (ORE).
On-stream ORE determination would eliminate the
need for a trial burn.
CHC13 and CC14 are strong infrared absorbers in
the 800-700 wavenumber region. This is beyond
the optimum region for the detector in the
XM21. Hence, the sensitivity of the instrument
is lower for detecting these two components
than would be the case with a detector
optimized in this region.
584
-------
Analysis of CHC13 consisted of five different
flows corresponding to concentration
pathlengths of 2.87, 5.12, 6.49, 7.85, and 8.87
ppm-m. The first and last values were used for
the calibration curve. The IR spectra obtained
are shown in Fig. 2. All absorbances in the
region 830-720 wavenumbers were used in
quantitation. The quantitative results are
shown in Table 2.
Table 2. Remote Detection of Chloroform Using a
Blackbody Background. Concentration-
pathlength units are ppm-m.
Cone.
Detected
X Error
5.12
6.49
7.85
4.85
6.15
7.10
5.3*
5.2
9.6
The relative intensity of the two peaks
associated with chloroform are clearly seen.
The absorbance in the 1220 wavenumber region
was not used for the data calculations shown in
Table 1. With a two-point calibration curve,
the percent error was within 10%. The
intensity of the 2.89 ppm-m absorption is
sufficiently strong to demonstrate sensitivity
in the high parts per billion concentration
range.
Data for CC14 showed similar sensitivity. Five
different flows of liquids were analyzed, which
correspond to concentration pathlengths of
3.47, 4.33, 5.77, 6.92, and 8.36 ppm-m. The
first and last values were used for the
calibration curve. The IR spectra obtained are
shown in Fig. 3. All absorbances in the region
810-784 wavenumbers were used in quantitation.
The quantitative results are shown in Table 3.
The quantitative data are accurate to within
10%, as were the chloroform data. The
detection levels are similar. The absorption
of CC14 is much sharper than that for CHC13.
The relatively low resolution of the
instrumentation gives the CC14 its sharp
features. At 2 wavenumber resolution, the fine
structure of most organic absorbances becomes
evident. At 4 wavenumber resolution, the fine
structure is lost, and the absorption degrades
to a curve or straight lines. The 4 wavenumber
resolution of the equipment employed in this
preliminary study limits the ability of the
software to identify and align peak absorbances
by their fine structure. The absorbance in the
780-760 region is background carbon dioxide.
This absorbance is also present in, and
overlaps with, chloroform. Higher resolution
data would show the carbon dioxide as sharp
bands superimposed on the chloroform
absorbance.
The laboratory work with CHC13 and CC14
demonstrated the potential of FTIR for remote
detection of organics. The next phase
consisted of determining the efficiency of the
instrumentation in the field. An experiment
was set up to detect MEOH released in front of
a brick wall. This experiment simulates the
detection of organics emitted from production
facilities or illegal drug laboratories. Data
collection is made more difficult because the
temperature of the brick wall and, hence, the
intensity of infrared energy being emitted are
changing during the day.
Data collected for MEOH are presented. The
three MEOH flows used correspond to
concentrations of 8.2, 18.3, and 27.8 ppm-m.
The spectra for these three concentration
ranges are shown in Fig. 4. Because the brick
wall was cooler than the released vapor
emission, spectra were obtained. The
concentration of the second sample was
calculated to be 22.5 ppm-m, or approximately
23% above the actual value.
Table 3. Remote Detection of Carbon
Tetrachloride Using a Blackbody
Background. Concentration-
pathlength units are ppm-m.
Cone.
4.33
5.77
6.92
Detected
3.92
6.28
7.58
Error
9.5Z
8.8
9.5
2. Analysis of Mixtures
A critical issue in demonstrating the potential
of remote FTIR is the ability of the
instrumentation to function in complex
environments. The technique must be able to
identify and quantify components in mixtures
under difficult and changing backgrounds.
Usually, components in mixtures absorb infrared
radiation at different energies. The
difference was readily observed in a simple
experiment in which 10 uL of MEOH and 5 uL of
ethyl ether were injected into an evacuated 10
cm cell placed in front of a blackbody
background. The spectral data are presented in
Fig. 5.
585
-------
a. Laboratory Release
Qualitative and quantitative data were obtained
for a mixture of CHC13 and CC14. These tvo
analytes have partially overlapping absorbance
peaks. This work simulates monitoring two
POHCs being emitted from an incinerator. The
blackbody temperature initially was set at 41°C
and allowed to slowly increase to 44° C by the
end of the measurements. This changing
background better simulates background
conditions found in an actual remote situation,
where data collection begins in the morning and
as the day progresses the temperature
increases.
Data were collected at six concentrations:
0.40, 0.81, 1.45, 2.02, 3.03, 4.04 for CHC1,
and 0.34, 0.68, 1.23, 1.71, 2.56, 3.41 for
CC14. CHC13 and CC14 were mixed 50:50 by
weight. The lowest concentration level (<0.5
ppm-m for each analyte) was below threshold
detection level and is not replotted. The IR
spectra of the other five solutions are shown
in Fig. 6. The quantitative data are shown in
Table 4. The 820-784 wavenumber region was
analyzed for CC14 and the 784-720 wavenumber
region was used for CHC13.
DEM in MEOH: 0.67, 0.98, 1.12 ppm-m DEM in
6.06, 8.82, and 9.72 MEOH, respectively. This
was immediately followed by three more flows,
which contained only MEOH: 5.93, 9.88, and
19.69 ppm-m. The objective was to determine if
the totally overlapped DEM could be identified
and quantified in the first three mixtures and
not misidentified (false positive) in the last
three flows. Figure 7 shows the IR spectra of
pure DEM, pure MEOH, and the mixture of both
components collected remotely. One cannot
visibly detect the presence of DEM in the
spectra of the mixture. The quantitative data
are presented in Table 5. The 1090-1000
wavenumber region was analyzed for DEM, and the
1100-975 wavenumber region was used for MEOH.
Table 5. Remote Detection of a Mixture of MEOH and DEM.
Concentration-pathlength units are ppm-m.
Cone.
8.82
5.93
19.69
MEOH
Detected
8.30
7.11
16. 46
Error
5.9X
19.9
16.4
Cone.
0.98
0
0
DEM
Detected
0.91
ND
0.17
Error
7.1Z
Table 4. Remote Detection of a Mixture of CHC13 and CC1,.
Concentration-pathlength units are ppm-m.
Cone.
0.40
1.45
3.03
CHClj
Detected
0.70
1.17
3.06
Error
75*
19
1
Cone.
0.34
1.23
2.56
CC1,
Detected
0.59
1.00
2.58
Error
74*
19
1
The second, fourth, and sixth samples were used
as calibration standards. The software was
able to identify the CHC13 and CC14 and
quantify their concentrations with the
calibration curve. The large deviation for the
lowest level is not unexpected. One cannot
visually identify CHC13 and CC14 in this
spectrum. The capability of the software to
identify the analyte under these conditions is
encouraging. The steadily increasing
background temperature did not result in a
degradation of the data.
The absorbances of CHC13 and CC14 only
partially overlap. It is necessary to study a
system in which the absorbances of both
components completely overlap. A diethyl
malonate (DEM) and MEOH mixture was studied at
six different flows. First, three flows were
studied, which contained a low concentration of
The first, third, and fifth samples were used
as standards. The software was able to
correctly quantify the low-concentration DEM
sample. The first pure MEOH sample was also
correctly identified, although the quantitative
data showed significantly more error. The
highest concentration MEOH sample, 19.69 ppm-m,
showed a low concentration of DEM present
(false positive). However, the concentration
detected was below the threshold detection
level of DEM.
b. Field Release
Field data were obtained at Aberdeen Proving
Ground. The XM21 was placed approximately 500
ft (~200 m) from the region where SF6 was
released. The angle of view, low sky
background, is the most difficult to work with
because of the infinite pathlength and greater
amount of atmospheric pollutants.
A methanol-DEM mixture was released while an
SF6 release was in progress. The raethanol
peaks were observed as emissions because they
were released at 42°C, which was above ambient
temperature. The SF6 was observed as an
absorbance spectrum because it was released
from a pressurized tank and the gas was below
ambient temperature.
586
-------
The spectral data are presented in Fig. 8, and
the quantitative data are presented in Table 6.
The spectral data show the steadily increasing
concentration of the methanol-DEM spectral
features and the reduction of the SF6
absorption as the gas disperses. The primary
band for DEM is completely overlapped by the
MEOH emission (Fig. 8). The secondary bands at
1200-1150 wavenumbers are readily observed.
Table 6. Methanol-DEM Release with SF6 Dispersion
in Lov Sky Background. All concentrations
are in ppm-m.
MEOH
Cone. Released Detected Z Error
4.6
11.7
17.4
23.5
STD
8.2
STD
25.5
29.9
-
9.4
DEM
Cone. Released
Detected X Error
2.2
5.8
8.5
11.6
STD
3.9
STD
12.5
32.8
_
7.8
The quantitative data in Table 6 were obtained
using only two standards. The percent error
was approximately the same as that obtained for
pure methanol. However, the analysis was more
difficult because only the spectral range of
1125-975 wavenumbers was used. The DEM and
methanol completely overlap in this region. No
pure components were entered into the
calibration file. Hence, these data indicate
that the analysis of mixtures is no more
difficult than the analysis of pure components.
SUMMARY AND CONCLUSION
This study has demonstrated that remote
infrared detection is a precise and reliable
technique for monitoring organic emissions.
The equipment is capable of detecting SF6
releases at 500 ft (-200 m) and low
concentrations of pure components and mixtures
released in the environment. Quantitation was
within 30% for these releases. Mixtures were
no more difficult to analyze than pure
components.
Several areas still must be addressed. The
limitation of the PLS software is the large
number of spectra required to reduce
quantitation error. For the data presented
here, it was not possible to obtain the number
of spectra that would reduce the quantitation
error. The classical least squares method
requires fewer library data and should improve
data quality. We hope to expand this work to
include advanced signal processing using
digital filtering in the time domain so that
variations in background and the need for
calibration spectra are eliminated. A
limitation of the existing equipment is the
lack of front-end optics, which are required
for analysis at distances of 1 km.
BIBLIOGRAPHY
1. J. C. Demirgian and S. M. Spurgash,
"Remote Detection of Chemical Agents by
Infrared Spectroscopy," ANL/ACL-90/1
Progress Report 1989.
*Work supported by the U.S. Department of
Energy under Contract W-31-109-Eng-38 and the
Chemical Research Development and Engineering
Center.
I I 00 10IIII
IfAVENUMDEUS
I in I. llctnolu III S|h.-clri\ i.f MROII ni»lainr,.'7'-', am! :n ,.|.iii..ii.
KKMOTK IIKTK*-Mi IN 111-' i;ill.OKOI--(iKM
I ;•(!() I Mill IIIIIO !l(l(l IIIKI
1VAVKNUM OEKS
iK 2 llmiiol- Ill Spi-Un ..fCHCI, OliUliml ll.lnu .. Ill.u-Llulr |]i..k|[r..ii,,il. C.»«-onlc»l»i
..i run-, w,-i< ».ii. r,.\,11.5, T.ti, ..n-i «•> ,T...,.
587
-------
RKMUTK MKI'Kt "I'll IN i 11-' UAKIMlN TKTRACI M.tlKl I)K
n;:u MOD
IV,\\'I-;NUMMKIC:S
-rrKt TIHM ••!•• ,\ MIXTMUK UK cnci.n AMU i:i;i.
Fi,;. -, llcm.iU- 1 i, i.-!,., , ,.| ., r [,.i.,i,- >! Cl.i'l, ..ml CCI j nl 1'ivc i1,,,,,,-,,',,,!,,,
lion I nnil
WAVKNIIMMKltS
[ ., I. Hi-iM-ili: in S|«-Llfn . I .' II r l[l OliliiiliL-il lu,,,,; » 111 irk. \Vlill ll.i< LI;..,-.H.| Ciil
.il Mi:nn w. i. a .', ia..i, ,,i,.i '2~.n ,,|..n n,
Rcniule Dclctlion uf Pure MCOII. DEM, and i Mijlore ot Dt.M tnj MEOII.
M FIELD, Sl'O RULI2ASE ITITII 2:L MUOII/DliM .|-20 PPM-h
(; ,,,,i,,,,, ,,f iKiiniK III S|.Tlr C.,i,l.»:,iic.|( III /.I. «' MKOII
MliOll wllli « 5 ,1. A I «( lill.jl lil.« Ad.tal 1.1 . 10 c... O.ll
1100 JOOO
WAV1SNUMDERS
MD30 ROS= -I
Fit. « MEOII-DEM Hclf»^ .III, Uw Sly U
[foul Llif liiilruinrnuiioii.
588
-------
DISCUSSION
EMILE HYMAN: Can you comment on the potential for stack emission
monitoring?
JACK DEMIRGIAN: We think there is tremendous potential forstack emission
monitoring, and that's one reason we ran carbon telrach lore form as an early set.
We are currently collecting, with the DOE Program, regular on-stream incineration
monitoring data. Once we get that data we will be giving it to Kroulil so that he
can calculate the coordinates for it. And then we can start to test this technique
on incinerations. There is a municipal incinerator on-site or near on-site, at the
Aberdeen Proving Ground. We have already collected some data there, before
some of the signal processing work was completed, and we wilt be analyzing that
data. We will be definitively moving, and hopefully we'll be presenting some on-
stream incineration monitoring data at the incineration conference this May.
Hopefully, after that we will present the remote equivalent of the on-stream
monitoring. So, stay tuned.
EMILE HYMAN: How about SOx and NOx and that kind of thing?
JACK DEMIRGIAN: Again, SOx and most of the NOx absorb out of the
detector window for this particular detector. SOx and NOx are a little more
difficult because they're between CXh and water in their absorption region. And
so the digital filler has to be done to eliminate water, or certainly reduce the effect
of water. This is not undoable. The technology is right there now. The new plateau
I discussed by Kroutil and Ditillo and Small, gives the potential to do that. 1 think
it's just more a function of time and staff than it is technologically difficult. Right
now the emphasis has been on organics as opposed to the SOx and NOx. There
are several good SOx and NOx monitors and we don't really want to compete. We
want to open up new areas. The new Clean Air Act has opened up a lot of areas
in which I think this remote passive technique will fill the bill. And I think FTIR
is right now about the only technique available that can do some of the
requirements of the new Clean Air Act.
DONALD GURKA: Those who know me know me to be skeptic. You said no
false positives and no false negatives? That suggests that the systems that you
looked at thus far are simple. Is that your conclusion?
JACK DEMIRGIAN: Yes, well, first off, keep in mind that the sophisticated
algorithms have just been developed this past September. What we have been
doing now is with conventional techniques. You're absolutely right. The first
system we worked with were pure components. Then we worked with two
component mixtures. We've worked with two component mixtures now with
some varying background. In this year we are funded to go ahead with more
complex backgrounds and more complex mixtures. In fact, we have ordered the
equipment so that we can now make multi-component mixtures and better
characterize the complex mixtures that would put more of the false positives and
false negatives to the lest, but your point is quite well taken. We presented data
with one and twocomponenl mixtures. The fact that we' ve chosen a difficult two
component mixture is a good sign of things to come. Had we failed miserably
with the methanol DEN, then I think you would have a very valid statement. I
think when we had succeeded with the very difficult methanol DEN case we have
justified going on to more complex mixtures. Again, the quantitative remote
passive as we're doing now is very new. And I guess if I'm here at the next
conference, you'll know that it worked.
DONALD GURKA: Can you visualize slanting this approach to all false
positives or all false negatives? Can you adjust your approach so that you screen
out only negatives or you screen out only positives?
JACK DEMIRGIAN: That's an interesting question. And I think that the key
on doing the false positives and false negatives is probably going to reside with
Kroulil's ability to digitally filter these things out. They have specific expertise
in filtering out false negatives, and they have a very great interest in filtering out
false positives. In order to do that properly with very complex mixtures, it is
better to ask Ditillo or Kroutil, because that's their specific expertise.
ERNIE TUAZON: You allude to methanol being a simple system. It isn't.
Underneath methanol, if you look at it very closely, there would be interferences
by ammonia. So, if you are in an area where there's fertilizer or a factory, or even
ammonia producing cows are around, it will be an interference. Also, CO2 is an
interference. These are the so-called lacing lines of CO2 that you don't see in the
laboratory, but at long parts you will see them. It's underneath those. Also, if you
are in the Los Angeles atmosphere, ozone will be an interference. So, it's not a
simple one as far as that is concerned, anyway.
JACK DEMIRGIAN: We specifically addressed the CO2 by collecting
chloroform data which totally enveloped the CO2. The Army has been working
on the ozone problem for quite a bit of time, and that's within their coordinate
system to filler out. Now, I don't think we have done ammonia yet or high
concentrations of...
ERNIETUAZON: No, I don't mean that. I mean just in analyzing that particular
band that you see, there are a lot of interferences underneath that.
JACK DEMIRGIAN: Yes, ammonia itself has relatively sharp bands and the
algorithm is able to discern a sharp band versus a smooth band. And I can show
you some data. If we get into the THAMA data where we're looking for
explosives which are nitrates, they absorb smack in the middle of the water
region. So, if you've got an interference, it's water in the soil samples. And we've
spiked the soil samples with 10% water, and looked at ppm explosives. Even with
the very sharp water bands versus the relatively broad explosive bands it does not
affect the algorithms' ability to quantify. We were very satisfied with these
results. A lot of the atmospheric gases that are small molecules have very sharp
bands, and the organics have much broader bands and that makes quantitation
and identification a good deal easier. So, as I said, methanol was relatively
straight forward. Most of the atmospheric gases are not going to be as big a
problem. I would guess in an application such as treaty verification where you
have someone deliberately trying to fool your system and put out components
that are very, very similar, that might be a tougher test. But, atmospheric is not
as bad as you would think.
589
-------
INTERPRETATION OF PPM-METER DATA FROM LONG-PATH OPTICAL MONITORING
SYSTEMS AS THEY WOULD BE USED AT SUPERFUND HAZARDOUS WASTES SITES
Thomas H. Pritchett
U.S. Environmental Protection Agency
Edison, NJ
Timothy R. Minnich, Robert L. Scotto and Margaret R. Leo
Blasland, Bouck & Lee
Edison, NJ
Recently, several groups have been evaluating the use of
long-path optical monitoring systems such as remote sensing
UV and remote sensing FTIR. Because of the potential
power of this field analytical technique, several of the more
active groups have been attempting to compare the results
from the long-path systems to Summa canister results. While
these comparisons have generally demonstrated that the two
techniques are indeed comparable, they generally have not
addressed how the long-path (or path-integrated) data could
be interpreted to meet the site manager's needs. Unfortu-
nately, many in the air toxics field are not familiar with the
downwind dispersion equations which can be used in con-
junction with path-integrated concentrations to define the
emission source to downwind receptor relationship. Instead,
most are usually used to interpreting only average concentra-
tion data from single locations.
The fundamentals of downwind transport illustrate how
ppm-meter data can be interpreted in three of the most
appropriate applications of long-path optical monitoring:
assessing of the baseline air emissions from an inactive site,
assessing the air impact of future cleanup operations during
pilot scale testing and fenceline monitoring during actual
cleanup operations. In all cases, the path integrated data can
be used to predict "worst-case" air concentrations for the
surrounding area — even for areas not covered by the
original monitoring study.
591
-------
DISCUSSION
DONALD (JURKA: My question is on regulatory aspects that are driven by the
technological slate of the art. As far as emissions go, the state of the art is usually
or always point sampling approaches. The question is, are there any gaps in the
point sampling approaches which are covered by open path IR, thus that open
path IR in that situation is now the state of the art?
TOM PRITCHKTT: There are two types of regulatory approaches. There are
air toxics people who work with point monitoring. Those point monitoring
techniques assume either a time averaged, a long term exposure average
concentration, or they assume a maximum concentration. Well, the point
monitoring approach is not a trivial matter to ensure that you're taking your point
monitoring sample at the maximum point in the plume. You have to know where
that maximum is before you take your sample. I've actually been out with a
mobile mass spec and was driving down the highway and watched a regulatory
agency take a measurement, a grab sample, that they're going to use to slam on
a company. Well, I drove another 50 yards down the road and found the plume
from (he facility, which they were actually trying to go after. They were taking
a sample in the upwind plume. So you've got to be very, very careful when you're
doing point monitoring to determine regulatory compliance. You can very easily
miss the maximum plume.
At this time, if you're trying to regulate a particular emission source based upon
a long-term average point monitoring concentration, there is a major gap in
guidance in how to interpret that data. Particularly, what do you do with
nondetects. Do you use a source receptor relationship and treat the upwind
nondetects as zeros, or do you use detection limits? That's particularly important
for risk assessment because the one-lo-two order of magnitude difference
between ihc instrument detection limits and the concentration of concern tends
to mask its risk.
The other type of regulatory situation is the one that is used in the Clean Air Act.
They regulate in terms of emission rates, grams per second. You cannot exceed
so many grams per second. I don't know what approach they're going to take in
the Clean Air Act, but I think ultimately they're going to have to go to grams per
second. To calculate an emission rate using point monitoring techniques under
simple gauging conditions takes anywhere from seven to ten Summa canisters,
and that's to get one reading which you then have to do in triplicate. So, that's 21
to 30 samples just to get one measurement you feel that you can use as an
emission rale.
With open path monitoring you can do that in probably about three minutes. You
may very well be driven not so much by the detection limit — how many spectra
you have to co-add to gel your detection I im il, as much as how many spectra must
you require before you have the gauging conditions which you're using to
interpret the data.
TOM PRITCHETT: The answer is that there's a lot of gaps in point monitoring
techniques which make it not as ideal for, let's say, health assessment purposes
related to specific source, and for determining emission rales related to a given
source that I think open path monitoring can solve.
GARY ROBERTSON: Tom, while I was listening to you I noticed that you were
doing most of your measurements perpendicular to the plume. It looked to me
that if you took those measurements at various angles to the plume, you could get
a lot of information about defining the plume and perhaps even the shape of the
plume.
TOM PRITCHETT: Under the straight gauging condition, if you're near
perpendicular, say within about 30 degrees perpendicular, you can use the
meteorological datalodefine the plume. It's a lot easier, believe it or not, lodefine
the plume using your meteorological conditions, as was shown by the Kansas
Slate study, than it is to try to multiplex by burying the angle that you're shining
the beam. It's also logistically a lot easier lo have one beam set up and just filter
out the data where you're essentially perpendicular, than it is to sit there and
continuously move your beam to try to gel your multiplex beam orientation.
CLIFF DAHM: I wanted to explore a couple of things. One is you talk about
meteorological measurements and conditions. What do you recommend routinely
be measured meteorologically, for example at a fence line? And secondarily, I
also want to know, whether, on any of these studies that you've been talking
about, if there's been vertical structure determined at some of the plumes that
you've been monitoring?
TOM PRITCHETT: In answer to the first question, you typically look for wind
speed, wind direction and sigma theta to calculate the stability class. And in
answer to the second question, looking at the vertical component of the plume,
no, we haven't. Again, we're looking downwind of small point sources. The one
thing we have done in that relationship is we actually used controlled releases.
Let me go back to one of my questions here; look at the bottom equation, equation
5.1t you have a control led release, and you're measuring the path of concentration,
and you know wind speed, the only thing that's unknown is sigma 9. What we've
done is used controlled releases at different distances to see whether or not the
sigma O's were, consistent with the predicted sigma 6's of the Clifford Path —
I guess it was Clifford that measured it or Pascal. But anyhow, we found very,
very good agreement for that stability class of the sigma 6's. So, in essence we
have not directly measured vertical dispersion, but we've shown that the vertical
dispersions being used in the gauging equations were experimentally confirmed.
592
-------
AWARDS CEREMONY
The sponsors of Second International Symposium — Field Screening Methods for
Hazardous Wastes and Toxic Chemicals were pleased to include an awards program.
Mr. John Koutsandreas, Florida State University and the Symposium Executive
Secretary organized the program and assembled the review panel that evaluated
nearly 60 platform presentations and over 60 poster session papers. The members of
the awards committee included:
Mr. Robert Booth, former director of U.S. EPA's Cincinnati EMSL
Dr. Steven Levine. University of Michigan, School of Public Health
Dr. David Nelson, Vice President. Perkin-Elmer Corporation
Dr. Roy Herndon, Director, Chemical, Biological and Toxicology Research, Florida
Stale University
Dr. Michael Dellarco, U.S. EPA, Office of Research and Development
Dr. Russell McAllister, U.S. EPA, Office of Solid Waste and Emergency Response
Dr. Joseph Leone Hi, Associate Director, Applied Electromagnetic and Optics Lab,
SRI International
Mr. David Bottrell, Department of Energy, Office of Technology Development
Dr. Richard Tinlin, Geraghly & Miller, Inc.
The Symposium organizers and sponsors arc grateful to this awards committee for
their time and effort expended in evaluating the presentations.
The panel judged the two best private sector (i.e..non-Federal) papers and the two
best Federal papers and awarded U.S. EPA engraved plaques to:
Private Sector
Susan Eberlein for "Space Technology for Application to Terrestrial Hazardous
Materials Analysis and Acquisition"
Hui Wang for "Comparison of Aqueous Hcadspace Air Standard Versus Summa
Canister Air Standard for Volatile Organic Compound Field Screening"
Federal
Donald Smilh for"A Study of the Calibration of a Portable Energy Dispersive X-Ray
Fluorescence Spectrometer"
Tom Spinier for "The Use of Field Gas Chromulogruphy to Protect Groundwater
Supplies"
U.S. EPA engraved plaques were awarded to Ihc two bcsl poster presentations as
determined by the Awards Committee. They were:
"A Field-Portable Supercritical Fluid Extractor for Characterizing Semivolatile
Organic Compounds in Waste and Soil Samples" B.W. Wright and J.S. Fruchter
"Real Time Detection of Biological Aerosols" P.J. Stopa. M.T. Good, W. Zulich,
D.W. Sickenburger. E.W. Sarver, R.A. Mackey
Mr. Larry Cottran from Hewlett Packard presented two eagle trophies, donated by
Hewlett Packard, for overall outstanding technical contribution and quality of
presentation. Hewlett Packard pays considerable attention to these two critical
elements: improving quality and increasing technical contributions. The winners
were:
For best technical contribution:
Fred Milanovich for "A Fiber Optic Sensor for the Continuous Monitoring of
Chlorinated Hydrocarbons"
For best presentation:
Steven Levine for "Fourier Transform Infrared Speclropholomctry for Monitoring
of Contaminant Gases and Vapors in the Workplace Air"
Certificates were also presented that recognized the most outstanding paper in each
of the ten platform sessions. They were:
Session 1
CHEMICAL SENSORS
Fred Milanovich for "A Fiber Optic Sensor for the Continuous Monitoring of
Chlorinated Hydrocarbons"
Session 2 ION MOBILITY SPECTROMETRY
Suzanne Ehart Bell for "Hand-Held GC-Ion Mobility Spectrometry for On-Site
Analysis of Complex Organic Mixtures in Air or Vapors over Waste Sites"
Session 3 ROBOTICS
Susan Eberlein for "Space Technology for Application to Terrestrial Hazardous
Materials Analysis and Acquisition"
Session 4 QA AND STUDY DESIGN
John Mateo for "A Quality Assurance Sampling Plan for Emergency Response
(QASPER)"
Session 5 AIR PATHWAY MONITORING AT SUPERFUND SITES
Steven Levine for "High Speed Gas Chromatography for Air Monitoring"
Session 6 FIELD MOBILE GC/MS TECHNIQUES
Mary Cisper for "Field Measurement of Volatile Organic Compounds by Ion Trap
Mass Spectrometry"
Session 7 PORTABLE GAS CHROMATOGRAPHY
Hui Wang for "Comparison of Aqueous Headspace Air Standard Versus Summa
Canister Air Standard for Volatile Organic Compound Field Screening"
Session 8 FIELD SCREENING METHODS FOR WORKF.R SAFETY
Gerald Moore for "Improvements in the Monitoring of PPM Level Organic Vapors
with Field Portable Instruments"
Session 9 X-RAY FLUORESCENCE
Donald Smith for"A Study of the Calibration of Field Portable X-Ray Fluorescence
Instruments"
Session 10 FOURIER TRANSFORM INFRARED SPECTROMETRY &
OTHER SPECTROSCOPY METHODS
Gary Small for "Pattern Recognition Methods for FTIR Remote Sensing"
593
-------
CONCLUDING REMARKS BY SYMPOSIUM CHAIRPERSON,
DR. LLEWELLYN WILLIAMS
As I lay in bed this morning, I looked back over the week as each of my senses awakened. First came my sense of touch, and
I felt that the Symposium was a success, and I was touched by the quality of the papers and of the posters. And I was almost
"touched" by a number of technology developers looking for Federal funds. And I recall the pain of stabbing my upper lip with
a toothpick holding two Swedish meatballs.
Next came my hearing. I heard a broad range of useful information from the bureaucrats. I heard of breakthroughs and research
advances from our researchers and technology developers. Fortunately, I heard few complaints about the papers. And I heard
the sound of two thousand Swedish meatballs being poured into a silver chafing dish.
Next to return was my sight. I saw a lit entry way and a table full of awards. I saw dim images on the screen during the Plenary
Session. I saw colorful exhibits and the sharp graphics of poster and platform sessions. And I saw two thousand Swedish
meatballs being poured into a silver chafing dish.
The next of my senses to return was that of smell. I detected the sweet smell of success that could be attributed to the enthusiasm
and energies of you, the participants. I smelted the various emergency deodorizing measures used on Wednesday morning when
the shower water didn't work in the hotel. And I smelled the odor of over two thousand Swedish meatballs simmering in a silver
chafing dish.
The last sense to return was my taste. I recall the good taste displayed by the exhibitors during our reception period. I can taste
a consistently good coffee that was provided on our breaks. And I fear I'll continue to taste the Swedish meatballs for days to
come.
The results of all of this sensory input was a series of visions. The first is a vision of us all returning to Las Vegas two years
from now to do it all again. The second vision is of the widespread acceptance of field methods and the data derived therefrom.
And finally the vision of fifty large Lutheran women feverishly molding Swedish meatballs.
At this time, as unprepared as I am, someone asked for another poem.
Now that you have seen it all and will set upon your way.
We'd love to get your feedback as we plan for number tres.
How'd you like the balance and the papers and the rest?
What things would you change?
What did you like the best?
Thanks to all the many folks who made this whole thing happen,
And bailed out the Chairman every time they caught him napping.
And thanks for sponsor monies, and support from all the brass.
It helped us build a program that was nothing hut first class.
My special thanks to Eric andJoAnn, and yes, to Kouts,
For pulling things together so give your horns some toots.
And if we had success in our attempts to make it work,
It was your participation here that really made it perk.
So, looking to the future 1 suspect we'll meet again.
As we catch up on developments in monitoring, and then
We'll see if the technologies have made it to the play off
Can bear the fruit, and stand the test, and over time will pay off.
Ladies and gentlemen, thank you very much.
595
-------
CALIBRATION OF FIBER OPTIC CHEMICAL SENSORS
W. F. Arendale and Richard Hatcher
Laboratory for Inline Process Analyses
Kennelli E. Johnson Research Center
The University of Alabama in llunlsville
HuiiLsvillc, AL l.vS'jy
and
Bruce Nielsen
H(| AFESC/RDVW
Tyndall AFB, FL 32403-6001
Fiber oplic chemical sensors to be used for monitoring envi-
ronmental pollutants have been extensively researched lor at
least ten years. Although excellent research results have been
presented, few if any systems have met the quality assurance
requirements such that they are now in production and avail-
able for general use. Many fiber optic sensors for monitoring
physical parameters as temperature, microslrain, and accous-
tics are available. Sensors for monitoring chemical processes
have been used successfully.
Sensors that meet QA/QC manufacturing requirements are
usually the result of careful modeling. For the past several
years a part of our research activity has been related to the
formulation of appropriate models for FOCS. Our studies
have shown that the chemistries of the materials placed on the
distal end of the fiber are adequately modeled. The deficien-
cies appear to be related to the physics and engineering ol the
optics, lack of quality control during the manufacturing proc-
ess, and/or lack ol sufficient information being collected to
assure reliable information in the presence of intcrlcmils.
Optical fibers are used as:
(1) Carriers ot optical signals - photons travel simul-
taneously in many directions.
(2) Sensor/Carriers - Optical properties of
provide the sensing medium
liber
(3) Components of integrated diagnostic systems.
Photons pass down the core of a fiber in several ways, hi
addition to the rays passing down the center of the fiber some
of the. photons are reflected at the boundary between the core
and the cladding. The cone that includes rays that pass
through the fiber is shown in Figure 1A. Energy can be lost to
the cladding when a bend occurs in the liber as shown in
Figure IB. To avoid loss due to microbends and to protect the
fiber from stress, the clad fiber is bundled into a protective
cable shown in Figure 1C.
Figure I. Light Transmission Through Fibers
The communication industry has spent millions of dollars on
obtaining high purity silica core, selection and application of
the cladding, and cabling of the liber. Some of the problems
that must be overcome are shown on Figure 2.
Figure 2. Sources of Energy Loss in Fibers
Our recommendation for quality sensors arc:
(1) Use the purist high quality fiber
(2) The cladding lor high quality liber is added to
the core while the fiber is still in the inert
atmosphere of the drawing furnancc. Do not
attempt to change the cladding except doing the
manufacturing process.
(3) If possible use only cabled or rigidly supported
fiber.
(4) Standardize the sensor using a sufficient number
ol measurement parameters such that all mean-
ingful variance is represented. It is not necessary
to quautitate all sources of variances, but a
measurement parameter must be included.
l-iguie 3 shows a test chamber that we have used successfully
lor determining the required number of measurement parame-
ters and standardizing our sensors.
597
-------
. ,
c n; n :ir z. s>n. cn< j
Figure 3. Inline MuHivariate Analysis Flow Apparatus
598
-------
GAS-CHROMATOGRAPHIC ANALYSIS OF SOIL-GAS SAMPLES AT A GASOLINE-SPILL
R.J. Baker, J.M. Fischer, N.P. Smith, S.A. Koehnlein, and A.L. Baehr
U.S. Geological Survey, 810 Bear Tavern Rd, W. Trenton, NJ 08628
ABSTRACT
The U.S. Geological Survey is studying
remediation processes at a gasoline-
spill site in Galloway Township, New
Jersey. A field-laboratory trailer was
equipped with a gas chromatograph (GC)
configured to analyze soil-gas samples
for gasoline hydrocarbons and inorganic
gases, such as oxygen, nitrogen, carbon
monoxide, and carbon dioxide. This
instrument was selected over other
analytical options because of its
versatility; it can be used to monitor
all significant organic and inorganic
components of unsaturated-zone gases.
Each of two chromatographic columns is
equipped with a vapor-sample injection
valve fitted with a sample loop. The
sample-loop volume determines the
injection size. A chromatography data
system and a micro-computer are used for
data acquisition, processing, and
storage.
A thermal-conductivity detector is used
in conjunction with a 3.3-meter-long
molecular sieve column for analysis of
inorganic gases. A flame-ionization
detector is used with a 30-meter-long
fused silica capillary column with a
dimethylpolysiloxane stationary phase
for analysis of vapor-phase gasoline
hydrocarbons. Inorganic and organic
species are identified by retention time
and quantified by linear-regression
standard-curve analysis.
A method for evaluating hydrocarbon
chromatograms that does not require
identification of specific peaks was
developed. Chromatograms are divided
into retention-time increments, each of
which contains peaks of compounds that
have the same carbon number (number of
carbon atoms). A sample can then be
described semiquantitatively in terms of
the number of compounds of each carbon
number, total mass of each carbon
number, or percent of mass represented
by each carbon number. The method is
based on the relation between the carbon
number of a compound and its boiling
point, and between boiling point and GC
retention time. By using this method,
retention time can be used to determine
the boiling point and most probable
carbon number of an unidentified
hydrocarbon compound. The margin of
error of the method was established by
determining the carbon numbers of 167
compounds from their boiling points.
Correct carbon-number assignments were
made for 131 compounds (78.4 percent),
and carbon-number was underestimated for
16 compounds (9.6 percent) and
overestimated for 20 compounds (12.0
percent). All over- and underestimates
were in error by one carbon atom.
The GC and the data-evaluation methods
used are providing excellent soil-gas
characterization during this field
study. Chromatogram analysis by carbon-
number determination can be used in
other studies where hydrocarbons are
detected but not specifically
identified.
599
-------
SIGNIFICANT PHYSICAL EFFECTS ON SURFACE ACOUSTIC WAVE (SAW) SENSORS'
David L. Bartley
National Institute for Occupational Safety and Health
4676 Columbia Parkway
Cincinnati. Ohio 45227
Surface acoustic wave (SAW)
devices are presently being
developed for applications in
chemical sensing as well as for
polymer characterization. SAW
gas and vapor sensors have the
potential for miniaturization
and high sensitivity to a wide
variety of substances. Polymer
characterization is applicable
to such diverse fields as
protective coating design and
decontamination of polymers.
Research was conducted to better
understand the physical
mechanisms behind SAW response.
Practical problems as to film
uniformity, thickness
measurement and environmental
controls such as temperature and
gas flow rates necessary in such
measurements were considered.
The effects of elastic
properties in comparison to mass
loading of polymer coatings on
SAW substrates were
investigated. A theoretical
basis for the effects of vapor-
induced swelling or of thermal
expansion was established.
Compressive tension and its
effect on SAW frequencies were
found to be simple to describe.
if there is no film slippage or
polymer flow. The response of
quartz-substrate SAW crystals
coated with polycarbonate and
polyimide (glassy polymers) upon
exposure to toluene and methanol
was measured and was found
consistent with theory in
predicting effects of the order
of the ratio of coating to
substrate elastic constants.
Bartley. D.L. and Dominguez. D.D.:
of Polymer Coatings on Surface Acoustic Waves."
Anal.Chem. 62:1649 (1990).
601
-------
AN EVALUATION OF FIELD PORTABLE XRF SOIL PREPARATION
METHODS
Mark Bcrnlck
Roy F. Weston, IncVREAC Contract
GSA Raritan Depot
Building 209 Annex
2890WoodbridgcAve.
Edison, NJ 08837
Donna Idler
Roy F. Weston, IncTREAC Contract
GSA Raritan Depot
Building 209 Annex
2890 Woodbridge Ave.
Edison, NJ 08837
Lawrence Kaelln
Roy F. Weston, IncTREAC Contract
GSA Raritan Depot
Building 209 Annex
2890 Woodbridge Ave.
Edison, NJ 08837
Mark Sprenger
USEPA/ERT
GSA Raritan Depot
Building 18
Edison, NJ. 08837
Dave Miller and Jayanti Patel
Roy F. Weston, IncTREAC Contract
GSA Raritan Depot
Building 209 Annex
2890 Woodbridge Ave.
Edison, NJ. 08837
George Prince
USEPA/ERT
GSA Raritan Depot
Building 18
Edison, NJ 08837
INTRODUCTION
The USEPA Environmental Response Team
(ERT) has been using field portable X-ray Fluores-
cence (XRF) spectrometers to characterize Super-
fund and hazardous waste sites. An Outokumpu
Electronics Inc. (OEI) X-MET 880 XRF
spectrometer equipped with a surface probe con-
taining Cm-244 and Am-241 radio-isotopes was
selected. Field portable XRFs have enabled the
ERT to estimate the extent of metal contamination;
support biological 'plant stress assessment" studies
in tidal wetlands; and, support a health and safety
assessment of the extent of contamination and pos-
sible human exposure of a network of hiking trails
and campsites contaminated by a smelting opera-
tion.
BATTERY BREAKAGE SITE
In September, 1989 the ERT deployed a XRF unit
to perform an extensive post remedial site lead
survey at an abandoned battery reclamation site.
Analyses were planned for both surface and sub-
surface soil samples to determine the extent of lead
contamination. The OEI X-MET 880 XRF
spectrometer was calibrated with a suite of 18 site
specific lead standards by an analytical chemist.
An in-situ analysis method was preferred since it
would reduce sample preparation time and the risk
of personal exposure to the contaminants. A litera-
ture search failed to find any studies demonstrating
the correlation of results from an in-situ XRF
analysis and the accepted method of preparing
soils by drying and sieving. The analyses included
29 in-situ and prepared soil samples in an effort to
evaluate the sample preparation methods.
Samples that were analyzed by the XRF in-situ
method were dried and sieved and reanalyzed by
XRF. The paired-difference t-test was used to
evaluate the results of both sample preparation
methods using a significance level of 5 percent.
PROCEDURE FOR IN-SITU SURFACE SOIL
LEAD XRF ANALYSIS
All organic matter and large rocks were removed
from the area (8" by 5") to be analyzed. The area
was then rendered flat with a stainless steel trowel
The XRF instrument was initiated for a 60 second
measuring time for the lead analysis while the sur-
face probe was held flush against the soil surface.
The sample area number, location (A) and XRF
lead result were logged into a field notebook. The
analysis was then repeated selecting a different
analysis location (B) within the prepared area. The
results of the two analyses were averaged and
reported.
603
-------
PROCEDURE FOR DRIED AND SIEVED SOIL
XRF ANALYSIS
The soil within the prepared area was removed to a
depth of a quarter inch. Large stones and organic
matter were removed prior to drying. The entire
sample was dried in an oven and sieved through 10
and 20 mesh stainless steel sieves with a stainless steel
spoon. All organic matter and stones were removed
and discarded. The sample was homogenized for one
minute by dividing the sample into quarters and
mixing opposite sides together.
A 31-mm, polyethylene X-ray sample cup was labeled
and Tilled with soil. The cup was sealed with 0.2-mm
thick polypropylene, X-ray window film. Prior to
XRF analysis, the sample cup was gently tapped
against the table top three times to pack the soil
evenly against the polypropylene window film. The
sample cup was placed directly on the XRF detector
window and the instrument initialized for a 60 second
lead analysis time. The result of the analysis was
reported.
CHEMICAL ANALYSIS PROCEDURE FOR
METALS IN SOIL
The XRF sample cup was submitted to the laboratory
for chemical atomic absorption (AA) analysis. Ap-
proximately 0.5 g of sample, weighed to 0.001 g ac-
curacy, was thoroughly mixed with 10 ml 1:1 nitric
acid:water, digested according to SW-846, Method
3050 and analyzed according to Method 7000.
RESULTS OF THE XRF AND AA METHODS
ANALYSES
Seven of the 29 samples analyzed had XRF results
below the XRF detection limit of 123 mg/kg or quan-
titation limit (QL) of 410 mg/kg. The XRF and AA
results of these samples are presented in Table 1.
The AA results show that all of the sample lead
concentrations fell below the XRF quantitation limit
of 410 mg/kg. The results of both XRF methods were
all below the XRF QL except for sample 5469B's
XRF in-situ result that was 543 mg/kg lead. This was
considered acceptable since priority samples with
XRF results at or near the QL would be sent to the
lab for AA analysis. A high frequency of false nega-
tive XRF results would have caused either XRF
method to be questioned.
Three of the 29 samples analyzed had XRF results
above the XRF linear calibration range ( 5300
mg/kg). The XRF and AA results of these samples
are presented in Table 2. The AA results show that
all of the sample lead concentrations are near the end
of the XRF linear calibration range.
Nineteen of the samples had XRF results above the
XRF quantitation limit and below the end of the XRF
linear calibration range. The XRF and AA results of
these samples are presented in Table 3. These 19
samples along with samples 5464B and 5469B from
Table 1 (n - 21) were used in a paired-difference
t-test analysis.
RESULTS OF THE XRF AND AA METHODS
PAIRED-DIFFERENCE T-TEST EVALUA-
TION
The goal of the paired difference t-test is to deter-
mine if the mean difference of two populations of
paired results, is different from zero at the 5-percent
significance level. In other words, the analyst is will-
ing to accept a 1 in 20 chance of saying that the
average difference of the two populations is sig-
nificantly different from zero when in fact it is not.
Additionally, the test makes two assumptions. First,
that each pair of measurements is independent of the
other pairs. Second, that the differences are from a
normal distribution. The populations used in this test
were normalized with a square root function.
The in-situ and dried & sieved XRF results were
analyzed by the paired- difference t-test The prob-
ability value for this test was 0279 and is greater then
0.05 (which is associated with a 5-percent significance
level). Therefore, the average difference of the
paired results of the two XRF analytical methods is
not significantly different from zero.
The in-situ XRF and AA results were analyzed by the
paired-difference t-test. The probability value for
this test was 0.671 and is greater then 0.05. Therefore,
the average difference of the paired results of the two
analytical methods is not significantly different from
zero.
The results of this statistical test enabled the project
manager to conclude that the two methods of
preparation were not significantly different and that
the in-situ XRF method and the AA lead analytical
methods were not significantly different. In-situ
analysis was then performed on 500 surface and sub-
604
-------
surface samples. Additionally, portable XRF was
used to support selection of soil samples for use in a
treatability study.
RESULTS OF THE SITE IN-SITU XRF AND AA
EXTENT OF CONTAMINATION ANALYSES
Seventy-one of the 500 samples analyzed by in-situ
XRF were collected and submitted to the REAC
laboratory for AA lead analysis. Twenty-six of these
samples had XRF results below the XRF QL of 270
rag/kg (the XRF lead calibration curve was modified
using site-specific standards from the method evalua-
tion work that resulted in lower XRF lead detection
and quantitation limits and an extended linear
calibration range). The XRF and AA results of these
samples are presented in Table 4. Four of these
samples had AA results above the XRF QL of 270
mg/kg.
Five of the 71 samples had XRF results above the
XRF linear calibration range of 12,000 mg/kg lead.
The XRF and AA results of these samples are
presented in Table 5. All of these samples had AA
results above or near the end of the XRF linear
calibration range.
Forty of the samples had XRF results above the XRF
quantitation limit and below the end of the XRF linear
calibration range. The XRF and AA results of these
samples are presented in Table 6. These forty
samples and the six samples in Table 4 with "J" XRF
values (n = 46) were used in a paired-difference t-test
analysis. The probability value for this test was 0.872
and is greater then 0.05 (which is associated with a
5-percent significance level). Therefore, the average
difference of the paired results of the two lead analyti-
cal methods is not significantly different from zero.
CONCLUSIONS
The paired-difference t-test can be used as a decision
tool in the evaluation of XRF soil lead analytical
methods. It showed that the average difference be-
tween these two methods was not significantly dif-
ferent from zero at the 5-percent significance level.
At also showed the average difference between the
in-situ XRF and the AA lead analytical methods was
not significantly different from zero at the 5-percent
significance level for both the XRF method develop-
ment data and the site extent of contamination data.
Additionally, the portable OEI X-MET 880 sup-
ported the following:
*Soil lead analysis in a densely wooded area of the site
initially assumed to be uncontaminated . Investiga-
tion of the area located battery casings mixed with soil
under the leaves. XRF lead analysis confirmed that
the area was contaminated with high mg/kg levels of
lead
'Selection of soil samples for a treatability study.
*320 XRF surface lead analysis results were used to
develop a site contour map.
OTHER XRF APPLICATIONS
In June, 1990, the ERT deployed a XRF unit to
perform cadmium and nickel analyses on sediment
samples from a tidal wetlands contaminated by a
battery manufacturing facility. The biological "plant
stress assessment" work plan called for the investiga-
tion of a minimum of one plot of vegetation for each
anticipated nominal cadmium concentration range.
The results of XRF analyses of the plot sediments
enabled the project manager to select the appropriate
plots. Selected samples were submitted for
laboratory AA analysis. The paired-difference t-test
evaluated the results of both XRF and AA analytical
methods and found the average difference of the
paired results of the two cadmium analytical methods
was not significantly different from zero (at the 5-per-
cent significance level).
That same month, an XRF unit performed lead and
zinc in-situ surface soil analysis in a network of hiking
trails and campsites contaminated by a smelting
operation for a health and safety assessment of the
extent of contamination and possible human ex-
posure. Samples were submitted for AA analysis and
the results of both analytical methods were evaluated
by the paired-difference t-test.
The average difference of the paired results of the zinc
analytical methods (most lead values were below the
XRF QL) were significantly different from zero (at
the 5-percent significance level). Almost all of the
XRF zinc results were higher then the AA analysis. It
is suspected that the contamination from the smelter
is concentrated in the top layer of the soil and was
diluted when the top quarter inch of soil was removed
for laboratory AA analysis.
605
-------
TABLE 1
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU AND DRIED & SIEVED SOIL XRF Pb ANALYSES,
AND AA Pb ANALYSIS IN mg/kg
SAMPLES WITH RESULTS BELOW THE XRF QUANTITATION LEVEL
SAMPLE NO. IN-SITU XRF DRIED & SIEVED XRF
5461B ND ND
5464B 125-J 148-J
5469B 543 125-J
5470B ND ND
4281B 190-J ND
4499B ND ND
4495B 138-J ND
XRF detection limit = 123 mg/kg. XRF quantitation limit = 410 mg/kg.
AA detection limit = 5 mg/kg.
ND denotes not detected.
J denotes the sample concentration is between the detection limit and the quantitation limit.
TABLE 2
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU AND DRIED & SIEVED SOIL XRF Pb ANAYLSES,
AND AA Pb ANALYSIS IN mg/kg
SAMPLES WITH RESULTS ABOVE THE XRF LINEAR CALIBRATION RANGE
SAMPLE NO. IN-SITU XRF DRIED & SIEVED XRF AA LAB
4332B 13800 * 6040 * 5000
4496B 7680 * 6520 * 5300
3319B 6590 * 7210 * 5100
XRF detection limit= 123 mg/kg. XRF quantitattion limit= 410 mg/kg.
AA detection limit- 5 mg/kg.
*- denotes sample cone, is above the XRF linear calibration range (> 5300 mg/kg).
TABLE 3
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU AND DRIED & SIEVED SOIL XRF Pb ANALYSES, AND AA Pb ANALYSIS IN mg/kg
SAMPLES WITH RESULTS ABOVE THE XRF QUANTITATION LIMIT & IN XRF CALIBRATION RANGE
SAMPLE NO.
5460B
5462B
5463B
5465B
5466B
5467B
5468B
5471B
4280B
4331B
4498B
5590C
4500B
5489B
5490B
5491B
5493B
5494B
5492B
XRF detection limit=
IN-SITU XRF
1950
1390
3180
838
771
540
416
1000
1170
1030
532
655
1730
2530
1430
1390
1370
910
1420
123 mg/kg. XRF quantitation
DRIED & SIEVED XRF
2320
1500
3080
1080
946
615
1450
2010
3040
1310
644
716
1990
680
2610
1350
1250
883
1270
limit= 410 mg/kg. AA detection limit=
AA LAB
1800
1200
3200
880
820
530
950
1100
2800
1300
470
460
1600
1000
1900
880
1200
700
1100
5 mg/kg.
606
-------
TABLE 4
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU XRF AND AA Pb ANALYSIS IN rag/kg ON SURFACE AND BORING SOILS
SAMPLES WITH RESULTS BELOW THE XRF QUANTITATION LEVEL
SAMPLE NO. IN-SITU XRF AALAB SAMPLE NO. IN-SITU XRF AA LAB
B2-1 ND 21 B4-3 ND 32
B6-1 ND 15 B8-2 ND 15
B19-2 ND 32 B20-2 ND 120
B20-4 ND 16 B22-2 ND 19
B24-1 ND 90 B25-1 ND 51
B26-2 ND 33 B28-2 ND 100
B29-3 ND 22 B30-4 ND 11
B32-6 ND 25 B34-6 ND 25
B39-0 ND 50 B45-0 ND 54
MW10-2 ND 24 MW11-2 ND 23
B4-1 81J 98 B42-6 152 J 370
B13-1 176 J 160 B2-2 190 J 790
B39-12 249 J 480 MW9-1 261J 1400
ND-denotes not detected. XRF detection limit=81 mg/kg' AA detection limit-5 mg/kg.
J-denotes concentration is between the detection and quantification limit. XRF quantitation limit = 270 mg/kg.
TABLES
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU XRF AND AA Pb ANALYSIS IN mg/kg ON SURFACE AND BORING SOILS
SAMPLES WITH XRF RESULTS ABOVE THE XRF LINEAR CALIBRATION RANGE
SAMPLE NO. IN-SITU XRF AALAB
B13-2A 22500 * 53000
B25-2 69500« 120000
B28-3 73000 * 11000
B30-2 54900 * 110000
B30-3 15700 * 170000
XRF detection limit= 81 mg/kg. AA detection limit= 5 mg/kg. XRF quantitation limit= 270 mg/kg.
J-denotes concentration is between the detection and quantification limit
•-denotes sample cone, is above the XRF linear calibration range (>12000 mg/kg).
TABLE 6
BATTERY BREAKAGE SITE
RESULTS OF IN-SITU XRF AND AA Pb ANALYSIS IN mg/kg ON SURFACE AND BORING SOILS
SAMPLES RESULTS ABOVE THE XRF QUANTITATION LIMIT AND IN XRF CALIBRATION RANGE
SAMPLE NO. IN-SITU XRF AALAB SAMPLE NO. IN-SITU XRF AALAB
B2-4 555 750 B4-2 1010 780
B5-2 412 1600 B6-0 2990 3100
B10-1 1290 2700 Bll-0 10700 6700
Bll-1 4230 4100 B13-2B 449 290
B14-1 1120 740 B14-2 11700 6100
B15-1 991 570 B17-1 1740 2500
B19-1 862 1400 B20-1 2460 4500
B20-3 648 610 B22-1 744 1300
B26-1 11000 8900 B31-2 4870 3200
B32-0 8390 12000 B34-0 1120 1500
B36^ 3710 2700 B39-18 844 1100
B41-0 10500 11000 B41-6 292 360
B44-30 730 800 MW9-2 945 810
MW11-1 1920 2700 SS165 579 380
SS170 1630 4400 SS173 1040 970
SS175 3080 2600 SS180 3010 2200
SS183 6930 5500 SS189 5060 2100
SS195 2220 1700 SS199 5690 8300
B5-1 301 140 B12-3A 657 140
B18-3 1303 270 B12-2 2797 170
XRF detection limit = 81 mg/kg. XRF quantitation limit = 270 mg/kg. AA detection limit » 5 mg/kg.
607
-------
DEVELOPMENT OF A FIELD SCREENING TECHNIQUE
FOR DIMETHYL MERCURY IN AIR
Brian E. Brass, Environmental Scientist
Roy F. Weston / REAC Contract
GSA Raritan Depot, Building 209 Annex
Edison, NJ. 08837
Lawrence P. Kaelin, Chemist
Roy F. Weston / REAC Contract
GSA Raritan Depot, Building 209 Annex
Edison, NJ. 08837
Thomas H. Pritchett, Chemist
USEPA, Environmental Response Team (ERT)
GSA Raritan Depot, Building 18
Edison, NJ. 08837
Many forms of organic and inorganic mercury are per-
vasive in the environment; with both natural and in-
dustrial sources contributing to the total environmental
mercury burden. Mercury can be biologically metabo-
lized to form various organomercurials. One of these,
Dimethyl Mercury (DMM) represents a potential
health risk via air exposure because of its high volatility
and toxicity.
Mercury contaminated soils and sediments are fre-
quently biologically active and have been demonstrated
to contain DMM. If left undisturbed, emissions of
DMM will be related to the biological generation rate
of DMM as well as the emission rate through the soil or
sediments as regulated by porosity, temperature, pres-
sure, and other physical-chemical factors. However,
when the soils are disturbed, the potential for elevated
emissions of DMM increases.
Research conducted by the USEPA's Environmental
Response Team (ERT), with the support of Roy F.
Weston, Inc. through the Response, Engineering, and
Analytical Contract (REAC), has resulted in a potential
real-time monitoring technique.
BACKGROUND INFORMATION PROMPTING
RESEARCH
The Army Corps of Engineers had been conducting a
cleanup of mercury contaminated soils; however, site
remediation was suspended due to the potential for
DMM emissions and the lack of a real-time air monitor-
ing method.
A preliminary investigation conducted by ERT and
Weston/REAC at the request of the Army Corps of
Engineering and USEPA Region I, indicated that
DMM could be present in the soils, especially in areas
where anaerobic activity was prevalent.
REVIEW OF REAL-TIME PORTABLE INSTRU-
MENTS FOR DETECTING DMM
The criteria established for selecting a real-time instru-
ment for monitoring DMM were quite restrictive. First,
the instrument had to be portable and permit operation
by non-technical personnel. Second, it had to have
real-time or semi-real-time monitoring capabilities.
609
-------
The third requirement was that it be specific to or-
ganomercurials.
A literature review indicated that the primary
methods for detecting mercury were atomic adsorp-
tion, gas chromatograph, infra-red analyzers, and
gold film technology. Atomic Adsorption instru-
ments are generally not portable, require technical
expertise to operate and provide only semi-real-time
results. Gas chromatography requires technical ex-
pertise and provides semi-real-time results. Infra-
red analyzers have too many interferences from
organic compounds. Gold film detectors are cross-
sensitive to sulfide compounds, however, the use of
an internal sulfide trapping pre-filter negates this
cross-sensitivity. The gold film technology thus ap-
peared to have the greatest potential for the required
application, and a gold film mercury vapor analyzer
was selected for detailed review.
GOLD FILM TECHNOLOGY
The Arizona Instruments Model 411, Gold Film Mer-
cury Vapor Analyzer was selected as the instrument
of choice. The Model 411 was originally developed
for monitoring elemental mercury in air, and operates
on the principal that mercury will form an amalgam
when it contacts a gold film.
The Model 411 detects the presence of mercury by
passing a stream of air across a thin gold film. As the
mercury in the air contacts the film an amalgam is
formed. The amalgamation causes an increase in the
electrical resistance of the film proportional to the
mass of mercury in the sample. The change in resis-
tance is compared to a sealed reference gold film and
processed by a microprocessor to provide a digital
read-out in milligrams of mercury per cubic meter
(mg/m3).
To eliminate the necessity of thermally desorbing the
gold film after each use, the Model 411 employs a
microprocessor which allows the instrument to
operate over a wide range of resistances while
remaining balanced with the reference film. Heating
the film to approximately 250°C, and subsequently
passing a stream of mercury free air across the film
desorbs the mercury and restores the film to its
baseline resistance.
EXPERIMENTS WITH SILVER-COATED
CHROMOSORB
Since DMM may be metabolized from elemental mer-
cury, the hypothesis was made that both DMM and
elemental mercury might be present during air monitor-
ing. Elemental mercury would interfere with the detec-
tion of DMM in monitoring air, therefore a means to
remove the elemental mercury from the sample without
affecting the DMM concentration was required. Silver-
coated Chromosorb was tested for this purpose.
The first test involved monitoring for elemental mercury
with the Model 411 using a silver-coated Chromosorb
tube pre-scrubber within a test vessel containing elemen-
tal mercury in which the vapor pressure had reached
equilibrium. Thirty samples of the mercury-saturated air
were collected without breakthrough occurring from the
silver-coated Chromosorb tube.
The second test of the silver-coated Chromosorb pre-
scrubber involved determining if DMM would pass
through it. This was accomplished by preparing a DMM
standard and measuring the concentration with and
without the silver-coated pre-scrubber. As Table 1 indi-
cates, the test results for the analysis with and without the
pre-scrubber are essentially the same.
PREPARATION OF DMM STANDARDS
DMM standards are not commercially available, there-
fore, it was necessary to prepare them in-house in Summa
passivated canisters by injecting a measured volume of
DMM and methanol solution into the canister. Due to
uncertainties in this procedure, DMM standards were
confirmed by select ion gas chromatography and mass
spectra analysis.
USE OF THE MODEL 411 FOR DETECTION OF
DMM
Initial experiments conducted with the Model 411 (con-
figured as per manufacturer's specifications) provided
erratic results and an inadequate detection limit for
DMM.
The Arizona Instrument's Model 411 was therefore
modified as follows:
l.The detector resistance was increased from ap-
proximately 60 ohms to approximately 98 ohms.
ZThe instrument's sample flow rate was increased from
720 cubic centimeters per minute (cc/m) to 866 cc/m.
3.The sampling duration was doubled from 10 seconds to
20 seconds.
610
-------
TABLE 1. COMPARISON OF ARIZONA INSTRUMENTS MODEL 411 RESPONSE WITH
AND WITHOUT A SILVER-COATED CHROMOSORB PRE-SCRUBBER TO 4.8 ppb-V and 24.6 ppb-V
STANDARDS OF DMM
DMM
Standard
Concentration
ppb-V
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
4.8
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
24.6
Date
Run
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/27/89
7/27/89
7/27/89
7/28/89
7/28/89
7/28/89
7/28/89
7/28/89
7/28/89
8/01/89
8/01/89
8/01/89
8/01/89
8/03/89
8/03/89
8/03/89
8/03/89
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/25/89
7/27/89
7/27/89
7/27/89
7/27/89
7/28/89
7/28/89
7/28/89
7/28/89
7/28/89
8/01/89
8/01/89
8/01/89
8/01/89
8/03/89
8/03/89
8/03/89
8/03/89
Model
411
Response
(unitless)
without
Pre-Filter
0.003
0.003
0.003
0.003
0.003
0.003
0.004
0.003
0.004
0.004
0.002
0.001
0.002
0.002
0.002
0.002
0.002
0.002
0.003
0.002
0.002
0.002
0.002
0.002
0.012
0.011
0.010
0.015
0.016
0.014
0.012
0.013
0.013
0.012
0.011
0.011
0.009
0.009
0.010
0.008
0.010
0.011
0.011
0.010
0.009
0.008
0.008
0.009
Model 411
Response
(unitless) with
silver-coated
Chromosorb
Pre-Filter
0.003
0.002
0.002
0.003
0.002
0.003
0.003
0.003
0.003
0.003
0.002
0.002
0.002
0.003
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.002
0.003
0.016
0.012
0.009
0.011
0.011
0.011
0.011
0.013
0.011
0.011
0,011
0.009
0.009
0.008
0.008
0.008
0.010
0.010
0.010
0.009
0.008
0.008
0.008
0.009
611
-------
4 j\ silver-coated Chromosorb tube was utilized as
a pre-scrubber to remove elemental mercury.
S.The calibration switches were adjusted to
calibrate the instrument to a known concentration
ofDMM.
SENSOR STATUS DRIFT IN THE MODEL
411
It was observed during method development that
the Model 411's sensor status would first increase
after the instrument detected DMM, then
decrease after a period of time. The sensor status
is an indication of the percent gold film saturation.
The increase and subsequent downward drift in
sensor status was not encountered with elemental
mercury. Sensor status drift resulted in the instru-
ment indicating readings lower than actual con-
centrations and was corrected for by allowing the
instrument to balance the Wheatstone bridge be-
tween the sample gold film and the reference film
prior to monitoring another sample. This was
accomplished by drawing four 20-second samples
into the instrument through a iodized charcoal
filter. The filter effectively adsorbs organic and
inorganic mercury resulting in mercury free sweep
air. The number of mercury free air sweeps re-
quired to permit the instrument to re-establish
baseline resistance was determined empirically to
be four.
LINEAR RANGE OF THE MODEL 411 FOR
THE DETECTION OF DMM
The linear range of the Model 411 for the detec-
tion of DMM was determined by diluting a 13.70
ppb-V DMM standard down to approximately
one half the Threshold Limit Value (TLV) of 0.01
mg(Hg)/M3. Dilutions were made to 0,0.64,6.40,
and 13.70 ppb-V and were validated by GC/MS
analysis.
The DMM dilutions were measured with the
Model 411 and the data utilized to generate a
calibration curve (Figure 1). The calibration
curve was found to be linear, with a critical cor-
relation coefficient (R ) of 0.9S. This was deemed
an acceptable linear range for this work effort.
DISCUSSION AND CONCLUSIONS
This study indicates that gold film mercury vapor
detectors have definite potential in monitoring for
DMM. The Model 411 appears promising be-
cause of its simplicity, stability and effectiveness
as a screening tool. However, as with any screen-
ing device, it should not be relied upon exclusively,
rather, it should be incorporated into a multi-
tiered sampling and monitoring program.
612
-------
a>
u
o£ E
0. 15
°-10
0.05-
0.00
FIGURE 1
MODEL 411 RESPONSE VERSUS
DIMETHYL MERCURY CONCENTRATION
(ppb-V)
GRAPH
I I I | I I I I | I I I I | I 1 I I | ''' I I ' ' ' ' 1^" ' ' 7 ' rT 1 I J T T I | I I I I | I I I I | 1 I I I | I I I I |
0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00
DMM CONC. (ppb-V)
Note: 1 ppb—V is approximately 0.01 mg/m3 for DMM
DMM CONC.
(ppb-V)
0
0.64
6.40
13.70
METER RESPONSE
(mg/m3)
o
0.008
0.043
0.126
LINEAR REGRESSION
VALUES:
R2 = 0.98
y Intercept = 0.00220
Standard error of
y = 0.010048
Slope = 0.008959
Standard error of
x = 0.000911
-------
APPLICABILITY OF THIN-LAYER CHROMATOGRAPHY
TO FIELD SCREENING OF
NITROGEN-CONTAINING AROMATIC COMPOUNDS
William C. Brumley and Cynthia M. Brownrigg
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory-Las Vegas
Quality Assurance and Methods Development Division
P.O. Box 93478, Las Vegas, NV 89193-3478
BACKGROUND
Nitrogen-containing aromatic
compounds (NCAC's) are toxic
and mutagenic environmental
contaminants of widespread
occurrence. Often, their
presence in soil or sediment
is a result of wood preserving
activities involving creosote.
Previously, we have used thin
layer chromatography (TLC) to
effect compound class
separations of NCAC's from
polynuclear aromatic
hydrocarbons (PNA's). Prior
to the TLC separation,
contaminants were extracted
from the soil sample and
divided into an HCL and
neutral fraction. The neutral
fraction in particular was
subjected to preparatory TLC
to isolate neutral NCAC's
without interfering
polynuclear aromatic
hydrocarbons (PNA's).
It seemed feasible to apply
TLC to field screening of
NCAC's and other compound
classes such as PNA's.
EXPERIMENTAL SECTION
TLC
E. Merck silica gel 60
preparative plates were used
that were 1 mm thick and 20 X
20 cm size with pre-
concentration zone.
Aldrich silica gel 60 plates
were used with 0.25 mm
thickness and 5 X 20 cm in
size.
Primary solvent systems were
30:70 methylene
chloride:hexane for neutral
NCAC's, 30:70:10 methylene
chloride:hexane:propanol for
the basic NCAC's, and 30:70:10
methylene chloride:
hexane:isopropyl ether for
combined fractions.
GC/MS
A Finnigan-MAT 4021 was used
in the electron ionization
mode with source temperature
270°C. The mass range scanned
was m/z 50-450 in 1 sec. A 30
m DB-5 column was used and
temperature programmed from
60-300'C
at 20*C/min.
615
-------
RESULTS AND DISCUSSION
Fractionation
Samples were available from
Soxhlet, sonication, or
supercritical fluid extraction
of soils and were analyzed in
methylene chloride solution.
A fractionation scheme was
used to separate NCAC's from
PNA's (Fig. 1). This scheme
afforded two fractions: the
basic compounds called the HC1
fraction and the neutral
compounds called the neutral
fraction. Both fractions were
free of interfering PNA's as
determined by GC/MS.
The neutral fraction had been
subjected to preparative TLC
in order to free the
cyanoarenes and
indole/carbazole derived
molecules from interfering
PNA's. The Rf range of 0.05-
0.32 was scraped (Fig. 2).
This region therefore provides
screening capability for
neutral NCAC's in soil. The
HC1 fraction could be
subjected to TLC determination
(Fig. 3) and was free of
interfering PNA's as
determined by GC/MS.
Validation by GC/MS
Fig. 4 and 5 provide the total
ion current chromatograms for
the HC1 and neutral fractions
of NCAC's from a soil heavily
contaminated by creosote.
Selected compound classes are
labeled in order to facilitate
comparison of retention
behavior and relative amounts
of NCAC's present.
Advantages of TLC
TLC offers a low cost, rugged,
simple, and efficient method
to screen for target compounds
such as NCAC's. Greater
selectivity can be achieved
than that illustrated by
incorporation of a third
solvent such as isopropyl
ether in place of propanol.
A great advantage of the use
of TLC is the multiple sample
capability. Up to 40 samples
could be run on a 20 X 20 cm
plate. This is a clear
advantage over HPLC methods.
By going to greater
complexity, automation can be
built into the methods.
Automatic spotting,
densitometry, and multiple
development are some of the
options available.
CONCLUSION
TLC offers a simple and
economical way to do field
screening of multiple samples.
This technique is by no means
limited to NCAC's.
Generalization to PNA's and
other aromatic compounds is
obvious. Non-aromatic
compounds can also be
determined through the use of
visualization reagents or in
situ derivatization reagents.
NOTICE
Although the research
described in this report has
been funded by the U.S.
Environmental Protection
Agency, it has not been
subjected to Agency review
and, therefore, does not
necessarily reflect the views
of the Agency and no official
endorsement should be
inferred. Mention of trade
names or commercial products
does not constitute
endorsement or recommendation
for use.
616
-------
Figure l. Fractionation scheme
to separate NCAC's (both basic
and neutral compounds) from PNA's.
M *_•<.: 1 v
(neul
nil
McCl dr led on A 1 pi-.i
|
o pll 14
*nd oxir.ict
McCl
T
McCl/HfK (
01 / 70%) Kxlr.u-ti'd w 1 1 li McCl
Scrape band
Analytes (Nl t riles, etc. ) 1
Ext racted with Hc-Cl
TLC
(Neutral Soil Sample)
Ml IHYlCMitiA/OI !•-
OA2OLE-
CYANOPHENANTHni Nl -
CYANONAPHIMAI I 'J!
I'NA1'.
DANCil. Sl.UAI'l..()
•CNA':-,
TLC
(HCI Standard Sample)
13ENZOOUINOLINE
METHYLOUINOLINC
V
o
-o o
o
ACRIDINE
OUINOLINF
Figure 2. Preparatory TLC isolation
of neutral NCAC's in the presence
of PNA's; solvent system 70:30
hexane:methylene chloride.
Figure 3. TLC of the HCL fraction
(basic NCAC's) illustrated with
standards; solvent system 70:30:10
hexane:methylene chloride:propanol,
617
-------
ALKYL OUINOLINES
Figure 4. GC/MS total ion trace of the HCL fraction (basic
NCAC's), m/z 50-450, electron ionization.
T
280
3:2!
CYANONAPHTHALENES
CVANOPHENANTHBENE
eit
16H8
i2ea sew
2«:B3 TIME
Figure 5. GC/MS total ion trace of the neutral fraction
(NCAC's), m/z 50-450, electron ionization.
618
-------
ASSESSING THE AIR EMISSIONS FROM A CONTAMINATED AQUIFER
AT A SUPERFUND SITE
Sella Burchette and Thomas H. Pritchett
U.S. Environmental Protection Agency - ERT
Edison, NJ
Steven Schuetz
IT Corporation
Edison, NJ
Kristen Harvey
Roy F. Weston, Inc.
Edison, NJ
The Environmental Response Team was asked by Region II
to assess the degree, if any, that vapors were migrating from
a contaminated aquifer through the vadose zone to the air. If
such migration was occurring, the Region also wanted to
know what would be the worst case long term average air
exposure to the surrounding residents. Even though the
emissions may not have been large at any point, the fact that
these emissions could be occurring over a large area raised
the possibility of an overall significant, long-term air emis-
sion problem.
The ERT's sampling approach involved the taking of flux
measurements over several transects. The flux measurements
were taken using a modified sampling system that would
easily switch from purging the system to filling a Tedlar bag
without any changes in the flowrates. The results from the
portable GC analyses of the Tedlar bag samples were then
used to compute flux rates (ug/sec/m2) at each point.
These flux values were then kriged, and the results were
plotted in order to determine the overall area of concern. The
intermediate kriging output file was then used to calculate an
average flux rate for the area of concern. This average flux
rate was then converted to an overall area source emission
term (total g/sec) that was then plugged into a long-term
exposure air dispersion model in order to estimate the long-
term average exposure of the nearby residences. This final
set of numbers were given to the Region for a subsequent
quantitative risk assessment.
619
-------
CALCULATION AND USE OF RETENTION INDICES FOR
IDENTIFICATION OF VOLATILE ORGANIC COMPOUNDS
WITH A MICROCHIP GAS CHROMATOGRAPH
K.R. Carney, E.B. Overtoil, R.L. Wong
Institute for Environmental Studies
Louisiana State University
Baton Rouge, LA
Introduction
A major difficulty in using elution time data
for component identification is the high
variability of such data with changes in
chromatographic conditions. The use of
retention indices improves the situation
somewhat but a degree of variability
remains. Reproducing Kovat's retention
indices with different instruments in a
laboratory setting can be difficult and such
difficulty are increased substantially during
field operations.
The Microsensor Technology Inc. model
M200 gas chromatograph is a
microprocessor controlled instrument
constructed from micromachined injector
and detector assemblies along with
microbore capillary columns. Independent
column heaters are capable of controlling
the temperature of each column to within
0.1 °C over a range of 30° to 180°C. The
resulting instrument is not only highly
portable but is capable of generating highly
reproducible retention data on both a
"between days" and a "between
instruments" basis.
The dual column capability reduces the
likelihood of coincident elution times for
different compounds and thereby increases
the reliability of component identifications.
A software package developed at LSU uses
a two tiered standardization technique to
provide even more consistent retention
data and then uses that data to generate
qualitative information.
Discussion
A time series of 220 separate analyses of
pentane, hexane and heptane was
performed over a period of 4 weeks.
Approximately 20 samples per day were
run for three one week periods having a 4
day interval between weeks. Variations in
retention time were below one percent for
the four weeks were on the order of 0.2
percent in any one week. This kind of
stability justifies the use of an external
retention index standard.
The corresponding retention index data
show extremely consistent values over a
several week period. Consequently,
retention indices may be determined in the
laboratory and used in the field without
bringing authentic standards of all possible
sample components. Variations in Kovat's
retention indices over the four week period
had standard deviations of less than 0.5
units.
621
-------
An M200 equipped with an OV-73 column
and an OV-1701 column provides
multidimensional elution time data . The
rather narrow dispersion of compounds
about a straight correlation line indicates
that the retention indices on these two
columns are moderately correlated; a
considerable area in the detection space is
empty. These two columns do not differ
greatly in polarity, thus one might expect
such a correlation. As a result the
dimensionality of the space is closer to 1.3
than 2. Note however, the substantial
increase in resolving power versus that for
either of the columns singly (i.e., the
projection of the space onto one axis). Using
a conservative estimate of a peak capacity
of 70 for a typical 100 second
chromatogram and a dimensionality of 1.3,
this system has a peak capacity of
approximately 250. Even with the
increased resolving power of two
columns,however, an unresolvable pair of
compounds is easily found. Such cases
could be further minimized by optimizing
the choice of which two stationary phases
used to obtain the detection space.
Estimation of the retention index from
retention time data requires an estimate of
the column dead time or gas holdup time.
The accuracy of this estimate can
substantially bias the resulting retention
index values for early eluting compounds.
A variation of ±2.5% in the dead time
estimate for column temperatures of 40°C
results in greater than 10% error in
retention index calculations at retention
indices of approximately 200. The value of
I at which the departure from linearity
becomes significant increases with
temperature (approximately 100 units per
20° change). Typical variance we have
observed in dead time estimates.have been
approximately 2.5%.
Significantly, using the elution time of the
air peak consistently overestimates the
dead time. Comparing the elution times of
air and hydrogen suggests that air is
actually retained to the extent of 0.1 to 0.2
seconds. This means that using air as to
estimate dead time consistently results in a
nonlinear function that systematically
overestimates retention indices.
Consequently, we use an iterative method
(1) which estimates to by linearizing the log
(t-to) vs I function. The result a linear
function which in essence recalibrates the
retention index library to current
conditions.
Temperature dependence of retention
indices were observed to range from near
zero for some nonpolar compounds to
approximately 2 units per degree for
alcohols. Oxygen containing compounds
showed negative correlations with
temperature, in contrast with non oxygen
containing compounds regardless of
polarity. This leads to the possibility of
using temperature dependence information
as a tool for identifying compound classes, i:
not specific compounds.
Conclusions
The use of retention index library concept
with the M200 should provide a reasonably
reliable screening tool for sample
component identification. The high degree
of reproducibility in retention data for one
instrument should make libraries prepared
in the laboratory field deployable. Based or
our experiences with the M200, there is a
good possibility that the 120 component
library created at LSU will work with a
large proportion of the M200 instruments
built to date.
An inherent limitation in the technique
presented here is a lack of dimensionality
due to the similarity of the two stationary
phases used. While the resolving power is
quite good as is, improvements can be mad<
by optimizing the choice of stationary phase
and/or implementing temperature
programming.
Implementation of temperature
programming is problematic with the
622
-------
micro-thermal conductivity detector used in
the M200. Baseline drift has been a
significant problem. We have recently been
successful in obtaining temperature
programmed chromatograms with the
M200 with thermal conductivity detection.
We expect this to result in a tremendous
increase in peak capacity and overall range
of analytes.
References
1. R.E. Kaiser, Chromatographia,2,
217 (1969).
2. G. Guiochon, C.L. Guillemin,
Quantitative Gas Chromatography,
Elsevier, New York (1988).
623
-------
Determination of PCB's bv Enzyme Immunoassav
MaryAnne Chamerlik-Cooper
Robert E. Carlson
ECOCHEM Research, Inc., Chaska, MN
Robert O. Harrison
ImmunoSystems, Inc., Scarborough, ME
ABSTRACT
A competitive inhibition Enzyme
ImmunoAssay (EIA) has been developed for
the determination of PolyCJilorinated
Biphenyls (PCB's). The test is capable of
analyzing for PCB's in the field in 15
minutes (from prepared sample), using no
specialized equipment. The test specificity is
restricted to PCB's, with high sensitivity for
Aroclor's 1016, 1232, 1242, 1248, 1254,
and 1260, and moderate sensitivity for
Aroclor 1221. Matrix and solvent
interferences are minimal. The test is capable
of direct analysis of PCB's at low ppb levels
in water. A rapid extraction technique using
DMF, DMSO, or methanol gave a mean
recovery of 70% of Aroclpr 1248 spiked into
non-oily soils. Soils spiked with Aroclor
1248 in transformer oil (to a final
concentration of 50 ppm) and extracted with
DMSO or 4 other solvents showed much
lower recoveries, though still adequate to
produce a strong test response. These spiked
soils were easily distinguished from mock
spiked soils (transformer oil only). Semi-
quantitative estimates of PCB levels were
made using an approximate correction factor
based on the oil-DMSO partitioning behavior
of Aroclor 1248. Improved soil extraction
methods are being developed, but the present
rapid extraction and EIA should be suitable
for PCB screening of soils in many field and
laboratory situations.
METHODS
Reagent Development
The development of the EIA for
PCB's followed these steps: 1) PCB
derivatives were synthesized for conjugation
to proteins; 2) one of these PCB derivatives
was conjugated to a carrier protein and the
resulting conjugate was used to immunize
animals, which then produced antibodies
recognizing both the PCB derivative and
PCB's; 3) a PCB derivative was conjugated
to horseradish peroxidase (HRP) to make a
conjugate which can be captured by anti-PCB
antibodies; 4) the PCB-HRP conjugate was
used to screen and select antibodies; 5) the
selected system was optimized for sensitivity
and matrix tolerance and characterized for
specificity. Also required but at present only
partially completed are the following steps:
6) develop sample preparation methods for
specific sample types; 7) validate using field
samples.
PCB EIA Procedure
The following procedure was used for the
analysis of samples containing PCB's: 1)
rabbit antibodies which recognize the PCB
structure are immobilized on the walls of
plastic test tubes or microwell strips; 2)
samples or calibrators are added to Assay
Diluent in tubes or wells, allowing PCB's to
be captured by the immobilized antibodies.
PCB's are retained on the solid phase when
the rest of the sample is washed away; 3)
PCB-enzyme conjugate is added to tubes or
wells and bound in the same manner as in
step 2. The unbound conjugate is washed
away and the amount retained by the
immobilized antibody is inversely
proportional to the amount of PCB bound in
step 2; 4) enzyme substrate and chromogen
are added to the tubes or wells for color
development by the bound enzyme. The
intensity of color is also inversely
proportional to the amount of PCB bound in
step 2. Therefore, more color means
less PCB.
625
-------
Field Soil Extraction
Soil samples were extracted for analysis by
the following procedure: 1) place soil into
syringe fitted with plastic frit prefilter in the
bottom of the barrel and a 0.2 uM filter. Tap
to allow soil to settle, insert plunger and
press lightly to tamp surface. Fill to 1.5 mL
mark for 2 g soil. Remove plunger and place
Luer cap on filter tip; 2) Add2mLDMSO
or other solvent, re-insert plunger, and shake
to break up soil plug. Time one minute from
plug breakup; 3) Remove Luer cap and
express solvent from syringe. Only a small
volume of filtrate is required since the extract
will be diluted for EIA. Capture filtrate in
clean glass tube or drip one drop (30 uL)
directly into Assay Diluent in antibody coated
tube (step 2 of EIA procedure).
RESULTS AND DISCUSSION
Matrix and Solvent Tolerance
The EIA interference of methanol
extracts of seven PCB-free soils was tested
using two EIA formats. The extracts were
diluted into Assay Diluent for EIA. The
sequential test was performed as described in
the EIA Procedure section above. The
simultaneous test combined the first and
second incubations (sample and PCB-HRP
conjugate on antibody-coated tube at the same
time). At the end of that incubation, the tubes
were washed and the normal procedure was
resumed at this point. The sequential test
was unaffected (>80% of control) by 1:4
dilutions of extract, while the simultaneous
test was strongly affected at 1:10. Similar
data were obtained for DMSO extracts of the
same soils. Additionally, the sequential
assay format tolerated DMSO up to 50% in a
similar experimental design.
These data show that the sequential
EIA described offers excellent resistance to
the effects of concentrated sample extracts,
superior to the simultaneous test. This in
turn means that extracts of low PCB samples
can be assayed with minimal matrix and
solvent effects, by increasing the amount
added to the test in the sample incubation
step.
Assay Precision
Standard solutions of Aroclor 1248 in
DMF were diluted 1:100 into Assay Diluent
for EIA analysis. The test was performed as
described in the EIA Procedure section
above. Data were calculated as a percent of
the control for each calibrator, then means
and standard deviations were calculated for
each calibrator. Precision estimates were
made based on 14 runs over 11 days. For
three calibrators of 0, 7, and 50 ppm, diluted
1:100 (final concentrations of 0, 70, and 500
ppb), the means and standard deviations were
respectively 100+/-4, 35+7-2, and 15+/-2; all
data are expressed as percent of the mean of
all negative control absorbance values. This
result shows that the EIA described offers
excellent reproducibility.
Test Specificity
The crossreactivity of the test for
seven commonly detected Aroclor's was
examined. Standard solutions of 200 ppm in
methanol (Supelco) were used to make serial
dilutions in methanol. These standard
solutions were diluted 1:100 into an aqueous
diluent for EIA analysis. Figure 1 shows that
the test recognizes most of the Aroclor's
nearly equally. Based on this 1:100 dilution,
the 500 ppb points (final assay concentration)
correspond to an initial extract concentration
of 50 ppm for soils. Thus, this test will
easily detect all seven of these Aroclor's at 50
ppm based on a 1:100 extract dilution.
Specificity was also tested for
selected specific congeners in the same
manner as for the Aroclor's of Figure 1. The
congeners most strongly recognized were
2,2',5,5' tetrachlorobiphenyl, 2,3',4,4',5
pentachlorobiphenyl, and 2,2',4,4',5,5',
hexachlorobiphenyl. These data show that
the Aroclor specificity reflects the congener
specificity. Biphenyl and several chlorinated
single ring compounds were also tested for
crossreactivity in the EIA. All of these
compounds demonstrated less than 0.5%
crossreactivity compared to Aroclor 1248:
1,2-dichlorobenzene, 1,3-dichlorobenzene,
1,4-dichlorobenzene, 1,2,4-trichlorobenzene,
biphenyl, 2,4-dichlorophenol, 2,5-
dichlorophenol, 2,4,5-trichlorophenol,
2,4,6-trichlorophenol, and
626
-------
pentachlorophenol. This means that more
than 200 ppm of any of these compounds
would be required to give the same test
response as 1 ppm of Aroclor 1248.
Test Sensitivity
Standard solutions in DMF were
diluted 1:100 or 1:250 into reverse osmosis
purified water for EIA analysis. The test was
performed as described in the EIA Procedure
section above. Figure 2 shows a typical
standard curve; data points represent the
means of two tests run on the same day.
Similar results were obtained for repeated
runs of Aroclor 1248 diluted in the standard
Assay Diluent. These data show that this
EIA is capable of direct analysis of PCB's at
low ppb levels in water.
Spike Recoveries Using Field Soil Extraction
Soil samples were spiked, extracted,
and analyzed by EIA to determine the ability
of the test to detect PCB's in soil. Standard
solutions in DMF, hexane, or transformer oil
were used as noted below.
Spikes of 1 mg/mL Aroclor 1248 in
hexane or DMF were made into 2 g samples
of 4 PCB-free soils, giving a final PCB
concentration of 50 ppm. Soils were
extracted using 2 mL of DMF, DMSO, or
methanol and recoveries were determined by
EIA using 1:100 extract dilutions.
Recoveries averaged 70+/-20% for a total of
15 samples. Mock spikes were performed in
the same manner, using the same soils and
solvents. For sixteen samples, the mean
percent of control was 102+/-15% (100% of
control means no PCB and no interferences).
These data indicate adequate recovery for soil
screening and minimal interferences from the
soils tested.
Mock spikes were performed in the
same manner as above, except the spike
material was clean transformer oil in a ratio of
100 uL/2 g soil (5% v/w). For seven
samples extracted with DMSO or methanol,
the mean percent of control was 77+/-8%
(100% of control means no PCB and no
interferences). This indicates interferences
from the soils tested greater than those non-
oily soils described above, but still giving a
signal approximately equal to 1 ppm of
Aroclor 1248 in the original extract (based on
the EIA Procedure step 2 dilution of 1:100).
Spikes of 1 mg/mL Aroclor 1248 in
transformer oil were made into 2 g samples
of 4 PCB-free soils, giving a final PCB
concentration of 50 ppm and a 5% v/w ratio
of oil to soil. Soils were extracted using 2
mL of DMSO and recoveries were
determined by EIA using 1:100 extract
dilutions. Recoveries averaged 12+/-3% for
a total of 4 samples. These values are similar
to those obtained for DMSO extractions of
the spiked transformer oil with no soil
present. Methanol, DMF, THF, and N-
methylpyrrolidone gave slightly less effective
extraction from oily soils than DMSO.
Significant reductions in recoveries of 50
ppm Aroclor 1248 spikes were observed at a
0.1 % v/w ratio of oil to soil.
The extracts of oily spiked soil and
oily mock spiked soil described above were
analyzed by EIA using extract dilution factors
ranging from 1:100 to 1:5. At a dilution
factor of 1:20, or 5% extract in PCB Diluent,
the mock spike response equated to much
less than 7 ppm in the original sample (the
color was much higher than the diluted 7 ppm
calibrator). At the same dilution, the spiked
soil response equated to nearly 50 ppm in the
original sample (the color was similar to the
diluted 50 ppm calibrator).
Based on the above results, oily soils
or suspected oily soils can be analyzed using
a direct DMSO or methanol extraction and
EIA of an increased concentration of the
extract, such as 5%, to partially correct for
inefficient partitioning from the oil phase.
Using this technique, samples could be
confidently screened in the field at the 50
ppm level. Oily soils containing 50 ppm
PCB would give a strong EIA response,
while oily soils with no PCB's would behave
as described for the mock spiked oily soils
described above.
CONCLUSIONS
1. The test is capable of analyzing for
PCB's in the field in less than 20 minutes
(15 minutes from prepared sample),
using no specialized equipment.
2. Test specificity is restricted to PCB's.
627
-------
Aroclor's 1248, 1254, and 1260 are
recognized best; 1242, 1232, and 1016
are recognized nearly as well; 1221 is
recognized significantly less well, but can
still be detected easily at 50 ppm.
Congener specificity of the test reflects
the Aroclor specificity.
The test is capable of direct analysis of
PCB's at low ppb levels in water.
Soils which are not oily can be analyzed
using a solvent extraction and direct EIA
of the diluted extract.
Oily soils can be analyzed using a solvent
extraction and direct EIA of an increased
volume of the diluted extract to correct for
inefficient partitioning from the oil phase.
8. Further work in this area will include
improved extraction from oily soils, oil
analysis, sediment analysis, biological
sample analysis, field testing for the
above, and quantitation of PCB's using
the strip-well method for lab analysis.
ACKNOWLEDGEMENT
The initial phase of development of
this PCB immunoassay was partially
supported by the US EPA through a sub-
contract to ECOCHEM from Mid-Pacific
Environmental Laboratories, Inc.
Figure 1. Crossreactivitv for 7 Aroclors
0 10 100 1000 10000
AROCLOR Concentration (ppb)
Figure 2. EIA of Aroclor 1248 In Water
100
90
o'°
c 70
8 6o
•5 so
C 40
O 30
|i 20
1 0
0
0.1 1 10 100 1000 10000
ppb AROCLOR 1248
628
-------
Practical Limits in Field Determination of
Fluorescence Using Fiber Optic Sensors
Wayne Chudyk
Carol Botteron
Kenneth Pohlig
Rose Najjar
Civil Engineering Department
Anderson Hall
Tufts University
Medford, MA 02155
Phone (617)381-3211
Fax (617)381-3819
Purpose
The long-term aim of our research program is
to develop an instrument useful for the field
determination of aromatic organic contami-
nants. Concentration ranges of interest are
in or below the ppb range. This approach may
be applied to ground water analysis, in both
the saturated zone and in the vadose zone,
and for water or wastewater treatment process
monitoring.
Our earlier work reported on the usefulness
of fiber optic sensors in detection of
aromatic organic ground water contaminants
such as the benzene, toluene, ethylbenzene,
and xylenes (BTEX) fraction of petroleum
fuels. We use a laser fluorimeter with fiber
optic sensors for in-situ measurements. The
lower limits of detection observed for com-
pounds excited in the ultraviolet (266 nm)
appear to be restricted by the optical
qualities of the field instrument, including
the sensor. For example, the dynamic
response range of fluorescence signal versus
concentration is narrow when excitation
occurs in the ultraviolet as opposed to
visible (532 nm) with dye tracer studies
(1, 2, 3).
Optical noise limiting the dynamic range of
the instrument could have sources inside or
outside the instrument. Optical evaluation
of the system has involved fluorescence life-
time analysis, which has produced results in
two areas. The first area is in determination
of the source of instrumental spectral noise
(1). The second area is in application of
fluorescence lifetimes as an identification
and measurement tool (4, 5).
Scope
This report outlines the suspected sources of
spectral noise that limit the dynamic range
of the instrument, and presents some applica-
tion of the use of fluorescence lifetimes
analysis as a measurement tool.
Methods
In-situ measurement of ground-water contami-
nants using unmodified fiber sensors has been
the focus of our work. Our research group
has built and field tested a second generation
prototype instrument. The prototype instru-
ment's theory, construction, and testing
results have been presented elsewhere (1, 2,
3, 6). In summary, it uses a Nd:YAG pulsed
laser (Laser Photonics MYLA laser operating
at 0.5 Hz), with an internal frequency
doubling crystal, as a 532 nm light source.
The 532 nm light is again frequency doubled
to 266 nm with an appropriate crystal (BBO,
Quantum Technology) and coupled to an optical
fiber (Superguide UV 600N, Fiberguide
Industries) that is placed in a protective
tefzel sleeve along with an identical emission
detection fiber. Both fibers terminate in a
stainless steel sensor, which holds the fibers
in the water to be analyzed.
Fluorescence and scattered excitation light
collected by the emission detection fiber is
carried back to the surface for analysis.
Either a monochromator (H-20, Instruments SA,
Inc.) or a set of glass cut-off filters can
be used for the requisite fluorescence light
isolation and detection. Light intensity is
measured either using a photomultiplier (PMT)
(Hamamatsu). A measurement of scattered
629
-------
excitation light is also made to use for power
normalization. When measuring fluorescence
intensity, the PMT current outputs are con-
verted to voltages using an electronic box-
car integrator (EG&G/Princeton Applied
Research), and the voltages are stored in a
portable personal computer (Compaq) used
as a data logger. For fluorescence lifetime
measurements, signals from the PMT's are fed
to a digitizing oscilloscope {LeCroy Model
9450, dual channel, 350 Mhz, 400 megasamples/
second per channel) with 50 ohm input imped-
ance at a sampling interval of 2.5 nanoseconds.
The oscilloscope is triggered by a photodiode
(Hamamatsu 1722 PIN photodiode) which is
illuminated by a fraction of the 266 nm
excitation laser light just before focusing
into the excitation fiber.
Sensor lengths, which correspond to useful
well depths, of up to thirty meters have
been used in the field, with most useful
data obtained with 10 m sensors. Laboratory
investigations have typically used shorter
sensors, with vapor analysis experiments
using fiber lengths of 1.85 m, while solution
analysis experiments usually use sensor
lenghts of 10 m.
Vapor phase analyses were performed by sus-
pending a sensor in a sealed glass dessicator
or flask over a solution of known concentra-
tion of the analyte. Henry's law could then
be used to predict the vapor phase concentra-
tion above the liquid (7).
Results
Fluorescence lifetime measurements on gasoline
samples show that such measurements may be a
useful means of determining solution concen-
tration. A log-log plot of fluorescence
lifetime values versus concentration for
unleaded gasoline shows a straight line for
data from less than 0.01 to more than 500 ppm.
A best-fit line through the data is described
by the equation log (lifetime) = 1.66 + 0.181
log (ppm).
Vapor phase analysis of phenol shows linear
response of fluorescence for phenol vapor
concentrations of 0.1 up to 1000 micrograms
per 1iter of air.
Interferences from turbidity are indicated by
the decreasing fluorescence seen from a
1 mg/1 phenol solution in water when increas-
ing amounts of silt were added to the
solution.
Conclusions
1. Spectral noise in the instrument, origi-
nating in the flashlamp of the laser in
addition to some fiber fluorescence, limits
the dynamic range of the instrument.
2. Fluorescence lifetime analysis may be an
additional parameter useful in determining con-
centration.
3. Vapor analysis suggests that vadose zone
analysis may be possible by performing analyses
on air instead of water.
4. Turbidity interferes with fluorescence of
aqueous solutions, and its scattering effects
may contribute to observed dynamic range
1 imitations.
References
1. Chudyk, W., and K. Pohlig, "Dynamic Range
Limits in Field Determination of Fluorescence
Using Fiber Optic Sensors," SPIE Annual
Conference, San Diego, CA, September 17-20,
1990.
2. Chudyk, W., K. Pohlig, L. Wolf, and
R. Fordiani, "Field Determinatoin of Ground
Water Contamination Using Laser Fluorescence
and Fiber Optics," Proceedings Volume 1172:
Chemical, Biochemical, and Environmental Fiber
Senors, SPIE Annual Conference, Boston, MA,
September 6-7, 1989.
3. Chudyk, W., K. Pohlig, N. Rico, and
G. Johnson, "Field Screening for Aromatic
Organics Using Laser-Induced Fluorescence
and Fiber Optics," EPA/EMSL First Inter-
national Symposium: Field Screening Methods
for Hazardous Waste Site Investigations. Las
Vegas, NV, October 11-13, 1988.
4. Birks, J.B., and I.H. Munroe, "The
Fluorescence Lifetimes of Aromatic Molecules,"
Progress in Reaction Kinetics, 4, 1967, pp.
239-300.
5. Pohlig, K., Design, Construction, and
Testing of a Remote Laser-Induced Fluores-
cence Fiberoptic Groundwater Contaminant
Detector, MS Thesis, Tufts University,
Medford, MA, February, 1991.
6. Nirmalakhandan, N.N., and R.E. Speece,
"QSAR Model for Predicting Henry's Constant,"
Environmental Science & Technology, £2, 11,
November, 1988, pp. 1349-1357.
Acknowledgement
The authors thank the continuing support of
the Alexander Host Foundation.
630
-------
The Colloidal Borescope.
A means of assessing local colloidal flux
and groundwater velocity in porous media
T. A. Cronk
Health and Safety Research Division
Oak Ridge National Laboratory
P. M. Kearl
Environmental Science Division
Oak Ridge National Laboratory
The colloidal borescope is a waterproof video
camera capable of viewing indigenous colloids
in a monitoring well. Through optical
magnification, the movement and density of
these colloids is easily assessed. The
instrument shows promise of providing an
improved methodology for determining both
local groundwater flow velocity and colloidal
transport potential.
Because observations are taken at the
microscale, data obtained are indicative of the
local transport parameters of the subsurface
flow system. By taking numerous
measurements at different spatial locations,
heterogeneities of the flow system in both the
vertical and horizontal sense may be defined.
Preferential flow zones and fractures may thus
be located. The ability to determine the
spatial variability of relevant transport
parameters is also necessary for the effective
use of stochastic transport models.
Field use of the colloidal borescope indicates
that colloidal parameters are very sensitive to
external perturbations. The slightest
disturbance induces pressure waves and
turbulent flow which greatly affect colloidal
density and migration rates. An interesting
field experiment confirming this sensitivity
involved dropping a slug into a well some 5
meters away from a well in which the
instrument was recording steady-state
movement. The results were dramatic as the
colloidal migration pattern literally exploded
into a turbulent flow pattern. One can
imagine the affects even the most gentle
pumping techniques must have on turbidity
and colloid density during sampling. These
observations indicate the instrument provides
a much more accurate method of assessing
natural colloidal densities and migration rates.
Observations to date indicate a steady, laminar
flow field in the borehole which has an
excellent directional correlation with specific
discharge in the surrounding aquifer.
Knowledge of groundwater flow direction
allows direct assessment of the affects
pumping and injection systems have on the
natural flow field. The effectiveness of
groundwater extraction/injection systems may
thus be investigated by observing the flow
direction near the radius of influence of the
system to see if the designed flow field is
being achieved.
Groundwater flow velocities in the
surrounding aquifer are inferred from
potential flow theory relationships between
flow through a wellbore and specific discharge
in the surrounding aquifer. While both field
and laboratory observations confirm a direct
relationship between these two parameters,
these same observations indicate wellbore
631
-------
velocities are consistently higher than those
predicted from potential theory. The
observed relationships suggest potential theory
may not be adequate to characterize the flow
patterns at the scale of observation. The
development of a consistent theoretical
explanation for these observed relationships is
sure to provide new knowledge about the
underlying physical processes involved. The
colloidal borescope thus provides an exciting
new means of investigating the physical
phenomena affecting flow in porous media at
the pore scale. These observations offer a
means of enhancing conceptual understanding
and of developing improved transport models.
632
-------
FIELDABLE ENZYME IMMUNOASSAY KITS FOR DRUGS AND
ENVIRONMENTAL CHEMICALS
Peter H. Duquette, Patrick E. Guire, Mclvin J. Swanson,
Martha J. Hamilton, Stephen J. Chudzik and Ralph A. Chappa
Bio-Metric Systems, Inc.
9924 West Seventy-Fourth Street
Eden Prairie, Minnesota 55344
Abstract
Immunoassays (e.g., RIA, EIA) have been demonstrated to
be useful for rapid, convenient detection and
semiquantitative analysis of drugs and various
environmental pollutants. Bio-Metric Systems, Inc. (BSI)
has developed a rapid, sensitive, self-contained, disposable,
EIA device designed to allow untrained personnel to
perform in field situations. This format has been developed
for drugs in urine or on surfaces and for environmental
contamination in soil or water. The analyte in the test
sample competes with an enzyme-analyte conjugate for a
limited number of immobilized antibody sites. This
AccuPress™ Test format can detect analytes at 10 ppb in
biological fluids, water, and soil, with positive results
indicated by clearly visible color development within 10
minutes. This format is designed to have all dry
components and to have an ambient shelf life of greater
than one year. The format is readily adaptable for use with
numerous low molecular weight analytes.
Introduction
Chemical pollution and the use of illicit drugs in the U.S.
are two of our most important social problems. Monitoring
environmental samples for various man-made hazardous
chemicals has become necessary in order to protect the
populace from the carcinogenic and toxic effects of such
materials. Of particular concern are certain polychlorinated
organics such as pentachlorophenol (PCP), chlorinated
dibenzo-p-dioxins (CDD's), and chlorinated dibenzofurans
(CDF's) which are known to be extremely carcinogenic.
Because of the known association of CDD's and CDF's
with PCP and the possibility that food-producing animals
are exposed to these compounds through treated wood, one
must be aware of the possibility that food and water
supplies can become contaminated by the chemicals.
Likewise, despite all the recent publicity regarding drug
abuse, evidence seems to indicate that the importation and
use of illicit drugs in the United States continues unabated
[1-5]. Contributing factors to this problem are the large
financial rewards generated by the illicit drug trade and the
seemingly insatiable demand by the American drug user.
Because of these factors, law enforcement agencies need to
know which drugs of abuse are most prevalent in
trafficking; and they must possess the analytical capabilities
for the detection of these illicit materials.
Most of the present day methods for analysis of
environmental hazards and illicit drug samples have been
based mainly on conventional chromatography techniques
such as gas chromatography (GC) [6,7], high performance
liquid chromatography (HPLC) [8,9], and gas
chromatography - mass spectrometry (GC-MS) [10,11].
Although these methods give excellent resolution and are
highly sensitive, they exhibit certain disadvantages such as:
1) the need for extensive cleanup treatment before analysis;
2) expense of solvents and instrument maintenance; 3) need
for trained personnel in a laboratory setting; and 4) length
of time for analyses.
The development of immunochemical techniques has added
a new dimension to the detection and identification of low
concentrations of pollutants and drugs. These methods are
based upon the displacement of, or competition with,
labeled analyte from an antibody-hapten complex by sample
analyte and the subsequent detection of the labeled analyte
by instrumental methods. These techniques provide high
sensitivity, good reliability, relatively fast results (usually
within a few hours), and require less expensive equipment.
At present, two different types of immunoassays have been
633
-------
developed: radioimmunoassay (RIA) [12-15] and enzyme
immunoassay (EIA) [16-19] for the detection of cocaine,
heroin, and environmental toxins. Both assays are quite
sensitive, require minimal sample pretreatment, and allow
several assays to be run daily. However, the major
disadvantage of most EIA's is the requirement for
expensive laboratory equipment in a laboratory setting.
Furthermore, with the radioactive technique there are the
additional problems of radioactive waste disposal and
specialized handling.
Experimental Approach
Bio-Metric Systems, Inc. (BSI) has developed prototype
AccuPress EIA test kits for both the analysis of PCP in
water and soil, and cocaine and heroin on surfaces. The
tasks required to produce the kits were: 1) obtain and
evaluate antibody specific to PCP, cocaine, and heroin; 2)
prepare enzyme-hapten conjugates; 3) investigate various
extraction and sampling methods; 4) optimize reagents; 5)
perform component and system stability testing; and 6)
conduct simulated field trials using fortified samples.
Antisera Procurement
Antibodies specific for PCP were purified from antiserum
that had been prepared by immunizing rabbits with either
keyhole limpet hemocyanin (KLH)-PCP or bovine serum
albumin (BSA)-PCP immunogen. The PCP-KLH
immunogen was produced by direct coupling of 2,3,5,6-
tetrachloro-4-aminophenol to KLH by diazotization. An
alternate immunogen was prepared by coupling PCP
through the phenol portion of the moiety to the BSA. This
was accomplished by coupling PCP-valeric acid with BSA
either by: 1) direct attachment of the free carboxylic acid
moiety to the amines on the protein by use of l-ethyl-3(3-
dimethylaminopropyl) carbodiimide (EDC) or 2) coupling
the N-oxysuccinimide ester (i.e., NOS) of the acid by
established methods [20]. Polyclonal antibodies which were
specific to cocaine and heroin were induced utilizing
immunogens prepared by similar procedures to those used
for PCP.
Enzvme-Hapten Conjugate Preparation
One of the most critical steps in the development of our
assay is the formation of the enzyme-hapten conjugate. The
appropriate enzyme-hapten conjugate must be able to: 1)
be bound by the immobilized specific antibody; 2) compete
with the analyte for antibody bonding; and 3) maintain
sufficient enzyme activity to generate a signal when low
concentrations of analyte are present in test samples. We
have developed assays which utilize the enzyme pair-
glucose oxidase (GO)/horseradish peroxidase (HRP). In
previous work we have used GO-hapten conjugates and
immobilized enzymes; however, recently we have prepared
GO-biopolymer conjugates to which were added various
haptens. We have found that this step allows better binding
of the enzyme-hapten conjugate to the immobilized
antibody. The GO-biopolymer conjugate can be prepared
by first oxidizing the biopolymer with periodate, then
adding the GO followed by reduction of the resulting Schiff
bases with NaBH, to yield a stable GO-biopolymer
conjugate. The carboxylic acid derivatives of the hapten
analogs prepared for the production of immunogens were
coupled to the enzyme (GO) or modified GO (i.e., GO-
biopolymer) by one of the following methods: direct
attachment of the analogs (i.e., free carboxylic acid moiety)
to the amines on the protein by use of EDC, or by coupling
of the N-oxysuccinimide ester (i.e., NOS) of the hapten by
established methods [20].
Sampling and Extraction Procedures
One of the biggest challenges to the development of EIA's
for the analysis of PCP in soil and drug residues on hands
and surfaces is to develop suitable sampling procedures for
extraction of the desired analytes. Although hexane or
toluene are used for extraction of PCP residues from soil,
these solvents are not compatible with an immune test
system, such as our AccuPress test format (Figure 1).
However, since PCP is highly soluble in MeOH [21], and
MeOH is compatible with our AccuPress reagents in less
than a 40% concentration [22], MeOH was chosen as the
extraction solvent for use in our assay. The basic protocol
as developed for PCP was the following: Ten gram soil
samples were spiked with 0,20 ppb, 100 ppb, and 500 ppb
PCP in methanol. The PCP was thoroughly mixed into the
samples which were then dried at 37°C for two hours in
order to mimic naturally occurring contaminated samples.
The soil samples were extracted by vigorous shaking for
one minute with 40% MeOH/PBS. The extracts were
filtered through a 1.2 pn filter and the filtrate (1.0 ml)
applied directly to the sample well of the AccuPress test
module. The antibody disk was washed with five drops of
PBS/3% PEG, after which one drop of conjugate was added
and incubated for two minutes. The two segments were
then pinched together and released. The color development
was read after five minutes. Based upon an arbitrary color
scale designation of 0-5, 20 ppb of PCP in soil gave a color
rating of 3. A positive response was easily distinguishable
from the 0 ppb sample which resulted in a colorless readout
(rating of 0).
During the prototype development of the AccuPress test for
drugs on surfaces, BSI developed a sampling vial which
served both as a sampling device and also as an application
device. As depicted in Figure 2, the sampling vial
consisted of a sample swab attached to the dropper top.
Also attached to the dropper top was a filter that would
634
-------
remove any extraneous materials from the sample solution
during application of the test (A). The user would simply
remove the dropper top with the attached swab from the
sample vial and swab the area to be tested (B). Next, the
dropper top was replaced and sampler vial shaken. If
testing a solid, a small amount of the solid was transferred
to the sampler vial, followed by replacing the top, and
shaking the sampler vial. The dropper cap was then
removed and 5-10 drops of the sample to the AccuPress test
(C). The development of this sampler made the prototype
AccuPress tests very convenient to use.
AccuPress Test Format
The AccuPress test format (Figure 1) consists of four parts:
1) antibody disks (A); 2) read-out disks (B); 3) absorbent
blotting reservoir (C); and 4) a crush vial containing
lyophilized antibody, while the read-out disks (B) contain
an immobilized enzyme (HRP), the chromogen (ABTS),
and a substrate (glucose) for the enzyme-hapten conjugate.
Also, a reservoir (C) containing an absorbent pad is located
beneath the antibody disks, and the enzyme-hapten
conjugate corresponding to the desired analyte is
lyophilized in a small crush vial (D). The user simply
reconstitutes the lyophilized conjugate (D) by squeezing the
tube, crushing the ampule, allowing the conjugate enough
time for complete dissolution (ten seconds). Next, five to
ten drops of sample are applied to the antibody disks and
the disks are rinsed with wash solution to remove any
extraneous material. One drop of the reconstituted
conjugate is added to the antibody disks (A) and incubated
for one to two minutes. The user then folds the top plate
containing the read-out disks (B) over the bottom plate
containing antibody disks (A), pinches for approximately
three seconds, and the results are read in five to ten minutes
with a positive result being indicated by color formation.
This assay format exhibits distinct advantages over other
enzyme immunoassay formats (e.g., ELISA). First, the
sample size is less limited since by exposing the antibody
disk to a large volume of sample (i.e., up to 800 |il), the
analyte can be concentrated on the disk, thus increasing the
sensitivity of the assay (Figure 1). Second, a wash step
with PBS allows any possible interfering substances still
present in the environmental sample to be washed off the
antibody disk. Third, the enzyme-hapten conjugate can be
added after the sample, which should also increase
sensitivity. Finally, this format allows one to use controls
more easily and could also be easily adapted as a multi-
analyte assay.
Reagent Optimization
Since the reagents for both the AccuPress test for PCP and
the test for drugs on surfaces were prepared by similar
methods, only generalized procedures will be described
using the PCP test as an example.
In order to carefully control the amount of immunoglobulin
(IgG) coupled to the antibody disks, the antiserum was
purified to >95% IgG by standard methods. The antiserum
was fractionated with saturated ammonium sulfate (SAS),
pH 7.8, by addition of an equivalent volume of SAS to neat
antiserum (50% saturation). After stirring for two hours at
room temperature, the antiserum was centrifuged. The
pelleted material was redissolved to one-half the original
volume with 20 mM phosphate buffer, pH 7.2, and dialyzed
exhaustively against 20 mM phosphate, pH 7.2. The
dialysate was then purified over Whatman DE-52 anion
exchanger. The IgG peak was pooled and characterized for
total protein (Pierce BCA Protein Reagents) [23] and for
total IgG (ICN Rabbit IgG radial immuno diffusion kits)
[24]. The IgG was then prepared for coupling to paper
disks or stored in PBS, pH 7.2, at -70°C, in 100 mg
aliqdots.
Antibody disks were prepared and evaluated to determine
the amount of specific antibody needed to be coupled to the
disk in order to obtain the desired sensitivity. Antibody
disks were prepared using ratios of DE-52 purified antibody
per 50 disks. The levels investigated were 20 mg/50 disks,
10 mg/50 disks, 5 mg/50 disks, and 2.5 mg/50 disks. A
study was done to determine which load of antibody could
be used in our test format to optimally achieve the desired
sensitivity. The coupling efficiency or the optimal amount
of specific antibody/disk was determined by measuring the
protein concentration [25] before and after coupling and by
radiolabeled uptake experiments using radiolabeled analyte.
This radiochemical procedure involved the incubation of the
antibody disk with [14C]-analyte for 45 minutes. The
amount of radioactive analyte bound to the antibody
allowed one to calculate the pmoles of analyte bound per
antibody disk.
As previously noted, one step in the development of our
assay is the production of the enzyme-hapten conjugate.
The amount of hapten on the enzyme-hapten conjugate is
critical for achieving the required characteristics. If
insufficient or excessive hapten groups are coupled to the
enzyme, the conjugate either will not bind adequately to the
antibody, or it will bind so well that native analyte cannot
compete effectively. Thus, hapten coupling experiments
had to be performed to determine the range of hapten
groups needed on the enzyme to achieve the desired binding
characteristic. Several conjugates were prepared by
coupling hapten-NOS to GO using various molar ratios of
hapten to enzyme (e.g., 20, 50, 100X) in the reaction
mixture. Analysis of the number of remaining amines by
a standard 2,4,6-trinitrobenzene sulfonic acid (TNBS) assay
[26] after coupling of the hapten-NOS to the enzyme, as
635
-------
compared to unmodified enzyme, indicated the approximate
load of the hapten on the enzyme (i.e., 50% amine
reduction). The enzyme-hapten conjugates prepared using
100 mg of hapten derivative to 50 mg GO consistently
produced the best conjugates for our EIA development with
regard to enzymatic and immunological activity.
Read-out disks were prepared using HRP covalently
coupled to chromatography paper disks through a diamine
spacer. The modified HRP was prepared by first oxidizing
the carbohydrate portion of the HRP with periodate, then
adding the diamine, followed by reduction of the Schiff
base with NaBH^ to yield a stable HRP-diamine derivative.
Enzymatic activity was measured before and after amine
modification (Table 1). After immobilization of the
enzyme, the disks were incubated with a solution of
glucose, chromogen (2,2'-azinobis[3-ethylbenzothiazoline-6-
sulfonic acid], ABTS) and proprietary stabilizer in PBS/1%
polyethylene glycol (PEG) 4000 for 60 minutes. The liquid
was decanted and the excess moisture was wicked away.
The disks were placed in trays, frozen at -70°C for 20
minutes, and lyophilized overnight. The disks were stored
at room temperature in a humidity controlled room (19%
R.H.) until used. Accelerated stability studies at elevated
temperatures (i.e., 37°C, 55°C) indicated that the read-out
disks are quite stable with only a 20 to 30% loss of enzyme
activity after 34 days at 55°C (Figure 3). The disks
performed well in the assay, showing little or no
background in the absence of free analyte, while giving an
easily observable response when analyte was present.
Component and System Stability
Our AccuPress EIA test format (Figure 1) consists of three
components: antibody disks, read-out disks, and lyophilized
enzyme-hapten conjugate, all containing biological reagents
whose activities can be affected by environmental factors
(e.g., temperature and humidity). Because of this, it is
difficult to prepare a rapid enzyme immunoassay suitable
for field use which would exhibit a minimum shelf life of
one year when stored at room temperature. Therefore,
stability testing was carried out on the individual
components and whole test kits at various time points
during storage at four different temperatures. Although
both AccuPress test kits and components for both drugs and
PCP were evaluated, only the PCP test kit data is presented,
which is quite representative of both test kits.
Antibody disks were prepared as previously reported,
lyophilized, and stored for two weeks before the stability
study was initiated. The disks used for the radiolabeled
uptake study were packaged in bilaminar foil pouches with
a desiccant pack and a N2 infusion immediately before heat
sealing. The packages were stored in the appropriate
temperature controlled environments for the duration of the
study. The testing protocol was as follows: A standard
stock solution of 10 dpm/nl of [14C]-pentachlorophenol was
prepared and stored at -20°C. On test days, an aliquot of
stock solution was diluted in 0.01 M phosphate, 150 mM
NaCl, pH 7.2. The antibody disks stored at various
temperatures were allowed to equilibrate to room
temperature. One ml of diluted radiolabeled PCP per
antibody disk was incubated in 1.5 ml microcentrifuge tubes
(five PCP disks and five control-normal rabbit serum
antibody disks) for two hours with shaking on an orbital
shaker. After incubation, 500 ^1 of the supernatant was
removed from each tube for liquid scintillation counting.
The quantity of bound PCP was expressed in pmoles/disk.
The stability of the enzymatic activity of the GO-hapten
conjugates was also investigated. A stock conjugate reagent
was packaged in crush vials for use in the stability testing.
The appropriate dilution of the conjugate was determined
using the appropriate antibody disks. A concentrated form
of the stock conjugate was added to the stabilization media
and aliquoted into polyethylene tubes for lyophilization.
After lyophilization, a reconstitution buffer, encapsulated in
an onion skin glass vial, was added to the tube. The
dropper top with filter was applied to the top of the tube.
The vials were packaged in foil pouches with a desiccant
packet and the test module, and stored at the selected
temperatures until tested by direct enzymatic analysis. Four
separate crush vials per storage temperature were
reconstituted and tested by the Worthington Kinetic
Glucose-Oxidase assay. The same vials were used for
performance testing in the whole kit stability evaluation.
The mean of the determined rates for four vials (per temp-
erature per time point) were converted to specific activity.
At each time point, an aliquot of native GO was assayed in
quadruplicate as a control. The stock of native GO was
prepared at 1 mg/ml, aliquoted and stored at -70°C.
The stability of the enzymatic activity on the read-out disks
at various elevated temperatures was also investigated.
Read-out disks were incubated with stabilizers and
lyophilized. The dry read-out disks were stored in foil
pouches with a desiccant packet and flushed with N2 before
sealing. The pouches were stored at the selected
temperatures until tested by either direct application of a
standard amount of enzyme or were assembled into kits for
performance testing. For the actual evaluation, the disks
were allowed to equilibrate to room temperature and were
stored in a desiccator until tested. A standard aliquot of GO
(the same as the control used for the Worthington kinetic
assay) was diluted to 1 |ig/ml. The disks to be tested were
laid out on a white sheet of paper and 15 Hi of 1 Jig/ml GO
was added to each disk. Color development was monitored
and recorded at 1 minute and 5 minutes compared to an
arbitrary color chart having five spots of progressing color
intensity.
636
-------
Assembled AccuPress test kits were prepared and evaluated
in an accelerated temperature study. The preparation of the
kits and testing protocol are presented as follows:
Preparation of Kits: Antibody disks were prepared using
ammonium sulfate preparation of antisera at an IgG level
corresponding to 50 disks per 5 ml of whole antisera with
four different stabilization formulations. These disks were
then evaluated for performance. Read-out disks used for
the whole kit testing were prepared as previously described.
Conjugate was titered to match antibody disks and
lyophilized in polyethylene tubes. Glass onion skin vials
with premeasured aliquots of PBS were placed in the tubes
and they were sealed with a Porex filter and dropper top.
Kits were assembled in the dry room (14% relative
humidity) and sealed in bilaminar foil pouches after N2
flush.
Testing Protocol: Stock standard PCP was diluted into
40% MeOH/PBS to a concentration of 100 ppb (= to 100
ng/ml). Negative control was 40% MeOH/PBS. The testing
protocol was as follows:
- 10 drops (0.25 ml) of positive control were added to
the positive well.
10 drops (0.25 ml) of negative control were added
to the negative well.
8-10 drops of neutral pH wash solution were added
to all wells. (Allergan Lens Plus®).
Conjugate was reconstituted by squeezing crush vial
and shaking vigorously.
- One drop of conjugate was added to each well.
It was incubated for 2 minutes.
The module was pinched together for 2-3 seconds.
Color development was monitored for 10 minutes,
and recorded at 5 and 10 minute intervals.
The results of our component and whole test module
stability studies using the PCP format as a model
demonstrated that we had excellent stability for both the
components and the whole test module. For example, we
were able to demonstrate a 95% retention of PCP antibody
activity after storage at 55°C for one month (Figure 4).
Similarly, the enzyme-hapten conjugate and read-out disk
also were stable when stored at 55°C for one month
(Figures 5 & 6). Examination of the performance of the
whole test modules (Figure 7) indicated that the modules
were quite stable at 55°C for 35 days. It is apparent that
the components and the whole test modules are stable and
exhibit a shelf life of at least one year when stored at room
temperature.
Reliability Testing
During the in-house testing, we investigated the
reproducibility of the test device in detecting PCP residues
in soil. The four types of soil (sand, clay, black dirt,
Minnesota river sediment) were collected and dried. Clay
was heat dried in a vacuum oven. Black dirt, sand and.
river sediment were air dried. All soil samples were sieved
through a #14 mesh screen. The soil samples were weighed
out at 10 gms per vial and spiked with PCP at varied
concentrations in 100 (0.1 of methanol to equal 0, 10 ppb, 50
ppb, 100 ppb, 1000 ppb. Ten samples were prepared for
each PCP level per soil type. Each soil extract was tested
in duplicate in a blind study.
Extraction & Assay Procedure;
To each 10 gram sample, add 10 ml of 40%
MeOH/PBS.
Shake the vials vigorously for 1 minute.
Allow the sediment to settle for a minimum of 5_
minutes.
Filter the supernatant using a syringe, through an
ED-141 prefilter placed in the syringe barrel and a
1.2 \i S&S Uni-Ro filter.
- Add 0.5 ml of filtered extract to the module sample
well.
- Wash with 8-10 drops of neutral pH
buffer.(Allergan Lens Plus).
Crush conjugate ampule and shake to reconstitute.
- Discard the first drop of conjugate in a waste
container.
Add 1 drop of conjugate to each well.
Incubate the conjugate for 2 minutes.
Pinch the module together for 2-3 seconds.
Monitor the color development for 10 minutes,
recording the 5 and 10 minute color. A chart with
5 spots of increasing green color intensity is used as
a reference.
Since the AccuPress test is intended as a qualitative screen,
we have used the following definition of positive and
negative results. If a color less than or equal to 1.0 on our
5 step color chart develops in 5 minutes, then the results are
classified as negative. Any test which develops a color
greater than a 1.0 in 5 minutes is classified as positive.
When applying this rule to the above samples, the following
conclusions were drawn:
1. For sand: No zero analyte samples generated color
greater than 0.25; there were no false positives; 90%
of the 10 ppb samples were positive (by definition);
and all samples >10 ppb were positive.
2. For Mississippi river sediment: No zero analyte
sample generated color greater than 0.25; 90% of
the 10 ppb samples were positive (by definition);
and all samples >10 ppb are positive.
3. For black dirt: No zero analyte sample generated
637
-------
color greater than 0.25; 100% of the 50 ppb samples
generated color of 1.0 or greater.
4. For clay: No zero analyte sample generated color
greater than 0.25; 60% of the 10 ppb samples
generated color of 1.0 or greater and 100% of the
50 ppb samples generated a color of 1.0 or greater.
Our results (Figure 8) suggest the greater extraction
efficiency for sand and river sediment compared to black
dirt and clay type soils at low levels of PCP contamination
(between 10 and 500 ppb). When testing soils suspected to
have high levels of contamination (>1 ppm) the results for
all soil types converge and the color generated with this test
is maximized.
Summary
Data presented demonstrates that BSI has developed an
easy-to-use enzyme immunoassay that can be used to
measure PCP in soil at concentrations of 10 ppb or greater.
The test kit has many distinct advantages over other
screening tests which are currently commercially available
for other small molecular weight analytes: 1) the assay has
a positive read-out system; 2) the use of wash steps
eliminates interfering substances; 3) no laboratory
equipment is needed, eliminating the purchase, calibration,
or maintenance of any equipment; 4) the assay is fast (less
than a total of ten minutes is needed for the results); 5) all
of the necessary reagents for the assay are present in the
assay kit, consequently the assay is very easy to use by
unskilled personnel; and 6) the enzyme immunoassay has
been miniaturized to maximize speed, portability, and ease
of use. Also, we were able to obtain evidence, through
accelerated time studies, that the test components and whole
kits were stable for one year when stored at room
temperature.
Acknowledgments
The authors wish to acknowledge the following support:
NIDA SBIR Grant No. 5 R44 DA03553; NIDA SBIR Grant
No. 1 R43 DA04372; NffiHS SBIR Grant No. 2 R44
ES04148; FBI Contract No. J-FBI-89-154; and USEPA
SBIR Contract No. 68D80035.
References
1. Morganthau, T., Miller, M. and Contreras, J., "Now
It's Bush's War," Newsweek. September 18, 1989,
22.
2. Morganthau, T., Miller, M., Sandza, R., Contreras,
J., Lane, C. and DeFrank, T.M., "Hitting the Drug
Lords," Newsweek. September 4, 1989, 18.
3. Baker, J.N., "The Newest Drug War," Newsweek.
April 3, 1989, 20.
4. Morganthau, T. and Miller. M., "The Drug Warrior,"
Newsweek. April 10, 1989, 20.
5. Hackett, G., "On the Firing Line," Newsweek. May
29, 1989, 32.
6. Heikes, D.L. and Griffitt, K.R., "Gas-Liquid
Chromatographic Determination of
Pentachlorophenol in Mason Jar Lids and Home
Canned Foods," J. Assoc. Off. Anal. Chem. 63,
1980, 1125.
7. Edgerton, T.R., Moseman, R.F., Lores, E.M., and
Wright, L.H., "Determination of Trace Amounts of
Chlorinated Phenols in Human Urine by Gas
Chromatography," Anal. Chem. 52, 1980, 1774.
8. Daniels, C.R. and Swan, E.P., "Determination of
Chlorinated Phenols in Surface-Treated Lumber by
HPLC." J. Chromatogr. Sci. 17. 1979, 628.
9. Ugland, K., Lundanes, E. and Greibrokl, T.,
"Determination of Chlorinated Phenols by High-
Performance Liquid Chromatography," J^
Chromatogr. 213. 1981, 83.
10. Paul, B.D., Mitchell, J.M., Mell, Jr., L.D., Irving, J.,
"Gas Chromatography/Electron Impact Mass
Fragmentometric Determination of Urinary 6-
Acetylmorphine, A Metabolite of Heroin," J. AnalL
Toxicol. 13. 1989,2.
11. Chen, B.H., Taylor, E.H. and Pappas, A.A.,
"Comparison of Derivatives for Determination of
Codeine and Morphine by Gas
Chromatography/Mass Spectrometry," J. Anal.
Toxicol. 14. 1990, 12.
12. Albro, P.W., Luster, M.I., Chae, K., Chaudhary,
S.K., Clark, G., Lawson, L.D., Corbett, J.T. and
McKinney, J.D., "A Radioimmunoassay for
Chlorinated Dibenzo-p-Dioxins," Toxicol. Appl,
Pharmacol. 50. 1979, 137.
13. Newsome, W.H. and Shields, J.B.,
"Radioimmunoassay of PCB's in Milk and Blood,"
Intern. J. Environ. Anal. Chem. U). 1981, 295.
14. Mule, S.J., Jukofsky, D., Hogan, M., DePace, A.,
and Verebey, K., "Evaluation of the
Radioimmunoassay for Benzoylecgonine (A Cocaine
638
-------
15.
16.
17.
18.
19.
20.
Metabolite) in Human Urine," Clin. Chem. 23.1977
796.
Spector, S. and Parker, C.W., "Morphine:
Radioimmunoassay, Science 168. 1970, 1347.
Newsome, W.H., "An Enzyme-Linked
Immunosorbent Assay for Metalaxyl in Foods," L
Aerie. Food. Chem 33, 1985, 528.
Kelley, M.M., Zahnow, E.W., Peterson, W.C., and
Toy, Stephen T., "Chlorsulfuron Determination in
Soil Extracts by Immunoassay," J. Agric. Food
Chem. 33. 1985, 962.
Rubenstein, K.D., Schneider, R.S. and Ullman, E.F.,
"Homogeneous Enzyme Immunoassay: A New
Immunochemical Technique," Biochem. Biophvs.
Res. Commun. 47(4). 1972, 846.
VanDyke, C., Byck, R., Barash, P.O., and Jatlow,
P., "Urinary Excretion of Immunologically Reactive
Metabolite(s) After Intranasal Administration of
Cocaine, as Followed by Enzyme Immunoassay,"
Clin. Chem. 23(2). 1977, 241.
Anderson, G.W., Zimmerman, J.E., and Callahan,
P.M., "The Use of Esters of N-Hydroxysuccinimide
in Peptide Synthesis", J.A.C.S. 86, 1964, 1839.
21. Benvenue, A., and Beckman, H.,
"Pentachlorophenol: A Discussion of its Properties
and its Occurrence as a Residue in Human and
Animal Tissues," Residue Reviews .19, 1967, 83.
22. Swanson, M.J., "Field Test Kits for Chemical and
Biological Warfare Agents," U.S. Army AMCCOM,
Contract No. DAAK11-83-C-0090, SBIR-Phase II,
Final Report, March, 1985.
23. Smigh, P.K., Krohn, R.I., Hermanson, G.I., Mallia,
A.K., Gortner, F.H., Provenzano, M.D., Fugimoto,
E.K., Goeke, N.M., Olson, B.J., and Klenk, D.C.,
"Measurement of Protein Using Bicinchoninic
Acid." Anal. Biochem. 150. 1985, 76.
24. Mancini, G., Carbonara, A., and Heremans, J.,
"Immunochemical Quantitation of Antigens by
Single Radial Immunodiffusion," Immunochemistry
2, 1965, 235.
25. Itzhaki, R.F. and Gill, D.M., "Micro-Biuret
Reaction," Anal. Biochem. 9. 1964, 401.
26. Habeeb, A.F.S.A., "Determination of Amino Groups
in Proteins by Trinitrobenzene-sulfonic Acid," Anal.
Biochem. 4. 1966, 326.
27. Worthington Enzyme Manual, L.A. Duker, ed.,
Worthington Biochemical Corp., p. 37, 1977.
Table 1. Modified HRP Activity after Purification
HRP-DADPA (Pre-purification)
HRP-DADPA (Post-purification)
Activity
Units/nq protein
700 ± 63
365 ± 49
n = 6
639
-------
NOTE: All components should be at room temperature.
1. Open foil package and remove test
module, color development tube, and
wash tube. (Just before use.)
2. Sample application: Remove red cap
from sample bottle and apply 10
drops (±5 drops) to the sample well of
the module.
A
D-
3. Wash application: Twist tab off wash
tube and squeeze entire contents into
sample well.
4. Color development tube application:
Hold tube upright and squeeze tube
where indicated to crush ampule
inside. Shake vigorously for 10
seconds.
Carefully apply ONE DROP of color
development solution to sample well.
Incubate for 1-2 minutes.
5. After incubation, press module closed
for 2-3 seconds. Release and open.
(Press only once.)
6. Open the module and monitor color
development. Record the result at 5
minutes.
A POSITIVE RESULT WILL SHOW A GREEN COLOR AS DARK OR DARKER
THAN THE REFERENCE COLOR.
A NEGATIVE RESULT WILL REMAIN WHITE OR BE LIGHTER THAN THE
REFERENCE COLOR.
FIGURE 1. AccuPress™ Test
Neg.
Pos.
640
-------
DROPPER CAP
DROPPER TOP
FILTER
SAMPLE SWAB
SAMPLE BUFFER
B
FIGURE 2. SAMPLER/EXTRACTION DEVICE
DAY
FIGURE 3. READ-OUT DISK STABILITY
641
-------
to
40
30
20
10
b
4°C.
R.T.
37°C.
45°C.
55°C.
5 10 15 20 25 30 35 40
STORAGE TIME IN DAYS
FIGURE 4. POP ANTIBODY DISK STABILITY
RADIOLABEL UPTAKE
4°C.
R.T.
37°C.
45°C.
55°C.
20 30
STORAGE TIME IN DAYS
FIGURE 5. POP CONJUGATE STABILITY
40
R.T.
37°C.
45°C.
55°C.
10 20 30
STORAGE TIME IN DAYS
FIGURE 6. READ-OUT DISK STABILITY
40
642
-------
4°C.
R.T.
37°C.
45°C.
o
55 C.
0
10
20
30
STORAGE TIME IN DAYS
FIGURE 7. STABILITY TESTING
POP ACCUPRESS TEST KIT
tr
3
8
UJ
3
i
RIVER SOIL
SAND
CLAY
BLACK DIRT
600
800
1000
FIGURE 8. POP ACCUPRESS TEST
VARIOUS SOIL TYPES
(N-10 EACH DATA POINT)
643
-------
XUMA EXPERT SYSTEM FOR SUPPORT OF INVESTIGATION AND EVALUATION OF
CONTAMINATED SITES
V. Eitel, R. Hahn
Landesanstalt fiir Umveltschutz
Baden-WUr 11 emberg
Abteilung Boden, Abfall, Altlasten
Griesbachstr. 3
D-7500 Karlsruhe 21, Germany
V. Geiger, R. Veidemann
Kernforschungszentrum Karlsruhe
Institut filr Datenverarbeitung
in der Technik
Postfach 36 40
D-7500 Karlsruhe 1, Germany
1. INTRODUCTION
In Baden-Wiirttemberg, programmes are
carried out to investigate and to regi-
ster the contaminated sites and to eva-
luate their environmental hazard. The
expert system will help in this vork.
The XUMA (German acronym for expert sy-
stem on environmental hazards of conta-
minated sites) expert system is being
developed within the framework of a
joint research project of the Institut
filr Datenverarbeitung in der Technik of
the Kernforschungszentrum Karlsruhe and
the Abteilung Boden, Abfall, Altlasten
of the Landesanstalt fOr Umveltschutz
Baden-Viirttemberg [1]. It is being im-
plemented on a Texas Instruments Explo-
rer II with the Inference ART develop-
ment environment and the RTMS database
system. The programs are written in
LISP and ART.
ports the user when a case-specific
analysis plan for the contaminates site
is prepared. The third function is used
for the input of the results of the
chemico-physical analyses into the
system. The fourth function supports
the assessment of a case, i.e.a comment
in form of an expert opinion. The fifth
function helps the user reconstruct the
derivation of the statements inferred
and the last function enables the
authorized experts to modify and com-
plete the domain knowledge acquired
within the knowledge base.
In the following, the central ap-
plication function of the system, the
assessment function, as well as the ex-
planation facility and first experien-
ces gained with these functions are
described in further detail.
of analysis re-
2. SURVEY
The1 following functions are covered
by the system:
1. Evaluation
2. Preparation of an analysis
plan
3. Acquisition
suits
4. Assessment
5. Explanation facility
6. Knowledge acquisition
The evaluation function primarily
deals with the determination of a num-
erical value for a first comparative
estimation of the environmental hazard
of contaminated sites. This value is
then used for setting priorities during
the investigation and sanitation of
waste sites. The second function sup-
3. ASSESSMENT
The basis of the assessment of a
contaminated site are the results of
investigations. The chemical and physi-
cal investigations are very important.
The expert has the problem to valuate a
lot of analysis data.
XUMA will help the expert in doing
this with the function "Assessment". On
the basis of the analysis results
statements are devired for the assess-
ment of the hazard level. Indications
of further investigations are given as
well as other assessment statements
like indications of inconsistencies in
the analysis data, control of plausibi-
lity. In an other function, that is not
yet realized, the local situation in-
cluding hydogeology will be considered
645
-------
[2]. The function "preparation of an
analysis plan", helps to find an indi-
vidual analysis plan with the investi-
gation parameters, that are of concern
for example on a contaminated soil
within an industrial plant. Contami-
nated soil, wastes, eluates of the
wastes, leakage water, groundwater,
surface water, air, soil will be ana-
lysed. For the assessment the analysis
data, information taken from the samp-
ling records and informations from the
record of the chemical analysis are of
concern.
XUMA contains rules for the assess-
ment of
- parameters
- one chemical analysis of a sample
- all chemical analysis of one sam-
ple, for example a contaminated
soil and the eluate
- all analysis results of sample of
discrete areas
- all analysis results and informa-
tions of the case.
More than 25 tables of limit values
or threshold values for water, soil and
air with more than 100 different para-
meters are taken into the knowledge
base of the system as well as rules ex-
plaining the special scope of a table.
Values of this limit values tables and
threshold values tables were associated
with 6 quality classes. Quality class I
is corresponding to the background va-
lues, quality class II is "tolerable",
class III means "further investigation
necessary", IV, V and VI medium, high
and very high hazard level. Example: If
the analysis result of a parameter of
groundwater is smaller or equal 0.2
times the limit value of the german
drinking water quality table, then it
is quality class I. The rules of asso-
ciating the concentration values to
quality classes are the result of ex-
periences with the risk assessment of a
great number of contaminated sites.
Assessment of the Analysis; 201.83
25.07.1983 Eluate
Analysis Results
Colour, qualitative » yellow
Electric conductivity - 430 uS/cm
Ammonium =» 0.200 mg/1
Chloride < 10 mg/1
Cyanide, total - 0.750 mg/1
Phenol, total - 0.900 mg/1
Dry matter - 1146 mg/1
Residue on ignition (550 C)= 1112 mg/1
Hydrocarbons (IR) 3.400 mg/1
Mineral oil - 3.400 mg/1
Loss on ignition at 550 C - 34 mg/1
Assessment Results
Assessments on the Basis of Limit Valu
Tables;
'Dry matter' is put into quality
class II - permissible
(TVO).
'Ammonium' is put into quality
class II - permissible
(TVO).
'Chloride' is put into quality
class I - within the
range of background
values (EG-TW).
'Cyanide, total' is put into quality
class interval IV to
VI (TVO).
'Mineral oil' is put into quality
class IV - medium
hazard potential
(NDL-GW).
'Phenol, total' is put into quality
class V - high hazard
potential (NDL-GV).
'Electric is put into quality
conductivity' class II - permissible
(TVO).
Definite Statements;
Theportion oforganic matter in the
dry matter is about 2 % (calculated). A
considerable portion of the substances
contained is not analyzed.
The value of 'dry matter' is normal.
The value of 'residue on ignition' is
normal.
The parameter 'colour, extinction at
436 nm' should be analyzed.
There are indirect indications of
'crude tar' to contained. Reason: co-
lour.
Potential Statements;
There is some indication that the orga-
nic portion in the dry matter might be
high. Reason: dry matter » electric
conductivity.
Total Result;
The analysis is put into quality class
interval V to VI.
FIGURE 1: Example of the assessment of
an analysis
646
-------
The tables were classified in groups
for groundwater, surface water, soil
etc. Within one group, .rules are de-
fined, for which purpose and with what
priority the tables are to be used. If
there is no possibility to find a value
for a parameter in one group of tables,
e.g. groundwater tables, rules are
given to use other groups of tables,
e.g. drinking water tables. If there is
no value in any of the tables for one
parameter, XUMA proposes different pa-
rameters with similar chemical charac-
ter (e.g. o-Xylol for p-Xylol). Examp-
les of rules for the summarization of
the assessment statements are shown in
figure 2.
Rules
les:
with regard to limit value tab-
If leakage water is to be assessed,
and the value of at least one para-
meter is quality class IV - VI,
then the sewage tables are to be used.
If a measured value x is compared with
the Dutch soil table
and B Value < x < C value is valid
for the B and C values of the para-
meter,
then the measured value belongs to qua-
lity class III.
Rule of assessing an individual analy-
sis parameter:
If pH value < 5,
then the solubility of heavy metals is
increased.
Rule of summarizing the results on the
sample level:
If the turbidness of a water sample is
clear in the sampling record and not
clear in the laboratory analysis,
then the sample has changed chemically
after sampling.
Rule for summarizing the results on the
case level:
If 'cyanide, total' or 'hydrocarbons
(IR)' is high or very high in leak-
age water samples,
then the ground water should be analy-
zed.
The assessment of a case gives sta-
tements on the hazard level (quality
class), the need of further investiga-
tions, statistic and definite and po-
tential statements.
XUMA can give some help for present
technical investigations or remedial
actions. The field screening analytical
data will be transmitted to the expert
system which gives an assessment for an
actual case. So the following actions
e.g. further sampling or remediation of
the waste will have a better basis.
4. EXPLANATION
It is of particular importance for
the acceptance of the system that its
conclusions are clear and may be recon-
structed by the user. For this purpose,
an explanation facility has been imple-
mented [3]. The derivation of the
statements is explained to the user by
means of texts written in the natural
language. Each statement displayed is
mouse-sensitive. The explanation faci-
lity is called by clicking on a state-
ment with the mouse. Now, the user can
choose between the local or global
justification of the statement. In lo-
cal justification (Fig. 2), the state-
ment itself is listed together with the
last rule that has led to this state-
ment and with the conditions fulfilled
(premises). In global justification,
the complete tree of derivation is
shown, i.e., the derivation of the
statement from the analysis results and
the facts and rules included in the
static knowledge base is represented.
The derivation structure is represented
by indentations.
FIGURE 2: Examples of assessment rules
647
-------
Local Justification
The fact to be explained is:
The portion of organic matter in the
dry matter, calculated from residue on
... ignition, is about 38 %.
It vas deduced by the rule G2-RESIDUE-
ON-IGNITION-2:
If loss on ignition and dry matter
are known,
... then the portion of organic mat-
ter can be calculated to be ap-
proximately:
... loss on ignition/dry matter.
The following premises are fulfilled:
The analysis '207.83 25.07.83 Elu-
ate' resulted in: dry matter =
210 mg/1.
The analysis '207.83 25.07.83 Elu-
ate' resulted in: loss on ignition =
80 mg/1.
FIGURE 3: Example of the local just-
ification of a statement.
5. EXPERIENCES
XUMA was tested in the LfU. It was
surprising how easy the system is to be
used even for a user without experience
in computers. The system is a good ass-
istant to help the expert in risk ass-
essment. Maybe that some of the rules
seem to be simple or trivial if they
are seen isolated. If rules are combi-
ned and used without any exception, the
statements are very helpfull for the
expert.
Even inconsistent assessments can
help to find errors in the analysis da-
ta or in the rules. The expert is then
able to create better rules or modified
quality classes. A risk assessment is
not only given on the basis of one dis-
crete value as a yes-no decision it is
furthermore the result of comparing va-
rious standard value tables with defi-
ned scope.
So the expert will have the necess-
ary tolerance for the special assesse-
ment of individual cases. Risk assess-
ments are transparent, standardized as
far as possible and reproducible. The
system can be used only by risk assess-
ment experts.
6. REFERENCES
[1] Veidemann, R. and Geiger, V.
(1989). XUMA - Bin Assistent fdr
die Beurteilung von Altlasten. In
A. Jaeschke, V. Geiger and B. Page
(Eds.), Informatik im Umweltschutz,
Informatik-Fachberichte 228, pp 385
- 394. Berlin: Springer-Verlag.
[2] Ministerium filr Umwelt Baden-Wurtt-
emberg (Ed.) (1988). Altlasten-
Handbuch, Teil 1, Altlasten-Bewer-
tung. Wasserwirtschaftsverwaltung,
No. 18.
[3] Huber, K.-P. (1988) Erklarungskom-
ponente fur das Expertensystem XUMA
unter Berlicksichtigung verschiede-
ner Benutzerklassen. Kernfor-
schungszentrum Karlsruhe, KfK 4478.
648
-------
A RAPID RESPONSE SAW-GC CHEMICAL MONITOR FOR
LOW-LEVEL VAPOR DETECTION
JOHN A. ELTON and JAMES F. HOULE
EASTMAN KODAK COMPANY
ROCHESTER, NEW YORK
INTRODUCTION
Chemical vapor monitors (CVMs) are
generally not sufficiently sensitive,
selective, or reliable enough to
detect a multiplicity of vapors in
less than 2 min. There is,
therefore, a need for a CVM which can
simultaneously detect a variety of
vapors in the presence of
interferents. In addition, the
detection of a specific vapor must be
conclusive so that false alarms are
minimized. Detection in less than 2
min requires either highly selective
multiple detection methods if several
vapors are present or separation so
that each vapor can be detected and
identified. The sensitivity for each
vapor must also be sufficient to
allow detection at desired or
required levels.
SAW sensors have been used to detect
vapors at low concentrations [1].
However, the SAW detection limits
reported to date for agents such as
GD and HD are much higher than the
limits other devices are capable of
reaching. Detection limits of
approximately 100 ppb (0.6 mg/m^) for
GD and 5 ppm (32 mg/m^) for HD have
been reported. It will be shown in
this paper that much lower levels may
be obtained for GD and HD when the
system described herein is used. In
addition, results on the detection of
methyl benzoate and phenyl acetone
using the same system as used for CW
detection will be provided.
BACKGROUND
The CVM unit contains major
modifications which allow
significantly improved response
times. Ambient vapors are collected
on a thermally desorbed type
concentrator by pumping air through a
glass tube packed with concentrator
material (Figure 1; concentration).
At the end of a fixed 20 sec
interval, the concentrator is heated
and the collected vapors desorbed
onto the GC column. Desorption
occurs in about 6 sec and provides
chromatographic peaks that are
compatible with the SAW detector
(Figure 1; injection). An additional
4 to 8 sec is typically needed,
however, in order to obtain complete
injection of the vapor plug onto the
GC column.
CARRIER GAS
GENERATOR
AMBIENT AIR
INLET
VAPOR
CONCENTRATOR
CONCBJTHATOR
VENT
CONCENTRATION
NJECTON
ANALYSIS
VALVES
Fig. 1. Kodak's Chemical Vapor
Monitor Showing
Concentration, Injection, and
Analysis Stages of Operation.
The GC column greatly enhances the
selectivity of the system by
separating the vapors (Figure 1;
analysis). Each vapor plug which
elutes from the GC column at a
different time is immediately
injected onto a SAW sensor. A
second, uncoated SAW sensor located
nearby is used as a reference. When
combined with a frequency mixer, this
configuration provides a frequency
649
-------
difference (Af) that is easily
measured and relates to
concentration.
The CVM has several other subsystems.
The sequencing of valves,
concentrator, pumps, and the
acquisition of SAW sensor data is
controlled by a Macintosh™ computer.
A second subsystem provides clean
air/carrier gas to the GC column with
a small pump that draws ambient air
through molecular sieve and charcoal
scrubbers. A solid-state mass flow
controller is used to guarantee a
stable carrier gas flow under varying
conditions of pump and scrubber
aging. A typical output
of the CVM is given in Fig. 2 and
shows the concentration, injection,
and analysis characteristics of the
device.
EXPERIMENTAL
Each SAW sensor was first tested as
an individual sensor with each vapor
of interest at one or more
concentrations. The sensor was then
incorporated into the CVM and system
testing performed.
Vapor Generation and Verification
Vapors were generated using a
Microsensor Systems, Inc. VG-7000
Automatic Vapor Generation System.
All vapors supplied to the CVM were
monitored using a Hewlett Packard
5890 Gas Chromatograph containing an
FID detector. For the SAW sensor
testing, periodic checks of the vapor
concentration were made every 11
minutes. For the CVM tests, vapor
concentration was determined by
sampling the final portion of vapor
which impinged upon the CVM
concentrator. This procedure was
performed in order to ensure
verifiable vapor concentrations.
SAW Sensor Preparation
Selective coatings of ethyl cellulose
(ECL) and fluoropolyol (FPOL) on SAW
sensors were prepared using
proprietary thin film coating
techniques. All coatings were
observed under a microscope to
CHEMICAL VAPOR MONITOR
RESPONSE TO CEES VAPOR
F
R
E
Q
S
H
I
F
T
700
500
300
100
-100
CONCENTRATION
6.4 mg/m3
INJECTION
10 sec
ANALYSIS
140
-20
10
20
30
TIME (sec)
40
50
Fig. 2. Typical CVM Output
650
-------
determine uniformity of the coating.
Only uniform and well adhered
coatings were used in this study.
The thickness of a coating was
determined by recording the frequency
shifts of the device both before and
after coating.
SAW Sensor Testing
Individual SAW sensors were tested
using various concentrations of the
vapors of interest. The SAW sensors
were exposed to the vapors for a
minimum of 20 minutes. A signal to
noise ratio of at least 3:1 was
chosen as a criterion for acceptable
data.
CVM Testing
The general operation of the CVM was
described above. Standard operation
is a 20 second concentration period,
a 12 second injection period, and an
88 second analysis time. Deviations
from this standard will be indicated
in the results section when
appropriate.
RESULTS
Results were obtained using two types
of SAW sensors. The first tests used
a modified version of an established
type of sensor (Type I). This was
followed by extensive testing of a
new type of SAW sensor (Type II).
Both SAW sensor test results and CVM
test results are reported below.
SAW Sensor Testing
Table 1 provides the frequency shifts
observed from selected experiments
when the ECL-I, FPOL-I, ECL-II, and
FPOL-II sensors were tested with
various concentrations of GEES, DMMP,
methyl benzoate, and phenyl acetone.
Chemical Vapor Monitor Testing
During CVM testing the vapor flow was
connected to the concentrator input
of the system; flow rates through the
concentrator were monitored. The
results of selected tests using the
first type of SAW sensor are shown in
Table 2.
TABLE 1
SAW Sensor Results
Vapor Concentration SAW
(mg/n»3) Response*
(Hz)
CEES**
DMMP**
CEES
DMMP
MB**
PA**
CEES
DMMP
MB
PA
2100
91
17.8
20.6
2.4
1.7
516
2.3 ±
19.2
6.48
2400
4,710
58
362
75
88
248
1.1 773
222
265 ± 60
Sensor
ECL-I
FPOL-I
ECL-II
FPOL-II
*80 Hz noise level (Type I);
5 Hz (Type II)
**CEES - chloroethyl ethylsulfide;
DMMP - dimethyl methylphosphonate;
MB - methyl benzoate; PA - phenyl
acetone
TABLE 2
Chemical Vapor Monitor Results
Sensor Vapor Concentration CVM
(mg/m-') Response*
(Hz)
ECL-I**
CEES
FPOL-I*** DMMP
ECL-II CEES
FPOL-II
DMMP
MB
PA
6.1
17.6
5.27
23.2
27.2
11.0
570
1600
82
495
213
150
*Response obtained in less than two
minutes; 2 Hz noise level unless
otherwise specified
**5 Hz noise level; 10 second
injection period
***10 Hz noise level; 14 second
injection period
DISCUSSION
These results indicate that detection
limits for GD and HD using the new
type of SAW sensor should be
considerably less than previously
reported [1-2]. Table 3 gives
extrapolated detection limits and
response times for both the
SAW sensor experiments and the CVM
experiments. The extrapolated
detection limits are determined from
651
-------
the values reported herein using a
3:1 signal to noise ratio.
TABLE 3
Extrapolated Detection Limits
Type Vapor Extrapolated Response
Detection Time
Limit (mg/m3) (min)
SAW Sensor
Type I CEES
DMMP
Type II CEES
DMMP
MB
PA
CVM
Type I CEES
DMMP
Type II CEES
DMMP
MB
PA
210
4.6
4.6
0.05
0.48
0.29
0.16
0.33
0.38
0.28
0.77
0.44
20
40
20
40
20
20
It should be noted that optimization
of coating thickness was performed
much more extensively with the Type I
sensor. Significant improvements in
the Type II sensor are expected in
the future as further optimization of
coating thickness and subsystem
parameters are performed. The much
lower noise level of the Type II
sensor is the principal advantage of
using this technology since both
sensors should provide approximately
the same response when the same
thicknesses and types of coatings are
utilized.
The results also reveal that the
greatly increased sensitivity of the
Type II SAW sensor is not carried
over to the detection of DMMP using
the complete CVM unit. FPOL coated
sensors do not equilibrate as quickly
with DMMP as with the other vapors.
The peaks observed during all DMMP
testing were much broader than for
all other cases. Significant tailing
of peaks was observed. Different GC
columns and higher temperature
operation of the SAW sensors may help
to narrow the peak width and improve
the detection limit for DMMP.
The results also show the response of
the two sensors to methyl benzoate
and phenyl acetone. It is believed
that this is the first time that
detection of such vapors with SAW
detectors has been reported. The
extrapolated detection limits
indicate that relatively low levels
of these vapors can be detected with
SAW sensors. No effort has yet been
made to develop special selective
coatings for these two vapors.
CONCLUSION
Our studies have shown that SAW
technology can be used to detect DMMP
and CEES at concentrations below 1
mg/m3 in less than two minutes. We
have also reported for the first time
the behavior of methyl benzoate and
phenyl acetone to SAW sensors
typically used for chemical agent
detection. Concentrations below 1
mg/m3 are also indicated for these
two vapors. Because of the
preliminary nature of some of the
data presented herein, we anticipate
even lower detection levels in the
future as operational parameters and
selective coatings are optimized.
ACKNOWLEDGEMENTS
The authors would like to thank
Arthur Snow of the Naval Research
Laboratory for providing us the
fluoropolyol used in these
experiments.
REFERENCES
1. Dennis M. Davis, Leon J. Schiff,
and John A. Parsons, "Detection Of
Chemical Warfare Agents Using A 158
MHz SAW Microsensor,"Proceedings of
the 1987 U.S. Army Chemical Research,
Development and Engineering Center
Scientific Conference on Chemical
Defense Research, 2., 939-945, 1988.
2. Dennis M. Davis, Raymond E.
Miller, and John A. Parsons, "Surface
Acoustic Wave Detection of
Organophosphorus Compounds,"
Proceedings of the 1987 U.S. Army
Chemical Research, Development and
Engineering Center Scientific
Conference on Chemical Defense
Research, 2, 917-923, 1988.
652
-------
PASSIVE CRYOGENIC WHOLE AIR FIELD SAMPLER
S teven J. Fernandez
Bill G. Motes
Joseph P. Ougan, Jr.
Susan K. Bird
Gary J. McManus
Westinghouse Idaho Nuclear Company
P.O. Box 4000, Mail Stop 2202
Idaho Falls, Idaho 83403
The portable, passive cryogenic sampler has
been designed by the Idaho National
Engineering Laboratory (INEL) for the
collection of whole air samples without the
loss or concentration of any atmospheric
constituents. The principle of operation is
the collection by bulk gas flow and
condensation of a whole air sample into a
previously evacuated cylinder held at liquid
nitrogen temperature using a reservoir. The
ability of the sampler to collect a highly
compressed gas sample without concentration
of condensable gases permits a large number
of gas constituents to be analyzed from a
single sample, even when analytes vary
widely in their boiling points.
Design criteria for the portable, passive
whole air sampler are listed in Table 1.
The sampler, constructed in-house of readily
commercially available components, is shown
in Figure 1.
The sampler evaluation was performed in
three phases. The first phase determined
sample flowrate, sample size, resultant
sample pressure, and sample collection
lifetime as a function of the liquid
nitrogen additions. The second phase
analyzed simulated whole air samples for
bulk composition, noble gases, selected
chlorofluorocarbons, and tritium before and
after collection in the cryogenic air
sampler. In the third phase actual field
samples were collected and analyzed for bulk
composition and chlorofluorocarbon content.
These samples were then concentrated,
separated, and analyzed for noble gases.
Also included in the third phase was the
analysis of altered whole air samples,
blindly and randomly introduced into the
sample analysis scheme as a means of
detecting sample tampering.
Results of the evaluation of the design
criteria for the sampler are listed in Table
2. A partial listing of past customers and
their application of the sampler is found in
Table 3. The need of a low cost passive
cryogenic sampler that can collect many
whole air samples at remote locations with
minimal logistical support will become
widespread in the future.
We have found the sampler to meet or exceed
all of the characteristics intended for it.
The sampler is capable of the collection of
samples without concentration or loss of any
sample constituents regardless of boiling
point.
The required sample volume of 100 L at STP
has been successfully achieved, and samples
as large as 131.2 L have been collected.
Most samples are between 70 and 90 L. The
volume of sample collected is dependent upon
the sample duration and flowrate. By
selection of the proper combination of
sample duration and flowrate, samples of
accurately known size from a few to 100
liters may be collected unattended within 30
minutes or over a period of time of 2 hours.
Laboratory tests on known standards
demonstrate that no concentration or loss of
atmospheric constituents occurs.
No electrical power is required for
operation of the sampler, which would enable
it to operate in hazardous environments such
as where potentially explosive mixtures of
hydrogen and oxygen are found.
The sampling lifetime can extend to 4 hours
with refilling of the liquid nitrogen
reservoir.
653
-------
TABLE 1
Desien Criteria for Whole Air Sampler
TABLE 2
Development of Cryogenic Whole Air Sampler
1) Sample Volume of 100 Liters
2) No Concentration or Loss of Constituents
3) Sampling Lifetime Greater Than 2 Hours
4) Small (50cm x 15cm) & Lightweight (20kg)
5) No Electrical Power
6) Operator Safety
7) Ease of Operation
Parameter
Sample Flowrate
Sample Volume
Sample Pressure
Sampling Lifetime
Sample
Concentration
Sample Loss Tests
Results
Controllable, 1 cc to 3 L/Min
Nominally 100 Liters, 130 L
Maximum
Nominal 2000 psi, 3650 psi
Maximum
30 Min to 10 Hours
Noble Gas Ratios, Unaltered
Chlorofluorocarbons, 100Z
Recovery
Tritium, 100Z Recovery
TABLE 3
Customer
DOE- Office
of Materials
US Air Force
DOE- Office
Waste Mgt
US Air Force
DOE- Defense Programs
DOE-Office
Arms Control
US Air Force
Facility
ICPP
Proposed for
White Sands,
KA-III Series
ICPP
TREAT Pulse
Reactor INEL
Advanced Test
Reactor INEL
INEL Research
ICPP
e vji.yugen.n; WIIUJ.B ftir aampiei
Application
Hydrogen-Rich
Off-Gas Study
Fuel/Air
Explosives
Environmental
Sampling
Environmental
Sampling
Off-Gas Studies
Evaluation of
Arms Control
Verification
Fission Products
in Ar Carrier Gas
Analvtes
Permanent
Gases
Combustion
Products,
Oxygen
85Kr,N2,02
Ar , C02
Kr, Xe,
Freon-11
Me thy 1-
chloroform
41Ar,Kr,Xe
CFC13,CH3CC13
Freon-113
XKr.Xe.He,
N2,02,Ar,H2,
Freon-12,
C02, Freon-113
Fission
Product Gases
654
-------
Variable Set-Point
Pressure Releif Valve
Fixed Set-Point
Pressure Relief Valve
Plate L1d
Sample Cylinder
Base Plate
V
Rotameter
Sample Inlet/Isolation
Valve
Hexagonal Nipple
SS Tubing
Liquid Nitrogen
Reservoir
ICPP-A-16925
(1-91)
FIGURE 1
PASSIVE CRYOGENIC WHOLE AIR FIELD SAMPLER
655
-------
Effectiveness of Porous Glass Elements for Suction Lysimeters
to Monitor Soil Water for Organic Contaminants
by
Stanley M. Finger, Hamid Hojaji,
Morad Boroomand, and Pedro B. Macedo
Vitreous State Laboratory
Catholic University of America
Washington, DC 20064
ABSTRACT
The objective of this effort is the development of
a porous glass suction lysimeter which can be used to
sample organic contaminants associated with unsaturated
soil matrices. Current ceramic suction lysimeters are
ineffective in sampling hydrophobic compounds since their
surface chemistry is hydrophilic, effectively repelling
organic species.
Methods for preparing porous glass elements with
controlled porosity have been developed. Elements with
air entry values (as measured by the bubbling pressure
method) corresponding to effective pore sizes as small as
2 microns with high saturated hydraulic conductivities
have been achieved.
The performance of porous glass elements in
sampling organic contaminants in aqueous media is being
evaluated. Aliphatic (1-octanol) and aromatic
(ethylbenzene) compounds dissolved in water were used
as the tests solutions. Tests are also being performed with
inorganic constituents in the test water to determine the
ability of the test elements to sample inorganics. Initial
results indicate that the porous glass elements are able to
effectively sample organic and inorganic constituents in
the test solutions. These data indicate that analyte concen-
trations in the water sampled through the porous glass
elements are within about 10% of the test solution concen-
trations.
BACKGROUND
The U. S. Environmental Protection Agency
(EPA) requires vadose zone monitoring at active
land treatment and disposal facilities for hazardous
wastes. The state of California has extended this
requirement to practically all active and closed
storage, treatment, and disposal facilities for haz-
ardous waste, solid waste, and underground storage
tanks. Routine analysis of samples collected with
suction lysimeters is considered an important ele-
ment in the vadose zone monitoring requirement.
Most of the suction lysimeters in use now were
developed for the agricultural industry to monitor
leachate from crops. These data are used to pro-
gram the application of fertilizers and soil amend-
ments. Another device, the tensiometer, is used in
conjunction with the suction lysimeter to monitor
soil moisture; this information is used to program
irrigation. This same equipment is now being used
to monitor the land treatment of certain hazardous
wastes, e.g. refinery separator sludge and wood
preservative waste. Many of the components of
interest in these wastes are organics and heavy
metals.
The suction lysimeter's porous element,
through which soil water is drawn under vacuum,
has been purposely designed to be hydrophilic to
facilitate the transport of the aqueous phase.
Porous elements currently in use are most frequent-
ly ceramic. However, TFE-fluorocarbon, nylon
mesh and alundum have also been used. The
porous element is typically treated with acid and
water to remove contamination and enhance hydro-
657
-------
philicity. Unfortunately, the resulting hydrophilic nature
of the porous element presents an effective barrier to
sampling of non-polar components. Organics, whether
dissolved in the aqueous phase or existing as a separate
phase are significantly under-sampled by existing suction
lysimeters. One recent field study of soil-pore water
sampling systems showed no correlation between organics
found by sampling compared with analysis of soil cores
(1). Additional studies have shown that xylene (2), DDT
(3), and fecal coliform (4) are not effectively sampled by
ceramic suction lysimeters. In addition, a number of
inorganic parameters, heavy metals in particular, are also
attenuated by ceramic suction lysimeters. Simultaneously,
a number of inorganic constituents are leached from
ceramic suction lysimeters into soil water samples. While
TFE-fluorocarbon porous elements are less prone to
significant adsorption or desorption of inorganics, they
also under-sample organic components. Additionally, the
large pore sizes of TFE media restrict their range of
operation to wetter soils than can be sampled by ceramic
suction lysimeters.
Ideally, a suction lysimeter should provide a
sample which accurately represents the soil liquid phases
at the sampling location. This would include all compo-
nents, organic as well as inorganic, dissolved in the soil
water and any non-aqueous, i.e. organic, phases. While
sampling all components representatively, the lysimeter
should also be inert so it does not leach any chemical
species into the sample. To achieve this ideal goal, the
porous element must be very stable over a wide range of
aqueous and organic conditions and be neither hydrophilic
nor hydrophobic. Such a perfect porous element is
probably unachievable. However, elements made of
porous glass could form the basis for approaching this
goal. Porous glass elements can be formulated which are
inert to organic and aqueous media over wide ranges of
pH and dissolved components. Also, the surface structure
can be controlled to moderate its hydrophilicity/hydropho-
bicity. This control can be achieved by modifying the
composition of the glass, modifying the thermal process-
ing of the glass, and, if necessary, by chemically treating
the glass to incorporate desired chemical species on the
surface. This paper describes the results of laboratory
studies aimed at the development of porous glass elements
for use in suction lysimeters to provide more accurate
sampling of organic as well as inorganic species.
EXPERIMENTS WITH HIGH SILICA POROUS
GLASS ELEMENTS
Preparation of the Elements
A series of porous glass discs were prepared
from powdered high silica borosilicate glass by
sintering. The solid state sintering mechanism for
different glass systems is well-known and, to a
large degree, applies to porous glasses. Densifi-
cation and the resultant reduction in porous volume
occurs in two separate regimes when high silica
porous glasses, as used in this research, are
sintered. The onset of the first stage starts above
750°C, at which point the micro-pores start to
disappear. The driving force for this process is a
reduction in the surface energy. Above 950°C, the
second stage of sintering begins. In this stage, neck
formation occurs between the individual grains of
glass, affecting the macro-pores. It is important to
control the overlap of the two stages, with more
emphasis on the second stage since this stage
controls the macro-porosity of the system.
A systematic study was conducted to evalu-
ate the degree of densification when sintering
powdered porous glass. The objective was to gain
control over the pore structure of the elements for
the porous glass suction lysimeter.
The porous elements were prepared by firing
at different peak temperatures. The glass was held
at the peak temperature for various times ranging
from 30 to 90 minutes. Heating and cooling rates
were maintained constant for all the samples.
Densities of the resulting glass discs were measured
and normalized against the density of the solid glass
having the same composition (the density of solid
high silica glass with 4-5 % boron oxide is approxi-
mately 2.25 g/mL). The densification and fraction-
al porosity as a function of firing temperature (60
minute firing time) are shown in Figure 1. As can
be seen in this graph, porous glass powder sintered
at 1200°C for 60 minutes achieves an 85 % densifi-
cation. A series of scanning electron micrographs,
showing the structure of porous glass elements
prepared at temperatures of 1100 and 1150°C for
60 minutes and 1200°C for 90 minutes are shown
in Figures 2a-c. These micrographs visually show
that the pore size and fractional volume decrease
with increasing firing temperature and firing time.
658
-------
Pore Size and Hydraulic Conductivity Measurements
Bubbling pressure, or air entry value, measure-
ments were performed on the porous glass elements
prepared. At first, a lucite disc holder was used. This
worked well at low pressures but leaked at higher pres-
sures. A second holder, made of stainless steel was
prepared which worked well over the full range of
pressures studied. Figure 3 shows both the lucite and
stainless steel holders.
The pore size corresponding to the air entry value
was calculated by the following equation:
d = 30 Y/P
(1)
where d is the pore size in microns, P is the bubbling
pressure (the pressure at which air first comes through the
porous disc) in mm Hg, and Y is the surface tension of
water in dynes/cm at the temperature of the experiment.
At room temperature, Y is 73.05 dynes/cm. It should be
noted that the pore size measured by this procedure is an
effective pore size; the actual pore sizes vary as can be
seen in the scanning electron micrographs (Figure 2).
Air entry value measurements were performed on
a number of porous discs prepared over a range of sin-
tering temperatures and times. The results are plotted in
Figures 4 and 5. In Figure 4, the effective pore size is
plotted as a function of firing time at three different firing
temperatures (1050, 1150, and 1200°C). Figure 5 shows
the effect of firing temperature on effective pore size
when the firing time is held constant at 60 minutes.
These graphs clearly show that the pore size can be varied
down to 2 microns (firing at 1200°C for 60 minutes).
The flow rate through a series of the porous glass
elements was also studied. These data were used to
calculate the hydraulic conductivity and determine the
relationship between effective pore size and hydraulic
conductivity. The same holder used to measure the air
entry value was used to measure the flow rate. For these
experiments, the flow was induced by maintaining a
vacuum on the porous glass disc. The experiments were
performed using a vacuum of 63.5 cm (25 in.) of Hg.
The fluid used for these experiments was deionized water,
which was drawn from a burette able to measure volume
to 0.1 mL. The flow through each disc was measured for
at least two runs and the results averaged.
Figure 6 shows the measured flow rate as a
function of sintering temperature (60 minute firing time).
It also plots the pore size against the same abscissa. This
graph shows that as the sintering temperature increases,
the flow rate decreases along with the pore size, as
would be expected.
The hydraulic conductivity was calculated
using the following equation:
K = (Q/t) * (L/A) * h
(2)
where Q is the volume of water flowing through the
element in time, t, L and A are the thickness and
cross-sectional area of the element, respectively,
and h is the pressure differential across the element.
Figure 7 plots the hydraulic conductivity against
pore size. The data indicate a linear relationship
between hydraulic conductivity and effective pore
size.
The hydraulic conductivity of the porous
glass elements appear larger than that of ceramic
suction lysimeter elements of the same pore size.
For example, the saturated hydraulic conductivity of
a Soilmoisture Corporation ceramic suction lysi-
meter with an air entry value of 1 bar (pore size 2.1
microns) is 3.36E-7 cm/s. Figure 7 shows the
saturated hydraulic conductivity of a 2 micron
porous glass element to be about 1E-5 cm/s, almost
two orders of magnitude greater than the ceramic
element.
A series of nominal 2 micron pore size
elements were prepared. The measured pore size
and hydraulic conductivities of the elements are
reported in Table 1. The average pore size was 2.1
microns with a standard deviation of 0.8 micron.
The average hydraulic conductivity was 1.8E-6
cm/s with a standard deviation of 0.6E-6 cm/s.
The pore size and hydraulic conductivity of the
elements range by a factor of about 2. The
hydraulic conductivities of these elements are
almost an order of magnitude higher than that of a
comparable Soilmoisture Corporation ceramic
suction lysimeter, although not as high as the
porous glass element reported in Figure 7.
Sampling Efficiency
Experiments were performed to determine
the permeability of the porous glass discs to inor-
ganics dissolved in water. Inorganic test solutions
contained sodium chloride, barium chloride, lead
chloride, and potassium chromate. The results of
these tests are shown in Table 2. For all tests, a
61.0 cm (24 in.) Hg vacuum was maintained across
659
-------
the elements. The concentration of the inorganics was
measured by Direct Coupled Plasma (DCP) Spectroscopy.
The table shows good correlation between the concentra-
tions in the sample solution as compared to the test
solution. The average ratio of sample to test concen-
trations was 0.88 with a standard deviation of 0.16.
Problems Observed
Several problems were observed with the borosili-
cate porous glasses used in the first set of experiments.
One problem was that when the porous glass powder used
to make the elements was exposed to air for extended
periods (hours), the resulting elements were very fragile
(they tended to crack easily). It was hypothesized that
this could be due to the formation of internal cracks
caused by drying or by formation of silica gel within the
pores. This problem was resolved by keeping the porous
glass powder in water until it was used to form the
elements.
A second, more serious, problem was clogging of
the elements over time. It was hypothesized that compo-
nents of the glass were leaching into and precipitating in
the interstices of the elements. To alleviate this problem,
the porous glass was modified by the addition of zirconia
to produce a more durable glass matrix. The results with
this zirconia glass are reported in the following sections.
EXPERIMENTS WITH ZIRCONIA GLASS POROUS
ELEMENTS
Preparation of the Zirconia Glass Elements
The composition of the porous glass powder used
to make the test elements was modified by the addition of
4-5% zirconia. This modification was made to produce
a more durable glass which would be more consistent and
less likely to clog. The glass was prepared by sintering
the powder at 1150°C for 60 minutes.
Pore Size and Hydraulic Conductivity
Table 3 provides the pore size and hydraulic
conductivity measured on several elements of the zirconia
glass. The consistency, in terms of pore size and hydrau-
lic conductivity, among elements was much better than the
earlier test elements. However, the pore size was approx-
imately 3 microns. Revised heat treatments should be
able to lower the effective pore size to the 2 micron
range.
Sampling Efficiency
Experiments were conducted to determine
the permeability of the porous glass discs to or-
ganics dissolved in water. Organics used in the test
solutions were ethylbenzene and 1-octanol. For all
tests, a 61.0 cm (24 in.) Hg vacuum was main-
tained across the elements. The concentration of
the organics was measured with a Total Organic
Carbon (TOC) analyzer.
The test organic solutions were prepared by
carefully placing a layer of the organic chemical on
top of a large beaker of water. The liquids were
allowed to equilibrate over several days. The water
in the bottom of the beaker was periodically sam-
pled (without disturbing the interface between the
two phases) and its TOC content measured. When
the TOC content of the water became constant, it
was carefully removed from the beaker so that no
droplets of organics were entrained.
The evaporation of the organic component
from the test solutions under vacuum was studied.
Figure 8 shows the significant decrease in the TOC
of the test 1-octanol solution as a function of time
when the solution was kept under a 61.0 cm Hg
vacuum. The data is linear when plotted against the
square root of time, indicating that the rate of
evaporation is controlled by the diffusion of organic
to the surface of the liquid. A similar experiment
conducted with ethylbenzene showed no decrease in
TOC as a function of time. The difference in the
rate of evaporation of the two compounds is due to
the (a) their volatilities, and (b) their polarity.
Since ethylbenzene is less volatile than 1-octanol, it
evaporates at a slower rate. Also, since ethyl-
benzene is more polar than 1-octanol, it forms
stronger hydrogen bonds with water molecules, also
retarding its rate of evaporation.
This observation is very important in the
development of a suction lysimeter for sampling
organics in soil water. Organic components which
tend to volatilize easily from aqueous solution could
be lost due to evaporation. This problem can be
corrected either through capture of the evaporated
organics on an adsorbent, such as carbon. Alterna-
tively, the TOC could be corrected mathematically
using calibration data such as Figure 8. Capture
and subsequent analysis of volatilized organics
would obviously be a more desirable approach.
660
-------
The performance of the zirconia porous glass
elements in sampling organic solutions is summarized in
Tables 4 and 5. The ability of the zirconia porous glass
elements to sample the ethylbenzene solution was excel-
lent. The difference between the TOC in test and sample
solutions was always less than 3 ppm, a error of about
4%.
The tests conducted with the 1-octanol solution,
Table 5, showed the effects of octanol evaporation.
However, when the TOC measurements are corrected for
the octanol evaporation using Figure 8, the results are
quite good. For the 1 hour suction period used in these
experiments, the correction factor is 1.36. This correc-
tion factor was used to generate the column of corrected
TOC's in Table 5. The average value of the corrected
TOC's is 305.0 ppm compared to 298.9 ppm TOC in the
test solution. This represents only a 2% error.
Thus, these data, while limited, demonstrate an
excellent ability to sample organic compounds in soil
water.
2. Barbee, G., "A Comparison of Methods for
Obtaining Unsaturated Zone Soil Solution Samples,"
M.S. Thesis, Texas A&M University, College
Station, Texas, 1983, p. 78
3. Stearns, R., R. Morrison and T. Tsai, "Validity
of the Porous Cup Vacuum/Suction Lysimeter as a
Sampling Tool for Vadose Waters," University of
California Engineering Laboratory, Report CE313,
1980, p. 11
4. Dazzo, F. and D. Rothwell, Appl. Microbiol..
27, 1172(1974)
CONCLUSIONS
This paper documents the significant progress
being made toward the development of a porous glass
suction lysimeter capable of sampling organic and inor-
ganic constituents in soil water. The ability to make
porous glass elements with pore sizes as small as 2
microns with high hydraulic conductivity has been
demonstrated. Also, initial experiments indicate that the
elements can accurately sample organics and inorganics in
water. Work is continuing to optimize the preparation,
including the composition and thermal treatment, of the
porous glass elements and to develop a comprehensive set
of data on the ability of the optimized porous glass ele-
ments to accurately sample soil water.
Future work will evaluate optimized porous glass
elements with simulated and real soils, leading to the
development of a suction lysimeter using porous glass
elements.
REFERENCES
1. Brown, K. W., "Efficiency of Soil Core and Soil-Pore
Water Sampling Systems," EPA Report No. EPA/600/2-
86/083 (September 1986)
661
-------
Table 1
Pore Size and Hydraulic
Conductivity of Porous Glass Elements
TABLE 2
TESTS WITH INORGANICS IN AQUEOUS MEDIA
Element
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Mean
Std.
Dev.
Pore Size,
microns
2.3
2.1
3.5
3.5
2.3
1.6
1.1
2.3
1.9
1.3
3.5
1.2
1.6
1.6
2.1
0.8
Hydraulic
Conductivity
cm/s
1.9 x 10*
1.6
2.8
2.6
1.2
1.2
1.2
1.2
1.6
2.5
2.6
2.6
1.2
1.4
1.8 x 10*
0.6 x 10*
Inorganic
NaCl
BaCl2
PbCl2
K2Cr2O7
PH
7.0
7.0
1.8
1.8
4.4
4.4
6.2
6.2
Test
Concentration
(ppm)
196
196
69.4
69.4
102
102
94.7
94.7
Sampled
Cone.
(ppm)
206
198
67.5
66.1
66.4
63.6
81.9
91.1
MEAN
STD.
DEV.
Ratio
1.05
1.01
0.97
0.95
0.65
0.62
0.86
0.96
0.88
0.16
662
-------
TABLE 3
PORE SIZE AND HYDRAULIC CONDUCTIVITY
OF ZIRCONIA POROUS GLASS ELEMENTS
Pore
Size,
microns
3.0
3.0
2.8
3.0
3.0
Hydraulic
Conductivity,
(cm/s)
7.4E-06
7.6E-06
7.4E-06
5.4E-06
5.3E-06
TABLE 4
SAMPLING OF AQUEOUS
ETHYLBENZENE SOLUTIONS
Porous
Glass
Element
1
2
3
4
5
6
7
8
Total Organic
Carbon, ppm
Test
Solution
72.95
72.95
72.95
72.95
79.07
79.07
79.07
79.07
Solution
Sampled
74.83
72.02
75.92
78.48
76.83
76.11
78.68
78.94
Table 5
Sampling of 1-Octanol Solution!
Porous
Glau
Element
1
2
3
4
5
6
7
Total Organic
Carbon, ppm
Ten
Solution
298.85
298.85
298.85
298.85
298.85
298.85
298.85
Solution
Sampled
214.2
211.6
209.6
247.0
220.8
230.4
236.2
Sample Corrected
for Evaporation
291.3
282.8
285.1
335.9
300.3
313.3
321.2
663
-------
Densification, and Volume Porosity
of Porous Glass
vs. Sintering Temperature
% Densification
% Volume Porosity
900 950 100O 1050 1100 1150 1200 125O
Temperature [C]
—•— % Densification —°— %Vol. Porosity
FIGURE 1
-------
FIGURE 2A
Scanning Electron Micrographs
of Porous Glass Lysimeter
Elements
Firing Temperature: 1100°C
Firing Time: 60 minutes
Firing Temperature: ]]50°C
Firing Time: 60 minutes
Firing Temperature: 1200° C
Firing Time: 60 minutes
665
-------
FIGURE 2B
Scanning Electron Micrographs
of Porous Glass Lysimeter
Elements
i
Firing Temperature: 1150°C
Firing Time: 60 minutes
Firing Temperature: 1100°C
Firing Time: 60 minutes
666
-------
FIGURE 2C
Scanning Electron Micrographs
of Porous Glass Lysimeter
Elements
Firing Temperature: 1200° C
Firing Time: 60 minutes
667
-------
FIGURE 3
Lucite and Stainless Steel Holders for Porous Glass Elements
(Scale: heavy grid lines are one inch apart)
-------
cr.
to
PORE SIZE VARIATION
as a (unction of
TEMPERATURE and TIME of SINTERING
PORE SIZE (microns)
20
40 60
SINTERING TIME (min)
80
• 1200
• 1150
•1050
FIGURE 4. Pore size was measured by the
technique.
Ave.Pore Size as a Function
of the Sintering Temperature
Sintering Time: 60 min
Pora Size (microns)
900 950
Figure 5
10OO 105O 11OO 115O 1200 1250
Temperature |oC]
~m~ pore size In microns
Relation of Pore size
and Hydraulic Conductivity
with Sintering Temperaure
Pora SIZ8 [microns]
8 ~~~*— — I
4 • ^--^S
.* ,
.
.
Hydraulic Cond Icm/s
:
3; j
*5ss*v^
:
950 10OO 1050 11OO 1150 1200 12
1.000E-O4
1.000E-05
1.00OE-06
1.00OE-O7
SO
Sintering TemperaturelC]
— •— Av». Por» Slzeluml
FIGURE 6
••— Hydraulic Cond. cm/s
Hydraulic Conductivity
as a Function of
Average Pore Size
Hydraulic Cond. [cm/s]
1.OOOE-O4F
2468
Pore Size [microns]
• Porous Glass
• Porous Ceramic
FIGURE 7
-------
Octanol-Water System
Selective Octanol Evaporation Rate
During Suction Lysimetry
TOO of the Liquid Sample [ppm]
0.5 1 1.5 2
(Suction Time] "I/2 Ihr]'1/2
—- TOC
2.5
FIGURE 8
-------
COMPARISON OF MOBILE LABORATORY XRF AND CLP SPLIT SAMPLE
LEAD RESULTS FROM A SUPERFUND SITE REMEDIATION IN NEH JERSEY
Jon C. Gabry, Ph.D.
ABSTRACT
A mobile laboratory X-ray fluorescent spectro-
photometer (XRF) was utilized to determine soil
lead concentrations 1n 2,725 samples obtained
during a Superfund Site remediation In New
Jersey. These sample results assisted In
guiding remedial excavation activities at the
site. One hundred twenty-five site soil
samples were split and analyzed for lead by
the on-slte mobile laboratory utilizing a XRF
and by a USEPA Contract Laboratory Program
(CLP) laboratory utilizing atomic absorption
(AA) and/or Inductively coupled argon plasma
(ICAP) methodologies. In general, XRF results
were usually higher than the CLP split sample
results. Although unknown, XRF spectral
emission Interference and/or Incomplete homo-
genlzatlon of the sample prior to splitting
are the most probable causes of these dif-
ferences. The XRF generated duplicate and
split sample mean RPDs that were comparable or
better than those obtained from the CLP
laboratory and another EPA funded study.
INTRODUCTION
The use of XRF spectrophotometers for
elemental analysis of soil samples In
analytical field screening programs at
hazardous waste sites 1s Increasing. The
purpose of this paper Is to present a
comparison of on-slte mobile laboratory XRF
and CLP split sample results obtained from a
Superfund site remediation In New Jersey. At
this site, a XRF was utilized to determine
soil lead concentrations In 2,725 samples
obtained to guide remedial excavation
activities. The site, a former used oil
reprocessing facility. Is situated on the
coastal plain with a uniform sandy soil type
across the entire site.
METHODS
A portable X-MET 840 XRF spectrophotometer was
used In an on-slte mobile laboratory to
determine soil lead concentrations. Prior to
the analysis of any site soil samples, the XRF
was configured to the on-slte soil matrix and
calibrated. This was accomplished by
obtaining a composite of clean native site
soil that was sent to a CLP laboratory which
subsequently generated 10 spiked native soil
calibration standards verified by AA and/or
ICAP CLP methods encompassing soil lead
concentrations ranging from 20 ppm to 1000
ppm. The on-slte XRF was subsequently
calibrated with these standards using the
L-beta spectral line of lead to avoid any
possible Inter-element Interferences by
arsenic present within the site soil. All
soil samples were dried and ground with a
mortar and pestle prior to XRF analysis which
followed the Instrument manufacturer's
Instructions and utilized a counting time of
300 seconds. Quality assurance protocols
performed during sample analysis Included the
analysis of native soil blanks and continuing
calibration verification standards, and
duplicate sample analysis at a frequency of 1
per 20 samples. Based upon the analytical
data obtained, a detection limit of
approximately 20 ppm lead In soil was
estimated for the XRF.
One hundred twenty-five site soil samples were
split and analyzed for lead with the on-slte
XRF and by a CLP laboratory utilizing AA
and/or ICAP methodologies. All split samples
671
-------
were homogenized 1n the field prior to
splitting. Additionally, multiple split
samples were submitted blindly to both
laboratories as part of the quality assurance
program.
RESULTS
Split sample relative percent difference (RPD)
values for mobile lab XRF versus CLP results
ranged from 5.2 to 173.2 with a mean RPD of
77.4 ± 48.7 (n-76). These results- were
comparable to an EPA funded study (1) which
exhibited XRF versus CLP RPD results ranging
from 16.6 to 131.5 with a mean RPD of 76.5 ±
45.7 (n-6).
For Intralaboratory duplicate analyses
performed on the split samples, the CLP
laboratory exhibited a mean RPD of 20.0 ± 26.7
(n-9) whereas the on-slte XRF had a mean RPD
of 6.6 ± 5.2 (n-4). Multiple split samples
submitted blindly to the laboratories had mean
RPD's of 163.7 ± 15.7 (n=3) for the CLP
laboratory and 65.7 ± 48.7 (n-7) for the
on-s1te mobile laboratory utilizing the XRF.
Duplicate analysis performed by an Independent
laboratory contracted by the state had a RPD
of 171.8 with the two analytical results
differing by a factor of 13.2.
In general, XRF results were usually higher
than the CLP split sample results by factors
ranging from 1.09 to 13.91 with a mean factor
of 3.01 ± 3.05
-------
Screening of Groundwater for Aromatics by Synchronous Fluorescence
R. B. Gammage, J. W. Haas, III, and T. M. Allen*
Health and Safety Research Division
Oak Ridge National Laboratory
P.O. Box 2008, Oak Ridge, TN 37831-6383
BACKGROUND, PURPOSE AND SCOPE
Pollution by petroleum, oil, and lubricants is a
ubiquitous national problem. The aromatic
constituents contained in these pollutants can
generally be induced to fluoresce. The problem of
identifying individual compounds, in what is often a
complex mixture of fluorescing constituents, can be
enhanced by resorting to a technique known as
synchronous fluorescence (SF) (1).
The first application of the SF technique for
screening polynuclear aromatic (PNA) contaminants
in groundwater was described in the proceedings of
the First International Symposium on Field Screening
Methods (2). In the interim, we continued to
develop the technique and continued long-term
examination of groundwater taken from specific wells
on the Department of Energy (DOE) reservation.
Our eventual purpose is to show that SF screening is
an acceptable field screening method at Levels I and
II (3). Our efforts in achieving this end have been
slowed by limited funding. Nevertheless we are able
to report worthwhile progress; reference spectra and
minimum detection levels (MDLs) were determined
for 17 PNAs, a rapid solid-phase extraction and
concentration method was developed and multi-year
screening of groundwater from a specific well was
continued.
EXPERIMENTAL
A detailed description of the method for making SF
measurements is contained in reference 1. In making
*Gordon College, Wenham, Massachusetts
the currently reported SF measurements, we used a
Perkin Elmer LS-50 spectrometer. In order to
optimize resolving power, compound selectivity and
sensitivity, slit widths of 2.5 nm were used for both
excitation and emission light beams. The wavelength
difference between the excitation and emission
monochromators was set at the minimum possible
value for this spectrometer, which was 5 nm. The
scanning speed used to obtain the reported data was
300 nm/min.
A solid-phase extraction and concentration procedure
was devised for lowering the MLD to 10 ppb or less
of each of the 17 PNAs investigated. A home-made
cartridge packed with C18 bonded-phase material
was employed as the solid extractant. The PNA-
containing water sample (250 mL) was first forced
from a syringe through the cartridge. The exiting
PNL-free water was discarded. N-propanol (2 mL)
was next passed through the cartridge to elute the
adsorbed PNAs which were now concentrated 125-
fold. The selection of n-propanol as organic solvent
was based largely one our having at hand, n-propanol
with a low fluorescence background between 250 nm
and 500 nm.
RESULTS AND DISCUSSION
After extraction from water and concentration by
125-fold into n-propanol, calibration curves of
concentration versus SF response were determined.
The data for the 17 PNAs are summarized in
Table 1. The noted SF is the wavelength at which a
single peak or the major one of multiple peaks
occurs; 9 of the 17 PNAs produced a single peak at
a A A. of 5 nm. The MDL for each PNA is based on
a signal strength three times that of the standard
673
-------
deviation in the background. The MDLs range from
about 1 ppt for benzo(k)fluorathene to about 5 ppb
for pyrene.
A complete screening measurement can be made in
about 5 minutes; the extraction and concentration
step takes about 3 minutes and the spectroscopic
measurement takes about 2 minutes. An example of
this concentration and SF measurement is shown in
Fig. 1 for a sample of groundwater spiked with 5 ppb
of naphthalene.
The EPA Contract Laboratories Program requires
participants to be able to quantitatively analyze PNAs
on the EPA priority list at 10 ppb (4). We have
shown that the SF method with a concentration stage
(125 x) is capable, in principle, of matching this strict
requirement.
The capability of the SF method for qualitative
screening of groundwater over a period of 4 years
can be visualized by referring to Fig. 2. Between
1988 and 1990, a constituent fluorescing at about 500
nm has appeared in the groundwater taken from well
GW15; the fluorescence in the region of 280 nm
stayed essentially unchanged. It remains to identify
the composition of the entities fluorescing at 280 nm
and 500 nm and determine whether they are of
natural or anthropogenic origin.
A field screening method based on UV fluorescence
has been described by Popp et al. (5) and is listed as
Method FM-25 in reference 3. A measure of the
total PNA concentration is made using two
wavelength pairs. The method was practiced at two
wood treating sites; some samples were analyzed by
both the UV-fluorescence screening method and the
conventional Contract Laboratory Program (CLP)
GC/MS method. There was an order of magnitude
relationship between the results of the UV-
fluorescing screening and the conventional CLP
GC/MS analysis for PNAs.
It remains to apply and compare the SF screening
with the UV-fluorescence screening and CLP GC/MS
techniques. One could then evaluate the advantages
that would accrue from making more compound
specific screening measurements using SF. The SF
technique should have the greater compatibility with
the CLP GC/MS method because the SF screening
can be tailored to measuring the sum of the 18 PNA
compounds on the hazardous substances list.
CONCLUSIONS
Progress continues in developing SF as a field
laboratory, quick-screening technique for Level I and
Level II analysis of PNA in groundwater. An easy
concentration step permits analysis of individual PNA
at concentrations of 10 ppb or less. Direct SF
measurements of groundwater taken from a specific
well over a period of several years show that
qualitative changes in fluorescing constituents can
readily be followed. The next phase of development
should include comparative testing against the
conventional UV-fluorescence screening and CLP
GC/MS methods.
ACKNOWLEDGEMENTS
Research sponsored by the Office of Health and
Environmental Research and Office of Technology
Development, U.S. Department of Energy under
contract DE-AC05-84OR21400 with Martin Marietta
Energy Systems, Inc.
REFERENCES
1. Vo-Dinh, T., Gammage, R. B.,
Hawthorne, A. R. and Thorngate, J. H.,
"Synchronous Spectroscopy for Analysis of
Polynuclear Aromatic Compounds," Environ.
Sci. Technol., Vol. 12, 1984, p. 477.
2. Gammage, R. B., Haas, III, J. W.,
Miller, G. H., and Vo-Dinh, T., "Improved
Luminescence Technique for Screening
Aromatic Contaminants in Environmental
Samples," Proc., 1st Intl Symp. on Field
Screening Methods for Hazardous Waste Site
Investigations," Las Vegas, NV, Oct. 11-13,
1988, p. 51.
3. U.S. Environmental Protection Agency, "Field
Screening Methods Catalog User's Guide,"
PB89-134159, EPA/540/2-88-005, Office of
Emergency and Remedial Response,
Hazardous Site Evaluation Division, USEPA,
Washington, DC, September 1988.
4. U.S. Environmental Protection Agency,
Contract Laboratory Program, "State of Work
for Organics Analysis: Target Compound List
and Contract Required Quantitation Limits,"
Document Number OLM01.0, U.S. EPA,
Washington, D.C., October 1986.
674
-------
Fig. 1. Synchronous fluorescence spectrum of naphthalene
at 5 ppb after 125-fold concentration
240 260 280 300 3K 340 360 380 400 420 440 460 460 500 i?0 MO 560 MC
Fig. 2. Qualitative SF screening of fluorescing constituents
in groundwater from a well (GW15) on the DOE
Oak Ridge Reservation
I
GW15 AA =5
nm
300 400
Emission Wavelength (nm)
500
1990
1988
675
-------
5.
Popp, S.A. and Motwani, J. N. "UV
Fluorescence Field Screening Technique
Developed and Utilized Under the Superfund
Program," Hazardous Wastes and Hazardous
Materials Conference, New Orleans, LA,
April 12-14, 1989.
Table 1. Synchronous fluorescence maximum and minimum
level of detection after 125-fold concentration
of 17 PNAs on the hazardous substance list
Compound
Fluorene
Naphthalene
Acenaphthylene
Acenaphthene
2-Methylnaphthalene
Phenanthrene
Chrysene
Pyrene
Anthracene
Benzo(a)anthracene
Dibenzo(a,h)anthracene
Benzo(b)fluoranthene
Fluoranthene
Benzo(h)fluoranthene
Benzo(a)pyrene
Benzo(g,h,i)perylene
Indeno( 1 ,2,3-c,d)pyrene
SF Maximum
(nm)
299.5
311.5
316.0
316.2
319.5
339.5
355.0
366.1
374.0
383.0
392.7
392.7
396.6
400.5
402.5
433.1
459.1
MLD*
3ppt
1.8 ppb
1.6 ppb
22ppt
0.4 ppb
0.4 ppb
0.8 ppb
5.2 ppb
13ppt
0.6 ppb
23ppt
4.2 ppb
5.5 ppb
Ippt
lOppt
3.8 ppb
30 ppb
*PNA in propanol
Perkin Elmer LS 50 Spectrometer
2-1/2 minute scan
2.5 nm slitwidths
= 5 nm
676
-------
IN SITU DETECTION OF TOXIC AROMATIC COMPOUNDS IN GROUNDWATER
USING FIBEROPTIC UV SPECTROSCOPY
J. W. Haas HI, T. G. Matthews, and R. B. Gammage
Health & Safety Research Division
Oak Ridge National Laboratory
P.O. Box 2008, Oak Ridge, TN 37831-6113
INTRODUCTION
Contamination of groundwater with organic compounds is a
common problem at Department of Energy (DOE) facilities
and other sites. Among the more prevalent contaminants
are benzene, toluene, ethyl benzene, and xylenes (BTEX)
which are used individually as solvents and are also major
components of gasoline and other fuels. Leaking
underground fuel storage tanks are significant contributers
to groundwater pollution. Polycyclic aromatic hydrocarbons
(PAH) are also possible groundwater contaminants,
originating from fuel leaks or other sources.
Because aromatic contaminants are so ubiquitous, two needs
related to their detection have arisen:
1. Rapid, cost-effective screening methods are
needed as an alternative to slow, expensive conventional
analyses.
2. The fate of aromatic contaminants in groundwater
needs to be determined. An understanding of paths and
rates of migration or biodegradation is crucial to the design
of effective remediation strategies.
One approach that can meet both of these needs is a sensor
that can detect aromatic compounds directly in groundwater.
Our previous experience with derivative ultraviolet
absorption spectroscopy, DUVAS (1-3), suggested that this
might be a useful tool for this application. At the first
meeting of this Symposium, we presented results
demonstrating the feasibility of performing groundwater
analysis in situ using a fiberoptic probe (4). Here we report
on the development of a field-portable spectrometer and a
fiberoptic probe based on that earlier work.
EXPERIMENTAL
Spectrometer construction. An Sin x 14in x 6in aluminum
box was used for the spectrometer (see Figure 1). Although
a more powerful xenon lamp was used for the light source
in previous benchtop experiments, a deuterium lamp
(Hamamatsu, Bridgewater, NJ) was employed in this work.
Higher UV output was obtained from the monochromator
as a result of the superior focal characteristics of the
deuterium lamp. A homemade power supply was used with
the lamp. An additional supply was included in the
spectrometer to power the CVI (Albuquerque, NM) Model
DK120 (Albuquerque, NM) monochromator (110mm focal
length) and the photodiode detector housed in the probe
(see description below). The monochromator was stepped
via an external DK-1200 controller, also manufactured by
CVI. Both power supplies were powered from a small car
battery linked through an inverter. Light from the
deuterium lamp was coupled into the monochromator
through an f/1 fused silica lens. Light emerging from the
monochromator was coupled into the optical fiber through
another f/1 lens. The signal voltage returning from the
probe to the spectrometer was sent to an IBM-compatible
personal computer containing a Data Translation
(Marlboro, MA) Model DT-2811 A/D board. Data
collection and processing was handled with modified
SpectraCalc (Galactic Industries, Salem, NH) software.
Fiberoptic probe. Previous experiments (4) showed that a
one-fiber, detector-in-probe design could increase
significantly the maximum sensing distance of a probe using
UV light over a more traditional two-fiber design. The
design was used to produce the in situ probe diagrammed in
Figure 2. The probe will be described in detail elsewhere,
however a brief description follows. A 25m, 600um core,
high-OH, all-silica optical fiber (Polymicro Technologies,
Phoenix, AZ) was used to bring UV light to the probe. The
light emerging from the fiber was focused through a 1cm
optical sample path onto a photodiode detector (United
Detector Technology, Hawthorne, CA) protected by an
optical flat. Water entered the optical path through slits in
the side of the probe. Photodiode power and the detected
signal were transmitted through 5-conductor cable. Both
the optical fiber and electrical cable were protected by a
677
-------
thick-walled air hose connected between the probe and the
spectrometer. The probe body was constructed of stainless
steel and was about 6in length and l1Ain diameter, fitting
easily into 2in or 4in diameter groundwater monitoring
wells.
Samples. Benzene standards were prepared by diluting
concentrated methanol solutions with distilled water. The
contaminated groundwater samples were collected from well
GW-15 at the Bear Creek Burial Grounds on the Oak
Ridge Reservation.
RESULTS AND DISCUSSION
Analytical capabilities. The probe was first tested under
laboratory conditions to determine its capability for long-
distance measurements. Improved signal processing and
reduced electronic noise allowed detection of benzene at
concentrations down to less than lug/mL when a 50m fiber
was used. PAH can be determined even more sensitively,
with lower detection limits ranging to below Ing/mL for
compounds such as anthracene. For benzene, a linear
calibration of absorbance vs. concentration was obtained
regardless of fiber length.
Figure 3 is a spectrum of lug/mL benzene spiked into
uncontaminated groundwater. A combination of Fourier
filtering, smoothing, and second derivative signal processing
was used to produce the characteristic benzene "fingerprint"
(Figure 3B) from the almost featureless transmission trace
(Figure 3A). One advantage of the second derivative
approach is that it tolerates reasonable levels of sample
turbidity without need for a double-beam design, which
would be difficult to incorporate into our probe. However,
if a sample is so turbid that little or no light reaches the
detector, then a measurement can not be made.
Groundwater samples. The Bear Creek Burial Grounds on
the Oak Ridge Reservation has a significant groundwater
contamination problem. Pollutants include chlorinated
solvents and light aromatic solvents. Over a period of 5
years we have been monitoring benzene and toluene
concentrations in groundwater wells at the site. Figure 4
tracks benzene concentrations in one well (GW15); the
trend is similar in other wells. Although benzene levels
dropped by a factor of 5 in the first 30 months, they have
changed little since 1989, levelling off at a disconcerting
20ug/mL.
In order to evaluate the performance of the fiberoptic
DUVAS system, we compared analytical results for GW15
water with results obtained using a laboratory DUVAS
instrument and independent gas chromatographic analysis
(Purge Method 5030 and GC Method 8000). The
comparison is summarized in Table I.
Table I
Benzene in GW15 - Comparison of Methods
Method
In situ DUVAS
Ex situ DUVAS
GC
1
2
1
2
Cone. (ug/mU)
18.8
17.9
17.2
24.2
17.4
Because the GC method involved several sample
preparation steps and relied on a one-point calibration for
quantitation, it is not surprising that the greatest variation
was observed in the GC data. Total GC analysis time was
about 3 hours for the two samples (including the standards).
In contrast, it took approximately 30 minutes to run a five-
point calibration curve and the two samples using the
DUVAS instruments (no sample preparation was required).
Clearly, either DUVAS approach offers considerable cost
savings over the GC method. Again, the laboratory
DUVAS method also demonstrated better precision than
the GC method. The fiberoptic DUVAS results were close
to those of the laboratory instrument, clearly demonstrating
that analytical performance is not compromised in the field
instrument. The results also suggest, at least for this well-
mixed shallow well, that samples collected with a bailer and
properly contained are representative of actual groundwater
concentrations. It is also notable that an exceptionally short
holding time (about 20 hours) was used prior to laboratory
analysis of the samples. This, of course, is not typical.
CONCLUSIONS
A fiberoptic DUVAS probe and field portable spectrometer
have been fabricated and tested. The instrument provides
reliable measurement of aromatic contaminants in
groundwater, as demonstrated at a local groundwater well.
In future work, we plan to use the device at a jet fuel spill
site where current analytical results are ambiguous. Depth
profiling within undisturbed wells will be conducted to help
locate the fuel, believed to be in a narrow subsurface zone.
ACKNOWLEDGEMENT
This research was sponsored by the office of Health and
Environmental Research and office of Technology
Development, U.S. Department of Energy, under contract
No. DE-AC05-84OR21400 with Martin Marietta Energy
Systems, Inc.
678
-------
REFERENCES
1. Hawthorne, A.R., Thorngate, J.H., Gammagc, R.B., and
Vo-Dinh, T., "Trace Organic Analysis Using Second-
Derivative UV-Absorption Spectroscopy," Proc. 9th
Materials Research Symposium, Gaithersburg, MD, April
10-13, 1978, p. 719.
2. Hawthorne, A.R., Thorngate, J.H., Gammagc, R.B., and
Vo-Dinh, T, "Development of a Prototype Instrument for
Field Monitoring of PAH Vapors," Polynuclcar Aromatic
Hydrocarbons, Ann Arbor Science, Ann Arbor, MI, 1979, p.
299.
3. Hawthorne, A.R., Morris, S.A., Moody, R.L., and
Gammage, R.B., "DUVAS as a Real-time, Field-portable
Wastewater Monitor for Phcnolics," J. Environ. Sci. Health,
A19(3), 1984, 253.
4. Haas III, J.W., Lee, E.Y., Thomas, C.L., and Gammage,
R.B., "Second-Derivative Ultraviolet Absorption Monitoring
of Aromatic Contaminants in Groundwater," Proc. 1st Intl.
Symp. on Field Screening Methods for Hazardous Waste
Site Investigations, Las Vegas, NV, Oct 11-13, 1988, p. 105.
TO BATTERY
A
8in
T
LAMP
POWER
SUPPLY
TO COMPUTER
POWER SUPPLY
DEUTERIUM
LAMP,
D-
LENS
MONOCHROMATOR
TO PROBE
14in
Figure 1. Portable spectrometer for use with fiberoptic DUVAS probe.
679
-------
PHOTODIODE DETECTOR
AND AMPLIFIER
FOCUSING
LENS
OPTICAL FIBER
PROTECTIVE
HOSE
OPTICAL PHOTODIODE POWER
FLAT AND RETURN SIGNAL
Figure 2. Cut-away view of fiberoptic DUVAS probe.
g
w
OT
I
200
250
WAVELENGTH (nm)
300
LLJ
m
DC
o
CO
m o
UJ
Q
200
250
WAVELENGTH (nm)
300
Figure 3. (A) Transmission and (B) second derivative spectra
of benzene in groundwater using the fiberoptic DUVAS.
680
-------
200-
0)
ID
LU
O
"Z.
O
O
100-
0
1986
1988
1990
1992
YEAR
Figure 4. DUVAS monitoring of benzene in groundwater well GW15.
681
-------
DEVELOPMENT OF FIELD SCREENING METHODS FOR
TNT AND RDX IN SOIL AND GROUND WATER
Thomas F. Jenkins and Marianne E. Walsh
U.S. Army Cold Regions Research and
Engineering Laboratory
Hanover, New Hampshire
Martin H. Stutz and Kenneth T. Lang
U.S. Army Toxic and Hazardous Materials Agency
Aberdeen Proving Ground, Maryland
INTRODUCTION
One of the most serious environmental problems facing the
Army is the presence of soil contaminated with residues of high
explosives at sites where the munitions were formerly manu-
factured, stored, used or demilitarized. TNT and RDX are the
two residues most commonly encountered because these ex-
plosives were extensively produced and do not rapidly decom-
pose. Since TNT and RDX leach through the unsaturated zone
with downward percolating water, they pose an immediate
problem to ground water; thus contaminated soil must be
treated or isolated. Though laboratory methods for analyzing
munitions residues in soil and water are now available (1,2),
reliable field methods are also desirable so that zones of high
contamination can be located during initial surveys and the
interface between clean soil and contaminated soil identified
during cleanup.
DESCRIPTION OF METHODS
The procedures for the soil (3,4) and water methods are
similar (Fig. 1). For the soil method about 20 g of soil is shaken
with 100 mL of acetone to extract the munitions residues and
the extract is filtered using a disposable syringe filter. The
methods then depend on the production of colored reaction
products when separate aliquots of these extracts are subjected
to two simple reaction sequences (Fig. 2). For TNT, a portion
of the extract is reacted with a strong base, and if TNT is present,
the reddish colored Jackson-Meisenheimer anion is produced.
Several other trinitroaromatics also produce reddish anions and
hence are potential interferences (3,5). For RDX another por-
tion of the extract is passed through a disposable anion ex-
change cartridge to remove any nitrate or nitrite. Then the
extract is acidified and reacted with powdered zinc (4). This
converts RDX to nitrous acid, which is detected by adding a
Hach NitriVer 3 powder pillow (Fig. 2). The development of a
red or orange color is indicative of the present of RDX or one
of several other military explosives that are potential interfer-
ences (HMX, nitroglycerine, PETN or nitrocellulose). The
intensity of the color produced can be measured with a battery-
operated spectrophotometer. The absorbances at 540 nm for
TNT and 507 nm for RDX are linearly related to concentration
(3,4). Detection limits are about 1 ug/g for both TNT and RDX.
SOIL SAMPLE
Extract With Acetone
TNT PR<
>CEDURE
| Obtain Initial Absorbance (540 nm) |
| Add KOH + Na2 SO3 |
| Fil
ter |
| Obtain Absorbance (540 nm) |
Janowsky Reaction (1 886)
I
RDX PROCEDURE
| Pass Through Anion Exchanger
| Add Zinc and Acetic Acid |
| Filter |
I Add Nitri Ver 3 Powder Pillow
| Obtain Absorbance (507 nm)
Griess Reaction (1864)
WATER SAMPLE
[Pass Through Solid Phase Extraction Cartridge |
| Elute With Acetone |
^P
Process Extract as Shown Above
Figure 1. Flow diagram for field methods.
683
-------
TNT Method
CH3
CH3
20%) of
water are present, the absorbance is also low. At intermediate
concentrations of water in acetone (1-17%), however, similar
absorbances (±15%) are obtained. If a 20-g sample of wet soil
is extracted with 100 mL of acetone, the 1-17% range of water
in acetone would correspond to soil moisture contents ranging
from 4-83% (on a wet weight of soil basis). This range of
moisture content is typical of the large majority of surface soils.
3. Reagent contact time. Several experiments were con-
ducted to determine if reagent contact time (1-20 min) had an
effect on the absorbance obtained. TNT solutions were pre-
pared in acetone containing 3.8% water. The results indicate
that the absorbance first increases and then declines with con-
tact time, the times being somewhat concentration dependent.
One interpretation is that the rate-limiting step is dissolution of
the solid reactants. If this is true, the concentration of water in
the acetone is also likely to have an effect, and the optimum
reagent contact time will be sample specific. The rate of decline
of absorbance for excess contact time, however, is relatively
slow. The reason for the reduction in absorbance for longer
contact times is discussed elsewhere (3,6).
Based on these tests, a reagent contact time of three minutes
was selected. For high TNT concentrations, three minutes may
be insufficient to attain the maximum absorbance, but the
absorbance will exceed 1.0 A.U. for this case and extracts will
have to be diluted anyway. For very low TNT concentrations or
solutions with less dissolved water, the measured absorbance
will be reduced but the reduction will be very small. For field
measurements, the ambient temperature can have an influence
on the proper reagent contact time. Observation of the rate of
color development for the standard will assist in selecting the
most appropriate time for a given temperature.
4. Stability of filtered solutions. A test was conducted to de-
termine if the colored anions formed from TNT were stable
with time after filtration. Filtration removes the colored anions
from further contact with the solid reactants and results indicate
that absorbance measurements are reliable for at least two hours
(3).
5. Comparison of TNT concentration estimates for soil ex-
tracts. A series of field-contaminated soils were extracted with
acetone and the extracts were analyzed using the field method
684
-------
Table 1. Comparison of colorimetric and RP-HPLC analysis of soil extracts.
TNT concentration TNB concentration
Colorimetric
Field Method
Sample origin (Hg/g)
Vigo Chemical Plant (Ind.)
Hawthorne AAP (Nev.)
Nebraska Ordnance Works (Neb.)
Nebraska Ordnance Works (Neb.)
Hastings East Indus. Park (Neb.)
Weldon Springs Training Area (Mo.)
Sangamon Ordnance Plant (II.)
Weldon Springs Training Area (Mo.)
Hawthorne AAP (Nev.)
Nebraska Ordnance Works (Neb.)
Raritan Arsenal (N.J.)
Nebraska Ordnance Works (Neb.)
Lexington-Bluegrass Depot (Ky.)
Chickasaw Ordnance Works (Tn.)
Hawthorne AAP (Nev.)
Weldon Springs Training Area (Mo.)
Hawthorne AAP (Nev.)
Raritan Arsenal (N.J.)
Nebraska Ordnance works (Neb.)
Nebraska Ordnance works (Neb.)
Nebraska Ordnance works (Neb.)
Nebraska Ordnance works (Neb.)
Nebraska Ordnance works (Neb.)
Nebraska Ordnance works (Neb.)
13.5
5.49
2.39
592
85.3
4.02
32.7
145
8.67
146
85.3
0.38
15.0
-------
6y the fab method for both RDX alone and the sum of RDX and
HMX (Table 1). The estimates of RDX concentration obtained
by the field procedure were not significantly different from
those obtained by the HPLC procedure for RDX alone or for the
sum of RDX and HMX.
Field Testing
The soil methods have been field tested at Umatilla, Oregon,
Newport, Indiana, Camp Shelby, Michigan, and Eagle River
Flats, Alaska. The methods were found to be usable under field
conditions and the estimates of analyte concentrations corre-
late well with estimates obtained by the standard laboratory
procedures (3,4).
REFERENCES
I.Jenkins, T.F., Walsh, M.E., Schumacher, P.W., Miyares,
P.M., Bauer, C.F., and Grant, C.L., Liquid chromatographic
method for the determination of extractable nitroaromatic
and nitramine residues in soil. Journal of the Association of
Official Analytical Chemists. 72: 890-899 (1989).
2. Jenkins, T.F., Miyares, P.H. and Walsh, M.E., An improved
RP-HPLC method for determining nitroaromatics and nitra-
mines in water. U.S. Army Cold Regions Research and En-
gineering Laboratory, Special Report 88-23, Hanover, New
Hampshire (1988).
3. Jenkins, T.F., Development of a simplified field method for
the determination of TNT in soil. U.S. Army Cold Regions
Research and Engineering Laboratory, Special Report 90-
38, Hanover, New Hampshire (1990).
4. Walsh, M.E. and Jenkins, T.F., Development of a field
screening method for RDX in soil. U.S. Army Cold Regions
Research and Engineering Laboratory Special Report, Ha-
nover, New Hampshire (in press).
5. Bost, R.W. and Nicholson, F., A color test for the identifica-
tion of mono-, di- and trinitro compounds. Industrial Engi-
neering Chemistry (Analytical Edition). 7:190-191 (1935).
6. Terrier, F., Rate and equilibrium studies in Jackson-Meisen-
heimer complexes. Chemical Reviews. 82: 77-151 (1982).
686
-------
QUANTIFICATION OF PESTICIDES ON SOILS BY
THERMAL EXTRACTION-GC/MS
T. Junk, T. R. Irvin
Institute for Environmental Studies
Louisiana State University
Baton Rouge, LA 70803
K. C. Donnelly, D. Marek
Agronomy Field Laboratory
Texas A & M University
College Station, TX 77943
Introduction
Site investigations and cleanup activities often
require the rapid analyses of soil samples for
semivolatile environmental toxicants. The
widespread use of pesticides makes the
development of rapid field-deployable
quantification methods for this class of
compounds particularly desirable. Recently,
thermal extraction techniques have been
investigated as a rapid alternative to classical soil
analyses by solvent extraction-GC/MS. Samples
are heated according to a preprogrammed
temperature profile and evolvig volatiles analyzed
by in-line GCMS methods. Thus, tedious wet
extraction procedures are completely eliminated.
However, the quantification of toxicants by this
procedure poses problems. Thermal extraction
efficiencies for toxicants do not necessarily reflect
solvent extraction efficiencies. Indeed, they are
typically lower and depend on the physical and
chemical properties of matrices (soils) as well as
those of analytes. Furthermore, toxicant extraction
and analysis are combined into one procedural
step and cannot be monitored seperately using
internal standards and surrogate standards.
Isotopic dilution mass spectroscopy, on the other
hand, postulates the free exchange of labeled and
non-labeled analyte molecules in a sample, an
assumption that is generally correct for solutions,
but questionable for solid samples. Thus, a
comparison study of thermal and wet extraction
procedures was undertaken using pesticide-
-containing soil standards with pesticide contents
ranging from 1 ppm to 1000 ppm. Analytical
results obtained with a thermal extraction system
were compared to those obtained by Soxhlet
extraction and subsequent analysis using a
conventional gas chromatograph coupled to an ion
trap mass spectrometer.
Instrumentation
Thermal extractions were performed on a Pyran
Thermal Chromatograph coupled tob a Finnigan
ion trap mass spectrometer. The system features
an all-quartz analytical flow path to minimize
catalytic sample decomposition and is fully
automated to eliminate operator-induced
variations. Soil samples were weighed into porous
quartz crucibles and heated in the pyrocell
compartment of the analyzer according to a
preprogrammed temperature profile. All volatile
components released during the heating phase
were flushed through a splitter assembly by
helium carrier gas and subsequently
cryo-condensed onto a fused silica GC column.
Analyte identification/quantification followed
conventional GC/MS procedures. The analyte
mixture was separated on a fused silica capillary
column and analytes were identified and
quantified by an in-line Finnigan ion trap mass
spectrometer. For this study, the system was
equipped with a 0.32 mm x 15 m x 2.5 urn DB-5
column.
Soil extracts were analyzed on a Varian 3500
Series gas chromatograph using the same type of
column as before. The identical Finnigan ion trap
mass spectrometer was used for analyte
identification and quantification (by
disconnecting-reconnecting the transfer line).
Experimental
A total of 18 soil samples was prepared and
analyzed in triplicate by both methods: six
samples each of Bastrop, Padena, and Weswood
soils contaminated with 1000, 500, 100, 10, 5, and
1 ppm pesticides and spiked with two surrogate
standards. Both methods were optimized
independently; different column temperature
687
-------
programming was used along with different
tuning parameters for the mass spectrometer.
Quantification for the thermal extraction system
was achieved independently by using the two
surrogate standards added to the soil samples
during their preparation and by spiking
isotopically labeled pesticide analogs onto the soil
samples. Quantification for GC/MS was achieved
with surrogate and internal standards in the usual
manner. All quantifications were based on peak
area ratios of analytes to standards for selected ion
chomatograms. The following analytes were
chosen: a-hexachlorocyclohexane, y-hexachloro-
cyclohexane, aldrin, endosulfan, bis-
(chlorophenyl)trichloroethane (DDT), and bis-
(chlorophenyl)dichloroethane (DDE). For
internal standards, 3,4,5,6-tetrachloro-2-xylene ,
9-bromophenanthrene, 2-bromofluorene, and
4-bromobiphenyl. For isotopic dilution
quantification, hexachlorocyclohexanes-13C6,
adrin-13C5, endosulfan-D4, DDE-D8, and
DDT-D8. Three soils were chosen for this series
of experiments, Weswood soil, a sandy,
organic-lean soil, Padena soil, organic-lean with
high clay content, and Bastrop soil, a clay rich
topsoil. The variation of pesticide contents in the
samples over three orders of magnitude
necessitated the use of variable split ratios
between 1:10 and 1:40 and variable sample
weights between 10 and 200 mg. Isotopic
standards were spiked as solutions in
dichloromethane directly onto the soil samples
immediately before thermal extraction. The
thermal extraction system was calibrated by
determining the peak area ratios for equal
concentrations of analytes and standards as
average over four runs. Quantifications for the
conventional GC system were based on the
average of three six point calibration curves (50 -
550 ng pesticides injected). Mass spectra were
acquired in full scan mode, 64-400 amu.
Conclusions
The Pyran system was shown to be capable of
providing rapid (35 min) analyses of different
soils for most of the pesticides included in this
study. Virtually no background signal from
organic materials contained in the soils was
observed and clean total ion chromatograms were
obtained. Problems with pesticide decomposition
were encountered for dieldrin, endosulfan, DDT,
and to a lesser extent for ODD. While DDT and
DDD underwent dehydrochlorination to alkenes,
endosulfan suffered loss of sulfur dioxide with
subsequent ring closure to the corresponding
isobenzofuran derivative, which was thermally
extractable and quantifiable. Dieldrin was not
thermally extractable; most likely due to its
conversion to the corresponding diol by traces of
water contained in the soils. DDE formed by
dehydrochlorination of DDT is indistighuishable
from DDE contained in the sample. As a result,
recovery values found for DDE in the presence of
DDT can be regarded as artificially high.
Comparison of recoveries and percent standard
deviations of recovery based on isotopic dilution
quantification and internal standard quantification
allowed us to distinguish between deviation of
found from actual pesticide content due to
variations of the thermal extraction process in the
pyrocell (e.g. uneven packing of the soils in the
sample crucibles, variable helium flow through
the soils etc., which will effect pesticides and
internal standards in the same manner, since both
are uniformly distributed throughout the soils) and
variations due to differences in the thermal
desorption behavior for different chemicals (wich
will not show for pesticides and their chemically
identical isotopic analogs). The results showed
that no free exchange exists of adsorbed pesticides
and their isotopic analogs spiked onto the soils
before analysis. Calculated recoveries based on
internal standard quantification generally
decreased as pesticide contamination decreased,
as shown for Weswood soil samples. However,
this did not hold for all pesticides: poor thermal
extraction of standards in combination with high
thermal extraction of analytes may result in
calculated recoveries in excess of 100 percent. A
pronounced dependence of extraction efficiencies
on the type of soil analyzed was demonstrated.
Soils with high clay content will generally allow
lower recoveries than sandy soils for internal
standard quantification.
Conventional solvent extraction was clearly less
dependent on the chemical nature of the extracted
analytes and their concentrations in the soils than
thermal extraction. These results etablish thermal
extraction as analytically useful tool for the rapid
semiquantitative, in some cases quantitative,
analysis of soil samples for semivolatile
pesticides; however, with somewhat lower analyte
recoveries and higher deviations than those
obtained in conventional procedures.
688
-------
A PORTABLE GAS CHROMATOGRAPH WITH AN ARGON IONIZATION DETECTOR
FOR THE FIELD ANALYSIS OF VOLATILE ORGANICS
Lawrence Peter Kapllnr rt.»mllf
Roy F. Weston, REAC Project
GSA Raritan Depot, Building 209 Annex
2890 Woodbridge Avenue
Edison, NJ 08837
Thomas Henry Prichett, Chemist
US EPA, Environmental Response Team
GSA Raritan Depot, Building 18
2890 Woodbridge Avenue
Edison, NJ 08837
ABSTRACT
The Environmental Response Team of the US EPA
(ERT-EPA) has been deploying field portable gas
chromatographs (CGs) for the characterization of Su-
perfund sites and landfill throughout the country. Port-
able GCs allow rapid determination of volatile organic
compounds (VOCs) and can both identify and quantify
these compounds. Several researchers as well as the
ERT have used data generated via portable GCs to
estimate and model contamination plumes, fugitive
emissions, and to direct remediation activities (1,2).
The Sentex portable GC, equipped with a high energy
ionization detector, has been shown to be invaluable for
determining the levels of VOCs at several EPA super-
fund sites.
The argon ionization detector (AID) has a high ioniza-
tion energy and will yield a response form a wide variety
of compounds whose ionization potential is at, or below,
11.7 electron volts (11.7 eV). These will include many
aromatic, chlorinated alkanes, chlorinated alkenes, and
nitrogen and sulfur containing compounds. Of par-
ticular interest arc the halogenated alkanes which do not
respond well using photoionization detector based port-
able GCs (3). The same AID can be easily converted to
a Electron Capture Detector (BCD). The ECD is very
responsive to halogenated compounds. The ECD works
best using nitrogen instead of argon as the carrier gas.
Carrier gases can be easily switched in the field to take
advantage of the dual detector capabilities of the Scntcx
Scentograph GC. For general screening on a wide
variety of compounds the ERT has used the Sentex GC
in the AID mode for most of the Superfund sites inves-
tigated.
INTRODUCTION
TheSentexSensingTechnologies,Inc.(Ridgciield,NJ.)
model Scentograph is a totally portable GC operating
with a 11.7 Ev AID. Internal cylinders of argon carrier
gas and calibration gas, as well as a 12 volt DC battery
pack allows the Sentex GC to operate without external
support from six to fourteen hours, depending on flow
and oven temperature. All operations are controlled by
the portable lap top computer (PC) interfaced with the
GC. The-PC permits the Scentograph to be automated
and therefore can be set up to run unattended. An
optional communications software package and modem
can be used to control and operate the GC remotely via
phone links. The PC will also archive all raw and
processed data as well as initial operating parameters.
The Sentex Scentograph GC software can identify a total
of 16 compounds stored in the current operating calibra-
tion library. A one point, one for each of the 16 possible
compounds, calibration is used to quantify the identified
compounds. A post analysis software routine allows
sample run to be compared to additional libraries ,
thereby allowing identification / quantification against
hundreds of compounds.
THE SENTEX GAS CHROMATOGRAPH
The GC system itself consists of three major com-
ponents: the programmable sampling pump and adsorp-
tion trap, the temperature programmable gc column and
detector block, and the PC data system.
The programmable sampling system and trap consists of
an internal sampling pump which can be programmed
via the PC to draw a sample for various periods of time.
Pump duration ranges from 1 to 999 seconds at a typical
flow rate of 100 cubic centimeters per minute. The
sample is drawn onto an adsorption trap of either Tenax
or Carbosieve where the sample components are con-
centration on the surface of the trapping material. The
trap is then heated from 1 to 4 seconds and backflushcd
to thermally dcsorb the concentrated sample com-
ponents off the adsorption trap and onto the GC separa-
tions column. Various trapping materials besides Tenax
and Carbosieve are available.
The temperature programmable block provides a stable
heated zone, from 30 ° C to 140 ° C for both the GC
analytical separations column and the GC detector. The
689
-------
Sentex block heater yields very stable temperatures,
reducing peak retention time drift present in other
field portable GCs. Since the heater is of a high mass
block design cool down times between run cycle can
be prohibitively long. The temperature ramping
routines available in the software have been found to
be impractical for most of the rapid screening needs
of the ERT. Consequently, all GC field operations, to
date, have used isothermal oven temperatures, typi-
cally 30 to 80° C. The GC oven is very small at 3" high,
3" wide, and 6" long. This constraint has made only
packed GC columns usable. Recent modifications of
the oven dimensions has allowed for the use of mega-
bore capillary columns. At present capillary columns
usable with the Sentex GC are only available through
Sentex.
The dual AID / ECD detector system will respond to
most compounds of environmental interest in either
one mode or the other. The AID has been used
predominately because it detects both aromatic and
chlorinated hydrocarbons down to the low parts-per-
billion, volume (ppbv) range. The AID has been
found to be very stable and equilibrates within one
half hour after initial setup. A grossly contaminated
AID can be easily reconditioned by baking out the
system at an elevated temperature for a short period
of time. Field experience with the GC configured in
the ECD mode has found the detector to take several
hours to stabilize. The ECD is also more sensitive
than the AID to the compound it responses to. It can
be more easily contaminated and may take several
hours at an elevated temperature to recondition a
contaminated ECD. Both the AID and the ECD have
a linear dynamic range of only 2 to 3 orders of mag-
nitude and be easily saturated at the higher parts-per-
million, volume (ppmv) concentration range. The
dual detector system operates best at the 10 to 1000
ppbv range. For most of the field screening needs of
the ERT this range is suitable. The AID / ECD detec-
tor uses a radioactive tritium foil (Ha) as a beta energy
source. A modified NRC license available through
Sentex is required No wipe tests are required and air
shipping is not a problem since the activity of the foil
is below DOT restrictions.
CONCLUSION
The Sentex Scentograph gas chromatograph has been
used for the analysis of volatile organic compounds at
various EPA Superfunds by the US EPA Environ-
mental Response Team. Detection limits have ranged
from 5 to 50 ppbv for various aromatic and
chlorinated compounds, when using the argon ioniza-
tion detector. Screening applications include ambient
air analysis, indoor air, stack emissions, and soil gas
surveys. A wide variety of aromatic and halogenated
hydrocarbons have been investigated ( Table 1). The
Sentex GC has yielded data that compared well to
other conformational analysis, such as GC / MS
(Table 2). In several cases the Sentex GC was the only
field portable GC that could detect chlorinated
alkanes, in the field, at the low ppbv range.
At present only vapor phase sample matrices have
been sampled by the ERT. Optional equipment can
be used to dynamically purge volatiles from soil and
water matrices for subsequent GC analysis. Initial
evaluation of this optional "purge and trap" apparatus
has shown detection limits for benzene, toluene and
total xylenes in soils to be in the low to mid ppbv range,
depending on soil matrix and GC operating condi-
tions.
REFERENCES
1) Spittler, Thomas and W.S. Clifford. " A New
Method for Detection of Organic Vapors in the
Vadose Zone." NWWA Conference Proceedings on
" rharacterJ7atinn and Mnnitnrinp of the Vadose
Zone.", Denver, CO, 1985
2) Clay, P.F. and T.M. Spittler. " The Use of Portable
Instruments in Hazardous Waste Site Charac-
terization." Proceedings of the National Conference
nn " Managpnipnt of Uncontrolled Ha7ardniis Waste
Sites.", HMCRI, Silver Springs, MD, 1985
3) Kaelin, Lawrence and T.H. Pritchett. " Analytical
Protocols for Portable Gas Chromatographs as Used
by the US EPA Environmental Response Team."
nf the " 198th ACS National Meeting.
Fnvirnnmental Chemistry Division.". Miami Beach,
FL.1989.
690
-------
Table 1
COMPOUNDS DETECTED VIA SENTEX AID
IN THE FIELD BY THE US EPA / ERT
Benzene
Toluene
o - Xylene
m,p - Xylene
Methyl chloride
Ethyl chloride
Vinyl chloride
Methylene chloride
1,1 Dichloroethane
1,2 Dichloroethane
1,1 Dichloroethene
trans 1,2 Dichloroethene
1,1,1 trichloroethane
Bis 2 chloroethyl ether
Trichloroethylene
Tetrachloroethylene.
Table 2
COMPARISION OF FIELD DATA FOR VINYL
CHLORIDE SOIL GAS SAMPLES
SENTEX GC GC / MS (tube)
1.15 ppmv 0.54 ppmv
1.01 ppmv 0.82 ppmv
2,45 ppmv 3.27 ppmv
ND(<0.005) ND(<0.01)
0.20 ppmv 0.42 ppmv
0.18 ppmv 0.79 ppmv
ND(<0.005) ND(<0.01)
0.82 ppmv 1.38 ppmv
7.43 ppmv 4.0 ppmv
0.006 ppmv ND(<0.01)
691
-------
SEAMIST — A Technique for Rapid and Effective Screening of
Contaminated Waste Sites
Carl Keller
Bill Lowry
The SEAMIST system was developed to allow
the insertion and removal of absorbent
collectors in long drillholes of marginal
stability. However, the technique has
such attractive attributes that its use
is being extended to many other aspects
of instrumentation and sampling from
drillholes. The name SEAMIST is an
acronym for Science and Engineering
Associates Membrane Instrumentation and
Sampling technique. The technique is
simple though not obvious.
The principle feature is a hole liner
made of a tubular fabric or film called
an "impermeable membrane" (Figure I).
The membrane lines the drillhole and is
pressed against the hole wall by a modest
internal pressure (1-3 psi). The bottom
of the membrane is gathered together
(inside out) and tied with a cord, "the
tether", which extends up the center of
the hole to a reel, in a canister, at the
surface. The top of the membrane is
attached to a short pipe extending from
the canister. The function is simply
that turning the reel winds up the tether
and inverts the membrane, peeling it
outside in from the hole wall. The
entire membrane can be wound onto the
reel, inside out. Reversing the reel
allows the membrane to reverse its
motion, extending down the hole under
pressure and everting as it descends to
re-1ine the hole.
CRANK
CANISTER
BASE PIPE
Figure 1. Components of the SEAMIST system.
693
-------
As the everting membrane descends/ it
provides support of the hole wall. It
also lines the hole like a continuous
packer and prevents flow into the hole.
The primary utility of that function is
that the membrane can carry instruments
into the hole by their being fastened to
the membrane (e.g./ thermo-couple/
absorbent pads, fiber optics, tubing,
electrodes, etc...) or, the larger
instruments can be carried down on the
tether (e.g., gamma logs, neutron logs,
resistance logs or a video camera (using
a clear membrane). The interior of the
membrane is isolated from the exterior,
except where ports and tubing allow
access to the geologic medium.
Since the membrane supports the hole
wall, a casing and backfill is not
required in many holes. Therefore, one
has access to the entire hole wall for
collection of water or gas samples or for
in situ measurements while the membrane
is supporting the hole wall and sealing
it against flow. The membrane insertion
into a drillhole can proceed as quickly
as 20 ft/min or faster. Since the
insertion supports the hole and
simultaneously carries instruments into
place, one can actually case and
instrument a 50 ft hole in under five
minutes.
For long term installations, the interior
of the membrane can be filled with water
or sand (even "dirty sand", since it
doesn't contact the medium to be
measured). Later, the sand or water can
be flushed or blown out of the hole, and
the membrane and instrumentation can be
removed or replaced.
The obvious utility for field screening
purposes is that the SEAMIST is fast,
relatively cheap, and removable. What is
also an advantage is that the membrane
nestles around each instrument or
sampling port forming a membrane blister
on the hole wall. The interior of that
blister, and the associated instrument or
port, is isolated from other such
blisters at other elevations in the hole.
In fact, one side of the hole is isolated
from the other side. In principle, a
reactive covering on the membrane can be
emplaced and pressed against the entire
hole wall to provide a two-dimensional
map (azimuth and elevation) of
contaminants in the wall material.
The instrument array shown in Figure 2
was designed for monitoring of a steam
flood experiment yet in the planning
stage. It is an example of instruments
that can be emplaced by SEAMIST. The
results are yet to come. The concept is
young and in need of field testing.
Current research of this concept is
funded by DOE (Argonne National
Laboratory) for vadose measurements and
by DOE (Sandia National Laboratory) for
geothermal drilling applications. Since
the SEAMIST system functions equally well
horizontally and in constricted and
crooked holes, that is probably its best
application yet to be developed and
tested.
694
-------
Electrical leads
*
\ Geologic
) medium
10'
2\
'•'.••'.••'.'•S«v
•
•^
.
&
V
Membrane
StainlesssteeK
screen section
-Electrical contact
(typical)
(•<—Gas sampling port
and thermocouple
(typical)
Gas sampling ports
Tubing to
surface
Port through
membrane
^-Glass screen
spacer
Figure 2. Membrane lined monitoring hole design for steam flow
experiment.
695
-------
PORTABLE GAS CHROMATOGRAPH FIELD MONITORING OF PCB LEVELS
IN SOIL AT THE ELZA GATE PROPERTY
Marty R. Keller and Gomes Ganapathi, Ph.D.
Bechtel National, Inc.
Oak Ridge, Tennessee
ABSTRACT
Bechtel National, Inc. (BNI) conducted radiological and
chemical surveys of the Elza Gate property in Oak Ridge,
Tennessee, as part of the U.S. Department of Energy
(DOE) Formerly Utilized Sites Remedial Action Program
(FUSRAP).
Based on site history and preliminary characterizations at
the site, it was determined that polychlorinated biphenyls
(PCBs) were present across the site. Because PCB analysis
with the use of a portable gas chromatograph (GC) is
relatively fast and inexpensive, soil sample analysis results
could be made available to help direct the field sampling
program.
This paper provides a discussion of the manner in which
PCBs were monitored in the field during ongoing sampling,
the cost of these analyses, and a comparison of portable GC
screening results with Contract Laboratory Program (CLP)
laboratory results (1).
SITE DESCRIPTION AND HISTORY
The 8.1-ha (20-acre) Elza Gate property is located in the
eastern portion of the city of Oak Ridge, Tennessee, now
known as Melton Lake Industrial Park. Access to the site is
off Melton Lake Drive, near its intersection with the Oak
Ridge Turnpike (Figure 1).
In the early 1940s, the site was developed by the Manhattan
Engineer District (MED) as a storage area for pitchblende
(a high-grade uranium ore from Africa) and ore processing
residues. Five warehouses were constructed on the site,
three of which were used to store radioactive materials.
The Atomic Energy Commission (AEC) used the site until
the early 1970s, when it was vacated. After a radiological
survey and appropriate decontamination activities were
conducted in 1972, the site was deemed acceptable for use
with no radiological restrictions (2). At that time, title to
the property was transferred to the General Services Ad-
ministration and then to the City of Oak Ridge. The prop-
erty was subsequently sold to Jet Air, Inc., and used for the
operation of a fabrication and metal plating facility.
In 1987, at the request of the Tennessee Department of
Health and Environment, Oak Ridge Associated Universi-
ties (ORAU) conducted a survey at the site because of the
possibility of contamination from the metal plating facility.
This survey confirmed the presence of heavy metals and
PCBs at the site.
In October 1988, a preliminary radiological survey of the
site was conducted by Oak Ridge National Laboratory
(ORNL) for DOE. The survey indicated that residual
radioactivity exceeded the criteria for declaring a site
eligible for remediation under FUSRAP. As a result, on
November 30,1988, the entire Melton Lake Industrial Park
was designated a FUSRAP site (3).
In 1988, ownership of the property was transferred to
MECO, a development company. The site is presently
under further development for use as an industrial park. In
addition to the five MED warehouses previously men-
tioned, smaller structures also may have been on site. None
of the original structures remain, but the concrete pads on
which the warehouses were built are still in place.
One building currently on the property was erected on an
existent concrete pad. A second pad adjacent to this build-
ing is used as a vehicle parking lot and material storage
697
-------
J 2
FEET (Thousands)
SCALE APPROXIMATE
Figure 1 Location of the Elza Gate Site
-------
pad. The site has undergone considerable modification
since 1987, and the building is currently occupied by a
manufacturer of storage containers. Modification of the
property is expected to continue as the parcels are sold or
leased.
SAMPLING LOCATIONS
Because PCBs were previously detected at low concentra-
tions over the site area, all samples collected during the
chemical characterization effort were analyzed for PCBs.
Both systematic and biased locations were sampled. Sys-
tematic samples were collected from the comers and center
of each 61-m (200-ft) grid block. Using the data from
previous characterizations and information from the prop-
erty history, biased sampling locations were selected. A
hand held auger was used to collect three samples from
each location for analysis. The samples were collected at
0.3-m (1-ft) intervals to a depth of 1 m (3 ft).
CHARACTERIZATION METHODOLOGY
Since conventional characterization using CLP laboratory
protocols is costly and turnaround time required for CLP
analyses is approximately 30 days, on-site screening of
PCBs in soil samples using a portable GC was considered
useful in making real-time decisions on the rationale for
additional sampling locations during the ongoing chemical
characterization.
A Hewlett-Packard 5890 portable GC equipped with a
capillary column and an electron capture detector for
monitoring PCB levels in soil was used on FUSRAP during
the Elza Gate site characterization. The ability to detect
PCBs on site while sampling is taking place is one of the
key advantages of this field screening method (4) which
was refined by Twomey, Turner, and Murray (5). The need
for additional samples can be evaluated using this strategy
while the sampling crew is still in the field. Another
advantage is that this method permits comparison between
the reproducibility of field data and that of CLP data
because similar equipment and techniques are used.
The extraction procedure used for the Modified Spittler
Method consists of placing 2 g of soil in a test tube and
adding 0.5 ml water, 2.0 ml methanol, and
2.5 ml hexane (6). The sample is then vigorously shaken,
and aqueous and organic phases are allowed to form layers.
The hexane layer containing PCBs is withdrawn from the
top of the mixture and injected into the GC.
While this extraction method is less efficient than the CLP
prescribed procedure, it is very rapid and cost-effective.
Using commonly available laboratory equipment, one
analyst can easily extract 20 samples in less than 2 hr.
The cost of analysis using this screening method (including
sample preparation, analysis, and data evaluation) is be-
tween $50 and $100, compared to $300 for the equivalent
CLP analysis. The savings in cost, coupled with the time
savings (25 min for the field screen vs. 30 days for the CLP
analysis), warrant the consideration of this screening
method to complement CLP analyses.
The results BNI obtained using this method correlate well
with CLP laboratory results from the same samples
(Figure 2). The field screening results, while generally
lower than values obtained by the CLP laboratory, give an
excellent indication of locations where PCB concentrations
are elevated and where additional samples should be
collected for laboratory analyses. A comparison of Modi-
fied Spittler and CLP predictions of PCBs in soils is shown
in Figure 3.
Major reasons for variability in results include the
following:
• Since percent moisture was not determined for the screen-
ing samples, these results were not calculated on a dry
weight basis.
• Even with the best efforts to homogenize the sample,
concentrations of PCBs vary within the same sample.
• The extraction technique used with the screening method
is less efficient than the CLP procedure in extracting
PCBs from the soil matrix.
REGRESSION FIT
0.10 0.30 050 0.70 0.80 1.10 1.30 1.50 1.70 1.80 2.10
SCREENING RESULTS (PPM)
Figure 2 Correlation between Screening Results and CLP
Laboratory Data for PCBs at the Elza Gate Property
699
-------
10000
1000
100
10000
1 10 100 1000
Modified Spinier Value (ppm)
Figure 3 Comparison of Modified Spittler and CLP
Predictions of PCBs in Soils and Sediments
(Adapted From Fowler and Bennett 1987)
CONCLUSIONS
The Modified Spittler Method (5), originally developed by
ABB Environmental Services chemists, has been refined to
determine PCB concentrations in soil that represent excel-
lent comparisons with results generated by CLP proce-
dures.
The Modified Spittler Method has proved to be a fast,
accurate, and cost-effective procedure for determining PCB
concentrations in soil at the Elza Gate FUSRAP site. It
permits collection and analysis of a larger number of
samples during a field characterization and provides direc-
tion during the sampling effort, indicating to field person-
nel where additional soil samples should be collected for
analysis. The result is a more thorough characterization
requiring fewer field sampling efforts.
REFERENCES
1. U.S. Environmental Protection Agency, 1985. "State-
ment of Work for Organic Analysis, Exhibit D:
Analytical Methods," USEPA Contract Laboratory
Program, 7/85 Revision, K1-D134.
2. U.S. Department of Energy, 1972. Letter from
S. R. Sapirie to J. L. Smith, "Disposal of Excess Real
Property-Parcel 1228" [Attachment: Authorization for
Remedial Action at the Melton Lake Industrial Park
(Former Elza Gate Warehouses)], BNI CCN 057470,
Oak Ridge, Tenn. (February 3).
3. U.S. Department of Energy, 1988. Letter from
J. J. Fiore (DOE-HQ) to P. J. Gross (DOE-ORO),
"Authorization for Remedial Action at the Melton Lake
Industrial Park (Former Elza Gate Area Warehouses),
Oak Ridge, Tenn." [Attachment: Summary for Desig-
nation of Melton Lake Industrial Park Site], BNI CCN
057470, Oak Ridge, Tenn. (November 30).
4. Spittler, T. M., 1984. "Field Measurement of Polychlo-
rinated Biphenyls in Soil and Sediment Using a Portable
Gas Chromatograph," Environmental Sampling for
Hazardous Wastes, American Chemical Society,
pp. 37-42.
5. Twomey, D. M., Turner S. A., and Murray W. A., 1990.
"The Modified Spittler Method for Fast, Accurate and
Low Cost Determination of PCB Concentrations in
Soils and Sediments," Proceedings of the Second
International Conference for the Remediation of PCB
Contamination, April 2 and 3, 1990, Houston, Texas,
pp. 83-89.
6. Fowler, B. A. and Bennett, J. T., 1987. "Screening for
Characterization of PCB-Containing Soils and Sedi-
ments," Proceedings of the National Conference on
Hazardous Wastes and Hazardous Materials,
March 16-18, 1987, Washington, D.C., pp. 204-207.
700
-------
Real Time Monitoring of the Flue of a Chemical Demilitarization Incinerator
S.N. Ketkar
Exlrel Corporation
575 Epsilon Drive
Pittsburgh, PA 15238
S.M. PENN
Extrel Corporation
575 Epsilon Drive
Pittsburgh, PA 15238
Introduction
Public Law 99-145 directs the Secretary of Defense to destroy
the nations stockpile of lethal unitary chemical warfare agents
and munitions stored throughout the continental United States
by September 30, 1994 [1]. The U.S.Army has selected
incineration as the best available technology for destroying
chemical warfare agents [2]. The National Research Council
has endorsed incineration as the method of choice for
chemical agent destruction. Maximum protection of the
environment, the general public and the personnel involved in
this destruction is required. For this reason very stringent
requirements are imposed for the maximum allowable
concentration of the chemical warfare agents in the effluent of
these incinerators. The allowable stack concentrations (ASC)
for the two nerve agents, GB and VX, are 0.0003 mg/m^, while
the ASC for the blister agent HD is 0.003 mg/m3. Moreover
analytical instrumentation is needed that can detect these
levels in real or quasi real time.
Normal analytical techniques employed to detect such low
concentrations use preconcentration and separation
techniques and are very time consuming [3]. The extreme
sensitivity of atmospheric pressure ionization makes it a
suitable technique to detect low concentrations of
contaminants in air [4]. Moreover, the ability of atmospheric
pressure ionizers to handle very large sample flow rates makes
it possible to use this technique for real time detection. The
specificity achieved by tandem mass spectrometry makes a
system based on atmospheric pressure ionization tandem
mass spectrometer (API-MS/MS) very attractive for monitoring
low concentrations of pollutants in complex matrices like stack
effluents. We report here the use of a commercially available
API-MS/MS system to monitor for chemical warfare agents GB
and VX, at concentrations near the ASC levels, in the flue of a
demilitarization incinerators.
The API-MS/MS system was tested on the flue of a
demilitarization incinerator at the Chemical Agent Munitions
Disposal System (CAMDS) at Tooele Army Depot in Tooele,
Utah. The incinerator was the liquid incinerator (LIC) which is
used to burn the liquid chemical warfare agents themselves.
Experimental
The system used was a commercially available EXTREL
Automatic Stack Sampling Mass Spectrometer (ASSMS). This
system uses an atmospheric pressure ionization source
coupled to a triple quadrupole mass spectrometer. This
system is described elsewhere in detail [5], so only a brief
overview will be presented here. A corona discharge operating
at atmospheric pressure, with a discharge current of about 5
uA, is used as a source of primary ions. A low pressure region,
operating at a pressure of about 1 torr is used to break up the
weakly bound water clusters which are always present in a
discharge operating at high pressures. The declustered ions
are injected into the entrance of a triple quadrupole mass
spectrometer. The mass spectrometer has three quadrupoles
each with 3/4" round and 6" long cylindrical rods. The middle
quadrupole is housed in a collision cell having end plates made
from a leaky dielectric material to improve transmission [6]. A
counting channel electron multiplier together with a
scalar/counter and a threshold discriminator serves as the
detection system. The triple quadrupole mass spectrometer
was used in a multiple reaction monitoring (MRM) mode to
monitor for the chemical agents.
A heat traced teflon transfer line was used to connect the inlet
ol the ASSMS system to the flue of the liquid incinerator. A felt
pad impregnated with silver fluoride was placed inside the stack
end of the transfer line to convert chemical agent VX to it's G-
analog. This was necessary because vapors of VX can not be
quantitatively transferred through a transfer line. The G-analog
of VX is structurally similar to GB and can be quantitatively
transferred through the transfer line. For detecting blister
agent HD, benzene charge exchange was used in the
atmospheric pressure ionization source to produce the
molecular ion of HD[7]. A mechanical pump is used to move
the stack effluent, at rates of up to 5 L/m, through the
ionization source. A syringe pump is used to introduce
solutions of the chemical agents in the transfer line, for
calibration purposes.
This system was used to monitor the two nerve agents GB and
VX and the blister agent HD. For the case of chemical agent
GB the transitions m/z=141 -* m/z=99 and m/z=141 -* m/z=81
were monitored. For the case of VX (in reality G-analog) the
transitions m/z=127 -> m/z=99 and m/z=127 -> m/z=81 were
701
-------
daughter ion spectra. For added specificity the transition to the
minor daughter ion has to be monitored. The minor daughter
ion is at m/z = 99 for the nerve agents GB and VX while for
blister agent HD it is at m/z = 63. For the case of the two nerve
agents the minor daughter ion is less than 10% as intense as
the primary daughter ion. Consequently this system has a
higher detection limit when monitoring this minor daughter ion.
For the two nerve agents the detection limit of the system for
the minor daughter ion is about 20 ASC. For the case of the
blister agent the minor transition is only slightly weaker than the
primary transition and consequently the detection limit for this
transition is 1.43 ASC.
This system was also tested on the exhaust ot a filter stack.
The filter stack provides a very clean matrix compared to the
matrix provided by the flue of the liquid incinerator. The results
of the tests performed on the exhaust of the filter stack are
summarized in Table 2.
Table 2
Results of Statistical Analysis ( Filter Stack)
Chemical Agent
GB
VX
HD
Decision Limit
0.08 ASC
0.12 ASC
0.05 ASC
Limit of Detection
0.15 ASC
0.23 ASC
0.13 ASC
Conclusions.
It has been demonstrated that a system based on atmospheric
pressure ionization tandem quadrupole mass spectrometry can
detect, in the flue of a chemical demilitarization incinerator,
nerve agents GB and VX and blister agent HD near the
allowable stack concentrations. In the absence of any matrix
effects, this system can delect these agents at concentrations
below 0.25 ASC. This has been demonstrated for the case of
the filter stack exhaust.
Credits
This work was supported by The Program Manager for
Chemical Demilitarization, Aberdeen Proving Grounds,
Edgewood, MD. under contract no. DAAA15-86-C-0107. We
thank Lanny Davis of Chemical Agent and Munitions Disposal
System, Tooele Army Depot, Tooele, UT for his assistance
during the course of this work.
monitored. For the case of HD the transitions m/z = 158 —> 109
and m/z = 158 —> 63 were monitored.
Results and Discussions
Calibration runs were performed on GB, VX and HD on four
consecutive days, to obtain the detection limit of the system for
detecting these agents in the flue of the liquid incinerator. On
each day the system was challenged with six different
concentrations of the chemical agents. We used six challenge
concentrations in the range of .5 ASC to 20 ASC. The ASSMS
system response at each challenge concentration was
measured in triplicate. The above procedure was repeated on
four days. The ASSMS system response thus obtained was
converted to a found concentration. Regression analysis was
performed on the resulting data to obtain statistical parameters
pertinent to describing the performance of the system.
We followed the procedure used by U.S. Army Toxic and
Hazardous Materials Agency (USATHAMA) to determine
certified reporting limits[8]. This procedure consists of
performing a weighted linear regression of the fount v/s target
concentration. Both upper and lower confidence limits, at any
desired confidence level, can then be obtained. In this work
we used a confidence level of 95%. Based on this analysis
statistical parameters like the limit of detection (LOD) and
decision limit (DL) can be calculated. LOD is the smallest true
concentration that will be consistently detected. If the analyte
is present in the sample stream at the LOD concentration level,
the probability that it will be detected is at least 95%. True
concentrations above the LOD are deemed detectable. DL is
the maximum found concentration that will result, with a
probability of 95%, from a stream containing no analyte.
However, since the DL is usually less than the LOD this will not
constitute a false positive. These two statistical parameters
contain all the information needed to assess a systems
detection performance. The results of this statistical analysis is
given in Table 1.
Table 1
Results of Statistical Analysis ( Liquid Incinerator Flue)
Chemical Agent
GB
VX
HD
Decision Limit
0.15 ASC
0.84 ASC
0.6 ASC
Limit of Detection
0.6 ASC
1.79 ASC
1.15 ASC
The results in Table 1 refer to the primary transition in the
702
-------
References
1. Public Law 99-145, " Department of Defense
Authorization Act 1986", Nov. 8, 1985, Title XIV, Part B
- Chemical Weapons, Section 1412 - Destruction of
Existing Stockpile of Lethal Chemical Agents and
Munitions.
2. Chemical Stockpile Disposal Program Report SAPEO-
CDE-IS-87005, Sep. 1987.
3. S.J. Smith, " Detection methods for highly toxic
organophosphates",7a/an/a, 1983, 10,725.
4. S.N. Ketkar, S.M. Penn, W.L. File, " Real time
detection of part per trillion levels of chemical warfare
agents in ambient air using atmospheric pressure
ionization tandem quadrupole mass spectromelry ",
Anal. Chem., To be Published, Feb. 1991.
5. S.N.Ketkar, J.G. Dulak, W.L. File, J.D. Buchner ,
Seksan Dheandhanoo, " Atmospheric pressure
ionization tandem mass spectrometer system for real
time detection of low level pollutants in air ", Anal.
Chem., 1989, 61, 260.
6. S.N. Ketkar and W.L. File," Transmission
characteristics of a triple quadrupole mass
spectrometer with leaky dielectric endplates", Rev.
Sci. Instrum.. , 1988, 59, 387.
7 S.N. Ketkar, J.G. Dulak, S. Dheandhanoo and W.L.
File," Benzene charge exchange at atmospheric
pressures for low level detection of pollutants in air",
Anal. Chimica Acta, To be published.
8. Department of the Army, U.S. Army Toxic and
Hazardous Materials Agency, "Sampling and Chemical
Analysis Quality Assurance Program", April 1982.
703
-------
FIELD EVALUATION OF THE BRUKER MOBILE MASS SPECTROMETER
UNDER THE U.S. EPA SITE PROGRAM
S.M. Klainer, M.E. Silverstein,
VJL Ecker, and D.J. Chaloud
Lockheed Engineering A Sciences Company
Las Vegas, Nevada
S. Billets
U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory
Las Vegas, Nevada
INTRODUCTION
The Mobile Environmental Monitor (MEM), afield-deployable
gas chromatograph/mass spectrometer (GO/MS)
manufactured by Bruker Instruments, Inc., was demonstrated
to assess the ability of this technology to perform in-field
analyses of organic contaminants in soil and in water. The
demonstration, conducted under the U.S. Environmental
Protection Agency (EPA) Superfund Innovative Technology
Evaluation (SITE) Program, took place at two Superfund
sites in Massachusetts during August and September 1990.
Detailed studies and quality assurance designs provided the
structure for each field demonstration. Real-world and
performance-evaluation samples were analyzed by the MEM
and by equivalent, standard EPA methodologies. Data
generated by the MEM were compared to that obtained from
the EPA methods and were used to assess specific
performance characteristics.
BACKGROUND
The Superfund Program was established by Congress in
1980 to identify, prioritize, and remediate the nation's
uncontrolled hazardous waste sites. Because the problems
associated with hazardous waste sites have proved to be far
more complex and diffuse than anticipated, Congress
enacted the Superfund Amendments and Reauthorization
Act of 1986 (SARA). Under SARA, EPA was charged with
effecting more timely and cost-effective remedies for
Superfund site remediations. The SITE Program satisfied
the requirement (SARA, Section 311[b]) that EPA establish
a program designed to accelerate the development,
demonstration, acceptance, and use of promising alternative
or innovative technologies targeted to meet the objectives of
the overall Superfund Program.
Two categories of technologies are recognized in the SITE
Program: (1) treatment technologies that may serve as
alternatives to land disposal of hazardous waste, and
(2) monitoring and measurement technologies for identifying
contaminants. Monitoring and measurement technologies that
are accepted into the SITE Program are evaluated as part of
the Monitoring and Measurement Technologies Program
(MMTP). Under the SITE Program, the MMTP is administered
by the EPA Office of Modeling, Monitoring Systems and Quality
Assurance (OMMSQA) through the Environmental Monitoring
Systems Laboratory in Las Vegas, Nevada (EMSL-LV). The
Bruker MEM was demonstrated under the MMTP.
The primary purpose of the MMTP is to provide developers with
the means to demonstrate innovative technologies that could
be used as alternatives to the current systems of detecting and
assessing the extent of pollutants at hazardous waste sites.
The focus of these demonstrations is to evaluate fully-
developed technologies, thereby making performance and cost
effectiveness data available to interested parties. Superfund
decision makers will thus have the information that is necessary
to consider whether or not these technologies are of potential
use in future site characterization or remediation projects. The
developers of the monitoring and measurement technologies
are identified from as many sources as possible, including
solicitations in relevant trade journals, periodicals, seminars,
and professional conferences. Once the developers reply to a
solicitation, the SITE Program representatives begin an
evaluation process to determine the feasibility, utility, and need
for each technology.
Bruker Instruments, Inc., of Billerica, Massachusetts, responded
to one of these solicitations and its MEM was identified by EPA
as a promising candidate for a field demonstration under the
MMTP. The MEM, designed for the on-site analysis of organic
contaminants, is a mobile mass spectrometer (MMS), optionally
coupled to a gas chromatograph (GC) or a thermal desorption
705
-------
sampling probe. Currently, full-size (therefore, nonmobile)
laboratory gas chromatography/mass spectrometry (GC/MS)
has been the preferred EPA approach to identifying and
quantifying organic contaminants at Superfund sites. This
technology analyzes compounds on the basis of the
molecular weight, retention time, and characteristic
fragmentation patterns of their chemical components. The
primary disadvantages of conventional GC/MS systems are
instrument size, power demand, and sensitivity to external
factors (e.g., temperature, humidity, and vibration). The
development of an MMS rugged enough to withstand a
variety of field conditions is of considerable interest to
parties responsible for contaminant monitoring and for
remediation of Superfund sites. Newly developed mobile
systems, such as the Bruker MEM, appear to have attained
satisfactory levels of stability, power usage, and
compactness for field applications.
MEM SITE DEMONSTRATION
The purpose of the demonstration was to evaluate the ability
of the MEM to analyze polychlorinated biphenyls (PCBs) and
polynuclear aromatic hydrocarbons (PAHs) in soils and to
analyze volatile organic compounds (VOCs) in water, under
field conditions at Superfund sites. The demonstration
focused on the capability of the instrument to generate
rapid, cost-effective, and reliable PCB, PAH, and VOC data
from real-world samples. The demonstration was used to
compare MEM performance to similar analytical method
performance as required under the EPA Contract Laboratory
Program (CLP) or the Resource Conservation Recovery Act
(RCRA). Detailed project and quality assurance plans were
prepared which defined the sampling and analysis protocols,
the experimental design, the quality assurance and quality
control (QA/QC) requirements, the data base management
system, the health and safety considerations, and proposed
data analysis and evaluation methods.
For this demonstration, real-world samples were collected
from two National Priorities List (NPL) sites in Massachusetts
(EPA Region 1). These sites were selected on the basis of
documented (Record of Decision) presence of analytes of
interest: e.g., PCBs in soil and VOCs in ground water at one
site and PAHs in soil at the other. A screening analysis by
the MEM identified the collection points (i.e., low, medium,
and high concentration levels) for five samples in each
compound class (PCB, PAH, VOC). Bulk samples were
collected, homogenized, and split (bottled) into replicates.
For each compound class, each of the five distinct samples
was split into seven replicates for analysis on site (or near
site) by the MEM, and off site by standard EPA methods. In
addition, standard reference materials (SRMs), and blank
samples were sent to all analysis locations for variability,
detection, and other data quality assessments. This process
worked well for the PCB and PAH soil samples; however,
remediation activities at the chosen site precluded the
collection of VOC-contaminated ground-water samples.
Instead, surface water collected from the other site was spiked
with different concentrations of VOCs.
ANALYSIS METHODS
The MEM analytical methodologies for field analysis were
developed by the Trace Analytical Chemistry Laboratory of
Tufts University. The off-site confirmatory laboratories used
EPA-approved methods for analyzing demonstration samples.
This process minimized intermethod biases because the EPA
methods were chosen for their similarity to the MEM field
methods. A brief overview of each method is provided below.
PAHs in soils: For the MEM, soils were first extracted with
methy lene chloride. The extracts were then thermally desorbed
onto a 3-m chromatography column interfaced with the mass
spectrometer. The data were collected and interpreted in a
manner similar to that used for EPA methods. For the off-site
laboratory, samples were first extracted by RCRA Method 3550
(methylene chloride, sonicatton extraction). The extracts were
then analyzed by RCRA Method 8270 (GC/MS analysis for
semivolatile organic compounds).
VOCs in water: For the MEM, analytes were purged from the
samples onto Tenex tubes. The tubes were thermally
desorbed onto a 30-m fused silica capillary column for
compound separation. Compound identification and
quantification were performed using quadrapole MS. The off-
site laboratory employed RCRA Method 8260 (capillary column
GC/MS for volatile organic compounds).
PCBs in soil: For the MEM, soils were first extracted with
hexane. The extracts were then thermally desorbed onto a 3-m
chromatography column connected to the mass spectrometer;
the final concentrations were calculated by quantitating
individual chlorination levels (mono- through
octachlorobiphenyl, only). For the off-site laboratory, samples
were first extracted according to RCRA Method 3550 followed
by GC/MS analysis in concordance with the CLP high-
concentration protocol. Like the MEM method, congener
counting was used in sample quantification. RCRA Method
3640 (gel permeation chromatography cleanup) was used
when necessary for high-concentration samples. (NOTE: The
conventional GC method for PCB analysis was not used; this
method measures arochlors and, thus, would not have
provided proper intermethod comparison.)
MEM EVALUATION
Data Analysis: The data from all analysis sites were compiled
into one fully documented data base. Data were then
subjected to a detailed verification process. Following
706
-------
verification, a variety of data analyses were performed,
including intermethod comparisons (between the MEM and
the off-site laboratory results), reproducibitty estimates (from
replicate analyses on the same instrument), and the
evaluation of data quality indicators. Direct comparison plots
and a variety of statistical routines were used to interpret the
data. Figure 1 and Table 1 represent selected demonstration
results. Figure 1 presents the comparisons of real-world and
SRM samples forthe PCS trichlorbiphenyl congener. Table 1
presents precision, accuracy, and bias information forPAHs
based on SRM analyses by the MEM and by the off-site
laboratories.
Figure 1. Linear regression of trichlorobiphenyl.
250H
200-
'150-
< 100-
50-
Plot Symbol*
A- Real-World Samplt Q- SRM
50
100
150
200
250
300
OFF-SITE ANALYSIS (ppm)
Table 1. Precision and accuracy data for PAHs based on SRM analysis.
-250
-200
-150
-100
-50
-0
350
Analyte
(Theoretical SRM Value)
Naphthalene
(50.7)
Acenaphthylene
(46.4)
Fluorene
(45.9)
Pyrene
(50.6)
Chrysene +
Benzo(a)anthracene
(97.9)
Lab
MEM
Off-Site
EMSL-LV
MEM
Off-Site
EMSL-LV
MEM
Off-Site
EMSL-LV
MEM
Off-Site
EMSL-LV
MEM
Off-Site
EMSL-LV
Mean*
28.6
32.8
29.7
10.9
20.4
24.6
29.2
31.1
28.0
40.7
34.6
32.4
86.8
66.1
63.3
%RSO
55.1
11.3
1B.S
39.1
11.9
20.1
25.1
8.0
19.6
16.2
9.6
23.4
26.5
9.0
23.5
%Bias from
Theoretical Value
-44.8
-34.9
-40.8
-76.3
-56.9
-46.1
-36.8
-32.5
-39.0
-19.0
-30.8
-36.8
-11.1
•32.6
-34.6
%Bias from
Off-Site Value
•15.2
-
•9.1
•45.0
-
+25.0
•6.5
-
•9.7
+ 17.1
-
-8.6
+31.8
-
-3.0
%Bias from
EMSL-LV Value
-6.7
+ 10.0
-
•56.0
•20.0
-
+3.6
+ 10.7
-
+28.1
+9.4
-
+35.9
+3.1
—
•Means bated on analysis ol: 40 replicates lor MEM; 30 replicates lor oil-site; 7 replicate* tor EMSL-LV. Units are in ppm.
%RSO » percent relative standard deviation
SRM - Standard Reference Material
EMSL-LV - Environmental Monitoring Systems Laboratory, Las Vegas. Nevada
MEM • Mobile Environmental Monitor
707
-------
Instrument Characteristics: The primary advantages of the
MEM are its portability and ruggedness. Rechargeable
batteries supply all power required by the MEM, and
logistical requirements are minimal and easily fulfilled. The
use of purified ambient air as the carrier gas eliminates the
need to transport compressed gas cylinders. The
demonstration plan called for the analysis of 13 samples per
day; the analysis team had difficulty meeting this sample
throughput requirement. Although the MEM is easy to
operate under normal conditions, a skilled operator is
required to correctly diagnose and repair malfunctions.
NOTICE
Although the research described herein has been funded
wholly or in part by the U.S. Environmental Protection Agency
under Contract Nos. 68-03-3249 and 68-CO-0049 to Lockheed
Engineering & Sciences Company, it has not been subject to
Agency review and therefore does not necessarily reflect the
views of the Agency, and no official endorsement should be
inferred. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.
FUTURE MOBILE GC/MS WORK IN THE SITE PROGRAM
On the basis of data collected and observations made
during this demonstration, several issues have been
identified that must be addressed before the MEM or other
MMS instruments can be employed in Superfund site
monitoring, characterization, and remediation activities.
These issues include: (1) method development and
procedural documentation, (2) development of standardized
QA/QC requirements and limits, (3) development of data
reporting standards for field analytical measurements, and
(4) development of a detailed troubleshooting guides and
training programs. These issues do not necessarily
represent problems with the technology itself; several are
external factors or policy issues that require attention before
the EPA can use MMS instruments as reliable field analytical
devices.
EPA is considering additional mobile mass spectrometer
evaluations at EMSL-L.V under the SITE Program. Future
laboratory evaluations will concentrate on (1) separating
variability associated with the instrument from variability
associated with the method, (2) formalizing QA/QC
procedures, and (3) establishing consistent data reporting
procedures for field applications. Additional field
demonstrations and evaluations will follow.
ACKNOWLEDGMENTS
The EPA would like to thank the following organizations for
their participation in the demonstration and evaluation of the
Bruker MEM under the SITE Program:
EPA Region 1
Bruker Instruments, Inc.
Tufts University
Lockheed Engineering & Sciences Company
S-CUBED
708
-------
THE DITAM ASSAY
A FAST, FIELDABLE METHOD TO DETECT
HAZARDOUS WASTES, TOXIC CHEMICALS, AND DRUGS
Dr. Cynthia A. Ladouceur
U.S. Army Chemical Research, Development and Engineering Center
Aberdeen Proving Ground, MD 21010-5423
PURPOSE AND SCOPE OF ASSAY
The DITAM (Diffusion Through A Membrane) assay
is fieldable, fast, extremely easy to use,
inexpensive, and can be used to detect one or
several substances simultaneously. The DITAM
apparatus was developed along with the DITAM
assay. This new assay is in the initial stages of
development for the detection of several small
molecular weight substances. Model systems have
been tested for the detection of progesterone and
Ricinus communis. a toxin. Based on the initial
experimental results, it appears that the DITAM
assay will be useful for the rapid detection of a
wide variety of substances in the field. Examples
of these substances include small molecular weight
hazardous wastes, toxic chemicals, and drugs.
Continuing research involves further modifications
of this assay to enable the detection of large
molecular weight substances such as proteins from
infectious organisms and antibodies directed
against these organisms.
MATERIALS AND METHODS
There are two version of the DITAM assay and
apparatus. The original DITAM apparatus consists
of a hand-held polystyrene "box" with two to four
chambers. The chambers are separated by
semipermeable membranes. Because of the
difficulty in producing these "boxes" in the
laboratory, the design of the DITAM apparatus was
modified. The modified version consists of a "bag"
in a 12x75 mm capped test tube or an alternate
pocket-sized vial. The "bag" is actually a
cylindrical semipermeable membrane which is filled
with reagents and tied or clamped at both ends.
These reagents consist of specific antibodies,
enzyme-labeled antigens, and a buffer solution. To
date, several membrane types have been tested for
their usefulness in this assay. The membranes
must be flexible and have pores which allow
molecules of a specific size to diffuse through
easily and rapidly. The appropriate molecular
weight cutoff of the membrane must be selected in
order to retain the antibody molecules within the
"bag" and enable the enzyme-labeled antigens and
antigens in the test sample to pass through the
"bag." All reagents and supplies needed for a
modified DITAM assay are illustrated in the figure
on the following page.
When performing a modified DITAM assay, an
individual is supplied with a test tube which
contains the reagent-filled "bag." To perform this
assay in the field, an individual only needs to add
the test sample to a fill line marked on the tube,
shake the tube for approximately one to two
minutes, add the enzyme substrate, shake the tube
again, and observe the tube for a color change. All
instructions can be printed on the tube.
The test sample can be liquid or solid. Solid test
samples, such as dust particles, can be
concentrated on cotton swabs and placed in test
tubes along with a buffer solution.
The color of the reaction solution depends on the
enzyme and the degraded enzyme substrate. When
horseradish peroxidase is used to label the antigens
and tetramethylbenzidine plus hydrogen peroxide is
used as the enzyme substrate, a turquoise colored
solution inicates a positive reaction and a clear
solution indicates a negative reaction.
709
-------
REAGENTS AND SUPPLIES NEEDED FOR
A MODIFIED DITAM IMMUNOASSAY
0
© 0
©
© ©
©
semi-permeable
membrane
test tube
specific
antibody
enzyme-labeled
antigen
enzyme
substrate
specific
antibody
bound to
enzyme-labeled
antigen
antigen from
test sample
• degraded
> • enzyme
• substrate
RESULTS AND CONCLUSIONS
Based on the initial experimental results, the
modified DITAM assay can be completed in three to
five minutes. Thus, it satisfies an initial
requirement of speed in obtaining assay results.
This assay is extremely easy to perform in both the
field and in the laboratory. Since all of the assay
instructions can be printed on the tube, little or no
training is required in order to perform this assay.
In order to achieve the maximum contrast between
the positive and negative reaction solution colors
(turquoise for the former and clear for the latter),
the appropriate concentration of reactants must be
employed. The use of an inappropriate
concentration of antibody molecules inside the
"bag" can result in false positive reactions (darker
blue coloring in negative test samples). Although
there may be pale blue coloring in the negative
samples due to background reactions, this can be
kept to a minimum if the concentrations of
reactants are carefully calibrated for the assay.
This procedure is standard when developing any
new immunoassay.
Due to the versatility of the assay, it can be
developed for use by many government agencies
and by the private sector as well. Possible future
applications include the following: (1) detection of
chemical warfare agents in the field and chemical
warfare treaty verification, (2) detection of drugs in
humans and animals, (3) protection of humans from
environmental contaminants (i.e., pesticides and
toxic chemicals in dust and water supplies), and (4)
detection of hormones (i.e., determining levels in
hospital patients and athletes). Continuing research
and further modifications of the DITAM assay and
apparatus should enable the detection of large
molecular weight substances such as biological
warfare agents and proteins from pathogenic
organisms (i.e., in food, humans, and animals).
710
-------
RAPID SCREENING OF GROUND WATER CONTAMINANTS
USING INNOVATIVE FIELD INSTRUMENTATION
Amos Linenberg
Sentex Sensing Technology,
Ridgefield, NJ 07657
Inc.
David Robinson
Sentex Sensing Technology, Inc.
Ridgefield, NJ 07657
With the increased use of on-site
monitoring equipment at hazardous
waste locations, an emphasis has been
placed on development of rapid
screening instruments designed
specifically to provide quick and
accurate ground water analysis.
Two instruments have been developed to
accomodate those needs. The first one,
the "TOP", provides quick assessments
of total concentrations of volatile
hydrocarbons in water. The second
instrument, the AQUASCAN, provides an
on-line analysis of individual volatile
hydrocarbons in water.
The "TOP"
The "TOP" (Total Organic Purgables) is
an instrument which monitors total
concentrations of volatile hydrocarbons
in water utilizing purge and trap
technology. Figure 1 is a block
diagram of the "TOP". The "TOP" is
designed around an argon ionization
detector (AID) and internal sample
purge and trap system. A computerized
control system activates an internal
sample pump, drawing water into the
purge cell. Argon gas is purged through
the purge cell, stripping the purgable
hydrocarbons from the water into the
purge gas stream. The gas containing
the purged hydrocarbons is routed
through an adsorption trap, where the
hydrocarbons are collected. Once
collected, the trap is heated, desorbing
the hydrocarbons into a blank capillary
tube and into the AID chamber for
quantification. The results are
OET.
&
CO
Y
-
••
TRAP
mmtrn
1
PURGE
CELL
20PPB
display
PUMP
— • mil
Figure 1. Block diagram of TOP instrument
automatically displayed on a LED screen
and stored on a memory chip. This data
can be transferred to a computer or a
printer for a permanent record.
Calibration of the system is carried out
by sampling a water sample containing
known concentration levels of hydro-
carbons. Figure 2 shows a typical trace
of a total purgable run. The dashed line
line is a calibration trace while the
solid line is the analysis trace. The
integrated area of the analysis is
compared to that of the calibration,
and a concentration value in the ppb
or ppm level is automatically calculated
and displayed on the LCD screen.
An AID is used because of its relative
uniform response to a broad range of
purgable hydrocarbons (Table 1.) This
includes halomethanes and haloethanes
which are not easily detected by other
total hydrocarbon detectors, such as
the photoionization detector (PID).
711
-------
X
Time (seconds)
Figure 2. TOP concentration trace
30
onizing Energy
etected cmpds.
Relative response
among detected
cmpds.
AID
11.6 eV
aliphatics
aromatics
halomethanes
haloethanes
1-10
times
PID
10.6 eV
aliphatics
aromatics
1-1000
times
Table 1. Comparison of AID and PID
The more uniform response of the AID
assures more accurate concentration
readings regardless of the compounds
chosen for calibration.
The AQUASCAN
The AQUASCAN is used for continuous
monitoring of a water source or stream
by purge and trap gas chromatography.
The instrument consists of an on-line
purge and trap sampling module attached
to a gas chromatograph (GC) (Figure 3).
The purge and trap module contains an
internal pump to draw calibration,
analysis, or clean water samples into
the purge cell. The sample is then
purged with inert gas with the
resultant vapor swept into an adsorption
tube. The trapped volatile organics
are then thermally desorbed and injected
into the GC column where they are
separated. Each compound is then
identified and quantified. The
resultant chromatograph (Figure 4) is
displayed on the computer screen and
stored on disk for later review.
The AQUASCAN is automated so that a
permanent operator is not needed. An
internal modem allows for remote
operation of the instrument.
Both of these instruments will aid in
the screening and analysis of contam-
inated water sources. The "TOP" can
rapidly and accurately determine total
hydrocarbon values in water. The total
PRINTS)
CIRCUIT
BOHR)
PIKE
OLi
PUHP
Figure 3. Block diagram of Aquascan; 1-drain;
2-ririse; 3-cal.; 4-anal.
analysis time from calibration to sample
acquisition to concentration display is
approximately two (2) minutes.
The AQUASCAN allows for complete, on-
line, automated, and accurate
chromatographic analysis of purgable
hydrocarbons in water. Each component
is accurately identified and quantified.
The AQUASCAN can serve as a continuous
monitoring system for traces of VOC's
in water, such as waste water streams
or water purification systems.
PUK
1
2
3
4
5
WE
two.
BEKZBE
ftt
H-m.Bf
WPIH
BT
%
SO
73
183
i300
1
cut.
875 p(t
1-3 H«
i 'H»
93? ^ti
1.8 pp.
S 3ZSP2IOO; «D. 95 'C
Figure 4. Aquascan chromatograph trace
712
-------
IMPROVED DETECTION OF VOLATILE ORGANIC COMPOUNDS IN AIR BY ON-LINE
SAMPLE CONCENTRATION IN A MICROCHIP GAS CHROMATOGRAPH
Aaron M. Mainga and Edward B. Overton, Institute for Environmental Studies, Louisiana
State University, Baton Rouge, LA 70803.
Abstract
Pre-concentration of dilute gas samples was
performed by adsorption on a 1.5" glass-lined
stainless steel Tenax trap, interfaced on-line, in a
Model 200 Portable Microchip Gas Analyzer
(Microsensor Technology Inc., Fremont,
California). Adsorption onto Tenax (2,6-diphenyl-
p-phenylene oxide polymer) was achieved by
passing 5 mL of dilute gas sample, at room
temperature, through the trap placed between the
sample loop (in a solid-state injector) and the
switch valve. The adsorbed analytes were
desorbed from the Tenax by rapid heating at
200°C followed by venting of the unconcentrated
analytes to the atmosphere before injecting a
concentrated plug of analytes into the analytical
column. Concentration ratios of up to 30:1 were
achieved for some analytes. Sample recovery
was affected by several factors, such as, the
amount of sample purged through the trap at
room temperature, amount of sample injected,
rate of desorption heating, final desorption and
cooling temperatures of the trap, sample volatility.
Sample recovery also varied according to
whether the trap contained Tenax alone or Tenax
with Spherocarb as the adsorbent.
Introduction
Throughout the history of modern
chromatography, there has been a consistent
trend to work with ever-decreasing amounts of
analysed materials and at increasing demands on
detection sensitivity. Only slowly has this direction
been translated into smaller chromatographic
columns and corresponding instrumentation (1).
Difficulties are frequently encountered in
attempting to directly analyze organic compounds
of interest, which are often below the part per
billion level. Despite the use of highly sensitive
instruments, detection of trace amounts of
substances in this range presents a technical
challenge, especially as regards to the use of
portable instruments (2). A good example is the
analysis of volatiles from human expired air to
seek distinctive differences between "normals"
and those afflicted by disease. More recently,
awareness has grown to the fact that minute
concetrations of chemical pollutants can have far
reaching effects as health hazards, further
underscoring the need for reliable analytical
techniques (3).
It is often stated that one of the possible
applications of high speed gas chromatography
(using a Microchip Gas Analyzer) is in the fields of
breath gas analysis (4), and on-site analysis of
hazardous volatile organic compounds (5). At the
part per billion level, it has almost always been
necessary to use some off-line cumulative or
concentrating technique to obtain measurable
amounts of sought-after compounds. Ideally, it is
preferable to elliminate as much as possible, the
unwanted background compounds (usually water
and air) while accumulating the desired
substances quantitatively. For most techniques,
the result is a compromise between these two
goals.
In off-line concentration applications with the
Microchip Gas Analyzer, sample components
trapped in a separate Tenax (a porous polymer of
2,6-diphenyl-p-phenylene oxide) concentrator are
manually collected in a gas tight syringe and
introduced to the Microchip Gas Analyzer for
separation (6). This off-line technique is generally
time consuming, operator-intensive, and difficult to
automate. In our work, we sought to put the Tenax
GC concentration trap on-line in a Microchip Gas
713
-------
Analyzer, by placing it between the silicon
injection wafer and the injection switch valve.
Experimental
GC System
This consisted of a model 200 Microchip Gas
Analyzer (Microsensor Technology Inc., Fremont,
CA ) which was equipped with a solid-state
sample injection system; two vacuum pumps
connected in parallel; a 4 m long x 0.10 mm i.d x
0.4 |im phase thickness DB-1701 capillary column;
a miniaturized thermal conductivity detector; and
an apple computer, using M2001 software.
Zero helium was used as carrier gas at a column
head pressure of 20 psi corresponding to a
flowrate of 1 mL/min (an average velocity of about
36 cm/sec.). An isothermal column temperature of
40°C was used in our experiments.
Sample Mixture
100 ppmv mixture of n-propane to n-octane
hydrocarbons in zero nitrogen.
1 ppmv mixture of acetone, benzene, toluene,
chlorobenzene, and bromobenzene in zero
nitrogen.
Traps
A 1 .5" long x ^" o.d.x 0.7 mm i.d. glass-lined
stainless steel tube containing 2.0 mg Tenax GC,
60/80 mesh, Applied Science Laboratories
lnc.,PA.
1.5"
1
long x yg" o.d.x 0.7 mm i.d. glass-lined
stainless steel tube containing 1.5 mg of Tenax
GC and 1.0 mg of Spherocarb 80/100 mesh,
Analabs, Norwalk, CT.
Trap Loading and Sample Injection
Five mL of sample mixture was adsorbed onto the
adsorbent in the trap by passing it through with the
aid of two vacuum pumps using a sampling time of
140 seconds. This was followed by rapid heating
of the trap to a temperature of 190°C for 90
seconds. With the trap at 190°C the switch valve
was turned to the carrier gas position followed by
venting of the the unconcentrated sample in the
sample loop to the atmosphere through the
sample valve and the injection valve opened for
200 msec., allowing about 0.2 jiL of the
concentrate to be injected into the analytical
column. Delay times of 400 msec, were applied
prior to venting and sample injection. After
injection, the trap was prepared for the next
sample by purging with carrier gas, while hot for
two minutes and allowed to cool to room
temperature with carrier gas purging continuing.
Results and Discussion
The main purpose of this work was to investigate
the possibility of placing an on-line Tenax trap in a
Microchip Gas Analyzer. For a glass-lined
stainless steel trap with its many advantages, it
has so far proved practically applicable (7).With a
Tenax GC or Tenax / Spherocarb trap on-line, the
two sample pumps were drawing the hydrocarbon
sample at the rate of about 2.5 mL/min.
For a trapped sample volume of 5 mL, an injection
volume of 0.2 uL (200 msec, injection time), and a
100 |xm i.d. analytical column, concentration
factors of 10 are quite common, depending on the
amount of venting done before the injection.
Concentration factors are dependent on the vent
times used.
For a trapped volume of 5 mL on a Tenax
adsorbent and an injection volume of 0.2 ^L (200
msec, injection time), the following were some of
the data obtained:
Cpd Name Area %RSD Increase
Propane Trace
Butane 411813 2 4
Pentane 783870 11 5
Hexane 3164633 8 17
Heptane 7152900 2 Large
Octane 5451000 4 Large
For a trapped volume of 5 mL (the rest same as
above):
Cpd Name Area %RSD Increase
Acetone 2205500 5 2
Benzene 13518667 4 15
Toluene 7751567 7 41
Chlorobenzene 1760633 14 Large
Bromobenzene 3075167 17 5
Note: An increase of large implies that the
compound was not detected before concentration
In general, the factors found to increase the
concentration factor include: the use of an
analytical column with an internal diameter larger
than 100 jam i.d.; use of a higher desorption
temperature (> 200°C); venting of an appropriate
amount of dilute sample in front of the adsorbent in
714
-------
amount of dilute sample in front of the adsorbent in
the sample line; use of the fastest rate of
desorption heating and use of an appropriate
adsorbent for the group of compounds under
investigation.The last factor which is still being
optimized is very crucial more so for the lighter
hydrocarbons which have a very low breakthrough
volume in a Tenax trap as compared to
Spherocarb.
Conclusions
Though still under investigation, our preliminary
results, so far, have indicated that placing a Tenax
or a Tenax / Spherocarb trap on-line in a
Microchip Gas Analyzer can enhance the
detectability of volatile organic compounds.
However, for better reproducibility and
quantitation, the stability of the trap heater still
needs to be improved and also a better
combination of adsorbents in the trap have to
optimized for better concentration of all the
analytes under investigation. According to Zlatkis
et al, (2) Tenax, with a surface area of 18 m2 g-1,
has a low retention for low molecular weight
compounds, especially water. Higher molecular
weight compounds with relatively low polarity can
be trapped and thermally desorbed (at 300° C) at
high efficiency. At this temperature Tenax does not
contribute to detectable artifacts, due to its unusual
thermal stability. Like any other chromatographic
stationary phase, the Tenax or Tenax /
Spherocarb trap must be evaluated as regards
partitioning of a particular compound between
adsorbent and carrier gas. As a consequence, our
resuts apply to the specific amount of adsorbents
employed in the trap. In this case, the beakthrough
volumes for compounds of interest, which are
proportional to the amount of adsorbent would
have to be determined for a particular trap before it
is put to use.
REFERENCES
1. Angell, J.B.; Terry, S.C.; Earth, P.W., Scientific
American, 1983, 44,248.
2. Zlatkis, A.; Weisner, S.; Ghaoui, L; Shanfield,
H., J. Chromatogr. Lib., 1985, 32, 449-459,
edited by Bruner, F.
3. Krotoszynski, B.; Gabriel, G.; O'neill, H., J.
Chromatogr. Sci.,1977,15, 239-244.
4. Saadat, S.; Terry, S., C.; American Lab.,1984, 5,
90-101.
5. Lee, G.; Ray, C.; Siemers, R., and Moore, R.,
Am. Lab., February 1989, 110-119.
6. Backhouse, T., "An Investigation of the
Performance of a Portable Concentration
Device for the Analysis of Air Samples, M.S.
Thesis, Institute for Environmental Studies,
Louisiana State University, 1989.
7. Kirk, B.D.; Zaffiro, A.D.; Westendorf, R.G., "Use of
Glass-lined Tubing in Purge and Trap
Concentrators",presented at the 39th Pittsburgh
Conference,February,1988, New Orleans, LA.
715
-------
ON-LINE SCREENING ANALYZERS FOR TRACE ORGANICS
UTILIZING A MEMBRANE EXTRACTION INTERFACE
Richard G. Melcher and Paul L. Morabito
THE DOW CHEMICAL COMPANY
ANALYTICAL SCIENCES
1602BLDG.
MIDLAND Ml, 48667
A unique membrane extraction interface has been
developed which enables automatic extraction of
selected trace organic compounds from aqueous
streams and samples. By selecting the type of
extractant flowing through the tubular silicone rubber
membrane, various classes of organic compounds
can be selectively extracted and concentrated, with
the exclusion of others. In addition to the advantage
of selectivity, the interface can be used for streams
with high dissolved solids and paniculate content.
Several types o.f on-line monitors have been
developed based on this interface which would be
suitable for screening and analysis of wastewater,
leachate, ground water and surface water.
The simplest membrane monitor is based on a flow
injection system where the analyte, which is extracted
from the injected sample by the membrane, flows
directly to the detector. This approach is useful in
screening for a major component or for the sum of a
class of compounds. For complex samples the
membrane/detector selectivity may not be sufficient to
isolate the single component and additional
separation is necessary. In one on-line application,
low ppb levels of chlorophenols are determined by
extraction through the membrane into a basic
extractant. This extract is then injected into a LC
system for analysis. In another application, a
capillary gas chromatograph is used to determine low
and sub-ppb levels of organic compounds extracted
through the membrane into a hexane extractant. An
automated large volume (25-250 n\.) injection
technique, which couples a retention gap with an air
actuated rotary valve, was developed to make on-
column injections.
Membrane. The membrane used in the systems
described was SILASTIC brand medical grade tubing
(Dow Corning, Midland, Michigan), a seamless
silicone rubber tubing designed for clinical and
laboratory applications. Silicone rubber is chemically
and mechanically stable and has a high permeation
rate for a large variety of organic compounds. A
single membrane has been used in a continuous
wastewater analyzer for over a year with no apparent
change. Depending on the membrane size,
pressures above 10-20 psi will cause the membrane
to expand and possibly rupture. Tubing of various
sizes can be obtained from medical supply houses.
The two membrane sizes used were 0.013 inches I.D.
by 0.025 inches O.D. (Dow Corning Catalog No. 602-
105), and 0.020 inches I.D. by 0.030 inches O.D.
(Dow Corning Catalog No. 602-135).
General principles of membrane extraction. The
permeation of compounds through a nonporous
polymer membrane occurs by a "solution-diffusion"
mechanism. The term "permeation" designates the
overall mass transport of the compound across the
membrane, whereas the term "diffusion" designates
only the movement of the penetrant molecules inside
the polymer matrix. For the compounds we have
worked with, the diffusion coefficients are similar and
the solubility of the compound in the membrane and
the extractant appears to be the major parameter for
selection.(1,2) The three major steps in the process
are:
1. The compound of interest in the sample
contacts the membrane and, depending on the
solubility parameters, extracts into the membrane.
717
-------
2. A concentration gradient forms and the
compound diffuses across the membrane.
3. When the compound contacts the extractant
on the other side of the membrane, it partitions into
the extractant depending on the solubility parameters.
Since the silicone rubber membrane has a similar
solubility parameter as hexane, membrane extraction
can be thought of as a combination of a two-step
liquid-liquid extraction with hexane. However,
because the organic phase is solid, many difficult
extraction procedures are now possible:
1. Samples which form emulsions can be
extracted.
2. Very small volume organic/sample extractions
can be performed.
3. Solvents such as acetonitrile, acetone,
isopropanol and methanol, which would normally be
miscible with the aqueous sample matrix, can be
used as extractants.
4. Automated on-line extractions interfaced with
analytical instrumentation can be developed more
easily.
Membrane/Flow Injection Analysis. This is the
simplest of the membrane systems. The detailed
parameters of this system have been reported (2,3)
and a brief description is given below.
In the Membrane/FIA system the membrane is
connected directly to an LC detector (Figure 1). The
membrane is contained in a glass flow-through cell.
For determination of phenol, a dilute caustic solution
is pumped through the tubular silicone rubber
membrane and water or a buffered carrier solution is
pumped around the outside of the membrane. The
sample loop is filled with sample and when the valve
is rotated, the sample is carried past the membrane.
Some of the phenol in the sample permeates the
membrane and forms a phenate salt when it reaches
the caustic. The phenate salt is no longer soluble in
the membrane and concentrates in the caustic. The
carrier flows the extracted phenol to the detector
where it is detected as a peak. Selectivity depends
on the membrane parameters and the detector
response. Since many neutral and basic compounds
are not very soluble in the caustic solution, they
prefer to remain in the membrane. Selectivity can
also be obtained between phenols if their pKa values
differ by 2 units or more. By making the sample a pH
of 9, phenol (pKa=10) will extract while phenols with a
pKa of less than 8 will show very little extraction. By
using a basic carrier and an organic extractant such
as methanol or acetonitrile, neutral compounds are
extracted with the exclusion of phenolics. Therefore,
by adjusting the pH of the sample and the
composition and pH of the extractant, the membrane
acts like a "chemical switch" which can select various
chemical classes.
Extractant
Detector
Sample
FIGURE 1. Membrane/Flow Injection System
Membrane/Liquid Chromatoaraphv. Although the
Membrane/FIA system will work in many situations,
complex samples require more selectivity. The
Membrane/LC system (1) is similar in principle to the
FIA system except the extract flows into a sample
loop of an LC valve as shown in Figure 2. The
contents of the loop are then injected into the LC
system where the extracted components are
separated and detected. Higher concentration factors
can be obtained by using a stop-flow extraction
technique. If the stop-flow valve shown in FIGURE 2
is rotated, the extractant is trapped in the membrane
and analyte continues to permeate and concentrate.
When the valve is rotated to the initial position, the
concentrated extract flows to the LC sample loop.
Phenols and neutral compounds can be determined
at low and sub-ppb levels using this technique.
718
-------
HPLC
Column
Sample Loop
FIGURE 2. MEMBRANE/LC SYSTEM WITH A
STOP-FLOW VALVE.
Another Membrane/LC has been developed (4,5)
which uses only one pump and only one solution that
serves as both the extractant and the LC eluent. An
eight-port valve is used to isolate the membrane from
the back pressure of the LC column and it is rotated
for a short period to allow the concentrated extract to
flow to a sample loop connected to the same valve.
Although this system is not as versatile, its simplicity
makes it useful for dedicated, on-line analyzers.
Membrane/Gas Chromatoaraphy. The
Membrane/GC systems use the same type of
membrane, however, the cell design is modified to
allow for the swelling of the membrane when
extractants such as hexane are use. Two types of
Membrane/GC systems have been developed.(6)
One design, shown in FIGURE 3., combines
membrane cell technology with a pneumatically
operated pressurized injection valve (POPSI). A
hexane extractant flows through the tubular
membrane and extracts permeated compounds. This
concentrate then flows to the injection valve with an
internal 1-3
-------
FIGURE 4. MEMBRANE/GC LARGE VOLUME
ON-COLUMN INJECTION:(a) membrane cell; (b)
extractant pump; (c) sample pump; (d) carrier gas Inlet; (e)
sample loop; (f) extractant waste; (g) oven; (h) retention gap; (i)
capillary union; (j) capillary analytical column; (k) detector.
6. Morabito, Paul; Melcher, Richard; Hiller, Joseph
and McCabe, Terrenes, "Method for On-Column
Injection Gas Chromatography," United States Patent
4,962,042, October 9, 1990.
7. Melcher, Richard and Morabito, Paul,
"Membrane/Gas Chromatographic System for
Automated Extraction and Determination of Trace
Organics in Aqueous Samples," Analytical Chemistry,
Vol. 62, No. 20, October 15, 1990, 2183-2188.
8. McCabe, Terrence; Hiller, Joseph and Morabito,
Paul, "An Automated Large Volume On-Column
Injection Technique for Capillary Gas
Chromatography," Journal of High Resolution
Chromatography, Vol. 12, August 1989,517-521.
9. Morabito, Paul; Hiller, Joseph and McCabe,
Terrence, "A Method for Automated Large Volume
On-Column Injection Technique for Capillary Gas
Chromatography with Solvent Diversion," Journal of
High Resolution Chromatography, Vol. 12, May 1989,
347-349.
References.
1. Melcher, Richard and Bouyoucos, Spiros,
"Membrane Interface for Automatic Extraction and
Liquid Chromatographic Determination of Trace
Organics in Aqueous Streams," Process Control and
Quality, Vol. 1, Issue 1, November 1990,63-74.
2. Melcher, Richard, "Flow-Injection Determination
of Membrane Selected Organic Compounds,"
Analytica Chimica Acta, 214,1988, 299-313.
3. Melcher, Richard, "Membrane Assisted Flow
Injection Analysis," United States Patent 4,819,478,
April 11, 1989.
4. Melcher, Richard and Cortes, Hernan, "Method
for Membrane Assisted Chromatography," United
States Patent 4,775,476, October 4,1988.
5. Melcher, Richard and Cortes, Hernan,
"Apparatus for Membrane Assisted Chromatography,"
United States Patent, granted May, 1990.
720
-------
Candidate Protocols for Sampling and Analysis of Chemicals
from the Clean A1r Act List
R. G. Merrill, Jr., J. T. Bursey, D. L.
Jones, T. K. Moody, and C. R. Blackley,
Radian Corporation, Box 13000, Research
Triangle Park, North Carolina 27709
W. B. Kuykendal, Office of Air Quality
Planning and Standards, U. S. Environmental
Protection Agency, Research Triangle Park,
North Carolina 27711
Clean Air Act (CAA) amendments of
1990 renew and intensify national efforts
to reduce air pollution. Title III of the
Amendments lists 189 hazardous air
pollutants (HAPs) and requires the
Environmental Protection Agency (EPA) to
promulgate new control standards for the
principal sources of such emissions. The
189 HAPs are chemicals not previously
regulated under the National Ambient Air
Quality Standards. These HAPs listed in
the CAA are not expected to be found in a
large number of areas nor in large
quantities. However, health effects may
occur at low concentrations because of
their high acute or chronic toxicity.
Measurement of pre- and post-control
emissions from a wide variety of stationary
sources will be required in order to
determine success in reducing emissions.
Prior to testing at a source, however, a
written sampling and analytical protocol
must be available to ensure that data
acquired during source testing are accurate
and of known quality. In order to optimize
the yield of information from any given
field effort, the sampling and analytical
methodologies which are selected should be
applicable to the broadest possible range
of compounds. Some sacrifice of accuracy
and precision of the methodology may be
necessary to extend the range of
applicability. This study reports the
results of a review and evaluation of
existing information on sampling and
analytical methods presently and
potentially applicable to toxic air
pollutants. Generic Methods which
simultaneously yield information for a
large number of HAP compounds have been
emphasized. The goal of using methods
which cover a broad range of compounds may
initially require the sacrifice of some
method accuracy. Many analytes have been
assigned to generic methods on the basis of
physical properties since individual
compound validation data are often not
available. Such validation data must come
from well designed programs using data
obtained by techniques such as dynamic
spiking of isotopically-labeled compound
analogs while sampling operating sources.
Extension of an existing methodology is
completely valid only when the
applicability of both the sampling and
analytical methodology to the specific
analyte has been established. If complete
field method validation has not been
performed, it is possible that the compound
can be analyzed using the analytical
methodology but not collected
quantitatively, or vice versa.
Some of the methods cited are written
specifically to address gaseous emissions
from stationary sources. Sampling and
analytical methods may be divided into
combustion and noncombustion stack methods,
since CAA requirements will cover both
types of stationary sources. For a
stationary source not related to a
combustion process, sampling and analytical
methodology used for ambient air monitoring
may be applicable. For some of the
compounds listed as HAPs, there is a choice
of methodology. Generic methods have been
emphasized. Specialized
sampling/analytical methodology for
Individual compounds or classes of
compounds has also been summarized when it
was available.
The following methods were selected
as having the broadest possible range of
application:
721
-------
Volatile Organic Sampling Train (VOST)
Methodology1
The VOST methodology is described in
OSW-SW846 Method 0030 for sampling and
Methods 5040 and 5041 for analysis.
Samples in the field are taken using a
specialized sampling train to collect
volatile organic compounds on a sorbent.
The sorbent sampling tubes are then
returned to the laboratory for thermal
desorption through water, collection and
concentration of the vapors, and ultimate
analysis by gas chromatography/mass
spectrometry. The sampling methodology
specifies that organic compounds with
boiling points between 30* and 100'C may be
sampled with the methodology. Compounds
with boiling points below 30°C may be
sampled using special care, and selected
compounds with boiling points above 100'C
may also be sampled in carefully selected
situations. Polar, water-soluble compounds
represent specific problems with this
methodology. The literature presents a
method for optimizing recovery of such
compounds , but even with the optimized
methodology, many problems remain.
Semi volatile Organic Samplino Train
(SemiVOSTl Methodology1
The SemiVOST methodology is describe
in OSW-SW846 Method 0010 for sampling and
Method 8270 for analysis. Samples in the
field are taken using a specialized
sampling train with sorbent to collect a
range of semi volatile organic compounds.
The train is recovered and components of
the train are returned to the laboratory
for extraction, concentration of the
extract, and analysis using gas
chromatography/mass spectrometry. The
sampling methodology specifies only that it
is applicable to compounds with boiling
points above 100'C. Although no upper
limit on boiling point Is posed, the method
is limited to compounds which can be
solvent extracted and analyzed by GC/MS.
Semivolatile organic HAPs in the CAA
Amendments constitute an extensive list,
including single entries that are groups of
compounds. Some members of these groups
will be observed with the Semi VOST
analytical methodology with poor detection
limits, but if higher specificity or
accuracy is required for polychlorinated
biphenyls or polycyclic organic material,
there are specialized applications of the
SemiVOST methodology available. Also,
since the range of compounds on the CAA is
so wide, the sample preparation methodology
will not be optimum for all of the
compounds simultaneously. The broad range
of the method must yield to specific
optimization of the methodology for a
particular class of compounds, such as
derivatization for carboxylic acids, or
adjustment of pH during extraction to
optimize recoveries of particular classes
of compounds.
Hultl-Hetal Sampling Train1
The multi-metal sampling and
analytical protocol is described in OSW-
SW846 Method 0012 for sampling and
analysis. The methodology for the
determination of multiple metals
incorporates a stack sampling train using
specialized aqueous solutions, with
ultimate analysis according to a series of
digestion and analytical methods which
include final quantitation with either
atomic absorption spectroscopy or
inductively coupled argon plasma
spectroscopy. The source samples are
withdrawn isokinetically from the stack
through a heated probe, with particulate
emissions collected on a filter in a heated
filter holder located outside the stack and
after the probe of the sampling train.
The analytical methodology detects and
quantifies metal ions, so an inorganic
compound is not detected as a molecular
species. For example, titanium
tetrachloride would be detected as
titanium; no identity of the molecule
would be retained in the acidic digestion
process by which the sample is prepared for
analysis.
Aldehyde/Ketone Sampling Methodology
Sampling and analytical methodology
for a variety of aldehydes and ketones is
described in OSW-846 draft Method 0011 and
Method 8315. Gaseous and particulate
pollutants are withdrawn isokinetically
from an emission source and are collected
in aqueous acidic 2,4-
dinitrophenylhydrazine (DNPH) solution.
Aldehydes and ketones present in the
emissions react with DNPH to form a
dinitrophenylhydrazone derivative which is
extracted, concentrated, solvent-exchanged,
and then analyzed by high performance
liquid chromatography.
722
-------
The sampling and analytical methods
described above will provide a means of
sampling and analyzing approximately 80-90
percent of the entries on the CAA List.
However, for many of the compounds listed,
a field test under controlled conditions
has not been performed to demonstrate and
evaluate the combined sampling and
analytical methodology. Demonstration
tests need to be performed to determine
recoveries of the compounds from stationary
source emissions exhibiting various matrix
conditions, which will cause the
performance of the methods to vary.
However, a relatively broad coverage of
compounds can be achieved with these
generic methods, saving the expense of
applying individual methods but potentially
losing some of the specificity and
sensitivity of individual methods.
Some of the entries on the Clean Air
Act List will require specialized
methodology because of problems with
reactivity or other difficulties. For
these compounds, some potentially
applicable methodologies are:
acetonitrile by Method 18 (GC/NPD),
bis(chloromethyl) ether by Method 18,
1,3-butadiene by Method 18,
carbaryl, by Method 0010 and Method
632,
carbonyl sulfide, by Method 15,
chloramben, by Method 0010 and Method
515/615,
• 2,4-D salts and esters, Method 0010
and Method 515/615,
t dimethyl carbamoyl chloride, by
Method 0010 and Method 531,
• 4,6-dinltro-o-cresol and salts,
Method 0010 and Methods 8270 and
515/615,
• ethylene oxide, CARB Method 431,
• hexamethylphosphoramide, Method 0010
and Method 632,
• hydrazine by Method 18,
• methanol by Method 18,
• propoxur by Method 0010 and Method
632,
• 2,3,7,8-tetrachlorodibenzodioxin by
Method 23,
• asbestos by CARB Method 427,
• chlorine by modified Method 26 and
OSHA Method ID-101,
• coke oven emissions, Method 0010 and
Method 8310,
• cyanide compounds, by modified Method
6 and NIOSH Method 7904,
t hydrochloric add by Method 26,
• hydrogen fluoride by Method 13A or
13B,
t mineral fibers by CARB Method 427,
• polycyclic organic matter by CARB
Method 429 or Method 5 G,
• radionuclldes, by Method 114.
Many of these entries on the CAA list
represent a very broad category, and no
single analytical methodology will be
equally effective for all possible members
of the category. Some of the categories
are undefined or poorly defined. In such
cases, the analytical methodology specified
will serve for some representative members
of the category.
No applicable sampling and analytical
methods could be found for the following
Clean Air Act List entries:
t diazomethane,
• phosgene,
• calcium cyanamide.
Many of the entries on the CAA List
are extremely reactive. Because of
physical properties such as boiling point,
these entries may be initially assigned to
a specific methodology. However, testing
will be required to demonstrate that the
compound can exist without serious
decomposition under the conditions of heat,
high water content, and possibly acid
content which may exist in a stack. Also,
a given compound may react completely when
it is present in emissions at trace levels,
but significant amounts may survive to be
sampled and analyzed if the compound is
present at levels of parts per million in
the emissions.
Most of the numbered Methods
Incorporate specific guidance for quality
control (QC) and quality assurance (QA) to
ensure that data obtained are of known
quality. Those who wish to apply the
methodology must establish their capability
and continuously train staff and
demonstrate the quality of their results.
Most of the guidance in the areas of method
performance relates to requirements that
the users of the methods:
• perform an initial demonstration of
capability with the method and
conduct ongoing demonstrations of
capability,
723
-------
• maintain accurate records, follow
Chain of Custody procedures,
• demonstrate control of Instrument
parameters,
• demonstrate that equipment 1s not
contaminated prior to use,
• perform appropriate QC daily for all
Instrumentation,
• establish the ability to generate
acceptable accuracy and precision,
• locate/correct any problems in
instrument operation,
• design and execute an appropriate
scheme of blanks of various types,
duplicates, matrix spikes, and matrix
spike duplicates,
• determine the accuracy and precision
of the methodology,
• qualify all data appropriately when
QC criteria not met and,
• participate in performance evaluation
studies, as available.
If the stationary source Is not a
combustion source, useful data can usually
be obtained from stack methodologies, but
methodologies originally developed for
ambient monitoring may also be useful,
with appropriate allowance for the fact
that emissions from a stationary source may
contain significantly different matrices
and higher than ambient levels.
Selection of sampling and analytical
methods is governed by many considerations.
Regulatory requirements dictate the
selection in many instances. The detection
limits which are required for the
analytical methodology dictate a selection
in many instances. Cost is frequently a
major factor in determining which
methodology will be used. Depending upon
cost considerations, for example, it may be
feasible to add an air toxics component to
a source test program with a different
primary mission. Selection of the best
applicable methodology from a wide variety
of potentially applicable methods is a very
difficult choice.
reflect the views of the Agency and no
official endorsement should be inferred.
References
1. United States Environmental Protection
Agency, Office of Solid Waste and Emergency
Response. Test Methods for Evaluating
Solid Waste. Third Edition. Report No. SW-
846. Washington, D.C.: 1986.
2. M. H. Owens, S.
Lachajczyk, Development
Analysis Protocol for
A.
of
Mooney, T.
VOST Sample
Water-Soluble
Volatile POHCs and PICs. EPA-600/8-87-008,
U. S. Environmental Protection Agency,
Research Triangle Park: 1987.
3. W. T. Winberry, Jr., N. T. Murphy, and
R. M. Riggin, Compendium of Methods for the
Determination of Toxic Organic Compounds in
Ambient Air. EPA-600/4-84-041, U. S.
Environmental Protection Agency, Research
Triangle Park: 1988.
Disclaimer
Although the information described in
this article has been funded wholly or in
part by the Environmental Protection Agency
under contract number 68-02-4286 to Radian
Corporation, it does not necessarily
724
-------
THE INVESTIGATION OF SOIL SAMPLING DEVICES AND SHIPPING AND
HOLDING TIME EFFECTS ON SOIL VOLATILE ORGANIC COMPOUNDS
J. R. Parolinl, V. G. King, T. W. Nail, and T. E. Lewis
Lockheed Engineering & Sciences Company
Las Vegas, Nevada
INTRODUCTION
Volatile organic compounds (VOCs) are the most often
encountered class of compounds at Superfund and other
hazardous waste sites. Many VOCs are considered
hazardous because they are mutagenic, carcinogenic, or
teratogenic and commonly are the controlling contaminants
in site remediation projects. Because decisions regarding
the extent of contamination and the degree of cleanup have
far-reaching effects, it is essential that these decisions be
based on accurate measurements of the VOC
concentrations present. VOCs, however, present sampling,
sample handling, and analytical difficulties, especially when
encountered in soils and other solid matrices. Sample
collection and handling activities can often introduce large
sources of random and systematic errors compared to the
analysis itself. Negative bias (i.e., measured value less than
true value) is perhaps the most significant and most difficult
error to delineate and control. This error is primarily caused
by volatilization losses during soil sample collection, storage,
and handling. Currently, no standardized procedures exist
for sampling soils for VOC analysis. Several different
samplers are available for collecting intact and disturbed
samples. Samples are usually removed from the sampler,
which often disturbs intact samples. Samples are then
placed in glass jars or vials and sealed with Teflon-lined
caps. Practical experience and recent field and laboratory
research, however, suggests that procedures such as these
may lead to significant loss of VOCs (1,2).
EXPERIMENTAL FEATURES
Experiments were conducted to evaluate sampling and
sample handling techniques for the collection of soil for
volatile organic analyses (VOA). Because natural soil
systems can be extremely heterogeneous, experiments
were performed by using large (18 in. i.d.) reconstituted soil
columns. The soil was contaminated by the upward diffusion
of VOCs from a glass-bead layer beneath the soil. This
approach produced very homogeneous material for the
evaluation of sampling devices and various sample handling
scenarios. Figure 1 shows the horizontal and vertical
homogeneity in bulk density and moisture content obtained by
this column packing procedure.
Four different sampling devices (treatments) were evaluated:
(1) acetate liner (4 cm i.d.) the contents of which were emptied
out and disturbed, (2) split-spoon sampler with a brass liner
(4 cm i.d.), (3) acetate liner (4 cm i.d.), and (4) acetate liner
(2.5 cm i.d.). Samples from each device were placed in either
a 40-mL VOA vial or a 125-mL wide-mouth jar.
Treatment 1 (disturbed vial sample) exhibited the largest VOC
concentrations (Figure 2). The disturbance resulted in a
homogenized material that had a higher concentration than the
original sample because the shallow, low-VOC-level soil was
combined with the deep, high-VOC-level soil. A vertical
concentration gradient in the soil column was the cause of the
elevated VOC levels in the disturbed sample. The vial-held
disturbed sample yielded greater VOC concentrations than the
jar-held disturbed samples, which indicates that VOC losses
occurred during the homogenization and separation into
aliquots as specified in EPA Method 8240. Of the undisturbed
samples, the jar-held samples collected with larger diameter
samplers (brass or acetate liner) exhibited higher VOC
concentrations than the jar-held samples collected with the
small-diameter, acetate-lined sampler. Samples collected from
the large and small diameter intact cores, using a subcorer,
yielded essentially the same VOC levels. Collection of a small
sample from an intact core with a subcorer and extrusion into
a 40-mL VOA vial maintained the integrity of VOCs betterthan
jar-held samples. The vertical concentration gradient, however,
725
-------
made direct comparisons difficult. Differences between jar-
held and vial-held samples were probably caused by sample
pretreatment rather than by leakage of VOCs from the
containers.
1.20
o
o<
1.15
1.10
1.05
1.00
fet
0-4 4-11 11-19 19-27
depth (cm)
90*
180' I I 270*
center
10.0
9.8
9.6
9.2
o
0 9.0
i)
a 8.8
'5 8.6
E8.4
8.2
8.0
0-4 4-11 11-19 19-27
depth (cm)
Figure 1. Horizontal and bulk density and moisture
content obtained by column packing procedure.
Air shipment of soil samples held in commercially available
sample containers was investigated. Samples shipped by
different air carriers underwent changes In pressure and
temperature. The results of pressure and temperature
monitoring on three commercial air carriers are presented in
Figure 3. This is obviously a small sampling of the
environmental conditions that occur in aircraft cargo holds.
These conditions will vary with the type of aircraft, the
altitude at which the aircraft flies, and the time of year. The
shipment container the monitoring devices were housed in
was not insulated, so the observed pressures and
temperatures are the actual ambient conditions inside the
cargo hold. In-flight intervals are indicated by negative spikes
in pressure. A pressure differential of as much as 2 psi was
exerted upon sample containers. The integrity of VOC soil
samples may be jeopardized when subjected to decreased
pressures in the cargo holds of aircraft.
i; 0.08
V 0.06
C 0.04
8 0.02
vial
a
jor«
1234
treatment number
Figure 2. VOC concentrations in vial-held and jar-held
samples collected with four different sampling devices.
The objective of the shipping effects study was to evaluate the
stability of VOC in soil samples shipped and held in various
commercially available containers. Five soil columns were
reconstituted and samples from each column were taken in the
following manner:
TjreatmentJM. - A1 -5 g aliquot was extruded into a tared 40-mL
I-CHEM amber-glass VOA vial and sealed with a Teflon-lined
septum cap.
Treatment #2 - A1 -5 g aliquot was extruded Into a taiwi 40^1
I-CHEM amber-glass VOA vial and sealed with a modified
purge-and-trap cap (Associated Design & Manufacturing Co.,
Alexandria, VA- ADMC). Prior to analysis the sample with an
ADMC cap was attached directly to the purge-and-trap unit by
pushing the sparger tube into the cap thus dislodging the
Teflon boili ng ball lodged in the bottom of the cap into the vial.
726
-------
Thus the sample was exposed to the atmosphere for only
fractions of a second.
Treatment #4 - A1 -5 g soil was extruded into a tared 40-mL
QORPAK amber-glass VOA vial and sealed with an ADMC cap.
Treatment #3 - A1 -5 g soil was extruded into a tared 40-mL
QORPAK amber-glass VOA vial and sealed with a Teflon-
lined septum cap.
o
0
(N
15
14
13
12
11
10
15
Airborne
Onlorls
pressure
temperature
i i i i
100
90
80
70
60
50
40
30
20
10
0
-50 5 10 15 20 25 30 35 40 45
n
a
12
,,
' '
10
Federal
Express
lot Vigot .!
Ut Vigo
• pressure
• temperature
i i i
60
50
40
30 3
10
-50 5 10 15 20 25 30 35 40
15
14
13
12
11
10
UPS-
l«l VlfOI
• pressure
• temperoture
i i i i
100
90
80
70
60
50
40
30
20
10
-50 5 10 15 20 25 30 35 40 45
elapsed time (hr)
Figure 3. Preliminary pressure and temperature
measurements.
Treatment #5 - the entire contents of the middle liner section
was extruded directly into an Eagle-Pitcher 125-mL wide-mouth
jar (Eagle Pitcher Industries, Inc.) and sealed with a solid
phenolic cap lined with Teflon. (Prior to GC analysis the
contents of the jar samples were prepared as per EPA Method
8240 specifications.)
Treatment #6 - the entire contents of the middle liner section
was extruded directly into an Eagle-Pitcher 125-mL wide-mouth
jar (Eagle Pitcher Industries, Inc.) and sealed with a solid
phenolic cap lined with Teflon.
For the shipping effects study one set of duplicates was placed
in an ice chest and held in the laboratory. Another set of
treatment duplicates was placed in an ice chest with several
Freeze Gel packs and shipped on Federal Express. When
samples returned to Las Vegas, NV, after two days, both held
and shipped samples were removed from the ice chests and
placed in the freezer until analysis.
All the containers evaluated adequately withstood the negative
pressure differentials exerted by air shipment (Figure 4). Jars
and vials may be equally suitable for shipping samples via air
carrier. The greatest VOC loss occurred when soil samples
were transferred from the sampling device to the container and
when the samples were prepared in the laboratory for purge-
and-trap analysis.
SUMMARY
The optimum soil sampling procedure reduces VOC losses by
minimizing sample disturbance during collection and transfer
to a container. The optimum scenario for maintaining the
integrity of VOCs in a sample was found to be collection of an
undisturbed sample with a tube-type sampler (split-spoon or
zero contamination sampler) that has a precut liner. The soil
In the middle liner section was used for sample collection
because it represented the least disturbed material. A 2-g
aliquot was taken from the center of the exposed soil surface
in the liner by using a subcorer. The contents of the subcorer
were extruded directly into a tared 40-mL VOA vial and the vial
was sealed with a modified purge-and-trap cap. The vial was
connected to the purge-and-trap unit without exposing the
sample to the atmosphere.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the University of Nevada
at Las Vegas Environmental Research Center for the use of
their facilities. We are grateful to the following employees of
Lockheed: Neil Amick for assistance in sample analysis; Phil
727
-------
c
o
o
£
c
V
u
c
o
u
1234
.1U
0.08
0.08
0.04
0.02
o nn
UIBK
-
,"?~"—
" — ••—
1234
20
IS
12
8
4
a
!>CE 1
•
'
' il , f\ '
itiiHiv
.loh
en
Jon
234
NOTICE
Although the research described herein has been funded
wholly or in pan by the U.S. Environmental Protection Agency
under Contract No. 68-03-3249 to Lockheed Engineering &
Sciences Company, it has not been subjected to Agency
review and therefore does not necessarily reflect the views of
the Agency, and no official endorsement should be inferred.
Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
treatment number
Figure 4. Shipping effects in containerized samples.
Malley, Mark Sweeney, and Heath Havey for sampling
support; and Dick Hannah for supplying the pressure
transducer used in these experiments. We thank to Roger
Shura of EPA EMSL-LV for instrumental support during the
shipping experiments, and we thank Bill Ahlert of Lawler,
Matusky, and Skelly Engineers for receiving and returning
air-shipped samples.
REFERENCES
1. U.S. EPA, Investigation of Techniques for the Analysis
of Volatile Organic Compounds in Soil, EPA/XXX, EPA
Environmental Monitoring Systems Laboratory, Las
Vegas, NV, 1991, 320 pp On preparation).
2. Siegrist, R, L and P. D. Jennsen. 'Evaluation of
sampling method effects on volatile organic compound
concentrations in contaminated soils," Env. Sci. Tech.
24:1387-1392,1990.
728
-------
DEVELOPMENTAL LOGIC FOR ROBOTIC SAMPLING OPERATIONS
Michael D. Pavelek
Micren Associates, Inc.
863 Tallyho Drive
Hershey, Pennsylvania 17033
(717) 533-8281
Chris C. Fromme
RedZone Robotics, Inc.
2425 Liberty Ave.
Pittsburgh, Pennsylvania 15222
(412) 765-3064
In the past decade great strides have been made in the
application of remotely controlled mobile robots. Uses for
one type of mobile robot known as a sampling robot has
been successfully demonstrated in the nuclear and space
industries. Two examples of successful applications of
sampling robots are the Remote Core Sampler used at
Three Mile Island to determine the depth and severity of
radioactive contamination in containment walls and the
Viking explorer which took and analyzed soil samples
from the surface of Mars. Both of these applications had
very similar developmental driving forces: the cost of
performing the operations with humans was prohibitive
and the environments were too hazardous to even consider
it. In the case of the TMI robot, existing technologies
were utilized in a cost effective manner to perform
required operations while for Viking, new technologies
were developed especially for the mission. In both cases,
however, the beneficial returns outweighed initial
development costs.
The most important factors to consider when undertaking
the development of robotic equipment for sampling
operations for field screening of hazardous wastes and
toxic chemicals are cost, options, and the scope of the
applications.
The primary factors that have an impact on the
development of robotic sampling operations include:
• Does a robotic system yield an obvious advantage?
• What are the specific tasks the robotic sampling
system must perform?
• Is it possible to perform these tasks with one
transporter equipped for bolt-on tooling?
• Where will controls be located? Is portability an
issue?
• What other tasks could the robotic sampling system
perform, or support through reconfiguration, which
would yield positive benefits?
• What unique or unevaluated conditions will be
encountered which may require engineering
evaluations, materials or component testing, or
safety systems? Who will put together the
expertise to properly address these issues?
• How will the robotic system be integrated into
operations? Who will provide overall project
management, design and fabrication and training of
operators?
• What precautions must be taken to control
contamination encountered by the robotic device?
The answers to these questions will begin to put the major
scope of operation into perspective. However, there are
many other items that will require attention both
throughout the evolution of the operation and
afterwards.
Cost Considerations
Cost must be evaluated thoroughly. The amount of money
saved through the elimination of protective clothing, the
improvement of worker safety, the reduction of manual
staffing requirements, and the reduction of waste
generated can rapidly reduce the initial estimates of the
cost of a robotic operation. Conversely, application of
new technology may be hindered by the unanticipated
behavior of materials in untested environments, the
failure to adequately research known limitations of
existing technology, and uncalibrated engineering
solutions to assure satisfactory performance. Any one of
these factors can drive the cost beyond acceptable limits.
In some cases, however, cost is not a factor. For example,
when a particular operation is required and there are no
existing methods to perform that operation, cost may
cease to be a primary factor in the decision making.
Technological Options
One of the first decision points in the development
process is whether to use existing technology or
729
-------
techniques or to develop new technology for the
application. This decision must be made in light of the
capabilities of the available technology, whether
existing equipment or techniques could be modified to
accomplish the task, the projected capability and cost of
new technology, and the risk of developing the new
technology.
The primary objective of most operations is to accomplish
a task reliably, safely, on schedule, and within budget.
The confidence in the capabilities of the equipment and
techniques to be used typically drives the time and
budget projections. The sensitivities of budget and
schedule make the technology development decision
particularly difficult. In some cases, a competitive
advantage in the marketplace can be established with
the development of a new technology, offsetting the
initial development costs. However, when the use of
existing techniques is possible and feasible with respect
to budget and schedule projections, it is difficult and risky
to choose to develop a new technology and place the
operational goals at risk. The development of new
technology becomes an attractive alternative when the
costs for existing techniques are exorbitant, the available
technology is insufficient for the task, or a competitive
advantage can be established in the marketplace.
Development Strategies
There are two strategies available when introducing new
technologies: the ideal strategy and the pragmatic
strategy. The ideal strategy presents the opportunity to
develop the application of a new technology in parallel
with the application of a proven technology. In this
case, the application of the new technology is gradually
phased into operations without creating disruption or
putting objectives at risk. With time, the new technology
permits the operation to become more effective and
increases the competitive position of developers,
rewarding them for their patience and vision through
economic savings and gain.
Operational reality is more pragmatic. New technology
should be seriously considered when necessary operations
cannot be completed with available technology, when
the necessary operational risk poses safety concerns that
cannot be overcome, or when regulatory agencies prohibit
the use of personnel or current equipment. Thus the
developers of successful technologies are rewarded for
their risk and vision through completion of operational
needs often preventing economic penalties and potential
losses. This defines the pragmatic strategy for
introducing a new technology. Note that the results
achieved by the two strategies are similar.
Development Approach
Idealistic and pragmatic development strategies require
different approaches. In the idealistic scenario, the
development should attempt to provide the greatest
possible gain with the least risk to the operation. For
example, a manufacturing operation that has several
similar plants located in key geographic areas should
attempt to identify problem areas in all plants, then
develop and prove robotic technology in one plant. After
the technology has been proven it may be applied in
other plants with much less risk and a greater potential
for gain. Each of the plants could be selected to develop
one application of new technology thoroughly, then
introduce it to the other plants and staff, achieving rapid
improvements in productivity.
Pragmatic development approaches generally do not
require extensive evaluation to determine the most
immediate area of need. For example, at Chernobyl, an
operation was required to move unshielded nuclear fuel
from the roof of one facility into the void left by the
disaster at an adjacent facility. Similarly, at Three Mile
Island, there was a need to remove the damaged core and
radioactive sediments from within the reactor facility.
At both of these accident sites, the operation staff knew
exactly what needed to be accomplished, the question
was how it could be accomplished.
A potential disadvantage of the pragmatic approach,
and of the development of new technology /equipment in
general, is the lack of clearly defined, ongoing objectives
for the new equipment. All too often, new technology or
equipment is developed for a very specific, one-time
operation without planning for other applications.
Developers should always look beyond the immediate
operational needs and attempt to tailor the development
process so that the new technology or equipment can be
easily reconfigured to meet future operational needs.
An example of this development foresight is provided by
Niagara Mohawk Power Corporation. This New York
electric utility recently developed a robotic device to
disassemble a conveyer system and to clean an area where
equipment malfunction resulted in stored radioactive
materials which could not be easily retrieved by plant
personnel. Although it was possible but undesirable to
utilize personnel to perform this task, the developer
decided to develop a specialized robotic system to
accomplish the task. They required that the design of
the device be reconfigurable to accomplish both the
specific task at hand and additional future operations.
The result of this endeavor is that the robotic device has
successfully completed the specific work required, and
will later perform several other planned tasks. This
example illustrates how a pragmatic situation was used
730
-------
as an opportunity to develop robotic equipment and
technology which achieved the immediate objective and
will provide ongoing benefit to the utility through
increased safety of their personnel, reduced requirement
for protective devices, and improved operational
effectiveness. W
Lessons from Experience
The design of robotic devices should incorporate
functional requirements for all conditions that can
reasonably be expected to be encountered during the
deployment of the device. Any additional capabilities
should be evaluated in light of their cost, the level to
which they would enhance system performance, the
extent to which the added capability would avert
catastrophic system failure, and the potential cost of a
catastrophic failure.
The following is an example of when enhanced
performance was justified. In the design of the Remote
Core Sampler, a break-off actuator was specified to
assure safe return of the device and teleoperated
transporter in the event the core did not break and the
drill became embedded in the concrete. During sampling
operations the drill was embedded in the wall at full
depth, the core could not be broken, and the drill could not
be removed. The break-off actuator allowed the
retrieval of the robotic sampling device and the
transporter without difficulty, effectively averting the
catastrophic loss of the robot. This was a cost effective
enhancement.
On another mobile robot developed for TMI, a 10
horsepower electric hydraulic pump powered the onboard
systems. A redundant pump was specified to assure total
operability in the event of pump failure. Only one fluid
reservoir could be fitted into the space available. The
redundant pump would also have been disabled in the
event of a fluid loss from the primary system. A much
smaller pump with a separate reservoir dedicated to the
operation of the primary propulsion units at a reduced
speed would have been a better choice.
The lesson from these examples is that redundant, or
backup, systems may be very effective and actually
salvage some operations and equipment. It is important
to have operations personnel and designers work closely
to achieve realistic functional requirements for the
robotic equipment in light of the environment in which it
will be deployed.
Another important lesson from our experience identifies
the immense value of the use of transporters with bolt-on
tooling instead of multiple dedicated tooling systems.
The Remote Reconnaissance Vehicle (RRV), used during
the clean-up of TMI was specifically designed and
constructed to permit attachment and manipulation of
teleoperated and robotic payloads to perform the entire
scope of operations required for completion of the project.
In addition, this equipment has general application to
other future needs. O)
Bolt-on tooling used with the RRV included:
•Radiation Survey Equipment
•Core Sampling Equipment
•Kraft Undersea Manipulator
•Sludge Sampling System
•Sludge Vacuuming and Pumping System
•High Pressure (Water) Flushing Equipment
•Ultra High Pressure (Water) Concrete Scarifier
•Abrasive Cut Off Wheels
•Rotary Impact Drills
It would have been impossible to equip the RRV with
dedicated systems for each of these tooling functions.
Given the performance record of the system, the
transporter concept with bolt-on tooling was a valuable
asset. We assert that the use of bolt-on tooling is also
appropriate for field screening robots — making them
capable, flexible, and reliable.
With a highly reliable transporter, the tooling does not
have to be infallible. In the event of tooling malfunction,
the transporter returns the equipment to a controlled area
for decontamination, repair, or replacement. In the event
that multiple sample types are required, tooling for each
type of sample could be exchanged after completion of
each sampling step and the program continued using the
same transporter.
Requirements for Robotic Sampling
One of the most frequent problems observed which results
in poor sample results has been the lack of attention to
basic technique. Common examples of conditions which
destroy the results of samples before they ever reach the
laboratory are the collection of gas samples by vacuum
pump into sealed vials not adjusted to standard
temperature and pressure for analysis, the collection of
soil samples which were deposited in the same pouch
without being individually sealed, and the use of the
same scoop for all samples collected.
To achieve accurate characterization of a site it is
imperative that samples are taken under conditions that
guarantee their integrity. This requires that individual
containers be maintained dean prior to sample collection
and be kept sealed with the contents isolated during
731
-------
transportation and storage. Also, control samples must be
taken frequently enough to verify the integrity of the
sampling system and the identity of the samples and
their location must be accurately maintained. This is a
tall order for a robotic sampling system and requires
intense attention to detail.
Some reliable techniques must be used to plot the exact
location of the individual samples. For a specific site,
position readings could be taken from fixed markers by
camera, sonar, or laser technology. In a more
sophisticated setting with larger distances needing
evaluation by a mobile robotic device, location may be
documented by satellite triangulation.
It is also critical that the robotic device does not cross
contaminate the sampling site. Specifically, the
tracking of surface materials from one location to another
or the loss of sample materials during collection could
potentially contaminate previously uncontaminated soil.
The specific contaminants to be investigated and the
degree of their toxicity determine what support will be
required. Some hazardous wastes and toxic chemicals
may be evaluated in simple field facilities while others
may require more sophisticated evaluation. The degree
of hazard present, the ease of analysis, the number of
samples to be analyzed, the requirements and cost for
transportation, and the size and needs of the project are
all factors which will determine support requirements.
Operational Safety
It is important to consider the safety of personnel in the
proximity of robotic devices. The area should be
restricted to exclude personnel not directly working with
the equipment. However, since robotic devices often draw
attention from admiring spectators, it should be expected
that people not working directly with the robots will
often be present. Measures must be undertaken to ensure
the safety of all those present.
Summary
A decade ago several people were faced with a unique
problem at the Three Mile Island reactor. They
envisioned employing a mobile robot that would perform
all of the operations necessary to solve the problem. It
took three years to make that vision a reality.
Similarly, last year a group of people at Niagara
Mohawk envisioned using mobile robots to help solve
some of the problems they had encountered. In this case
it only took six months to transform that vision into
reality.
The amount of time necessary to make a reality of such
visions has decreased tremendously during the past
decade. Mot surprisingly, the technology available has
had a tremendous increase during the same timespan.
Relevant applications of such technology are increasing
each day with the enforcement of more stringent
regulations and with an increased public awareness of the
effects of exposure to hazardous substances. Currently,
the nuclear and space industries are at the forefront both
of robotic technology and the applications of that
technology. Robotic technology is now at a point where it
can be effectively applied to the characterization and
remediation of hazardous waste sites.
Applications with a greater volume of repetitive
operations will be more effective than those which have
a once and done scope. @) The use of personnel involved in
previous projects provides a continuity of experience
which also increases effectiveness, even for the once and
done applications.
For companies with several sites which require
characterization, the greatest effectiveness of mobile
robotic operations is projected through use of a dedicated
team which provides management, supervision, and
continuity of all sites from a central location.
References
RedZone Robotics, Inc., Design and Operations Manuals
for the Tethered Remote Operating Device, Unpublished
(1989)
Pavelek II M.D., Giefer D., and Hine R., "Remote
Reconnaissance Vehicle Program," NP-4265, Research
Project 1544, Electric Power Research Institute (Sept '85)
Vallem RJ. and Jobe E.G., "Decontamination Using
Ultra High Pressure Water at TMI-2," Trans. Am. Nucl.
Soc.,54,90(1987)
732
-------
PRACTICAL PROBLEMS ENCOUNTERED IN REMOTE SENSING
OF ATMOSPHERIC CONTAMINANTS
Kirkman R. Phelps
Michael S. DeSha
Chemical Research, Development, and Engineering Center
Detection Directorate
Aberdeen Proving Ground, MD 21010
Recent sensing technology is now
ready and able to play a significant
role in Environmental Protection
Agency (EPA) programs. The U.S. Army
Chemical, Research, and Development
Center (CRDEC) has been involved in
remote sensing of environmental
contaminants since 1951. During this
time much practical information has
been gathered concerning designing,
building, and testing remote sensing
systems. This paper briefly examines
the practical development of remote
sensing systems which could benefit
the EPA in its mission to detect
potential environmental contaminants.
Perhaps the least exciting and
most tedious of any remote detection
program is data base development. But
for an instrument which will be used
to detect and discriminate thousands
of specific chemical species among a
plethora of natural and manmade
interferents this is the most
important first step in any standoff
detection program. The detection and
discrimination of environmental
pollutants is an exceedingly difficult
problem because a clear base line
measurement is almost impossible to
get. This makes a complete data base
very important.
There are two types of data bases
used in the research and development
of any remote detection instrument.
The initial working data base consists
of spectra, interferents, and
backgrounds acquired either from
in-house laboratory spectrometers
or purchased from spectral data
base houses. The second type is an
instrument specific data base
developed with a first generation
sensor (a crude, working instrument
designed from first principles).
The initial data base is used to
determine the type of instrument,
the spectral band in which this
instrument will work, the spectral
resolution and some initial
detection and discrimination
parameters. This data base need
not be quantitative (i.e.
calibrated to a specific instrument
response) but quantitative data
will save some steps in future
development. The instrument
database is used to fill in details
of instrument development and
"finetune" the instrument for the
work for which it is being
designed.
For the most part two
categories of remote detectors
exist, active and passive. Active
systems are based on the Light
Detection And Ranging (LIDAR)
concept. The detection scheme,
whether it is differential
absorption (DIAL) differential
scattering (DISC), Raman, or laser
induced fluorescence (LIF), is
dependant on the detection
733
-------
requirements of the system. The LIDAR
emits LASER radiation, at frequencies
appropriate for the chemical under
investigation, and the radiation is
scattered back through a telescope to
the detector for analysis. Passive
systems are based on either a grating
or an interferometrie spectrometer.
Interferometric systems are usually
employed for portable field
instrumentation because of the size
and weight advantage over grating
systems of comparable sensitivity.
Passive systems use the ambient
radiation emitted or absorbed by the
chemical vapor under investigation as
the basis for detection. The
detection analysis is similar to that
used in a laboratory Fourier transform
infrared (FTIR) spectrometer. The
choice of either a passive or active
system is a consequence of the
spectral band of interest, resolution,
physical state of the contaminate, and
the use concept.
The selection of spectral band
and resolution is, at least initially,
a consequence of the chemical species
you want to identify. In order to
detect a specific chemical you must
select a band and resolution which
affords the best chance of
identification of the chemical. You
must also consider all possible
interferents against all backgrounds
you may encounter in your use concept.
This a particularly important decision
since all subsequent development will
proceed from this decision point. You
must also determine what resolution
you require to completely discriminate
the chemical among possible
interferents encountered in your use
concept.
We have used the phrase "use
concept" twice in the above paragraph
- what is use concept? Use concept is
nothing more than a notion as how,
where, and under what conditions you
will use your remote sensor. You can
make your initial conditions something
along the line of; I want to use the
detector everywhere, under any
conditions, and operated by an
untrained chimpanzee. You then use
some computer modeling and knowledge
of current technology to back away
from this initial use concept. This
is where a quantitative data base
comes in handy. If you have
absorptivity coefficients or
scattering cross-sections available
for your chemical species and
interferents and spectral
emissivities for the backgrounds
then you may "build" computer
models of detectors and estimate
their sensitivities. Computer
models are very convenient for
estimating performance for
instruments with widely different
design concepts including changes
in fieIds-of-view, detector element
response, LASER power, LASER
frequency agility, etc. The
results of these models can then be
incorporated into a first
generation instrument which can be
used to develop a database that is
instrument specific.
Unless you have ample
resources in electro-optics, opto-
mechanics, optics and system
integration you will run into some
difficulty building your first
instrument. You have four choices,
design and fabricate the instrument
in-house; design in-house, have the
parts fabricated by contract and
integrate the instrument in-house;
design in-house but fabricate by
contractor; or find a contractor to
do the whole job. The first option
requires considerable in-house
expertise and fabrication
facilities but usually very little
money. The second option requires
the same in-house expertise but
almost no fabrication capability
and only slightly more funds. The
third option requires some in-house
technical capability and more money
and the last requires in-house
technical capability, a good
contracting officer and lots of
money. For a research/development
operation we suggest one of the
first two options and of these we
think that the second choice is the
best. Our reasoning goes like this
all of these options require at
least some in-house technical
knowledge and if you don't have any
in-house fabrication capability
then the only other requirement is
some integration capability. In
addition you can rely on some of
734
-------
the expertise you will obtain from the
various parts vendors. This
integration capability can be
initiated with a minimum of startup
time and allows your in-house
personnel to become intimately
familiar with the instrument they are
integrating. This intimacy will
become vitally important in the later
stages of development.
You've got your first instrument
in-house and you've worked out most of
the electronic and mechanical bugs.
What's next? - calibration,
characterization, and collection!
Calibration is a basic precept to
understanding instrument response and
most of the operational attributes of
the instrument. An unknown instrument
response hinders the ability to make
any type of confident detection and
completely destroys any discrimination
capability. The initial calibration
method is dependant on the type of
instrument. For example most passive
FTIR based instruments are calibrated
against some reference source which,
within some practical error, mimics a
blackbody, this permits an
understanding of total instrument
response.
Instrument characterization is
essentially a "calibration" taking
into account use concept, optical
parameters (i.e. fieId-of-view, field-
of-regard, etc.), measurements against
contaminants and interferents in a
controlled environment, and a
measurement of how instrument response
changes in the operating environment.
These measurements permits the
operator to gain a complete
understanding of the instruments
capabilities or lack of capabilities,
the logistics of maintenance and
operation, and its usefulness as a
contaminate detector.
Data collection in order to
define an instrument database is a
time consuming, tedious, and expensive
endeavor but is vitally important to
complete instrument development.
There are two stages of data
collection, open air testing of
interferents, simulants, and natural
backgrounds and controlled chamber
testing of obnoxious and dangerous
contaminants. The chronological
development of this instrumental
database is of little technical
importance and is based solely on
an established or changing use
concept. Chamber testing of the
obnoxious contaminants is necessary
to establish a sensitivity to known
quantities and is performed in
conjunction with simulant
sensitivity measurements to
corroborate instrument responses to
open air tests. Open air testing
of interferents, backgrounds, and
simulants establishes instrument
performance in the operating
environment.
Although detection and
discrimination have distinctly
different definitions in theory
they are practically impossible to
separate in practice. They are
intimately linked simply because
you cannot make a proper detection
of any contaminate without the
ability to discriminate it from
background clutter, for this reason
we will discuss them together.
The initial ability to detect
and discriminate particular
contaminants or class of
contaminants rests solely on the
human perception. The operator
must separate the contaminate from
the interferents and backgrounds
based on the use concept, the
operator's knowledge base, the
depth of the database, and the
operator's ability to understand
and adapt to changing environmental
conditions. By the time instrument
development reaches this point the
use concept has, hopefully, been
established and this leaves us with
the task of developing some type of
detection and discrimination
algorithm based on the remaining
parameters. You can, given an
infinite amount time, money and
manpower, develop an empirical
solution to this problem. For all
practical purposes this is
impossible, and any solution based
on a subset of parameters has at
best a very limited success rate.
Then what is your alternative?
Statistics! The saviour of the
735
-------
working scientist and the bane of the
absolutists. Unfortunately there is
no single statistical method which
defines every detection and
discrimination problem - indeed there
are as many methods as there are
problems. The method is defined by
all of the ingredients in the
development recipe, the use concept,
instrument type, knowledge base, and
database. It is just a matter of
finding a method that fits (e.g.
filtering, database matching, etc.) or
developing a new method from a
combination of previously defined
methods. What ever your choice we
strongly urge you employ the services
of an experienced statistician from
the beginning of your development
effort. It will save you
considerable time and frustration.
There you have it. A
practical, albeit brief, recipe for
the practical problems encountered
in the development of remote
sensing instrumentation. CRDEC has
developed several remote sensors
since 1951, some have successfully
managed the development cycle and
some have not. But despite all the
requirement and funding vagaries,
CRDEC has acquired an extensive in-
house capability based on years of
practical experience. This
experience and capability is
available to the EPA for its own
detection programs.
736
-------
A SI/LI BASED HIGH RESOLUTION PORTABLE X-RAY ANALYZER
FOR FIELD SCREENING OF HAZARDOUS HASTE
Stanislaw Piorek and James R. Pasmore
Outokumpu Electronics, Inc.
P.O.Box L1069, Langhorne, PA
INTRODUCTION
Only four years have passed since
the first publication describing the
application of a portable x-ray
analyzer, (XRF), for on-site
chemical characterization of
contaminated soil [1].
During that period, field portable
x-ray fluorescence (FPXRF) has
established itself as the most
useful technique for a broad range
of environmental applications. Its
well known attributes such as
ruggedness, nondestructiveness,
minimal sample preparation and speed
of analysis are indisputably the
factors contributing to its growing
success. However, it was
technological advances in the
proportional detector (high
resolution) and in microprocessor
technology (computing power and
portable architecture) which really
made feasible a small, truly
portable, battery operated device
with analytical capabilities similar
to the laboratory XRF systems.
FPXRF ANALYZER CONCEPT
The most successful implementation
of the FPXRF for the on-site
screening and analysis of inorganics
in hazardous waste is based on the
aforementioned microprocessor
controlled analyzer connected to a
hand-held probe.
The probe contains an x-ray
source(s), a detector and a means of
reproducible presentation of the
sample for measurement. The
electronic unit accepts the signal
from the probe, processes it and
displays the result. It also
contains power supplies and
interfaces for communicatiion with
the operator and peripheral devices.
A sealed radioisotope capsule
emitting x-ray or low energy gamma
rays is a preferred source of
primary radiation for portable
instruments. Such sources are
rugged, compact, light weight and
drift free.
A high resolution, gas filled
proportional detector has been for
years an integral part of the most
successful FPXRF analyzer available,
the X-MET 880. Its much improved
energy resolution of 10 to 12% as
compared with conventional
proportional counters (20%), made
possible abandonment of mechanical
means of element separation (so
called nondispersive XRF, using a
pair of balanced filter for each
measured element) in favor of energy
737
-------
dispersive XRF, based on electronic
separation of elements according to
their characteristic x-ray energies.
More recently, the probe of the
analyzer has been modified to accept
two excitation sources and thus has
extended the range of elemental
analysis of the probe.
Fig. 1. FPXRF Analyzer X-MET 880
with a gas filled detector
probe.
QUANTITATIVE ANALYSIS
Quantitative analysis is
accomplished by employing empirical
calibration methods. Usually a set
of 15 to 20 samples is required to
develop calibration curves for up to
six analytes per calibration program
(model). The instrument can quantify
six elements in each of its 32
calibrationh models. Availability
of calibration samples may pose a
problem especially in situations
where not much is known about the
site to be analyzed. Since XRF
CLP analyzed
the site to be
cal1ed site
samples). An
is calibration
i set of spiked
technique, it is important that the
calibration samples match in matrix
composition the unknown samples to
be analyzed. This condition can
rarely be met, although the most
accurate results have always been
obtained when the analyzer was
calibrated with
samples collected on
investigated (so
specific calibration
alternative solution
of the analyzer with
soil samples, so called site typical
samples [2]. This approach results
usually in a systematic error (bias)
in the XRF measurements. However, it
can be easily corrected as it is a
common practice to submit 10 to 20%
of all samples measured on the site
with the FPXRF for verification by
contract laboratory program (CLP)
analysis. By correlating the XRF
with the CLP results one is able to
correct for the bias in the
remainder of the XRF results.
This approach has been successfully
used for screening and preliminary
evaluation of levels of contaminats
on a number of sites where FPXRF
could be accepted as a Level I
analytical method (that is
inaccuracy up to +/- 50% relative
and precision up to +/- 10%
relative) [3].
HIGH RESOLUTION SI/LI PROBE
While the
configuration
enables one
limits down to
elements such
etc. [2],
calibration
FPXRF analyzer
described above
to reach detection
100 to 200 mg/kg for
as Cu, Zn, Pb, As,
it has demanding
requirements when
handling the diverse sample matrices
common in analysis of hazardous
waste. To address this problem a
new, Si/Li based, hand-held probe
was designed. The probe combines
unsurpassed energy resolution with
portability and ease of operation.
The heart of the probe is a Si/Li
detector featuring 30 mm active
area and an energy resolution
better than 170 eV for the K-alpha
line of manganese at 1000 cps.
The detector is cooled by a small
738
-------
LN2 capacity with a holding time of
8 hours. Dewar construction enables
operation of the probe in any
position making it truly portable.
There were no adverse effects
observed due to thermal cycling of
the probe. The probe can accommodate
two radioisotope sources to cover
the elemental range from K to U.
Fig. 2. A prototype Si/Li probe.
The probe is equipped with
interlock mechanisms which prevent
source exposure and high voltage
supply to the detector, whenever
the amount of LN2 in a dewar is not
sufficient. The probe can be easily
set-up directly on the soil surface
for true in-situ measurements, or it
can be, after turning it over, used
as a sample probe to measure samples
presented in cups.
Perhaps the most important feature
of the probe is that it can be used
directly with the existing
population of X-MET 880's. The Si/Li
probe is therefore a useful addition
to the many types that already are
used with this analyzer.
Fig. 2 shows a photograph of the
prototype Si/Li probe.
PROBE PERFORMANCE
The advantage of the state-of-the-
art energy resolution of the probe
can be seen in Fig. 3 . The figure
shows two simulated spectra as would
be generated in a sample with Cr to
Fe concentration ratio of 1 to 20.
COMPOSITE SPECTRUM OF 1:20 Cr/Fe RATIO
HIOH PFS. WOP Dr. AND SI/LI DrTTCTO*
/ u V.
CHANNEL NUMBCR
SI/LJ; rWHM -16D.V
Fig. 3
Comparison
resolution.
of detector
It is clear that with a Si/Li
detector it is possible to
distinguish a minor Cr peak from a
massive Fe peak, whereas even with
a high resolution proportional
detector such a faint Cr peak can
hardly be seen.
Fig. 4 illustrates a typical soil
spectrum excited with a Cd-109
source and collected with the Si/Li
probe connected to the X-MET 880
FPXRF analyzer.
As expected, all peaks are clearly
resolved except for the notorious
pair of As K-alpha and Pb L-alpha.
However, it is important to note
that the resolution of the detector
is not the only parameter
determining its overall performance.
For example, a gas proportional
detector has much higher detection
efficiency than a small Si/Li diode.
This is due to the fact that a
typical proportional counter
collects radiation from a much
larger solid angle than a typical
Si/Li detector. However, a
739
-------
HEISTC 2HETUIH091«IOI HOOK SHED KUGI1 8-NUIIKIEM1HUII
Fig. 4. Spectrum of soil sample.
proportional counter will usually
exhibit also a higher background
which somewhat offsets its
efficiency advantage. The
improvement in sensitivity and
detection limits achievable with a
Si/Li detector comes mainly from the
low background of this detector.
Although the proportional detector
exhibits excellent performance with
conditions of optimal separation of
more than Z+2 atomic number spread,
when adjacent elements (or
overlapping spectral lines) are
present, enhanced resolution is of
importance. In such cases of severe
spectral overlap and unfavorable
ratio of analyte concentration to
interfering matrix element the
resolution factor plays a critical
role.
Another important implication of
superb energy resolution is the
ability to separate coherent and
incoherent backscatter peaks of
primary radiation. This enables one
to implement a more sophisticated
data treatment, such as those based
on a fundamental parameters
approach, which can better handle a
diversity of sample matrices.
At present, the Si/Li probe can be
used directly with the X-MET 880 in
an empirical calibration mode. An
extensive development program is
being completed to implement a
fundamental parameters based mode,
initially in a PC connected to the
FPXRF analyzer via its serial port.
Typical detection limits obtained
with a Si/Li probe for a
multielement matrix such as Cu, Zn,
As, Pb are on the order of 30 to 80
mg/kg as opposed to a 100 to 200
mg/kg with a gas filled proportional
detector.
Further work is in progress to
further refine the final probe
design and mathematical algorithms
for data treatment. These results
will be reported in the near future.
[1] Chappell R.W., Davis A.O., and
Olsen R.L. - "Portable X-Ray
Fluorescence as a Screening Tool for
Analysis of Heavy Metals in Soils
and Mine Wastes", Proc. Natl. Conf.
on Management of Uncontrolled
Hazardous Haste Sites, Washington,
D.C., pp. 115-119, HMRCI, Silver
Spring, MD, 1986.
[2] Piorek S. and Rhodes J.R. - "A
New Calibration Technique for X-Ray
Analyzers Used in Hazardous Waste
Screening", Proc. 5th Natl. Conf. on
Hazardous Hastes and Hazardous
Materials, pp. 428-433, HMRCI,
Silver Spring, MD, 1988.
[3] "U.S. EPA Data Quality
Objectives for Remedial Response
Activities Development Process",
EPA/540/G-7/003, US EPA Washington,
D.C., 1987.
740
-------
Measurement and Analysis of
Adsistor and Figaro Gas Sensors Used for
Underground Storage Tank Leak Detection
Marc A. Portnoff, Richard Grace, Alberto M. Guzman, Jeff Hibner
Carnegie Mellon Research Institute
Carnegie Mellon University
4400 Fifth Ave.
Pittsburgh, PA 15213
ABSTRACT
Gas sensor properties are measured with the purpose
of comparing two sensor technologies used for
underground storage tank leak detection. Four types of
Figaro gas sensors and the Adsistor gas sensor were
tested in simulated underground storage tank
environments using the Carnegie Mellon Research
Institute (CMRI) automated gas testing facilities. This
automated system monitored the sensors' responses
while dynamically exposing them to various mixtures
of methane, butane and xylene. The sensors were also
tested to determine the effects of humidity on their
responses. Sensor responses were characterized by
sensitivity, selectivity, and speed of response and
recovery to selected test concentrations of methane,
butane and xylene. The test results are presented as a
list of sensor specifications to allow the potential end
user a direct comparison of these two different types
of sensors.
INTRODUCTION
This study was initiated to help the users of
underground storage tank (LIST) vapor phase product
leak detectors to better understand the capabilities and
limitations of commercial vapor sensors. The study
was limited to characterizing two types of commercial
vapor sensors, the Figaro [1] sensor and the Adsistor
[2] sensor.
Four types of Figaro gas sensors, models number 812,
813, 822, 823, and the Adsistor gas sensor were
tested in simulated UST environments using the CMRI
automated gas testing facilities. The characterization
of these sensors resulted in a set of specifications that
allows direct comparison between the different sensor
types. The Figaro sensors are metal oxide
semiconductor devices that operate at elevated
temperature [1]. The Adsistor sensor operates at
ambient temperature and it works on the principle of
gas adsorption [2] in a polymeric material.
The selection of test gases was based upon a study
performed by Geoscience Consultants, Ltd in 1988
[3]. This study detailed the hydrocarbon vapor
concentration at 27 gasoline service stations from
three diverse geographic regions in the United States.
Their findings indicated that:
• all the surveyed locations had some evidence of
underground methane and gasoline vapor products.
• methane existed in high concentrations at many
locations.
• tracking butane concentrations would be useful in
detecting recent gasoline leaks or spills.
• m-xylene was a large component of gasoline product.
Based on this study, methane was chosen as a potential
interference that may cause false alarms for UST
monitors. Also iso-butane and m-xylene were chosen
as tags because they represent major chemical
constituents in gasoline.
The sensors were tested to determine their sensitivity
and cross sensitivities to methane, butane, and xylene
and humidity to help the UST leak detector
manufacturers to better understand how to use these
sensors. For example, 1) if a sensor responds to
methane but the instrument's user is unaware of this
sensitivity, then, this instrument placed in the field
could produce false alarms due to methane
interference. 2) The humidity level underground at
UST sites is considered to be near saturation [4], If a
monitor is calibrated with dry gas, and the sensor is
placed in the damp underground environment, this
also could lead to false alarms, or worse, no alarm
will be set when a real leak is occurring.
Response time is not a critical sensor parameter for
this application since leaks in USTs generally occur
slowly and site monitoring is done on time scales of
days and not minutes. However, recovery time can be
important in situations where an accidental spill
741
-------
occurs. In this case, if a sensor takes too long to
recover from the spill, the detection of a true leak
could be masked.
Sensor responses were characterized by sensitivity,
selectivity, and speed of response and recovery to
selected test concentrations of methane, butane and
xylene. The test results are presented as tables of
sensor specifications to show the potential end user
the advantages and disadvantages of using various
sensor types for monitoring underground storage
tanks.
EXPERIMENTAL
The data presented was collected using the CMRI
automated gas sensor characterization facility. The
facility has been designed to study the behavior of gas
sensors and characterize their response in terms of
sensitivity, selectivity, speed of response and
recovery, and stability. A computer controlled gas
delivery and data acquisition system (GDS) creates the
test atmosphere in the sensor test chamber and
records the corresponding sensor responses. The GDS
controls and sets proper levels of oxygen, nitrogen,
and water vapor to create a clean baseline
environment through a network of mass flow dilution
modules. This clean air can then be contaminated with
up to five different vapor compounds. For this study,
the facility was modified to independently set
concentrations for methane, (CHU), butane (C4Hs),
and m-xylene (CsHio). The GDS maintained a
constant flow rate of 1 liter/minute.
A second gas system, delivering clean humidified air,
was used to maintain the sensor atmosphere when the
sensor chambers were not connected to the GDS.
An on-line gas chromatograph was used to verify the
delivery of gases to the test chamber both during and
in between tests.
Three test chambers were built to house the sensors.
One chamber was built to test 9 Adsistor sensors and
two chambers to house 12 Figaro sensors, 6 of each
type. All the materials used in the construction of the
chambers were chosen to minimize contaminating the
test atmosphere. The chambers were built to power
the sensors and monitor their responses in accordance
with manufacture literature. The volume of each test
chamber was 1.2 liters.
The test chamber temperatures were monitored
during testing. The recorded room temperature and
that of the Adsistor test chamber temperature was
22°C ±. 1°C. The temperatures of the Figaro test
chambers were 33°C ± 1°C.
TEST DESCRIPTIONS
Several types of tests were performed to characterize
sensor response. These tests include:
• Gas concentration ramp tests to determine sensor
sensitivity and selectivity to individual test gases.
• Target gas excursion test to determine sensor
response to the presence of multiple test gases.
• Water vapor excursion tests to determine sensor
humidity response in the presence of multiple test
gases.
• Response and recovery time tests to determine how
fast a sensor responds to changing concentrations of
test gas.
Gas Concentration Ramp Test
Ramp tests expose the sensors to a single test gas at a
time. The sensors are exposed to five different test
gas concentrations for each test gas. The ranges were
50,150, 500, 1500, 5000 ppm for methane and
butane and 10, 30, 100, 300, 1000 ppm for xylene.
Each concentration was held for 30 minutes before
preceding to the next level. The sensors were exposed
to clean air for two hours between each ramp.
Each of the ramp tests was performed at two humidity
levels. The first set was conducted at 15,000 ppm of
water vapor. This level was chosen to represent the
humidity present at underground storage sites (97%
RH at 55 °F). The second set was done in dry air (less
than 50 ppm water vapor) to simulate sensor
response when exposed to dry calibration gases.
Target Gas Excursion Test
This test was designed to show sensor behavior in the
presence of all three test gases. The sensors were
exposed to relatively small concentrations of the three
gases, as a background level. Then each gas was
separately raised to 10 times its background level.
The background gas concentration level was set to 500
ppm CH4, 500 ppm C^Q, and 100 ppm CsHio in air
containing 15,000 ppm of water vapor.
Water Vapor Excursion Tests
This test was designed to show how changes in
humidity effect sensor response in the presence of all
the three test gases. The background level used was
the same as in the mixture excursion test. The water
vapor concentration was then changed in thirty minute
steps from 15000 ppm , to 5000 ppm, to 1667 ppm,
to 0 ppm water vapor, and then set back to 15,000
ppm.
742
-------
Response and Recovery Time Tests
These tests were performed to determine the speed of
response and recovery to set levels of target gases.
The tests were performed in air humidified to 15,000
ppm water vapor. The sensors were measured at one
minute intervals during the test. The xylene
concentration changed in thirty minute steps from 0
ppm, to 1000 ppm, to 100 ppm, to 1000 ppm and
back to 0 ppm.
The response time is defined as the time from when
the new gas concentration is first introduced into the
chamber until the sensor reaches 95% of its final
reading. The recovery time is defined as the time
from when the new gas concentration is first
introduced into the chamber until the sensor reaches
95% of the total change in the sensor reading. The
final reading for both recovery time and response
time is defined as 30 minutes after the new gas
concentration has changed.
SENSOR MODEL EQUATIONS
To simplify direct comparison of these sensors,
mathematical model's were used to convert sensor
resistance (ohms) into gas concentration (ppm). The
model chosen for the Adsistor is the one suggested by
the manufacture [2]. The model selected for the
Figaro sensors is commonly used in the literature [5].
Adsistor Sensor Model Equations
Adsistor data was collected by measuring the sensor
electrical resistance. The resistance is related to
concentration for most gas vapor, concentrations by
equation 1.
Eqn. 1 R-RD10c/k
where R - Measured resistance Rb - Resistance in
clean air, k = Gas constant at ambient temperature,
and c = Gas concentration (ppm)
The Adsistor sensor resistance versus concentration is
reported to be a straight line when plotted on a semi-
log graph [2].
For this paper, because the sensors did not respond to
the lower test concentration, a two point fit between
the 100 and 1000 ppm xylene were used to determine
Rb and k in equation 1. Solving equation 1 for c yields
equation 2 which is used to translate the measured
Adsistor resistance into a measured gas concentration.
Eqn. 2 c - k log10(R/Rb)
Figaro Sensor Model Equations
Figaro sensor data was collected according to
manufacturer's recommendations and converted to
sensor resistance using equation 3.
Eqn. 3 R = R| (VB - VR)/VR
where R - Resistance (ohms). R| = Load resistor
(3920 ohms), VB - Voltage bias (10 volts), and
VR - (10- VB) = Sensor voltage
The resistance concentration curve was observed to be
approximately linear on a log - log plot. Therefore a
power law model was adopted for these sensors as seen
in equation 4.
Eqn. 4 (a) Log(R) - log(Ro) = Blog(c)
(b) R/Ro = CB
where R = sensor resistance, c - gas concentration
(ppm), B = power law slope, and Ro • sensor
resistance when c-1
The two parameters RO and B are determined by
considering measurements taken at c =100, and
c=1000 ppm for the gas in question. Once the
parameters are determined, the sensor resistance is
translated into concentration by inverting equation 5
Eqn. 5
RESULTS and DISCUSSION
For this abstract only data comparing the Figaro 823
sensor and the Adsistor will be presented. The poster
board data presented shows that the Figaro 812, 822,
and 823 sensors all have comparable responses. The
Figaro 813 sensor is more sensitive to methane than
butane or xylene and is of questionable use for UST
product monitoring.
The test results are presented in terms of sensor
specifications related to sensitivity, selectivity,
response time, and reproducibility. The data
presented in this paper are shown in Tables 1-3. The
data is the average of nine Adsistor sensors, and six
Figaro 823 sensors. The data are reported as the
average measured sensor response along with the
standard and percent standard deviations.
Sensitivity
The ramp tests were used to determine the test gas to
which the sensors were most sensitive. The sensors
were then modeled for this target gas.
743
-------
The Adsistor sensors and the Figaro 823 sensors were
clearly more sensitive to xylene than either the
methane or butane vapors. Thus, these sensors were
all modeled and calibrated for xylene.
Selectivity
The Figaro 823 sensors respond to both butane and
xylene, but are more sensitive to xylene that butane.
The ramp and excursion tests indicate that these
sensors are basically insensitive to methane at the
levels tested.
The Adsistors are sensitive to xylene levels greater
than 100 ppm. These sensors are basically
insensitive to the tested levels of methane and butane.
Water Response
The response of Figaro 823 sensors is strongly
affected by changes in humidity. Changes in reading of
more than 50% were observed when the humidity
varied from dry to wet conditions. This is seen both in
the ramp tests and the water excursion tests.
The Adsistor sensors readings show little effect due to
short term changes in humidity.
Speed of Response and Recovery
Both the Figaro 823 and Adsistor sensors respond and
recover more quickly when changing from one xylene
concentration to another than from clean air to a
xylene concentration level.
Reproducibility
All the Figaro 823 serssors tested in this study showed
wide variations in the sensor parameters and
responses. The spread in response ranged from 15%
to 100% of each other.
The Adsistors sensors tested had model parameters and
sensor responses with in 11% of each other.
CONCLUSIONS
Sensor specifications for direct comparisons of the
two different sensor types, the Figaro MOS sensor and
the Adsistor adsorption sensor, has been presented.
Both sensor types appear to have sufficient properties
to be used for UST leak detection. Both respond well to
xylene, with the Figaro sensor being more sensitive to
lower levels than the Adsistor. Both sensor types are
relatively insensitive to methane, which is the
primary interfering compound underground. The
observed butane response for the Figaro sensor is not
a serious problem since butane js also a component of
gasoline. The Adsistors as a group were more
reproducible, and had a much smaller humidity
interference in comparison to the Figaro sensors.
These two properties make the Adsistor easier to
calibrate and work with from an instrumentation
point of view. However, the Adsistors were observed
to have longer xylene recovery times than the Figaro
sensor.
Stability is a major sensor specification not yet
studied. It plays an important role in determining
how a sensor is employed in UST monitoring. If a
sensor changes with time, independent of the actual
conditions, it could lead to false alarms and/or not
being able to detect a leak. It is recommended that
stability test be undertaken to determine the
calibration periods of the sensors and how their
characteristics change with time.
ACKNOWLEDGEMENTS
This research was funded by the U. S. Environmental
Protection Agency, Environmental Monitoring
Systems Laboratory, Office of Underground Storage
Tanks, Las Vegas, Nevada.
REFERENCES
1) Figaro Taguchi sensors are a product of Figaro
Engineering of Japan represented by Figaro USA,
Inc., P. O. Box 357, Wilmette, IL 60091.
2) Adsistor Vapor Sensors aro products of Adsistor
Technology, P. O. Box 51160, Seattle, Washington
98115.
3) Schlez, C., "Background Hydrocarbon Vapor
Concentration Study for Underground Fuel Storage
Tanks", Draft Final Report for U.S. EPA, Contract
No. 68-03-3409, February 29, 1988.
4) Personal communication with Philip B. Durgin,
PhD, U. S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory,
Las Vegas, Nevada, November 1990.
5) Grace, R., Guzman, M., Portnoff, M., Runco, P.,
Yannopoulos, "Computational Enhancement of MOS
Gas Sensor Selectivity", P-33, Proceedings of the
Third International Meeting on Chemical Sensors,
Cleveland, OH, September,1990
744
-------
Table #
Model P
Xylene F
Calibrati
Xylene F
Cross S
Methane
Butane
1: Figaro £
arameters
B
Ro
123 Sensor Specifications
Calibrated @15K ppm H20]
Average
0.56
9.1E+04
Std. Dev.
0.12
3.6E+04
% Dev.
21 .4%
39.6%
leadings (ppm) @ 15K ppm H2O
id at 100 and 1000 ppm Xylene
Actual Cone
10
30
100
300
1000
leadings (p
Actual Cone
10
30
100
300
1000
ensitivlty (p
5000 ppm
5000 ppm
Average
10.7
43.5
100.0
239.9
1000.0
Std. Dev.
5.8
10.0
0.0
36.1
0.0
% Dev.
53.8%
23.0%
0.0%
15.0%
0.0%
am) @ 0 ppm H2O
Average
0.2
1.4
5.9
38.8
437.8
Std. Dev.
0.3
1.4
4.5
21.5
136.7
% Dev.
141.3%
100.0%
75.0%
55.4%
31.2%
>pm Xylene) @15K ppm H2O
Average
23.5
793.4
Std. Dev.
8.6
792.9
% Dev.
36.6%
99.9%
95% Response Time (Minutes) <§> 15K ppm H20
0 to 1000 ppm
100 to 1000 ppm
95% Recovery Time
1000 to
100 ppm
1000 to 0 ppm
Average
15.30
10.18
Std. Dev.
6.7
7.4
% Dev.
42.3%
68.7%
(Minutes) @ 15K ppm H2O
Average
3.33
4.08
Std. Dev.
1.0
0.9
% Dev.
31 .0%
23.1%
Table #2
Model Pa
Xylene Re
Calibratec
Xylene Re
Cross Se
Methane
Butane
Adsistor Sensor Specifications
rameters [Calibrated @ 15K ppm H20]
K
Rb
Average
2987.72
3.5E+02
Std. Dev.
308.26
3.5E+01
% Dev.
10.3%
10.0%
adings (ppm) @ 15K ppm H2O
1 at 100 and 1000 ppm Xylene
Actual Com
10
30
100
300
1000
Average
61.5
67.9
100.0
233.3
1000.0
Std. Dev.
2.8
2.3
0.0
3.7
0.0
% Dev.
4.6%
3.4%
0.0%
1.6%
0.0%
adings (ppm) @ 0 ppm H20
Actual Coru
10
30
100
300
1000
nsltlvity (pp
5000 ppm
5000 ppm
95% Response Time
0 to 1000
ppm
100 to 1000 ppm
95% Recovery Time
1000 to 100 ppm
1000 to 0
ppm
Average
118.9
126.4
139.0
251.3
997.6
Std. Dev.
13.1
12.5
12.1
10.7
9.7
% Dev.
11.1%
9.9%
8.7%
4.3%
1 .0%
m Xylene) @15K ppm H2O
Average
62.9
61.8
Std. Dev.
4.0
3.2
% Dev.
6.3%
5.2%
(Minutes) @ 15K opm H20
Average
7.29
7.86
Std. Dev.
1.5
1.8
% Dev.
18.6%
20.5%
Minutes) @ 15K ppm H20
Average
>30
>30
Std. Dev.
0.0
0.0
% Dev.
0.0%
0.0%
Table #3: Figaro 823 and Adsistor Sensor Response to Target
Gas Excursion Test and Water Vapor Excursion Test
Calibrated for Xylene @ 15 K ppm H2O
Actual
H20
(Dom)
15002
15002
15002
15002
15002
15002
4999
1667
0
15002
Actual
Methane
(ppm)
500
4999
500
500
500
500
500
500
500
500
Actual
Butane
Jppm)
500
500
4999
500
500
500
500
500
500
590
Actual
Xylene
(ppm)
100
100
100
1000
100
100
100
100
100
100
F
Average
306,7
321.2
1042.6
1720.5
284.7
272.8
157.4
100.1
57.6
318.0
Igaro 8:
Std. Dev.
219.2
234.9
1086.7
696.1
223.5
186.2
102.4
59.3
30.3
247.0
3
% Dev.
71.5%
73.1%
104.2%
40.5%
78.5%
68.3%
65.0%
59.3%
52.7%
77.7%
i
Average
141.2
142.2
142.3
940.6
213.4
134.4
137.1
131.0
127.1
111.3
idsistor
Std. Dev.
8.3
8.6
8.5
10.3
12.1
11.1
13.0
13.9
14.5
10.8
5
% Dev.
5.9%
6.1%
6.0%
1.1%
5.7%
83%
9.5%
1 0.6%
11.4%
9.7%
745
-------
Extraction Disks for Spectroscopic Field Screening Applications
Edward J. Poziomek
Environmental Research Center
University of Nevada, Las Vegas
Las Vegas, Nevada 89154-4009
DeLyle Eastwood, Russell L Lidberg,
and Gail Gibson
Lockheed Engineering and Sciences Co.
Las Vegas, Nevada 89119
Introduction
Field screening methods for hazardous waste site
investigations need to be rapid and low cost to
support on-site monitoring and characterization
activities. The challenges are enormous because of
the number and variety of chemicals that could be
encountered. Detecting and monitoring
contamination of water is one scenario which could
benefit from the availability of a relatively simple field
screening method. The data could be used to decide
whether to apply more rigorous analytical methods in
the field and/or to send samples back to the
laboratory.
The research results described in this paper bring
out the potential of utilizing solid phase extraction
membranes as part of a field screening method.
Concept Description
The idea involves using commercially available solid
phase extraction membranes to preconcentrate
pollutants onto the membrane by sorption from
aqueous solution followed by nondestructive
spectroscopic measurements on site using man
portable or fieldable instruments. Depending on the
analytes being sought and the systems' parameters,
the measurements could involve ultraviolet/visible
luminescence directly, colorimetry/fluorometry with
appropriate reagents, X-ray fluorescence analysis,
and/or radioactivity determination.
The solid phase extraction membranes normally
serve as alternatives to column chromatography in
preconcentrating analytes from dilute solution. The
use of solid-phase extraction techniques to replace
conventional liquid-liquid extraction for isolating
analytes has gained much popularity. Two reviews
on water analysis cite various examples (1, 2). The
usual approach is to use short columns or cartriges
containing various solid sorbents. Such columns are
prepacked and readily available from a number of
manufacturers. The use of solid phase extraction
membranes for preconcentrating analytes is also
gaining popularity. The type of sorbent used to
concentrate trace materials can vary widely
depending on the analyte and the medium. The
sorption theory behind the process relates to removal
of components from both gases and liquids.
Organic analytes preconcentrated on the supports
are usually extracted with an appropriate solvent.
The extract is then analyzed using an appropriate
laboratory method such as gas chromatography
(GC), liquid chromatography (LC) or some
hyphenated technique, e.g., GC-mass spectrometry
(MS). The concept pointed out in this paper involves
examining the extraction membrane directly using
solid state spectroscopy. Laboratory analysis would
be an available option after field screening.
Method Description
A variety of information is available in the literature
on solid phase extraction methodology. For
example, the use of solid phase membranes in the
form of 25- or 47- mm disks for the extraction of
pesticides, polychlorinated biphenyls (PCBs), and
phthalates at the microgram per liter level was
reported recently (3). The purpose of the work was to
replace liquid-liquid extraction with a more rapid and
less labor intensive technique. Standard filtration
equipment (a laboratory suction flask) was utilized.
Groundwater, surface water, and laboratory tap water
were used for pesticide, PCB, and phthalate
analysis, respectively. Adsorbed organic species
were eluted from the disks with a small volume of an
747
-------
appropriate solvent for subsequent chromatographic
separation. Recoveries usually exceeding 80 to 90%
were obtained for the classes of compounds
examined. The membranes were obtained from
Analytichem International under the trademark
Empore with a typical composition of 90% (by
weight) of octyl (C8)- or octadecyl (C18)- bonded
silica particles and 10% polytetrafluoroethylene
(PTFE).
The concept described in the present paper extends
the application described above by examining the
solid phase extraction disks in a nondestructive
manner utilizing solid state spectroscopy before the
elution step. Sufficient information may be obtained
from the spectroscopic examination to often eliminate
the need for any further work thus saving additional
time and resources.
Further savings of time and costs are possible if the
filtration step was also eliminated. For example, the
solid phase extraction disks could be used in a dip
stick mode. Alternatively, tabs of the extraction disks
could be placed into a sample of the water being
examined.
The concept is illustrated in Figure 1. The surface of
the tab, modified with long alkyl chains, attracts the
analytes. After a specified amount of time the tab is
removed, allowed to dry, and examined with an
appropriate portable instrument such as a
spectrofluorometer, depending on the analytes being
sought
The use of solid phase extraction media in a static
configuration in which the analytes must diffuse to
the surface has not been reported previously.
However, we have determined that this is not only
feasible but can also provide semiquantitative
information. An experiment is describe below which
simulates scenarios in which a dip stick is used with
a water sample or in which a tab is inserted into a
well.
Experimental
C18 Empore (TM) solid phase extraction disks were
examined for the sorption of anthracene from water
and then analyzed nondestructively using solid-state
fluorescence spectroscopy. Tabs (1 cm x 2 cm) were
cut from the disks and suspended without stirring in
40 ml of aqueous solutions containing ppb
concentrations of anthracene at room temperature.
The tabs were alowed to stand in the solutions for
given time intervals at different concentrations of
anthracene. The tabs were then withdrawn, allowed
to dry in air, and examined front surface using solid-
state fluorescence spectroscopy. A Spex laboratory
spectrofluorometer was utilized. Figure 2 shows a
SPECTROSCOPIC
FIELD MEASUREMENT
Figure 1. Illustration of the concept of using tabs from solid phase extraction disks to sorb
analytes from aqueous solution followed by nondestructive solid state spectroscopic
examination.
748
-------
CO
UJ
UJ
u
g
o
UJ
c
o
u?
350
EMPORE C18
100ppb
/SO ppb
30 ppb
',10 ppb
1ppb
/BLANK
2 HOUR EXPOSURE TO
ANTHRACENE IN WATER
EXCITATION= 254 nm
380 395 410 425 440 455 470 485 500
WAVELENGTH (nm)
Figure 2. Solid-state fluorescence emission curves of Empore C18 tabs that had been
allowed to stand for two hours in water containing 1-100 ppb of anthracene. (Excitation
wavelength 254 nm; band passes 4 nm and 1 nm for the excitation and emission
monochromators, respectively)
series of solid-state fluorescence curves of Empore
C18 tabs that had been allowed to stand for two
hours in water containing 1-100 ppb anthracene.
The intensity of the emission peak at 380 nm versus
anthracene concentration, was found to be linear.
Various relationships were also found in other
experiments, e.g., a linear increase in solid-state
fluorescence intensity was observed of tabs taken at
various time intervals (minutes to a day) from
solutions containing 10 ppb of anthracene.
Discussion
Though the results reported in this paper are
preliminary, the basic idea of using solid phase
extraction disks in combination with solid-state
spectrqscopy is attractive to pursue for field
screening applications. The individual technologies
have strong scientific bases and do not need
extensive development work, although the use of
solid phase extraction membranes in a dip-stick
mode is new.
Attractive features are listed below:
0 The method is nondestructive.
0 Extraction disks are commercially available.
0 The potential exists for at least semi-quantitative
analysis.
0 The method is relatively simple.
0 The opportunity exists for screening a variety of
organic and inorganic compounds.
0 The method is readily adaptable to decision-
making in the field.
There are no apparent barriers to overcome in
extending the techniques to environmental
monitoring in aqueous media for a variety of
analytes. Nevertheless, the combination of
extraction and nondestructive spectroscopic analysis
749
-------
using solid phase membranes has not been
examined sufficiently to allow limitations to be
defined thoroughly.
Future studies will focus on concept validation.
Various analytes (organic and inorganic) will be
examined using solid phase extraction
disks/membranes both in dip-stick and filtration
modes.
Notice
Although the information in this paper has been
funded wholly or in part by the U.S. Environmental
Protection Agency under Cooperative Agreement
No. CR814702-01 with the University of Nevada -
Las Vegas, and under Contract 68-CO-0049 with
Lockheed Engineering and Sciences Co., it does not
necessarily reflect the views of the Agency and no
official endorsement should be inferred. Mention of
trade names or commercial products does not
constitute endorsement or recommendation for use.
References
1. MacCarthy, P., R. W. Klusman, and J. A.
Rice, "Water Analysis," Anal. Chem.
61(12), 1989, 269R-304R.
2. MacCarthy, P., R. W. Klusman, and J. A.
Rice, "Water Analysis," Anal. Chem.
59(12), 1987, 308R-337R.
3. Hagen, D. F., C. F. Markell, and G. A.
Schmitt, "Membrane Approach to Solid-Phase
Extractions," Anal. Chim. Acta, 236,1990,
157-164.
750
-------
FIELD ANALYTICAL SUPPORT PROJECT (FASP) DEVELOPMENT OF HIGH-PERFORMANCE
LIQUID CHROMATOGRAPHY (HPLC) TECHNIQUES FOR ON-SITE ANALYSIS OF
POLYCYCLIC AROMATIC HYDROCARBONS (PAHS) AT
PREREMEDIAL SUPERFUND SITES
Andrew Riddell, Andrew Hafferty, and Dr. Tracy Yerian
Ecology and Environment, Inc.
101 Yesler Way, Suite 600
Seattle, Washington 98104
INTRODUCTION
Active and inactive voodtreating facilities
employing creosote are one of the classes
of industry most often investigated during
the preremedial or site assessment phase of
hazardous waste investigations and cleanups
in the Pacific Northwest. Creosote is
composed almost exclusively of polycyclic
aromatic hydrocarbons (PAHs). This group
of organic compounds is listed in the U.S.
Environmental Protection Agency (EPA)
Target Compound List (TCL), and significant
numbers of samples are submitted annually
to the CLP for semivolatile (GC/MS)
analysis, which includes the PAH fraction.
Turnaround time between sample collection
and receipt of validated data is generally
7 to 9 weeks.
The Field Analytical Support Project (FASP)
program developed by Ecology and Environ-
ment, Inc. (E & E) is utilized when project
data quality objectives (DQOs) include any
of the following as goals:
o Rapid turnaround of data results;
o Extensive sampling for site
characterization;
o Optimization of sampling location
selection while investigators are
on-site; and/or
o Prioritization of samples for more
expensive CLP analyses.
FASP data are utilized routinely to
supplement and enhance the more rigorously
analyzed CLP results. Use of FASP during
site investigation activities has been
demonstrated to provide both significant
project cost savings and improved
descriptions of contaminant distribution.
E & E's previously developed gas chroma-
tographic method with flame ionization
detection (GC/FID) for analyses of PAHs in
contaminated soil has been demonstrated to
provide results of good comparability with
samples analyzed through the CLP. However,
high performance liquid chromatography
(HPLC) with in series ultraviolet/visible
(UV/Vis) and fluorescence detectors offers
numerous advantages over early FASP
methodologies:
o HPLC instrumentation requires fewer
gases for field analysis;
o Two detectors provide real-time
confirmation of target analytes;
o HPLC allows injection of larger sample
volumes, yielding lower method
quantitation limits; and
o HPLC methodology provides better
resolution than the GC methodology for
comparable analysis times.
The HPLC method developed for analysis of
PAHs in contaminated soil utilizes small
volumes of sample and solvents, and
disposable glassware to minimize the
generation of investigation-derived waste
in the field laboratory. Rapid extraction
751
-------
and analysis techniques are employed to
allow the shortest possible turnaround time
for on-site samples.
SYSTEM SELECTION
Five commercial systems were evaluated for
FASP use: Hewlett Packard, Shimadzu,
Spectra-Physics, Dionex, and Waters. The
primary considerations for purchase were
ruggedness, size, and simplicity (ease of
operation and maintenance). Secondary
considerations were cost, warranty,
compatibility with Nelson analytical data
processing system, and technical training/
support. Finally, potential future analy-
tical uses of the system (other analytes of
interest that may be analyzed with the
chosen HPLC) were investigated.
The Spectra-Physics system was chosen as
the most appropriate and cost-effective
instrumentation for field applications.
The physical space constraints for field
laboratories are met by the system, and the
components operate with standard llOv
power. The system is equipped with a
universal system organizer, which facil-
itates securing the instrumentation during
mobilization for field use. All mainte-
nance (except electrical) is performed
through front entry into the pumps and
detectors, which in E & E experience, is a
critical necessity for field repair or
maintenance. The Spectra-Physics SP8800
gradient pump is equipped with an automatic
maintenance log, automatic cleanup cycle,
and self-diagnostic information on elec-
tronics and flow performance. In a cost
comparison of price quotations, the Spectra
Physics system was the least expensive
overall, with a total system cost of
$30,399.00. The warranty on the Spectra-
Physics pump and the UV/Vis detector is
5 years. Shimadzu, Dionex, and Waters each
offered a 1-year warranty, and Hewlett
Packard offered a 90-day warranty in the
base purchase quote. The field technical
representative for the Spectra-Physics
system is based in Portland, Oregon;
technical support is also available through
an '800' telephone number. Technical
support includes system installation and
on-site training for all chemists.
Specific applications support is also
available. Finally, this system is
currently in use at the National Oceanic
and Atmospheric Administration and the
Federal Drug Administration laboratories;
both laboratories require instrumentation
of rugged, durable quality.
EXTRACTION AND ANALYSIS
One ± .01 gram of soil was weighed into a
12 mL disposable culture tube. The sample
was then extracted twice. Consecutive 5-mL
volumes of acetonitrile were repipeted into
the culture tube, vortexed 1 minute, cen-
trifuged 10 minutes, and combined in a
10-mL graduated centrifuge tube. The
sample extract was evaporated to 1 mL under
a gentle stream of nitrogen. Aliquots of
the concentrated extracts were injected
into the HPLC column for analysis.
The PAHs were analyzed with a Spectra-
Physics Gradient HPLC System equipped with
in-series fluorescence and UV/Vis detec-
tors. A stainless steel chromatography
column (25 cm x 4.6 mm, 5 mm octadecylsilyl
stationary phase) under isothermal
conditions was employed. Analyte separa-
tion was achieved using an acetonitrile/
water mobile phase with initial flow
conditions of 35:65 v/v acetonitrile:water
for 2 minutes followed by a 14 minute
linear gradient to 100% acetonitrile. The
mobile phase composition was then held at
100% acetonitrile for 9 minutes. Flow rate
during the analysis was 1.5 mL/min and the
analytical run time was 25 minutes.
Samples were quantitated based on a
five-point initial external calibration of
all target analytes. The linear regression
coefficients of all analytes routinely
exceeded .995. Samples with large
interfering areas were diluted and re-
analyzed. Samples were analyzed for the
following PAHs:
Naphthalene
Acenaphthylene
Acenaphthene
Fluorene
Phenanthrene
Anthracene
Fluoranthene
Pyrene
Benzo(a)anthracene
Chrysene
Benzo(b)fluoranthene
Benzo(k)fluoranthene
Benzo(a)pyrene
752
-------
Dibenzo(a,h)anthracene
Indeno (1,2,3,-c,d)pyrene
Benzo(g,h,i)perylene
RESULTS AND DISCUSSION
Method quantitation limits for HPLC/UV and
GC/FID are presented in Table 1. For
routine use, HPLC/UV/Vis results vere used
for identification and quantitation of the
PAHs; fluorescence detection allows prac-
tical quantitation limits approximately 10
times lower than the reported UV quanti-
tation limits, and was used primarily for
confirmatory analysis of the quantitative
data.
For this study, a representative number of
samples from a site previously investigated
(and known to be contaminated with PAHs)
were split for sample analysis, matrix
spike analysis, and duplicate analysis, by
both HPLC and GC methodologies in order to
compare the analytical results.
HPLC sample results showed reasonable
agreement with GC comparison analyses.
HPLC data and GC data for three soil
samples are presented in Table 2. Analytes
listed in the method that are not reported
in Table 1 were not detected above the
method quantitation limits by either
analytical system.
To illustrate the efficiency of the HPLC
extraction technique, results from three
matrix spike events are summarized in
Table 3. Three aliquots of a PAH-free soil
sample were spiked and subsequently
analyzed by HPLC to generate the matrix
spike recovery data.
HPLC matrix spike results showed consis-
tently higher recoveries than the matrix
spike analyzed by GC. This difference is
probably due in part to the loss of
analytes during the cleanup procedure
performed as part of GC sample preparation.
Duplicate analysis results of three
contaminated soil samples are reported in
Table A. Duplicate sample results from
both HPLC and GC displayed substantial
variability. This phenomenon was due
primarily to the non-homogenous nature of
the soil matrix at the site; the GC results
from the initial site investigation also
demonstrated this variability. Variability
of the GC results also could be influenced
by the cleanup step of the GC sample
preparation.
CONCLUSION
Recent developments in HPLC, including
gradient elution and dual in-series
detectors, have been introduced into
E & E's FASP arsenal of instruments and
techniques. Chemists may now provide
reliable data of known and documented
quality on PAHs in a near real-time mode to
site investigators. Use of multiple
detectors provides supplemental information
regarding the accuracy of both the quali-
tative identification and quantitative
measurement of target analytes. The data
presented in this study document the
accuracy and precision of the method for
both standards and real world samples.
FASP analyses employing HPLC for PAH
measurements are designed to meet the DQOs
and data use guidelines for the needs of
preremedial site investigators. With
appropriate alterations, this method also
can be an effective analytical option for
other types of investigations involving
screening activities.
753
-------
A FIELD COMPARISON OF MONITORING METHODS
FOR WASTE ANESTHETIC GASES AND ETHYLENE OXIDE
Stanley A. Salisbury, National Institute for Occupational Safety and Health (NIOSH), Atlanta
Region. Atlanta. GA; G.E. Burroughs. NIOSH. Cincinnati. OH; William J. Daniels. Charles
McCammon. Steven A. Lee, Denver Region, Denver, CO.
Purpose and Objectives
Several electronic direct reading Instruments,
that have been or could be used for monitoring
toxic gases and vapors in hospital environments,
were evaluated by Investigators from the National
Institute for Occupational Safety and Health
(NIOSH) in a series of three field studies. The
selected Instruments were used to measure waste
anesthetic gases and vapors released during
surgical procedures 1n operating rooms, and to
monitor ethylene oxide (EtO) concentrations
during the operation of gas sterilizers.
Instrument readings were compared with results
from conventional Industrial hygiene air sampling
methods. The objectives of these field studies
were to: (1) compare calibration and operating
techniques for several types of direct reading
Instruments, (2) compare Instrument operational
advantages and disadvantages during actual field
survey applications and, (3) compare accuracy and
precision of direct reading instruments to
conventional a1r sampling methods. The locations
chosen included two separate field surveys at a
small community hospital, and a third, and more
extensive survey at a large medical university
teaching hospital.
Methods and Procedures
Battery powered air sampling pumps, configured
with bag filling outlet ports, and 40 Liter (L)
Tedlar* sampling bags were used to collect air
samples. The bagged air samples were then
comparatively analyzed by direct reading
Instruments and conventional sampling and
analytical methods. Sampling pumps modified for
bag filling were typically set to a 1 liter per
minute (Lpm) flow rate. During actual surgical
procedures, breathing zone air samples were
collected either from the "Scrub Nurse" or from
the anesthetic cart. Samples were collected
through vinyl plastic tubes connected to air
sampling pump and bag assemblies. Area air
samples were collected near the operating room
exhaust vents.
Sterilization gas samples were also collected in
40L and SOL Tedlar* bags using bag filling
sampling pumps. Air samples were collected from
several points near EtO gas sterilizers during
gas purge cycles and after hospital personnel
cracked open sterilizer doors to dissipate
residual gas before removing a load to an
aeration chamber. Other air samples were
collected from inside mechanical enclosures and
near floor drains. Air samples collected In
Tedlar* bags were then analyzed on-site using the
selected direct reading instruments. Direct
reading Instruments were also used for continuous
and sequential real-time monitoring of EtO and
Freon 12 (d1chloro-d1fluoromethane) during purge
cycle operations and during unloading of EtO
sterilizers.
Where sufficient sample remained in sample bags,
and after direct analysis with instruments showed
measurable concentrations of either halogenated
anesthetics (isoflurane or halothane), EtO, or
Freon 12} air samples were withdrawn from the
bags using conventional NIOSH sampling and
analytical methods'1"3' for the analytes of
Interest. To obtain analytical precision data.
three samples from each selected bag were
collected and subsequently analyzed by the NIOSH
contract laboratory. A three-outlet manifold was
connected to the bag valve and each outlet of the
manifold was connected to the Inlet of the
appropriate sorbent tube. Known volumes of the
sample were pulled from the bags through the
sorbent tubes using pre-calibrated battery
powered air sampling pumps. NIOSH recommended
sampling rates were used. Based on results from
direct reading measurements, sufficient air
sample volumes were pulled from the bags to
ensure the analytes collected on the sorbent
755
-------
tubes were above the NIOSH published analytical
limits of quantitation. Results from laboratory
analyses of the air samples collected from air
sample bags were compared to instrument readings
from direct analyses of those same bag samples.
Direct Reading Instruments Tested
Real-time instruments evaluated were the Briiel &
Kjaer (B & K) Multi-gas Monitor Type 1302, the
Mi ran 103 Specific Vapor analyzer and Mi ran 1B2
Portable Ambient Air Analyzer manufactured by The
Foxboro Company, the Photovac Model 10S50
Portable Gas Chromatograph (GC), and the Summit
Interests, Model SIP-1000 Portable GC.
The Bruel and Kjaer (B & K) Model 1302 gas
monitor uses a photoacoustic spectroscopy
detection technique to measure simultaneously up
to five gases or vapors, plus water vapor, down
to the part-per-billion (ppb) range.
Microprocessor control allows the instrument to
compensate for water vapor and other gaseous
interferences such as carbon dioxide. The
photoacoustic spectroscopy technique uses an
infrared (IR) light source focused through a
chopper which pulses the IR light beam through
one of six optical filters rotated into position
on a filter carousel. Light transmitted by the
optical filter at the predetermined wave length
is selectively adsorbed by the gas being
monitored. The gas sample analyzed is
automatically pumped into a hermetically sealed
analysis cell. The modulating expansion and
contraction of the gas in the cell caused by
heating and cooling of the gas as it is
irradiated by the pulsed infrared light beam
generates pressure waves in the cell that are
detected by sensitive microphones mounted on
opposite sides the cell. The amplitude of these
pressure waves is proportional to the
concentration of the measured gas. After the
first analysis, the filter carousel turns to
bring the next optical filter into position so
that other gases in the cell which adsorb
infrared light at different wavelengths can be
subsequently analyzed. The 1302 is operated by
using the push-buttons and the two-line digital
display on its front panel. Measurement results
are automatically stored in the instrument's
"display memory" and can be permanently stored in
one of ten "background memory" locations.
Display memory data also can be transferred to a
printer or personal computer.
The Mi ran 103 and Mi ran 1B2 analyzers are
single-beam infrared spectrometers. The air
sample analyzed is pumped through the analysis
cell at a flow rate from 25-30 Lpm. Quantitative
analysis of the gas in the cell is accomplished
by electronically detecting and comparing the
energy of an infrared light source with the
energy of the light after passing through the gas
in the cell. Infrared energy lost through
absorption by the gas is proportional to
concentration of the gas in the cell. The Mi ran
103 Specific Vapor Analyzer can monitor several
gases and vapors. However, to change from one
gas or vapor to another, a different filter and
meter scale must be installed. The use of the
Mi ran 103 for monitoring nitrous oxide has long
been the established sampling and analytical
method used by NIOSH investigators.'4' The Mi ran
1B2 Portable Ambient Air Analyzer is a portable
microprocessor-controlled infrared spectrometer
configured with an internal library of 116
precalibrated compounds and ten user selected
compounds. It uses interactive programming to
prompt the operator through available choices and
functions.
Both the Summit and Photovac portable GCs use a
photoionization detector, and each was equipped
with a Carobopak BHT packed column. The carrier
gas used was ultrapure air. Both GC columns were
operated at ambient temperature. Samples and
calibrations standards were injected using
gas-tight syringes. Injection volumes ranged
from 10 to 500 microliters. Gas concentrations
were detected by measuring the peak height of an
injected sample from a recorder output, and
comparing the result to a calibration curve.
Standards were periodically injected between
sample injections.
During the surveys all instruments were subjected
to many span calibrations using known
concentrations of the gases measured. The B & K
1302 was configured with filters to measure Freon
12, halothane, isoflurane, ethylene oxide,
nitrous oxide, carbon dioxide, and water vapor.
Before use, the B & K 1302 was zero calibrated
and humidity interference calibrated according to
manufacturers recommendations. The 1302 was then
subjected to a single point span and
cross-interference calibration for each gas or
vapor to be measured before each series of sample
measurements were made. The Mi ran 1B2 user
library parameters for nitrous oxide, EtO, or
isoflurane were used to set up the instrument.
To optimize accuracy of the 1B2 and 103 when
monitoring EtO or nitrous oxide, a five-point
span calibration was performed using a
closed-loop calibration system. Pre-calibration
data stored in the user library of the 1B2 was
used for measuring isoflurane concentrations in
spiked samples. Calibration of Photovac and
Summit portable GCs was done through microliter
injections of known concentrations. Throughout
the surveys, considerable time and effort was
devoted to calibrating and verifying instrument
accuracy through testing of prepared standards.
Nitrous Oxide, EtO and Freon standards were
prepared from dilutions of pure gases mixed with
clean air or nitrogen in gas sampling bags. A
purchased cylinder containing 9.8 ppm EtO in
nitrogen was also used. Halogenated anesthetic
756
-------
standards were prepared from liquid anesthetic
agents supplied by the hospitals surveyed.
Measured amounts of liquids were injected into
gas sampling bags and mixed with mete red volumes
of clean air or nitrogen to prepare standards
that were then diluted to the desired
concentrations. A mixed Standard of 9.8 ppm EtO
and 30 ppm Freon 12 was also used to test
instrument accuracy when measuring EtO in the
presence of Freon 12.
Using the appropriate concentrations for
performing span calibrations was critical for
obtaining accurate results when measuring EtO
with the 1302. Freon interference over
compensated EtO readings when the instrument was
span calibrated with a 1 ppm EtO standard. B & K
recommends using a span calibration standard of
at least 100 times the detection limit, which for
EtO is 0.2 ppm. When the 1302 was recalibrated
using a 20 ppm EtO standard, over compensation
effects were eliminated. When calibrating for
analytes that are (^ compensated, room air could
not be used for preparing standards because of CO,
build-up in indoor air. Although the Miran 1B2
can measure both isoflurane and nitrous oxide, to
switch from one gas to the other required
time-consuming rezeroing of the Instrument. The
relatively large volume of air sample required
for analysis by the Miran 103 and 1B2 (about 20L)
permitted only one measurement from each sample
bag. It was therefore not possible to measure
both halogenated anesthetic and nitrous oxide
concentrations from the same bag sample using the
Miran 1B2 or 103. To allow both the 103 and 1B2
to obtain a reading from the same sample bag, a
tube from the sampling outlet of the 1B2 was
connected to the inlet of the 103. Of all the
instruments evaluated, the only instrument tested
that could make simultaneous measurements of more
than one gas or vapor from the same sample bag
was the 1302. Neither the Photovac nor Summit
GCs would respond to samples containing
halogenated anesthetics, and the Summit GC would
not detect nitrous oxide. Difficulty identifying
the EtO peak detected on the Photovac GC rendered
all EtO readings from this instrument invalid.
An interference peak from Freon 12 or some other
source made quantitative analysis of low-level
EtO concentrations difficult with the Summit GC.
Measurement Results
Nitrous Oxide
All the instruments used for measuring nitrous
oxide, which included the B & K 1302, Miran 1B2,
Miran 103, and Photovac GC, gave similar
readings. For nitrous oxide concentrations above
10 ppm, instrument responses relative to Miran
103 readings were within ± 5% for all instruments
used. At concentrations below 10 ppm, nitrous
oxide readings on the 1302 averaged 1.82 times
higher than Miran 103 readings. Photovac
readings were within ± 25% of readings obtained
on the Miran 103 for concentrations ranging from
5-110 ppm. The Mi ran 1B2 did not detect nitrous
oxide in sample bags containing Miran
103-detectable concentrations of less than 10
ppm.
Halogenated Anesthetics
Laboratory results from nine Isoflurane samples
collected during surgical procedures ranged from
0.09 to 0.95 ppm. B & K 1302 results from gas
bag samples collected side-by-side with charcoal
tube samples were within ±0.3 ppm of the lab
results. The average relative response of the
B & K 1302 when compared to the average
laboratory results was 0.98. B & K 1302 readings
from two sample bags spiked with isoflurane
averaged 0.16 ppm higher than laboratory analysis
of those same samples. Miran 1B2 readings
average 0.23 ppm lower than the laboratory
results. More comparisons made from two gas-bag
collected samples on a follow-up survey at
another hospital showed B & K 13O2 readings
averaging 0.13 ppm lower than the laboratory
results. Less satisfactory results were obtained
for halothane when comparing B & K 1302 readings
with laboratory results. The average response
from the B & K 1302 analysis of three sample bags
was 3.8 times lower than the laboratory results.
Laboratory results ranging from 1.1 to 1.5 ppm
halothane ranged from 0.3 to 0.5 ppm on the B & K
1302. No other direct readings for halothane
were measured or detected on the other
instruments.
EtO/Freon 12
The Miran 1B2 and 103 EtO readings from gas bag
samples collected near operating gas sterilizers
were consistently higher than the EtO readings
from the B & K 1302, Summit GC, and analytical
laboratory. Of the nine bag samples collected,
only four were subjected to follow-up laboratory
analyses. Laboratory results for one of the four
samples was 4.9 ppm EtO. Direct readings from
analyses of this sample were 4.2 ppm for the
B & K 1302, 5.3 ppm for the Summit GC, 11 ppm for
the Mi ran 1B2, and 9.1 ppm for the Miran 103.
The other three laboratory analyzed samples were
compared only with the 1302 and the Summit.
Average response to EtO for the 1302 and Summit
relative to laboratory results was 1.2 and 1.1
respectively. The 1302 gave a 1.3 relative
response to Freon 12 when compared to laboratory
results for four samples ranging in concentration
from about 1 to 80 ppm.
Concurrent monitoring of EtO and Freon 12 during
real-time measurements and analyses from
collected bag samples showed considerable
EtO/Freon ratio variations in both analytical
laboratory results and B & K 1302 readings.
Although an 88/12 mixture contains a
volume-to-volume ratio of 73% Freon 12 and 27%
757
-------
EtO, only two of 23 samples tested came close to
this ratio. Most samples showed the Freon
component well above the expected 73% level. In
two of the samples the EtO component was greater
than 70% of the total mixture.
Conclusi ons
Any of the instruments tested will give
satisfactory performance for monitoring nitrous
oxide. Until additional field testing shows
consistent accuracy and comparability with
laboratory results, direct reading instruments
may not yet be suitable for monitoring all
halogenated anesthetic gases. Both the B & K
1302 and Summit GC gave satisfactory performance
for the monitoring of both short term and long
term exposures to EtO. The Photovac GC has been
shown to give satisfactory performance for
monitoring EtO,'5' but unknown operational problems
caused the instrument to fail during this field
testing. Freon 12 interferences from the 88/12
sterilization gas will likely give false positive
EtO readings or readings with a high positive
bias on IR spectrometers like the Mi ran 1B2 and
103. The considerable variation in EtO/Freon
ratios noted from various locations near and
during the operation of gas sterilizer equipment
should prohibit the use of 88/12 sterilization
gas as a calibration standard.
References
1. National Institute for Occupational Safety
and Health. NIOSH Manual of Analytical
Methods, 3rd. Ed.. (with supplements).
(NIOSH Publication No. 84-100). Cincinnati,
OH: 1984. (Method 1003).
2. National Institute for Occupational Safety
and Health. NIOSH Manual of Analytical
Methods, 3rd. Ed., (with supplements).
(NIOSH Publication No. 84-100). Cincinnati,
OH: 1984. (Method 1614).
3. National Institute for Occupational Safety
and Health, NIOSH Manual of Analytical
Methods, 3rd. Ed., (with supplements).
{NIOSH Publication No. 84-100). Cincinnati,
OH: 1984. (Method 1018).
4. National Institute for Occupational Safety
and Health. NIOSH Manual of Analytical
Methods, 3rd. Ed., (with supplements).
(NIOSH Publication No. 84-100). Cincinnati.
OH: 1984. (Method 6600).
5. Cummins, Kevin, G.E. Burroughs, Julie
Tremblay. Field Comparison of Sampling
Methods for Ethylene Oxide. (manuscript in
preparation).
758
-------
On-Site and On-Line Spectroscopic Monitoring of Toxic Metal Ions Using Ultraviolet
Absorption Spectrometry
Dr. Kenneth J. Schlager Bernard J. Beemster
Biotronics Technologies, Inc. Beemster & Associates
12020 W. Ripley Ave. 10062 N. Sunnycrest
Wauwatosa, Wl 53226 Mequon, Wl 53092
(414)475-7653 (414)242-9101
I. The need for on-line monitoring of heavy metals
Heavy metals are common by-products in Industrial operations and can thus enter the environment from wastewater
discharges or as leachate from industrial wastes.[1] Wastewater discharges from industrial operations must be
periodically tested for compliance with permit requirements, including limits for several heavy metals.[2] Groundwater
and drinking water quality testing also includes measurement of several heavy metals.[3,4]
Although Atomic Absorption Spectrometry Is the standard method of analysis required for compliance reporting, it is
not a method that is easily adapted for on-line monitoring in factory or field screening applications. Reliable and
affordable methods are needed to detect and measure specific heavy metals In multi-constituent effluents and to detect
specific heavy metals in surface or ground waters.
II. Detection of absorption spectra
Heavy metals tend to form anions that bond with water molecules into compounds known as ligands. These
compounds contain bond structures where electrons can become excited upon exposure to electromagnetic energy
of a specific frequency, resulting in absorption of light in the ultraviolet-visible wavelength range (200 nm to 800
nm).[5] Chemical analysis of liquids using uv-vis absorption spectra does not rely upon detection of a single peak
wavelength as with other forms of spectroscopy, but Instead makes use of an absorption signature across a range of
wavelengths. This signature is a function of all absorbing components in the solution. Special apparatus and techniques
are required to detect the spectra and interpret the information.
III. Apparatus required for detection of heavy metal spectra
Absorption spectra attributable to individual elements can be observed by recording the signature for the element
dissolved in a transparent solvent, such as pure water. Spectra for metals such as chromium, copper, iron, mercury
and zinc have been recorded for the applications discussed in this paper. Figure 1 represents the spectra for several
concentrations of iron, ranging from 0.1 to 2.0 ppm (the actual spectra being mathematical values for absorption at
numerous wavelength intervals). It is possible to characterize an unknown substance In pure water as iron if the
absorption signature matches the pattern observed for Iron. Furthermore, it is possible to estimate the concentration
of iron by comparing the Intensity of the signature for unknown concentrations to the relative intensity of the signatures
for known concentrations.
759
-------
The apparatus required to perform absorption spectroscopy in the laboratory is well known.[6] Basic elements include
a light source for the wavelengths of interest, a transparent cell to hold the sample, a detector to measure the light
remaining after transmission through the sample, and a means to process analysis models for interpretation of the
detected information. Many simple laboratory analyzers such as colorimeters use an optical system that is limited to
one or a few specific wavelengths, which limit the instrument to detection of a specific substance. Other laboratory
instruments have a wider wavelength range, but only look at one or a few wavelengths at a time, requiring mechanical
adjustment to the optics in order to step through a wide range of wavelengths. These instruments are slow and
unsuited for use outside of the laboratory.
ABSORB
1
I
11 ^
|
r
\
^V— -—
.
"> •••
• v'
!.0 pp. fa
X
1.0 • ft
0.5 mm re
0 . 1 1>J"» Fo
,.^' --,
\^
:-:^
130 L8Q 230
ELEMENTS
330
Figure 1. Iron in pure water.
IV. Technology advances for on-line absorption spectroscopy
On-line spectroscopy for field use must be able to rapidly detect a wide wavelength range in a flowing sample or in
a dynamic environment. Several recent technology advances make this possible:
FIBER OPTICS make it possible for there to be distance between the analyzer and the liquid to be
analyzed, with the light source and detector remaining in the analyzer. Transmission of light through
the liquid occurs in a device known as an optrode, which may be immersed in a process tank or flow
stream, or may be designed to permit a sample line to flow through a special optical cell.
ARRAY DETECTORS contain a series of photodiodes, each connected to its own storage capacitor. Each
element in the detector is responsible for a specific wavelength interval, with as many as 1024 intervals
possible in the most advanced version. A fixed grating is used to separate the detected light into
wavelength intervals and to project to light onto the detector. The system used for ultraviolet-visible
absorption spectroscopy (UVAS) can simultaneously scan 1024 intervals from 200 nm to 800 nm.
CHEMOMETRICS is the name collectively given to the statistical and mathematical models used for
chemical analysis of multi-component liquids. These models make it possible to perform qualitative and
quantitative analysis by establishing the contribution that an individual chemical constituent makes to
the overall absorption spectra of the liquid.
V. Chemometric analysis of absorption spectra
Heavy metals must often be analyzed in waters that contain numerous components, resulting in overlapping or closely
grouped spectra. The overall absorption spectra for the liquid is a smooth pattern that results from the effects of
absorption by these individual components. There are three basic steps involved in the process of using absorption
spectra for chemical analysis:
760
-------
QUANTIFICATION involves converting detected spectra for calibration solutions and unknowns into
numerical values that can be processed using mathematical and statistical procedures.
PREPROCESSING of raw data reduces the effects of noise and transforms absorption information into
forms that permit more efficient analysis.
ANALYSIS of absorption values identifies individual components and calculates an estimate of their
concentrations in the liquid.
These three steps are the result of a process that is performed at the beginning of a monitoring project to select the
combination of wavelengths, preprocessing techniques and analysis models that are capable of providing the most
accurate analysis of the analytes of interest in a specific application. This process uses information from several site
specific samples that contain known concentrations of the target analytes. These samples, known as a "learning set"
are used to perform a parallel calculations using combinations of techniques to find the model that produces the lowest
error when actual and predicted values are compared. Several "test sets" are then processed to verify the model.
The quantification step is fairly straightforward. Absorption of light is governed by Beer's Law, which relates absorption
to the absorptivity of the media, path length through the media, and concentration of the absorbing components within
the solution. When all of the absorbing components in the media are known, total absorption at each wavelength is a
function of the sums of all of the absorbing components. A series of simultaneous equations can be used to calculate
absorption. Most often, however, all of the absorbing components are not known, in which case an inverse technique
that defines concentration as a function of absorbance must be used.[7]
Preprocessing of spectra is often done for multi-component solutions or to adjust for noise or drift. Typical techniques
include the use of first or second derivatives of the absorption spectrum, the use of Fourier or Walsh transformations,
and the use of Principal Components Analysis (PCA). PCA uses statistically determined quantities to rotate the
coordinate system such that the original information that may have been aligned on several axes becomes aligned on
only a few axes. In effect, the variables that are highly correlated with one another can be treated as a single variable,
thus simplifying the analys!s.[8,9]
The analysis techniques currently used include multiple linear regressions (using least squares techniques) and
discriminant analysis. Discriminant analysis is a clustering process which defines linear decision boundaries between
information clusters for known concentrations of analytes, and assigns unknowns to an appropriate cluster based upon
detection of significant characteristics for the unknown.[10]
Emerging techniques for analysis include experimental methods such as inductive learning and neural networks,
especially for problems that cannot be simplified through principal components analysis. A technique that shows great
promise is the Lattice-K Nearest Neighbor technique, where known values for variables are organized into the nodes
of a lattice. Predicted values for an unknown are based upon relative distances of variables for the unknown with those
of the nearest neighbors in the lattice.
VI. Application of Chemometrics for Analysis of Heavy Metals
Several recent applications have demonstrated the ability of ultraviolet-visible absorption spectroscopy (UVAS) to detect
various heavy metals in multi-component solutions.
Industrial process (boiler) water was analyzed for the presence of iron and copper. Copper was detected over a range
of 1.0 to 5.0 ppm with an error of 0.047 ppm, while iron was detected over a range of 0.5 to 10.0 ppm with an error
of 0.014 ppm. These were the lowest errors achieved, using Walsh transformations and discriminant analysis.
Iron was analyzed over a range of 0.0 to 10.0 ppm in a complex nutrient solution containing random concentrations
of copper, nitrates, phosphates, calcium, magnesium, sodium, chlorides and other compounds. Figure 2 shows several
spectra for iron in the nutrient solutions. Figure 3 plots actual versus predicted iron values for 20 samples, using linear
regression of untransformed absorbance values which produced an error of less than 0.03 ppm. Nitrates were also
successfully analyzed for this application.
761
-------
inf: WCSnR Fa solutions
i.eeee
mm pith length 8 to 18 pp«
3/1/98
188.8 isa.a see
Figure 2. Iron in nutrient solutions.
PREDICTED VALUE. PPM F«
Figure 3. Actual vs. predicted iron values.
Other applications to date include trace levels of mercury in wastewater (range: 0.0001 to 0.01 ppm), molybdate in
cooling water (range: 1.0 to 2.2 ppm), zinc in wastewater (range: 0.85 to 3.65 ppm), and chromium in wastewater
(range: 0.85 to 4.45 ppm).
VII. Conclusion
Ultraviolet-visible absorption spectroscopy (UVAS) is an emerging technology that is currently being demonstrated for
on-line analysis of heavy metals and other chemical substances to monitor water quality in complex multi-component
solutions without the need to chemically alter samples prior to analysis.
REFRENCES
[1] Lund, Herbert, editor, "Industrial Pollution Control
Handbook", McGraw-Hill, Inc., New York, 1971
[2] Clean Water Act, 40 CFR 121 to 135 and 403
[3] Water Pollution Control Federation, "Understanding Hazardous
Waste Management", Operations Forum, Volume 8, No. 1,
January 1991, pages 22-23
[4] "National Drinking Water Regulations: A Summary of the
Latest Listings", Water Technology, Volume 13, Number 10,
October 1990, pages 30-32
[5] Thompson, Clifton, "Absorption of Radiation",
Ultraviolet-Visible Absorption Spectroscopy, Willard Grant
Press, Boston, Mass., 1974, pages 17-27
[6] Thompson, Clifton, "Instrumentation", ibid, pages 29-45
[7] Thompson, Clifton, "Spectroscoplc Applications", ibid,
pages 47-64
[8] Jolliffe, IT., "Principal Component Analysis",
Springer-Verlag, New York, 1986
[9] Zupan, Jure. "Transformations". Algorithms for Chemists,
John Wiley & Sons, Cichester, 1989, pages 87-142
[10] Goldstein, Matthew, et al, "Discrete Discriminant Analysis",
John Wiley & Sons, New York, 1978
762
-------
RAPID SCREENING OF SOIL SAMPLES FOR CHLORINATED ORGANIC COMPOUNDS
H. Schlesing, N. Darskus, C. Von Hoist, R. Wallon
Biocontrol Institut for Chemische und Biologische Untersuchungen Ingelheim GMBH
West Germany
Cleanup of an industrial site contaminated with chlorinated
organic compounds requires methods for the rapid assess-
ment of many soil samples. An estimate of soil content of
chlorobenzenes, chlorophenols, and hexachlorocyclohexanes
is the EOX value. This determination requires solvent
extraction of the soil, which generally takes at least 2 h and
is therefore too lengthy for the present purpose. In the present
work we have compared this method with the following more
rapid ones:
•Thermal desorption of organic compounds from soil,
followed by combustion in an oxygen atmosphere
(Organochlortest A-P-E; supplier: Burger)
• Measurement in the headspace over a soil sample with a
photoionisation detector (supplier: TIS) test kit based on
extraction and reduction of chlorinated compounds from
soil ("Chlor-N-Soil"; supplier: Dexsil)
The main characteristics of these four methods were compared.
Nine soils of different type and degree of contamination
were examined with the results.
Our provisional method of thermal desorption, which is
still under development, almost always yields higher values
than the EOX method, even though the former have been
corrected for ionic chloride in the soil. Possibly thermal
desorption is more efficient than soxhlet extraction for the
compounds in question. However, except for sample 92/03,
both methods yield the same relative order for the degree of
contamination. This result suggests that the thermal desorp-
tion method merits further development.
No such correlation was obtained for the PID. For the Chlor-
N-Soil test kit, results were obtained for only three samples
because of limited availability of reagent sets. In principle,
this test appears to be applicable within the limited scope of
its specification, but a correlation of the colour change
(violet: "little", yellow-brown: "strong") with approximate
contamination has yet to be established.
763
-------
DEVEDDIMENT OF A MZCRDBORE CAPIIIARY COLUMN GC-FOCAL PIANE MASS SPECTROGRAPH
WITH AN ARRAY DETECTOR FOR FHU) MEASUREMENTS
M. P. Sinha
Jet Propulsion laboratory
California Institute of Technology
Pasadena, CA 91109
A gas chzomatograph-mass spectrograph (GO-MS)
system using a microbore capillary colunn (50
fan i.d.), and a miniaturized focal plane mass
spectrograph (Mattauch-Herzog type) with an
array detector has been developed. The ex-
tremely, snail carrier gas flow rate (0.05 atm
cmmin of helium) through the colunn permits
its direct coupling to the ion source, and
reduces the pumping needs of the MS. The mass
spectrograph with an array detector measures
the intensities of all masses simultaneously.
Analysis of mixtures of compounds, each at a
concentration of 1 ppmv has been performed with
high signal-to-noise ratio. The Tnirmaim de-
tectable quantity of benzene is determined to
be 7.5 x 10 g which corresponds to a concen-
tration of 40 ppb for an injected sample volume
of 0.5 /il. Lower analyte concentration can be
determined by increasing the sample volume
and/or the signal integration time. The system
is .found to have a linear dynamic range of
>10 . Because of its low weight, power, and
high sensitivity, the combination of a micro-
bore GC column and a miniaturized plane mass
spectrograph is uniquely suited for field
analysis.
INTRODUCTION
The combination of a gas chromatograph with a
mass spectrometer (GC-MS) is one of the most
powerful instruments for the analysis for
complex mixtures. GC-MS is eminently suited
for the measurement of environmental pol-
lutants. However, in its present form it has
remained largely confined to the laboratory
because of its mass and power requirements.
Our own interest lies in the development of a
field-portable GC-MS instrument. Such an
instrument is much needed for the real-time,
on-site measurement of pollutants, e.g., at
toxic waste dump sites and for fugitive emis-
sions from various sources. This instrument
should also be fast and possess high efficien-
cy and sensitivity in order to analyze com-
pounds present at lew concentration levels.
In the hyphenated technique of GC-MS, the
speed of analysis is determined by the GC
separation time. Fast separation with high
efficiency can be achieved by the use of a
narrow-bore capillary column (e.g., 50 fan
i.d.) of short length. Also, the carrier
gas flow rate through such a column is very
low which offer the advantage of reducing the
pump-size (often requiring large mass and
power) needed to maintain the proper operating
vacuum conditions in the MS.
Such microbore columns, however, put important
restrictions on the sample size for analysis,
and on the detector: used for measuring the
eluted compounds. ' Extremely narrow and
closed spaced peaks are produced from the use
of microbore columns, particularly in the
early part of the chromatogram. The detector
must, therefore, have a high sensitivity and
a low-time constant for signal measurement.
To maintain the column efficiency, the dead
volume needs to be minimized. These consider-
ations have prohibited the application of
columns of <100 /on i.d. in commercial GCs.
The fast rate of data acquisitions needed to
measure peaks from a microbore column makes it
incompatible with a scanning type mass spec-
trometer.
The aforementioned problems in exploiting the
advantages of a microbore column can be over-
come by the use of a mass spectrograph (non-
scanning) . The capability of a mass spectro-
graph for measuring the intensities of all
masses at the same time confers on it an
almost unlimited speed for obtaining mass
765
-------
spectra. Its sensitivity also is inherently
greater that of a scanning-type MS because the
latter measures the signal at a given mass peak
only for a short dwell time. However, in the
past, the lack of a sensitive ion detector has
been an important reason for not using a nonsc-
anning MS for measurements that required high
sensitivity. Recently, an array detector known
as an electro-optical ion detector (EOID) has
been developed in our laboratory for a focal
plane mass spectrograph (Mattauch-Herzog
type). ' The BOID possesses the simultaneity
of a photoplate (used in focal plane MS) and
the high gain of an electron-multiplier. The
EOID can integrate signals continuously for a
wide range of time (25 ms - 30 s) and, by an
appropriate selection of integration time,
multiple mass spectra from transient samples
(like a narrow GC peak) can be obtained without
sacrificing sensitivity.
Our approach towards the development of a high
performance field-portable GC-MS instrument
consists of combining a short microbore column
and a miniaturized focal plane mass spectro-
graph. In this paper, the new GC-MS system
developed in our laboratory is described. Some
of the results obtained on this system for the
analysis of a mixture of priority pollutants
are also reported.
II. EXPERIMENTAL
B.
Mass Soectroaraoh
A.
Gas Chromatooranh
The experimental arrangement is shown schemati-
cally in Fig. 1. The fused silica microbore GC
column (3.0 m, 50 /an i.d.) with a 0.2 /an bonded
DB-5 stationary phase (J. & W. Scientific,
Folsom, Ca.) was housed in a temperature pro-
grammable oven. The outlet end of the column
was directly led into the ion source of that
mass spectrograph. A sample injector valve
(Valco Instruments) with an internal volume of
0.5 nl was used to inject the sample onto the
column. A pneumatic actuator along withjpilot
valves and a digital valve interface was
incorporated into the sample injector for fast
injection. Samples could thus be injected
reprcducibly in less than 14 ms. GC-grade
helium was_used as a carrier gas at a flow rate
of 40 on s . Because of the small volume flow
rate of the carrier gas (0.05 atm on min ),
it was possible to connect the GC column and
the MS without any interface. The direct inlet
of the column effluents into the ion source
eliminated the dead volume that usually arise
from GC-MS interfaces and allowed for the
complete utilization of the analyte sample.
Two miniaturized focal plane mass spectro-
graphs, one with 2.0" long focal plane and the
other with a 5.0" long focal plane have been
designed and fabricated at JPL. The 2.0"
focal plane covering a mass range of 40-250
amu is destined to be used for field measure-
ments. A photograph of this MS is shown in
Fig. 2a. The magnetic sector of this analyzer
was fabricated from new magnetic materials
having high energy product value, and high
magnetic flux permeability for reducing the
mass of this sector. The 5.0" focal plane MS
covers a mass range of 28-500 amu.
C.
Array Ion Detector
The details of the EOID have been reported
previously. ' In short, it consists of a
microchannel electron multiplier array, a
phosphor-coated (P-31) fiber optic window, and
a photodiode array (PDA). In the BOID, an ion
exiting the magnet impinges on the microchan-
nel array and initiates an electron cascading
process along the channel length. The elec-
trons coining out at the other end of the chan-
nels produce photon images of their parent
ions on the phosphor window (shown in Fig.
2b). The intensities of these images are then
measured by the photodiode array (2.0" long
active region) having a center-to-center
distance of 25 /on between its two adjacent
diodes.
The photodiodes are integrating detectors and
accumulate the photon signal (proportional to
the ion signal) for the desired period of
integration. The position of the photodiode
along the focal plane determines the mass of
the ions producing the ion image at that loca-
tion. The signal stored in the photodiodes
are read (at a rate of 220 kHz) serially by a
computer after a predetermined integration
time. Each readout, called a frame, provides
a mass spectrum of all the ions accumulated
during the integration period. Each diode
accumulates the signal continuously except for
its read-out time ("4 jus) when it is reset and
resumes signal integration. This allows for
the complete mass spectral measurement of GC
effluents at a high frequency without any loss
of sensitivity in the process.
Both of the mass spectrographs described above
are equipped with their own array detectors.
The computer interface electronics for the
small MS has not been cotpleted at this time
and, therefore, the results reported in the
paper were obtained on the 5.0 in. focal plane
MS. For laboratory measurements, this did not
create any complications and demonstrated the
analytical capability of the MS-EOID system.
766
-------
Moreover, it is expected that the new 2.0-in.
array detector will have better performance
because of the minimization of the signal
losses at the PDA-fiber-optic window interface
in this design.
A mixture having a concentration of 1 ppmv in
air of each of the compounds listed in Table 1
was prepared. The internal volume of the
injector valve was filled with this mixture and
injected on the GC column for analysis.
RESULTS AND DISCUSSIONS
The mass chromatogram of a mixture of the
compounds listed in Table 1 is shown in Fig. 3.
Each component in the mixture had a concentra-
tion of 1 ppmv in air. The GC column was
maintained at the room temperature and a signal
integration time of 250 ms for the array detec-
tor was used in the measurement. Complete mass
spectra of the components eluting into the ions
were recorded every 250 ms. In obtaining the
mass chromatograph, the sum of the intensities
of all masses (>45 aim) in each record (frame)
is plotted against the corresponding frame
member (time).
The chromatogram shows that the components
(dichlorodifluoromethane, chloromethane, bromo-
methane and chloromethane) correspond to peaks
2-5 are narrow and closed spaced. For example,
the peak-to-peak separation between 2 and 3 is
less than 700 ms and the full width of peak 2
is about 300 ms. Quantitative measurement of
such GC peaks are made possible by the simul-
taneous measurement of all ions and by the
proper selection of the signal integration
time.
The continuous measurement by the EOID with a
short integration time (>25 ms) can be used to
perform time-resolved mass spectral measurement
and can be applied to resolve otherwise over-
lapping GC peaks. Figure 4 demonstrates the
effect of measurement time on resolution of
compounds by the microbore column. It is seen
in Fig. 4a that bromomethane and chloromethane
corresponding to frame numbers 89 and 95,
respectively, are well separated when an in-
tegration time of 100 ms is used for their mass
spectral measurement. For 250 ms integration
time, the chromatcgraphic separation is barely
adequate (Fig. 4b) but the separation is lost
when spectral measurements are made every 500
ms (Fig. 4c). The time resolution capabilities
of the MS-BOID make it particularly useful for
short columns of moderate resolving power.
Their combination reduces the analysis time and
renders it suitable for a field-portable GC-MS
analyzer.
It should be noted that the quantitative nature
of measurement is not conpromised by the
number of mass spectra (frames) obtained from
a GC peak because of the continuous and simul-
taneous measurements of ion intensities.
Figure 5 shows that some of the intensities
contained in all the frames of a GC peak
(corresponding to dichlorodifluoromethane) is
independent of the integration time used in
recording these frames. The sum of intensi-
ties determines the amount of the compound.
The mass chromatogram (Fig. 3) demonstrates
that this GC-MS system can readily analyze
mixtures of compounds present at the 1 ppmv
level without preconcentration of the analyt-
ical sample. From these data, the minimum
detectable quantity (MDQ) was calculated for
each compounds. For benzene this amounts to
7.5 -10 g, which corresponds to a concentra-
tion of 40 ppb for an injected volume of 0.5
/il (results of 100 ppb mixtures of benzene and
chloroform are included in Fig. 6). lower
analyte concentrations (<40 ppb) can be deter-
mined by increasing the sample volume and/or
the signal integration time. However, larger
volumes (>2 /il) cannot be injected without
degrading column resolution. The problem can
be overcome by sweeping the sample from an
injector valve and cryofocusing the volatile
organic compounds at the head of the column,
thus, removing the air. The temperature of
the column can then be programmed for subse-
quent analysis.
A series of mixtures of chloroform and benzene
of various concentrations (0.1 - 100 ppmv) in
air was prepared to study the dependence of
mass spectral intensity on concentration.
These mixtures were injected onto the GC
column and their mass spectra were measured.
In Fig. 6, the sum of the intensities of a
single mass (m/z = 83, characteristics of
chloroform) and also of a group of masses (76-
78 amu characteristics of benzene) contained
in frames of the respective GC peaks have been
plotted. The intensity is found to increase
linearly with concentration showing a linear
dynamic range of >10. This is the range with
a constant integration time of 250 ms. It is
possible to further extend the dynamic range
by suitably adjusting the signal integration
time. The straight lines in Fig. 6 are the
least square fit through the data points. A
linear-correlation coefficient equal to 0.99
is found for mass spectral measurement of
benzene showing an excellent correlation
between concentration and intensity.
CONCLUSIONS
A GC-MS system using a microbore column (50 pm
i.d.) and a miniaturized mass spectrograph
767
-------
with an array detector has been developed. The
performance of this system in the analysis of
mixture of priority pollutants has been demon-
strated. A short microbore column (50 /an i.d.,
3.0 in. long), when combined with the MS-EOID,
resolves the early eluted gases satisfactorily.
The GC-MS system described above possesses hioh
sensitivity and a linear dynamic range of >10 .
The minimum detectable quantity (MDQ) for
benzene is found to be 7.5 x 10 g which cor-
responds to a concentration of 40 ppmv in a
sample volume of 0.5 /Ltl. larger sample volume
can allow measurement of lower concentrations.
The combination of a microbore column and a
miniaturized focal plane MS is eminently suited
for field measurements. The extremely small
carrier gas flow rate drastically reduces the
mass and power needs of the mass spectrograph.
ACKNOWLEDGMENTS
The work described in this paper was performed
at the Jet Propulsion Laboratory, California
Institute of Technology and was supported in
part by the U. S. Environmental Protection
Agency (Grant No. R-814410-0-01-0).
REFERENCES
1. Guiochon, G. Anal. Chem., 50, (1978)
1812.
2. Schutes, C. P. M., Vermeen, E. A. Rijks,
J. A., and Cramers, C. A. "High Speed
Profiling of Complex Mixtures by Means of
Gas Chromatography in Narrow Bore Capil-
lary Columns" in Proceedings of the 4th
Symposium on Capillary Chromatoqraphy.
Kaiser, R. E. (Ed.), Institute of Chroma-
tography, Bad Durkhein, Germany, p. 687.
3. Trehy, M. L. Yost, R. A. and Dorsey, J.
G., Anal. Chem., 58, (1986) 14.
4. Holland, J. F. Enke, C. G., Allison, J.,
Stuffs, J. T. Pinkston, J. D., Newcome,
B., and Watson, J. T. Anal. Chem, 55,
(1983) 998.
5. Leclercq, P. A. Schutjes, C. P. M, and
Cramers, C. A., J. Chromatofr. Libr., 32.
(1985) 55.
6. Boettger, H. G., Giffin, C. E., and
Norris, D. D., "Electro-optical Ion
Detector in Mass Spectrometry: Simul-
taneous Monitoring of All Ions Over Wide
Mass Ranges," in Multichannel Image
Detector. Talmi, Y. (Ed.) ACS Symposium
Series No. 102, ACS, Washington, D. C.
(1976) p. 292.
Sinha, M. P., in "Particles in Gases and
Liquids II: Detector, Characterization
and Control," Mittal, M. L. (Ed), Plenum
Publishing Corp., 1990 (in press).
Harvey, M. C. and Stearns, D. D., Anal.
Chem 56 (1984) 837.
Compounds
TABLE 1
Peak No. (Fig.
air
dichlorodifluoromethane
chloromethane
bromomethane
chloroethane
dichloromethane
1, 1, 1 - trichloroethane
chloroform
benzene
trichloroethylene
1
2
3
4
5
6
7
8
9
10
768
-------
ELECTRIC
SECTOR
MAGNETIC
SECTOR
MICROCHANNEL
ELECTRON
MULTIPLIERS-
PHOSPHOR
FIBER OPTICS
PHOTODIODE
SAMPLE
INJECTOR
VALVE
ELECTRO-OPTICAL
ION DETECTOR
•MICROBORE
COLUMN
(50 urn ID)
WASTE
SAMPLE
CARRIER GAS (HELIUM)
Fig. 1 Schematic of the microbore capillary column gas chromatorgraph and the local plane mass
spectrograph assembly. The sample injector is pneumatically actuated and is provided with
pilot valves and a digital valve interface for fast sample injection.
-------
-ELECTRIC SECTOR ELECTRQ~OPTICAL ION DETECTOR7
-MAGNETIC SECTOR /
/
5* V *«A
Fig. 2a: Photograph of the focal plane (2.0-in) mass spectrograph
with an electro-optical ion detector.
Fig. 2b: Photograph of ion images
770
-------
2100
50
100
150 200
FRAME NUMBER
250
300
Fig. 3: Total ion chromatogram obtained from a mixture of compounds listed in Table 1. Each component
in the mixture has a concentration of 1 ppmv. A sample volume of 0.5 \i I was injected, and a
signal integration time of 250 u.s was used for each frame.
771
-------
2400
1600
Z 1200
(a)
20
60
SO
40 60
FRAME NUMBER
89
80 100
2000
1600
Z 1200
400 -
(b)
28
10 20
FRAME NUMBER
30
30
2400
2000
I 1200
800
400
(c)
10
9 12
FRAME NUMBER
15
17
15 18 21
Fig. 4 Effect of signal integration time on resolution of GC peaks. Integration times of 100,250, and
500 us were used fora frame in (a), (b), and (c), respectively. The peaks corresponding to
dichlorodifluoromethane and chloromethane, and bromomethane and chloroethane are not
resolved with 500 us integration time.
772
-------
2
55
z
LLI
I
Fig. 5 Sum of the intensities of the
frame comprising the last two
peaks in Figs. 4 a, b, c are
plotted against their frame
integration time. The sum is
found to be independent of the
integration time.
200 400
INTEGRATION TIME (ms)
600
Fig. 6 The straight line plots show a
linear dynamic range of >103
Os represent the sum of intensities
in various frame of mass 83
(characteristic of chloroform) where
as As represent the sum of
intensities for a group of masses
76-78 (characteristic of benzene).
CONCENTRATION (ppm)
773
-------
APPLICATION OF A RETENTION INDEX APPROACH USING INTERNAL STANDARDS TO A
LINEAR REGRESSION MODEL FOR RETENTION TIME WINDOWS IN VOLATILE ORGANIC ANALYSIS
Russell Sloboda
NUS Corporation, Wayne, Pennsylvania 19087
The retention time (RT) of an analyte can be predicted by
measuring its fractional distance between the RT of the
internal standard (IS) eluting before the analyte and the RT of
the IS eluting after the analyte. The development of RT
windows using the retention index method involves
calculation of a prediction interval that is derived using linear
regression statistics. This approach can also be adapted to
the relative retention time (RRT) method, which uses one
rather than two internal standards to predict an analyte's RT.
Linear regression equations were derived and software was
developed for linked operation with a gas chromatography
(GC) data system. RT windows were established to compare
the performance of the RT index method versus the RRT and
external standard techniques. Data sets were generated
under a variety of conditions for purge and trap analysis of
calibration standards using megabore capillary and packed
columns with a dual detector system. RT windows generated
using the RT index method were much narrower than those
obtained using the RRT method in variable-temperature
environments and slightly narrower than the RRT method in a
controlled-temperature environment. The RRT and the RT
index methods generated much narrower windows than the
external standard method under all conditions. All methods
were successful in terms of speed of calculation, minimal QC
failures, and few interferences. Because the RT index method
achieves the narrowest prediction windows of the three
techniques, it offers increased specificity of analyte
identification without changing the probability of missing an
analyte that is present.
INTRODUCTION
When GC analysis is performed in a mobile laboratory,
fluctuations in ambient temperature and other factors can
cause greater variations in analyte RTs than under more
controlled laboratory conditions. The use of ISs can improve
the accuracy of RT predictions under such conditions. The RRT
method successfully compensates for RT shifts when the
analyte elutes very close (within ± 20 percent) to the IS but
provides a less accurate estimation of RTs for analytes that
elute farther away. Whereas the RRT method assumes that an
analyte's RT will be increased or decreased in proportion to
the ratio of the RTs of the IS in the sample divided by the
standard, often analyte behavior is intermediate between a
proportional RT shift and a constant, additive shift. The RT
index method offers the advantage of compensating for
either type of shift by means of a linear interpolation in the
predicted magnitude of RT shifts in the region from the IS that
elutes before the analyte to the IS following the analyte. If
the general 2 parameter linear equation for prediction of RTs
is employed (ISi • A + IS2 • B s predicted RT), then it can be
shown that a range of predicted RTs will occur as the fitting
parameters A and B are varied, with the predictions based
upon the RRT method using the first IS (B = 0) or the second IS
(A = 0) at either extreme and with the RT index method
(having the side constraint A + B = 1) yielding a prediction that
is intermediate between the two RRT methods. In addition,
the RT index method is more practical than other enhanced
prediction techniques. This is because it is only a one-
parameter method and so does not require multiple
calibration runs to compute the prediction coefficients as is
necessary with the true two-parameter linear model
referenced above.
GC conditions were selected for rapid and simultaneous
separation and identification of 33 analytes using purge and
trap sample preconcentration, DB-624or 1% SP1000 analytical
columns, and a PID/ECD dual detector system. Three IS
reference peaks were used for each detector. Standard
operating procedures (SOPs) and computer programs were
written so that the width of the RT window for each analyte
was computed as plus or minus the square root of the variance
of the observed minus the predicted RTs within a set of
standard analyses, multiplied by the student's t-value using a
99% confidence level. The width of RT windows calculated in
this manner was considered valid for up to 60 days, as long as
GC conditions remained constant. The center of each
analyte's prediction window was calculated every day using
the RTs from a single standard run in the 12-hour period prior
to the sample. Predicted RTs were computed for all three
techniques (external standard, RRT, and RT index). The RRT
technique and the RT index technique were designated as the
primary methods for analyte identification, with the external
standard technique applicable only to those (rare) situations
where interferences precluded the use of ISs for RT prediction.
The GC system was operated in the Region 3 EPA Field
Investigation Team (FIT) mobile laboratory and also in the FIT
base-support facility. The effect of ambient temperature
fluctuations was investigated by analyzing a series of
standards and calculating and plotting RT window widths
under several different temperature conditions and
environments. In addition to demonstrating the overall
superiority of the RT index method over the RRT and external
standard methods, this investigation provided insight into
several factors that influence RT variance.
775
-------
THEORY
All three RT prediction techniques discussed above utilize RT
windows that are confidence intervals for the predicted minus
the observed values of the retention time. The RT window
width is computed as a 2-tailed 99% confidence interval using
the t-distribution coefficient multiplied by the square root of
the variance of the observed minus the predicted RT:
(1) RT window = ±t0.995,df x ^Var(RTobs-RTpred)
The above equation is valid provided that the observed RTs
exhibit constant variance and observed minus predicted RTs
exhibit a normal distribution. These assumptions could not be
strictly verified because of the small data sets (typically, n = 6
standards) employed to calculate the variance in each case;
however, moderate departures from normality should still
produce reasonable estimations.
The derivation of the variance of the observed minus the
predicted RT is outlined below for the RT index method.(1)
The variance for the RRT and the external standard methods
can be derived in an analogous fashion.(2,3) The RT index
prediction formula is as follows:
(2)
Where X = sample, S = standard, C = the analyte, and I,
and \2 are the bracketing internal standards.
(3) Algebraic rearrangement of equation 2 leads to:
bs-hs
(4) Introduce the change of variables Z1 and Z2:
Zi =
(5) Substituting (4) into (3): Cx = ZiX * 22X
Z2S
(6) Equation (5) happens to be the linear regression
solution for the linear equation with one calibration
run only: Cx = ZiX + Z2XA
(7) To derive the slope of the regression line (A), the sum
of the square of the errors in the regression (observed
minus predicted values) is minimized by taking the
first derivative and setting this equation equal to
zero. This yields:
NP
A =
NP
Z Z2,2
(8) To derive the variance of the observed minus the
predicted retention time, it is necessary to derive the
variance of "A." This is illustrated in reference 1 and
yields:
Var(Q)
Var(A) =
NP
Z
(9) Var [RTobs-RTpred] = Var(Cr Z1X-Z2X A)
(10) Because Z]x and Z2X are treated as constants:
Var [RTobs-RTpred] = Var (Q) + Z2X2 Var (A)
(11) Substituting equation (8) into (10):
Z2X2
Var [RTobs-RTpred]
NP
Var(Q
(12) From chapter 11, section 11.3 of reference no. 2, the
unbiased estimate of the variance of the observed
(retention time) value about the regression is related
to the sum of the residuals from the fitted regression
by the following equation:
(13)
(14)
Var(Cj)=
1
|