EPA-600/9-83-007
                                                    May 1983
                       PROCEEDINGS:
         NATIONAL SYMPOSIUM ON RECENT ADVANCES IN
POLLUTANT MONITORING OF AMBIENT AIR AND STATIONARY SOURCES
    Proceedings from the National.Symposium held at the
                    Mission Valley Inn
                  Raleigh, North Carolina
                       May 4-7, 1982
        ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
            OFFICE OF RESEARCH AND DEVELOPMENT
           U.S.  ENVIRONMENTAL PROTECTION AGENCY

-------
                      NOTICE'

This document has beert reviewed in accordance with
U.S. Environmental Protection Agency policy and
approved for publication.  Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
                        11

-------
                                 ABSTRACT

     The second  national  symposium to explore recent  developments  that  may
improve the state-of-the-art  for monitoring  techniques was  presented by the
U.S.  Environmental  Protection  Agency,   Environmental  Monitoring  Systems
Laboratory  (EMSL)j  May 4  through  7,  1982,  at  the Mission  Valley Inn  in
Raleigh, North Carolina.

     This  symposium  is part of  a  continuing  effort  to  explore  recent
advances  in pollutant  monitoring  of  ambient  air  and stationary  sources.
Approximately 300 engineers and scientists from industryj  universities,  and
ciontrol agencies attended  the meeting.

     The symposium served  as  a forum for  exchange of  ideas  and information.
The presentations addressed both source emission monitoring and ambient  air
monitoring.   Included were presentations  on gaseous  organics,  particulate
pollutants, and  personal monitoring.   Also presented  were  findings relative
to  sampling . and  analytical  methods  as  well  as  to  a broad spectrum  of
organic chemicals in  Outdoor  and indoor air.

     This  publication is   intended  for those interested  in  air  monitoring
and who  were  unable  to attend the  symposium.    This  report  includes  only
those  papers  submitted  voluntarily  by  speakers.   An  agenda is  included
listing all the  speakers who  participated in the symposium.
                                    iii

-------
                                  CONTENTS
Disclaimer	'  .  .	....'.  ii
Abstract	iii
Figures  	.....  vi
Tables	  .  ....  .xvi

Validation of EPA Reference Method  25—Determination of Total
  Gaseous Nonmethane Organic Emissions as Carbon,'Gary B. Howe,
  Santosh K. Gangwal, R.K.M. Jayanty, Joseph E. Knoll, and M.
  Rodney Midgett	1

Evaluation of 2-Propanol as a Liquid Absorbent for Hazardous     •'  '-'•'-•"•
  Pollutants in Stationary Source Gas Streams, Craig M. Young
  and Larry E. Trejo	•..'..'	  19

Formaldehyde Surface Emission Monitor, T.G. Matthews, A.R.
  Hawthorne, J.M. Schrimsher, M.D.  Corey, and C.R. Daffron	30

Correlation of Remote and Wet Chemical Techniques for the
  Determination of Hydrogen Fluoride Emissions from Gypsum
  Ponds, Howard F. Schiff, Daniel Bause, Mark McCabe, Verne        .  '
  Shortell, William F. Herget, and  Mark Antell	44

Results of the Synthesis and Solid  Sorbent Evaluation of
  Some Porous Copolyamides, Sajal Das, Louis A. Jones, John
  E. Bunch, and James D. Mulik	  81

Synthesis and Evaluation of a Porous Polyetherimide for the,
  Collection of Volatile Organics,  Sajal Das, Louis A. Jones,
  John E. Bunch, and James D. Mulik 	 ............  93

Advanced Concentrator/GC Methods for Trace Organic Analysis,
  S.A. Liebman, T.P. Wampler, and E.J. Levy . .	  .103

Air Analysis by a Nondispersive Infrared Method, Philip Hanst  ....  .117

Sampling Variability and Storage Stability of Volatile Organic
  Contaminants, Harold G. Eaton, Frederick W. Williams, and
  Dennis E.  Smith	  .136

Exposure to Perchloroethylene Associated with the Use of
  Coin-Type Dry Cleaning Machines,  R.H. Jungers and S.J.  Howie	153
                                    IV

-------
Preliminary Results from the Wide Range Aerosol Classifier,
  R.M. Burton, Dale A. Lundgren, Brian J. Hausknecht, and
  David C. Rovell-Rixx	162

Detection of Graphitic Carbon in Collected Particulate Matter,
  W.A. McClenny	184

Status of Sampling and Analysis of Ambient Nitric Acid,
  Nitrates and Ammonia, Robert K. Stevens, Robert W. Shaw, Jr.,
  Robert Braman, and C.W. Spicer	.197

A Simple Design for Automation of the Tungsten VI Oxide
  Technique for Measurement of NH3 and HN03, W.A. McClenny,
  P.C. Gailey, R.S. Braman, and T.J. Shelley.	202

Ozone Precursor Monitor (0PM) for Investigating Air Pollution, ,
  Gordon C. Ortman. ."..,,	.2.08

A New Reliable Ambient Air Chlorine Monitoring System, Eric
  F. Mooney ..................  	  	  .237

The Development of Standard Reference Materials Containing
  Selected Organic Vapors in Compressed Gas Mixtures, W.P.
  Schmidt and H.I/. Rook .	.246
                  '                                                      •*

Human Exposure- to Vapor-Phase Halogenated Hydrocarbons:
  Fixed-Site ^VS_ Personal Exposure, E.D. Pellizzari, T.D.
  Hartwell, C. Leininger, H. Zelon, S. Williams, J.J. Breen,
  and L. Wallace	  .264

A Personnel or Area Dosimeter for Polynuclear Aromatic
  Vapors, T. Vo-Dinh		  .289

The NBS Portable Ambient Aerosol Sampler, R.A. Fletcher,
  D.S. Bright, and R.L. McKenzie.	301

Development of a Prototype Active Personal Monitor  for SOa,
  N02, and Airborne Particles, Tahir R. Khan, Jean  C. Meranger,
  and Belinda Lo.	315
                                                             '.*'-,
Development of SPE Diffusion Head Instrumentation,  J.A.
  Kosek, J.P. Giordano, and A.B. LaConti	333

Laboratory Studies of a Passive Electrochemical  Instrument
  for Measuring Carbon Monoxide, Vincent A. Forlenza.  ........  .358

Results of Testing Diffusion-Type Nitrogen Dioxide  Personal
  Monitors at Low Concentration, James B. Flanagan  and
  Joseph Ryan	  .369
                                     v

-------
                                 FIGURES
Number
Page
Howe, Gangwal, Jayanty, Knoll,  and Midgett

  1  —  Single trap sampling train	    3

  2  -  Modified dual  trap  sampling  train	    4

  3  —  Condensate recovery and conditioning system  .  .  .  .  .  ...  .   '6

  4  -  Byron Model 401'NMOA schematic.  .  .	.'.,-..    7

  5  -  Experimental setup  for  C02 interference effect  study.  .  .  ,  .    8

  6  —  Relative responses  of the NMOA  for several organic compounds.    9

  7  —  Dry gas volume sampled  before plugging versus water vapor
        content	  .........    16

Young and Trejo

  1  —  Bench—scale test system	    21

  2  -  Eight-inch pilot plant  incinerator.  ....  	    22

  3  -  Analysis of 1.6 yg CCl^/ml in 2~propanol.	    25

  4  -  Analysis of 1.3 yg  1,2 - Dichlorobenzene by  HPLC	    25

  5  -  Flow chart of  sample analysis	    27

Matthews, Hawthorne, Schrimsher, Corey, and Daffron

  1  —  Conceptual design of the Formaldehyde Surface Emission
        Monitor 	 ........,,,,....    31

  2  —  Diffusion model for the CH^O concentration at the  surface
        of the test medium  as a function of sorbent-test medium
        separation and the CH20 emission rate of the test medium.  .  .    33

  3  -  Formaldehyde sampling characteristics of passive samplers
        containing water (o) and'molecular sieve (A) sorbents  ....    35
                                    VI

-------
Number
                                                              Page
  4  -
Comparison of test results for the Hardwood Plywood
Manufacturers Association-National Particleboard
Association (HPMA-NPA) dessicatpr and surfape moni-
toring methods	
  5  -
Formaldehyde emission rate of edge-coated wood paneling
samples as a function of ambient CH20 concentration  .  .
        Urea-formaldehyde foam insulation  (UFFI) panel  construction  .

        Comparison of .CH^O emission  rates  from urea-formaldehyde
        foam insulation (UFFI) panels using dynamic  flow  an$  surface
36


37

38
Schiff,
i _
2 -
3 -
4 -
5 -
6 -
7 -
8 -
9 —
10 -
11 -
12 -
Bause, McCabe, Short ell., Herget, and Antell
Sampling points at Agrico Chemical Company's gypsum ponds . .
HF sampling trains utilized in £he program (a) prefilter
and alkali-treated filter (b) spdium bicarbonate-coaled
glass tube, and (e) vacuum system for 'sampling trains t . . .


Concentration calibration curve for the R(5) lipe of HF . . .
Clean air and gypsum pond spectra (average of Agrico
ROSE Runs 32-38) 	 	 , •
Adjusted clean air and gypsum pond absorbance spectra . , . . .
Transmittance of HF line after background subtraction ....
Determination of HF line center absorbance by elimination
of HoQ. interference ....... 	 ..........
47
48
49
52
60
61
63
66
67
69
70
71
  13  -
  14  -
Concentration  of HF  in  air  assayecj  by ROSE  and manual,
methods  at various time periods  ...........
Composite  comparison  of  HF  concentrations  measured by two
techniques.  ........................
                                                                         75
                                                                         76
                                     Vli

-------
Number

Das, Jones, Bunch, and Mulik

  1  —  Polymers A and B - Copolyamides I,  II,  and III	

Das, Jones, Bunch, and Mulik

  1  -  Proton NMR of cis-g-phenoxystyrene	.  .  .  .

Liebman, Warapler, and Levy

  1  —  CDS 320 sample concentrator	............

  2  -  Advances in concentrator technology using thermal desorption.

  3  -  Remote sampling of solvents  in air.	•	  .  .  .

  4  -  Experimental conditions for  remote  sampling of  solvents  in
        air	•	

  5  —  Test air mixture analysis -  Tenax  cartridge	

  6  -  Test air analysis - Tenax cartridge for  diesel  fuel  	

  7  —  Test air mixture analysis -  sorbent comparisons  	

  8  -  Test mixture analysis -Tenax/Ambersorb XE-340 cartridge.  .  .  .

  9  -  Test mixture analysis 	  ........

 10  -  Total organic carbon analysis  	  ...........

Hanst                                                .

  1  -  Nondispersive analyzer in negative  filter configuration  «.  .

  2  -  Spectra and detector signals in nondispersive  analyzer.. .  .  .

  3  —  Emission spectra of ammonia  in nondispersive analyzer  .  .  .  .

  4  -  Emission spectra of sulfur 'dioxide  in nondispersive  analyzer.

  5  -  Detection of S02 by nondispersive  analyzer	

  6  -  Spectra in nondispersive analyzer  tuned  for sulfur dioxide.  .

  7  -  Spectra in nondispersive analyzer  tuned  for nonmethane ;
        hydrocarbons	

  8  —  Detection of butane by nondispersive analyzer  	
 84



 99



104

105

107


108

110

111

112

114

115

116



118

120

125

127

128

129


131

132
                                    viii

-------
Number

  9  -  Spectra of hydrogen chloride gas	

Eaton, Williams, and Smith

  1  -  Sampling flask with critical orifice valve.

  2  -  Sampling manifold with one bottle attached.

  3  -  Gas-handling and analytical systems  .  .  .  ,
Page

 133



 139

 140

 141
  4  -  Subsystem-I. component analysis.  .	   143

 '5  -  Subsystem-II component analysis  .  .  .	  .   144

  6  -  Response averages from high-concentration mixture sampling
        bottles	146
  7  -  Response averages from low-concentration mixture sampling
        bottles  .	.  .  .	
 147
Burton, Lundgren, Hausknecht, and Rovell-Rixx

  1  -  Schematic diagram of the mobile sampling  system  	   165

  2  -  WRAC cumulative mass distributions	170

  3  -  WRAC mass distribution histograms  for Birmingham,  Research
        Triangle Park, and Philadelphia 	   171

  4  -  WRAC mass distribution histograms  for Phoenix  and  Riverside  .   172

  5  -  Normalized mass distributions for  Birmingham and Riverside.  .   173

  6  -  Normalized mass distributions for  Research Triangle Park,
        Philadelphia, and Phoenix 	   174

  7  -  Mass distribution curves for Phoenix and  Riverside	175

  8  -  Comparison of sampled vs computed  (WRAC)  Hi-Vol  mass
        Birmingham, RTP, Philadelphia, Phoenix, and Rubidoux.  ....   179

 ' 9  -  Comparison of sampled vs computed  (WRAC)  SSI-IP  mass
        Birmingham, RTP, Philadelphia, Phoenix, and Rubidoux	   180

 10  -  Comparison of sampled vs computed  (WRAC)  BIGOT-IP  mass
        Birmingham, RTP, Philadelphia, Phoenix, and Rubidoux.  .  .  .  .   181
                                     ix

-------
 Number

 McClenny

   1   -  Schematic of detector cell used in optical measurements
         of particulate samples (a) front view of disassembled
         parts,  (b) back view.	
, Page
   2a -  Calibration curve relating optical absorptance and optical
         absorbance (BL)  to sample weight loading of soot in yg per
         cm2 of filter area	  .  .  .   •	
   2b  -   Calibration curve relating photoacoustic signal to sample
         weight  loading of soot in yg per cm2 of filter area 	

   3   -   Ratio,  R,  of soot estimates by optical absorption and
         photoacoustics versus  total mass loading per filter .  .  .  .  .

   4   -   Schematic  representation of model used in calculating
         light transmitted through and reflected from particle '
         layers  in  a sample	
  5  -  Model  predictions  of  absorbance  (ln(lo/l)),  versus  soot
        loading with mass, M,  of  scattering  particles  as  a  parameter.

  6  -  Model  predictions  of  absorbance,  BL,  versus  experimentally
        measured values  for 18 samples collected  in  Houston,  Texas.  .

Stevens, Shaw, Braman, and Spicer         -

  1  -  Schematic of diffusion denuder nitrate-nitric  acid  sampler.  .

McClenny, Gailey, Braman, and  Shelley

  1  -  Schematic of system designed for  monitoring  HN03  and  NH3.  .  .

  2  -  Schematic of electrical system designed for  automated
        sampling and analysis  using time  delay relays  (TDR)  .  .  .  .  .

  3  -  Three-day monitoring sequence using  the automated system.  .  .

Ortman

  1  —  Schematic diagram of ozone precursor  monitor (0PM)	

  2  -  Irradiation chamber with  reaction vessel.	  .

  3  -  Function of integration vessel.	  .  .

  4  -  GE F40BL lamp spectral output	

  5  -  Fluorescent lamp energy maintenance  .	  .  .
  186



  187


  187


  190



  191


'  193


  194



  200



  203


  204

  206



 211

 214

 216

 217

 218

-------
Number                                                                 Page

  6  -  Fluorescent  lamp mortality	,	219

  7  -  Ozone yield  vs. irradiation  time.	   222

  8  -  Ozone precursors and ozone in  a  small  city.  	   226

  9  -  Ozone precursors and ozone in  a  large  city	228

 .10  -  Analytical agreement of OPM's  on ambient  air	   229

Mooney

  1  -  Cross-section of probe.	239

  2  -  Basic chlorine monitoring system using local control  or
        central processing unit  	   241

  3  -  LCD alarm graphics on LCU	242

  4  -  Probe response to 5 ppm chlorine	244

Schmidt and Rook

  1  -  Effect of 100 cc/min sample  flow rate  on  equilibration
        time of cylinder control valve for VCM.  .	253

  2  -  Effect of 100 cc/min sample  flow rate  on  equilibration
        time of cylinder control valve for CHClg	254

  3  -  Effect of sample flow rate and flow volume  on equilibra-
        tion time of cylinder control valve for benzene  	  .  .   255

  4  -  Effect of 100 cc/min sample -flow, rate  on  equilibration
        time of cylinder control valve for C2Clit.  .	256

  5  -  Effect of sample flow rate.and flow volume  on equilibra-
        tion time of cylinder control valve for chlprobenzene ....   257

  6  -  GC-FID chromatograms of low  concentration mixtures  pre-
        pared by diluting primary standards 	   260

Pellizzari, Hartwell, Leininger, Zelon,  Williams, Breen,  and  Wallace

  1  -  Stratified populations in Baton  Rouge  and Geistnar,  LA .  .  .  .   267

  2  -  Vest equipped with Tenax GC  sampling cartridge,  prefilter
        for particulate and personal pump (in  pocket) for collect-
  •      ing vapor-phase halqcarbons  in personal 'air  .........   268

  3  -  Sampling system depicting filter, Tenax GC  cartridge
        and pump for collecting fixed-site air samples	269
                                    xi

-------
Number                                                                Page

  4  -  Replicate samples for Tetrachloroethylene - Greensboro,
        NC study	273

  5  -  Replicate samples for 1,1,1-Trichloroethane - Greens-
        boro, NC study	.  .   273

  6  -  Replicate samples for Chloroform - Greensboro, NC  study  ...   274

  7  —  Replicate samples for Carbon  tetrachloride - Greensboro,
        NC study	274

  8  —  Replicate samples for 1,2-Dichloroethane - Baton Rouge/
        Geismar, LA study	275

  9  —  Replicate samples for 1,2-Dichloroethane - Greensboro,
        NC study	275

 10  —  Frequency distribution  for 1,1,1-Trichloroethane	 .  .   278

 11  —  Frequency distribution  for Tetrachloroethylene, 	   279

 12  -  Frequency distribution  for Trichloroethylene	280

 13  —  Comparison of percent detected for fixed-site vs.  personal
        air samples - Greensboro, NC	281

 14  —  Comparison of percent detected for fixed-site vs.  personal
        air samples - Baton Rouge/Geismar, LA	  .  .   282,

 15  —  Spearman Correlation for personal vs.  fixed-site air levels
        of carbon tetrachloride - Baton Rouge/Geismar, LA  study  .  .  .   286

 16  —  Spearman Correlation for personal vs.  fixed-site air levels
        of 1,2-dichloroethane - Baton Rouge/Geismar, LA study  ....   286

Vo-Dinh

  1  —  Photograph of the PNA dosimeter worn by a worker at a
        synfuel facility	   291

  2  —  RTP signal of phenanthrene vapor collected by the  dosimeter
        after two-hour exposure 	   294

  3  —  Typical response of the dosimeter to various exposure
        periods of pyrene vapor	,  .   295

  4  —  Room temperature phosphorescence spectra of quinoline  and
        isoquinoline	297

-------
Number
                                                                       Page
  5  -  Differentiation  of  isomeric  quinolines  by synchronous
        excitation.  .  .	298

  6  -  RTF response of  various  dosimeters  exposed at  different
        locations  inside a  synfuel  plant.  	  299

Fletcher, Bright,  and McKenzie

  1  -  Schematic  of the sampler	302

  2  -  Filter collection efficiency as  a  function of  aerodynamic
        diameter.  .  .  '.	  305

  3  -  Schematic  of the inlet	307

  4  -  Inlet cut  curves for  15  ym,  10 vim,  and  7  pm cut-points. .  .  .  308

  5  -  NBS wind tunnel  test  facility	308

  6  -  Horizontal particle concentration  profile in the  wind tunnel.  309

  7  -  Inlet sampling efficiency as  a function of wind velocity
        in the wind tunnel	310

  8  -  Summary of the N02 collection rate  for  the Dupont Pro-Tek
        badge	311

Khan, Meranger, and Lo

  1  -  Schematic  of NO/N02 flow system  for  testing solid sorbent  .  .  317

  2  -  Schematic  of flow system for  generating SOa +  NC>2 +  humid-
        ified conditions  for  the testing of  sorbent  tubes/sampling
        pumps	318

  3  -  Sampling arrangement  at  the Teflon manifold  	  319

  4  -  A typical  chromatogram of blank  (a),  1  ppm standard  (b),
        and duplicate injection  of sample  (c)	   322

  5  -  Photograph of assembled  prototype	   326

Kosek, Giordano, and'LaConti

  1  -  Pumped sensor cell assembly	   335

  2  -  Schematic of SPE Diffusion Head Gas  Sensor	338

  3  -  Remote diffusion head .transducer monitor	339
                                   xiii

-------
Number
Page
  4  -  Transducer and Control Modules Carbon Monoxide Monitor System  340

  5  -  Effect of a porous metal disk on diffusion sensor cell
        response	342

  6  -  Effect of porous metal on diffusion cell response	  343

  7  -  Response vs. concentration, diffusion sensor cell 	  344

  8  -  Application of the SPE Diffusion Cell for the detection
        of high concentrations of CO	345

  9  -  External view, CO diffusion dosimeter ...  	  346

 10  -  Exterior view, CO diffusion dosimeter with interference
        filter removed	  347

 11  -  Flow dependence, CO diffusion dosimeter	348

 12  -  CO diffusion dosimeter linearity data .....  	  350

 13  -  Response of a typical diffusion dosimeter as a  function of
        temperature	•	351

 14  -  Exterior view, NO diffusion dosimeter 	  	  352

 15  -  NO diffusion dosimeter flow studies	353

 16  -  NO diffusion cell response as a function o£  NO  detector
        response	•  •  •  354

 17  -  Temperature dependence of NO diffusion dosimeter response  .  .  355

Forlenza

  1  -  ECOLYZER 210 CO  sensing  system	360

  2  -  Instrument  reading vs. face velocity  at  constant gas
        concentration	362

  3  -  Ambient  readings from the ECOLYZER 210 vs.  face velocity
        at constant CO concentration	363

  4  -  ECOLYZER 210 reading  at  constant CO concentration with and
        without  convection barrier. ............ 	  364

  5  -  Percent  signal vs. time  for ECOLYZER 210-averaged data
        from five  instruments	  365

  6  -  CO vs. LCD  readout  (ppm)	367

                                     xiv

-------
Number




Flanagan and Ryan




  1  -  Calibration of Palmes  tubes analysis  system with  standard
Page
        NaNO,
                                                                        372
  2  -  Ratio of Palmes result/GEL versus dose	374




  3  -  Palmes results versus dosage	>	376
                                    xv..

-------
                                  TABLES
Number
Page
Howe, Gangwall, Jayanty, Knoll, and Midgett

  1  -  Field Test Sampling Conditions	    5

  2  -  NMO Analyzer Linearity Study Results. 	    9

  3  -  Condensate Recovery System Performance Check Results	   10

  4  -  Textile Plant Field Test Results	   11

  5  -  Effect of Leak Check Scheme on the Total Measured PPMC
        for the Plywood Veneer Dryer Field Test Samples 	   13

  6  -  Veneer Dryer Plant Corrected Field Test Results 	   14

  7  -  Effect of Sampling Train Configuration on C02 Interference.  .   15

Matthews, Hawthorne, Schrimsher, Corey, and Daffron

  1  —  Measurements of CH20 Emission Rates from Major Surfaces
        in Three Occupied Homes Using the FSEM	   41

  2  -  Comparison of Measured CH20 Concentration Levels with Values
        Predicted from Combined FSEM and ACH Measurements ......   41

Schiff, Bause, McCabe, Shortell, Herget, and Antell

  1  -  Results of Laboratory Phase	 .....   54

  2  -  Precision and Accuracy—Laboratory Phase	  .   55

  3  -  PPB HF	   56

  4  —  Preliminary Field Phase:  Results and Intersampling Device
        Precision at Same Site	-...*.   57

  5  -  Preliminary Field Phase Comparison of Results for Analysis
        by 1C and Autoanalyzer	;   58

  6  -  Analysis of Citrate Filters for Fluoride.	  .   58

  7  -  HF Concentration Data Grouped in Sequence Obtained	   74

                                    xvi

-------
Number
Page
  8. -  Statistical Analyses Based on Difference Values for
        Manual Sampling Data and ROSE Data.	   78

Das, Jones,  Bunch, and Mulik

  1  -  Breakthrough Volume of Copolyamide at 25°C (Liter/Gram) ...   88

  2  -  Reaction Conditions for the Preparation of Copolyamide-III. .   89

  3  -  Breakthrough Volumes for Copolyamide III Preparations
        at 25°C (Liter/Gram)	   90

  4  -  Copolyamide III - Preparation and Properties	   91

  5  -  Copolyamide III - Breakthrough Volumes (Gm/L) 	   91

Das, Jones,  Bunch, and Mulik

  1  -  Pertinent Information for the Selected Compounds	   95

  2  -  U.S; Production/Pollutant Information 	   95

  3  -  Porous Polymers Arranged in Order of Increasing
        Polarity, Determined According to Walraven	   96

  4  -  Properties of Polyimide Ether 	  100

  5  -  Breakthrough Volumes of Polyetherimide at 25°C (Liters/Gram).  100

Liebman, Wampler, and Levy

  1  -  Test Air Mixture Analysis Tenax Cartridge 40 PPB.  ......  109

Hanst

  1  -  Detection Limit as Function of System Parameters	124

Eaton, Williams, and Smith

  1  -  Contaminants Investigated in the Experiment 	  138

  2  -  The Observed Values of the F-Statistics Obtained from
        the Analysis of Variance	148

  3  -  Amounts and Sources of Variance in a Single Observation
        of Contaminant Concentration (High Levels)	149

  4  -  Amounts and Sources of Variance in a Single Observation
        of Contaminant Concentration (Low Levels)	 .  149
                                   xvil

-------
Number                                                                Page

  5  -  Approximate 95 Percent Confidence Intervals for the True
        Contaminant Concentration in the Test Gas Bottle Based on
        the Observed Average Concentration in JT Flasks with JI
        Analyses for Each Flask (High Concentration Levels) 	  150

  6  —  Approximate 95 Percent Confidence Intervals for the True
        Contaminant Concentration in the Test Gas Bottle-Based on
        the Observed Average Concentration in N^ Flasks with tl
        Analyses for Each Flask (Low Concentration Levels)	151

Jungers and Howie

  1  -  Summary of Ambient PERC Concentrations Inside and Outside
        Laundries (PPB) 	  158

  2  -  Comparison of Indoor PERC Between Laundry D and the Apart-
        ment Upstairs	159

  3  -  Summary of Ambient PERC Concentrations Inside a Closed
        Living Space Where Dry Cleaned Clothing Was Introduced. .  .  .  160

Burton, Lundgren, Hausknecht, and Rovell-Rixx

  1  -  Sampler Vs. WRAC Concentration Expressed as Ratio  	  176

  2  -  Comparison of Measured Vs WRAC Modeled Mass Loading	182

McClenny

  1  —  Summary of Mean Values for Parameters Characterizing
        Particulate Matter Collected on Teflon Substrates  	  189

Stevens, Shaw, Braman, and Spicer

  1  -  Levels of Particulate Nitrate and Nitric Acid 	  200

Ortman

  1  -  Repeatability of Ozone Yield	223

  2  -  Agreement Between OPMs	223

  3  -  Line Voltage Effect on Ozone Yield	224

Mooney

  1  -  Stability of Calibration	245
                                   xviii

-------
Number
                                                                      Page
Schmidt and Rook

  1  -  Toxic Organic Primary Mixtures	247

  2  -  Comparison of Gravitnetrically Calculated and Analyzed
        Concentrations for Primary VCM Mixtures  ....  	  248

  3  -  Comparison of Gravimetrically Calculated and Analyzed
        Concentrations for Primary.Toluene Mixtures 	  249

  4  -  Comparison of Gravimetrically Calculated and Analyzed
        Concentrations for Primary Chlorobenzene Mixtures 	  249

  5  -  Comparison of Gravimetrically Calculated and Analyzed
        Concentrations for VCM Mixtures:  Modified Vs. Original
        Preparative Procedures	250

  6  -  Comparison of Gravimetrically Calculated and Analyzed
        Concentrations for Primary VCM Mixtures Prepared by the
        Successive Dilution Technique and the Microtube Technique .  .  251

  7  -  Multi-Component Mixtures With Total Number of Organic
        Components and Number of Components Showing Significant
        (±2% Rel) Deviation Between Calculated and Analyzed Con-
        centrations:  Modified Vs. Original Preparative Procedures.  .  252

  8  -  Comparison of Gravimetrically Calculated and Analyzed
        Concentrations for Low-Concentration «200 PPB) and High-
        Capacity Cylinder Mixtures	259

  9  -  Calibration of Benzene Permeation Tubes at 25.0°C 	  262

 10  -  Comparison of Nominal and Analyzed Concentrations for
        Toxic Organic SRM's	263

Pellizzari, Hartwell, Leininger, Zelon,  Williams, Breen, and Wallace

  1  -  Geographical Areas and Halocarbons Monitored in Personal
        and Fixed-Site Air	265

  2  -  Number of Samples Obtained by Category for Each Geograph-
        ical Area	270

  3  -  Quality Control/Quality Assurance 	  270

  4  -  Recovery of Halogenated Chemicals From Control Sampling
        Cartridges - Greensboro, NC Study ..... 	  271

  5  -  Recovery of Halogenated Chemicals From Control Sampling
        Cartridges - Baton Rouge/Geismar, LA Study	272

-------
Number

  6  -  Halocarbon Levels (yg/m3) in Air for Greensboro, NC Study:
        Fixed-Site Vs. Personal Samples 	  276

  7  -  Halocarbon Levels (yg/m3) in Air for Baton Rouge/Geismar,
        LA Study:  Fixed-Site Vs. Personal Samples. .... •-.•'.  . .•  .  276

  8  -  Percent Detection in Ambient Air Samples Matched by          -..•.•-
        Participant - Greensboro, NC Study	283

  9  -  Percent Detection in Ambient Air Samples Matched by
        Participant - Baton Rouge/Geismar, LA Study ....... -.-.  . -284

 10  -  Significant Spearman Correlations Between Period No. 1   •
        and Period No. 2 Measurements	285

 11  -  Significant Spearman Correlations for Fixed-Site Vs.
        Personal Air Samples	285

Vo-Dinh

  1  -  Limits of Detection (LOD) for  Several PNA  Compounds by
        Room Temperature Phosphorescence	293

Khan, Meranger,  and Lo

  1  -  Collection Efficiency  of Tea-Silica Gel Sorbent  for N02  •  •  •  323

  2  -  Sorbent  Tube Breakthrough at High Loadings	323

  3  -  Effect of Relative Humidity  (RH) on Collection  Efficiency
        at 1.0 PPM Level	324

  4  -  Life Tests of Candidate  Sampling Pumps	327

  5  -  Replicate Testing of  Sampling  Pumps at  1.0 PPM  Level  and
        75 Percent Relative Humidity (RH)  	  328

  6  -  Results  of Final Testing	330

Kosek, Giordano,  and LaConti

  1  -  Transducer Module Features	341

  2  -  Typical  Response Data, CO Diffusion Dosimeter 	  349

  3  -  CO Diffusion  Dosimeter Linearity Data	350

  4  -  Dosimeter Coulometer  Data	351

  5  -  NO Diffusion  Dosimeters, Coulometer Data	356

                                     xx

-------
Number
                                                                      Page
Forlenza  *




  1  -  Instrument Error Due to the Effect of Face Velocity  	  361




  2•„ -  Interference Equivalents for CO Diffusion Sensor. ......  366




Flanagan and Ryan




  1  -  Results of NaN02 Calibration	371




  •2  -  Palmes Tubes N02 Exposure Results Stationary Exposure  ....  373




  3  -  On-Subject Exposures, in Chamber	  377




  4  -  Exposure of Palmes Tubes in GEL Chamber	378
                                     xxi

-------

-------
       VALIDATION OF EPA REFERENCE METHOD 25—DETERMINATION  OF  TOTAL
              GASEOUS NONMETHANE ORGANIC EMISSIONS AS  CARBON
           Gary B. Howe, Santosh K. Gangwal, and R.K.M. Jayanty

                        Research Triangle Institute
                        Research Triangle Park, NC

                                    and

                   Joseph E. Knoll and M. Rodney Midgett

                   U.S. Environmental Protection Agency
                        Research Triangle Park, NC
                               INTRODUCTION

     On October 5, 1979, under Section  111 of  the Clean Air  Act  as  amended,
the U.S. Environmental  Protection Agency (EPA) proposed standards  limiting
the emissions  of  volatile organic compounds  (VOC)  from new, modified,  and
reconstructed  automobile  and light-duty  surface  coating operations  within
assembly  plants  (1).    The  standards  were  based  on  the  Administrator's
determination  that such emissions  contribute  significantly  to air pollution
by  producing  ozone  and  other  photochemical  oxidants  that  result  in  a
variety of adverse impacts on health and welfare.   The standards included a
description of a  method known  as  EPA Reference Method 25 for measuring  VOC
emissions.  This  method,  adapted  from a procedure  first introduced  by  Salo
et al. (2) has undergone  modifications  since  that date (3) and  is  the  sub-
ject of the present study.

     The essential feature  of  this method is  that  it allows specific  mea-
surement of  the  carbon  contained  in all  types of VOC emissions, including
hydrocarbons and  oxygenates.   However,  when the method was used to  analyze
identical  standard organic  test  mixtures,   a wide  range  of   results  was
obtained by  four  different  laboratories (4).   Another shortcoming  (labeled
as the  "C02  interference effect") was  observed  by researchers at  Midwest
Research Institute  (5),  who reported that  the presence of  high concentra-
tions of  C02 and water  vapor  in  effluent streams  resulted in  erroneously
high values of organics when measured by Method 25.

     The Research Triangle  Institute (RTI) undertook a systematic  evalua-
tion of Method 25 to  identify  its deficiencies and implement modifications
to eliminate  these deficiencies.   This  paper presents the  initial  results

-------
 in an ongoing  investigation of Method 25.  Presented  here  are descriptions
 of  the  experimental apparatus,  the procedures used,  the  results  obtained
 for preliminary  laboratory and  field testing, and  further examination  of
 the "C02 interference effect."
                                EXPERIMENTAL

      The basic  principles of EPA Method 25  are  as follows.   An  emission
 sample is withdrawn from  the source stack through  a chilled  condensate trap
 and  into an  evacuated  sample  tank.    Total gaseous  nonmethane  organics
 (TGNMO) are determined by combining  the  results  obtained from  independent
 analyses of the condensate trap  and sample  tank  fractions.   The  organic
 contents of the condensate trap  are  oxidized and  quantitatively  collected
 in an  evacuated vessel,  and a portion  of  the resulting  C02  is reduced  to
 methane and measured  by a flame  ionization  detector (FID).   A portion  of
 the gas  collected  in  the  sample  tank  is  introduced into  a gas  chromato-
 graphic (GC) system  to achieve separation of the nonmethane organics  from
 CO, C02, and CH^.  After  this  separation,  the nonmethane organics  are  oxi-
 dized to CO2,  reduced to CH^, and measured by an  FID.   in this way, varia-
 tions in FID response  to  different  compounds  are  eliminated and all carbon
 is measured as  methane.

      The basic  equipment  required for Method  25  consists of a VOC  sampling
 system, a  condensate  recovery  and  conditioning  system,  and  a nonmethane
 organic analyzer (NMOA)  capable  of  separating 'fixed gases,  C02, methane,
 and NMO, and of measuring C02 and NMO.   For  preliminary  laboratory  evalua-
 tion,  the equipment assembled was similar  to  that specified in the  Federal
 Register (3).    Two  field tests  of  the  method  were  also  conducted, and
 experiments  were performed to examine the "C02 interference effect."  Brief
 descriptions of the equipment for each phase of the investigation are given
 below with  special  reference  to  features  that represent  modifications  to
 Method  25  as  described  in the  Federal Register  (3).   Unless  otherwise
 specified all parts of the equipment were constructed from 1/8-inch  or  1/4-
 inch, 316 stainless steel tubing or fittings.

 Field Sampling  Trains

     The sampling system (train) used during  preliminary laboratory evalua-
 tion and the first  field test  (carried out with the effluent from a textile
 drying  plant) is shown in Figure  1.   This tr,ain involves  two minor modifi-
 cations  from the  configuration  prescribed in the  Federal Register   (3).
 First,  a bellows seal  shutoff  valve was used  on the sample  tank instead of
 a  quick disconnect/connect fitting,  and second,  the sample tank  used was
 electropolished  on  its  internal  surface  to minimize sample degradation.

     During  the second  field  test (on  the  exhaust  from a  plywood veneer
 drying  plant),  it was  known "a  priori" that  the exhaust  could contain up to
 30  percent  water vapor.   Thus,  in order to avoid  plugging of the  dry ice
 trap, an ice water  trap  was added  to  the sampling  train  upstream of the dry
 ice trap, as shown  in Figure 2.   Three  additional  modifications  were  made
—a  rotameter  was  added  to  the  system  to   facilitate   maintenance  of  a

-------
PROBE
                               TOGGLE
                                VALVE
NEEDLE
 VALVE
                            CONDENSATE
                                TRAP
                                                                VACUUM
                                                                 GAUGE
                                                                  TANK
                                                                SHUTOFF
                                                                  VALVE
           2 LITER
           SAMPLE
             TANK
Figure 1.  Single trap sampling  train.
constant flow rate  during  sampling; a  glass  wool  filter was added on  the
probe to prevent particulates  from  entering the train, and the toggle  valve
(suspected of  leakage  during  the first field  test)  was removed from  the
train.  It is to be noted that this  toggle valve was near the center of  the
first sampling  train (Figure  1)  and was  not  the  same as the one used  for
sample tank shutoff, which had a screw shut  type  stem rather than a toggle
type stem.

Field Sampling Procedure

     The principal objective of the field  tests was to determine the preci-
sion of Method  25  under  actual  field  conditions.   Thus  two identical sam-
pling trains (Train A and Train B)  were used for  each field test, and sam-
ples were collected for the same point in the stack with each train.   Four
sets of duplicate  samples were collected during each field test under con-
ditions of sampling shown in Table  1.   Pretest  and  post  test procedures, as
specified  in  the Federal Register   (3), were  carried out  for  all  samples
except  that the  initial  and final  sample  tank pressures were measured  for
convenience with a Heise compound gauge with 2-mm graduations instead of a
mercury manometer, and overall leak check  procedures were modified for Runs

-------
                                                               VACUUM
                                                                GAUGE
     PROBE
PART1CULATE
   FILTER
                       CONDENSATE
                          TRAPS
                ICE WATER
                                                                      TANK
                                                                    SHUTOFF
                                                                      VALVE
                                                               2 LITER
                                                               SAMPLE
                                                                TANK
Figure  2.  Modified  dual  trap  sampling  train.
3 and 4  in  the  second  field  test.   This  is  discussed  further  in  the Results
and  Discussion  section.   A  detailed  description of the sampling  procedure
is available  elsewhere (6).

Condensate  Recovery and Conditioning  System

     A schematic  of  the  system for recovering the condensate trap contents
by  heat  desorption  and  catalytic  oxidation is  shown  in  Figure 3.   The
system is essentially  similar  to  the  one specified in the Federal Register
(3) except  for  two modifications.   Again, a Heise  compound gauge  (with 2-mm
graduations)  was  used  to measure pressures  instead  of a mercury manometer
and, instead  of two  four-port  valves  for diverting gas  flow to  the oxida-
tion catalyst or  to vent, one six-port Valco zero  volume valve was used.  A
detailed description of the  system  is available elsewhere (6).

Nonmethane  Organic Analyzer

     Model  401, manufactured by Byron Instruments, Inc., Raleigh,  NC,  was
used as the NMOA  for this  study.   Although descriptions  of  previous models
of this instrument have been published (7,8), it has recently been modified
and  its  present  configuration  is  shown  in Figure 4.   The  instrument  is
capable of  measuring total organic  carbon,  CH^,  nonmethane  organics (NMO),
and COg semicontinuously every 12 minutes.  Only the measurement of C02 and

-------
                 TABLE  1.  FIELD  TEST  SAMPLING CONDITIONS
              Sample   Trap(s)
              number    type
                                   Sample  train A
  Flow
  rate
(mL/min)
Sampling
  time
  (min)
Sample train B
Flow     Sampling
rate       time
(mL/min)   (min)
Field test 1     1       D*          35

Textile          2       D           38

Dryer            3D           54

Plant            4      AIBDt        46
              30

              20

              20

              20
              51

              48

              54

              50
   *D = Dry ice.
tAIBD = Ice water for sample  train A, dry  ice  for  sample  train B.
      = Dual trap system; ice water and  dry  ice  traps  in  series.
             30

             20

             20

             20
Single trap
sampling
Field test 2 1 ID$ 25 20 25
Veneer 2 ID 25 20 25
Dryer 3 ID 25 30 25
Plant 4 ID 20 30 40
Dual trap
sampling
20
20
30
30



NMO are required for Method 25.

Field Sample Analysis Procedure

     All field  samples  were analyzed according  to the procedure  described
in the Federal Register  (3), except  for  the  changes  made  below.   Instead of
a separate intermediate  collection tank  for  the  condensed organics from the
dry ice and ice water traps, the matched sample  tank itself  was  used as the
intermediate collection  tank after  undergoing,NMO analysis  and  evacuation.
This repesented a significant improvement  of  the method since  the  necessity
of measuring tank volumes  was  eliminated (a time-consuming  and  error-prone
task), because  the  intermediate  collection tank volume and  the  sample  tank
volume cancelled out in  the calculations (6).

-------
                                 CONDEN-
                                  SATE
                                  TRAP
                      INTERMEDIATE
                       COLLECTION
                         VESSEL

Figure 3.  Condensate  recovery and  conditioning  system.
     The gas  used  to bake out the trap consisted  of  30 cm3/rain, 5  percent
02  in N   added to 35 cm3/min pure  oxygen for samples from the  first  field
test.  Due to  low  flow and high  oxygen  content,  slight  carbonization of  the
sample was suspected.   Thus  for samples  from the  second field  test,  dilu-
tion and total flow  were  increased  with the bake—out gas now  consisting  of
85  cm3/min,  5 percent 02 in N2 and 15  cm3/min pure  oxygen.   A detailed
description of the analytical procedure is  available  elsewhere  (6).

CO2 Interference Effect Study—Apparatus  and  Procedure

     As stated in  the introduction,  erroneously  high  values of  NMO resulted
when Method  25 was  used  for samples containing a large  percent  of  water
vapor and C02«  Two  possible reasons for  this  are  dissolution  of C02 in  the
water condensate  of  the  ice  water  trap  or entrapment  of  CO2  in  the  ice
condensate of  the  dry  ice trap.   The dissolved  or  trapped  C02 is unlikely

-------
STRIPPER
COLUMN


1 — >-
SEPARATION
COLUMN
                              CO/CH4/CO2/C2's
                                                   CO2/C2's-
Figure  4.   Byron  Model 401 NMOA schematic.
to  be  fully released  during the condensate  recovery C02  purge (since  the
trap  is  maintained  in ice  water  or dry  ice  during this  purge to  prevent
organics  from escaping) and would end  up being reported  as NMO  following
bake out  and  measurement.

     Further  examination of  this  effect was attempted  in this -study  using
the experimental  set up shown in Figure 5.  An 11.6 percent  C02  in N2  stan-
dard bottle was  used as the C02 source,  since  this concentration approxi-
mately represents the C02  content of  some  power plant  stacks.   Water  was
pumped by a syringe  pump into a heated manifold  (120°C), where  it  was  mixed
with the  C02  containing gas.   Samples  were collected  from the manifold  at
50 cm3/min  through both the  single  and the dual trap  sampling trains  (the
evacuated sample  tank was  replaced with a vacuum pump  for both trains)  for
a  series  of  water  vapor  concentrations ranging   from  0  to  25  percent.
Further  details  of   these   experiments  are  presented  in  the  Results  and
Discussion section.

-------
        VENT
        JL
               HEATED STAINLESS STEEL
               'MANIFOLD (120° o
 5 mm
GLASS
BEADS
                                              METHOD 25
                                              SAMPLING TRAIN
                                                                 10 cm3 SYRINGE
                                                   SYRINGE PUMP
   MASS
   FLOW
CONTROLLER
 Figure  5.  Experimental setup for C02 interference effect study.


                            RESULTS AND DISCUSSION

      A  preliminary evaluation of the equipment used was completed according
 to  specifications of the  Federal  Register (3).  A  detailed description of
 the preliminary evaluation is available  (6).   In the  interest of brevity,
 only the highlights  are reported here,  followed by  the results  from the
 field tests  and the C02 interference effect study.

 Preliminary  Evaluation

      The linearity and  compound dependence of the  NMOA  response were mea-
 sured  for a  variety of gases  and organic vapors.   The  instrument  showed
 excellent linearity for several compounds  subjected to the linearity  study
 over wide concentration ranges, as  shown in Table  2.   The compound depen-
 dence results are shown in Figure 6.  The response  variability to different
 organics is  expected to result in inaccuracies for  measurement of the  vola-
 tile portion of  the  sample,  which is  not  captured by  the  condensate trap
 and ends  up in  the  sample  tank.    The  variability of  response, however,
 should  have  no  effect  on the precision  of results from samples collected
 via the dual sampling train if all other parts of the sampling and analysis
                                       8

-------
 system are performing properly.
               TABLE 2.  NMO ANALYZER LINEARITY  STUDY RESULTS
Compound
Propane
Benzene
Ethylene
Propylene
Methyl acetate
Ethane
CO,
Concentration range
59.7
49.2
9.5
14.8
52.5
590.0
57.3
- 9,150
- 2,090
- 40,912
- 2,050
- 1,350
- 1,430
- 10,000
Correlation
coefficient*
0.999
0.999
0.999
0.999
0.999
0.962
0.999
 *Based  on linear least squares fit.
   2.0
RELATIVE RESPONSE
p =» _i
o bi b w
-
DECANE
CARBON TETRACHLORIDE

NONANE

TOLUENE

AMYL ACETATE

NAPHTHALENE

CO
t—
O
UJ
oc
u_
BENZENE

ETHANE

TRICHLOROETHYLENE

ETHYLENE

PROPANE

PROPYLENE

TETRA-
HYDROPYRAN
TETRA-
HYDROFURAN
HEXANE
ACETYL
ACETONE

ISOPROPYL
ALCOHOL

r-i
METHANI
METHYL
ACETATE
                                 COMPOUND
Figure 6.
Relative responses of the NMOA for several  organic  compounds.

-------
     The observed  response variability could  be  due to the  differences  in
the oxidation efficiency  of  the oxidation catalyst  of  the NMOA for differ-
ent organics.   This was  investigated further using a propane  (59.7  ppmC)
and a  toluene  (128  ppmC) standard.   An oxidation  catalyst  identical  in
design to  the one  used internally in  the NMOA was used externally to oxi-
dize the standard  propane and toluene gases.   Each gas was  flowed through
the catalyst  at 150 cm3/min into  a sampling manifold from which samples
were drawn into NMOA  for analysis.   Both  CO and  C02  resulted  from  the
oxidation,  and  residual  nonmethane organics  were  also  observed  to  break
through the catalyst with either propane or toluene.   The  most significant
difference  in the  oxidation  products  was an  18  percent difference  in  the
C02 formed  per  ppmC of  organic.   Although the  higher production of CO2 from
toluene as  opposed to propane  correlated with its  higher  response (Figure
6), it was  hardly  enough  to  account  for  the  actual response difference that
existed between the two  gases.  Furthermore,  since the actual continuous
flow conditions of  the experiments  were not  representative  of  the  pulse
flow conditions within the NMOA,  no definitive conclusions  could be drawn
from these  experiments.   Further studies  are needed to pinpoint the reasons
for the response variability of the NMOA.  However, as pointed out earlier,
the response  differences  affect the accuracies  but should  not  affect  the
precision.   For example,  evaluation  of  the field  sampling trains  in  the
laboratory  showed  a 99 percent  recovery and a relative  standard deviation
of 0.08 percent in repetitive sampling of a  603 ppmC propane standard.

     The  condensate  recovery  and  conditioning   system was   evaluated  by
making  liquid  injections of  hexane  and  toluene  at  its  gas  inlet  via  a
heated  injection  port.    These  recoveries   are  shown  in Table 3  and were
acceptable  for  the 9 yL injections.  For the  100  yL injections, the recov-
eries were still  better  than 80 percent.   The lower  recovery is probably
due  to  a  combination  of factors,  including  incomplete  recovery  and  the
slight  nonlinearity of the NMOA for the high C02  concentrations  generated
(3.2 percent using n-hexane  and 4.5 percent  using toluene).
     TABLE 3.   CONDENSATE RECOVERY SYSTEM PERFORMANCE CHECK RESULTS
                               Injections
 Compound
                               Percent
                               recovery
 n-Hexane

 n-Hexane

 Toluene

 Toluene
100.Q

  9.0

100.0

  9.0
82.8

96.4

81.1

94.0
                                     10

-------
 Field  Test Results

     The conditions for field sampling were  shown earlier in Table 1.   The
 flow rate values shown for the first  field test are actually average  esti-
 mates  calculated from  the  sample  tank initial  and  final pressure, and  the
 tank volume since  a rotameter for  visual  observation and  control  of  the
 flow rate was  not  used (Figure 1).  For the second field  test, however,  the
 flow rates were calibrated on a rotameter  (Figure 2).  The results from  the
 first  field test are  shown in Table  4  along with  standard  deviations  for
 dual train sampling of each run.
               TABLE 4.   TEXTILE PLANT FIELD TEST RESULTS



Run
number

1


2


3


4



Sample
train
A

B
A

B
A

B
A

B

Noncondensable
organics
(ppmC)
8.7

7.7
<3.0

3.5
4.3

<3.8
160.0

55.6

Condensable
organics
(ppmC)
627

320
244

193
642

360
579

703


Total
(ppmC)
636

328
244

196
646

360
738

759
Relative
standard
deviation*
(percent)

45


15


40


2


inn
*Relative standard deviation
x = Total ppmC.
x = Mean for each run.
n = 2.
     Another way  to  look  at  the  precision of the data is  on  the  basis of a
pooled relative standard  deviation,  Sp,  shown in Equation (1)
100v/Zcr/2n  /x
                                                                        (1)
where  d  is the ^difference  between the  paired values,  n is the  number  of
runs (=4), and  x is  the overall mean  value.   The pooled  relative standard
                                     11

-------
deviation for the  four  runs  in Table 4 is  31 percent.  Two factors probably
contributed most  to  this  imprecision.  First,  it  was discovered during the
third  run  that  the toggle valve  was leaking.   This  would normally  not be
discovered  during  the leak  check  procedure,  because  the   probe   end  is
plugged.  However,  in the process of properly  inserting  the  dual probes in
the stack,  leakage would be  allowed into the  train.  Another  factor that
was even more critical  was that constant  flow rate over the entire duration
of  the run could  not be  maintained because of  an increase   in  the  sample
tank pressures during the run.   The  effect of this would be higher when the
stack  concentration is highly variable,   as  seen in Table 4.  Run  4 gave
good precision after  the  toggle valve problem  was corrected.   It  is to be
further noted  that an  ice water  trap and a dry ice trap were  used  during
Run 4  for sample  Trains  A and B,  respectively (Table  1).   This  did not
result in a  reduced precision for the total ppmC  value,  although obviously
the ice water  trap allowed  more  organics to end  up in the  sample  tank as
noncondensables.

     The problems  observed during the  first field test  were corrected for
the second  field  test,  which was  carried  out using  a  plywood veneer dryer
plant  exhaust.  Also,  as  noted in the Experimental section, due to the high
water  vapor concentrations of this exhaust,  an ice water trap and a  dry ice
trap in series were used  to  prevent  plugging (Figure 2 and Table 1).

     Again, four  runs with dual samples were carried out, but this time the
Federal Register  (3)  leak test specifications were not precisely adhered to
for Runs  3  and 4.   The Federal Register  specifies a  pretest as  well as a
post  test  leak  check  for  the  entire  sampling  train  under  sample  tank
vacuum.  Thus  air contained  in the  sampling train is  allowed to enter the
sample tank during the  pretest leak check with  the probe  end plugged.  The
plug has to be removed following the leak check before sampling can'begin.
Air thereby enters the  train a second time and is again drawn into the sam-
ple tank during sampling.  Thus two  complete dead volumes effectively enter
the  tank  before  sampling.   During  the   post  test  leak  check,  the train
supposedly  contains  the last stationary portion of the sample.  This should
not affect  the precision as  long as  the mixing  with ambient  air  of this
stationary  sample is  identical for  both  trains  during  the   time  taken to
remove the  probes  and plug them for  the post test leak check.

     The leak checks  are  not representative  of the actual conditions exist-
ing during  sampling since only a small portion of the  train from the needle
valve  exit  to the  sample  tank (Figure 2)  is  under the  influence of the high
tank vacuum, the  rest of  the train being  at  close to ambient   pressures.  In
order  to evaluate  the effect of leak checks, the- leak  check procedures were
modified and varied,  as shown  in  Table 5, for Runs 3  and  4.   A correction
was applied to the  'total measured  ppmC  NMO on the basis of ' the measured
dead volume of  the sample train  (0,86 cm3).   The  actual  sampled volume was
thus reduced by  86 x 2 =  172 cm3  (which  represents  the  dead volume effect
of  the pretest leak check) for Runs  1 through 3.  The  actual  sampled volume
was reduced by  86 cm3  only  for Run 4, where a pretest  leak check was not
performed  on the  entire  train.   (This would  then result in  only one dead
volume entering  the  train,  as discussed  previously.)   The corrected total
NMOC  was  then  calculated by  dividing  the  measured  NMOC  with  the ratio
                                      12

-------
 of the corrected volume to the uncorrected  volume.
      TABLE 5.  EFFECT OF LEAK CHECK SCHEME ON  THE  TOTAL  MEASURED PPMC
                FOR THE PLYWOOD VENEER DRYER FIELD  TEST SAMPLES


Leak
Run check
number scheme













1 t


2 t


3 $


4 §

Uncorrected values
Sample
train
A

B
A

B
A

B
A

B
V*
(cm3)
672

672
672

672
836

836
600

1,200
Total
(ppmC)
4,820

3,900
3,490

3,860
4,920

4,710
5,550

6,200
Corrected values
y*
(cnf3)
500

500
500

500
664

664
514

1,110
Total
(ppmC)
6,470

5,230
4,680

5,190
6,200

5,920
6,480

6,670

*Vq = Volume sampled.
t
$

§
= Pretest and
= Pretest leak
valve .
= Pretest and
posttest leak
checks of
check of entire train,

posttest leak

checkup to
entire train.
posttest leak

needle valve.

checkup to



needle


     The  dead  volume correction would result  in  the same factor for Trains
A  and  B  for Runs  1 through  3,  where equal  volumes were  sampled  into the
train.   Different factors  would  result  when  different  volumes are sampled
as  for Run 4.    Improvement in precision  of  the  results  from the modified
leak check procedure is  reflected  in the  standard  deviation  for  Run  4 in
Table  6,  where data are also presented  for the portion  of  the NMOC parti-
tioned into the  sample  tank,  the  dry  ice trap, and  the ice  water trap.
Notice that the precision  of  the data  is  improved  considerably  for  Run 4
after  the  leak  check  correction  is  applied  (without  the  correction,  the
relative  standard  deviation is  7.8 percent, as opposed  to 2.0 percent with
the correction).   Also,  the method is seen to be rugged  enough that a flow
rate variation of a factor of 2  between  the  two   trains  in Run  4 caused
little  effect  on  the precision  of  the data.  Furthermore, as expected, at a
higher  flow rate,  a  larger  carryover of  the organics into the dry ice trap
is observed for Run 4.

     The  dead  volume would  have a  stronger influence on the accuracy of the
results  for smaller  volumes  as   collected in  this study.    The  Federal
Register  (3) specifies a  4  to 6-liter sample  tank as opposed to the 2-liter
                                    13

-------
       TABLE 6.  VENEER DRYER PLANT CORRECTED FIELD  TEST  RESULTS



Run
number

1


2


3


4



Samp le
train
A

B
A

B
A

B
A

B
Noncon-
densable
organics
(ppmC)
352

612
151

444
887

677
808

529
Condensable organics
(ppmC)
Ice water
trap
2,370

1,300
842

1,020
1,160

1,030
1,170

894
Dry ice
trap
3,750

3,320
3,690

3,730
4,150

4,210
4,500

5,250
Total
(ppmC)
6,470

5,230
4,680

5,190
6,200

5,920
6,480

6,670
Relative
standard
deviation*

15.0


7.3


3.3


2.0


inn .... ... — 	 _ 	
*Relative standard deviation
x s Total ppmC.
x - Mean for each run.
n - 2.
xv/E(x-x)2/(n-l),
sample tank  used  here for convenience.  Assuming  a 2 to 3—liter  sample  is
actually collected  in practice using Method 25 procedures,  the  dead volume
error would  be  approximately 6 to 8 percent with  a dual trap system (^172
era3  dead volume),  and 3 to  4 percent with  a single trap  system (0.86  cm3
dead volume).

     The pooled standard  deviation calculated  as  before  by Equation (1) for
the data in  Table 6 was found to  be 8.3 percent,  as opposed  to  31 percent
calculated for  the first field  test.   This is  evidence that the modified
procedures  (including  precise  flow  control  with  a  rotameter,  modified
sampling  train,   the  dead volume  correction,  and  the  modified  leak check
procedure)  resulted  in significant  improvement   of  the  precision  of  the
method.

COa Interference  Effect

     Trap  samples  were collected  using  the  experimental setup  described
earlier  (Figure 5) with  both the single  dry  ice  trap  system and the dual
ice water and dry ice trap  system.  These traps were then subjected to the
Method 25 procedure to determine the amount of C02  that  would end up being
measured as  NMO because of  incomplete recovery of  the trapped or dissolved
C0£  during the C02 purge.    These  results  are shown in  Table 7.   The C02

                                     14

-------
 TABLE 7.  EFFECT OF SAMPLING  TRAIN CONFIGURATION ON C02 INTERFERENCE
Single dry ice
trap sampling
Mole
percent
water
vapor
10
15
20
25

Sampling
time
(min)
60
46
34
23
CO 2
inter-
ference*
(ppmC)
20.5
17.6
49.3
16.3
Dual dry ice and ice water trap sampling

Sampling
time
(min)
60
60
60
60
CO 2
Ice water
trap
(ppmC)
1.4
30.4
26.9
22.6
interference*
Dry ice
trap
(ppmC)
0.0
8.0
2.0
,0.0

Total
(ppmC)
1.4
38.4
28.9
22.6

*Corrected for background.
NOTE:  Sampling rate =  50  cm3/min dry basis;  sample gas 11.6% C02 in N2 dry
       basis.
interference ranges  from  16  to  50 ppmC for single trap sampling and 1 to 39
ppm for  dual trap sampling.   These values  do  not represent  a significant
interference in  sampling  sources- containing several  hundred  ppmC or higher
concentrations  of NMO.    However,  they  would be  significant  in sampling
sources containing NMO  concentrations  of a few hundred  ppmC  or lower.   The
extent  of  C02 interference  appears to  peak out  and then decrease  as  the
water  vapor concentration  is  increased  with  the  dry gas  concentration
fixed.   Although no  explanation can  be  offered for  this  behavior,  it is
presently being  further investigated.   A mathematical simulation of the COa
condensation  phenomena using  transient mass  balances  and  Henry's  law is
being  attempted  to  predict  the amount of  C02  that.  will be  trapped.   The
simulation  is complicated by the  fact  that,  during the C02  purge for Method
25 analysis, portions  of  the trapped C02 could be  released  and carried out
with the inert purge.

     The most significant conclusion that can be drawn  from  Table 7 is the
clear  superiority of  using  a  dual trap sampling  train as  opposed  to  a
single  dry  ice  trap sampling  train in the sampling  of  high  C02  and water
vapor-containing  samples.   It  is  to  be  noted  that   a  forced  shutdown of
sampling resulted each time  with the  single  trap  system because  of plug-
ging.   The  dry gas  volume that caused plugging  is  shown as  a function of
the water vapor  content  of  the  sample  gas  in Figure 7.   No plugging was
observed for the  dual  trap system even after sampling for 60 minutes at 50
cm3/min dry gas  with  25 percent  water  vapor concentration.  Also,  even
though  the  water admitted into the dual  trap system at 20  percent water
vapor concentration  was almost  twice as  much as  the single  trap system, the
measured CO2 interference was less  than three fifths.
                                     15

-------
  3,000 r
CD
H 2,250
CD
CD

1
UJ
cc
   1.500
CQ
Q
3
UJ
O
    750
                               _L
_L
                               10          15

                                PERCENT MOISTURE
            20
25
Figure 7.   Dry gas  volume  sampled before plugging  versus water vapor  con-
            tent.
     Most of the C02 interference for dual trap sampling  ends  up  in  the  ice
water  trap,  which is expected to trap  the  water as a liquid rather  than as
a  solid.   This may  be  one  explanation of the  superiority of  the dual trap
system, because it is probably easier to strip the COa  out  of  liquid water
during the  purge  than it  is  to purge  it  out while  it  is  trapped  in  ice
formed within the dry ice trap.   Further research  into  the  "C02 inteference
effect" is  continuing.
                       CONCLUSIONS AND RECOMMENDATIONS

     With careful  experimentation,  good  precision  can  be  obtained  for
source  measurement of  NMO using Method 25.   The accuracy  of the  results
obtained,  on the  other hand,  depends  on  such additional  factors  as  dead
volume  of  the  sampling  train  and  the  relative  response  of  the  NMOA  to
different organics.    Several  modifications  were made  to  the  procedures
specified in the Federal Register  (3),  and these modifications in  combina-
tion were shown to improve the overall  precision  of  the  method.    The  best
                                     16

-------
relative standard deviation obtained was 2.0 percent  for  field  sampling and
0.08 percent for laboratory sampling.
     The  "COa   interference effect"  appears  to  be  minimal  when  sampling
stacks containing several hundred ppmc  of NMO,  but  it is  necessary  to  use a
dual trap sampling system to prevent  plugging  during  the  sampling of  stacks
containing high concentrations of water vapor.

     Additional studies are required  to make theoretical .predictions  of  the
C0£  interference  effect,   for  the  design  of   an  NMOA  that  would produce
nearly equal  responses to  all  organics and  on the  behavior of thermally
labile compounds during condensate  trap bake out.   Further studies are  on-
going at RTI on  various  aspects  of  Method 25  with  a  view to improving  its
accuracy.
                              ACKNOWLEDGMENTS

     This work at the Research Triangle Institute  was  supported  by U.S.  EPA
Contract No. 68-02-3431.
                                REFERENCES

1.   Federal Register 44:57792-57822  (1979).

2.   Salo, A.E., W.L. Oaks, and R.D.  MacPhee.   1975.   Measuring  the  organic
     carbon  content  of  source  emission for air  pollution control.    JAPCA
     25:390.

3.   Federal Register 45:65956-65973  (1980).

4.   Midwest Research Institute..  1981.   Evaluation  of  Method  25 for  air
     oxidation  processes—volume  I.   EPA  Contract  No. 68-02-2814,  Assign-
     ment 36.   Emission Measurement  Branch,  U.S. Environmental  Protection
     Agency, Research Triangle Park,  NC.

5.   Midwest Research  Institute.    1981.   Investigation  of carbon  dioxide
     interference with  Method  25.   EPA Contract No. 68-02-2814,  Assignment
     41.  Emission Measurement Branch,  U.S Environmental  Protection  Agency,
     Research Triangle  Park, NC.

6.   Howe, G.B., S.K. Gangwal,  and R.K.M. Jayanty.   1982.  Validation  and
     improvement  of EPA  Method  25—Determination  of  gaseous   nonmethane
     organic  emissions  as  carbon:    manual  sampling  and  analysis   proce-
     dure—Draft report.  EPA Contract  No. 68-02-2341.

7.   Hilborn, J.C., and N.  Quickers.   1975.   Evaluation  of the  Byron Model
     233A air quality analyzer.  Chemical Instrumentation 6:75.

8.   Scott Environmental  Technology,  Inc.   1978.   Evaluation of  the Byron
     Model 401  Hydrocarbon  Analyzer.    EPA  Contract  No.   68-02-2813.    Emis-
                                     17

-------
sion  Measurement   Branch,
Research Triangle Park, NC.
U.S.  Environmental   Protection  Agency,
                               18

-------
            EVALUATION OF 2-PROPANOL AS A LIQUID ABSORBENT FOR
           HAZARDOUS POLLUTANTS IN STATIONARY SOURCE GAS STREAMS
                     Craig M. Young and Larry E. Trejo

                            Energy Incorporated
                              Idaho Falls, ID
                               INTRODUCTION
General Description

     Energy  Incorporated is  an engineering  firm working  in  the  area  of
fluidized  bed  combustion systems  and,  with  its  manufacturing  subsidiary,
Energy Products of Idaho, has 40 commercial fluidized  bed systems  currently
in  operation.    Energy  Incorporated  (El) has  recently devoted its  pilot
plant facility to the  investigation of  the destruction of hazardous  wastes.
Most  studies have  been  conducted with  carbon  tetrachloride  (CCl^)  and
1,2-dichlorobenzene (DCB) as  representatives  of chlorinated wastes.

     El needed a method  for analyzing off-gas samples  from the  pilot  plant
tests that was simple, reliable,  and  would give rapid  turnaround  on sample
analyses.  For these  reasons, liquid  absorption systems were  investigated.
From the work  done  in this  study,  it was  concluded  that 2-propanol was  an
acceptable liquid absorber.

Background

     Both  liquid and  solid absorbers have been  used  for sampling  off-gas
streams.  Xylene has  been used  by  Continental Can as  a liquid absorber for
PCBs  in combustion gas  in  boilers fueled  with PCB  contaminated oil  (1).
Ferguson  and others  used  a  sampling train  with  a  glass  probe and  three
midget  impingers to  sample  for  pesticides in the combustion  gases  from the
burning  of pesticides.  They used benzene, isooctane,  2-propanol,  or  water
as  the  liquid  absorber,  depending  on  the pesticide  under  study  (2).
Ethylene glycol  was  used to sample for DDT  and PCBs  at General Electric's
Pittsfield,  Massachusetts,  liquid  injection  incinerator (3).   Guilford and
Brandon  (4)  also used ethylene glycol to sample  for  PCBs in  a  pilot  plant
incinerator.  They were  able  to achieve an  overall absorption efficiency of
99.64 percent  with  two  impingers.   Guilford  and  Rosenblatt  (5)  reported a
97.7  percent efficiency with  ethylene glycol when   sampling  for PCBs  in
combustion   gases.      Komaniya   and  Morisaki   (6)    reported  using   a
three-impinger  system  to  sample  for PCBs.   Water  was contained  in the

                                      19

-------
 first irapinger, followed  by hexane in the  other two impingers.   Shoemaker
 (7) has reported  the  use  of a solution of  5 mg/ml  aniline in  isooctane  to
 sample for bis(cyclohexyl isocyanate) in ambient air.

      Probably the most  widely accepted method  at  present for  sampling  for
 organics in off-gas streams is the use of solid  absorbers  in  a  modified  EPA
 Method 5 sampling train.  Several different solid absorbers have  been used.
 Ackerman  (8)  reported  using  XAD-2  for  absorbing  PCBs  in  a  water-cooled
 sorbent trap.   The  gas  sampled was  the  combustion gas  from  a rotary kiln
 incinerator fed with PCB contaminated capacitors.  MacDonald  and  others  (9)
 reported using  a  modified Method 5  sampling  train with Chromosorb 102 as
 the solid absorber for sampling the  off-gas of  a cement kiln burning PCBs.
 Haile and  Baladi (10)  evaluated Florisil, XAD-2,  and Tenax-GC as solid
 sorbents for sampling off-gas streams  for  organics.   They determined their
 optimum configuration to  be  two  impingers   containing  water,   an empty
 impinger,   the  solid  sorbent, an  impinger containing  10 percent  sodium
 hydroxide,  and finally,  an impinger containing silica gel.  The U.S.EPA has
 recommended in  their  interim  sampling and  analysis  manual to use Florisil
 for sampling for PCBs and XAD-2 for  other  chlorinated organics and  general
 organics  (11).    Bombaugh  and  Lee  (12)  used  Tenax  GC cartridges  when
 sampling ambient air around synfuel plants.   The Source Assessment Sampling
 System has  been developed  to sample for  organics in  gas streams.   This
 system uses cartridges of  Florisil,  XAD-2,  or Tenax GC (13,14).
                               SYSTEMS STUDIES
 Testing  Equipment
      Energy  Incorporated carried out its hazardous waste  destruction test-
ing  in  two stages.   The first stage was bench-scale studies in an apparatus
shown schematically in  Figure 1.   For the CCl^  studies,  the  water vapor
source  was  a  steam generator.   In the  DCB studies,  the  water  vapor  was
supplied  by  bubbling air  through heated water.   The organic  vapor gener-
ator,  in  both cases,  was  a  three-neck flask that was  held at  a constant
temperature  and  air bubbled  through the liquid.   The  packed bed varied from
inert particles  to  catalytic particles.  The number  of  impingers and their
contents were  varied according to the  goals of  the test  being run.

     The  second  testing  facility,  a  fluidized  bed  pilot  plant,  is  shown
schematically  in Figure  2.    Sampling  of  the  gas stream  was   carried  out
before  and after the spray tower  assembly.

Liquid  Absorbers Studied

     To assure a timely pilot plant development,  it  -would be  necessary to
analyze the  off-gas stream several  times  per day, since combustion condi-
tions were varied in an effort to optimize destruction  conditions.   There-
fore, a large  number of samples would need to be  analyzed  in  a  short  time
to supply the  data  necessary  to  set conditions  for subsequent  pilot  plant
runs.   This  restriction eliminated the use of  solid  absorbers,  with their
extensive clean-up  requirements.
                                     20

-------
                                                      PACKED BED
                                                    ORGANIC
                                                     VAPOR
                                                   GENERATOR
Figure 1.  Bench-scale  test  system.
     The  use of  liquid absorbers  was  also  limited  to  solvents  that  are
soluble  in  water.   Water  vapor would  be present  in the off— gas  from  the
pilot plant  facility and also in the  gas stream of  the  bench-scale assem-
bly.   When  the  destruction of CCl^  was  studied,  steam would  be  injected
directly  into  the system.    These  large amounts  of water would  form emul-
sions or  two phases  unless a waste-soluble  solvent was  used.   These emul-
sions or  two-phase systems would interfere with  the  direct  analyses of  the
impinger  solution.   Direct analysis of  the  impinger solutions was the major
objective of the  sampling  studies.

     After  these evaluations,  ethylene  glycol  was  chosen  for use  in  the
impingers.   Figure 1  shows a schematic  drawing of the bench— scale apparatus
that  was used  in the   first  studies  with CCl^  destruction.   The  first
impinger was filled  with a sodium hydroxide solution and the next  two were
filled with  ethylene glycol.   No  CCl^ was found  in the three  impingers.
The CClif  was found  in  the knock-out flask ahead  of  the vacuum pump.   The
first  test   proved  ethylene  glycol to be  an  unsatisfactory  absorber  for
carbon tetrachloride.

     At  this point,  other  alcohols  were considered.   Methanol proved to be
too volatile, and it is also toxic.  Ethanol,  butanol,  and  higher alcohols
were too  expensive and  were not  readily available.   2-propanol was found to
be inexpensive and readily available in a pure form.   Amyl acetate was also
tested,  but  its  low  solubility  for water  and  unpleasant odor  resulted in
its removal  from  consideration.
Absorption of
     The  test results  obtained from  the bench-scale  studies  with carbon
tetrachloride  showed  a chloride  closure  of 96  percent  on  the  material
balance.   This  was  obtained  by  analysis  of  the  chloride  in the  first
                                    21

-------
                                                               a
                                                              •H
                                                               O
                                                               a
                                                              4J

                                                               O
                                                              •s
                                                              ft
                                                               too
                                                              
-------
 impinger and the CCl^ concentration in the final impingers.

      When the tests were conducted  in  the pilot plant facilities,  as  shown
 in Figure 2, three impingers  in  an  ice-water bath  were  used to sample  for
 CCl^.   The samples were pulled from the  off-gas duct through a  sample  port
 located ahead of  the spray  tower.   When  the  impingers  were  analyzed  for
 CCl^,  29.3 percent of  the  CCl^ recovered  was  in the  first impinger,  46.8
 percent was  in  the  second  impinger,  and 23.9 percent  was in the  third
 impinger.   The high concentration of CCl^  in the  third impinger led to  the
 conclusion that  all of  the  CCl^ was  not trapped.

      The next experimental  run used  four impingers, with  the first  impinger
 cooled  in a dry  ice/acetone  bath.   The remaining  impingers were cooled  in
 an ice-water bath.   The  analysis of these impingers showed:  67.2 percent
 trapped in the first  impinger, 23.3  percent trapped in the  second impinger,
 7.3  percent trapped in the third  impinger,  and 2.2  percent trapped in  the
 fourth  impinger.   The CCl^  concentration  decreased  by a factor  of  three  in
 each impinger; it  was projected that more than 99 percent  of  the CCl^ was
 being  collected  in the off-gas.   These numbers  were the  average  of  three
 sets of  off-gas  samples.    This  test  was  run at  a  low  temperature  to
 minimize destruction  of  CCl^.   The off-gas samples accounted for 94 percent
 of  the  CCli,. fed  into  the  system.

 Absorption of DCS

     When  the bench-scale  tests were run  on DCB,  the air was blown through
 the  system and not pulled with a  vacuum  pump.   The  second  change  was  that
 saturated  water  vapor was used instead of steam.   The tube furnace was run
 at  a  low  temperature  as  an attempt  at  recovering all the  DCB  in the
 impingers.    Four  impingers  were used,  and  they  were   all  cooled  in  an
 ice-water  bath.  This resulted in an average  of 97 percent of the DCB being
 recovered.   Of the DCB  recovered, 98.8  percent was  in  the  first impinger,
 1.2  percent in  the second  impinger, and no DCB was found  in the  third  or
 fourth  impingers.

     The  pilot  plant off-gas samples  for DCB  were  taken  by  using  three
 impingers  cooled in an  ice-water  bath.   The  first  two  impingers contained
 2-propanol.   The  third  impinger  contained a sodium hydroxide  solution  to
 prevent  any HC1 from  going  through to  the  vacuum pump.  The recovery of DCB
 was  92  percent in  the first impinger and  8 percent  in the second impinger.

 Absorption of HC1

     The  absorption  of  HC1   by   2-propanol  was  studied  along with  the
 destruction of DCB  on the  bench-scale  experiments.   The  recovery of HC1 in
 the  impingers was  97.5  percent in the  first impinger,  2.3 percent  in the
 second  impinger,  and  0.2 percent  in  the  third impinger.

     The DCB  off-gas  samples   from the  pilot  plant  were  also analyzed for
HC1.  The  recovery of HC1  in these  samples was 98.3 percent in the  first
 impinger,  1.7 percent in the  second  impinger, and  <0.1 percent  in the third
impinger.
                                     23

-------
Absorption of Phenols

     Two pilot  plant runs were  made  with mixed  phenol  feed.   The  mixture
was  50  percent phenol,  25  percent pentachlorophenol,  and 25 percent  fuel
oil.  There  were very  small amounts  of  phenol and  pentachlorophenol  col-
lected  in the  impingers.   Therefore,  a  phenol mixture  was  run  in  the
bench-scale  apparatus  to determine  the  collection efficiency  for  these
phenols.  Approximately 0.1  gram of  each  compound was  placed in  a  combus-
tion boat and placed in the  tube ahead of the  packed bed.   Three  impingers
containing 2-propanol and one  containing  sodium hydroxide were used to col-
lect the samples.

     The   compounds  used   were  phenol,   pentachlorophenol   (PCP),   and
0-,0'-diphenol.   It  was found  that  100 percent of the phenol was trapped in
the  first impinger,  and no phenol was found in the second, third,  or fourth
impingers.  With  PCP, 74 percent was  collected in the  first impinger, 11.5
percent in the  second impinger,  13.1  percent in the third impinger,  and 1.2
percent  in the fourth  impinger.   With  0-,0'-diphenol,  76.3 percent  was
trapped in the  first impinger, 12.4 percent  in the  second,  11.4 percent in
the  third, and  none  was detected in the fourth impinger.

Absorption of  Other  Aromatic Compounds

     The  possibility  of  the  formation   of  certain additional  compounds
during  destruction   testing  was  considered.    Therefore,  the  collection
efficiencies  for  several   compounds   in  the  bench-scale  apparatus  were
examined.  Approximately 0.1 gram of material was weighed into a combustion
boat.   The boat  was placed  into the  tube  ahead of  the  packed  bed  and the
temperature  was  raised from 200°C to 400°C over  a period of  35 minutes.
The  compounds  used  were acenaphthylene,  biphenyl,  and  phenanthrene.   The
collection  efficiencies were  97  percent,  90  percent,  and 90  percent,
respectively.   Of the  acenaphthylene trapped, 87  percent  was collected in
the  first impinger,   12  percent  in  the second  impinger,  and 0.8 percent in
the  third impinger.   Of the biphenyl trapped, 87 percent  was in the first
impinger,  11 percent in the  second  impinger, and  1 percent in  the  third
impinger.    With  phenanthrene,  83  percent  was  collected  in  the  first
impinger,  15 percent in the  second  impinger, and  2 percent in  the  third
impinger.

Compatibility of 2-Propanol With Analytical Measurement  Procedures

     The  analysis  of  CCl^  was  performed  on  a  gas chromatograph  with an
electron  capture detector.   An  electron  capture detector was necessary to
achieve the sensitivity required to  verify  that  99.99  percent  of the  CCl^
fed to  the burner was  indeed destroyed, as  required by U.S.EPA regulations.
Figure  3 shows the  analysis  of a 1.6 yg/ml CCl^  standard in 2-propanol.
The column used  was a  1/8" x  6' Chromosorb 101.   The  attenuation was X5.
There   is  tailing  from  the 2-propanol  peak,  but  the  CCl^  peak  is  not
seriously interfered with.

      The  analysis of DCB was  performed by liquid chromatography, with  a UV
detector  set at  210 nm.  Figure 4 shows  the analysis  of  a 1.3 yg/ml DCB
                                      24

-------
Figure 3.  Analysis of 1.6 ug CCl^/ml  in  2-propanol.
           Chromasorb 101 Column - Ni63 Electron Capture Detector.
Figure 4.  Analysis of 1.3 ygl,2 - Dichlorobenzene by HPLC.
           Reverse phase C-18 column.  UV detector set at
           210 nm, 0.05 absorbance units full scale.
                                    25

-------
standard  in  2-propanol.    The  column  used  was  a  reverse  phase  CIS.
Absorbance  range  was  0.05  AUFS.    At  210  nm,  the  absorbance  by  the
2-propanol is very  small  and  the  2-propanol elutes at about two minutes for
an elution volume of 4 ml.  The DCB  elutes at about six minutes,  so there
is no interference  from the 2-propanol.   At 210 nm,  there are no signficant
impurities in the 2-propanol  that  interfere with the DCB analysis.

Comparison of 2-Propanol  With Solid Absorber System

     There are  several advantages  to  the  use of 2-propanol as the absorber.
The  greatest  advantage is  the  sample analysis time  saved  with this system
over the  solid absorbers.  Figure  5  shows  the steps associated  with each
method.   Essentially no sample preparation  is  needed after the gas samples
are  taken using 2-propanol.   With use of  solid  absorbers  for chlorinated
organics,  the XAD-2 must  be soxlet  extracted before sampling to clean it up
sufficiently  and  then again after the sample  is, taken to desorb the chemi-
cal  compounds.    The impinger  contents  must  be  extracted and  the organic
layer  combined  with the soxlet extract and evaporated down before analysis
by the  appropriate  method.

     Another  advantage  is the adaptability  of   the  2-propanol  to highly
volatile  species.   Very good  trapping  efficiencies were  demonstrated for
CCli, and DCB,  which are  very  volatile  when  compared  to PCBs.   Haile and
Baladi  reported decreased  trapping efficiencies  for the more volatile PCBs
(10).   When the solid  absorber system was  tested with DCB  in the El bench
scale  system,  only 82 percent  of  the DCB was  recovered.   This compares to
the  97 percent that  is  recovered using 2-propanol.   The  solid  absorber
system was  also used to sample the gas stream in  the pilot plant.   A sample
was  pulled using 2-propanol,  then immediately a sample was pulled  using the
solid  absorber  system.  Immediately  after  that,  a second sample was pulled
using  2-propanol.    The  results using  the  solid  absorber  system  were   a
factor of  ten  lower  than the  average   of  the  two  samples  taken  using
2-propanol.    These  findings  agree  with  Haile  and Baladi  that,  as the
volatility of  the  chemical species  goes up,  the  collection efficiency for
solid  absorbers goes down.

     Another advantage  of this system over  the  Source  Assessment  Sampling
System (SASS) is that of  cost.  The  SASS cost  is  currently  in  the  neighbor-
hood of $14,000 to  $25,000,  whereas  this  modified  EPA  Method 5 system can
be  set up for a few hundred dollars.

     One disadvantage  of this system is  the volatility of  the  2-propanol.
When sampling  times  run  past  30  minutes,  evaporation  of  the  2-propanol
begins to  be  a  problem.   Generally,  this  can  be  avoided,  but  if  sample
times  longer than  30  minutes  are required,  there will  be  a loss  in  volume
of  the 2-propanol that must be accounted for.  Although there is  a loss  in
volume of  the  2-propanol, the compounds  being  sampled in  the gas  stream
will not necessarily migrate  from one impinger to  the  next.   An impinger
containing  130   yg DCB/ml   was  placed  in  the  first  position  in the
bench-scale apparatus.   The  tube furnace was  heated to 400°C, and  air was
pulled through the  system for  60 minutes.   There  was an appreciable loss  in
volume of the  2-propanol, but only  1 percent  of  the DCB had  migrated from
 the first impinger  to the second.
                                      26

-------
            IPA METHOD
                                                        SOLID ABSORBER METHOD
         SAMPLE GAS STREAM
              (20 MIN)
CLEAN ABSORBER
     11/a HR
     MEASURE VOLUME OF ABSORBER
                                                            SAMPLE GAS STREAM
                                                                 (20 MIN)
           FILTER SAMPLE
              (10 MIN)
         ANALYZE DIRECTLY BY
              GC OR LC
              (20 MIN)
                                                             CLEAN IMPINGERS
                                                          COMBINE WITH CONTENTS
                                                                  V* HR
                                                             SOLVENT EXTRACT
                                                       IMPINGER CONTENTS AND RINSES
                                                                  1/a HR
         TOTAL TIME= 1 HR
                                                             SOXLETT EXTRACT
                                                             SOLID ABSORBER
                                                                  1 HR
                                                     COMBINE EXTRACTS AND EVAPORATE
                                                                  V2 HR
                                                           ANALYZE SAMPLES BY
                                                            GC OR LC (20 MIN)
                                                       TOTAL TIME= 4 HRS 40 MIN
Figure 5.   Flow chart of  sample  analysis.
                                         27

-------
     One advantage  of  using 2-propanol over  other absorbers is  that  it  is
probably less hazardous  to handle than the  other liquid absorbers.   There
is  still a  possible  fire hazard,  but  this  is  encountered  whenever  an
organic  solvent  is  used.   The  use of  solid absorbers  is also  hazardous
because of the extended method steps involving  organic solvents.
                                 CONCLUSIONS

     From  El studies,  we have  concluded that  2-propanol is  a very  good
absorber for hazardous  pollutants.   With proper cooling,  it can even sample
volatile compounds such as  carbon  tetrachloride.   It offers a technique for
direct analysis  after sampling a gas  stream.   With this  system,  four  sets
of gas samples can be  taken and  analyzed in  one day by one person.

     Absorption  efficiencies  are  generally  very  good,  with  most of  the
compounds  being  trapped in  the first impinger.   The number of  impingers to
be used is dependent  on the volatility  of the  compound  being analyzed  for.
Most organic compounds  show good solubility  in 2-propanol.

     2-propanol  has minimal impact on the analysis of CCl^  and DCB.   Some
tailing of the 2-propanol does overlap  low-level CCllt peaks.  This tailing
is not serious and  ug/ml levels of  CCl^ can be easily analyzed.

     The use of  2-propanol  has many advantages,  such  as  the  time  saved per
analysis.  It is  easily adapted  to  sampling  of the more  volatile compounds.
The cost of  the  sample  system is much  less  than that of  systems such as the
Source Assessment Sampling  System.   The  major disadvantage of the  system is
the volatility of the  2-propanol when sampling periods are  longer than 30
minutes.   Even  with  this   loss in  the 2-propanol  volume,  the  chemical
compound being trapped  does not  tend to  migrate with the 2-propanol.
                                 REFERENCES

1.   Anonymous.   1976.   Emission testing  at  Continental  Can Co., Hopewell,
     VA, July  14-23,  1975.  EPA-330/2-76-030,  October.   U.S.  Environmental
     Protection Agency,  Office of Enforcement.

2.   Ferguson,  T.L.,  F.J.  Bergman,  G.R.   Cooper,  R.T.  Li,  and  F.I.  Honla.
     1975.   Determination of  incinerator  operating conditions  necessary
     for safe  disposal  of pesticides.   EPA-600/2-75-041,  December.

3.   Leighton,  I.W. ,  and J.B. Feldman.   1974.   Demonstration  test  burn of
     DDT in General Electric's liquid  injection incinerator.  U.S. Environ-
     mental Protection  Agency Region I, December.

4.   Guilford, N.G.H.,  and R.J.  Brandon.    1978.   Pilot  scale evaluation of
     the destruction  of waste PCB by  incineration.   Ontario Research Foun-
     dation, Mississauga, Ontario,  Canada, August 28.
                                     28

-------
5.   Guilford, N.G.H.,  and  G.  Rosenblatt.   1977.  Polychlorinated  biphenyl
     source   testing   survey,   Ontario  region.     MS   Report  No.   OR-9,
     September.

6.   Komaniya, K.,  and S.  Morlsaki.   1978.   Incineration of  PCB  with  an
     oxygen burner.  ES&T 12 (10) : 1205-1208.         ;

7.   Shoemaker, R.A.  1981.  J. Chromatogr. Sci.  19:321.

8.   Ackerman, D.G.   1977.   Destroying chemical wastes in commercial  scale
     incinerators.  Final report phase II.  EPA  Contract No.  68-01-2966,
     November.

9.   MacDonald, L.P.,  D.J.  Skinner,  F.J.  Hopton, arid G.H.  Thomas.   1977.
     Burning waste  chlorinated hydrocarbons  in  a cement  kiln.   Report  to
     Fisheries and Environment Canada.  Report No. EPS 4-CoP-77-2,  March.

10.  Halle, C.F.,  and  E.  Baladi.    1977.   Methods for  determining  the PCB
     Emissions  from  incineration  and  capacitor  and transformer  filling
     plants.  EPA-600/4-77-048, November.

11.  Beard,  J.H.   Ill,  and  J.   Schaum.    1978.     Sampling methods  and
     analytical procedures  manual for PCB  disposal:   interim report.   U.S.
     Environmental Protection Agency, Office  of  Solid Wastes, February.

12.  Bbmbaugh, K.J., and K.W. Lee.  1981.  ES&T  15 (10) :1142-1149.

13.  Acurex Corporation.   1978.   Source assessment  sampling  system design
     and development.   EPA-600/7-78-018.

14.  Arthur D.  Little, Inc.   1978.   EPA/IERL-RTP  interim  procedures for
     level 2 sampling and analysis  of organic materials.  EPA-600/7-78-016.
                                    29

-------
                    FORMALDEHYDE  SURFACE  EMISSION MONITOR
              T.G. Matthews, A.R.  Hawthorne,  J.M.  Schrimsher,
                       M.D.  Corey,  and  C.R.  Daffron

                    Health and Safety Research Division
                       Oak Ridge National Laboratory
                                  ABSTRACT

     A  formaldehyde  surface emission monitor is under  development  for  pas-
sive,  non-destructive measurement  of  formaldehyde  (CH20)  emission  rates
from  flat  surfaces  of  solid CH20 sources.   The monitor  utilizes a  solid
sorbent, 13X molecular  sieve,  that provides  excellent chemical  stability
for sorbed CH20 and  permits  the monitor  to  be  used  in any  physical orienta-
tion.  With  a 0.032  mutest  area, a =0.01 mg CH20/m2'hr detection limit can
be achieved  with a  3-hour  sampling  period  and pararosaniline  colorimetric
analysis.  Preliminary  results  indicate that the monitor  could  be  used for
1)  quality   control measurements   of   commercial   CH20   resin-containing
materials  such  as pressed—wood  products,  and  2)  in—situ measurements  of
CH20 emission rates  from a  variety of  CH20  sources  in domestic  environments
such  as  pressed-wood,   textiles,  and  urea-formaldehyde   foam  insulation
products.
                                INTRODUCTION

     Formaldehyde  (CH20)  is  currently recognized as an  important  pollutant
in a  variety of  industrial  and domestic  environments.    The  100 ppb  CH20
concentration  guideline  recommended for  indoor  air  by  several  European
Nations  (1)  and  the  American Society  for  Heating,  Refrigeration,   and
Air-Conditioning  Engineers  (ASHRAE) (2)  reflects  a  serious  concern  for
prolonged public  exposure at low concentrations.  As  a  result,  two general
needs  in  monitoring  technology  have  arisen:   first,   in-situ  analysis
methods  to  identify dwellings  with  high  CH2C>  levels and  the  major  CH20
sources within these environments, and second,  quality control  test methods
for a  variety of consumer and  construction  products  containing  CH20-based
resins.

     Despite  a  plethora  of  ambient  vapor monitoring  techniques  for CH20,
source strength  analyses for CH20-resin-containing products are  currently
limited  to  sample-destructive  tests  of   selected pressed-wood  (3)   and
textile products  (4).   As a result, the  development  of  a  non-destructive

                                     30

-------
 surface  emission monitor  for CH20  emission from  consumer  products  could
 have  a major  impact o^ improved quality  control  and  in-situ measurement
 capability.    For  quality  control  applications,  a.  broader  testing  of
 construction  and consumer products could be performed,  resulting  in selec-
 tive  grading  of  materials  prior  to sale to  the  consumer.  For in-situ moni-
 toring,  precise  source  identification  could  be  accomplished without  the
 destructive   removal  of   numerous  samples   for   subsequent   laboratory
 analysis.

       This  report addresses  the  development of  the  new  surface  monitoring
 methodology.   The basic  design  and  specific  applications  to  pressed-wood
 products,  urea-formaldehyde  foam  insulation   (UFFI)   in  simulated  wall
 panels,  and  CH20 source  identification and  potential  quantification  in
 three homes is also  addressed.
                               MONITOR DESIGN

      The  physical  design  of  the  formaldehyde  surface  emission  monitor
 (FSEM) is  conceptualized in  Figure 1.   It is a  passive flux  monitor  for
 CH20 emission  from a selected area of  a solid flat surface.   Formaldehyde
 is collected on  a 13X molecular sieve  (1.6 mm pellets),  suspended  in uni-
 form close proximity  to  the surface of  the test media.   The  CH20 content is
 then  analyzed  via  a  water-rinse  desorption  and  colorimetric  analysis
 methodology (5).   The  average CH20 emission rate of the  test  media  is cal-
 culated as a function  of  the sorbed CH20,  sampling period, and test  area of
 the monitor (Equation  1).
 Emission Rate
(mg ,CH20/m2-hr)
	CHaO  (ing/ml)-Sieve Rinse Volume  (ml)    	
Test Period (hrs)-Test area (mz)-CH20 Desorption Efficiency ^  '
            FORMALDEHYDE SURFACE EMISSION MONITOR
                               COVER
.. J

MESH CONTAINER
T
a
-1
p

/SOFT
GASKET
 Figure 1.  Conceptual design of the Formaldehyde Surface Emission Monitor.

                                      31

-------
The  sieve also  sorbs  sufficient  quantities  of water  vapor  to  eliminate
potential condensation  problems  within the monitor.

     The sensitivity  of  the  surface  emission measurement is proportional to
the  test  area of the monitor  and  the duration  of  the sampling period.   A
large  test area  is  desirable to limit  the effects of  the  inhomogeneity of
the  test  medium.   In contrast, a  short sampling  period  is  desirable  for
convenience and  to  limit any physical changes  in the test  medium,  such as
dehydration,  that could  occur  as a  function of  sampling period.

     The  separation between the sorbent  and emitting surface  (distance a,
Figure  1) is  an additional  important design parameter,  which has  a major
impact  upon the  steady-state CH20 concentration  profile within the  passive
monitor.  In  an  ideal measurement,  the CH20  concentration  within  the moni-
tor and in the  environment surrounding the monitor should  be equal  because
of the  potential for  CH20 concentration dependence of  the  emission  rate of
the test media.   In practice,  the CH20 concentration  within the  monitor is
dependent upon the  source strength  of the test media.   With an appropriate
design, the concentration can  be  limited to a  range consistent  with most
living  environments.    A  simple  mathematical  model   (Equation 2)  for  the
steady-state  CH20 concentration at  the surface  of the test  media as  a func-
tion  of emission  rate  and  sorbent—test  media  separation  is  represented
graphically in Figure 2.
[CH20] (ppm)
                            Emission
                  0.8[ppm/(mg/m3)] • Rate
Diffusion Coefficient (m2/hr)
                   Sorbent-Test Media
               (mg/m2«hr) • Separation  (m)   (2)
The primary  assumptions  of this model  are simple diffusion  behavior  and a
100 percent  efficient  CH20  sorbent.    For emission  rates  of  0.0 to  1.0
(mg/m2«hr),  typical  of most  consumer and construction products,  a CH20 con-
centration range  of  0-0.2  ppb is predicted for a separation of =:2 cm.

     The current  prototype of the  FSEM  is constructed from a 20-cm mechani-
cal sieve with  a #20 mesh.   The load factor  within the monitor  is  =20  (m2
test area/m3  air volume).    The sorbent-test media  separation  is  2.2  cm.
For horizontal  operation, the  sieve sample  (10  g) is  sprinkled uniformly
over the screen of  the mesh  container.   For vertical  operation,  the  sieve
is held in several screen  trays attached directly to the screen in the mesh
container.   With a  typical  test  period  of  3  hours,   the  lower  limit  of
detection is =0.01 mg/m2«hr.  This  is  adequate for  all materials normally
incorporated in living environments that  could  have a  significant  impact
upon the ambient  CH20 concentration.
        QUALITY CONTROL  (QC)  APPLICATION FOR PRESSED-WOOD PRODUCTS

Comparison to Current QC Methods

     The  current  analytical method used in the United States  for measure-

                                    32

-------
     0.6
     0.5
 o
 *
 LU
 CJ
 z
 o
 o
 I
 LU
 Q
 oc
 o
 U-
 X
 <
     0.4
     0.3
     0.2
     0.1	— —
                         0.25              0.50              0.75

                         FORMALDEHYDE EMISSION RATE (mg/rr\2hr)
                                                                          1.00
Figure 2.   Diffusion model  for  the CH20 concentration at the surface  of  the
            test medium  as  a function  of  sorbent—test  medium separation  and
            the CHpQ  emission  rate  of the  test  medium.   The  dashed  line
            corresponds  to  the  ASHRAE  guideline of  0.1  ppm  for indoor  air
            (2).
                                      33

-------
 ments  of  CH20  release  from  pressed-wood  products  is  the  2-hour  static
 dessicator test of  the Hardwood  Plywood Manufacturers'  Association-National
 Particleboard Association  (HPMA-NPA)  (3).   It is a  sample  destructive  test
 method  that  is  applied to  as  few as  1 of  every 1,000 to  10,000  sheets  of
 particleboard decking  and  plywood panels used principally  in mobile  homes.
 Prior to  testing,  the  pressed-wood samples  (5x13 cm) are conditioned for  1
 to  7  days in  a 22°C,  50  percent  RH environment.   The CH20  sorbent  and
 colorimetric analyses used  are water  and chromotropic  acid, respectively.

      The  proposed  non-destructive method  using  the   FSEM is  modeled  as
 closely as possible to  the HPMA-NPA method to  promote its  acceptance  as  a
 quality control technique.   The  primary differences are 1) application  to
 all pressed-wood  products, 2) inclusion  of new  conditioning  criteria, 3)
 use of  a  solid  CH20 sorbent,  13X molecular sieve,  and 4)  substitution  of
 the pararosaniline  (PA) colorimetric  analysis.

      Sample  conditioning  is  performed for >2 weeks  in an  atmosphere at
 22°C,  50 percent RH, and  0.10 ppb CH20.  This allows   the  emission rate of
 the test media to stabilize under  conditions consistent with typical  living
 environments.  In practice,  any  significant change in  environmental  condi-
 tions  (e.g.,  ±5°C, ±25 percent RH) can cause  large fluctuations (e.g., >50
 percent)  in  the CH20  emission rate,   which can  require 1  to  3  weeks to
 stabilize.

     The use  of  the 13X molecular  sieve  instead  of  water as a CH20 sorbent
 has several advantages.  These include 1) the  option to use the monitor in
 any physical orientation,  2) the  elimination  of water condensation  prob-
 lems,  and  3)  long-term stability of   sorbed  CH20 (5).   In addition,  the
 sieve  does not  appear to suffer from the sampling rate  anomalies as a func-
 tion of  sorbed CH20 concentration as does the water sorbent (Figure 3).   In
 preliminary studies  conducted under  near-static conditions,  the  sampling
 rate  of  water  in  a  =:10-cm-diameter  open tube sampler  demonstrated  a
 non-linear CH20  uptake as  a function  of  concentration  x time.   In contrast
 to  the  experimental results,  which  showed an  enhanced collection  rate,
 diffusion  theory would  predict  a rollover  in  the  sampling  rate at high
 CH2P*H20 concentration  levels due to the  increased  CH20  vapor  pressure
 above  the  sorbent.   A clear explanation of this sampling  rate phenomenon
 has  yet  to be made.   The sieve exposure was  performed using  a =12-cm-diame-
 ter  tubular sampler  containing a #20 mesh screen  to  support the sorbent in
 the  center of the  unit.  Under similar exposure  conditions  to  the experi-
 ment with  the water sorbent,  a  linear uptake  of CH20 was observed as  a
 function of concentration x time.

     The use  of  the PA  analysis  results  in a  three-fold increase in CH20
 sensitivity over the chromatropic  acid (CA) method (5).  No difference in
 the  degree of  CH20  selectivity  between the  PA  and CA  methods  has  been
 observed in  the measurement  of  particleboard,   plywood,   fiberboard,  and
 paneling samples.

     Comparative measurements of four  types  of pressed-wood products were
 performed  using  the 2-hour  dessicator and  FSEM methods (Figure  4).  All
measurements  were  taken 1  to 3  weeks  following sample   preparation  and"

                                     34

-------
 (J
 UJ
 O
 O
 UJ
 Q
 >-
 UJ
 Q
  cc
  O
     10.0
      8.0
      6.0
4.0
      2.0
                                         O   HPMA-NPA SAMPLER,
                                             WATER SORBENT

                                         A   MOLECULAR SIEVE
                                             SAMPLER
                           1.0                 2.0

                              FORMALDEHYDE EXPOSURE (ppm hr)
                                                               3.0
Figure 3,
     Formaldehyde  sampling characteristics of passive samplers
     containing water  (o)  and molecular sieve (A) sorbents.
conditioning.    For low  emitters,  an  excellent  correlation  is  observed
between the two  test methods.  For the  strong emitter (paneling  sample),  a
rollover is observed that  is  consistent with the CH20 sampling  rate anomaly
of the water sampler used  in  the  HPMA-NPA test.

Comparison to Dynamic  Chamber Tests

     Environmental   chamber  analyses   represent   the  closest   laboratory
approximation  to real-world  conditions.   Such tests are  therefore  useful
for  intercomparison  of simple   static  methods  used  for  quality  control
applications.   The  chamber method  involved the testing of edge—coated wood
product samples  at  24°C and 50 percent  RH  under dynamic flow conditions  in
a 0.2  m3  Teflon-lined chamber.    The air  exchange rate  (ACH)  and  loading
factor  were .varied in  the  ranges  of  0.7-4.4 hr~   and  0.02-1.70  m2/m3,
respectively.   The  concentration of  the CH20 vapor exiting the chamber was
measured with  a modified  CEA Instrument (6).   The Cl^O  emission  rate was
then determined  using  Equation 3.

     One  important  result  of ' this analysis was  a comparison,  of  the CH20
emission rates  of the  test media as a function of  CH^O vapor  concentration.
                                     35

-------
        6.0
   1
   oc
   o
   g
   in
   CO
   ui
   oc
   CO
   UJ
   o

   111
   o
   
-------
I
CO
CO
Q
oc
o
    1.2
    0.8
    0.4
                                          TEMPERATURE = 23 ± 1°C
                                          RELATIVE HUMIDITY = 50 ± 5%
SURFACE EMISSION
MONITOR
                                   HPMA-NPA TEST
                                    13X SIEVE SORBENT
                                                     HPMA-NPA TEST
                                                     H2O SORBENT
                                                            2 hr
                   0.2           0.4          0.6           0.8


                          FORMALDEHYDE VAPOR CONCENTRATION (ppm)
                                                                      1.0
Figure 5.   Formaldehyde emission rate of  edge-coated wood paneling samples
            as  a  function of ambient CtLjO  concentration.
test media-CH20 sorbent separation  inherent in the dessicator results in a
larger CH20  concentration at the  surface  of the test  samples.  This causes
the  enhanced   suppression  of  the  CH20   emission  rate  of  pressed-wood
products.
                   IN-SITU SURFACE MONITORING  APPLICATIONS

Simulated House  Wall Test Panels Containing UFFI

     The measurement of CH20 emission from UFFI through gypsum board repre-
sents a separate class of applications for the FSEM where the primary emit-
ting material  is only indirectly accessible.   To test  the  efficacy of this
type  of  measurement,  a  comparison  study was performed applying  the FSEM
and  laboratory methods  to  a selection of  simulated house—wall  test panels
(Figure 6)  containing UFFI.   The panels  were foamed according  to  manufac-
turers' specifications at  the  Franklin Research  Institute  in March-April,
1980  (7).   The  reference  dynamic-flow determination  of the  CH20  emission
                                      37

-------
17'/2X
Vz GYPSUM
 WALLBOARD
                    ' 17V2 x 93
                     'J Mil MYLAR
                     COVER (FRT & REAR)
                     COMPARTMENTS,
                     SEALED TO PANEL
                     w/ACRYLIC LATEX
                     BAULKING
                    1 '•:• *3
-------
 rate was performed by measuring  the CH20 vapor concentration  in  air  passed
 through an external  chamber  sealed to the gypsum  board side  of  the  panel.
 This chamber was  used  to simulate  the  effects of  air  exchange  rates  that
 are typical of indoor environments.   The FSEM measurement was  performed  by
 attaching the monitor directly to  the surface  of the  gypsum  board.

      In a previous study (8) the emission rate  through  the gypsum board was
 found to  be  a strong  function  of the  CHgO  concentration  in the chamber
 exterior to the gypsum board.   Below -0.2 ppm,  the CH20 emission  rate was
 maximized.   At concentrations  increasing from -0.2 to 2-7 ppm (static  con-
 ditions),  the CH20 emission  rate  decreased  to near-zero levels where  emis-
 sion from the UFFI was  highly suppressed.

      For the dynamic flow measurements,  the  air  flow  through the chamber
 was sufficient  to maintain a sub-0.2 ppm concentration  level.   The  tempera-
 ture and relative  humidity  of the  laboratory  at  the time  of  the  measure-
 ments were  16 to  18°C and 30 to 50 percent RH,  respectively.   This  accounts
 for the low  CHaO  emission levels  in  comparison to previous work  that was
 conducted  at  25°C and 50 percent RH (8).

      The linear correlation  between  the  FSEM  and  dynamic flow measurements
 is  excellent (Figure 7).   This  is  strong  evidence  that  both measurement
 protocols  maintained sub-0.2  ppm  CH20 concentration  levels  at the surface
 of  the  gypsum board,  thereby maximizing  the CH20 emission rate.  The slope
 of  1.3  is  consistent  with a slightly larger test area for the  FSEM  than the
 specific opening  of  the  monitor.    This is  probably  the  result  of  CH20
 transport  through gypsum board adjacent  to  the opening  of the monitor.  A
 14  percent  increase  in  the  radius of the  effective  test area (i.e.,  ^1.3
 cm,  which  is  equal to the thickness  of  the gypsum  board) would account for
 a 30 percent  increase in the  CH20 sampling rate.

 FSEM Measurements  of  Surfaces of Walls and Floors in Homes

      FSEM  measurements  were  performed  on the gypsum board surface of
 interior-partition and exterior walls, and carpeted and tiled floors of one
 UFFI and two non-UFFI homes   (Table  1).   A  primary  goal  was   to  test the
 efficacy of  the FSEM for localizing  and  ranking the important  CH20 sources
 in  individual  environments.   In  dwelling 1,   no  significant  CH20 sources
 were found.    In  dwelling 2,  the  strongest  source  was  the  UFFI in  the
 exterior walls.  The  primary  emitter in  dwelling 3  was probably the plywood
 decking; the  CH20  emission could be detected through the carpeting.  In all
 three  dwellings the  floor  tile was  an  effective  blocking  agent  for CH?0
 emission from the  wood  flooring.    In dwellings 2  and  3, the off-gassing
 from interior  gypsum board walls  with  empty  wall cavities  indicates  the
 strong  sorbtive and  desorptive  potential of  construction materials  with
 significant water  content.

     A second objective was a comparison of the measured CH20 concentration
 levels with concentrations  estimated from FSEM  and  air  exchange rate  (ACH)
measurements.   A  good correlation  between the measured  and  predicted con-
 centrations  would  support  the  quantitative  capability  of  the  FSEM  for
 in-situ  environmental analyses.  For  dwellings 1,  2, and  3,  the  areas of

                                    39

-------
     0,25
     0.20
.c
r*

O)
E.
cc
O

is
O
5
2
O
CO
00
5
LU
UJ
O
£

-------
   TABLE  1.   MEASUREMENTS OF CH20 EMISSION RATES FROM MAJOR SURFACES
   	IN THREE OCCUPIED HOMES USING THE FSEM
                          Formaldehyde emission rate (mg/m2«hrT
  Dwelling
 Exterior gypsum Interior  gypsum    Carpeted
   board wall	board wall          floor
   Tiled
   floor
   1 non-UFFI*  0.014  ± 0.005   0.014 ± 0.005

   2 UFFI       0.30   ± 0.03     0.10  ± 0.010

   3 non-UFFI   0.18   ± 0.02     0.22  ± 0.02
                                   0.011  ±  0.005    0.00  ± 0.005

                                   0.095  ±  0.01     0.022  ± 0.005

                                   0.27   ±  0.03     0.024  ± 0.005
   feUrea-formaldehyde  foam insulation.
  the  principal emitting  surfaces,  including the  interior walls  (and  ceil-
  ing), external walls,  and carpeted and tiled floors,  were  estimated for an
  average room.  The  emission rate from gypsum board  surfaces  was  reduced by
  a factor  of  1.3  to  account for the effective  test area  of  the FSEM deter-
  mined from the simulated wall panel work.  .No analogous  type of  correction
  factor was applied  to the  floor  data.   With the  final CH 0  emission rate
  data for  each surface and  a  measured ACH,  a  predicted  CH20 concentration
  was determined via Equation 4.
[CH90](ppm)
(zSource Area (m2) •  Emission Rate(mg/m2'hr) ) •  0.8  (ppm/Cmg/m3)")
                 Room Volume (m3) • ACH (hr-1)
            (4)
       A comparison  of  measured and predicted  CH20 concentration levels  for
  dwellings 1, 2, and 3  is  given in Table 2.   An encouraging correlation  is
  observed for all three dwellings despite the  inherent  complexity and  numer-
  ous experimental variables associated with the  comparison.


  TABLE 2.   COMPARISON OF MEASURED CH20 CONCENTRATION LEVELS WITH VALUES
  	PREDICTED FROM COMBINED FSEM AND ACH MEASUREMENTS
  Dwelling
                                       Formaldehyde concentration (ppm)
                      Estimated
                                                                  Measured
  1  non-UFFI*

  2  UFFI

  3  non-UFFI
                     0.04 ±  0.02

                     0.16 ±  0.05

                     0.21 ±  0.14
0.036 ± 0.01

0.15  ± 0.05

0.18  ± 0.02
  bUrea-formaldehyde  foam insulation.
                                      41

-------
                                  SUMMARY

     Quality control applications of  the FSEM  to pressed-wood  products  have
been  examined.   A  comparison of  the FSEM  and a  conventional  dessicator
method used  in the United  States  showed  a strong  linear correlation  for
weak  emitting  boards.     A comparison test  with a  dynamic chamber  method
showed that  the surface monitor  gave results  consistent with  chamber  data
at 0-0.2 ppm concentrations.

     The  feasibility  of  certain  in-situ monitoring  applications with  the
FSEM have  been  studied.   An encouraging quantitative  relationship  has  been
observed  between  the FSEM  and a reference  method for  CH20 emission  from
UFFI through gypsum board in simulated wall  panels.   The results  of  surface
emission monitoring in three homes  were  also supportive  of  the quantitative
capability  of  the  monitor.    A close  correlation  was  observed  between
measured  CH?0  concentrations  and CH20 concentrations  estimated from  FSEM
and ACH. measurements.

     The potential uses  of  FSEM for  both quality control  and  in-situ moni-
toring  applications  have  yet  to be  fully  developed.   A  non-destructive
quality  control method  for CH20 is  also of  interest  to  the textile  and
non-UFFI insulation industries.   The  monitor could  also  be used to  evaluate
the long-term decay of CH20 release from consumer  and construction  products
incorporated  in living  environments  and  the  efficacy  of  source  removal
methods.
                               ACKNOWLEDGMENTS

     Research  for  this paper was sponsored jointly  by  the  Consumer Product
Safety  Commission  under  Interagency Agreement  CPSC-IAG-81-1360  and  the
Office of Health and Environmental  Research, US Department  of Energy,  under
contract W-7405-eng-26 with the  Union Carbide  Corporation.
                                 REFERENCES

1.   Borzelleca,  J.F.   1980.   Formaldehyde-an  assessment  of  its  health
     effects.    Report  to  the  U.S.  Consumer Product  Safety  Commission,
     March.

2.   American  Society  for  Heating,  Refrigeration,  and  Air-Conditioning
     Engineers Standard 62-1981.

3.   Tentative  test   protocol   for  emission  of  formaldelhyde  from  wood
     products.   1981.   Hardwood-Plywood Manufacturers'  Association,
     National Particleboard Association.

4.   AATCC  Test  Method  112-1978.   Formaldehyde odor in  resin-treated  fab-
     ric, determination of:   sealed jar method.
                                     42

-------
Matthews, T.G., and T.C. Howell.  In press.  Solid sorbent methodology
for formaldehyde monitoring.  In Analytical Chemistry.

Matthews, T.G.   In press.   Evaluation  of  a modified CEA Instruments,
Inc. , Model  555  Analyzer for the monitoring  of formaldehyde vapor  in
domestic  environments.     American  Industrial  Hygiene  Association
Journal.
Osborn,  S.W.    1981.    Technical  report  F-C5316-01.
Research Center, Philadelphia, PA.
The  Franklin
Hawthorne, A.R.  An evaluation of formaldehyde emission potential  from
urea-formaldehyde foam insulation:  Panel measurements and modeling.
ORNL/TM-7959.                                  ;
                               43

-------
         CORRELATION OF REMOTE AND WET  CHEMICAL  TECHNIQUES  FOR THE
      DETERMINATION OF HYDROGEN FLUORIDE EMISSIONS  FROM GYPSUM PONDS
      Howard F. Schiff, Daniel Bause, Mark McCabe,  and  Verne  Shortell

                          GCA/Technology Division
                                Bedford, MA

                             William F. Herget

                   U.S. Environmental Protection Agency
                Environmental Sciences Research Laboratory
                        Research Triangle Park, NC

                                Mark Aritell

                   U.S. Environmental Protection Agency
                     Office of Air Noise and Radiation
                 Division of Stationary Source Enforcement
                              Washington, DC
                                 ABSTRACT

     Data  on concentrations  of  gaseous hydrogen fluoride  in  air near  an
extended  area  source were  collected simultaneously  by  the Remote Optical
Sensing of Emissions (ROSE) system using longpath,  high-resolution, Fourier
transform  infrared  (FT-IR)  absdrption  spectroscopy  and  by  standard  wet
chemical techniques.  The program was  divided  into  five  phases,  including a
literature  review,  pretest surveyj  sampling and analytical trials  in  the
laboratory,  prelimiiiary field  phase,  and  the  final,  collaborative  field
phase.  Precision and accuracy of standard  techniques were evaluated  in the
laboratory and preliminary  field phases.

     Field sampling  efforts were  conducted along gypsum ponds at  two  phos-
phate  fertilizer facilities.    Point  sampling efforts  utilizing  both  the
double filter cassette  and  sodium bicarbonate-coated tube  methods  were con-
ducted simultaneously  with the operation  of the ROSE  system.   Four  point
sampling  sites  were  located  at  approximately  equal intervals  along  the
optical path from the ROSE  system  light source to the ROSE  system van con-
taining  the FT-IR  system.    The fluoride  collected  by  the  wet chemical
methods  was  analyzed  cblorimetrically using  a  semiautomated  method  with
Lanthanum-Alizarin  complexone reagent.  The FT-IR  data  were collected  at
0.125 cm-1, spectral  resolution,  and the HF concentration  was determined by

                                     44

-------
measuring the peak absorption  of  the R(5) line of HF at  4174  cm-1,  after a
minor correction  for  water vapor interference.  In  32  independent tests of
comparable ambient HF concentrations, the overall average  HF  concentration
was 36.1  ppb (ROSE system)  and 36.4  ppb (wet  chemical  techniques).   The
standard  deviation between  the ROSE  system data and  the manual sampling
results was 9.7 ppb.
                                INTRODUCTION

     For several  years,  the U.S. Environmental Protection  Agency (EPA)  has
used the Remote Optical  Sensing of Emissions  (ROSE)  system to characterize
the  gaseous  pollutants  emitted by  a variety  of point  and  extended  area
sources  (1).   The  ROSE  system consists  of  a  Fourier transform  infrared
(FT-IR) interferometer  with telescopic  optics  and has been  installed  in a
van.   The system is used  either with  a  remotely  located  infrared  light
source to make longpath  (up to  1.5 km)  atmospheric  absorption measurements
or  in  a single-ended  mode  to  measure  the infrared  emission signal  from
gases exiting industrial  stacks at  elevated temperatures.

     For the purpose of  developing  a technical basis for enforcement action
to abate human health  hazards,  it may be  necessary  to determine  concentra-
tions  of  toxic gaseous  pollutants  in the vicinity of  sources.    The  ROSE
system used in the  "active  longpath  mode"  is  conceptually capable of evalu-
ating  the  breathing zone pollutant  concentrations.   The  purpose  of  this
project was to extend  the data base of  this versatile and  promising pollu-
tant sensor  by comparison of data  generated by  the ROSE  system with  data
generated by  standard  techniques for  the  measurement: of hydrogen  fluoride
(HF).

     A five—phased  approach  was adopted  for conducting the  project  as indi-
cated below:

        • Literature review                           ,

        • Site survey

        • Sampling  and analytical trials in the laboratory

        • Preliminary field  phase                                        ,

        • Collaborative  field sampling and analytical 'phase

     Prior to the commencement  of the sampling progra.m,  it  was necessary to
determine the wet  chemical techniques for  sampling" and  analysis  that would
facilitate the comparison with  the  ROSE system data.   Ambient sampling  for
HF is complicated by the low concentrations (ppb level), reactivity of  the
compound, and  the  effects  of   interfering  species.    The  selection  of  wet
chemical methods  was  based on compatibility  with  the  sampling  program;
factors  studied   included sensitivity  (minimum  sampling   time),  reproduc—
ibilityi ease  of  handling,  and freedom  from interferences.   The  literature
was  reviewed with respect to the above  sampling requirements  (1-10).   The

                                     45

-------
extensive  review of  Jacobson and Weinstein (2) for procedures prior to 1976
and NERAC  computer search for post-1976 procedures  formed  the basis of the
literature review.   The sampling  and analytical  procedures  selected are
published  as  ASTM methods (3) and were modified  to  fit  the requirements of
the  project.    In addition, discussions with Dr. Jay  S.  Jacobson  and Mr.
Richard  Mandl and Larry  Heller  of the Boyce Thompson Institute  for  Plant
Research  at  Cornell  University  were very  helpful  in the  selection  of
procedures.

     The sampling procedures selected were ASTM D3266-AISI  double-tape sam-
pler  with  citric  acid  and  sodium   hydroxide-treated  filter papers  and
D3268-bicarbonate-coated tube samplers.  D3266 was  modified  in that circu-
lar filters in  37-mm personal sampling cassettes were used  in place of the
AISI tape  samplers.   The analytical procedures selected were ASTM D3269 ion
exchange  separation  followed   by  conductometric  detection—i.e.,   ion
chromatography  and D3270-semiautomatic microdistillation spectrophptometric
method.  The  procedures  are described in  detail subsequently.
                               SITE  SELECTION

     The site  selection  was  based on criteria delineated below, and  onsite
surveys  of  facilities  in  the  Bartow,  Florida,  Phosphate  complex.    The
criteria were:

        • The  site should have  geography  compatible with the  ROSE  van and
          chemical sampling  methods.  This  included an  access  road for the
          ROSE van; an unobstructed line of sight of at  least 400 meters at
          the  edge of  gypsum pond (this  provides for a high signal-to-noise
          ratio  for the ROSE  system),  and no interferences  from emissions
          from the facility  or surrounding facilities.

        • The  facility  could  permit the use  of the wet  chemical sampling
          systems  utilizing  gasoline-powered electrical  generators.

        • The  facility would have onsite laboratory space  and  instrumenta-
          tion for fluoride  analyses.

        • The  concentrations  of  fluoride  in  the  ponds  and  the  pH of  the
          pond water  must be  sufficient to release ppb levels  of  HF  into
          the atmosphere.

     Two  gypsum  ponds,   one  in the  CF  Industries,  Inc.  (CFI),  Bartow,
Florida, plant,  and  the other  in  the  Agrico  Chemical Company (Agrico),
South  Pierce,  Florida,  plant, were  found  to meet  the  criteria and  were
selected for the conduct of  the project.   The plot  plans of each facility
are  shown  in Figures  1  and 2.   Sampling  line  B at  both facilities  were
utilized  for  the  lines of   sight  in  the preliminary  and  collaborative
phases.
                                     46

-------
                        POND FEED
                                                    CF INDUSTRIES
                                                     PLANT SITE
ffl
m
                            STACK—
             A
             L.
                      COOLING
                       POND
                                      SETTLING
                                        POND
                                                   STACK
r\    POND   r~\
                                 ELEVATION
Figure  1.   Sampling points  at CF Industries gypsum ponds.

                                        47

-------
                          LOWER
                          STACK
                          ROAD
                            B I
                                   ELEVATION
Figure 2.   Sampling points at Agrico Chemical Company's gypsum ponds.

-------
                             LABORATORY PHASE

     The   laboratory  phase  was   designed  to  determine   the   precision,
accuracy,   and   sensitivity  of   each   sampling,  method  under   controlled
conditions  of  hydrogen fluoride  concentrations.    To  do  so required  the
generation  of  precise  levels  of  HF,  as   described  below.   Analyses  were
conducted  by  ion chromatography with conductometric  detection.
                            HF GENERATION SYSTEM

     The  design  for  the HF  Generation  System and  the  injection box  was
supplied  by  Dr.  Jay S. Jacobson of the  Boyce  Thompson Institute  for Plant
Research  (personal communication, of Jacobson  and Heller,  letter  of  April
10,  1979).    An  HF generation  system  was  constructed  as  illustrated . in
Figure 3.  Air was pumped  through an indicating  silica gel drying trap,  a
tube packed  with  glass  wool, and a Whatman 42  filter  into heated  Teflon
tubing at a  flow rate of 1.5  dscfm.   An aqueous  HF  solution was  pumped at
0.05 ml/min  into the  heated  Teflon  tubing  (located  in the injection box)
through  which  the  filtered  air  flowed.   The injection box  was kept  at
175°F.   The  fluoride-laden air  was then cooled  to  room temperature  in an
ice bath  and  divided.   Sampling took place at two points downstream of  the
flow division.   The portion of air that  was not  sampled was exhausted to  a
laboratory hood.
   AIR    O
 COMPRESSOR
     HF RESERVOIR
(NALGENE GRADUATED CYLINDER)
                                          O°C
                                     (COOL AIR TO ROOM
                                        TEMPERATURE)
POINT OF
SAMPLING
Figure 3.  HF generation system.
     The amount  of  fluoride put  into  the system  is  dependent on  flow rate
and the concentration of the HF  solution,  i.e., the concentration of  solu-
                                      49

-------
tion.  in the  reservoir  times  the flow rate (pg HF/ml  x ml/min = pg HF/min).
Adjustment of  the  air  flow rate through the  system alters the concentration
of  fluoride  per  unit volume  of  air,  but  not   the amount  of  fluoride
delivered  through the system.    The latter  is  controlled  by varying  the
aqueous HF solution concentrations.    The  concentration of  HF in  the  air
stream is  determined by the  following equations:
   TTT^/J ^^o
   HF/dsf t3
                                   HF ml~l x ml min~l
                                  _ - ,    , - - -
                                  Vm, dsft3
where Vm, dsftS  is  at  77°F  and 29.92 in Hg
   TTT./J  o
pg HF/dsm3
                                Vg HF dsft-3
                                .If! _
                                0.02832 ft3 m-3
                    HF  (ppb)  =
                                    pg HF dsm-3
                                0.818 pg dsm~3
                           PRESAMPLING PREPARATION

     All  glassware and polyethylene  sampling  bottles utilized in  the lab-
oratory phase  were cleaned with  an Alconox solution and rinsed with tap-
water  and distilled,  deionized  water.    The  glassware  was air-dried  and
capped with parafilm.

     The  sodium  bicarbonate-coated tubes were prepared as  outlined in ASTM
D3268.  The tubes  were cleaned with detergent,  alcoholic KOH solution,  and
distilled water.   While the  inner surface was  still  wet,  a 5 percent  (by
weight) NaHC03  solution was  poured through the  tube to  coat  the  internal
surface.   Hot, fluoride-free air  (prepared by  passing air  through coiled
copper tubing  heated  by a heating  tape) was blown through  the tube to  dry
the sodium bicarbonate on  the inner wall.

     The  Whatman  42   prefilter  and  Whatman 4  filters were treated  with
citric acid  and sodium hydroxide, respectively,  according  to ASTM D3266.
The filters were immersed  in the appropriate solution  (either  0.1  m citric
acid in 95 percent ethanol or 0.5  N NaOH  in 95  percent ethanol and 5 per-
cent glycerin)  and dried  under an  infrared lamp.   All  filters and  tubes
were sealed until  sampling occurred.   At  the  completion of  each  sampling
run, the filters or tubes  were resealed  until  recovery.   The dry gas meters
were calibrated according  to  procedures  in APTD  0576.
                             SAMPLING  PROCEDURES
Double Filter Cassette
     The  double  filter cassette  sampling train  is  a modification  of  ASTM

                                     50

-------
D3266—i.e.,  a  double  filter  cassette  is  used  in  place  of  the  AISA
Automatic Tape Sampler.  The  constituents  of  the  train  (Figure 4) were a 37
mm Millipore  filter cassette  containing the Whatman  42 filter  pretreated
with  a  citric  acid  solution  back  to  back with  the  Whatman  4  filter
pretreated  with a  sodium hydroxide  solution; a modified  Greenburg-Smith
impinger  containing indicating-silica gel; a dry gas  meter and  an orifice
meter; and a leakless  lubricating  vane  pump.

     The  sampling  rate was  0.5 cfm.    Leak  checks of  all  sampling  trains
were conducted prior to  and  after each sampling  run to  determine whether a
leak  rate greater  than  0.02  cfm existed.   The  cassettes  were  capped  to
prevent exposure to  the  ambient air.   After  sampling,  the  inlet  and outlet
were again plugged.  The used filters were placed in clean sample bottles,
and  10.0  ml of  distilled  deionized  water  and 0.1 ml  of  1.0 N  NaOH were
added.  The bottles were sealed tightly until analysis.

Sodium Bicarbonate-Coated  Tube

     Sampling  with  the  sodium bicarbonate-coated  tube was  performed  as
described in  ASTM D3268.  The train (Figure 4)  consisted  of a  4-ft glass
tube  (7-mm ID) evenly  coated with sodium bicarbonate,  connected directly
to  a 47-mm polypropylene  filter  holder containing  a  citric acid-treated
Whatman 42  filter.   The tube was  followed  by the same  drying,  vacuum, and
metering  equipment as  described  for the  double filter cassette sampling
train.

     Both ends  of the  collecting  tube were  sealed  until  sampling took
place.  After  sampling,  the  ends were capped until recovery.  The air to be
sampled  was  drawn  through the tube  at  a  rate of 0.5  cfm.   Each sampling
train was leak-checked before and after  the  sampling  run to determine that
no  leak greater  than 0.02  cfm existed.   The collected fluorides were  eluted
with  8-9 ml  of distilled deionized  water.   One  drop  of  1.0 N  NaOH was
added, and  the  solution  was  diluted to 10.0 ml.   The samples were stored in
clean bottles  and  sealed until analysis.
                            ANALYTICAL PROCEDURES

      The  samples  from  the laboratory  phase  were analyzed  for  F   on a
Dionex Model  14  Ion  Chromatograph.    This   automated  ion  chromatograph
incorporates  the  ion-separating  capabilities  of  the  ion  exchange column
with a conductimetric detection system.

      The  column  system employed  for the  fluoride  analyses  consisted of
Dionex pre-column  (3  x 150 mm)  to remove particulates,  strongly  retained
anions,  and organic species; a  separator column (3 x 250 mm)  in the HCO~3
form; and a suppressor  column  (6 x  250 mm)  in  the  H  form  to remove  the
background  conductivity of the  eluent.  The eluent, which was a solution of
0.003 M NaHC03  and 0.0024 M Na2C03, was  pumped through the  column at a  rate
of  150 ml/hr.   The injection loop had a capacity of  100 pi, and the sample
was introduced  from a 5-ml disposable syringe  fitted with a 0.22-iam Milli-
pore filter to  remove pafticulate matter.  A IN fc^SO^ solution regenerated

                                     51

-------
a. PREFILTER AND ALKALI TREATED FILTER
25 MM 	 >
PLASTIC
HOLDER







|j< 	 SHORT TEFLON PROBE
J, j I CITRIC ACID TREATED
= = = = = cS 	 PREFILTER
II NaOH TREATED FILTER
1
TO VACUUM





b. SODIUM BICARBONATE COATED GLASS TUBE
T
4ft


_1 (
<— 7 MM ID GLASS TUBE
INSIDE COATED WITH,
SODIUM BICARBONATE



1
in
/ \^ 	 pnt YPROPYI FNF
4/ MM CIIHIC ACID — >L^3*^ FILTER HOLDER
TREATED WHATMAN 42 *^ II ' "ULUtH
FILTER U
TO VACUUM
             c. VACUUM SYSTEM AND SAMPLING TRAINS
                                                                CHECK VALVE
                                                               THERMOMETER
                   ORIFICE
                 MANOMETER
                                                                       FROM
                                                                     •*• COLLECTION
                                                                       SYSTEM


                                                                    IMPINGER
                                                                    WITH SILICA GEL
Figure 4.   HF sampling  trains  utilized  in  the  program  (a)  prefilter and
            alkali-treated  filter  (b) sodium bicarbonate-coated glass  tube,
            and (c) vacuum  system for sampling  trains.
                                        52

-------
the suppressor column after  an 8-hr  period.

     Blank  filters  and tubes  were  prepared  and analyzed  as  previously
described for the  samples.   Values for all samples were blank-corrected.

     A  series  of  tests  was  conducted  to determine  the intra-  and inter-
sampling device  precision  and  accuracy at approximately 20,  30,  50, and 60
ppb HF and the sensitivity of  the procedures  with a sampling duration of 15
minutes.  This interval corresponded to  the  integration period of the ROSE
system.  Two  sets  of experiments were  performed,  one with two  of the same
sampling device  connected  to the generation  system and the  other with one
of  each sampling  device.    The results  are  presented in  Table  1.    The
accuracy—i.e.,  percent  recovery, and inter-  and intra-method precision are
presented in  Table 2.    The  mean percent  recovery of  both methods  was 100
percent over  the range  of HF  concentrations  generated.  The  mean percent
recovery for the filter  method was 99.3,  and  for the  tube  method was 102.3.
The overall  precision as  measured  by  the relative standard  deviation was
less than 8 percent.
                      PRELIMINARY FIELD SAMPLING PHASE

     The preliminary  field  phase was  designed to determine:

     1.   The compatibility of  the  selected manual sampling procedures with
          the sampling  location, i.e., the presence  or  lack of interfering
          substances, sensitivity levels,  etc.

     2.   The range of  ambient  HF concentrations at the two ponds.

     3.   If a  sampling period  of 15 minutes is  compatible with the sensi-
          tivity requirements of the  analytical techniques.

Sampling Locations

     The lines  of  sight and associated sampling  sites  for  the CFI pond and
Agrico pond are shown in Figures 1 and 2.   The length of  the  CFI line was
415  meters  and  the  Agrico line,  630 meters.    The  lines  of sight  were
divided into four  equal segments,  and  the sampling sites  were situated at
the  center of each segment.

Sampling Protocol

     Initially,  the  double  filter  cassette was  utilized at  the  four sam-
pling  sites  at CFI to  determine the  ambient HF  concentration.   All four
sites  were  sampled simultaneously.    The  sampling  rate  was 0.5 cfm,  for a
duration of 15 minutes.   Three  sets of samples were  obtained.   The filters
were treated as described  in the laboratory phase  and were sent to the GCA
laboratory for  analysis  by  1C.   The samples were analyzed the next day, and
the  results were  transmitted to the field  team.   The results  are given in
Table  3.  The values  obtained showed that  the  ambient  HF concentration was
at a satisfactory  level for the sampling  and analytical methods.   Succes-

                                     53

-------
TABLE 1.  RESULTS OF LABORATORY PHASE

HF
generated
ppb
18.2
18.2
18.2
30.8
30.8
30.8
30.8
50.1
50.1
50.1
50.1
50.1
50.1
50.1
50.1
57.5
57.5
57.3
57.3
57.3
57.3
61.5
61.5
61.5
61.5
61.5
HF found
Filter Filter Tube Tube
1 212 Filter
19
17
25
32
30
31
30
50
51
50
51
50



54 58
58 58
54 52
54 57
55 60
58 57
59 60
67 63
82 72
62 61
63 70
ppb
Tube
19
19
18
28
32
31
29
48
53
42
49
52















Tube Filter







48 47
55 48
49 52
49 50
50 49
48 50
49
50











                 54

-------
PM

|X
£
o
H
^•»


I
o


Q
§
M
U
CN


W
                                         s  s
                                          s  -
                   0 5 0
                       ! 1   s   s
            SB
1
51
                                     3 i'
                                       11111
                                                 55

-------
                             TABLE  3.   PPB;HF

Run Site
1
2
3
A
50
53
65
B
29
35
36
C
71
74
76
D
95
73
79

sive samples  showed  good reproducibility.   A concentration  gradient  along
the line of sight was also  shown to be present.  The wind  was  blowing from
the northeast with a speed  of  7-10  mph.

     The inter-sampling  device  experiments  were  then performed  as  fol-
lows:  at both ponds, two  sampling  trains were set up,  one for each device
at a sampling  site.   Because  of equipment  and  power  constraints,  sites  A
and B were sampled  together and then  followed by  sites   C  and D.    Five
replicates were  run.   The  samples  were recovered  and  returned to  GCA for
analyses by 1C.   Some  samples were  also analyzed by  the Agrico  Environ-
mental Laboratory  using  a  Technicon Autoanalyzer and  the  semiautomated
spectrophotometric procedure (ASTM  D 3270).

     The citrate-treated prefilters used  in the double  filter cassette were
also analyzed for fluoride.

Results and Conclusions

     Results  from the 1C method of  analysis are presented  in Table  4.  The
data in Table  5  indicate that the  filter  and  tube results  are comparable
when analyzed  by both the 1C  and spectrophotometric methods.   The  citrate
filter results are presented in Table 6.

     Several  conclusions  can  be  drawn from the results of the preliminary
field phase.

     1.   A sampling  period of 15 minutes was  adequate for both the filter
          and tube collection  methods for measuring the HF concentration at
          each of the gypsum ponds.

     2.   No  interferences  were observed  when either  filter or tube samples
          were  analyzed  by  either  the  1C  or  autoanalyzer  methods.    A
          previous ROSE study  has  found that a possible interferent,  SiF^,
          was not present  in the atmosphere above the gypsum ponds (10).

     3.   The  citrate-treated  prefilter was intended to remove particulate
          matter and was  not supposed to remove any HF.   The results indi-
          cate  that  essentially  no fluoride was  collected  on  the  citrate
          prefilter.

                                     56

-------
TABLE 4.  PRELIMINARY FIELD PHASE:  RESULTS AND INTERSAMPLING DEVICE

Filter
Group ppb HF



CFI
Site A
_Z
X



CFI
Site B
£_
X



CFI
Site C
_Z
X



CFI
Site D
_£
X

Group
Grand
n = 39
CFI -
n = 20
Agrico
n = 19
59
50
28
17
21
175
35
43
19
37
18
, 18
135
27.0
14
32
35
14
21
116
23.2
57
36
17
36
42
188
37.6


Z_
X
Z_
X
Z_
X
Tube
ppb HF
41
24
23
25
24
137
27.5
42
19
36
15
17
129
25.8
17
27
30
21
18
113
22.6
30
17
36
17
17
117
23.4
Filter
ppb HF
1282
32.9
614
30.7
668
35.1
d
(F-T)
18
26
5
-8
-3
38
7.6
1
0
1
3
1
6
1.4
-3
5'
5
*— 7
3
3
0.6
27
19
-19
19
25
71
14.2








Filter
ad Group ppb HF

14.2

Agrico
Site. A
Z
X


1.6
Agrico
Site B
2
X


4.8
Agrico
Site C
1
X


22.1
Agrico
Site D
£_
X
Tube
ppb HF
1074
27.5
496
24.8
578
30.4

23
29
27
23
102
25.5
54
24
32
39
36
185
37.0
35
35
39
40
40
189
37.8
19
37
43
46
47
192
38.4








Tube
ppb HF

21
24
28
30
103
25.8
32
24
23
29
38
146
29.2
32
25
31
42
37
167
! 33.4
16
24
39
38
45
162
32.4
d
(F-T)
208
5.3
118
5.9
90
4.7
d
(F-T)

2
5
-1
-7
-1
-0.3
22
0
9
10
-2
39
7.8
3
10
8
-2
3
22
4.4
3
13
4
8
2
30
6.0








°d



4.4





11.6






6.1






7.1





ad

11.3

12.5

6.5
                                57

-------
      TABLE 5.  PRELIMINARY FIELD PHASE  COMPARISON OF RESULTS FOR
                  ANALYSIS BY 1C AND AUTOANALYZER

Site
CFI A
CFI A
CFI D
Agrico A
Agrico A
Agrico B
Agrico B
S
X
°d
h=a*
— = 0.407
°d
t.975 = 2.447
Sampling
device
T
F
T
F
F
T
T







ppb/HF
1C
24
17
36
29
38
32
24
200
28.57





AA
21
38
34
27
26
35
30
211
30.14





d d2
+3
-21
+2
+2
+12
-3
-6
-11 647
-1.57
10.2
3.855




Note:  No significant difference  between the two methods of analysis
            TABLE 6.  ANALYSIS  OF  CITRATE FILTERS FOR FLUORIDE

Sample
1
2
3
4
F~ (yg/mL)
0.37
0.38
0.36
0.35
Blank (yg/L)
0.36
0.36
0.36
0.36
Net F- (yg/L)
0.01
0.02
0
0

     4.  The precision between the two manual  sampling  devices  is shown in
         Table  4.   In most  of the runs,  the  quantity of ppb  collected by
         each method  is comparable;  both methods  can be used  for  HF sam-
         pling,  since the quantities  of  HF collected are similar for sam-
         ples collected  simultaneously.
                             FORMAL  FIELD PHASE

     The objective  of  the formal field phase  of  the  project  was to compare
the  results  of  the  simultaneous   measurement  of   ambient  HF  levels  as
                                     58

-------
obtained by manual  wet chemical sampling methods  with the EPA ROSE system.
Both  sampling systems were  located  along the  edge  of the  gypsum ponds at
both  CF  Industries  and Agrico Chemical  Co.   Sampling  at  CF Industries was
conducted  on July  24th  and 25th,  1979, while  samples  were  obtained at
Agrico on July 26,  1979.

Sampling Locations

      The  sampling  line  of  sight  was  adjacent  to each  pond,  as  shown in
Figures  5 and 6.   The ROSE  van  and  light source were  aligned  visually and
the distance  between  them measured with a laser rangefinder.   At CF Indus-
tries, the  line  of sight  for the ROSE  system was 3  feet east  of  the wet
chemical sampling  line.   At Agrico, the  line of sight for  the ROSE system
was bracketed on  each side  with  the  positions of the  sampling sites being
determined by the configuration  of the road.   The  line of sight established
at each pond  was  divided into four equal segments.   One  manual sampler was
situated at  or as  close as possible  to the  center  of each segment.   The
height of the inlet of  each of  the manual sampling devices  was at the mid-
point of  the  light beam,  but did not interfere with the beam.   The loca-
tions designated  A, B,  C,  and  D were determined  by  the  restraints  of the
terrain and positions  of the electrical  generators.

Sampling and  Analytical  Procedures

Manual Wet Chemical Methods

      Two  collection  devices were used  with  the manual sampling  trains
(Figure  4);  i.e.,  (1)  a filter  cassette  containing a  citric  acid-treated
prefilter followed  by  a  sodium hydroxide-treated filter (designated F), and
(2)   a   sodium  bicarbonate-coated   pyrex   tube   (designated   T).     The
vkeuum/metering  system  was  a Research  Appliance  Corporation  (RAG)  meter
control console, which was  calibrated  according to procedures delineated in
EPA  publication  APTD  0576.    The  filters   and  tubes  were  prepared  as
previously  described.   Regardless  of  the   collection  device  (filter  or
tube), the sampling train was used at all four sample locations.   The air
was sampled at a  rate of 0.5-0.6  acfm for a  period of 16 minutes.   Twenty
runs  were conducted at  CF Industries.   All odd-numbered  runs  were executed
with  the filter cassettes,  and the even-numbered runs used tubes for sample
collection.   The  initial 10 runs  were conducted  on July  24, 1979,  and the
remainder of  the samples  at CF  Industries  were  collected  the  next  day.
Test  samples  for  18 runs  (No. 21-38)  were collected at  the Agrico  gypsum
pond  on July  26, 1979.   Two sets of samples  were  collected  with the filter
cassette,  followed by  one  run  with  the  bicarbonate-coated  tube.    This
sequence was  repeated six times.   The sampling rate and duration  was  the
same  as for the CFI runs.

     After completion of  the sampling  runs,  the  collection  devices  were
removed and the samples  were recovered as  follows:

     Filter   -  The sodium  hydroxide  filter  was  placed  in a  125-ml  LPE
                bottle;  10 ml of  distilled deionized  water and  0.1 ml of IN
                NaOH were added.   The  bottle  was  capped and swirled.

                                     59

-------
                                            W)
60

-------
                                                              o
                                                              a
                                                              4-J

                                                              so
                                                              •H
                                                              CO
                                                              (U
                                                              S-i
                                                              s
                                                              60
61

-------
      Tube    -  Two 5-ml portions  of  distilled deionized water were  poured
                 onto  the  inner surface  of  the tube;  the  tube was  swirled
                 and  the liquid collected  in  a  125-ml LPE  bottle.    To
                 preserve the  sample, 0.1  ml  of  IN NaOH  was added.    The
                 bottle was capped and swirled.

 Blank filters and tubes were also subjected to the above procedure.

      All samples  were analyzed the  day after  they  were  collected  at  the
 Agrico  Analytical Laboratory  using  the  semiautomated  spectrophotometric
 procedure with a Technicon Auto Analyzer system (ASTM D3270).  The  remain-
 ing aliquots of the CFI samples were brought back to the GCA Laboratory  for
 analysis by 1C.   Because of  difficulties  with the  ion  chromatograph,  the
 samples were analyzed approximately 4 weeks after they were collected.   The
 data for the 1C  analyses correlated  well  with the  data for  spectrophoto-
 metric  analysis,  as  was  also  shown  for  the  preliminary  phase  data.   A
 t-test indicated no significant difference between the two data sets.

 Remote Optical Sensing of Emissions (ROSE) System

      The EPA Remote  Sensing  of  Emissions  (ROSE)  system  consists  of a
 commercial   Fourier transform  infrared  (FT-IR)  spectrometer system   and
 auxiliary equipment installed in a van.   The ROSE system has been used in a
 variety of  source  characterization studies and  is described  in  detail in
 the literature  (1,11,12).   For the gypsum pond measurements, the system was
 operated in the longpath absorption mode, as indicated in  Figures  5 and 6.
 The optical system inside the van is  shown in  Figure 7.   The light source
 is  a 1000-Watt  quartz-halogen lamp.  Energy from  the  lamp  is collimated by
 a  Dahl-Kirkham  f/5  telescope with  a 60-cm-diameter  primary  mirror.   The
 collimated  infrared beam is  transmitted through the atmosphere  and passes
 through a port  in the  side  of  the  van, where an identical telescope focuses
 the beam onto the  aperture  of  the  interferometer.   The interferometer is
 part  of a standard Nicolet  Instrument Corporation Model 7199  FT-IR system
 configured  to fit into the  van.

     Major  components of the  interferometer system  consist of a  computer
 with  40K memory;  dual-density disk  with a 4.8 million,  20-bit word capac-
 ity;  teletyper;  paper tape  reader;  oscilloscope interactive  display unit;
 and a  high-speed digital  plotter.    Each traverse  of  the  interferometer
 mirror  produces  an interferogram, which is actually  the Fourier-transform
 of  the  spectrum of the incoming infrared signal.   A  liquid-nitrogen-cooled
 InSb/HgCdTe  sandwich  detector  is used to convert the  infrared  signal to an
 electrical  signal for processing.   The  InSb  detector, which  is  sensitive
 over  the 1800-6000  cm-1  region,  was  used for  the  HF measurements.   For each
 16-minute data  set,  100 interferograms  were  averaged.   The data were col-
 lected  at a  spatial resolution of  0.125 cm-1.  The computer requires about
 12 minutes to calculate  the  spectrum from the interferograms;  each  spectrum
was permanently  stored on disk for  later analysis.

 Sampling Protocol

     The manual  collection  devices were  set  into the four  sampling  trains

                                     62

-------
                                                                        e
                                                                        0)
                                                                       4J
                                                                        CO
                                                                        >»
                                                                        CO
                                                                        cC
                                                                        O
                                                                       •H
                                                                       4-1
                                                                        Cu
                                                                        O

                                                                       w
                                                                       I--

                                                                        
-------
and  initial system  readings  obtained.    The four  samplers  and the  ROSE
system  were  then  started  simultaneously.    The  sampling  and  spectral
accumulation  proceeded  for  16  minutes.     All  systems   were   stopped
simultaneously.   The final  readings  for  the manual samplers  were  obtained
and  the  spectra obtained by the ROSE were  checked.    Coordination  between
the  two  sampling methods was  handled through two-way  radio  communications
and  recording of  run start  times.

Calculations
Manual Methods

     1.
     2.

     3.
  The volume of dry gas  sampled  is converted to  standard  condi-
  tions, 777°F and 29.92  "Hg  (25°C nd  760  mm Hg).
                  VMstd = dry  std  ft3  =
                                         537  (Y)  (VM)   PB +
                                                  PM
                                                13.6
                                              (29.92)  (TM)
where    Y   = dry gas meter  calibration factor

        VM   = sample gas  volume,  ft3

        PB   = barometric  pressure "Hg

        PM   - AH, pressure at DGM "HaO

        TM   =,temperature at dry  gas  meter,  °R (°F + 460)

       537°R = 77°F -1- 460

 Vmstd, dry std., m3 ~ Vmstd  (dry  std.  ft3)  x 0.02832 m3 ft~3

 The concentration of HF in the  ambient air  is determined by:
                           -n_ /
                           F  /
            yg,
                            20.006  (HF)
                   il    L   18.998  (F)
                    -Vmstd  (m3)
            where VT  = volume  of  sample,  ml

            The concentration  in  ppb is:

                                g  HF/m3
                   , yT,,._
                 ppb HF -
                                      5  ppb
            T,OT.0  n  0,0--  20.006  yg/  mole x 109  L/m3
            where  0 . 818 -   24.45  yL/  mole x 109 ppb-l
                                     64

-------
     5.      The  four  results from each  manual sampling run  were averaged
            arithmetically  and geometrically.
                             X Arithmetic = EX.
                                            n-11

                X Geometric =  [(XA)(XB)(XC)(XD)

Calibration of ROSE System Data

     Calibration of ROSE system  absorption spectra obtained in the field is
normally  done by  measuring  the  absorption  spectra  of  known amounts  of
various  gases  contained in  a  calibration  cell (shown in  Figure  7).   The
transmittance of the gas sample  as a functipn of  wavelength (or wavenumber)
is related to the cell  length and  gas concentration by Beer's Law:

                                 .,  ,    -K(v)CL
                              T  (v)  = e
where  v    = wavenumber  (cm~l)

       C    = concentration  (ppm)

       L    = path length (meters)

       K(v) = spectral absorption  coefficient  (ppm meters)   .

Once K(v) has  been determined from  the  calibration spectra,  its  value can
be used with longpath absorption spectra  obtained in the field to calculate
the path-average gas concentration (again using Beer's  Law).   For the field
spectra, the path  length  L from the source telescope to.the  receiver tele-
scope is measured using a laser range-finder.

     Because  of  its high reactivity, HF requires  a  special  gas  handling
system for  filling  calibration cells.  Such  a system was not  available at
the EPA laboratory.  Therefore, another method, which  is based oh measuring
the area under the  absorption curve  of the  spectral line in  question, was
used  to  assist  in determining  K(v)»   The  particular  advantage of  this
method is  that  the area  un4er  the absorption  curve is independent  of the
spectral resolution used, which allows the use of HF  data obtained  at low
resolution  (13,14) to be  used  as calibration  data.  From the  Ipw resolution
data, the  relationship  between the  area  under  an absorption  line and the
igas optical depth  (product C x L)  was determined  (private  communication of
D.E.  Burch and  D.A.  Gryvnak,  Aeronautic Division of  Ford  Aerospace  and
Communications Corporation), and  the resulting  calibration  curve for the
R(5)  line  of  HF at  4174  cm-.l  is  shown  in  Figure  8.    The  R(5)  line was
selected for the concentration  calculation and because  it provides the most
suitable compromise  between maximum  line  strength'and minimum water vapor
interference.  The spectral  signatures of clean, air and the'. gypsum pond are
shown in the  4168  to  4178 'cm-1 region in Figure 9.   It is seen that there

                                    65

-------
o'ro
                                                                                0)
 q

 ii

 J.


 Q
 LU


 UJ

 Q

 <

 O
 cc
 CO
E
Q.
_

O
        m
        •JS
        O

        M-l
                                                                               o
                                                                               •H
                                                                               •H
                                                                               .H
                                                                               O
                                                                               •H
                                                                               •U


                                                                               2
                                                                               a
                                                                               o
                                                                               o
                       3Auno NOiiduosav aaaNn vaav
        oo

        0)
        J-J

        §,
                                       66

-------
                         CLEAN AIR  L = 900 METERS
          100
        LU
        O



        1

        ^
        CO
        •z.
        <
        on
           25 -
          TOO H
           75 -
        LU
        O
        E 50

        ^
        C/3
        DC
           25 -
                           GYPSUM POND L = 630 METERS
4168     4170
                                  4172      4174

                                 WAVENUMBERS
4176      4178
Figure 9.  Clean air and gypsum  pond  spectra (average of Agrico ROSE Runs

           32-38.
                                     67

-------
 are weak  H20 lines at  approximately 4173.6 and  4173.9 cnr1.   The  method
 used to eliminate  the HaO interference and  determine  the  value of K(v)  is
 described below.

      First,  the  spectral  data  from each  of the  four measurement periods
 (two mornings at CFL, and  a  morning and afternoon at Agrico) were  averaged
 to  produce  four  single  spectra  with  excellent  signal-to-noise  ratios.
 (Figure 9  shows  the average of ROSE  system runs  No. 32-38.)   Next,  the
 clean air  and  gypsum pond absorption  spectra were converted to absorbance
 [-LogT (v)].   The clean air spectrum was then multiplied by  a factor  deter-
 mined so  as  to make  the  maximum  absorbance  of   the  water vapor  line  at
 4176.4 cm-1  equal  in  both spectra  (Figure  10).   In this  way,  the optical
 depth (C  x L)  of  HaO was  made equal  in each  spectrum.    The background
 spectrum was  then  subtracted from  the pond spectrum  and  the  result  con-
 verted back to transmittance  (Figure 11).   (All of the mathematical  opera-
 tions  were carried  out with  standard system software.)  The area under  the
 HF absorption line  was then measured manually for  each of  the four  spectra,
 and an HF  concentration was  determined from Figure 8.   A value  of K(v)
 could then be  calculated from  Beer's  Law for  each of the  spectra,   since
 T(V),  C,  and L were now  known.   The four values  obtained  were 5.14,   5.04,
 5.29,  and  5.09 x 10-3 (ppm meters)-!.   The  average of  these values, 5.14 x
 10~3,  was  taken as  the value of K(v) for the R(5) line of HF.  Each indivi-
 dual pond  spectrum  could then be calibrated in the following manner.

     Each  individual  pond spectrum  was expressed  in  absorbance  over  the
 4168-4178  cnrl region, and  the clean air  (background)  absorbance spectrum
 was adjusted as described  above to make the H20  line  at  4176.4  cm"1 have
 equal  optical depth for each pond and  background  spectra.   The two spectra
 (each  pond and the background)  were then superimposed, as  shown in Figure
 12,  in such a manner  as  to best match the spectral features on each side  of
 the HF line.    The  absorbance at the HF line center was  measured directly
 from the superimposed  plots (difference between pond and background spectra
 at  the  HF line  center).    For the example shown (ROSE  Run  No.   38  at
 Agrico):

                     Absorbance at  line center  = 0.061

                                             T  = 10-°-°61  - 0.869

                                           KG L  = -Ln T     = 0.140

                                             KL = 5.14 x 10-3 x 630 = 3.24

                                              C = 43 ppb

The  principal error in  this  measurement  is  in the -baseline determination
 (i.e., the mismatch of the  spectra  on each side  of the  HF  line).   The  maxi-
mum  error  is  approximately ±0.005 absorbance units, which  is equivalent to
approximately ±3.5 ppb.
                                     68

-------
                             CLEAN AIR ADJUSTED
         2.0 -i
         1.5 -
      o
      g 1-0

      O
      to
      CD
          .5 -
          0
         2.0 -i
                                   I          r

                          GYPSUM POND
                                             A
           4168       4170       4172      4174

                               WAVENUMBERS
                                                     ABSORBANCE OF

                                                     THESE TWO LINES
                                                     MADE EQUAL
                                                          I
 A
  I           I

4176       4178
Figure 10.  Adjusted clean air and gypsum pond  absorbance spectra.


                                      69

-------
             100-n
          -I 94 -
          LU

          o
          CO


          <
          oc
                                                       BASELINE ERROR

                                                       RANGE = ±2 PPB
              4173.0      4173.5      4174.0  I    4174.5


                                  WAVENUMBERS
4175.0
Figure  11.   Transmittance of HF  line after background  subtraction.




                                      70

-------
                      .-  05 O

                      O  CO *—
                      d  o o

                      ii  n n

                      CD v-» 7-?
    CO
    ro
    •z.

    DC
            Q
O
cn
            o
            2
0>
u
c
CO
.n

o
                                                                                             o

                                                                                             (U
                                                                                             0)
                                                                                             4-)
                                                                                             C
                                                                                                         o
                                                                                                          CM
                                                                                                         EC
                                                                                                         d
                                                                                                         o
                                                                                                         •1-1
                                                                                                         4J

                                                                                                         §
                                                                                                         •rl
                                                                                                         r-l
                                                                                                         0)
                                                                                                         0)
                                                                                                         o
                                                                                                         CJ
                                                                                                         ctf
                                                                                                         O
                                                                                                         CD
                                                                                                        •s
                                                                                                         (!)
                                                                                                         O
OL'O
    80'0
           90'0           WO

              aoNvaaosav
                                                                                   o
                                                                                                         o
                                                                                                        •H
                                                                                                        J-l
                                                                                                         cS
                                                                                                         CJ
                                                                                                        •H
                                                                                                         a
                                                                                                         S-l
                                                                                                         0)
                                                                                                        +J
                                                                                                         0)
0)
S-i

§,
•H
                                                  71

-------
ROSE System Sources  of Error

     Potential  sources  of error  in  ROSE  system measurements  have  been
studied extensively  under  both laboratory and field conditions.  The labor-
atory studies have  addressed system reproducibility  and  calibration error.
The reproducibility  of the system  was  tested by collecting 10 separate sets
of interferograms at 0.125 cm-1  (50 in each set) on  the  same gas sample in
a cell.  Each set was  transformed  to give a single spectrum.   It was found
that  the  peak  height  of  individual absorption  lines  was  reproducible  to
within  ±1 percent   of  the peak  height  value.   Conventional  calibration
procedures  were  evaluated by filling  the calibration  cell  to  the  same
nominal pressure  several times and collecting a set of  interferograms after
each fill.   In this case the error  was determined by  the  accuracy  with
which the pressure gauge in the  gas  handling system could be read (about ±5
percent).   The error  in  the  area method  used for  the  HF  calibration  is
certainly no worse than  this value.

     Error  in  field measurements  tends  to be  greater than  in  laboratory
measurements  simply because,  as  the  light  source  is  moved further  and
further  from   the   van,   less  energy  is  collected,  and  the  S/N  ratio
decreases.  The most reliable  test of the  system  for field measurements is
to compare from run  to run the spectra obtained for  the  gases CO2 and N20.
Since these species  have essentially constant  concentrations, comparison of
spectra  from different  runs  gives  a measure  of  the overall  instrument
reproducibility.   Eight runs  from Agrico  were studied.   At  the  spectral
region of maximum S/N,  the maximum 'variation  in peak  height  for eight C02
and eight N20 lines was  5.5  percent  of  the  peak  height.   Because  of the
fall-off  in  detector sensitivity  toward  shorter wavelength,  the  S/N ratio
is about four times  less at the  region of HF absorption than at the regions
of C02 and N20  absorption. Another C02 band, located  where  the  S/N is the
same as HF, was similarly  studied.   For  eight spectra,  the  maximum varia-
tion was  14.2 percent  of  the peak height.   A combination  of  the reproduc-
ibility  error with  the  calibration  error leads  to  an  estimate  that  the
error in any single  HF measurement (average of 100 interferograms) would be
±15 percent.

     The  two  manual sampling  methods for  the collection  of HF  have  been
Studied in  the  laboratory  and field and  the  results  of  these experiments
are presented above.  The  percent  recovery of HF  for each  method was shown
to be about 100 percent, and the precision for each sampling device for the
laboratory phase  is  given  in Table 2.   For both sampling  devices, the rela-
tive  standard  deviation  was  less than  10 percent  for  HF  concentrations
above 18  ppb.   The  percent recovery and precision,  as shown  in  the above
tables,  are  a  reflection  of  the  sources  of  error  in  both  the  sampling
devices and the analtyical method.  In the  laboratory  phase,  ion chromato-
graphy was used to analyze the samples  for HF.

     The precision  obtained in the preliminary field phase is presented in
Table  4.    Again,  ion  chromatography was  used  to determine  HF   in  the
samples.

     For the  formal  field phase,  the  samples  were analyzed  by the colori-

                                     72

-------
 metric method with  an Technicon Autoanalyzer.   The precision  and  accuracy
 of the semiautomated  method have been documented  in ASTM-Method D3270.   A
 collaborative  study   by  nine  laboratories  using  the  method   for  the
 determination of HF in vegetation gave relative  standard deviations ranging
 from 4 to 13.4  percent  of  different   types  of   vegetation.     Replicate
 analyses of  standard  NaF  solutions  by  four  laboratories  had  relative
 standard deviations of 11.4, 3.9,  and 3.0 percent  for  solutions  containing
 0.28, 1.41, and  2.81  yg/ml, respectively.   Replicate analyses of  standard
 NaF solutions  by four  laboratories  showed average recoveries  of  101.8,
 101.4, and 100.7 percent  for  solutions containing  0.28,  1.41,  and 2.81  ua
 F/ml, respectively.
                           RESULTS AND DISCUSSION

      The results are  presented in  Table 7  in the  sequence  in  which  the
 samples were  collected.    Graphical   representations   of  the  data   are
 presented in the following figures.

         • Figure 13 plots the HF concentration measured at each site  by  the
           manual methods, the arithmetic average of the manual methods,  and
           the ROSE  system.

         • Figure 14 is  a plot of  the HF concentration as measured by  the
           ROSE  system  vs.  the  HF  concentration  as  determined  by   the
           arithmetic average  of  the manual  methods.   The  theoretical  1:1
           correspondence line  is  indicated.

      The ROSE system measures  the average concentration  of HF molecules in
 the  30-cm-diameter   cylinder   extending   through  the  atmosphere  from   the
 source  to the receiver telescope.    The  point sampling systems  measure  the
 point concentrations of  HF.   Both  the optical and point method measurements
 were  averaged over  16-minute  time  intervals  for each sampling  period.   If
 the HF  concentration were relatively  uniform  along  the  sampling  path,
 fairly  small  differences would be  expected  between values obtained  at  the
 four  sampling  sites  during   a  given   sampling  period,  and  reasonable
 agreement  between  an  average  of   the   sampling site  values  and  a  ROSE
 measurement   would   be  expected.    If,  on  the  other  hand,  there  were
 appreciable  concentration gradients  along  the path, then the  two  methods
 could give widely differing  results without  either being  "incorrect."   The
 situations that  best illustrate this are:  (1)  a  spatially small but  high
 concentration HF pocket  could slowly  traverse the area  of a  single  point
monitor  (the  result  would  be a high  reading at one  site,  but  no appreciable
affect  on the   ROSE   data);   and  (2)   an   extended   pocket  of  high  HF
concentration  that  slowly passed  through the optical  path but missed  the
point monitors  (obvious  results).

     Inspection  of data  shown  in  Figure  13,   with the  above in mind,  shows
the following:

     (1)  During  the  sampling at  CFI  on July 24  the  HF   concentration
          spread between  sampling  sites  is at maximum about ±35 percent  of

                                    73

-------
       TABLE 7.  HF CONCENTRATION DATA  GROUPED IN SEQUENCE OBTAINED

Concentration (opb)
Manual sampling
Dace
7/24/79
(CFI)





Mean
7/25/79
(CFI)







Mean
7/26
(Agrico)














Mean
Grand
mean
GCA
run
no.
4
5
6
7
8
9
10

12
13
14
15
16
17
18
19
20

21
22
23
24
25
26
27
28
30
31
32
33
34
35
36
37



Sampling
start
time
1043
1124
1200
1224
1246
1307
1330

0830
0940
1005
1027
1050
1115
1145
1205
1225

1040
1106
1131
1150
1210
1231
1252
1311
1550
1608
1630
1635
1722
1742
1809
1827



Site
A
27
32
44
35
46
44
42
38.6
19
34
14
32
22
52
17
43
27
28.9
20
21
22
34
34
31
35
37
30
31
23
32
34
26
25
26
28.8

31.0
B
44
32
36
57
24
48
40
40.1
15
58
98
34
106
37
7
54
31
48.8
19
29
27
29
26
29
34
—
36
27
—
33
30
30
22
26
28.4

39.1
C
52
35
•: 	
50
40
43
30
41.7
22
43
25
51
—
47
44
60
17
38.6
—
29
27
32
36
34
35
—
32
46
27
32
33
30
29
30
32.3

36.1

D
27
32
36
50
30
45
40
37.1
51
48
29
75
24
42
24
53
24
41.1
31
40
42
43
45
42
46
45
50
43
38
52
47
39
34
35
42.0

40.7
Arith-
metic
mean
38
33
39
48
35
45
38
39.4
27
46
42
48
51
45
23
53
25
40.0
23
30
29
35
35
34
37
41
37
37
29
37
36
31
28
29
33.0

36.4

Geo-
metric
mean
36
33
39
47
34
45
38
38.0
24
45
32
45
38
44
19
52
24
34.1
23
29
29
34
35
34
37
41
36
36
29
36
36
31
27
29
32.3

34.0

Collec-
tion
device*
T
F
T
F
T
F
T

T
F
T
F
T
F
T
F
T

F
T
F
F
T
F
F
T
F
T
F
F
T
F
F
T



ROSE
method
42
39
40
43
30
43
32
38
27
28
32
39
: 28
29
38
38
43
33
21
23
32
35
36
41
34
38
39
41
37
46
46
37
41
38
36

36







.4









.6
















.6

.1

*F = Filter cassette.
 T = Bicarbonate - treated tube.
                                       74

-------
CO
O
co
              C/5
              Q
              O
              I

              LLJ
                                                                   UJ
                                                                   ^1

                                                                   111
                                                                        II  "  "
                                                                        x O <1
                                                          b. —
                                                                                 U.

                                                                                 U.
                                                                                 b.


                                                                                 U.
                                                                                 U. —
                                                       LU  I- '
                                                       CQ
                                                       =>  "- —i


                                                       Q  >-•
                                                       LU
                                                                             LU  |- -
                                                                             QC
                                                                           LU I-

                                                                           Em
                                                                         "Ji to b
                                                                         yw<  H-
                                                                         — _ —  U.
                                                                         S^?,
                                                                         O
                                                                         O
                                                               co
                                                               to
                                                               oo
                                                               us
                                                               co
                                       co
                                       CO
                                       CO
                                       CM
                                       CO

                                       CO

                                       o
                                       CO
                                                               00
                                                               CM
                                                               r~
                                                               CM
                                                               CO
                                                               CM
                                                               LO
                                                               CM

                                                               "fr
                                                               CM
                                                               CO
                                                               CM
                                                               CM
                                                               CM

                                                               CM
                                      o
                                      CM

                                      O)

                                      00



                                      CO

                                      in


                                      T—
                                      co

                                      CM
                                                                                      CO
                                                                                      co
                                                                                      LO
                                                                                              o
                                                                                              •H
                                                                       0)

                                                                       a
                                                                       4-1

                                                                       CO

                                                                       O
                                               cd

                                               CO

                                               o

                                               4-1
                                                                                              cd

                                                                                              g
                                                                                              "S
                                                                                              cfl
                                                                                              w
                                                                                              w
                                                                                              •S:
                                                                                           ^ CO
                                                                                           r» co
                                                                                              ctf
                                                                       ca
                                                                       d
                                                                       •H

                                                                       m
                                                                       d
                                                                       o
                                                                       •H
                                                                       4-1
                                                                       2
                                                                      . o

                                                                       g.
                                                                       co
          o
          
-------
                 1°          20          30           40           50          60

          CONCENTRATION (ppb) - MANUAL (ARITHMETIC AVERAGE AVERAGE OF SITES)
Figure 14.  Composite comparison of HF concentrations measured  by two  tech-
            niijues.


                                    76

-------
           the  average value for  a sampling period.   The agreement between
           the  ROSE and  average point values  are all within the estimated
           ROSE  error, and furthermore, each method  follows the same up and
           down  trends.

      (2-)   During the sampling  at  CFI on July 25,  the fluctuations between
           point  measurement  sites during a given sampling period were much
           greater  than  on  July 24,  and the  highs  and  lows  vary between
           sites  for different  periods.   The  point measurements  thus indi-
           cate  widely varying HF concentrations.  As might be expected, in
           this  case the  agreement between the two methods is not as good as
           on July  24.                                                      ;

     Analysis of the  measurements at Agrico shows an appreciable difference
from CFI in that a true  HF gradient  along the sampling path is indicated by
the point  measurements.   Sites  A and B are generally the lowest and Site D
always  the highest  (by  an appreciable  amount)   in  HF concentration.   The
readings at Site D can perhaps be explained  by  the  fact  that  the  site was
next  to a  small  stream of liquid leaking from an upper  gypsum pond.  Also,
it should  be  noted that the truck holding  the ROSE  light  source and teler
scope  was  in an area moistened  by  pond runoff.   The  data fall  into  two,
categories:

      (1)   During the  first  eight sampling periods (1038  to 1311 hours)  the
           agreement  between  the methods  is  excellent.

      (2)   In the eight  late  afternoon sampling periods (1548 to 1830 hours),
           the ROSE data  gave HF concentrations always slightly higher than
           the average of  the  point sampling data.                         ':

     The original  ROSE data  for the  last  six runs  at  Agrico  were repro-1-
cessed  a  number of times to  eliminate  the  possibility  that  human  error
could have caused  the difference -between the point and ROSE values.  Also/
the C02 and  N20 concentrations  measured by  the  ROSE system were  compared'
for the last six runs.  No mistakes  in data reduction were found for the HF
measurements.  The C02 and N20  concentrations  were all within ±5 percent of
each other (respectively) when  measured  in  spectral  regions of maximum S/N.
In the spectral  region,  where the signal-to-noise ratio  was about  the same
for C02 as for HF,  the apparent  C02  concentration varied about ±15  percent.
The internal checks on the spectral  data provided by C02  and N20 show that
there  is  nothing  abnormal about the  spectra  measured  by  the  ROSE system
during the last six  sampling  periods at Agricb.   The  consistently higher
values obtained by the ROSE  system during the warm afternoon may have been
due to the pond runoff in the vicinity of the  light  source.

     To determine  whether any  statistical  differences existed  between  the
sampling sites at  CFI and Agrico,  an analysis  of  variance at the 95 percent
level was  calculated  for  the  data obtained  each  sampling  day.   The results
of these analyses  indicated no  differences  among  the  sites at CFI;  however,
a significant difference  existed  among the  sampling  sites at Agrico.

     Statistical analyses based on the differences between  the  manual sam-
                                     77

-------
   TABLE 8.  STATISTICAL ANALYSES BASED ON DIFFERENCE VALUES FOR MANUAL
             SAMPLING DATA AND ROSE DATA

Concentrat ions

All data
CFI
7/24
CFI
7/25
Agrico
Filters
Tubes
n
32
7

9

16
17
15
d*
+8
+1

+6

-3
+1
-0
.3
.0

.4

.6
.2
.53
Sdf A$
9
4

14

5
8
10
.7
.8

.5

.6
36.4
39.4

40.0

33.0
.96 37.5
.3
35.07
R§
36.
38.

33.

36.
36.
35.
1
4

6

6
6
6
1.
0.

0.

1.
1.
1.
in ppb •
• 1.18

0.55 -> 1.72

0.92 •*• 1.46
0 . 64 ^ 1 . 46
0.55 -> 1.72

*d (average
of
tsd (standard
differences) =
deviation
£(A-R)/n
of differences)

= [S




(d-df^n-l]^
$(A) - mean of data for manual methods

§R « mean of ROSE values
    (R/A) - Z(R/A)/n
                                    78

-------
pling data and the ROSE values are  presented  in Table  8  for all of the data
and for  each  day  of  sampling.  The difference between  the  arithmetic mean
of the manual methods (A) and the  ROSE data (R)  was computed  from the data
in Table 7.

      The mean HF concentration determined by the manual  sampling  methods
for all  of the data  was 36.4 ppb,  compared to a  concentration  of  36.1  ppb
calculated by  the ROSE system.   The  standard deviation of the difference
was  9.7  ppb,  with  the  variation  at  CFI  on July  25,  1979,  contributing
significantly to  raise  the overall  standard deviation.

      Theoretically, random  errors  should  give an overall  d that is zero or
very close to  zero.   The small 3  indicates  a good  correlation  between the
ROSE system  and  the manual  sampling  methods.   In addition,  the  overall
average  of  the quotient  (R/A) should  be  about  one if the data  from  the
manual sampling  methods and  the  ROSE  system is   comparable for  any  given
run.  The values  for (R/A) approach one for the data.
                              ACKNOWLEDGMENTS

     Research herein  was  supported by U.S. Environmental  Protection Agency
Contract  68-01-4143,  Task Order  59.    This  paper  is  abstracted  from  U.S.
Environmental Protection Agency publication  340/1-80-019.

     The  cooperation  of Mr.  Harold Long,  Manager of Environmental  Control
and Mr. Maurice  Johnson,  Environmental Control, Agrico Chemical  Co.,  South
Pierce,  Florida;  and  Mr.  William  Schimming,  Director  of  Environmental
Affairs,  CF Industries, Bartow, Florida,  is  gratefully  acknowledged.

     Dr.  Jay  S.  Jacobson and Larry Heller of  The Boyce Thompson Institute
for Plant Research  at  Cornell University  supplied advice  and the HF genera-
tion apparatus used in the laboratory  phase  of  the program.

     The  analyses  of  the formal  field  phase  samples were conducted in the
Agrico Environmental  Laboratory  by Mr. Ed Germain and Mr.  Charles  Kinsey.
Their help is gratefully appreciated.
                                REFERENCES

1.    Herget, W.F.,  and  J.D.  Brasher.   1979.   Applied Optics 18:3404-3420.

2.    Jacobson,  J.S.,  and L.H. Weinstein.   1977.   Journal  of  Occupational
      Medicine  19:79-87.

3.    American  Society for Testing  and  Materials.   1978.   Annual  book of
      ASTM  standards,  D3266-D3270.

4.    Intersociety  Committee.   1972.   Methods  of  air  sampling and analysis.
      American  Public  Health  Association.
                                     79

-------
5.



6.



7.



8.




9.



10.




11.,


12.

13.



14.
Kommers,  F.J.W.
Institute    for
N79-16435/6WP.
 1976.    Determination  of  fluorides.    Research
Environmental    Hygiene,    Delft,    Netherlands.
Israel,  G.W.   1974.  Evaluation and comparison of  three  atmospheric
fluoride  monitors  under  field  conditions.   Atmospheric  Environment
8:159-166.

Okita, T., K. Kaneda,  T.  Yanaka, and R. Sugar.  1974.   Determination
of  gaseous  and  particulate chloride  and  fluoride  in the  atmosphere.
Atmospheric Environment 8:927-936.

Glabisz,  U.,   and  Z. _ Trojanowski.     1976.     Metody   oznaczania
nieorganicznych  zwiazkow  fluoru  w  gazach  odlotowych  z  instalacji
kwasu  i  nawozow fosforowych oraz  w powietrzu atmosferycznym.   Prace
Naukowe Akademii Ekonomicznej  We Wroclawiv No.  9/113/:253-257.

Jacobson, J.S.,  and L.I.  Heller.   1976.   Evaluation  of probes  for
source  sampling ,of  hydrogen  fluoride.   APCA  Journal  26  (11):1065-
1068.

Boscak,  V.,  N.E.  Boune,   and  N.  Ostojic.    Measurement  of  fluoride
emissions from gypsum ponds.   Draft final report.   TRC  -  The Research
Corporation.    Prepared  for  U.S.  Environmental  Protection Agency.
Contract No. 69-01-4145.
Herget, W.F.,  and J.D. Brasher.
514.
               1980.   Optical  Engineering 19:508-
Herget, W.F.;  1982. . Applied Optics  21:635-641.        •

Randall,  C.M.   1975.   Line-by-line calculations  of  hot gas  spectra
including HF  and HC1.   SAMSO  Report  TR-75-288.   Air Force  Systems
Command, Los Angeles, CA.  December.

Meredith, R.E., and F.G.  Smith.   1974.   Broadening  of  HF  lines  by H2,
D2, and Na-  J. Chem. Phys. 60:3388-3391.
General Reference

Ferraro,  J.,  and  J.  Basil,  eds.    1979.    Fourier  transform  infrared
      spectroscopy,  applications to  chemical  systems,  vol.  2.    Academic
      Press, New  York,  NY.  Chapter  2  - Trace gas  analysis,  by P.  Hanst;
      and Chapter  3  - Air pollution:   ground based sensing  of  source  emis-
      sions, by.W.R. Herget.
                                     80

-------
                 RESULTS  OF  THE SYNTHESIS AND SOLID SORBENT
                    EVALUATION OF SOME POROUS COPOLYAMIDES
                 Sajal  Das,  Louis  A.  Jones,  and John E. Bunch

                            Department of Chemistry
                        North Carolina State University
                                  Raleigh, NC

                                James D.  Mulik

                    U.S.  Environmental Protection Agency
                       Environmental  Protection Division
                          Research Triangle  Park,  NC
      Though  Tenax-GC is currently  the  polymer-of-choice  for  air sampling,
 it  has  demonstrated poor adsorptive  capabilities for  low molecular weight
 compounds  (1).   Of  particular  concern is  its  low affinity  for aldehydes,
 nitriles,  and amines.   Within these  classes  are acrolein,  acrylonitrile,
 and  dimethyl  amine;  the  adsorption  of  these  compounds  by  Tenax-GC  is
 observed  to  be  insufficient (1).   Acrolein and acrylonitrile  are  known to
 be  mutagenic, while dimethylamine is suspected  to be  a potential precursor
 to  dimethyl  nitrosamine formation in the atmosphere' (2).   Thus, there is a
 need for  formulation of new, selectively adsorbing materials  to permit the
 determination  of these  trace  low molecular weight  toxic'ants  from ambient
 air.    Similarly,   Tenax-GC does  not  efficiently  trap  propylene 'oxide,
 actaldehyde,  allyl  chloride,  vinyl  chloride,  vinyl  bromide,  1-hexene,
 methyl  chloride, and  methyl bromide.   These  toxic  organic  compounds have
 been found in trace  amounts  in  ambient  air  samples  (1).

      The  mechanism  of adsorption is a  complex phenomenon and  has  not been
 completely  elucidated.   Physical  adsorption   occurs  by  impaction of  the
 vapor on  the  surface of the  adsorbent.   Polymer porosity,  rigidity, surfape
 area, pore volume,  pore size distribution,  and particle §ize of the ambient
 air  "aerosol"  have  been specified as  controlling  factors  that can affect
 the  separation  performance of  porous  polymers (3).    It has also' been
 observed  that  separation is more a functiph of  the  surface  nature  of the
 porous  polymer  other  than related  to .its  microppre volume or  average
 micropore  size  (4).    Alternatively,  chemisorpti'pn  can  be  related  to
 interactions  involving TT-electron density,  dipole-dippl^  interactions, and
 London  dispersion forces, all  of  which  Depend  on the 'chemical nature of the
 sorbent  as  well as  the adsorbate.    From thermodynamlcal  consideration,
/ chemisorption is stronger  than physical adsorption.,  .The  ability pf Tenax-

                                      81

-------
 GC to adsorb  the higher molecular  weight alkenes  and  aromatics, aromatic
 halides,  alcohols, acids,  and  amines suggests  that  the  chemisorption pro-
 cess  is  operative and  important,  but  is  not  sufficiently  controlling to
 adsorb  compounds with diminished ir-bond or  dipole-dipole interactive prop-
 erties.

      In considering the desirable properties  of Tenax-GC,  the high thermal
 stability,  ability  to  be  easily  handled,  lack of  demonstrated catalytic
 activity,  and  hydrophobicity are of  importance  (5).  As indicated previous-
 ly, the lack of  affinity for lower molecular weight toxicants suggests that
 a  higher ir-system/polar  group  polymer  is  needed.   The argument  might be
 made  that  the  2,6-diphenyl substitution on the polyphenylene oxide provides
 this  needed  it-system.    However,  the  resonance polar  property  of  these
 phenyl  groups  is low,  as reflected by the low Hammett cr-value of the phenyl
 group (crm » 0.06, CTp = -0.01)  compared  to that of nitro group  (crm - 0.71,
 Op =  0.78,  o^J  «  1.27)  or the dimethylamino group  (oi  =-1.7)  (6).  A porous
 polymer with either strongly electron-withdrawing  substituents  (as  a nitro
 group)  or  strongly electron-donating  (as  a  dimethylamino  group)  should
 exhibit  a  higher polar  affinity for  the  lower unsaturated ^and/or  polar
 compounds  that constitute a part of  the organic toxicants of ambient air.

      A  desirable porous polymer  should  contain or possess  (a)  an aromatic
 system  (for thermal stability),  (b)  polar groups  (for  enhanced chemisorp-
 tion),  (c) high molecular weight  and  tractability,  (d) high  hydrophobic
 properties,  (e)  non-catalytic  or chemical activity  relative to  the  toxi-
 cants,  (f)  narrow  molecular  weight distribution,  ,(g)  highly  amorphous
 structure,  (h)  non-homogeneous  surface  (7),   (i)  minimum  micro-void  to
 obtain  reproducible results, and  (j)  facile  incorporation of  functional
 groups  to modify specificity.

      Our  approach to a systematic study of  the  effect  of  polar functional
 groups  on the sorptivity  of low  molecular  weight  polar compounds was  to
 prepare  a series  of  copolyamides incorporating N,N'-bispropargyl-4,4'-di-
 aminodiphenylmethane  (8),  the corresponding Qbenzophenone and 4,4'-diamino-
 diphenylamine.   Thus the  effect of  -CH2~,_U_,   and  -NH-  could  be  deter-
mined, and  this  report  describes  the  preparation and sorptive properties of
 these new porous polymers.
                                EXPERIMENTAL

     The  copolyamides were  prepared  by a  modification  of  the  procedure
reported by Wolfe  (8) and will  be described in detail at  a later  date.   Of
importance to  subsequent  copolyamide preparations (Copolyamides I,  II,  and
III) was the  compound N,N'-bispropargyl-4,4'-diaminobenzophenone  (VII)  and
the  synthesis  utilizing  the  oxidation  procedure of  Bell  (9),  which  is
summarized in Reaction Scheme I.   Polymers  A and B were kindly made avail-
able by Dr. James  Wolfe of Virginia  Polytechnic Institute  and State Univer-
sity.  The structures  of  the polymers  and  copolyamides  used in this  study
are shown in Figure  1.

     In general, to  prepare  the bispropargylated copolyamides,  a mixture of
                                     82

-------
                         Reaction Scheme 1
   H2N
D)
                                ACOH/AC-O
-NIL
   H3C-C-HN
                        II
                           Cr00/ACOH
                              -J
                             [0]
                                                  H
                                                  r+
                                               Hydrolysis
                       III
                           THF
                                                       (i)  NaH

                                                       (ii) G
HN
 K
 CHCH
          VII
  NH ,;KOH
  I     EtOH
                     C=CE
' CF-^-CHN.

    CH,

    CECH
                            VI
                                    0
                                    II
                                  NHCCF
                                     HC=C
                                 83

-------
                       CH,0
                       I 3!l
                     I  If,  \     f,  \  I 3
                    _N-/'  V-cH2—'/  y-*t-
                               Polymer A
  0 OH,

rc-*\_j~™z~
                                   CH.O
                                   I 3U
                                    — O
                               Polymer B
                   CH,-C=CH
                                          tfCHC-CH.
               ..-/Vc^yN-
                               l
                           HCHC-CE,
                                                    CH2-C=CH
             1   /
       H  0

       N-C
'c-lH^T\t/r\-*-l
                                Copolyamide I
                                                      0      0
                       H P
                                   CH2
                                     -CHCH
                                                  CH2-C=CH
                       N—C
                 C—N
                                Copolyamide II
                  H      HO
                                  II     /?
                                  _N_/
                                Copolyamide
                                                   CH,-C=CH

                                                     0
Figure 1.   Polymers A and B - Copolyamides  I,  II, and III.


                                    84

-------
VII and  the  desired  aromatic diamine (ca.  3:1) was dissolved in redistilled
N-methylpyrrolidinone  (NMP)  and to  this was  added isophthaloyl chloride in
NMP at  elevated temperatures.  When gel formation was  observed,  the reac-
tion  was  quenched  with  methanol,   the polymer  filtered,  washed  and/or
extracted  with methanol, dried under vacuum,  and sieved.   The  60/80 mesh
particle size  was  collected, packed  in  1/8"  nickel columns  (6  or 3 ft) and
the columns  conditioned overnight at  150°C under 5 ml/min  helium.   Break-
through  volumes  (BTV's) were determined by the method  of  Brown and Purnell
(10) and consisted of  determining the retention  volumes  from  110°-160° at
flow rates of  10 ml/min  of  helium  for  each  10°  increment, plotting  log i
vs.  T  (K°)  and  extrapolating to  25°G.
v
                           RESULTS  AND DISCUSSION

     Although Polymer A does  not  contain N-propargyl groups, it was consid-
ered  a necessary  "base compound"  that  could be  compared in  its  sorptive
characteristics  to Tenax-GC.   This  compound, a fibrous  polymer,  could not
in itself be packed  in  a chromatographic (1/8" x  6')  nickel column for BTV
evaluation  due  to its   non-rigid  properties.   As  a  consequence, it  was
coated on Gas  Chrom Q (15 percent load  level) and its  BTV  values  evaluted
for  hexane, benzene,  chloroform, methanol,   and  acetone.   Comparison  of
these values with  those of the Tenax-GC  showed that,  with the  exception of
methanol, Tenax-GC was   at  least  100  times  more  sorptively  effective than
Polymer A.  Polymer  B was  deposited  on Gas Chrom  Q  (15  percent)  and evalu-
ated  as  before and  proved  even  less sorptive  than  Polymer A,  relative  to
Tenax-GC.   It was  apparent from these results  that  the  coating of  prepared
copolyamides  on  Gas Chrom  Q could not  meaningfully   be  compared  with
Tenax-GC, which  was  used in the pure form.   Thus  future comparison was made
between Tenax-GC and the pure polymers prepared.

     Wolfe  had reported (8) that heating Polymer  B led to cross-linking via
the  propargyl  groups,  resulting, in  a more  rigid copolyamide.   Reasoning
that  such  heating  might  produce  a  more  suitable   GC   packing  material,
researchers in this  study  converted  Polymer B  to  the  crosslinked Polymer C
by heating  it at 270°C  for 6  hours under nitrogen.   The  resulting  slightly
tan solid was ground, sieved  (30-65  mesh),  packed in a GC column, and  eval-
uated as before.   The only noteworthy difference  in the  BTV values was the
twenty-fold  decrease in  the sorptivity of  methanol,   all  other  values
remaining essentially unchanged and  inferior  to those of Tenax-GC.

     The previous  polymers A, B,  and indirectly  C,  were considered  to  be
the base  from  which changes  could be made and related  to.   In  accordance
with the aforementioned objective  of introducing  electron-donating  or  with-
drawing functionalities into  the polymer system,  the next  step  would have
been the replacement of the methylene group  in  B and/or  C with  a  carbonyl
group  to  modify  the electron  density  in  the  polymer  system.    Although
subsequent  literature searches  revealed   that  in  the  curing process of  B ->
C, the -CH- group  was  simultaneously converted  to  the  8   group  (11,12),
the  predicted  need  required  the independent synthesis  of 4,4'-diamino-
benzophenone, followed  by  propargylation (Compound VII)  as described in the
Experimental.
                                    85

-------
     The  first synthesis  of  a new  copolyamide,  Cop-I, designed  to modify
the  above results  by changing  the  degree  of aromaticity  of  the  polymer
system  while  retaining  the  thermally  stabilizing  carbonyl  group,  was
accomplished  by  the reaction of  non-methylated ortho-nitrophenylenediamine
with  N,Nl-bispropargyl-4,4'-diaminobenzophenone  (75:25)  and reacting  the
mixture with  isophthaloyl chloride to produce  Cop-I.   This  copolyamide was
evaluated  as  before and the BTV values  determined  as  previously described.
The significant  improvement  in  the performance of Cop-I relative to Poly A,
B, and C  prompted the extension of  the  list of  toxic  organic  compounds to
include acrolein  and acetonitrile.  The BTV values  (relative to Tenax) for
raethanol  (104.9),  acetone (1.53), acrolein  (4.80),  and  acetonitrile (3.66)
suggested  that the addition of the polar nitro-group  considerably enhanced
the  ability  of  the copolyamide  to  absorb  the  polar  compounds  while  not
particularly affecting  the  sorptivity  for  the  non-polar compounds.
           H
                     H  0
                  NO,
     To  further test  this  hypothesis,  Copolyamide II,  (Cop-II)  was  pre-
pared, identical to Cop-I except  that  no ortho-nitro group was present.  As
anticipated,  the  BTV  values for  the  polar compounds decreased, while  non-
polar compound BTV's  remained  essentially unchanged.

     To  determine  the effect  of  strongly  electron-donating  groups  on the
sorptivity of the  test compounds, a new  copolyamide, Cop-Ill,  in  which the
diaminobenzophenone  moiety  in Cop-II  was  replaced  with  4,4'-diaminodi-
phenylamine,  was prepared  using the same proportions.   The BTV values for
the  non—polar molecules were  predictably  unchanged,  while the  sorptivity
for  acrolein and acetonitrile  was  thirty  times  larger than  that  observed
with Tenax-GC.
     H
                                   HCEC-CH,.
                                  Cop-Ill
                                                      HC=C-CH,
                                                                           n
                                    86

-------
      The results are summarized in Table 1.

      The dramatic improvement in the BTV values  for Cop-Ill prompted us  to
 repeat  the synthesis four more times, varying the reaction temperature, gel
 time,  and purification  solvent;  Table  2  outlines  the  conditions employed
 for  each preparation.

      The BTV values  of each preparation were determined and, as can be seen
 in Table 3,  large variations  occur  in  the BTV's  of  the different prepara-
 tions,  that  of  C being  the  most  efficient  for  acrolein, and  second most
 efficient for acetonitrile.  The  reason for the inversion of efficiency for
 these two compounds  between A and  D versus B and C is not as yet known, but
 may  become  more  apparent  after  extensive  characterization  studies  are
 complete.  However,  the  data only  emphasized the need for careful synthetic
 control,  a problem not yet solved  even for Tenax.

      To  determine if the adsorption  properties  of  Copolyamide  III could  be
 replicated by  preparing  the polymer  at  one   temperature,  the  synthetic
 procedure was  modified  such  that   the  propargylated  diaminobenzophenone
 (VII)  and  the  4,4'-diaminodiphenylamine   were   (in  N-methylpyrrolidinone)
 heated  to 140°C.  To this  was  added  isophthaloyl chloride with stirring
 until  gel  formation  was  observed,  usually  occurring  within  7  hours.
 Reaction Scheme  II summarizes the  procedure.
                             Reaction Scheme II
                                                        (75%)
                 HCEC-H2C
CH2-C=CH

NH
                                                        (25%)
                             NMP
                                                  -Cl
                               Copolyamide-III
The  final  product was  extracted  with methanol,  vacuum dried,  and  sieved.
The preparation and properties are  summarized  in  Table  4.   After packing in
a column, the BTV values  were determined and the results  are  summarized in
                                    87

-------
I
       0)
       pq
           (Tv
CM    .-I

O    O
                          CM
                                       OO
                                             00
                                             m
                                                    CO
             \o
             CM
             O
                          OO
                                             CO
                                                    CO
                   CM
                          CO

                          00
                                OO
                                       m
                                       m
                                             CO
             o    —i
             co    o
                          vO
                          VD
                                CO
             CM
                   CO
                     »

                   CM
                                CO
                                CM
                                  •

                                0\
                                             CM
                                       in
                                       oo
                                       CM
                    PQ
                                       PX
                                              vu

                                             •a
                                                          pq
                                                     88

-------













H
M
M
1
M
C
S
r-
P-
O
U
pc-l
o

£2
o
1— 1
H
•
rC

JT{ r^J
o cu
0) &
S o
I t*H
Q_j rH
^ o
P ^




CU
4-1
Cd 4-1
cd
TJ
CU CO
J-l f-(
J-l ,1^
d
O LO
o •
O ^H




CM
rH
0
CM
U

P-I
5rJ
&




X-N X-X
CM CM
*"•— •* X^r*

u u
0 0
o in
1 CM
^ CD
1







•^J



x-s
ci
J_)
> 4J
w
(U

4-)
 — i o r^» « — i



cu
0
Q)
^
0
4-1
^^
p t p.
s s3
s §


m

 u u
O o O O o o o
o in o o o m o
O CM O O CM CM CO
II-* —l| rH
O O i — 1 | O
in o o o
1 rH







M O



w
o
ffi cu
o s
CU 1
a c\i
T 0 rH
PH cu o

a 4J tt
1 u
0 1
<4 p.
p i



S_i
cu
4_J
Cd 4-1
cd

CU CO
Vl J-l

3 o
o m o
o • ^~
O CO i— 1




CM
1 — 1
ir

p
s
g



/^N
m
y~> CM CO
CM ^rf' ^-x
\^x
C_) o
C_> 0 0
o m o
O CM -*
1 1 -i
m o
1







Q



89

-------

-------
          TABLE 4.  GOPOLYAMIDE  III  -  PREPARATION AND PROPERTIES
Solvent
               Temperature
                    Reaction time
                  Yield
NMP
CH2C12
80
20
125-130
7 hrs
70%
Inherent Viscosity - dl/c = 0.95
(0.5% in NMP)

Tg (from DSC) - 277°C	
Table 5.  Although the replicability  of  the  BTV's  is  acceptable,  it is felt
that the performance  of  the polymer can be  improved  by further  study.   Of
concern is the large BTV for  methanol and,  by analogy, water.  It  has been
shown  that  the  polymer  should  have  a  low  affinity for  the water  vapor
present  in ambient  air  samples  (13);  otherwise,  displacement   chromato-
graphy, in which water vapor  displaces  the  already  adsorbed  organic  mole-
cules, may become an important factor  in the adsorption process.   Addition-
ally, adsorbed  water vapors  can cause  condensation  problems (ice)  during
the  cryogenic  concentration  phase  of  the  desorption-analytical  process
(1).
          TABLE 5.  COPOLYAMIDE III -  BREAKTHROUGH VOLUMES  (GM/L)
Compounds
              Acetonitrile
                     Acetone
                Benzene
Cop II1-6

Cop III-7

Cop III-8

         I

Compounds
                  107.2

                  110.9

                  111.8
                       32.84

                       31.96

                       59.16
                  1.55
                  1.18
 Relative breakthrough volumes of copolyamide III/Tenax GC
               Acetonitrile
                     Acetone
                Benzene
Cop III-6

Cop III-7

Cop III-8
                  228.6

                  236.5

                  238.6
                       55.07

                       53.60

                       99.21
                  0.202
                  0.162
     The next  step  then will be to modify  Copolyamide III to  diminish  the
affinity for  methanol,  with minimum  adsorptive modification  for the  com-
pounds of interest.
                                     91

-------
                              ACKNOWLEDGMENTS

     This  research was supported  by U.S.  Environmental Protection  Agency
Grant No. CR-807922-01.
                                REFERENCES

1.   Krost, K. J. j  E.D.  Pellizzari, S.G.  Walburn,  and S.A. Hubbard.   1980.
     Submitted  to  Environmental  Science & Technology for  review.   See  also
     EiD.  Pellizzari,   Publication  No.  EPA-605/2-74-121,  Contract   No.
     68-02-1228, July 1974.

2.   Fishbein, L.  1972.  Chromatography of  environmental  hazards:   carcin-
     ogens, mutagens, and teratogens.   Elsevier, NY.

3.   Hollis, O.L.  1966.  Anal. Chem. 38;309.

4.   Johnson, J.F., and E.M. Barrall.   1967.  J. Chromatogr.  31;547.

5.   Pellizzari,  E.D.    1975.    Development  of  analytical  techniques  for
     measuring  ambient  atmospheric  carcinogenic  vapors.   Publication  No.
     EPA-60012-75-076, November.

6.   Johnson, C.D.  1973.  Chapter  2 in The  Hammett  equation.   Cambridge at
     the University Press, London.

7.   Sakodynskii, K., L. Panina,  and N. Klinskaya.   1974.  Chromatographia
     7:339.

8.   Greenwood,  T.D.,   D.M.  Armistead,  J.F.  Wolfe,  A.K.  St.Glair,   T.L.
     St.Glair, and J.D. Barrick.   1982.   Polymer 23:621.

9.   Bell, V.L.  1976.  J. Polymer  Sci.,  Polymer Chem.  14:2275.

10»  Brown, R.H., and C.J. Purnell.  1979.   J. Chromatogr.  178:79.

11.  St.Glair, A.K., T.L. St.Glair, and  J.D.  Barrick.   1980.  NASA  techical
     memorandum 81918, January.

12.  Androva, N.A.,  M.I. Bessonov,  L.A. Laius,  and A.P.  Rudakov.    1970.
     Polyamides,  a new class  of  thermally  stable  polymers.    Techomic
     Publications, CN.

13.  Pellizzari, E.D.,  J.E.  Bunch, B.H.  Carpenter,  and E. Sawicki.   1975.
     Environ. Sci. Tech. 9:559.
                                    92

-------
            SYNTHESIS AND EVALUATION OF A  POROUS  POLYETHERIMIDE
                  FOR THE COLLECTION OF VOLATILE  ORGANICS
               Sajal Das, Louis A. Jones,  and John E.  Bunch

                          Department of Chemistry
                      North Carolina State University
                                Raleigh, NC

                              James D. Mulik

                   U.S. Environmental Protection Agency-
                   Advanced Analytical Techniques Branch
                  Environmental Monitoring Division, EMS
                        Research Triangle  Park, NC
                               INTRODUCTION

     In  ambient  air,  there  are many  carcinogenic  and  mutagenic  volatile
polar organic  compounds  present at extremely  low concentration  (parts  per
trillion)  (1-3).   Most  analytical  techniques presently  available are  not
sensitive enough  to  detect  these trace quantities.  Chromatographic  analy-
sis*' of  these  contaminants requires first  a  concentration step  followed  by
thermal  desorption  (4,5).   The  commercially  available porous .polymers  are
not sufficiently polar to concentrate  (high-breakthrough—volume) low-molec—
ular-weight polar substances (6,7).  Presently,  Tenax  GC  (2,6,diphenylpoly-
phenyl  ether)  is used  extensively as  a sorbent  material for  trace-level
organic pollutant analysis, but  since  it is a  nonpolar polymer (8),  it  does
not effectively trap lower-molecular-weight polar  compounds.   The polymeric
materials  required  for  concentration and  thermal  desorption  should  meet
several  specifications.   The most  important  features  for a  polymer  adsor-
bent to have are  (a) an  aromatic system (for  thermal  stability),  (b) polar
groups (for specific interaction),  (c) high molecular weight  distribution,
(d) hydrophobic properties, (e)  non-catalytic  or chemical activity relative
to the toxicants, (f)  highly  amorphous structure, (g) non-homogeneous  sur-
face, (h) large surface  area, (i) minimum micro-void to obtain reproducible
results, (j) uniform pore volume distribution, (k) high durability to avoid
compression  and   fragmentation   under  the  stress  of  high flow rates   and
handling,  and (1)  facile  incorporation  of   functional  groups  to  modify
specificity.

     The aim  of  the present research  is to  prepare a polyetherimide  con-
taining  several  polar  groups (carbonyl,  imide,  and ether).   The  starting

                                    93

-------
material for this polyetherimide  is  an acetylene-containing iraide oligomer,
Therraid-600, commercially available from Gulf  Oil Chemicals  Company,  USA.
The  modification of  this  oligomer  is  achieved  through  the addition  of
bisphenol compounds at  the  endcap acetylene  groups of  Thermid-600.
                               EXPERIMENTAL

Materials

     Thermid-600 was  obtained  from Gulf  Oil Chemicals  Company,  and was  used
without further purification.  Bisphenol A was  obtained from Aldrich Chemi-
cal  Company,  USA,   and  was  recrystallized  from  toluene.    The  solvent,
N-methyl-pyrollidinone  (NMP),  was distilled  over  calcium hydride  prior  to
use and stored over  4 A molecular sieves.

Polymer Synthesis

     The polymer may be  synthesized by  either melt  or solution polymeriza-
tion  technique.    A  typical  solution polymerization   can  be  described  as
follows:  20 gram (0.087 mole) of bisphenol A was  dissolved in 50 ml of NMP
in a mini-resin  kettle;  20 gram  (0.018  mole) Thermid-600 was  dissolved  in
50 ml  of  NMP and a few drops  of triethylamine was  added to  the  mixture.
This  mixture  was  added  gradually  over  a period  of  20  minutes  to  the
bisphenol solution  at 150°-155°C.   The reaction  was  kept  at  160°C for  2
hours.  The pot temperature was  then raised to  230°C,  at which time a high-
ly viscous liquid formed and NMP  began to  distill  off.  The pot temperature
was dependent  on the  amount of NMP distilled off.   Upon reaching 230°C, the
reaction was  then  run  for 2-1/2  hours.   Toward  the  end  of  the  reaction
(last half-hour), a slight vacuum was applied to  accelerate  the removal  of
NMP.  Total distillate collected  at  the  end of  the reaction was 75 ml.   Gel
formation was  noticed in the fourth  hour of reaction,  when the reaction was
quenched  by  pouring into methanol  in a  blender and  the  polymer  thus
obtained was purified by using NMP as  a  solvent  and methanol as nonsolvent.
The polymer was further purified  by  Soxhlet extraction with methanol for  24
hours.  The polymer  was dried  in  a vacuum  dessicator at 120°C.

     The IR spectrum was  obtained  with a  Perkin-Elmer model  521 in KBr
pellet.  Viscosity  measurements  were determined using  a 0.5  percent (g/ml)
solution of  the  polymer  in NMP at 20°C  using a Ubbelohde viscometer.   Gas
chromatographic  measurements  were performed with a  Perkin-Elmer  900 gas
chromatograph.    Breakthrough  volumes  were  obtained  for  the  polymeric
material  at several  temperatures,  whereupon a  plot   of  In  1/V versus  T
produced extrapolated results  to  ambient temperatures  (7).
                          RESULTS  AND  DISCUSSION

     The compounds  of  interest for concentration are tabulated  in  Table  1,
along with their  structure,  molecular  weight,  boiling  point,  and  dipole
moment.  Molecular weight data  show  that all of  the compounds listed  are
low  molecular  weight,  while boiling  point  and  dipole-moment  data  are

                                     94

-------
        TABLE 1.   PERTINENT INFORMATION FOR THE SELECTED COMPOUNDS

Name
Acrolein
Acryonitrile
Acetonitrile
Acetone
Benzene
1 ,4,Dioxane
Ethylene oxide
Chloroform
Methanol
Structure
CH2=CHCHO
CH2=CHCN
CHgCEN
^C=0
C6H6
pC2HttOC2H^
. CH-2H20
CHClg
CH3OH
M.W.
(g/mol)
56.05
53.06
41.0
58.08
78.11
88.12
44.05
119.38
32.04
B.P.
52.5
78
81.6
56.2
80.1
100.1
10.7
61.3
64.6
Dipole moment
(liD)
2.90
3.51
3.94
2.73
0
0
1.89
1.06
1.67

indicative of the volatility and polarity of the compounds, respectively.




     Table 2 shows the U.S. production and pollutant information (9).






              TABLE 2.  U.S. PRODUCTION/POLLUTANT INFORMATION

Compound
Acrolein
Acryonitrile
Benzene
1 ,4,Dioxane
Ethylene oxide
Chloroform
U.S. production
(106 Ib/yr)
60
1410
1400
14
3960
260
Pollutant
Carcinogen
Possible
Probable
Probable
Probable
Possible
Probable
information
Emission*
Kg
1.1 x 104
3.7 x 105
5.9 x 107
1.3 x 106
3.8 x 104
	

*Emissions, Kg = source emissions, Kg/yr (365 da) x half  life,  I da.
                                     95

-------
Tenax GC is a polyphenylene  oxide,  prepared by oxidative coupling reaction,
with  cuprous chloride  as  catalyst  and  pyridine  as  solvent,  as  follows
(10):
                                ;  CuCl
                              pyridine
          X = H;  Cl,  Br,  1
Tenax GC is  a product -of AKZO  Research  Laboratory,  Netherlands, and  it  is
distributed  through Applied  Science,  Inc., USA (8).  It  is  a thermostable,
hydrophobic  polymer (11), has been  shown to  be  an excellent  chromatographic
packing  material  (12),  and  has been  extensively  used  in the  field  of
environmental  pollution  concentration/analysis  (13-15).   One  of the  most
important  limitations  of  Tenax GC  of   present  concern  is  its   defined
nonpolar character and,  presumably because  of  this, its  inability  to  trap
low-molecular-weight  volatile organic polar compounds (listed in Table  1).
The nonpolar  character of Tenax GC  is calculated  from the retention  indices
for  benzene  and  ethanol  (8)  and,  according  to  Walraven,  the polarity
increases  with  decreasing  ratio  of  the retention  index  of  benzene  and
ethanol.   From Table 3,  it can  be concluded  that  Tenax GC is the  least
polar polymer  on the  list.
         TABLE 3.  POROUS POLYMERS ARRANGED  IN ORDER OF  INCREASING
                   POLARITY, DETERMINED ACCORDING TO WALRAVEN

Porous polymer
Tenax GC
Chromos orb
Porapak QS
Porapak Q
Chromos orb
Porapak P
Chromos orb
Porapak S
Chromos orb
Porapak R
Porapak N
Porapak T
Chromosorb

102


101

105

103



104
Ifienzene/lEthanol
1.64
1.56
1.56
1.56
1.52
1.47
1.47
1.41
1.40
1.36
1.35
1 . 28
1.21
                                     96

-------
       Recently, a  series  of copolyamides was  synthesized in the  laboratory
  that were  shown  to be very promising  for  the present  research (16).   The
  structure of the copolyamide for which  the best  breakthrough volume results
  were obtained is shown below:
                                         CH^CSCH
                                      CO-N

  The polar groups  in  this  polymer are the  amide group, the carbonyl  group,
  the secondary amine  group,  and the acetylenic  bond.   Although  the  copoly-
  amide  shown  above  gives  good  results,  some  of  the  problems  encountered
  include a high  affinity for methanol  (and presumably  water)  and the  fact
  that  the  N,N'-bispropargyl-4,4'-diaminobenzophenone   requires  a  multistep
  synthesis (18).   However,  a commercially  available; oligomer,  Therraid-600
  (Gulf Oil Chemicals  Company, USA)  has  similar  functional  groups, as  does
  the above copolyamide, as well as the polyether  linkage of Tenax.
 Hca
                                                                    N-
                                                                           C=CH
where n = 1-2.                                         -


  The high softening  point  (185-200°) and minimal  water adsorption «1 per-
  cent after  1000  hours at  125°F,  100 percent  R.H.)  (19)  suggested  that  a
  polymer could be prepared  having  polar  character with minimal affinity  for
  water.   Thermid-600 by itself  could not  be evaluated as a packing material
  since the powder form is too fine for a packing material by  itself.

       Chemical modification of Thermid-600 was  the object at  this point.

       In 1965, Monsanto Chemical Company, USA,  disclosed a  patent describing
  the addition of the hydroxyls  in  glycols,  alcohols,  and phenols across  phe
  triple  bonds of  diacetylenic diesters (20).   Recently,  a polyether ketone
  was synthesized by the following reaction  (21):
                                CH CHCH
                                      97

-------
  NMP
-HC=HC-H2C-HN
                 NH~CH2HC=CH~°  \O
Encouraged  by  the results  of  these  reactions,  researchers  reacted  the
Therraid-600 and  bisphenol A similarly, and a polymer was  obtained  with the
following suggested structure:
                 HCEC /^[Thermid]—. C=CH   +   OH
                  - -HC=HC /-w-lThermid}-— CH=CH-0
The structure of Thermid is:
                                                                 ^-(6)
The  addition of  aliphatic  OH across  the  triple  bond  is  well  documented
(22), but addition of phenolic OH across  the  triple -bond is  not  well  estab-
lished.  To  determine  if the latter reaction was feasible,  phenylacetylene
was reacted with  phenol with K2CO_ as  a base  catalyst.
           C=CH
:o
                                   NMP     '

                               reflux, 40 hours
                                      .a
                                                      H   H
                                                      I   I
                                                      C = Cv
                                      a
                                                 sis 3-phenoxystyrene
                                     98

-------
The product obtained was cis g-phenoxystyrene,  characterized by IR,JSIMR and
mass spectra.   The  IR spectrum showed bands  at 1580 cm~l ,  1240  cm"!,  1040
cm'1,  and  770 cm-l,  corresponding  to  -C=C-,  0-0,  C-0,  and  cis-HC=CH-
groups,  respectively.   Figure 1  shows  the proton NMR.   The  observed  cou-
pling constant  for  the cis protons  is  7Hz,  consistent with cis  configura-
tion.    The  mass  spectrum indicated a  molecular  ion of  M/Z  =  196,  and
fragmentation was consistent with the structure assigned.
                                                Ha   Hb
                                                 \   /
                                                  C=C
                                J =7Hz   c/s-j3-Phenoxy Styrene
              10   98
                                                         10
 Figure 1.   Proton NMR of cis-g-phenoxystyrene .


      N-methyl-2-pyrrolidone  was  selected  as   the  polymerization  solvent
 because of  its  superior  physical  and  chemical  properties  compared with
 other aprotic  liquids.   The fact that  it is highly  polar  is reflected  in
 its high  boiling point  of 202°C  and  high dipole moment  value,  4.09  (yD).
 Therefore, one would  predict  that it would  be  a suitable aprotic,  dipolar
 solvent, which is  one important criterion  for  its  use under  the  demanding
 reaction  conditions.    Moreover,  because  of  the  polar  character  of  the
 monomer, polymer,  and solvent,  there  is a strong tendency for association,
 which  facilitates the  polymerization  reactions  (23).   The  N-substituted
 carboxamide ring has  good hydrolytic  stability,  except in the  presence  of
 strong  aqueous base  or  acid at  elevated temperatures.   Also, it is  rela-
 tively  nontoxic  and non-corrosive.

      Table 4 shows the  general  characteristics  of  the polymer.
                                      99

-------
                    TABLE 4.   PROPERTIES OF POLYIMIDE ETHER
                 Color

                 Inherent viscosity*

                 Yieldt.%
Pale yellow

0.91

65 (SOL)
 *0.5% solution  in NMP  at  21°C.
 fYield is  calculated on the  basis  of Thermid-600.
 From inherent  viscosity  data,  it can be inferred  that  the  molecular weight
 of the polymer is high.
 GC.
      Table 5 shows the breakthrough  volumes  of  the polyetherimide and Tenax
              TABLE 5.  BREAKTHROUGH VOLUMES  OF  POLYETHERIMIDE
             	AT 25°C (LITERS/GRAM)
 Polymer	Methanol  Acetone  Hexane  Benzene  Acrolein Acetonitrile

 Polyetherimide    1.758    288.91  0.2774    8.61      92.20        124.1

 Tenax GC	0.041	0.85  4.6       3.9       0.48          1.20
 Breakthrough volumes of all polar compounds like acetonitrile, acetone,  and
 acrolein are much superior compared with Tenax GC.  The higher breakthrough
 volumes  for  polar compounds  suggests  that there  is  a  strong interaction
 between  the polar organic compounds and the polymer,  n-Electron-containing
 compounds,  like benzene, are  also  better retained on  the  new polymer than
 on  Tenax GC,  possibly  because  the polyetherimide  contains  many aromatic
 rings,  and  the  presence of  double bonds  contributes  to  the  TT-electron
 density  of  the  polymer.   The methanol  retentivity is greater than Tenax GC,
 but much  smaller than for copolyamides  (19).   From methanol retentivity, it
 can  be anticipated  that water  adsorption for  the  polyetherimide  is  much
 less and  that it may be  considered  as  a hydrophobic  polymer.


                                 CONCLUSION

     The  present research has  developed a polar  porous polymer,  which  is
useful for  trapping  both polar and  nonpolar  compounds.   The water retentiv-
ity of this polymer is  low.   The starting materials  for this polymer are
available commercially,  so  extensive monomer synthesis and  purification  is
not necessary.   The  polymerization  technique is  also  simple, and  chances  of
side reactions are minimal when  the  reaction is  run  in  equimolar  ratio.
                                     100

-------
                              ACKNOWLEDGMENTS

     This  work was  financed  under U.S.  Environmental  Protection  Agency
Cooperative Agreement Grant No. CR-807922-01.   The  authors  thank Mr.  Rodney
Beaver of this Department for helpful discussions during  the  course of  this
study.
                                REFERENCES

1.   Sydor, R., and D.J. Pietrzyk.   1978.  Anal.  Chem.  50:1842.

2.   Bertsch, W., R.C.  Chang,  and A. Zlatkis.   1974.   J. Chromatogr.  Sci.
     12:175.

3.   Pellizzari, E.D.,  J.E.  Bunch, B.A.  Carpenter,  and E. Sawicki.   1975.
     Environ. Sci. Tech. 9:552.

4.   Pellizzari,  E.D.,   J.E.  Bunch,  R.E. Berkley,  and  J.  McRae.    1976.
     Anal. Lett. 9(1):45.

5.   Versino,  B. ,  M.   deGroot,   and F.  Geiss.    1974.   Chromatographia
     7(6):303.

6.   Krost, K.J., E.D.  Pellizzari,  S.G.  Walburn,  and  S.A. Hubbard.   1980.
     Submitted  to Environmental  Science & Technology for  review.   See also
     E.D.  Pellizzari,   Publication  No.  EPA-605/2-74-121,   Contract  No.
     68-02-1228, July 1974.
7.   Brown, R.A., and C.J. Parnell.   1979.  J. Chromatogr.  178:79.

8.   Daemen,  J.M.H.,  W.  Dankelman,  and  M^E.  Hendriks.     1975.
     Chromatogr. Sci. 13:79.
                                                                          J.
9.   West, D.S., F.N. Hodgson, J.J.  Brooks,  D.G.  DeAngelis,  A.G.  Desai,  and
     C.R. McMillin.   1981.  In Potential  atmospheric  carcinogens  phase 2/3.
     Analytical   technique  and   field   evaluation.     Publication   No.
     EPA-600/2-81-106, June.

10.  Cooper, G.D., and A. Katchman.   1969.   Advan.  Chem.  Series  91:660.

11.  Pellizzari,  E.D.   1975.    Development  of  analytical  techniques  for
     measuring  ambient  atmospheric  carcinogenic vapors.   Publication  No.
     EPA-60012-75-076, November.

12.  VanWijk, R.   1969.  J. Chromatogr. Sci.  7:389.

13.  Bunch,  J.E.,  and  E.D.  Pellizzari.    1979.     J.   Chromatogr.   Sci.
     186:811.

14.  Versino,  B.,  M.  deGroot,   and  F.  Geiss.    1976.     Chromatographia
     7(6):302.
                                    101

-------
15.  Bertsch,  W.,  R.C.  Chang, and A.  Zlatkis.   1974.  J.  Chromatogr.  Sci.
     12:175.

16.  Jones,  L.A.   1982.   Porous  polymer-adsorption  of  volatile  organic
     toxicants  in ambient  air.   The  synthesis  and evaluation  of  aromatic
     copolyamides   containing   N-propargyl  groups.      EPA  Report   No.
     CR-807922-01-1, January.

17.  Deanin,  R.D.   1972.   Polymer  structure, properties  and  applications.
     Cahners  Books, MA.

18.  Wolfe, J.D.  NASA Grant  #NSG-1524.   Patent  applied for 1980.

19.  Gulf  Oil Company,  Technical  Bulletin,  Thermid-600  addition  curable
     polyimide resin.

20.  Butler, J.M., L.A.  Miller,  and G.L. Wesp.   1965.  U.S.  Patent  3201370
     (to Monsanto Co.).

21.  Das,  S., L.A.  Jones, J.E.  Bunch,  and  J.D.  Mulik.    1982.    Polymer
     Bulletin 6:509.

22.  Miller,  S.I.  1956.   J. Am.  Chem.  Soc. 78:6091.

23.  Sroog, C.E.  1976.  Macromolecular  Review 11:161.
                                   102

-------
        ADVANCED  CONCENTRATOR/GC METHODS FOR TRACE ORGANIC ANALYSIS
                 S.A.  Liebman,  T.P.  Wampler«,  and E.J.  Levy

                        Chemical  Data Systems,  Inc.
                                Oxford,  PA
                                INTRODUCTION

     Gaseous  organic  pollutant analysis  requires  reliable  and  versatile
instrumentation.   Sampling procedures have been developed  with  such equip-
ment designed  and engineered to  provide  analysts  with  the needed informa-
tion.   The CDS 320 Concentrator  is a microprocessor-based analytical sam-
pling system (1,2) for air, water,  and  solids  (Figure 1) that permits abso-
lute repeatability throughout  the  operational  sequence to ensure accurate
results under  appropriate conditions.  The role of  modern instrumentation
is  to  provide  cost-effective analyses  with minimal operator training.   In
trace   organic  studies,   traditional  enrichment  methods   have   employed
cryogenics,  solvent  extraction,  derivatization,  or  other  time-consuming
steps  that  are  difficult to  automate.    The  rationale  for  emphasizing
thermal desorption is  outlined  in Figure  2  and,  as  in any analytical situa-
tion, the  disadvantages  must be  evaluated  relative to  the advantages—for
example,  thermally labile compounds  are  not successfully  handled  in this
approach.  Also,  standardization  procedures,  although basic  to  any method-
ology,   are  particularly  critical for trace organic studies  at  the ppm-ppb
levels handled  by  gas  chromatographic (GC)  and  concentrator technologies.

     This paper emphasizes method development,  with the  concentrator inter-
faced  to  a  high resolution  capillary  GC   unit   using  controlled  and
wide-ranging  experimental parameters,   including  sampling  flow  rates,
thermal desorption/time/temperatures, and varied sorbent trap packings  and
matrix  effects.    Additionally, coupling  of the  concentrator with on-line
oxidation/reduction reactors  further  provided total  organic carbon analysis
for  organics  from  air-sampling  cartridges.     The  sorbed   organics  are
desorbed in the thermal desorber module of  the  320  Concentrator  and carried
in a helium  stream to the microreactors  and  on-line TCD/FID GC system  for
analysis of hydrocarbons at the 10  ppb  level.
                          EXPERIMENTAL  CONDITIONS

     For  sampling,  organics were  sorbed onto  a  3" x  1/4" I.D.  stainless
steel  cartridge  containing about  100  mg of  adsorbent held  in place  with

                                    103

-------
UL
UJ
UJ
UJ
UJ
                UJ
                     UJ
                o   o
                                                 o
UJ
o
CO
UJ
          CO
                UJ
co    D
  ~   o
cc    o
UJ    -5
      CO
CC    CO
UJ    UJ
CO    O
cc    o
o    <
CO    ^
UJ    
-------
                          CO
                          UJ
                          0
              H
              CO

          CO  Z  ^
          UJ  Q  *"
          «•      w
          5  H  CO
                             -J
                             <
CO
u
<

£
UJ
      CO

w    r
z    H

      o
     o
     CO

     o
     z
     <  o
o

H

CC
UJ
Q.
O

UJ
-I
m

o

Q
O
CC
Q.
Ul
CC
CO
UJ
o
Z
<
>
Q
CO m
< -I
O a,

M
     CO  -
             CL


             O

             Q
       CO
       UJ
                  O
                  Ul
                  -J
                  UJ
                  CO
m  CO
^  UJ
>  5

PS
s  H
                      CO  O
                      UJ  CO

                                 6
                                 o
                                 •
          O
       £0   |
       CU  O



       si
       IL  O
       CO  O
       o  i-
              OL  o
< o  z
       UJ
       m
       cc
O  o  S  y
co     2  <
ui  o      u-
Q  Z  Q  K
a  =:  iuj  uj
-I  O  H  H
<  z  <  z
                                               z
                                               o
                                    N

                                    a
                                u.
                                u.
                                      CO  Z
                                      o  cc
                                      Z  Ul
                                      3  I-
                                      o  z
                                      CL  ""
                                      2  ui

                                      0  S
                                      O  <

                                      ui  tt

                                      3  S
           O  UJ
           UJ
O     CO

£ g  -
Si
(C  3  O  O
UJ  O  H  UJ
Z  UJ  3  H
I-  CO  <  <


 •   •  "  o
01  «  -1  P
                           CO
                           Ul
                           o
                           <

                           CO
                                       CC  -3
                                       ui  O
                                                  c
                                                  o
                                                  •H
                                                  4-1
                                                  CX
                                                  J-J
                                                  O
                                                  CO

                                                  •s
                                                  60


                                                 •H
                                                  CO
                                                             Ml
                                                             O
                                                  O
                                                  0)
                                                  4-J


                                                  S-i
                                                  o
                                                 Cl
                                                 <0
                                                 o
                                                 a
                                                 o
                                                 a

                                                 a
                                                 •H

                                                 CO
                                                 CU
                                                 O
                                                 d
                                                 n)


                                                 I
                                                 CU
                                                 t-l
                                                 3
                          105

-------
quartz  wool.    The  sorbents,  as  denoted  in the  respective  figures,  were
Tenax-GC  (obtained  from Supelco,  Beliefont, PA) CDS modifier,  or  Ambersorb
XE-340  (from Rohm and Haas, Philadelphia,  PA).   Test mixtures were  gener-
ated in glass  vessels  by means of syringe  transfer dilutions at ambient  or
elevated  temperatures,  as  required.
320 Concentrator:
Sample flow:
Trap flow:
GC carrier:
40 ml/min helium
30 ml/min
30 ml/min
Thermal desorber:  260°C, 5 min
Tenax traps:  250°C, 1 min or
                 as indicated •
Transfer line:  275°C
Valve oven:  280°C
GC-Varian 3700 FID:
Inj. and Det.:  285°C
Direct Capillary  (splitless) injection
30M x 0.32MM,! micron film; programmed  55°C,  hold  3 min,
5%nin to 175°C or as indicated
Att'n:    10-H range
Chart speed:  1 cm/min

Total Organic Carbon Analysis;

Sample:  10 ppb hydrocarbon (Cjo)  sampled  at  0.5 1/min  for
         2 min onto Tenax  cartridge

320 Concentrator:  Thermal desorber  225°C  for 5 min
                   Tenax trap 225°C/2 min

GC:  CDS FID/TCD  14' x  1/8":  mixed  porous polymer ss column
                              140°C  isothermal
Mlcroreactors:
  320 Concentrator effluent from traps led into
  the CDS 820WP reaction system containing CuO
  reactor (850°C) followed by Ni catalyst reactor
  (400°C, hydrogen 5 ml/min).  The reaction prod-
  ucts, CHij and E^O, were detected by the internal
  TCD and FID.
                          RESULTS AND DISCUSSION

Industrial Hygiene Applications
     The role of  air  monitoring in the industrial hygiene field has  inten-
sified over  the  past several years.   Mandated monitoring programs  require
accurate sampling and analysis  of  trace (ppm—ppb) organics in the work  and
plant environment.   The  onsite  sampling process using the 320 Concentrator
is demonstrated by the results  shown in Figure 3 with the .conditions  given
in Figure  4.   A test air/mixture  was loaded onto a  Tenax cartridge  by  an
                                    106

-------
   CDS 320 CONCENTRATOR
   SIGMA II QC FID
   SAMPLE TUBE
   3' X 1/4" OD SS
   TENAX-GC

   ON-SITE AIR
   2 MIN AT 11/MIN
MHHYL •

   THICK
                                   TOUIINK
                               N-MiTHYI.liOIIPHOI.IME
HYL KITONI

OMOITNYLINI
               MITHANOL
     ATT'N: 32 X  10-11
Figure 3. Remote sampling of solvents in air.

                        107

-------
CDS 320 CONCENTRATOR

    REMOTE SAPLING:  CN-SITE AIR, 2 MIN, 1 LITER/MIN,  3* x W CD SS TENAX
                     CARTRIDGE.
    320 SWUNG:
    INJECTION;
PURGE 1 MIN, 20 ML/MIN HELIUM; No HEAT,  HEAT THERMAL     ;
DESORBER, 200°C,, 10 MIN,,  20ML/MIN TO TRAP, CbOL DOWN^  6 MIN,

BftCKFLUSH TRAP 25 ML/MIN HELIUM. 200PC. 4 MIN TO GC,
     P-E SIGMA II FTP
    PRECOL, :
    COL,:
   DET:
   CHART:
12" x 1/8" 01) SS 3% CARBOWAX 1500 ON CHROMOSORB W,

8' x 1/8" OD SS 0,2% CARBOWAX 1500 ON CARBOPAK C, 25 ML/MIN
HELIUM,
     , H MIN, THEN 1CP/MIN TO 16CPC, 10 MIN,


32 x 10'11

0.5 CM/MIN.
 Figure 4.  Experimental conditions  for remote sampling of solvents  in  air.


                                       108

-------
 air  sampling pump,  and the cartridge was then inserted into  the  320  thermal
 desorber module.   Packed column  chromatography was  satisfactory for  this
 purpose, and  sampling  a test  air  mixture directly (with  GG  injection)
 compared well to  the results seen in Figure 3.   In  this  manner,  typical
 industrial  solvents are analyzed at low ppm-ppb concentrations.

 Priority Pollutants - Quantitative Results

     The need to examine sorbents  used in the  air  sampling  method  is  evi-
 dent,  since sorption/desorption efficiencies vary significantly  forv organic
 mixtures.   Figure  5  shows  the results  of  such  test  mixture components  at
 the  40 ppb  concentrations sampled and analyzed as indicated.  The  reprqduc-
 ibility of   the  analyses is  shown in Table  1 to 'have a  .'relative "standard
 deviation of better  than 5  percent.   Literature  results  (3)  using manual
 methods for such analyses  at the  ppb  level rarely  achieve  this  reproduc-
 ibility.  Clearly,  the  closely controlled  automated instrumentation of  the
 320  has provided  the much-needed  improvement  in  a  cost-effective/" manner.
 Furthermore,  Figure 6 extends the  range  of analyses to complex  hydrocarbon
 mixtures with high-molecular-weight components such as those in  diesel  fuel
 with the same methodology  as that  shown  in Figure  5.    Analysis  of trace
 organics in water  with  the  purge and trap, module  were  reported earlier,
 with similar results  (4).
                     TABLE 1.   TEST AIR MIXTURE ANALYSIS
                               TENAX CARTRIDGE
                               40 PPB

Sample
Benzene
Toluene
Chlorobenzene
Heptane
0-dichlorobenzene
Dodecane
Peak
94
123
120
68
167
148
height, mm
90
112
111
72
161
142
96
115
115
65
172
147
95
129
123
73
178
140
Avg.
4 runs
93.8
119.8
117.3
69.5
169.5
144.3
Std.
dev.
2.3
6.7
9.7
3.2
.6.3
3.3
Average
Std. dev.
rel. %
2.4
5.6
8.2
4.6
3.7,
2.3
4.7%

Varied Adsorbent

     Instead of Tenax adsorbent alone serving as  the  sorbent  bed,  compari-
son cartridges with  other  solvents were prepared  to increase  the  range for
monitoring polar and  low-molecular-weight  compounds.   Figure 7, compares the
                                    109

-------
                                         O  =
                                         =:  "•
                                         o=*
                                         o  »-

                                         «  E  °
                                         z  £  o
a
z
Zi
a
2

0)
            o>
            E

            o z
            o «
 5»

 S°
 OS  I-
 O  <
 (0   .

 X  °
 **  m

 z  w
 U  o
 i-  o
        cc
        H

        lil
                    O
                    CM

                    CO


                    (/)

                    O

                    O
           »°    i
            O
            CO    *-
            CM
                        ac
                        ui
                        ca
                        cc
                      10
                 §5".
                 > co O

                 
-------
                  a:
                  o
                  flC
                  H
                  Z
                  Ul
                  O
                  z
                  o
                  o

                  til
Ul
_l
0.
     IO
u   o


15
Ul   z

u   i
5   N-
o
CM
                  b
                  o
         z   -
 o
o
 o

 CM

 oc
 Ul
 03    .   Z
 OC  Z
 o  -   ^

 2  ?0°
 Q   E   O
 _,  O   w
                       O
                       O
                       a   ±
              
              o.
                   o
                   Ul
                   oc
                            CM

                            CO
                             •

                            O


                            X
                            o
                                              oa
                                              Q
                                     o
                                S  °o
                                o   o
OS-
03       Z
<   CO  ^

OC  o   ^

Q   m  to
                                                                              ui
                                                                0)


                                                               M-l
                                                                                  0)
                                                                                  CO
                                                                                  a)
                                                                                  •H
                                                                                  o
                                                                                  4-1
                                                                                  a)
                                                                                  60
                                                                                  S

                                                                                  X
                                                                                   w
                                                                                  •H
                                                                                   to
                                                                                   s
                                                                                   cfl
                                                                                  •U
                                                                                  CO
                                                                                  0)
                                                                                  EH
                                                                                   60
                                                                                   •H
                                      111

-------
                                                (JO
                                                •H
112

-------
results  from Tenax  cartridges  only;  Tenax  with CDS  modifier  (1:1);  and
Tenax + modifier  + Ambersorb XE-340  (1:1:1)  used for  the  analysis of  a 1
ppm methanol, methyl ethyl ketone,  and dioxane  (1:1:1)  test  air  mixture.
It is shown  that  the Tenax with these  added  sorbents makes  it  possible to
trap polar organics  for improved detection and  analysis.   The  performance
of a Tenax/Ambersorb  XE-340 sorbent  bed  is  illustrated  in Figure 8,  with
methyl  formate,   butyl  benzyl  phthalate,  and naphthalene   for.  a  range  of
polarities  and  molecular  weight   components.    Likewise,  low-molecular-
weight,  nonpolar  alkanes  are  poorly retained by Tenax sorbent  alone,  but
the excellent  retention shown  with  the  Tenax/Ambersorb  bed  is  shown  in
Figure 9 at  the low-ppm level under the nominal  sampling and analysis  con-
ditions indicated.

Total Organic Carbon  Determination

     Interfacing  the 320  Concentrator  in the above  studies was  achieved
with a  direct  capillary GC  configuration for high  resolution  analysis  of
the organics.  Additionally,  the effluents from  the  trap were directed via
a heated transfer  line  to  a CDS  microreactor  system  designed 'for conversion
of the  organics  to C02  and,  subsequently, to CHt,. for FID detection.    In
this manner, high sensitivity detection is shown by the  data  in Figure 10
for a 10-ppb hydrocarbon sample sampled as indicated.  - Also^ a  TCD was in
line so that the  CH^  and H20  reactor  conversion  products  could be chromato-
graphed and  detected  with  the internal  GC/TCD system.
                          SUMMARY  AND  CONCLUSIONS

     The versatility of modern  instrumentation  tq  sample  and analyze a wide
range of   organics—low—and—high-molecular-weight,   polar  and   non-polar
compounds—has  been demonstrated  in this  study  using the  CDS 320  Sample
Concentrator.  Variations in  trap  contents  and  instrumental  parameters make
possible  accurate, "reproducible  analyses  with minimal  operator  training
using the microprocessor-based  system.   Development will  continue  in order
to provide cost-effective methods  to meet the challenge of analyzing diffi-
cult trace organics.                                         >  •
                                REFERENCES

1.   Gargus, A.G.,  and  C.R.  Watterson.   1980.  A microprocessor  controlled
     purge & trap GC concentrator.  American  Laboratory,  February.

2.   Gargus, A.G.,  S.A.  Liebman,  and C.R.  Watterson,   1981.   Applications
     of advanced concentrator/GC  technology.  American Laboratory,  May.

3.   Application Laboratory.  Concentrator  bibliography.   CDS,  Oxford,  PA.

4.   Gargus, A.G., W. Bowe, W. Dodson,  T.P. Wampler,  S.A.  Liebman,  and  E.J.
     Levy.  1982.   Applications  of concentrator/direct capillary GC  system
     with a purge &  trap  autosampler.   Paper No. 349,  Pitts. Conf.,  Atlan-
     tic City, NJ.
                                    113

-------







SAMPLE






FORMATE
METHYL


H

_J

ENZYL PHTH
VLENE
BUTYL B
NAPHTH/
oc
O
1-
^
OC
H
IPLE CONCEN
2
O
CM
CO


O
O
(O
CM
. DESORBER
THERMAL






OmI/MIN. He
z
i
m






z
i
o
o
o
m
TRAPS 2






IPILLARY GC
DIRECT Ci
m
i
m
o
o
DURABON




Z
_i
H
T-
2
2
CM
CO
d
X
2
6
CO






Q
O
X
z
i
CO
o






o
m
CM
O
H
Z
i
o
T-
                             CM    CO
                                                                                                                                 .
                                                                                                                                • 60
                                                                                                                                 W
                                                                                                                                 •H
                                                                                                                                 V<
                                                                                                                                 4-J
                                                                                                                                 M
                                                                                                                                 rt
                                                                                                                                 o
                                                                                                                                 CO
                                                                                                                                  I
 o
 CO
 M
 tu

I
CO
                     CM
                                                                                                                         UJ
H

 CO
•rl
 CO

iH

 §
 cO

 a)

 3
 j-i
 X
                                                                                                                                  CO
                                                                                                                                  Q)
                                                                                                                                 H
                                                                                                                                 oo

                                                                                                                                  a)
                                                                                                                                  .M
                                                                                                                                  S
                                                                                                                                  t)0
                                                                                                                                 •H
                                                               114

-------
  Z
  <  (O
  X  Ul
     S -
     - -
     < O
Ul
-I
Q.
  ui
UJ
"1 o  S
uipi-  <

ui
CO t-  CM
             CM
DC
1-
Z
UJ
O
Z
o
o

o
CM
CO
0)
O
O
CC
UJ
03
OC
O
0)
Ul
Q
	 1
mJ
<
CC
Ul
X
1-
0)
Z
•
Z
•e
2

£
O

CM
£
^
cc
1-
                       O
                       ol
                      o
                      o
-I

QL

O

H
o
UJ
K
                        CM

                        CO
                         •

                        O


                        X
o
CO

10 <
m 2
o oc
« w
i K

5 1
CC o""
=  30
00^
                              O

                              X


                              Z
                              Ul
                              H
                                     1
           Ui
  CD
                                       Ul
                                       rt

                                       0)
                                         
-------
                                                           oc
                                                           UJ
           V)
UJ
O    n    .
ta         z
a:    ui   =
<    O   S
o    Q   ^
O    a   3j
OC    (-
ooco

x    2   5

2    x   ^
£    2   I
o    in
T-    I-   W
                      03
                      CL
                      Q.
                                                                                               Q
                                                                                               O
                                                                                                       CO
                                                                                                       •H
                                                                                                       CO
                                                                                                       c
                                                                                                       o
                                                                                                      -(3
                                                                                                       cfl
                                                                                                       00
                                                                                                       O
                                                                                                      H
                                                                                                       3
                                                                                                       (JO
                                                                                                      •rl
                                             116

-------
              AIR  ANALYSIS  BY A NONDISPERSIVE INFRARED METHOD
                                Philip Hanst
                                INTRODUCTION

     A nondispersive  analyzer responds to  infrared  radiation,  but does not
spatially  separate the various  radiation  frequencies.  When  compared with
instruments  that  use gratings  and  slits,  a nondispersive  instrument shows
advantages  of  high energy throughput, a high  degree of spectral multiplex-
ing,  and an  unlimited  degree  of  spectral resolution.    These  advantages
result from the use  of  optical  components  that are inherently responsive to
spectral  characteristics of compounds  being  measured, while  passing broad
bands of  frequencies.   The performance of  the nondispersive optical system
depends  on correlations  between  the infrared  spectrum of  the  gas  being
measured and the  spectral response  characteristics of the system elements.

     For any gas,  infrared absorption and  emission  are spectrally similar.
This allows  a  gas-filled microphonic detector to respond  more  strongly to
radiation  from a  heated  sample of  that  gas   than  to radiation  from other
gases.  Using  a  continuum source,  such a detector will also see absorption
by  the  chosen gas more  strongly  than absorption by  other  gases.   These
correlations are  the  operating  basis of widely used nondispersive analyzers
called  positive-filter systems  and  described,  for  example,  by  Luft  (1).
The performance  of these systems has been mainly limited  by interferences
from absorbing species other than  the one being measured,  especially from
water vapor.

     When  the  components of a  nondispersive  analyzer are  arranged in what
is  called  the  negative filter  configuration,  a high  degree  of  discrimina-
tion between the  object gas and other  gases  is obtained.   These negative-
filter  systems have  not  yet had  as  much  commercial  use  as the  positive
filter systems (based  on the  microphonic  detector),  but  they  have  proven
their worth in a number  of  development  projects  and  field tests  (2-7).
Ambient  air pollution  measurement  requires   the  high  selectivity  of  the
negative filter system.
                                    117

-------
      All  nondispersive  analyzers  have  as  basic  components  a  radiation
source, filters,   a sample cell, a radiation detector, and  electronics  that
process and  display the detector signal.   Nearly all systems have  at  least
one  rotating light  chopper.    The  systems currently  in  use  do not  differ
basically  from   those  studied  by  Pfund  and  others  in  the   1940s   (3).
Although the principles  of operation of the systems have not  changed  in the
last   35  years,   capabilities  of  system  components  have  been  greatly
expanded.     Filters,   detectors,   and  electronics   have   been especially
improved.  When  these  improvements are coupled with gas-filled  filter  cells
for  interference  removal,  it  is  found that the  sensitivity of  the nondis-
persive technique becomes  great enough for direct measurement of pollutants
in the ambient air.
                        SYSTEM OPERATIONAL PRINCIPLES

     A  nondispersive  instrument  in  the  negative filter  configuration  is
diagrammed  in  Figure 1.  Radiation sources  in instruments described  in  the
literature  have included  hot wires,  globars,  Nernst  glowers,  heated gas,
and  the  sun.   The source spectrum may  be altered  by spectral filters that
can  be positioned  nearly anywhere in the  optical train.  Typically,  source
filtering might be done with a  narrow  band  interference filter in  combina-
tion with  a   gas-filled  filter  cell.    The  emitted  radiation  is  passed
through  a sample  cell,  which may  be  a single-pass' type or  the many-pass
type (White cell).  In field studies,  the sample has  been  the open  atmos-
phere.   The  gas  correlation filter  cells  have usually  been just  a  few
centimeters in length, containing the  filter gas at  a partial pressure  of
several  torr.   The optical trimmer is a neutral  density filter that  can  be
adjusted to balance  the two beams. Wedges or screens can be used for this.
Beam-combining optics and  a photoelectric  detector  complete  the  train  of
optical  components.   The  beam  alternator was  the only moving  part  in  the
system studied in  the present experimental program.  Instead of beam  alter-
nation, one could  choose  to modulate  each beam at a different frequency  and
to monitor the intensity  ratio  of the  two frequency components.
  SOURCE
        FILTERS
                 SAMPLE CELL
                              BEAM
                            ALTERNATOR
\-
                                                                  ELECTRONICS
                                                                 AND RECORDER
Figure 1.  Nondispersive  analyzer in negative filter configuration.
     System  operation  can  be  understood by  considering  the  spectra  and

                                    118

-------
 detector signal  in  the  absence  and  presence  of  the  gas  being  measured
 (object  gas).   Let it be the case that the source and its  filters  produce  a
 range  of frequencies that encompasses two spectral  lines of  the  object  gas.
 Figure  2  (upper  half)  shows  the  spectrum  in  the  two  channels  of   the
 instrument  when there is no object gas in  the sample cell.  The  absorption
 lines  appear only in the  filter  gas  channel,  and  the  optical trimmers  are
 adjusted to give  equal  total  intensities  with a  consequent zero  detector
 signal at the beam alternation  frequency.   Then, when  the object  gas  does
 appear in  the  sample cell  (lower part  of  figure),  its  absorption lines
 appear in  both  channels,  the  intensities  at  the  detector become  unequal,
 and  a  signal with  the  beam alternation  frequency appears.   The system is
 called negative filter  because the intensity reduction is  greater in  the
 reference channel  than  in the gas filter channel.

     Interfering gases will increase  the signal when their absorption lines
 overlap  with the lines  of the object gas and  will decrease the signal  when
 the  lines do not  overlap.  If  the interfering  gas has many  lines in  the
 spectral region being used,  the  positive and  negative  interferences might
 nearly cancel each other.

     The following symbols are defined:

          Vj, V2 = Detector output for the  filter cell and reference
                    channels

          Tl> T2 = Transmission of the optical trimmers in the filter
                    cell  and  reference  channels (will be between 0 and
                    1)

          I(v)   = Source  intensity as  function of infrared radiation
                    frequency,  v

          R(v)   = Response  of detector

          Tp(v)  = Transmission of source filter

          TG(v)  = Transmission of gas-filled filter cell

          Tg(v)  = Transmission of sample under study

     For  the filter  cell channel,  V^ is  then given by an integration over
all frequencies:

                        CO
              Vl = TX  /   I(v)R(v)TF(v)TG(v)Ts(v)dv
                      o
If at  a given  frequency any  of  the  factors  has  a  zero  value,  then  that
frequency does not  contribute  to  the signal.

     For  the reference channel, the detector  signal is given  by  a similar
integration without  the factor  TQ(V):
                                    119

-------
       en
       o
       G   tr
       UJ   I
       LU   J=-=
              ^
                                                          Q.



                                                          O

                                                          OC
                                                          O

                                                          O
                                                          UJ

                                                          UJ
                                                          Q
                                                                                                                      UJ
                             A
                                                                                           A
s
UJ
D_
S
<
CO
uu
3
CD

O
UJ
CO
O
00

LU
oc
LU
           CJ
           cc
           UJ

                  /
                  \
D_X
CO O
UJ
o
z
UJ
GC
UJ
U.
UJ
CC
           oc
              UJ
coo
LU
O

LU
CO
oc
<
LU
a.
D_
<
CO
                                                     o
                                                     LU

                                                     CO
                                                     o
                                                             (D

                                                             CC
Q.X
CO O
UJ
o

LU
EC
UJ
LL.
LLJ
CC
                                                             UJ *l
                                                             0- I
                                                             CO O
                                                            H
                                                             (fl
                                                             a
                                                                                                                  to
                                                                                                                  }-)
                                                                                                                  tu
                                                                                                                  &
                                                                                                                  to
                                                                                                                 •H
                                                                                                                 T3
                                                                                                                  S
                                                                                                                  s
                                                                                                                             a
                                                                                                                            •H
                                                                                                                             en
 o
 (U
•P
 
-------
                   V2 =
I(v)R(v)TF(v)Ts(v)d
 Since  the missing factor,  TG(v),  is always  smaller than unity,  the  signal
 balance  V]_  =  V2  will  require  T^ to   be  greater  than  T2.    In  other
 words,  the optical trimming is done mainly in  the  reference  channel.

     Detector,  filters, and source  must  be  chosen to  give a  spectral  pass
 band  that favors  the  object gas  and discriminates against  possible  inter-
 fering  gases.   Because the sample  transmission,  Tg(v),  will  be  less  than
 unity,  filling the  sample cell will  cause  a  decrease  in both  Vj and  V2.
 If  Ts(v)  correlates  with  TG(v),  then   the  decrease  in  V2,  (AV2)  will
 be  greater  than   the  decrease  in V^,  (AV^).    If the  object  gas   is  the
 only species in the sample that absorbs in the band pass  of  the  system,  the
 correlation between  Ts(v)  and TG(V)  will  be  so  high  that  AV-L  will  be
 near  zero  and AV2 will   be  nearly  the  whole  signal.    AV2/V^  is  then
 the apparent integrated absorptivity in the  band pass region.

     Proper choice of  source, detector,  and filters will  center the  band
 pass around a  chosen  band of the object gas, thus  maximizing the  integrated
 absorption coefficient.  If the object  gas has a broad absorption band  in a
 spectral  region  relatively free  from  interference,  then the best  system
 design  would include  a filter with a  band pass  slightly  wider than  the
 spectral  band  of the  object   gas.   This would  be  the  case for  N02  or
 S02, or when using the C-H band to measure hydrocarbons.

     If the  object gas has a band with only  a  small number of lines  widely
 spaced, then most of  the  energy would  fall  between  the  spectral  lines  and
 the use of  a continuum source  with a band pass filter will not yield  a  high
 absorption coefficient.  In this  case,  a  heated  sample  of object  gas could
 be  used as  source.    The  emitted  lines  would match the  absorption lines
 quite well,  with  little energy falling  between the lines, and  therefore  the
 integrated  absorption  coefficient would  be  high.    This  technique  can  be
 used  for   detecting  HC1,  CO,   NH3,  NO,  and  other  thermally  stable  small
 molecules.
                          REMOVAL OF INTERFERENCES

     When another  gas  has no lines within the band pass selected for detec-
tion of the object  gas,  there will of course  be  no  interference.  If  there
are such lines,  they  can be blanked out of  both  beams  by placing enough of
said gas in a  filter  cell located in the combined beam portion of the  opti-
cal  train.    The  system will  then be  "blinded" to  the  interfering  gas.
Interference  between  pollutants  is  only an occasional problem  because the
main  absorption bands  of  the  major  gaseous pollutants  do  not generally
overlap each  other and  besides,  the absorption  by  each  pollutant  is  very
slight.   The absorption  by water  vapor,  however,  is  very great,  and the
main  bands  of  several  important pollutants,  such   as  N02,  NO, and  S02,
do fall within  strong  regions of  absorption by water.  For the detection of
these pollutants,  the  system must be blinded to  water vapor,

     A small cell  cannot  be used  to blind  the system to water vapor because

                                     121

-------
at  ordinary temperatures  the vapor  pressure  of  water  is  no  higher than
about  0.03 atmosphere.    Instead,  a  long  path  filter  cell  is  required.
Preferably, the product  of water vapor concentrtion times pathlength should
be  greater in the  filter  cell  than  in the  sample  cell.   Thus  the system
requires two  long path cells  in  tandem.   On first, consideration, it appears
that  two  long path cells  would  invalidate  the system design  by  being too
cumbersome and  too  sensitive  to misalignments.  Therefore,  a major aspect
of  this research  has  been to  find a  way to build long path cells  that are
compact,  inexpensive, and permanently  aligned.    This  effort has  been  a
success.   A cell  design  has been evolved  that  has  such a degree of stabil-
ity and permanency  of alignment  that  having two long  path  cells  in tandem
is not a limit on system performance.


                           LONG PATH  CELL DESIGN

     The multiple-pass optical technique of J.U.  White has been used in the
cell  design (8).   Three  telescope  mirrors,  all with  the  same  radius  of
curvature,  are  arranged to  reflect   the  beam  of  radiation back  and forth
Within a  tube.   The  beam is  re-focussed on  every  other pass  so  that the
energy is  conserved.

     The  only major  source  of   energy  loss  in  the  mirror system  is  the
absorption by the  mirror .surfaces.    If  R  is  the  reflectivity  of  the
mirrors,  the  fraction of  the  input  energy  that  will be  lost  at  n reflec-
tions will be (1 - Rn).   At  each encounter,  a clean mirror  will reflect
about 97.5  percent  of  the  incident energy.   For  such a mirror, the optimum
case is to  have 40  reflections,  which allows about 37 percent  of the energy
to  be  transmitted.    Adding more  reflections  than  this  will  decrease the
energy faster than  the pathlength is  increased.

     Normally, the  mirrors of  a  White cell- are held at the ends of the cell
body  in  spring-loaded  kinematic mounts  with precision  micrometers  for
adjustment.   Such  cells are  not only costly but are  subject  to  misalign-
ments from vibrations and  temperature changes.  In the present work, it has
been  found possible  to  avoid  these   limitations  by  cementing the mirrors
permanently within  a  glass pipe.  Alignment  is achieved  with  assistance of
a laser beam that  is reflected  back  and  forth between  mirrors  during the
cementing  process.  The  mirrors  are  trimmed to closely-fit inside the glass
pipes and  are fixed in place  with epoxy cement.  Each cell carries a trans-
fer optics  assembly for  coupling the  radiation in and out.   Cells of this
type  have  been  found to  maintain  their  alignment   over  a  two-year test
period.   Since the cell  design avoids precision  mechanisms,  there  is  no
great  expense involved  in  building   two  such cells  into  a   nondispersive
analyzer.
                            DETECTION SENSITIVITY

     Detection sensitivity  may be examined in terms of the absorption equa-
tion:  In  IO/I  = k-L-c.    Io  is  the  incident   energy within  the  pass
band,  and  I is  the  transmitted  energy.   k  is  the  absorption coefficient

                                     122

-------
averaged  over  the pass  band,  L is  the pathlength  through the  sample (in
centimeters),  and c  is  the  concentration  of the  absorbing  species  (its
partial pressure  in atmospheres).

     The detection limit  is assumed  to  be  reached when the signal reduction
due to sample absorption  equals  the  noise level  (N) of  the  total  signal
(S).  Since one is always working  in the  small absorption range, In  IO/I
at the detection  limit  is approximately equal to  the noise-to-signal ratio,
N/S.  There is ho apparent reason  why the  N/S of  a properly designed system
cannot be  10"1*  or smaller.   In the  laboratory breadboard system used for
tests described here, the N/S  was  approximately  10~3.   Detector  noise was
not the limitation.  Rather, mechanical stability of the optical system was
the limitation.   This can be  improved by better construction.

     In addition  to optical system stability, the second major objective of
system design is  to develop the  largest possible  absorption coefficient, k,
for, the object gas..  The higher  the  value  of k,   the shorter  the required
pathlength, L.  For a strongly  absorbing  species,  when the  pass band  is
properly centered on the absorption  band,  the value  of  k  can  be  as high as
50 cm~l atm~l.    If  the absorption  is  weak,   or  if  there  is  difficulty in
controlling the pass band, k might be as low as 1 cm"1 atnf *.

     Because the  system always  operates in the small absorption  range, the
output signal is  directly proportional  to  the  concentration  of  the  object
gas.

     The detection limit, c,  is  listed  in Table  1 as a function  of  noise-
to-signal level,  absorption  coefficient, and  pathlength.   Tests  conducted
in the present study have shown  that,  for  many important  air  contaminants,
the bottom line detection limit  of  10~9  atmospheres  is   achievable.    If
the system designer is  only  seeking  parts-per-million  sensitivity  rather
than parts-per-billion, then he  can  allow  the indicated shorter pathlengths
and higher values of noise-to-signal ratio.
                             LABORATORY  TESTS

Equipment

     An experimental nondispersive  analyzer system was  set up as diagrammed
in Figure 1, using the following  components in various  combinations:

     1. Sources:          Nernst  glower,  heated gas-filled cells

     2. Filters:          Long  path cell with humidified air,  10-cm cells
                          with  such filter  gases as methane, C02,  etc.

     3. Sample cell:      Multiple-pass   cells  of  the  type  described  in
• •..  .   .                   text.

     4. Beam alternator:  P.A.R.,  two-channel  lock-in  chopper  with  four-
                          aperture  blade

                                     123

-------
        TABLE  1.   DETECTION LIMIT AS FUNCTION OF SYSTEM PARAMETERS

Ratio of
noise-to-
signal
(N/S)
10-3
10-1*

Absorption
coefficient
k
(cm"1 atm"1)

1 	 1

i n 1
1U |

1 1
i I

	 - . 10 	 1

Path-
length
L
(cm)
lh+3
10+^
10+3
jn+^
io+3
I"*
10+
m4-3
+^
	 10 	
Detection
limit
C
(atm)
10
10
10
io
TO
TO
f A
IU
	 10
_6
-7
_7
-g
•— 7
-8'
~8:
g


     5.  Filter gas  and
         reference  cells:  Cylindrical gas absorption  cells 10 cm  long and
                           2.5 cm in diameter

     6.  Optical  trimmer:  Optical trimming was done by defocussing one beam
                           or the other at the detector
     7.  Detectors:


     8.  Electronics:


     9.  Recorders:

Ammonia Test Results
Mercury-cadmium-telluride  and  indium  antimonide
at liquid nitrogen temperature

Hughes  and  Digilab  pre-amplifiers  with P.A.R.
lock-in amplifier

Tektronix scope and Hewlett-Packard  recorder
     Ammonia  is  a compound  that is  thermally stable  and  has  many  widely
scattered  lines   in  its  spectrum*    It therefore  is  a  candidate for  the
measurement  technique  that  uses  a.  heated  gas  sample  as  the  radiation
source.   A standard  type of 10 cm  infrared  absorption  cell wrapped  with
heating  tape  was used  as the test  source.    Barium fluoride windows  were
cemented on with sili'cone cement.   About 50  to>rr  of N%; gas was  placed in
the cell, and  the temperature was  raised to  300°C.   The  resulting emission
spectrum is  shown in Figure  3,  bottom half.   The emission  continuum  that
underlies the  gas lines probabl3f came  from  the BaF2 windows,, whose emissiv-
ity would increase in the lower  frequency direction, as seen in the figure*
The upper part of the figure shows the spectrum of  the  radiation after it
had passed through the  gas filter  cell containing  a few torr of  ammonia gas
at  room temperature.   It  can  be  seen  that  the  strong  lines  have  been
removed from the emission spectrum,,  while the  weak  lines are only slightly

                                    124

-------
                                                                               CU>
                                                                               N
                                                                               •H
                                                                               CO
                                                                               !-i
                                                                               0)
                                                                               P-
                                                                               CO
                                                                               «H
                                                                               T3
                                                                               fi
                                                                               O
                                                                               a
                                                                               •H
                                                                               e?
                                                                               O:
                                                                              m-
                                                                               o
                                                                               4-J
                                                                               O'
                                                                               CD
                                                                               ft.
                                                                               CO
                                                                               CO
                                                                               CO
1VM9IS
1VNOIS
                                                                               00
                         125

-------
attenuated.  In  the  test  system,  the beam alternation, was therefore compar-
ing the reversed spectrum against the unreversed.   When ammonia appeared in
the sample  cell, it  caused an imbalance  between the two beams  with a high
absorption  coefficient.   If  the  underlying continuum had been eliminated by
the  use of  windows   that did not absorb  or emit  in  the  pass  band,  the
absorption  coefficient would have been still higher.

     Since  there are only a few  water vapor lines  in  the  ammonia emission
region, there was  no need for a  water vapor  filter  cell.   Furthermore, the
detailed structure in the ammonia band allowed a highly  selective ammonia
detection,  with  practically no interference  from  other compounds.  Ammonia
detection was tested using an early version  of  the  test optical system, in
which  the  pathlength was 30 meters  and  the  noise-to-signal ratio  of  the
system was  about 10~2.  In this  test, 5 ppm of ammonia in air gave a signal
about 30 times greater than  the  noise.

Sulfur Dioxide Results

     Sulfur  dioxide  was detected in tests using two radiation sources:  (1)
heated S02  gas,  and  (2)  the Nernst  continuum source with band-pass filter.
The strong  S02  band  that  falls  between 1300 and 1400  wave  numbers appears
to be  the  best  band  to use, but  this  band falls within a  region of strong
water  vapor absorption.    To prevent  interference  by  atmospheric humidity
changes, it was  therefore  necessary  to  blind the  862  system  to  water by
including in the optical  train a water vapor-filled long path cell.  Figure
4  (lower half)  shows the emission spectrum  of  the S02-filled  source cell
heated  to  about  300°C.   The  upper  part  of  the  figure shows  the altered
spectrum after the radiation was passed through a sample of S02  gas at room
temperature.

     The system  was  operated using the SO.  gas source, but without the long
path water  vapor filter  cell.   A sample  cell  of  30 meters  pathlength was
used.  Ambient  air was  slowly flushed through this cell and small amounts
of S02  were introduced periodically.  This  produced  an  S02 concentration
that rose  and fell  between 0 and 0.1 ppm.   The rise  and  fall  in output
signal under these conditions  is shown in Figure 5 (upper portion).

     The second radiation  source for S02 detection  was  tested  using  two
long path  cells  in  series,  the  first  being  the water-filled  filter cell.
In this case, the  Nernst  glower  was  operated at a temperature of about 1700
K, and  its  radiation between about  1230  and  1390 cm-1  was  isolated by use
of a  band-pass   interference filter.   Figure 6 shows  the  spectrum  of  the
radiation  at  three   places   in  the  optical  train:  (A)   after  it  passed
through  the long  path filter cell  containing water   vapor,  but  not  the
sample   cell,  (B) after  it  passed through both  the long path  filter cell
and  the long  path  sample  cell  with room  air,  and   (C)  after  it  passed
through both long path cells  and the S02-filled  filter gas  cell.   Nearly
all the absorption lines  in  spectra (A) and (B)  are due to water vapor.  It
is to  be noted  that spectrum (B) shows  only slightly  greater absorption
than spectrum (A),  illustrating  the blinding  of  the  system  to humidity
changes.  Spectrum (D) is a ratio plot of  spectrum  (C) divided  by spectrum
(B), showing the absorption  by the S02 in the filter gas cell.  This is the

                                     126

-------
                                                      SO2 EMISSION, REVERSED
                                                     SO2 EMISSION, UNREVERSED
               1 300 CM'1
1400
Figure 4.  Emission  spectra of sulfur dioxide in nondispersive  analyzer.
                                    127

-------
      Q.
      Si  0.1
      C
      _o
      *•*
      2
      4-»
      
      o
      o
      o
      CM
      o
      CO
           0 -
               0
*
5
                                       10
15
?0
                   Time (Minutes)
Figure 5.  Detection o_f SO2 by nondispersive analyzer.
           Path =  30 meters.
                Upper - Heated S0_ gas source, no water filter cell;
                         detection time constant = 3  seconds.
                Lower — Nernst glower source with band—pass filter;
                         30-meter sample cell in series  with 30-meter
                         water vapor filter cell; detection time
                         constant = 10 seconds.
                                    128

-------
            i
            to
            z
            LU
            PC
            CD
            z
            CO
            CC
            o
             1200 cm-1
                              1400
            I
            w
            z
            LU
            g
            o
            <
            oc
            cc
            o
                                            1200 cm-'
                                                             1400
1.0
                                           g
                                           5
             1200cm-'
                              1400
                                            1200 cm-1
                                                             1400
Figure 6.  Spectra  in nondispersive analyzer  tuned for sulfur dioxide.
           A -  Source radiation through 30-meter  filter cell
                with water vapor; sample cell  empty.
           B -  Source radiation through 30-meter  filter cell
                with water vapor and 30-meter  sample cell con-
                taining one atmosphere of ambient  air.
           C -  Radiation shown in (B) after passing through
                sensitizer filter cell with S02  gas.
          . D -  Spectrum (C) divided by spectrum (B).
                                     129

-------
absorption that is compensated by the optical trimming.

     This  second  862  detection system was  tested in  the  same  way  as the
first, by  periodically introducing small amounts of S02 into the air stream
flowing  through the sample  cell.   In this  case,  the S02  amount peaked at
about 0.7  ppm,  giving  the signal changes  shown  in the  lower part of Figure
5.

Hydrocarbon Results

     Hydrocarbons  were  measured using  the  C-H absorption band  near  2900
cm-1.   In  this case,  the Nernst  glower  source  was  used  with  a band-pass
filter.  For hydrocarbon detection,  the  analyzer  was sensitized by placing
butane in  the filter cell.  The infrared spectra at the  several points in
the  optical train  are shown in Figure 7.   Part  A is  the  spectrum  of the
source after the radiation passed through the 30-meter filter cell contain-
ing water  vapor;  B is  the spectrum after  the  radiation passed through both
the  30-meter filter cell  and  the 30-meter  sample  cell;  C  is  the spectrum
after the  radiation passed  through both  sample  cells and the butane-filled
filter gas  cell;  and D is the ratio  plot  of  spectrum C divided by spectrum
B, showing the  butane  band that must  be  compensated by optical trimming.

     This  dual  cell hydrocarbon-tuned  system  was   tested  by  flowing air
through  the  sample cell while  periodically  introducing  small  amounts  of
butane into the stream.   The changes  in  output signal for a 5—ppm peak con-
centration of butane are shown  in Figure  8.   It can be seen  in the  figure
that the signal-to-noise ratio was quite  high,  even  for  a detection system
response  time  of  3 seconds,  and that  it became  higher  as  response  time
lengthened.   Tests  showed that  system response to an  individual molecular
species  was  approximately  proportional  to  the  number of carbons  in the
molecule.   Thus,  1 ppm of hexane would give  about twice the signal given by
1 ppm of propane.

     The hydrocarbon detection sensitivity could have been further improved
by narrowing the  band-pass.   Figure 7  shows  a band-pass  of  approximately
900 cm-1,  while the butane  band itself  was only  about 100 cm-1 wide.   If
the band—pass filter had  a  total  width  of  about  200 cm-1,  the absorption
coefficient  would  have been  about  five times higher.

     The hydrocarbon-tuned system will not  respond to  methane.   The  system
is  blinded  to  methane  as   well  as   to  water vapor  because  methane  is  a
constituent  of  the air in the long path filter  cell.   Different choices of
filters could make the system sensitive to  methane and blind  to nonmethane
hydrocarbons.

Hydrogen Chloride

     The measurement  of  hydrogen chloride  gas  was  found  to  be difficult
because  of the weakness  of  its one  infrared absorption  band.    This  band
does fall  in a clear  region of the  spectrum  so  that there is  very  little
interference  from  other  gases,  but  the  band has  only a  small  number  of
absorption  lines  spread  across a wide spectral  region.  HC1  is  a compound

                                    130

-------
          tg
          03
        0-1-
0-1-
           2400
                                3200
                                         24001
                                                              3200
                                                    cm"1
          eg
          to
         O-1-
                                      0.1-r
                                        o
                                        s
                                       0-L
           2400
                                3200
                                         2400
                                                              3200
Figure 7.  Spectra  in nondispersive analyzer  tuned for nonmethane
           hydrocarbons.
           A - Source radiation through 30-meter  filter cell
               with water vapor; sample cell  empty.
           B - Source radiation through 30-meter  filter cell
               filled with water vapor and  30-meter sample cell
               containing one atmosphere  of ambient air.
           C - Radiation shown in (B) after passing through sensi-
               tizer filter cell with butane  gas.
           D - Spectrum (C) divided by spectrum  (B).
                                    131

-------
 t
 CO
 01

 1
 CO
                                           15

                                     MINUTES
20
            25
                                            30
                                                                 10
                                    SECONDS
                               TIME CONSTRUCT

   Figure  8.   Detection of butane by nondispersive analyzer.
              Path length = 30 meters
              Band pass = About 700 cm-1 (2500 to 3200)
              Each peak resulted from adding butane to flow
              of  room air through cell,  giving about 5 ppm
              butane  at the peak.


that  should  best  be  measured using as  its  radiation  source a cell contain-
ing HC1 vapor at  a high temperature.

      Tests  with  such a  source did  indeed show  substantial increases  in
sensitivity  over  tests using a continuum radiation source.   Several  diffi-
culties were encountered with this source, however,  preventing  a  full test
of the potential  of  the method.

      For  a gas-filled cell to be  successful  as  a source,  its windows  must
be both nonabsorbing and nonscattering.   If  the windows absorb,  they  also
emit  unwanted continuum  radiation that drastically  lowers  the absorption
coefficient  exhibited by the  object  gas.   If  the windows  scatter,  they will
introduce into the  optical  path continuum radiation  from the  cell  walls,
This  radiation will  also drastically reduce  the absorption  coefficient  of
the object gas.

     The  fogging  of  sodium chloride windows  on  the HC1 source  cell  proved
to be a  serious  problem.   Figure  9 shows  the emission from an HCl-fiJled
cell  (lower  spectrum), plus the reversed  spectrum (upper) after  the  emis-
sion was passed through  a cell containing  HC1 at room temperature.   It  can
be seen  in  the  figure  that  there  is  an  underlying  radiation continuum.
This  continuum largely nullifies the advantage  of having the HC1  emission
                                    132

-------
                                                                   o
                                                                   o

                                                                   CO
                                                            DC

                                                            O
                                                            LU
                                                            Q-
                                                            CO

                                                            UJ
                                                            O
                                                            DC
                                                            .13

                                                            O
                                                            C/D
                       O
                       O
                       O
                       CO
                       O
                       O
                       o>
                       CM
                                                    I
                                                    7
                       O
                       O
                       00
                       CM
E
o
                                                                   CM
                                                                   O
                                                                   O
                                                                   CD
                                                                   CM
                                  CO
                                  OJ
                                  60


                                  0)
                                 13
                                 •H
                                  ?J
                                  O
                                 r-f
                                                                              oo
                                                                              o
                                                                   O
                                                                   O
                                                                   in
                                                                   CM
                                                                              o
                                                                              01
                                                                              a.
                                                                              CO
A1ISN31NI
A1ISN31NI
     60
     •H
                                   133

-------
lines  in the source.   An attempt  to use CaF2  windows on  the  source cell
failed  because  the windows  caused  the HC1 gas  to disappear.   Perhaps the
HC1 was  attacking the windows and  being  exchanged for HF.   Success  in HC1
measurements  is  clearly dependent on source  cell improvements.

Other Compounds

     Chlorofluoromethanes,  vinyl chloride,  nitrous  oxide,  and  other com-
pounds  were mesured  in the nondispersive  analyzer  with  a high  degree  of
sensitivity.   These  experiments were  qualitative in  nature, but  in each
case,  the measurement  confirmed that  the nondispersive  method  is  easily
applied  to  the  measurement  of molecules  that  are  thermally stable and have
strong infrared  bands.
                                 CONCLUSIONS

     Nearly  all gaseous  air  pollutants can  be  measured by  their infrared
absorption,  using  the  nondispersive technique.   Filters that select desired
regions  of  the spectrum are  the key  optical  elements  that  permit  the
measurement  of  selected gases within  a  mixture.  It  has  been demonstrated
that  appropriate  combinations  of  radiation  sources,  gas-filled  filter
cells, band-pass  filters and cooled photo-detectors  can  yield much greater
detection  sensitivity  than has  been  available  to date.   It  has  also  been
shown that a filter  cell containing humidified  air will  blind a nondisper-
sive infrared system to  any  humidity  change in  the air being studied.

     These new  developments  will  allow the  construction  of  nondispersive
infrared  analyzers that  will measure pollutants  in   the  ambient  air  in  a
passive, nondestructive  way.   Automatic filter  changes will  allow a single
train of optical components  to  measure a multiplicity of  pollutants.  These
studies should  be  followed up  by  the building  of a  prototype  instrument,
based on  the physical  principles  discussed in  this article.   Success  with
the prototype could  then lead  to  commercial  realization of  a new improved
class of ambient air analyzers.
                                 REFERENCES

1.   Luft,  K.F.    1943.    Uber  eine  neue  methode  der  registrierenden
     gasanalyse mit hilfe  der  absorption ultraroter'strahlen ohne spektrale
     zerlegung.  Z. Tech.  Physik,  Bd.  24(5);97-104.

2.   Bartle, E.R., S. Kaye,  and  E.A. Meckstroth.   1972.   An in-situ monitor
     for HC1 and HF.  J. Spacecraft and  Rockets  9(11):836-841.

3.   Fastie,  W.G.,  and   A.H.   Pfund.     1947.     Selective  infrared  gas
     analyzers.  J. Opt. Soc.  Amer. 37(10):762-768.

4.   Fowler,  R.C.    1949.   A  rapid  infrared  gas  analyzer.    Rev.  Sci.
     Instrum. 20(3) :175-178.
                                    134

-------
5.   Hill, D.W., and T. Powell.   1968.  Nondispersive  infrared  gas  analysis
     in science, medicine  and  industry.   Plenum  Press,  New York, NY.   212
     pp.

6.   Sebacher, D.I.   1977.   A gas  filter  correlation  monitor  for  CO,  CHi^,
     and HC1.  NASA Technical Paper No. 1113, NASA Scientific and Technical
     Information Office.  28 pp.

7.   Wright,  N.,  and L.W.  Herscher.   1946.   Recording infrared analyzers
     for butadiene and  styrene  plant streams.   J. Opt. Soc. Amer.  36:195-
     202.
                                    135

-------
                 SAMPLING  VARIABILITY AND STORAGE STABILITY
                      OF VOLATILE ORGANIC CONTAMINANTS
                 Harold  G.  Eaton and Frederick W.  Williams

                          Naval Research. Laboratory
                               Washington,  D.C.

                                     and

                               Dennis E.  Smith

                               Desmatics,  Inc.
                              State  College,, PA
                                INTRODUCTION

     The Naval Research Laboratory (NRL) has been active  for  many years  in
the  detection,  analysis,   and  control  of  atmospheric   contaminants  in
enclosed inhabited spaces.   These  contaminants may  be  produced  by  a wide
variety of sources,  including  materials of construction,  instruments, human
effluents,  cigarette smoke,  refrigerant  gas  leakages,  paints,  decreasing
solvents, cooking, diesel fuel,  aerosol cans,  and  others.

     NRL initiated a program in 1956 to determine the identity  and concen-
tration  of contaminants found  in  the  atmosphere  of  nuclear  submarines.
Over 200 individual  components varying  greatly in  volatility and concentra-
tion were  identified,  and  the bulk of  these  proved to  be  hydrocarbons
(1,2).  Most  of  this earlier work involved vacuum or  steam desorption from
exposed  charcoal,  followed by  analysis  of  the  carbon  desorbate  (3,4).
Although this procedure produced invaluable information as  to  the identity
of the contaminants  present,  it was  found that the lower  molecular  weight
(MW), and hence  higher  volatility type  components,  were not  as  effectively
adsorbed on the  carbon  as  were the higher MW  low  volatility types' (5>»   A
review of  this  earlier  work has  been published (see  reference 6  and pre-
vious publications listed as references).

     At  present,  some  Navy submersibles  have on board  in-situ  type  ana-
lyzers that are  designed to  detect  a  few individual  components periodically
as well  as  total  hydrocarbon; concentrations  (7,8).   These  analyses  are
complemented  with  later laboratory analysis  of  gas  samples  collected  in
evacuated  stainless  steel bottles.   These methods  are  valuable  in  deter-
mining  contaminant  concentration  at  a definite  point   in time,  but  in

                                    136

-------
 practice do not provide  time-integrated contaminant  data over an  extended
 period,  such as for an 8-hour exposure.  For  this,  a time-integrated  sam-
 pling method is needed.

      Commercially  available  sampling  methods that are  applicable for  col-
 lecting  atmospheric  contaminants on  a time  basis include  the  use  of  an
 adsorbent  such as  charcoal,  Tenax GG,  and  others.   They usually  consist  of
 a  small  pump to draw the  air through an adsorptive tube to trap atmospheric
 contaminants.   Sampling  times,  however, are generally  limited from one  to
 two  hours  for  each  tube  exposure.   In addition,  as  discussed previously,
 low  MW components  ate not  quantitatively trapped  on  charcoal (5).   Holzer
 (9)  and Bertsch (10)  found  that,  in general,  the lower  limit  for effec-
 tively trapping contaminants on adsorbents was in the Cg  to C? range.

      NRL has designed, built,  and field-tested  a  time-integrated  sampling
 method capable  of collecting  contaminants,  including  low  MW components,
 over various time  periods (11-13).  This  procedure does  not  employ a pump,
 but  rather,  an evacuated  stainless steel container equipped with  a  critical
 orifice.   As  discussed  in  detail  below,  air  to  be  sampled  is  allowed  to
 leak into  the  container across a small  orifice until  a one-half  atmosphere
 pressure is  obtained.  Time of sampling is  controlled  by the  size of the
 orifice.

      Because of the  fact that in  many instances  these  samples  cannot  be
 analyzed for days  or  months after  samples  are  collected,  a statistically
 designed experiment  was   set  up to  study  the  sampling variability  and
 storage  stability  of  several volatile organic comtaminants within the time-
 integrating  sampling  containers.   These contaminants were representative  of.
 those found  in closed  inhabited environments (6).
                           EXPERIMENTAL PROCEDURE
General
     In  this  statistically designed study, a matrix  of  eleven contaminants
in  air,  listed  in Table  1,  were  evaluated  for  storage  stability  in the
time-integrated  gas  sampling bottles.   These contaminants  will be referred
to  by the numbers  given  in Table  1  in  this report.  Each contaminant in the
matrix was  evaluated at  a high  and  low  concentration.   In  general,  high
concentrations were  in the range  of 60 tb 110  ppm by volume per contaminant
and the  low levels were  in the  range of 5 to 12 ppm  by  volume per contami-
nant.  The use of  the two levels of concentration permitted the evaluation
of  concentration levels  as a  possible effect  on storage  stability  of  a
component.

     The original statistical  design for this  study was to  have a total of
12  bottles  each  for  the  high-  and  low-level  concentration matrices.   The
samples were  divided into three groups of four bottles  for each concentra-
tion.   In  the   procedure described below,  each  group  was  attached  to  a
heated  manifold   and the  test  gas  allowed  to   enter  the  four  sampling
bottles  through  a critical orifice until a one-half atmosphere  pressure

                                    137

-------
           TABLE  1.   CONTAMINANTS INVESTIGATED IN THE EXPERIMENT
1  Methane, CH^
2  Carbon monoxide, CO
3  Dichlorodifluoromethane  (R-12),  CC12F2
4  l,l,2,2-Tetrafluoro-l,2-dichloroethane
   Vinylidene chloride, CH2  = CC12
   l,l,2-Trichloro-l,2,2-trifluoroethane (
   Hexane, CHgCCH^^CHg
   Methyl chloroform, CH3CC13
   Benzene, C6H6
   Trichloroethylene, CHC1  = CC12
   Toluene,
 5
 6
 7
 8
 9
10
11
                                                   ,  CF2C1CF2C1

                                                  , CC12FCC1F2
was obtained.  The  bottles  were analyzed immediately to obtain the zero-day
data, and  then every  10  days  thereafter until 80 days had elapsed.

Time-Integrated Gas Sampling  Bottle

     The sampling flasks used in this  study were 1.64—liter type 304 stain-
less steel bottles  (Alloy  Products Corporation, Waukesha, Wisconsin).   As
seen in Figure 1, atop the  bottle at a right angle to the "T" connection is
an on-off  valve (Whitey,  part No. B-14DKM4).  This valve was used either to
evacuate the  sampler  when connected to  a  vacuum system or to  connect  to a
gas chromatograph (GC) for  evaluation  of the contaminants entrapped after a
sampling period.  On  top of  the  "T"  connection is the  gas  sampling valve,
which houses  the  critical  orifice  and protective filter  through which the
atmospheric sample  enters.    This valve was actuated by  one  complete coun-
terclockwise  turn.    A  complete  detailed  description  of  this  valve  and
analytical system  used  to evaluate  the  contaminants  has  been published
(11-13).

     In general,  the  gas sampling  bottle was  prepared  for atmospheric samr-
pling  by  first  attaching  the  bottle via  the  off-on  valve  to a  vacuum
system.  The  bottle was  evacuated  to  less  than one Torr  pressure.   At the
time of  sampling,  the sampling  valve  was  opened by  turning  counterclock-
wise.    The  gas  flow  into  the  container  through  the  critical  orifice
remained constant as  long as  the criteria for  critical  flow  existed (14).
In order for  this to  occur, the internal pressure of  the container remained
less than  one half  of the external  atmospheric pressure during the sampling
period.  Thus, the  length of  sampling  time depended on both the size of the
orifice and the volume of the sample container.
                                    138

-------
Figure 1.  Sampling flask with  critical  orifice  .valve.
     The equation that governs  the  gas  flow across  a critical orifice is as
follows (14):
                      m
                         =  10~5CAP
                                             Y±i
                                    RT
                                                                         (1)
where m = mass flow rate, g sec-1
      A = orifice throat area, cm2
      P = external pressure, Ncm~2
      Y = specific heat ratio, Cp/Cy
      M = molecular weight of the gas, g
      T = ambient temperature, °K
      R = gas constant, 8.31 joule mole~l  °K~1
      C = discharge coefficient (accounts  empirically  for  boundary layer
          effect having a range of 0.8 to  0.95 for  small orifices).
For this  study,  a sampling time  of  one hour  was  selected, which  required

                                    139

-------
 the use of a 38-micron orifice with, the  1.64-liter  bottle.             :;'."

 Sampler Preparation and Sample Collection

      Each group of four bottles  was  prepared for this study  by  alternately
 evacuating and  purging  the bottles  several  times  with helium.   With  one
 atmosphere of  helium left  in  the bottles  for 24 hours,  they were  then
 analyzed for potential  residual contamination  as   described  below.    Next,
 the four bottles  were  again evacuated  and attached  at  the  sampling  valve
 with 6.3 mm (1/4 in.) o.d. Teflon  tubing to a heated  (80°C)  stainless-steel
 manifold, as  shown schematically in Figure 2.
                     HEATED
                     MANIFOLD
                SHUT OFF
                VALVE
Figure 2.  Sampling manifold with one bottle attached.
     One end of  the  manifold was connected via Teflon  tubing to a test gas
bottle containing  the contaminants listed in  Table  1 (Union Carbide Corpo-
ration, Linde  Division, Keasby,  New Jersey).   After  purging  the manifold
for several minutes  with the test gas,  the  opened end  of  the  manifold was
closed.   After  an in-situ  analysis  by GC  of the  manifold gas,  the four
sampling valves  were opened.

     One atmosphere  pressure of test  gas was  maintained within the manifold
by adjusting the flow of gas from the  test  gas  bottle.  At  the  end  of the
sampling period,  the sampling  valves  were closed.   The bottles  were then
                                    140

-------
  analyzed to establish  the Day 0  point and every  10 days  thereafter to Day
  80.


       The above  procedures were  conducted  on the  remaining groups  for both
  high— and low-concentration  levels.



                           GAS CHROM&TOGRAPHIC ANALYSIS

       A   Beckman  (Beckman  Instruments,  Inc.,  Fullerton,  CA)   Model  GC-5
  equipped with  dual  hydrogen  flame  ionization  detectors  (FID)  was  used  to
  evaluate the  storage  stability  of  the gases collected  in  the  stainless
  steel  bottles.    This  analyzer  was  modified  considerably to  analyze  the
  subambient  pressured bottles.

       A  schematic of  the gas  handling  and analytical system is presented  in
  Figure  3.  Subsystem I was  designed  to  analyze  all  of the  organic compo-
  nents with the  exception  of  methane  and carbon  monoxide, which were  ana-
  lyzed on Subsystem II.
                      SAMPLE
                       INLET
   AUXILIARY
      He
MAIN
He
SUPPLY
  AUXILIARY
                                                10% DC-20O pkOOcstks]
                                                 CHROMO.G., 45/60.

                                               3.O5m x 6.35mm (10ft it '4in.)
                                                                        JAMPLIFIER
               RECORDER  INTEGRATOR
    PORAPAK T, 80/fOO    [[—["[—[  |—
0.6rnri)c3.l75rnm(2ftic
                                              BACKFLI
                                              VALVE.
                         H  COMPUTER  TTY


                           l AMPLIFIER
                                                             VENT
                   AUXILIARY
 	             He
 PORAPAK T, 80/100   . l—'
 0.6tm x3.IT5mm {^ft x /sin.) ,	,
 ' UUUUUUOU	1=	1    I.  I—
                     VACUUM
                                               MOLECULAR
                                             SIEVE 5A, 7O/80,
                                             t.83m.x6.35mm
                                               (6ft * 1/4. rn)
                   NICKEL
                  CATALYST
                   (330-C)
 Figure  3.   Gas-handling and analytical systems.
                                       141

-------
     The  gas  sampling  and  backflush  valves  were  manufactured  by  Carle
Instruments  (Carle Instruments,  Inc.,  Fullerton, CA).   Other valves  used
were stainless-steel  ball valves (V1} V5,  V6,  V7 ),  toggle valves  (V2,  V3,
Vt,), and  fine metering valves  (Vs,  Vg).    All  tubing was  stainless  steel.
Tubing connections  to and from  sampling  valves,  Subsystems I and  II,  were
1.6 mm (1/16  in.)  o.d.  The  sample  loops  were 6.3 mm  (1/4  in.) o.d.  and of
8 cm3 and 40  cm3 volume for  Subsystems I and II,  respectively.   The  tubing
from valves  V2 and V,(  was 3.2  mm  (1/8  in.)  o.d.   To  measure  the  sample
pressure  accurately,  a Wallace  and  Tiernan (Belleville, NJ), Model 62A-4A-
0100D, 0  to  5000  Torr,  pressure  gauge was  connected by a  cross,  via valve
75, to the sample  line.   A vacuum system was connected  at  valve  V6 leading
also to the  cross.

     Helium  was used as  the  carrier  gas  and  was purified  by  the  use of
tandem columns  of  molecular  sieve and charcoal immersed in a liquid  nitro-
gen container.  Flow  controllers in the  main helium  supply to  the sampling
valves were  standard  parts of the GC.  Secondary or  auxiliary  helium flows
were brought in downstream  of  the  flow  controllers  through Va  and Vt,. to
provide  a high-volume  flow . for  rapid   compression  of  the  sample  during
injection (15).   In  Subsystem  I,  valve Vg  was   used  as  a  restrictor  and
employed  to  prevent the FID from blowing out when the  backflush valve  was
actuated.  In Subsystem II,   the detection  of CO  with' the  flame  ionization
detector  was accomplished by  converting  CO  to CH^ over a  nickel catalyst
(16).  To prevent  the  contamination  of  the molecular sieve  columns  by  the
other contaminants in the sample, a backflush  valve  equipped  with two pre-
columns,  both Porapak T,  80/100 mesh,  0.6 m  by 3.2 mm,  was  used.   Valve V9
served as a  restrictor to minimize  pressure fluctuations when  this back-
flush valve  was actuated.

     The  sampling  valves  for  both  subsystems were  operated at  80°C,  the
nickel catalyst at 300°C, and the detectors  at 150°C.   The backflush valve
for Subsystem II was  operated at room temperature.

     The  flow rate of the carrier gas through  both  Subsystems  I  and  II was
60 ml/min.   To obtain  ideal  detector  performance from the flame ionization
detector,  an  additional  60  ml/min of  helium flow  (a  total  flow  of  120
ml/min) was  passed through the  detector  (not  shown  in Figure 3).  Hydrogen
flow for  the FID was  50 ml/min  in each system.   As  shown,  H2 for Subsystem
II was added at the catalyst, whereas for  Subsystem  I,  H2 was  added at the
detector.    Air flow to  the detector  was  300  ml/min.   These  operating
conditions   were   maintained   throughout  the  evaluation  of  the  sampling
bottles.

     The  response from  the  FID through  the  amplifier  was  fed  into   a
Hewlett-Packard,  Model 3370B,  integrator  (Avondale,  PA)  and a  1-mV strip
chart  recorder.   The digitized  signal  was  processed  by  a Hewlett-Packard
minicomputer, Model 2116C.
                                     142

-------
                              SYSTEM PROCEDURE

Preparation of  the  Sampler for Analysis

     The  sampler was prepared for  evaluation of its contents  by connecting
the off-on valve to  the  gas handling  system at the inlet  port.   With the
sampler  closed,  valves  Vj,  ¥3, V5,  and  Vg were  opened  to  evacuate the
sampling  system to  less  than 1-Torr pressure.   Valve  Vg  was then closed.
By opening the  off-on valve  of  the sampler, valve V}  was  used  to  meter in
the sample to a pressure  of 100 Torr to both Subsystems I and  II.   Valve Vi
was then  closed.  Valve V3 was  then closed to  separate  the  sample  loops of
both subsystems during analysis.

Subsystem-I Analysis

     With the  sample  contained  in the  sample  loop at a  pressure of 100
Torr, the sampling  valve  was rotated to place the loop  in the  carrier  flow.
At the  same time,  valve  V2 was  opened  to rapidly  compress the sample and
force it   into  the  smaller-diameter tubing leading  to  the   backflush  valve
and column.  After  20 seconds the  sampling valve was returned to  its  orig-
inal position and valve V2 was  closed.   The flow controller now controlled
the flow  of  the carrier gas.   The  analysis  of  components  was conducted by
program temperature, as indicated in Figure 4.
                   IDENTIFICATION
               PEAK NO.   COMPONENT
                  1
                  2
                  3
                  4
                  5
                  6
                  7
                  8
                  9
                  10
          METHANE+AIR
          R-12
          R-114
          VINYLIDENE CHLORIDE
          R-113
          HEXANE
          METHYL CHLOROFOM
          BENZENE
          TRICHLORO-
          ETHYLENE
          TOLUENE
10
                         6   8 10 12 14 16 18 20  22  24  26  28  30
                          TIME — MINUTES
024
rt	
            60
                                         ,  ,  ,  ,
                                         I  I  I  I        I
                                         708090100    110
                                         - START TEMP. PROGRAMMING
                                           TEMPERATURE, °C
Figure 4.  Subsystem-I  component analysis.
                                    143

-------
 Subsystem-II Analysis

      Subsystem II was designed to  detect  CH^  and CO.  Although  the  detec-
 tion of  these components  was fairly  routine and  was  accomplished on  a
 molecular sieve column, caution was  exercised so that this  column did  not
 become  contaminated with the other components  of  the sample.   To  avoid this
 contamination, two  precolumns, Porapak T, 80/100 mesh,  0.6  m  by 3.2  mm,
 operated at  room temperature were  used to separate CH^ and CO from  the
 other gases in the injected sample.  After  the CEL  and CO passed through,
 the  other gases still on the precolumn  were  backflushed.

      The sequence of operation (Figure  3) was as follows.  With  the  sample
 enclosed in the sample loop at a  pressure  of  100 Torr, the  sampling valve
 was  rotated to place the loop  in  the carrier gas  stream.   At  the  same time,
 valve V^  was  actuated  for 35  seconds  to  rapidly  inject   the  sample  as
 described for Subsystem I.   At the end  of this time  the injection valve  was
 returned to  its  original  start position and  valve  Vij was closed.   At  55
 seconds  the valve was returned to its original start position and valve V^
 was  closed.  By this time,  the CH^  and  CO had passed through the top pre-
 column  and were  on the molecular  sieve column.   The  backflush  valve  was
 actuated,  turned  clockwise, thereby allowing the auxiliary carrier to flow
 in the  opposite direction  on  the  top column carrying  the contaminants  out
 through  valve Vg.   The  main carrier flow continued on the  bottom  precolumn,
 now  in the opposite direction, to  the molecular sieve column..  The reversal
 of flow on  the precolumns   did not  affect  the normal  forward flow  of  the
 molecular sieve column.

      To  detect the  response  of  CO with an FID, as  stated  previously,  a
 nickel  catalyst was used to convert  CO to CHv  Figure  5 illustrates  the
 results  obtained  at a column temperature of  60°C.
                INJECT  AIR
                           2468   10  12

                                TIME-MINUTES
Figure 5.  Subsystem-II component analysis.
     At the conclusion of this analysis cycle, valves Vy and YS were opened
to vent the positive pressure.  Valve Y7 was then closed.  Valves V5 and Vg
were opened to prepare both Subsystems I and II for the next analysis.
                                   144

-------
                           RESULTS AND DISCUSSION

      As discussed previously, the sampling bottles  were  divided into groups
 by high-  and low-concentration.    Each group  consisted  of  four  bottles.
 They  were filled  to  one-half atmosphere pressure,  stored at room  tempera-
ture,  and analyzed,  but at  different  times.   The number  of  days  between
 filling and  analysis, however, was  the  same  for all groups.  Due  to equip-
 ment  defects associated with the  sampling bottles  during the  experiment,
•.such  as leaks into the subambient pressured bottles,  several  bottles had to
 be rejected.  As a  result,  9 high- and  7 low-concentration  samples  were
 evaluated.     This   situation  resulted  in   a  statistically  unbalanced
 experiment,   since  the  number   of   samples   analyzed  were  different  for
 different', groups of the contaminants.         ,           ,

      Two analyses  were made  of  each  bottle  on  Day 0  and  every   10  days
 thereafter,  until Day  80.   The  response  average  of  the contaminants  from
 all  of the  groups  is presented in  Figures  6  and 7 for  the high- and  low-
 concentration mixtures,  respectively.    At  the  low-concentration  level,
 Figure 7,  contaminants #1 (CHt,.) and #2  (CO)  both responded high on  Day 70,
 but returned to  a reasonable concentration response on Pay  80.   In  general,
 at both .concentration levels  the  concentration  of  the  contaminants,  and
 hence  detector  response, fluctuated very little during the  80-day  period,

 ;, ... A more  detailed examination  of the data base  was  conducted using the
 analysis  of  variance,  a versatile  statistical  technique  for  studying  the
 overall variation in a set  of observations.  Analysis of variance  permitted
 an examination  of the effects of possible sources of  variability.   In addi-
 tion,  it was used to test whether any  portion of the overall variation was
 directly  attributed -to any of these sources.   This type  of test was  based
 on the observed value  of  an F-statistic computed from the data.    In  gen-
 eral,  the larger  the  observed  value,   the  more  evidence   that the  source
 being  tested had a contributing effect  to  the overall variation.  However,
 the number of observations  on which each  specific test was  based  and  the
 associated degrees of freedom had to be taken into account.

     The  observed  F_ statistic  value  was  compared  with  tabulated  values
 (17,18).   An observed value  that was  larger  than  the   tabulated  value
 corresponding  to  a  0.05 significant  level,   for example,  was   said to  be
 statistically  significant at  the  0.05  level.   This  indicated  that  if  the
 source  being tested had no  real  contributing effect  to  the overall  varia-
 tion,  there  would be  less  than a 5 percent  chance of  observing  an Jj-statis-
 tic  value that  large.    Thus, such  an observed  value  would  offer  strong
 evidence  that the source being  tested  does,  in fact, "have a  contributing
 effect.

     Table 2 presents a summary of  the  observed values  of  the Jj-statistics
 obtained from the  various analyses  of variance.   This table indicated  those
 values  that  were statistically significant at  the 0.05  significance  level.
 Althpugh  the results  in the  table  provided no  evidence  of variability  due
 to time of analysis  (i.e.,  no evidence  of  time trends or loss of components
 in the gas mixture due- to  adsorption)   during  the 80-day  period,  they  did
 reveal  significant group and  flask  variability.   It  was  concluded,  there-

                                   .145

-------
                       0   10   20   30  40  50  60  70  80
                                   DAY ANALYZED
Figure  6.  Response   averages   from  high-concentration  mixture   sampling
            bbttles.   For contaminant number identification, see Table  1.
                                     146

-------
                           10   20  30  40  50  6O  70  80
                                  DAY ANALYZED
Figure  7.  Response   averages  from   low-concentration  mixture   sampling
            bottles.   For  contaminant  number identification, see Table  1.

                                    147

-------
        TABLE 2.  THE OBSERVED VALUES OF THE  F-STATISTICS  OBTAINED
                  FROM THE ANALYSIS OF VARIANCE

Source of
variation
High concen-
trations
Croups
Flasks
TiBG
Lou concen-

Groups
Flasks
Tlae
Contaminant
1

4.12
1.91
1.44


7.79*
4.61*
1.51
2

1.60
1.91
1.08


6.87*
3.39*
1.73
3

12.25*
3.70*
0.20


0.00
2.40*
0.27
4

0.65
4.44*
0.50


0.15
4.50*
0.26
5

33.83*
32.5*
2.46


3.29
11.07*
0..-43
6

19.09*
4.83*
0.33


0.62
9.01*
0.33
7

10.73*
6.16*
2.12


5.48
20.74*
0.39
8

10.30*
7.48*
2.72


1.87
3.73*
1.71
9

5.12
4.70*
0.45


0.12
3.49*
0.78
10

11.46*
4.92*
0.85


1.42
1.92
0.94

11

6.90*
5.61*-
0.53


20.72*
13.31*
0.68

Total

13.15*
6.04*
1.25


8.96*
8.51*
1.08

fore, that  there was noted variability  between different  observations  due
to the following three sources  of variation:

          Source 1:  Variation  between groups  (fillings)

          Source 2:  Variation  between flasks

          Source 3:  Variation  between analyses (measurements)

     The variation  between groups reflected overall differences  in average
concentrations in the groups.   For  example, if for  one particular  group of
high-concentration  observations,  the  average was  83.0  ppm while in a second
group the  average  was  84.2 ppm,  this  difference would  be  ascribable  to
between-group  variability.   Similarly,  variation between  flasks  reflected
overall differences  in average  concentrations  in the flasks within the same
group.  Likewise,  the  variation between analyses  reflected overall differ-
ences between  analyses (measurements) performed on the same flask.
     If 02  denotes the variance  of any  observed  concentration based  on a
single  analysis,  then a2  = a2-  +  a|  + a| .   The  three components  on the
right-hand  side denote the  port ion "of variance  due to  group differences,
flask  differences  and analysis  differences, respectively.   Each  of these
three  components were  estimated statistically from the analysis of variance
based  on  the available experimental data.  (See (17,18) for a discussion of
estimation  process.)

     Tables 3  and  4 exhibit the  sources  and estimated  components  of vari-
ance for  each of  the  eleven contaminants  considered.    These figures  also
indicate  the  percentage  of  the  variance a2  attributable  to each  of  the
three  sources.   The  estimated  components  (denoted  by  a2  a2, and  cr2V can be
used to construct  approximate  confidence  intervals base§" ofr~a single obser-
vation or on a  number  of  observations.

     In general,  shipboard  sampling  will  involve  taking  samples  of  an
atmosphere  in  a number of flasks.   Because each of these  samples  is taken
independently  of each other, variability  in sampling  includes both  filling

                                    148

-------
     TABLE 3.  AMOUNTS AND  SOURCES  OF VARIANCE IN A SINGLE OBSERVATION
               OF CONTAMINANT  CONCENTRATION '(HIGH LEVELS)
Contaminant
                    Estimated
                    variance
    Percent of estimated variance
attributable to differences between:
Groups
Flasks
Analyses
1
2
3
4'
5
6
7
8
9
10
11
2,49
3,80
7.18
12, 10
4.06
3.77
6,12
1.30
4.73
3.81
4,79
0
0
5
0
19
19
18
35
2
22
26
48
49
72
83
50
57
46
27
67
36
29
52
51
23
17
31
24
36
38
31
42
45

     TABLE 4»  MOUNTS AND  SOURCES  OF VARIANCE IN A SINGLE OBSERVATION
               OF CONTAMINANT CONCENTRATION (LOW LEVELS)



Contaminant

1
2
3
4
5
6
7
8
9
10
11
Estimated
•variance
(a2)

0,45
0.31
0,15
0.11
0.04
0.08
0.06
0.03
0,10
0.06
0.07
Percent
attributable
Groups
<«fc
8
9
0
0
12
0
18
7
0
2
48
of estimated
variance
to differences between:
Flasks
<-4)
52
62
80
81
64
68
71
21
67
16
38
Analyses
C0I)
40
29
20
19
24
32
11
72
33
82
14

and  flask  variability.  Thus,  the sampling variance  is  equal to cr^  + crz.
It should  be noted  that, by the manner in which the time—integratedexperi—
ment  was  conducted,  each   of  these individual  components  was  able  to be
estimated.   In  other words,   it  can be  Judged  how  much  of  the  overall
.sampling variability  is  due to differences  in filling and  how  much is due
to differences in flasks.

     The ultimate aim of  sampling is to obtain  an accurate estimate of the

                                    149

-------
contaminant  concentrations  within the  environment  of  interest.    In the
experiment,  this  is  the  test gas bottle, while  in  the "real world" this is
some specific  enclosed environment.   To increase the accuracy  of an  esti-
mate, the number  of  flasks (the samples  taken)  may  be increased and/or the
number  of analyses  done on each flask  may  be increased.   Of course,  there
will be trade-offs to  be  made between greater accuracy and greater cost and
time for a larger number  of  samples or analyses.

     Based on  the data from  this  sampling  experiment, Tables 5  and  6 pre-
sent  approximate  95  percent  confidence  intervals  (C.I.)   for  the  true
contaminant  concentration based on the  average  observed  concentration iii N_
flasks where _M analyses are  done for  each flask.  Thus,  the entry  for N_ = 1
and M^ -  1 indicates  the  estimated  accuracy of  a single  observation, while
the other entries indicate how increasing  the number of  flasks  or analyses
affects  the  accuracy.   The  formula  for  the  approximate  95  percent  confi-
dence intervals is:
                   95%  C.I.  = ±2
                                      + a2/N]+[~2/NM]  %
                                         a  —    A —
(2)
where the values  of
                         ~>  and Q-2  are obtained from Tables  3  and 4.   As a
                       , G~>      Q-                                    .
point of interest, it  should be noted that  regardless  of how many analyses
are done for each flask,  the confidence  interval can never be smaller than
                              ±2
                                    +
                                       F   —
(3)
    TABLE 5.  APPROXIMATE  95  PERCENT CONFIDENCE INTERVALS FOR THE TRUE
              CONTAMINANT  CONCENTRATION IN THE TEST GAS BOTTLE BASED ON
              THE OBSERVED AVERAGE  CONCENTRATION IN N_ FLASKS WITH M
              ANALYSES FOR EACH FLASK.   (HIGH CONCENTRATION LEVELS)

Con-
tami-
nant
1
2
3
4
5
6
7
8
9
10
11
Total
Avg.
cone.
(ppm)
74.1
79.7
77.7
84.3
81.4
78.4
103.5
57.9
84.5
74.4
80.9
876.8
Approximate 95 percent confidence intervals (ppm)
N = 1
H = 1
±3.2
±3.9
±5.4
±7.0
±4.0
±3.9
±5.0
±2.3
±4.3
±3.9
±4.4
±25.4
5
1
±1.4
±1.7
±2.4
±3.1
±1.8
±1.7
±2.2
±1.0
±1.9
±1.7
±2.0
±11.4
10
1
±1.0
±1.2
±1.7
±2.2
±1.3
±1.2
±1.6
±0.7
±1.4
±1.2
±1.4
±8.0
1
5
±2.4
±3.0
±4.8
±6.5
±3.5
±3.5
±4.2
±1.9
±3.8
±3.2
±3.5
±21.5
5
5
±1.1
±1.3
±2.2
±2.9
±1.6
±1.6
±1.9
±0.8
±1.7
±1.4
±1.6
±9.6
10
5
±0.8
±0.9
±1.5
±2.0
±1.1
±1.1
±1.3
±0.6
±1.2
±1.0
±1.1
±6.8
1
10
±2.3
±2.9
±4.8
±6.4
±3.4
±3.4
±4.1
±1.8
±3.7
±3.1
±3.4
±21.0
5
10
±1.0
±1.3
±2.1
±2.9
±1.5
±1.5
±1.8
±0.8
±1.7
±1.4
±1.5
±9.4
10
10
±0.7
±0.9
±1.5
±2.0
±1.1
±1.1
±1.3
±0.6
+ 1 7
i • ^,
±1.0
±1.1
±6.6

                                     150

-------
TABLE  6.   APPROXIMATE 95 PERCENT CONFIDENCE INTERVALS FOR THE TRUE
           CONTAMINANT CONCENTRATION IN THE TEST GAS BOTTLE BASED ON
           THE  OBSERVED AVERAGE CONCENTRATION IN _N FLASKS WITH M
           ANALYSES  FOR EACH FLASK.   (LOW CONCENTRATION LEVELS)

Con-
tami-
nant
1
, 2
3
4
5
6
7
8
9
10
11
Total
Avg.
cone.
(ppm)
9.1
11.2
8.6
6.5
7.2
8.0
10.5
4.7
7.4
7.3
6.1
86.8
Approximate 95 percent
N = 1
M = 1
± 1.3
± 1.1
± 0.8
± 0.7
+ 0.4
± 0.6
± 0.5
± 0.3
± 0.6
± 0.5
± 0.5
±3.8
5
1
± 0.6
± 0.5
± 0.4
± 0.3
± 0.2
± 0.2
± 0.2
± 0.1
± 0.3
± 0.2
± 0.2
± 1.7
10
1
± 0.4
±.0.3
± 0.2
± 0.2
+ 0.1
± 0.2
± 0.2
± 0.1
± 0.2
± 0.2
± 0.2
± 1.2
1
5
± 1.1
±1.0
± 0.7
± 0.6
± 0.4
± 0.5
± 0.5
± 0.2
± 0.5
± 0.3
+ 0.5
± 3.5
confidence intervals
5
5
± 0.5
± 0.4
± 0.3
± 0.3
± 0.2
± 0.2
± 0.2
± 0.1
± 0.2
± 0.1
± 0.2
+ 1.6
10
5
± 0.3
± 0.3
± 0.2
± 0.2
± 0.1
± 0.2
+ 0.1
± 0.1
± 0.2
± 0.1
± 0.2
+ 1.1
1
10
± 1.1
± 1.0
± 0.7
± 0.6
± 0.4
± 0.5
+ 0.5
± 0.2
± 0.5
± 0.2
± 0.5
+ 3.5
(ppm)

+
+
+
+
+
+
+
+
+
+
+
+
5
10
0.5
0.4
0.3
0.3
0.2
0.2
0.2
0.1
0.2
0.1
0.2
1.6
10
10
± 0.3
± 0.3
+ 0.2
+ 0.2
± 0.1
+ 0.1
± 0.1
+ 0.1
+ 0.2
±0.1
+ 0.2
+ 1.1

The  data  in Tables  5  and 6  show that  increasing N_  is  more  effective  in
narrowing the limits than  increasing M.


                                   SUMMARY

     We have  developed a time-integrated  sampling method which  is  capable
of collecting contaminants, including  low  MW components,  over variable time
periods.  Storage  stability of  several  components at one-half  atmospheric
pressure within  the  stainless-steel flasks  has  been proved  over  an 80-day
period.  An analytical  technique based on GC has  been developed  to  analyze
the subambient pressured flasks.
1.
2.
3.
4.
                                REFERENCES

     Piatt, V.R.   1960.   Chemical constituents  of  submarine  atmospheres.
     Chapter 1 in NRL Report 5465.

     Carhart,  H.W. ,   and  V.R.  Piatt.    1963.    Chemical  constituents  of
     nuclear submarine atmospheres.  Chapter 8 in' NRL  Report  6053.

     Nester, F.H.M, and W.D.  Smith.   1958.   Submarine habit ability— Atmos-
     phere sampling and analysis.  NRL Report 866.

     Nester, F.H.M.   1960.   Trace  contaminants, sampling,  and analysis.
     Chapter 4 in NRL Report 5465.
                                    151

-------
5.
Umstead,  M.E.,  J.C. Christian,  and J.E. Johnson.   1960.   A  study of
the  organic vapors  in the atmosphere  of the USS  SKATE.   NRL Report
1057.
     Saalfeld,  F.E.,  F.W.  Williams, and R.A. Saunders.
     tion of  trace  contaminants  in enclosed atmospheres,
     tory 3:8-15.
                                                     1971.   Identifica-
                                                      American  Labora-
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
Saalfeld,  F.E.,  and  J.R.  Wyatt.   1976.   NRL's central atmosphere moni-
tor program.  NRL  Memorandum Report 3432.

Eaton, H.G., M.E.  Umstead,  and W.D. Smith.  1973.   A total hydrocarbon
analyzer  for use in nuclear submarines  and  other  closed environments.
J. Chromatogr. Sci.  11:275-78.

Holzer, G., H.  Schanfleld, A. Zlatkis,  W. Bertach, P.  Juarez,  and H.
Mayfield.   1977.  Collection and analysis of  trace  organic emissions
from, natural sources.   J.  Chromatogr.  142:755-64.

Bertsch,  W., R.  Chang, and A. Zlatkis.   1974.   The  determination of
organic volatiles  in air pollution studies;  characterization  of pro-
files.  J. Chromatogr.  Sci.. 12:175-82.

Eaton, H.G., J.P.  Stone, and F.W.  Williams.   1976.    Personal atmos-
pheric gas sampler with a critical orifice:  Part  1  — Development and
evaluation,  NRL  Report  7963;  Part  2  -  System  for  handling  and
analyzing  the gas  from  the  sampler, NRL Report  7960.

Stone,  J.P.,   H.G.  Eaton,  and  F.W.  Williams.    1975.    Atmospheric
sampling;  description of a small  flow-control  valve unit.   Rev.  Sci.
Instrum. 46:1288-89.

Williams,  F.W.,  J.P. Stone,  and  H.G.  Eaton.   1976.    Personal atmos-
pheric gas sampler  using  the critical  orifice concept.   Anal.  Chem.
48:442-45.

Anderson,  J.W.,  and  R. Friedman.   1949.   An  accurate gas  metering
system for laminar flow studies.  Rev.  Sci.  Instrum.  20:61-66.

Umstead,  M.E.    1974.   A sampling  technique for  subambient  pressure
systems.   J. Chromatogr.  Sci.  12:106-08.

Porter,, K.,  and &.H. Volman.   1962.    Flame ionization detection of
carbon  monoxide  for   gas  chromatographic   analysis.     Anal.  Chem.
34:748-49.                                                           ~~

Bennett,  C.A.,   and  N.L.  Franklin.    1954.   Statistical  analysis  In
chemistry  and the  chemical industry.   John  Wiley and  Sons,  Inc.,  New
York, NY.

Box, G.E.P.,  W.G. Hunter,  and J.S.  Hunter.   1978.   Statistics  for
experimenters.    John Wiley and Sons, Inc., New  York,  NY.
                                    152

-------
                  EXPOSURE TO PERCHLOROETHYLENE ASSOCIATED
               WITH  THE  USE OF COIN-TYPE DRY CLEANING MACHINES
                                R.H.  Jungers

                    U.S.  Environmental Protection Agency
                         Research Triangle Park,  NC

                                 S.J.  Howie

                          FEDCo  Environmental,  Inc.
                              Cincinnati,  OH
                         THE  PURPOSE  OF THIS STUDY

     Nearly  all  coin-type  dry cleaners use  perchloroethylene  (PERC)  as  the
cleaning  solvent (1).   Since PERC  is  a suspected  carcinogen  (2,3),  the
potential  of  human  exposure to it by  such widespread uses  as  dry cleaning
is an  area of great concern.   Previous studies have shown  that  relatively
high  concentrations may be found in  self-service  laundries  (4) and  that
exposure to PERC may be expected  in residences  near its  use  as well (5).

     This  study was initiated  as  the  result  of  a request by  the Interagency
Regulatory Liaison Group, and  was'sponsored  by  the Environmental Monitoring
Systems Laboratory  of the  U.S.  Environmental  Protection Agency  (EPA),  in
conjunction with  the  U.S.   Consumer Product  Safety Commission  (CPSC).   The
purpose was  to obtain  data  demonstrating the potential  for  public exposure
to PERC that  may result from  coin-type dry  cleaning in  self-service  laun-
dries.
                            GENERAL DESCRIPTION

     The  major  data  base  collected during  this study  consisted of  daily
indoor and outdoor  PERC  concentrations measured for seven  consecutive  days
at each of six  selected  self-service  laundries.   Additional  data were  col-
lected to  demonstrate PERC levels  in a  residence  above one  of  the  laun-
dries, and in a residence  where  clothing was brought in from being  freshly
dry cleaned in a coin machine.

     Testing at the laundries and at an  overhead apartment  was  conducted  in
the Washington, DC  vicinity  during August and early September,  1980.  The
clothing study  was  performed  in a private  residence near  Cincinnati,  Ohio

                                   153

-------
during early October, 1980.

     All  testing involved  the use  of  standard charcoal  sorption  tubes,
through which  measured  volumes of air  were drawn.   The adsorbed PERC  was
later  desorbed at  analytical laboratories  and  analyzed  by gas  chromato-
graphy.  Volume  of  PERC per volume of  air  was determined for  each  sample,
and all  results  were reported in terms  of parts per  billion (ppb).   The
field  work was performed  by PEDCo Environmental,  Inc., under contract  to
the EPA.   The  majority  of  the analytical work was also  performed  by PEDCo,
with  some analyses  and support  provided  by Research  Triangle  Institute
(RTI) and  TRW  Corporation.
                         EXPERIMENTAL  APPLICATIONS
Technical Approach

Sampling Methods

     Previously  tested methods  of  sampling ambient  air  were used  in this
study  (6,7).   These  methods involved  sampling  of air with  NIOSH-approved
150-mg  charcoal  tubes  for  measured periods  of  time  at  constant  air flow
rates.

     Indoor sampling  in the Laundries  was performed with air flow rates set
at 75 ml/min  for periods of 8 hours.  This allowed  a sufficient air volume
to  be  collected  so that  the adsorbed  PERC  would  be  measurable  by  flame
ionization detection.   This method of sampling was  also  applied to testing
inside  the apartment  above  one  of the  laundries.

     Outdoor  testing  was performed using air  flow rates of  250 ml/min for
periods of 24 hours.   This  allowed a  sufficient air  volume to be collected
so that the adsorbed  PERC in these samples would be detectable.   Since very
low  levels were  present  at  outdoor locations  relative to  inside the  laun-
dries, more sensitive detection methods were  employed.  This method of sam-
pling  also  applied to  testing  in the home into  which freshly  dry cleaned
clothing was  brought.

     In all   cases, sampling systems  were  constructed to  allow the tested
air  to enter directly  into  the sorption  tube.   All valving,   pumps,  and
vacuum lines  were  located downstream from the sample collection point.

Analytical Methods

     All analyses  of  samples were made by desorbing the sample into a suit-
able solvent  (carbon  disulfide  or   carbon  disulfide with  methanol)  and
analyzing the solvent by gas chromatography.   The indoor laundry and apart-
ment analysis instruments  were  equipped with  flame  ionization detection
(FID),  while  the  outdoor  and  clothing  study analysis   instruments  were
equipped with electron  capture  detection (ECD) or  mass  spectroscopy  (MS).
In  this particular application, ECD and MS are much  more  sensitive to low
levels than  the  less  expensive  FID,  making  their  use  mandatory  in this

                                    154

-------
case.  Determinations  of ppb PERG,  using the analytical results and the ,air
volume  field data,  yielded  time-weighted  average  (TWA)  results  for each
appropriate  sampling period.

Field Application

Site Descriptions

     The  six laundries were selected to  be  representative of various types
and sizes  of operations.  All  of  them had  coin-type  dry  cleaning machines
on  the  premises  as  well as  regular  customer-operated washers  and dryers.
The physical layouts, ventilation  methods,  machine operating  methods,  and
number and type  of  dry  cleaning machines  varied from laundry  to laundry.
The apartment above one  of  the laundries was  selected because  it had  one
room  directly  above several  dry   cleaning  machines  and  it was   generally
unoccupied during testing hours.   The  home  that was  used  for  the clothing
test was  chosen because  it was in  a semi-rural  area,  which minimized back-
ground PERC  levels, and because it  had an  unoccupied bedroom,  which  was
used in this  test.   All  outdoor test  sites  in the vicinity of the  laundries
served as  background level  indicators.   Since  these  levels  were  generally
much below the  tested laundry  levels, detailed  descriptions  of these sites
are not  necessary to describe  the  significance  of  the laundry  data.   The
sites were located from 50 to  1000 meters  away from  each  laundry in order
to provide representative background  data  for each location.   Each indoor
location was  assigned  a letter identification for  ease in presentation and
discussion of results.

     Laundry  (A)—This   spacious,  well-ventilated  laundry  contained  24
customer-operated dry  cleaning machines.   All of the  clothing  dropped  off
by customers  for  attendant dry cleaning was  sent  off  premises,  however,  so
machine use  was generally low.  This  laundry had good cross-flow ventila-
tion, which was assisted by  opposing,  open  doors.

     Laundry  (B)—This  small  laundry  had   very  good  ventilation and  two
attendant-operated machines.  Machine  usage  was  high due to the convenience
of the attendants' service.   One door was kept  open during business hours,
and large  fans  provided  constant, fresh air  inflow.

     Laundry  (C)—This  moderately  large laundry  had  fair ventilation  and
eight dry  cleaning machines.  Attendant  service  was  provided for dry clean-
ing, which led  to  regular use of the machines.   One  door was  kept open,  and
some fans  were  used  to assist ventilation.

     Laundry  (D)— This  older,  medium-sized  laundry  had four customer-oper-
ated  dry  cleaning machines  that   were  regularly  used during  the  study.
Ventilation was not  consistent  throughout the laundry because of the rather
complicated  floor plan.   Although  a  customer waiting  area was next  to  an
open door  that  allowed plenty of fresh air,  the  dry cleaning machines were
in a  somewhat  isolated  room with  little  direct ventilation  to  the  out-
doors.

     Laundry  (E)—This medium-large laundry  had  four attendant-operated  dry
                                    155

-------
cleaning machines  that were regularly  Used  during the  test.   This laundry
was the only  air-conditioned one that Was tested,  and  hence can be assumed
to have the poorest  overall ventilation of the laundries tested.

     Laundry   (F)—This  medium-large   laundry  had   eight   customer-  or
attendant-operated dry cleaning machines.   Use of the machines was somewhat
below  normal  during  the test,  since a few scheduled  maintenance repairs
were  being performed.   The floor  plan was unusual  compared to  the  other
laundries, since  the  shop  occupied  a narrow property between other shops in
a large building.  This may have impaired  the circulation of fresh air from
the open  front door  to the back of the  shop where  the dry  cleaning was
located.

     Apartment—This  apartment was  located directly  above  the dry cleaning
machines in Laundty D*   It  was not  occupied during testing, since the resi-
dents, who had  no children, worked during normal  business  hours.   Because
the apartment was  closed when they were working,  ventilation was very poor
during the test.

     Bedroom  in private home—The  bedroom  that  was  tested for  clothing
emissions  was  an unoccupied room in a  semi-rural, one-family  home.   This
room was generally closed off  during the  test arid hence had poor ventila-
tion.

Testing Strategy

     To obtain  data  that would  present  a  clear picture  of  potential  human
exposure caused by the various environmental  factors  tested,  it was neces-
sary  to  design  a comprehensive  sampling  scheme.   Background  PERC levels
needed to be established for each test  to  ensure that the exposure data did
not represent  existing ambient levels.   The tests also had  to  be designed
to provide data generally  representative of  each tested environment, rather
than flukes caused by peculiarities in  the vicinity of sampling points.

     Background tests—Except  for   the  home  clothing  emission test,  all
background data were  obtained  by operating three to four outdoor monitoring
stations in the vicinity of each indoor test.  Each  outdoor station was at
least 50 meters  away  from known potential PERC  sources,  and within 1  kilo-
meter of the indoor  testing site.

     A variety of monitoring   locations were  chosen  to obtain background
data.  Most background sites were at nearby commercial or retail establish-
ments,  and at  least  one  representative  residential  site  per  test  was
chosen.  Background  data for  the home  clothing emission test was obtained
simply by  monitoring the  test room prior  to  the  introduction  of  clothing
into the room.

     Exposure tests—All indoor tests were  designed  to show representative
levels of PERC  that  might be inhaled by people  occupying the room environ-
ment  during  the  monitoring  periods.   In  laundries,  this  required  that
sampling  points be  located in areas  of  high  traffic  and  use,  and  that
specific points of suspected high and low  exposure levels be identified and
                                     156

-------
 tested as well  as  areas  probably representative of the room air as a whole.
 The specific goals in  each laundry were to  test  one  point at least halfway
 from  the  dry  cleaning  machines  to  the  open front door,  if  applicable, one
 point in a customer lounge area, one point  within 2  meters  of dry cleaning
 machines, one point near customer working areas,  and one centrally located
 point.   Since  only three  points in each  laundry were tested,' these goals
 had to  be met   by  combining purposes at  single  sampling  points;  occasion-
 ally,  specific  types of  samples could  not  be  obtained. • In  the apartment,
 testing  points  were  obtained  to  show  levels of  high potential  exposure
 (directly above the machines)  and  "background"  levels  in the  rest  of the
 apartment.   The home  clothing emission test  was similarly  arranged,  with
 one sampling  point  near the  closet  where  the  clothing  was  stored  (high
 exposure level) and the  other  two at points  of high  probable  exposure dura-
 tions:  the desk and the bed.

 Test Schedule

      Testing was scheduled to  collect  consecutive  daily data  at  each  sam-
 pling  site.   Exposure  tests  at laundries  were  scheduled to run for  8 hours
 each  day  during normal  business hours,  while the  home clothing  emission
 test ran continuously,  collecting 24-hour  samples.

     Background outdoor tests  collected 24-hour averaged  daily results,  and
 were scheduled  to  coincide with the nearby  indoor tests.   All tests'were
 run until seven consecutive daily sets  of  samples were  obtained.


                                   RESULTS

 Perchloroethylene Levels  in Laundries

     The  indoor  ambient  PERC  levels  varied considerably from laundry to
 laundry  and  from day to day inside each laundry.  The average  levels  ranged
 from 90  to 14,000  ppb, which  compares  to a background range  of  0.1 to  7
 ppb.

     Table 1 presents a summary of the  high, low,  and 7-day mean PERC con-
 centrations  for  each laundry and for each  network of background monitoring
 sites.   This table shows a  fairly  wide spread  of values between  the six
 different  laundries, with 7-day mean values  ranging  from a low of 130 ppb
 to a high  of 8600 ppb.  Each laundry also  showed  a great  deal of variation
 on a day-to-day  basis,  with a  high  to  low  ratio of about  10:1  for five of
 the six.

     Although most  of  the  information  gathered does  not  explain the daily
variations in  each   laundry,  some of  the differences  between  laundries  can
be understood by reviewing  the physical layout and operations  of each.   In
making these comparisons, it  is helpful  to  separate the laundries into low,
medium,  and  high  PERC  concentration  categories, based   on  the  relative
average levels found in each.
                                    157

-------
              TABLE 1.  SUMMARY OF AMBIENT PERC CONCENTRATIONS
                        INSIDE AND OUTSIDE LAUNDRIES
             	(PPB)
              Indoor PERC concentration  Outdoor PERC concentration    N_
              Lowest   Highest           Lowest   Highest              ~
              daily     daily   7-day    daily     daily   7-day     7-day
              average  average  mean	average  average  mean       I/O*
Laundry A
Laundry B
Laundry C
Laundry D
Laundry E
Laundry F

91
75
250
464
1,900
330

200
700
2,700
3,800
14,000
3,100

130
320
1,300
1,500
8,600
1,300

0.86
0.34
1.2
1.3
0.13
0.27

6.7
3.9
4.5
4.4
5.0
4.0

3.3
1.4
3.6
2.5
1.5
2.0

59/62
28/29
33/30
33/33
62/49
30/35

*I/0  signifies  "N" for 7-day means as  Indoor/Outdoor,  where N is the total
 number  of  measurements used in arriving at mean values.
Low PERC  Levels

     Laundries  A and  B  demonstrated  average  PERC levels  from 100  to 300
ppb,  which is much  lower than  the  other four  laundries.   Laundry  A, the
lowest  of all,  probably  had  the best ventilation  of  the six  due  to  cross
air flow, and the lowest on—premises use of  dry cleaning machines  due to
its practice of  sending dropped-off  clothing out  for dry  cleaning  else-
where.  Laundry B, with  somewhat higher PERC levels than A,  also had excel-
lent  ventilation.   Its  higher  PERC levels  are  explainable by  the regular
use  that  its dry  cleaning  machines  experienced—this  laundry  provided
convenient  operation  of  machines by  a full-time  attendant.

Medium PERC Levels

     Laundries C, D,  and F all  averaged between  1300  and 1500 ppb,  which is
much  higher than Laundries A and B, but still much  lower than Laundry E.
The relatively  high levels found in these  laundries  is probably  due  to  a
combination of factors generally related  to  machine usage and ventilation.

     All  three generally kept their  front doors  open  during  business hours,
but in Laundries D and F, the benefits of this may have been  reduced since
their physical layouts were not  conducive to good air  circulation.   Laundry
C,  which  most  likely had  better  air  circulation,   probably  had  higher
machine usage than D and  F.   Laundry C had attendant-provided  dry  cleaning,
unlike Laundry D  which was solely customer operated.   Finally, Laundry  C
did not experience any scheduled repairs  as  did Laundry F.
                                    158

-------
High PERC Levels

     Laundry  E  showed  consistently  higher  PERC levels  than  any of  the
others that were  tested.   Daily average  levels  ranged from  1900  to 14,000
ppb, with a 7-day mean of  8600 ppb.   Because the operations  were similar to
those  in  Laundries B and  C, with  attendant-provided  services,  operations
alone  cannot  explain  the  high levels  that  were  found.  The  major physical
distinction was that Laundry E was  air conditioned while  none of the others
was.   This may have led  to a concentration  of PERC  caused by recirculation
of  indoor  air without dilution  makeup from the outdoors.   In short,  the
ventilation was very poor.

Perchloroethylene Levels in  the Apartment

     In the apartment  tested above Laundry  D, PERC  levels were practically
identical  to  those  downstairs.    Table  2   shows  a  reasonable  consistency
between the daily averages found in this  apartment and the laundry.
           TABLE 2.  COMPARISON  OF  INDOOR PERC BETWEEN LAUNDRY D
                     AND  THE APARTMENT UPSTAIRS
Location
                                       Daily average,  ppb
Day 1   Day 2   Day 3   Day 4   Day  5     Day  6    Day  7
Dry cleaning
room of laundry
Apartment
1550
1100
920
680
5550
5800
2500
720
645
2300*
1300
620
1450
1200

*No explanation for  this  occurrence  was  recorded in the field notes.  This
 high level may be evidence  of  reentrainment  of  exhaust fumes from the dry
 cleaning machines vented to  the  outdoors.
The  field  notes taken during  the  test indicated an  interesting occurrence
on Day 3 that probably explains  the  high  PERC levels  that were monitored on
this  day.   A PERC  spill  had occurred near  the  machines, and had  not  been
cleaned up  immediately.

Perchloroethylene Levels  in  the  Private Home

     A  definite rise  in  the  indoor  ambient PERC  levels  was  noted  when
freshly  dry cleaned clothing  was  brought  into  the  bedoom and allowed  to
hang in the  closet  undisturbed.   Table 3 shows  a steady background concen-
tration  of 0.2  ppb that  had  been  observed prior  to  the  arrival  of  the
clothing,  after which  an immediate rise  to 100  ppb  was  measured.   The
levels  dropped  quickly the  next day  to  about 20  ppb,  and  then  gradually
declined.  After 1  week of exposure, the  PERC levels  in the room were still
6 ppb, which is distinctly higher  than the  background amounts.
                                     159

-------
     TABLE  3.   SUMMARY OF AMBIENT PERC CONCENTRATIONS INSIDE A CLOSED
                LIVING SPACE WHERE DRY CLEANED CLOTHING WAS INTRODUCED


Test
day
1
2
3
4
5
6
7
8
9


Comments
No dry cleaned clothing*
No dry cleaned clothing*
First day of exposuret
Second day of exposure
Third day of exposure
Fourth day of exposure
Fifth day of exposure
Sixth day of exposure
Seventh day of exposure
Mean PERC
concent rat ion j
ppb
0.21
0.18
102
22
17
25
16
14
6.2

*Values obtained were  used  to  establish baseline PERC concentration.
tDry  cleaned  clothing was  introduced  at the  beginning  of this  24-hour
 period.
                                 CONCLUSIONS

 This study showed  that  the  use  of  coin-type dry cleaning machines can lead
to a definite potential  human exposure to PERC,  and that there are indirect
as well  as  direct routes in which  this  exposure may occur.   The  data also
demonstrated that,  at  least  in terms of  direct  exposure potential in laun-
dries,  ambient  PERC  levels  can differ  enormously,  depending on  various
physical and environmental factors.   The specific findings of the study are
discussed in detail below.

PERC Levels in Laundries

     Significant  levels  of  PERC may be  encountered in laundries  that  use
coin-type dry  cleaning machines.   Ventilation  appears  to be a key  factor
regulating the indoor  concentrations,  as  well as relative machine use.  Air
conditioning, which reduces  ventilation,  can apparently lead to  a concen-
tration of PERC well in  excess of that  found where direct air flow from the
outdoors is present.   This  is especially  important  in evaluating  the expo-
sure potential  for  winter operations,  when  ventilation to the  outdoors  is
reduced.

PERC Levels in Properties Adjoining Laundries

     PERC shows an  apparent  high ability to  permeate into adjoining proper-
ties, at least  insofar as apartments  above  laundries are  concerned.   More
extensive studies of  this nature are  required  to fully  characterize this
potential.   Since  only  one  apartment  was  tested during  this study,  the
possibility   exists   that   the  data   obtained   represent   an   isolated
circumstance, not a generally prevalent problem.  In addition, it should be
emphasized that  this  was in an  older building.   The newer  laundries were

                                    160

-------
 all one-story affairs  in  nonresidential areas and  it  may well  be  that in
 the future,  laundries with overhead dwellings will become phased out.

 PERC Emissions  from Clothing

      PERC is apparently very  persistent in  clothing  once it has  been dry
 cleaned.   The levels measured  during this  study  showed  a day-to-day trend
 that demonstrates  a definite,  long-term emission pattern.  These results do
 have limitations,  however,  since  they  represent data • obtained during  a
 single  test  series.   More  extensive tests  of  this nature  are needed to
 establish generalities  concerning clothing  emissions.
1.
2.
3. .
4.
5.
6.
7.
                            REFERENCES

 Fisher,  W.E.    1977.    International  Fabricare  Institute,  Research
 Division,  Silver Springs,  MD.  Paper presented  at the EPA Hydrocarbon
 Workshop,  Chicago,  IL.   July 20.

 National Cancer Institute.   1977.   Bioassay of tetrachlproethylene for
 possible carcinogenicity.   Publication  No.  77-813.  U.S. Department of
 Health, Education,  and  Welfare, Public  Health Service, National Insti-
 tute  of Health.                           ,

 Blair,  A.    1979.   Causes  of  death among laundry  and dry  cleaning
 workers.   American  Journal  of Public Health 69(5):508-511.

 Sykes,  A.L.,  and  J.E.   Bumgarner.  1979.   Analytical  results of  the
 perchloroethylene study of a coin-operated dry  cleaner.  U.S.   Envi-
 ronmental  Protection Agency Contract  No.  68-02-2688,  Task 13.

 Verberk, M.M., and T.M.L.  Scheffers.    1980.   Tetrachloroethylene  in
 exhaled  air  of residents  -near  dry cleaning shops.    Environmental
 Research 21:432-437.                                      ~	
U.S.  Environmental   Protection  Agency.     1979.
perchloroethylene in  ambient air.  August.
                                                            Measurement   of
U.S.  Department  of  Health,  Education,  and  Welfare.    1977.    NIOSH
manual of analytical methods.   Second edition, Volume 3,  5335.
                                    161

-------
                  PRELIMINARY  RESULTS  FROM THE WIDE  RANGE
                            AEROSOL  CLASSIFIER
            R.M. Burton, Dale A.  Lundgren,  Brian J.  Hausknecht,
                         and David  C.  Rovell-Rixx
                                 ABSTRACT

     The  Wide  Range  Aerosol  Classifier  (WRAC)  was  built  under  a  U.S.
Environmental  Protection  Agency  cooperative  agreement  with Dr.  Dale  A.
Lundgren of  the University of  Florida (Gainesville).   This  sampling system
provides a mechanism for  size-separating and  collecting with high efficien-
cy the full  spectrum of ambient  particulate  sizes  to include  fugitive dusts
in the range from  15 to 200um  aerodynamic  diameter.

     The WRAC  was  designed with  a very  large  inlet  and  a high  sampling
flowrate to  permit collection  of very large particles  with  nearly  100 per-
cent efficiency.   The  system consists  of a series  of single-stage impactors
operating at high  flow rates in parallel,  thus  eliminating  the  problems of
wall loss and  transport that  are  associated  with cascade impaction.   This
sampling  concept  was  demonstrated  to  be  viable,  as  described  in  Air
Pollution Control  Association Journal  24:(12),  December,  1975.   The  sam-
pling  results   are presented  as  mass  concentration distributions  over  a
complete particulate size  range  from 0.04  to  200)jm.

     WRAC field sampling  has been  conducted  at a variety of  sites to obtain
valid  particulate  size  distribution data.    Sites  in  Birmingham,  Alabama
(industrial);  Research Triangle Park, North  Carolina  (background);  Phila-
delphia,  Pennsylvania (metropolitan);  Phoenix,   Arizona  (high  fugitive
dust)j and Los  Angeles,  California (automotive  generated  smog and  fugitive
dust)  have   provided  a  wide  range of  atmospheric  particulate mass  size
distributions for  monitoring.

     Preliminary analysis  of  the  data  shows  the WRAC  to  be   capable  of
determining  the total  atmospheric  particulate  matter mass size  distribution
together with   predicting  and validating  the fractions  collected  by  the
Inhalable Particulate  Network  samplers.
                                    162

-------
                                INTRODUCTION

     The  present  air quality standard for  particulate  matter is based upon
the total amount  of  suspended particulate matter collected by the reference
method, a high volume air sampler.   This sampler has  been  assumed  to col-^
lect particles less than ^lOOjim  diameter (Stokes equivalent).   Tests con-
ducted  by  Wedding  and  co-workers  indicate,  however,  that the  sampling
efficiency  of  the high volume  air sampler may  be  as low as  7  percent for
50ym particles and  18 percent  for 30ym  particles  in moderate  winds.   The
high volume air sampler is  also sensitive to  orientation, showing a 20 per-
cent  drop  in  collection  efficiency  for  particles  15ym  in diameter and
larger  with a 45  degree  shift in  wind  direction.   Because the mass  of  a
particle  increases  as  the  cube  of its  diameter,  the  mass  concentrations
measured by the high volume  air sampler  can also vary widely.

     The U.S.  Environmental Protection Agency (EPA)  in 1979  defined inhal-
able particles (IP)  as those  less than 15ym and is currently considering an
upper size  limit  of lOym  (1).  The EPA has also  considered  establishing  a
fine particle  standard consisting  of  particles  less  than ^2 to  3ym diam-
eter.   Consideration of  an IP standard has generated considerable interest
in defining the total  atmospheric particulate size distribution.  Only then
will it be  possible to  determine  what   fraction  of  the  total  atmospheric
aerosol is  being  collected  and what fraction would be  desirable to  collect
by existing or proposed sampling  devices  for  inhalable  or other particulate
measurements.

     Lundgren  and Paulus  previously described a  stationary  sampling system
that effectively  sampled  large atmospheric  particles (up to  ^lOOym);  they
determined  the  large particle size distribution  and  total atmospheric mass
concentration,  then compared  their results with  collections by  dust  fall
plates  and  with a standard high  volume air  sampler  (2).   That  study pro-
vided  the background  for  the  present  project   to  design,   construct,  and
field-test a mobile  large particle  sampler that  determines the mass  distri-
bution  of large (10  to 200ym) atmospheric particles.   With this  sampler, it
is possible to characterize  the total  particulate mass  size distribution of
ambient aerosol and  to compare the results with  particulate  mass  data col-
lected  simultaneously  by  the Total Suspended Particulate  (TSP)  Hi-Vol, the
Size Selective Inlet Hi-Vol  (SSI), Small Particle  Cascade  Impactors,  and
the Dichotomous  Samplers.   The sampling  system will be  especially useful
for evaluating areas with  high fugitive  particulate concentration.   Data
showing  the  relationship  between  total  suspended  particulate,  inhaled
particulate, fine  particulate,  and  other  measures of  particle concentration
can be  obtained  while the total  atmospheric  particle mass  distribution is
being measured.   Differences in  the  quantity of particulate  mass measured
by  the  various   devices  can  then  be rationally explained  and properly
related to the aerosol size distribution  in the  atmosphere.

     The  objective of this  project was  to  design and  construct a mobile
particle sampler  capable  of  collecting size-separated mass  samples  provid-
ing mass  distribution of  the  total atmospheric  particle  size  range  up  to
200ym aerodynamic diameter,  and  to  use  the instrument  to  perform  field
sampling for defining  the size  distribution of  the total particulate matter

                                    163

-------
mass.
                                  PROCEDURE

     The  sampler,  as shown in Figure  1,  is  fitted into  a  trailer and con-
sists  of  a large,  high-flow-rate (1380  CFM)  inlet from which  five isoki-
netic  samples  are  withdrawn.    Four  of the  samples  are passed  through
single-stage  impactors  with different  cutpoints,  while  the fifth is passed
through a total particulate matter filter.  The four impactors are designed
to  collect particles  greater than  9.6ym, 18ym,  34ym,  and 57ym diameter,
respectively.    Smaller  aerosol  particles  are  sized  by attaching  cascade
impactors  following the  single-stage impactors.

     An  accompanying analysis  lab  is  set up  in a mobile van.   Analysis
equipment  includes  a precision  balance, optical  sizing microscope,  and a
sample equilibration chamber.
                                INLET DESIGN

     The major  task in measuring large atmospheric  particles  is  transport-
ing  a  true or  representative sample  of  the  particles  into  a  measurement
device.   This  difficulty  is particularly  serious  for  the  collection  of
particles  larger than  'vSOym diameter  because of their  great inertia and
high settling rate.

     Particle settling  velocity  (Vs) is  determined by a balance  between the
force of gravity acting on a particle and  the fluid drag  force  exerted  by
the medium through  which  the  particle falls.   Simply put,  settling velocity
increases rapidly with  increasing  particle  size.   Sampling tube  inlets that
are  pointed  upward  will  capture  a greater  proportion  of  large  particles
than are in the  actual  distribution due to the settling of large particles
into the  inlet.  The  increase  in  the relative  number  of  large  particles
sampled is equal to 1  +  [Vs/Vo] ,  where Vo equals the sampling  tube  inlet
velocity.  The  error  can be kept  reasonably small  if  the inlet  velocity
(Vo) is made  several times greater  than  the  settling velocity (Vs)  of the
larger  particle desired.   Therefore,  a  Vo  of  25  times  Vs  would give  a
permissible  error  of   only  4 percent  overestimation  for  a  tube  pointed
upward.   The  Wide  Range Aerosol  Classifier meets  this   criterion  for  a
55ym-diameter  particle,  and  an  11  percent   overestimation  for  a  lOOym-
diameter particle is predicted.

     Particle inertia is  a function of particle mass.  From Newton's  first
law, where' force equals mass  times  acceleration,  a larger  force  is  required
to accelerate (or decelerate) a larger  (heavier)  particle as quickly  as  a
smaller (lighter) particle.   Because  the size of a particle also  affects
the drag it experiences from  the air,  the relaxation time  (T)  is  used as  an
indication of its ability to  accelerate or  decelerate.  Particle  relaxation
time (T ) has  been defined as:  T  = Dp2p /18n, where Dp  is the diameter  of
the particle, p  is  its  density, n  is the  viscosity of air,  and T has  units
of time.
                                    164-

-------
               1  Raincap
               2  Wind Shroud
               3  Main Inlet
               4  Shroud Support Pole
               5  Air Plenum
6  WRAC Sampler
7  Flow Controller
8  Flow Recorder
9  Auxiliary Blower
10  Coarse Filter
Figure  1.   Schematic diagram of the mobile  sampling  system.

                                      165

-------
      It  is  also convenient to define  particle  stopping distance, &, as the
distance a particle will  travel when  decelerated  in a  fluid  medium  (air)
from  an  initial velocity (V)  to  rest.   It hs been shown that £  = VT.

      Davies  uses 5, to evaluate  the effect of  particle  inertia on sampling
efficiency  for  two  situations  (3).   When sampling  at  a known  inlet  flow
rate, Q,  a  suction velocity (Vx) will  be  produced;  Vx  is  a function of the
distance from the  inlet  orifice.  To examine the case where Vx is evaluated
at a  distance i  from the center  of the orifice, Davies uses the equation Vx
-  Q/4ir£,2, where Q  is  the flow  rate  of the  sampler inlet and b^H2  is the
surface  area of a  sphere of radius H.

     More appropriate might  be  an equation determined  by Dalle  Valle  from
measured velocity  contours of exhaust  hoods.   For  a point at  a  distance £
along the center line  from the hood face,  Vx = Q/10£2+A, where A equals the
inlet face  area.   Davies  reasons  that if the  radius  of the  inlet  (R)  is
several  times larger than £,  the effects of inertia will be negligible.,

     Davies  also  has  examined  the  situation  of  sampling in  a  cross-wind
with  some velocity  (Vs)  and reasons  that the  inlet  radius  (R)  should  be
much  greater than  the  stopping distance associated  with the wind (i.e., a =
Vw T).    He  suggests that  R  should be  at least 5£.   For the mobile  WRAC
sampler,  the condition that R =  5H requires that £  be less than 6 cm.  At a
distance  of  6  cm  from the inlet  orifice, the  maximum velocity  using the
latter equation would be  2 m/s  (^4.5 mph).    The  stopping distance for  a
lOOpm-diameter particle at this velocity  is  6.15   cm.   This  nearly meets
Davies'  criteria and suggests that  the mobile  sampler  inlet  is  dimension-
ally  adequate to efficiently  sample lOOym-diameter  particles  in winds up to
^2 m/s (4.5  mph).

     Several researchers  have noted that  Davies' theoretical  criteria  seem
overly  restrictive  for  efficient  sampling  and  are  not met by  several
commercially  available  sampling  instruments.    A  theoretical  study  of
sampling  efficiencies  by  Agarwal  and  Liu took into account  both particle
inertia  and  settling (4).   They  determined the flow field around  a verti-
cal,  thin-walled inlet from the  Navier-Stokes equations and then calculated
particle  trajectories.   They  determined a critical  particle trajectory that
is a  distance,  Re,  from  the inlet  axis, such  that the particles inside  this
radius enter the  inlet.   Sampling  efficiency  was   calculated  by comparing
the concentration  of particles entering the inlet versus the  actual concen-
tration  of  the air sampled.

     Agarwal and  Liu  noted  that  the  sampling efficiency  was a  function
independent  of  both the  Stokes  number  (STK = p CDp2Vo/9r)D)  and  the relative
settling  velocity  (Vs'  =  Vs/Vo).   The characteristic  length  and  velocity
are in the inlet diameters, D,  and the inlet velocity,  Vo.   Their research
indicates that  if  the  product (Stk) (Vs?) is less  than 0.10,  then sampling
efficiency  will be  greater  than 90  percent.    Agarwal  and  Liu  caution,
however,  that their  criterion  for  efficiency  regards  all particles  that
enter the inlet  as sampled, whether or not they impact and  possibly stick
on the inside wall  of  the inlet.   The value  of  (Stk) (Vs1)  for  sampling
lOOym particles with  the  mobile sampler  is  0.025.   Solving  for  particle

                                    166

-------
diameter with  (Stk)  (Vs1) equal  to  0.10 indicates that  the  mobile sampler
can sample particles  as large as 130ym  in  diameter with an  accuracy of 90
percent.  These values  assume,  however,  that Stokes Law holds for particles
this size.

     The shroud of the  WRAC was  designed to provide a calm air space around
the inlet  orifice so  that  the  sampler  would be  less sensitive  to cross-
winds.  To be  effective, it was  made  large enough so  that  particles would
not impact  on it  but  would  flow around and  over the  shroud.    To ensure
this,  it  was  designed  large  in comparison  to £.   Using the  same  ratio of
1:5, this  criterion  was  met  for a particle  with a  stopping  distance  less
than or  equal to  one—fifth the  diameter  of  the  shroud, or  30 cm.    This
condition was met  for a lOOym-diameter particle  in a 9 m/s (20 mph) wind.

     A rain shield designed for use when the  sampler is  operating consists
of  a  90 cm  flat  disk  supported  30 cm  above  the  top  of  the  shroud and
centered above the inlet.  It prevents  rain from  falling directly into the
inlet  under  light wind  conditions.   A wind-driven  rain should  fall  at an
angle  and  impact  on the  side of the  tube.   From there  it  should run  down
the side and .into  the air'plenum box and not into the samplers.
                          SELECTIVE  SAMPLER DESIGN

     The  size-selective sampler inlets  were designed to  extract  an isoki-
netic  sample  from the  air flowing  through the  inlet  tube.   The samplers
were constructed  of  0.10  mm aluminum.   The sampler inlets are of equal area
so  that  an equal volume  of  air  is sampled  through  each  sampler.   Each
impactor  inlet  measures 17.8 cm x 6.55  cm.   The  inlet  area was determined
by the slot width required for the  first  impactor;  therefore, the Number 1
impactor  is a straight  nozzle  (neither  converging or diverging).

     To  minimize  losses   and  facilitate  construction,  a  rectangular  jet
design was used.   The impactors  were designed with the jets 17.8 cm long in
order  to  leave  more room for  the airflow  over  the impactor  plate  to turn
down into the filter.   This reduces  the particle losses on  the walls of the
impactor.

     The  fifth  sampler, which  collects  a total particle sample, was design-
ed with  a square inlet  so  that  it would  fit  better between the impactor
inlets.   The  sides of  all  sampler  inlets  are straight,  to  reduce any loss
of particles  onto the  walls.   The  inlet  area  of the fifth sampler  is the
same as for the other samplers.

     An  additional advantage  of  using  high volume air  sampler blowers is
that  flow controllers  are readily  available on the market.   These flow
controllers ensure a constant  flow rate  through  the  sampler  so  that the
total air volume  for the  sampling period is known and that  a  representative
sample is drawn for  each  hour over  the normal  24-hour  period.   Therefore,
if the  total  sampling  time is known  and  the flow rate  is  known and con-
stant, the total  air volume sampled is  known.   Aerosol mass  concentrations
can  then  be  computed.   A constant  flow rate is  additionally critical for
                                     167

-------
impactors  because  the aerosol  impaction efficiency  is a  function of  the
flow rate.  Field  sampling  and  evaluation at selected sites has  shown  that
the sampler  can provide  reliable  information on  the  size distribution  of
particle mass,  including particles up  to 200ym  aerodynamic diameter.   The
system  is  self-sufficient,  with  the  accompanying mobile  laboratory,   for
processing and analyzing  the collected  samples.
                      FIELD SAMPLING SITE DESCRIPTION

     WRAC  field sampling was  conducted in  five  locations  throughout  the
country, each  location chosen to represent  a distinctively different  type
of ambient aerosol loading.  The five  locations  are  described  below.

          1.  Birmingham, Alabama (industrial)
          2.  Research Triangle Park,  North  Carolina (background)
          3.  Philadelphia, Pennsylvania (metropolitan)
          4.  Phoenix, Arizona (high fugitive dust)
          5.  Los Angeles (Riverside), California  (automotive-generated
              smog and fugitive dust)

     Each site  was selected at an existing EPA  Inhaled Particulate Network
site in order that Total Suspended Particulate Hi-Vol,  Size  Selection Inlet
Inhalable  Particulate Hi-Vol, and  Dichotomous  Inhalable  Particulate  mea-
surements could be made  simultaneously and concurrently with  the WRAC mea-
surements.   Sample duration  time  was  24 hours  for sampling  at all  loca-
tions..   Since  the WRAC  gives particle  mass by  size distribution  up  to
200pm, cut—point characteristics of the conventional particle  samplers used
in the study could be  evaluated from  the  WRAC data.  Also,  a  better  under-
standing of the total and inhalable particulate mass particle  size distri-
butions around  the country would be acquired from  the WRAC data.

     The  analysis  van  containing  an environmentally  controlled   filter
equilibration chamber and a precision  balance for  filter weighings accompa-
nied the WRAC to each  of  the  five  sampling sites  for sample processing and
weighing.
                         RESULTS OF FIELD  SAMPLING

     Data  from  each sampling site are  reported separately for  comparison.
At each site, several  similar runs were selected and averaged  to  determine
a representative  average  distribution for site-to-site comparison.   Selec-
tion for  averaging was based on similarities  in total particulate  concen-
tration and  distribution.   Based  on  these two  parameters,  data  from  Bir-
mingham and Riverside  strongly  suggest  that two  distinct  distributions  were
present,  depending on the local  wind  and  weather  conditions.   On  this
basis, a  total  of seven data sets were used for comparison  purposes:   one
set  each  for Research Triangle Park,  Philadelphia, and  Phoenix; and  two
sets each for Birmingham and Riverside.

     In the  following  data reporting, each set  is  referred  to  by the  site

                                    168

-------
204.6 yg/m3
113.0 yg/m3
48.4 yg/m3
100.1 yg/m3
111.6 yg/m3
255.1 yg/m3
89.5 yg/m3
NB-High
NB-Low
RTP
PA
PHX
RBX-High
RBX-Low
 acronym.   Where two data sets were used,  the  descriptive  suffix "-High" and
 "-Low" is  used  after  the  acronym.    The  following  is  a  list  of  sites,
 average total concentrations and appropriate  acronyms:

                       Birmingham:
                       Birmingham:
           Research Triangle Park:
                     Philadelphia:
                          Phoenix:
                        Riverside:
                        Riverside:

      Calculations were made  for each data  set  to determine the  cumulative
 particle  size distribution.  The  points  were plotted on  logarithmic  normal
 graph paper as % < Dp  versus Dp (impactor  cutpoint).   The results for  the
 seven data  sets  are  shown  in  Figure 2.   A set  of points  generating  a
 straight  line  on a  cumulative  probability  graph  describes  a  log-normal
 distribution.   The  shape of a  best-fit   curve  through  each  set  of  points
 indicates  that the  distributions are not   log-normal but bimodal in  shape.

      Each  data set  is also  illustrated as a histogram in Figures 3 and 4.
 The  partial concentrations  are weighted  for the particle  interval  width.
 Because of  the weighting  procedure, the area under the histogram  represents
 the  total  mass of  the aerosol;  the  percentage  of  the  area  under  a  given
 size  interval  segment  is  equal  to the percentage of mass  in that  given size
 interval.

      A comparison  of  the  shapes   of  the  different  distributions  is made
 possible by  normalizing the concentration for each data set.  For each size
 interval,  the  partial  concentration  is   divided (weighted)  by  the   total
 concentrations  before  being divided by  the  particle  size  interval   width
 term.   This allows each  interval value  to  be represented  as  a  percentage
 (decimal value) of  the total concentration.  Superimposing  the  histograms
 of the different  data  sets  allows  direct  comparison of the shapes,  indepen-
 dent  of the  total concentrations.   These  histograms  are  shown in Figures 5
 and 6.

      The  use  of  small  particle   «10ym)  impactors  in  the WRAC  further
 describes the  particle  size distribution  below  10pm.   Data  from  a Univer-
 sity  of Washington  impactor is  used to  illustrate the average distributions
 obtained in  Phoenix and Riverside.  These distributions  are illustrated in
 smooth,curve plots  in Figure 7.   The  curves were generated from the normal-
 ized  histograms for  the respective WRAC and Washington  impactor data sets.
 This  plot is important  because  it  shows the distinct small particle mode of
 the RBX-High data below lOym, which cannot  be determined from  the  plot in
 Figure 4.

     Data for  the collected IP  Network  samplers were averaged  together to
 correspond to  the WRAC  run  groupings.   Where  several  data points  (measure-
ments) were  not  obtained,  the  average  was  omitted.    To readily  compare
 results site-to-site,  the sampler data are  reported as  ratios  to the WRAC
 total aerosol concentration.  These ratios are presented  in  Table 1.   Also
                                     169

-------
                         1—i—i—i—i   r
                                            CM
                                            o
                                            o
         Cd
         O
                                             o
                                             Ifl
                                               3
                                               Q.
                                               Q
                                               DC
                                               111
                                               LLJ
                                               O
                                             U)
  I  I  I -I
  CB  O>  pj OJ

  cq  q  ^ ^
     §co  co Q
     ^—    co
«- 01  CO
XXX
X m  m
a. cc  cc
n M  ii
  b6n<3OOO
                                                   2
                                                   CO
                                                   •H
                                                   ^3
                                       CO
                                       CO
4J

A


I
O
0> CO  IO
 • ^^   •
G> O)  O>
O> O)  O)
                                        CD
                                        t-J
                                        3
                        da> SSVI/M %
                         170

-------
                                                      o

                                                      DC
                                                      Q

                                                      HI
                                                      _4
                                                      o
(Euj/Brf) dQ eon VA/JMV
                                                              60
                                                             •H
          171

-------

                                    (O   r-

                                    r^   1O
                                    r-   ta
                                         X    X
                                         m    m
                                         cc    oc
                                                                                                        G
                                                                                                                                o
                                                                                                                                §
                                                                                                                                8
                                                                                                                                §     i
                                                                                                                                      £
                                                                                                                                      <
                                                                                                                                      Q
                                                                                                                                      LU
                                                                                                                                      fe
S
                                                                                                                                                cu

                                                                                                                                                •H
                                                                          S
                                                                          tfl
                                                                          H
                                                                          60
                                                                          O
                                                                          JJ
                                                                          CO
                                                                          •H
                                                                                                                                                •0
                                                                                                                                                 O
                                                                                                                                                •H
                                                                                                                                                4->
                                                                                                                                                 CO
                                                                                                                                                 CO
                                                                                                                                                 cfl
                                                                                                                                                c_>
                                                     8
            S     g
o
t
o
CO
euj/6;/)
                                                                                       VA/IMV
                                                                                                                                                 60
                                                                                                                                                •H
                                                                                  172

-------
                      it
                          f
I
CO
§
I
in
in
CM

X
      X

      cc.
                                       8
                                         cr
                                         LU
                                     Q


                                   2  I
                    g
                  dd 60-] v I/\IZ/1/MV
                                             •H

                                             CD ,
                                             •a
                                             •H
                                             e
                                             O
                                             M-l
                                             O
                                             •H
                                             4J
CO

3

CO
CO
                                             CU

                                             N

                                             •H
                                             t-i
                                             O
                                             25
                                             oo
                                             •H
                    173

-------
                                                                                            -I
                                                                                                            pi
                                                                                                            a)
                                                                                                            o
                                                                                                            ,£
                                                                                                            PL,

        s
I   <   *
CC   Q.
                 £
-     D<3O
                                                                                                            oo
                                                                                                            d
                                                                                                            ctf
                                                                                                    -3     H
                                                                                                    a.
                                                                                                    Q     &

                                                                                                    «=      M
                                                                                                    £      Id
                                                                                                    w      cu
                                                                                                    •3      CO
                                                                                                    ^      
-------
                                                     1

                                                     o.
                                                     O

                                                     tr
                                                     o

                                                     13
                                                     o
                                                                           •i-l

                                                                            CO
                                                                            Vl
                                                                            (3
                                                                            O)
                                                                            O
                                                                           ,£5
                                                                           PH
                                                              CO
                                                              (U

                                                              M

                                                              3

                                                              a
                                                              o
                                                             •H
                                                             4-1
                                                              J-l
                                                              4J
                                                              CO
                                                              CO

                                                              CO
(em/erf)
   VA/IMV
 cu
 f-J
 3
 W)
•H
175

-------
          TABLE 1.   SAMPLER VS. WRAC CONCENTRATION EXPRESSED AS RATIO







Sampler concentration/" C"

Site
NB-High
NB-Low
RTF
PA
PHX
RBX-High
RBX-Low
Total*
cone "C"
204.6
113.0
48.4
100.1
111.6
255.1
89.5

TSP
0.84
0.95
—
1.08
0.87
0.92
0.95
SSI
<15wm
0.50
0.69
—
0.72
0.62
0.94
0.74
Dicot
<15um
0.57
0.66
0.81
0.71
0.55
—
—
Dicot(F)
<2.5um
0.13
0.26
0.46
0.35
0.12
—
— •
Dicot(C)
<2.5um
0.44
0.40
0.35
0.36
0.43
—
—


Site
NB-High
NB-Low
RTP
PA
PHX
RBX-High
RBX-Low

Total*
cone "C"
204.6
113.0
48.4
100.1
111.6
255.1
89.5


>57ym
0.10
0.07
0.02
0.06
0.07
0.01
0.02
WRAC concentration

>34jim
0.17
0.11
0.05
0.11
0.12
0.02
0.05
> Dp/"C"

>18um
0.34
0.26
0.15
0.25
0.28
0.06
0.20


>9.6um
0.45
0.37
0.24
0.33
0.41
0.14
0.31

*Total WRAC concentration expressed as
presented,  for  comparison purposes,  are  ratios of the  particle  concentra-
tion  collected  on  a  given  WRAC  impactor  to  the  total  concentration
measured.
                     DISCUSSION  OF MASS DISTRIBUTIONS

     Particle  mass distributions  obtained  from all  of  the  sampling  sites
exhibit a  biomodal shape.   The  position and magnitude of  the  modes  varies
from one location  to another  and even varies at different  times in the same
location.  The cumulative  probability distributions  in Figure  2 are  curves
that indicate  multimodal distributions.  The histograms in Figures 3  and 4
clearly  show the  various  positions of  the  large particle  modes,  together
with their magnitudes.   The distribution curves  in Figure 7 show  that  the
small particle mode also may  vary in position and magnitude.

     The great  difference  in  the RBX-High and RBX-Low small particle  modes
is  easily  explained  by  examining  the  respective  sampling   conditions.
During  the  days  averaged   for  the  RBX-Low  distributions,  the winds  were
predominantly  from the east,  or  off of  the Santa  Anna Desert.  The  particu-
late mass was  primarily  composed of large particles  entrained  in the  wind.

                                     176

-------
The  small particle mass mode was  small because most  of  the mobile  sources
and  other condensation aerosol  sources  are located to  the west, and  their
contributions  were not carried in by the wind.

     During  the  sampling  runs  used for RBX-High averaging,  the  wind  was
primarily from the west,  out qf the Los Angeles Basin.  The wind velocities
were,  on the average,  lower than winds from  the  east.  The  air was  heavy
with small-diameter particulate matter due to the abundance of  condensation
aerosol   (including  photochemical  aerosol) being blown  in  from  the  Los
Angeles  basin  area.   This  boosted the spall particle .concentration by  about
a  factor of  eight.  An interesting observation is that  the large particle
mode appeared  to  increase  in  magnitude  with a downward  shift  in mean
particle size.    The  Los Angeles  basin  is  a moderate  source  of   larger
particles, due to the many dispersion sources such as the extensive  roadway
system.   However, with lower wind speeds,  the air mass  that  starts in  Los
Angeles  takes  a considerable amount  of  time to travel  to  Riverside.  This
may  have allowed  most of the  larger particles to  be removed  (or  settled
out)  from the  air mass  and produce  a downward  shift  in the  mean  large
particle size.

     In  contrast  to  this,  the  observed distributions  in  Birmingham were
characterized  by  a fairly  uniform  increase  in  the  whole  size spectrum.
This condition was usually  the result  of  air stagnation  in  the immediate
area,  resulting  in  an overall  increase  in concentration  of all  sizes  of
particles.

     A  final  comment  on   the  actual distributions  refers   to  the Research
Triangle Park  site.   The  overall low concentrations observed were generally
expected because  of  the lack of significant anthropogenic pollution  sources
in  the  area.   It is  interesting  to note  that the largest  size fractions
show magnitudes  similar   to  those  found  in Riverside.    This  leads  to a
suggestion  that  these  particles •,  which included  things like  insect  parts
and  pollen,  are  the  result  of  natural  generation and exist  in relatively
clean areas  around the country.

     By  normalizing the mass distributions with  respect  to concentration,
the  shapes of  the distributions can be  compared as in Figures 5 and 6.   The
figures  indicate  that, while  the general  shape  of  the distributions  are
similar,   the   individual   interval  values   can vary  greatly.     Each
distribution measured is  the result  of  a unique,  set  of  meteorological  and
geographical conditions that influence  the presence  and residence  time  of
each size range of particles.   There is no simple pattern  that all  of  the
distributions  follow.   One"  cannot  simply  have a  generalized distribution
with a  correction factor  to apply  for  actual  concentration.   Indeed,  the
distribution shapes  can be  quite  different when  the  concentrations  in  two
areas are similar.   The  result is  that ambient  particulate  distributions
must be  considered as  a function of location,  time, etc.

     When the  concentration  obtained by the WRAC is compared to the concen-
trations  obtained by  other  conventional samplers, as  in Table  1,  several
facts are revealed.   The Hi-Vol sampler normally collects  between  85 per-
cent and 95  percent  of the  total  suspended aerosol mass  present, depending

                                    177

-------
on  the location.   The amount  of  material collected  by the  >  57ym  WRAC
sampler can  be  1 percent  to 10 percent of  the  total mass.   However,  these
differences  are  not  a  function  of  concentration,  but  a   result  of  the
distribution  characteristics.    Aerosol  distributions   that  exhibit a  low
concentration  in  the  large  particle  mode  give  much  closer  correlation
between the Hi-Vol and WRAC  total  aerosol measurement.
  COMPARISON OF MEASURED VS WRAC MODELED  TSP  HI-VOL,  SIZE SELECTIVE INLET
                    HI-VOL AND  DICHOTOMOUS MEASUREMENTS

     Measurements  by  the  conventional  TSP  Hi-Vol,   Size  Selective  Inlet
Hi-Vol,  and Dichotomous samplers  were made  simultaneously with  each  WRAC
measurement.   The three  conventional  sampler  types have  been  recently
characterized  for particle  size  separation  and  collection  efficiency
definition  under EMSL extramural programs.  With  the  particle  size cut-off
characteristics of the conventional samplers applied  to  the  total particle
mass distribution  defined by WRAC measurements, it is possible  to compute
mass  loadings  that  each  of  the  three  conventional samplers  should  be
collecting.  The results  of  the modeled predictions are  compared to actual
measurements  by the  TSP  Hi-Vol,   SSI  Hi-Vol,  and Dichotomous  samplers,  as
shown in Figures 8, 9,  and 10,  respectively.   Correlations are good for all
three  sampler  types.   A  statistical  analysis  was   performed,  with  the
results shown in Table  2.  No significant  difference  at the 95 percent C.L.
was  found between TSP Hi-Vol  modeled and  measured   samples  for  all  five
cities  sampled.    SSI  Hi-Vol  measured  vs  modeled  showed no  significant
difference  at the  95  percent  C.L.  for all measurements in  four of the five
cities  sampled.    Birmingham SSI  Hi-Vol measurements  were somewhat  lower
than  predictions.    Dichotomous  sampler measured  values  were  lower  than
predicted in each  of  the five cities.


                                CONCLUSIONS

     System testing,  along with field sampling and evaluation,  have  shown
that the Wide Range Aerosol  Classifier is  a  viable instrument  for collect-
ing  size-separated particulate mass samples to  include  the  total ambient
aerosol size spectrum.  WRAC field  data reveal  that  the  amount of particu-
late matter present  in the ambient  air greater  than.  lOym diameter can vary
by as  much as a  factor of 3,  with a maximum percentage  obtained  in this
study of 45.

     Ambient  aerosol distributions  tend  to display bimodal shapes  under a
variety of  situations such as  sampled with the  WRAC.   In reference to time
and  location,  implementation of a standard will  affect  some  areas  of the
country more radically than  others, due to the  large  variation in the size
distribution of the  ambient  aerosol.

     By use of WRAC particle mass  distribution  data,  precise  estimates for
the  TSP Hi-Volume  and  Size  Selective  Inlet Hi-Volume  samples of  a  given
total ambient aerosol mass can  be  made.
                                    178

-------
<. t "t 	 1 	 1 	 « 	 1 	 1 	 1— I 	 1 	 1 	 1 	 T 	 1 	 1 	 1 	
A
\
i
'\
. \
\.


.

\
A
\"
\
\
*\
\
* *
j|f *!*
' \ "
* \ *
*^^
&
\ ^
\Jfi
0) %^
O5 A 4.A
So *C
in °° .V«
^- ,v °5 II *=^
II S •* E jft,
» 0> ,, — *\A
^ ° *• s *.
' /? " S" 'o ^ "
QJ O »P 9lt
2 | S3 'S \
co 2 »± o j
Q co £ o *.
*v '
\
tltllllll'llllllv
f r r iifrfififiir

u o
*


CO
E
\
.8 §
CO
(O
g
cc
h-
LLI
O
0
o
o W
1 O v/J
tN <;
^

^~
Q

0-
5-
O
o
•8 o
*~ ^
QC
, in >
r-
> o

. in
tN

• o
o o o omo^x-^
p o o or-uj<">'c'

(ELU/Bf/) 'SNOI1VU1N30NOD SSVI/M 1OAIH QaHdlAIVS














•
co > p^
i .
/^ •
^ *
Cs S
N^ g

o> PM
I**"
O XI
o a.
T-H
CO 0)
cO
TJ r^
O *r*t
CX 0-1
a
W P-l
S
"o
a e"
O ctf
CO J3
•H 60
S-l Pi
cl 's
O -H
CJ PQ
•
OO
co ,
3
60
•H
179

-------
-i—i—i—i—i—i—i—»—i—i—i—i—i—i—i—h
                                                        4-8
                                                           CO
       O o
   f»   CO
   •* oo n "

   " CD 7 £
   ftp
   Q CO .£ O
    I—I—I—K—I—I—1—{—I—I—I—I—I—I—I-
                                                           O
                                                           o
                                                           CN
                                                           o
                                                           in
E

1
co
z
o
DC


LU
O
z
o
o
CO
10
                                                                  CO
                                                                  CO

                                                                  Q
                                                                  LLJ


                                                                  Q.

                                                                  S
                                                                  O
                                                                  o
                                                                              co  X
                                                                              co  d
                                                                              Cti  O
                                                                              a ^
                                                                                •H

                                                                              M  3
              8
                            °
                                              §to  O  in
            S S
               O
            13 ,d
            CU P-,
            4-1
            3  **
            &, CB
            S -rj
            O rC!
            O P!


            CO 0)
            ^> ^

               cfl
            CO P-l

            m PS!
            o
                **
            c  e
            o  n)
            co ^c!
            •H  60
            H  C3
            cd -H
            &• S
            0  !-<
            O -H
                                                        _
                                              r~-  in  es ' o
             'SNOI1VH1NE30NOD SSVI/M dhlSS a31dl/\IVS
            2
                                180

-------
I—t    <    <—I—I—I-H—I—I—h-1—I
Data Points = 46
Slope = 1.0308
Intercept = -20.3028
Coefficient (R) = 0.9547
i 	 r— i 	 r-
o
o
T — i — T-
o
8
—i — i — r— 1 — r-
8
CM
— i — r-
8
— r-
in
—T~
o
IO
— r-
u>
CM
—r
o
                                                                     CO



                                                                     I

                                                                     CO
                                                                     <
                                                                     cc.
                                                                     LU
                                                                     O
                                                                     z
                                                                     o
                                                                     o
                                                                     CO
                                                                     CO
                                                                     o
                                                                     o
                                                                     Q
                                                                     Q
                                                                     LU

                                                                     a.
                                                                     S
                                                                     O
                                                                     O
                                                                     o
                                                                     <
                                                                     cr
f(£UJ/6r/) 'SNOI1VU1N30NOO SSVIAI
                                            a31dl/M\zLS
                                                                                       en
                                                                                       co   •
                                                                                       cd  X

                                                                                       B  §
                                                                                       PL, T3
                                                                                       M 'H
                                                                                       I J*
                                                                                       H  S

                                                                                       8*
                                                                                       M T3
                                                                                       Q  C
                                                                                      ^— ' 0)
                                                                                          o
                                                                                      rfl J3
                                                                                       0) CM
                                                                                       4J
                                                                                       3   •>
                                                                                       PU ctf
                                                                                       S -H
                                                                                       O ^
                                                                                       cj  a.
                                                                                         iH
                                                                                       CO  0)
                                                                                       > T3
                                                                                          cfl
                                                                                      *O rH
                                                                                       0) -rl
                                                                                   PM
                                                                                 c  s
                                                                                 O  ctf
                                                                                 CO rG
                                                                                •H  60
                                                                                 a.  s
                                                                                 a  n
                                                                                 O iH
                                                                                CJ PQ
                                                                                o
                                                                                I—I

                                                                                
-------





















o
•3
H
£]
O
CO
CO
is
o

o
o
S
o
i

j^j
^

d
iyi
3
pzf
rn

|J
0
^j
o
CO
M
p2
— 1 CO
•— i cr» c^ f^i
rH t— 1







oo cri r-.
«d" OO CT* VO
"-I OO OO OO
i-H



CO
i-l rH 3
O O O JJ
M > CO > CO B CU
CU 1 CU 1 CU O rH
t— 1 T-l TJ -H -H T-l 4-1 P.

> w T-I T-I -H ^3 aj
1 H O MO MO OCO
•H C/3 C/3 T-l
(B in co m co -* o











































































182

-------
                            REFERENCES
1.
2.
3.
4.
Miller,  F.J.   1979.   Size  considerations for establishing  a standard
for inhalable particles.  APCA Journal 29(6).

Lundgren,  D.A. ,  and  H.J.  Paulus.   1975.   The  mass distribution  of
large atmospheric particles.   APCA Journal  25:1227-1231.

Davies,  C.N.   1968.   The entry  of aerosols  into  sampling  tubes  and
heads.  Brit. J. Appl. Physics  2(1) ;921-932.

Agarwal,  J.K. ,   and B.Y.H.  Lui.    1980.   A  criterion  for  accurate
aerosol sampling in calm air.   Am.  Ind.  Hyg. Assoc.  J. 41:191-197.
                              183

-------
       DETECTION OF GRAPHITIC CARBON  IN  COLLECTED  PARTICULATE  MATTER
                               W.A. McClenny

                   U.S. Environmental Protection Agency
                        Research Triangle Park, NC
                                 ABSTRACT

     Concentration estimates  of  ambient light-absorbing carbon  in  particu-
late matter collected on Teflon  filter  substrates  have  been made,  using the
optical, nondestructive  techniques  of differential transmission and  photo-
acoustics.  Based  on a comparison with similar measurements  on  laboratory-
generated samples, analysis  of ambient  samples indicates  that  differential
transmission  measurements  overestimate  absorbing carbon  by an  increasing
amount  as  the  total sample  mass  increases.    Photoacoustic measurements
appear  to  be  more accurate,  but are subject  to  the  influence  of  thermal
properties of the  sample and  substrate.
                                INTRODUCTION

     This  research attempted  to  measure  absorbing  carbon in  particulate
samples  collected for  X-ray fluorescence  analysis.    Absorbing carbon  is
assumed  to  be synonymous with  graphitic carbon  in  terms of visible  light
absorption  (1).    To  be  compatible  with X-ray  fluorescence analysis  (2),
ambient  samples  were  collected over periods  of  12  to 24 hours  by  dichoto-
mous samplers  (3)  using Teflon filter substrates.  The  analytical  approach
used to  detect absorbing  carbon on these samples has  been to make simultan-
eous  photoacoustic  and  transmission  measurements  using  a visible  light
source.   Measurements  were  interpreted using  calibration  curves  relating
response  to  loading of  flame-generated  soot.


                          EXPERIMENTAL  PROCEDURE

     We  have used an experimental arrangement similar  to that  described by
Lin  and   fellow  researchers  (4)  to  record  transmission and  photoacoustic
signals  for ambient  samples collected  on  Teflon  filter  substrates  (Ghia
Types  501 and 504).   To  obtain the transmission value,  a  fraction  of  the
forward  scattered light is  measured before and  after  loading  with  particu-
late matter.  The relationship  of  the  signal  differences  to  the  sample
loading  of  absorbing  carbon has  been  examined.   We  did  not   observe  the

                                     184

-------
 sampling  constraints  specified by  Lin;  other considerations  related  to
 compatibility with X-ray  fluorescence analysis dictated  the  sample charac-
 teristics.  The  experimental arrangement also  used  a  photoacoustic cell to
 provide an analytical  signal  related  to  sample absorption.

      Figures la  and  Ib show  the  cell-holding  Teflon  filters.   Light  from
 either a broadband or  laser  source  is directed through a glass  window onto
 the side of the Teflon filter containing the  particulate  matter.   The light
 is scattered  and absorbed within  the sample.   A portion  of the  light  is
 forward-scattered  through the Teflon  filter  and opal  glass  diffuser  to  a
 photodiode detector  (RCA, 1P39) and recorded  as the  transmitted  signal.  In
 the latest configuration, a  15-mW HeNe laser beam is  expanded and directed
 onto a beam-defining aperture of  1  cm diameter just before passing through
 the optical  cell's front window.   The  filter is  usually  placed  with  the
 aerosol  deposit  facing  the  light  source,   although  it   can   easily  be
 reversed.

      Pressure seals  are made with  0 rings  for those Teflon  filters  bonded
•into frames, or  with the retaining ring in filters not  bonded.    The  opal
 glass  is  used  to  reduce  the effect  of  scattering,  and of  substrate  and
 sample inhomogeneities, by diffusing  the incident  beam to form  the exiting
 pattern of a diffuse radiator.  An aperture in  the aluminum retaining plate
 for the opal  glass  limits  the  exiting  beam  before  it  reaches  the  photo-
 diode.

      The  light  absorbed  in  the sample   causes  a slight increase  in  the
 pressure of the closed volume in  front of  the  Teflon  filter.  Because  the
 light beam  is  mechanically  chopped,  the  pressure  variation is  periodic.
 This  variation is  detected by a Knowles  electret microphone  (BT  1759)  and
 processed  by a lock-in amplifier.

      To relate  photoacoustic  and   transmission signals  to  the   graphitic
 carbon content of  the  particulate  sample,  the following sequence is  per-
 formed.  The blank Teflon filter is mounted in the  cell  and  a transmission
 measurement  is recorded.   The  photoacoustic  signal is  also  measured,  but
 since  the  light  absorbed by  the blank  filter  is quite small  (approximately
 10    times smaller than  the  signal for  typical ambient  loadings), a  pre-
 loading measurement of the  photoacoustic signal is  not  usually  necessary.
 After  the  filter is  loaded,  the sequence  of  measurements is  repeated."  A
 value   of  absorbance,  -ln(l/I0),   is  formed   from  the  two  transmission
 measurements.   For  the photoacoustic  analysis, the  same before  and after
 measurements  are  subtracted   to  form  a  net  signal  value.   These system
 responses  are used  to find the amount  of a reference sample that  would have
 given  an  equal   signal.    This  correspondence is  made  using  calibration
 curves  that  relate  system response to  reference sample loading.   The cali-
 bration curve  is  established  by   plotting absorbance   and  photoacoustic
 response as  a function of carbon  loading for  filters  on  which  known load-
 ings  of soot have  been placed.    Typical calibration  curves  are  shown in
 Figures  2a and  2b.   Figure 2a gives  the  value  of absorbance, (1-I/IO), as
 well  as -ln(l/I0).    The measured  response  of  either system  is  located on
 the ordinate  scale  of  the corresponding  calibration curve, ' and  a value of
 graphitic  carbon  loading  in  micrograms  per centimeter squared is  located on

                                     185

-------
            AIR CHANNEL
           TO MICROPHONE
MICROSCOPE SLIDE
                       ELECTRET
                      MICROPHONE
                      37mm FILTER
                   BONDED TO HOLDER
             OPAL GLASS
              DIFFUSER
             PHOTODIODE
Figure  1.   Schematic  of  detector  cell  used  in  optical  measurements  of
            particulate  samples  (a)  front  view  of  disassembled parts,  (b)
            back view.
                                      186

-------
                                                                  - 10.0
                                      80     100
                                     LOADING, /ug/cm2
             2.0
             1.5
           1
           £•
           85 1.0
             0.5
         "i	1	1	r
                            o   o
-  6°
  o
                       J	L
                                                n	1	\	r
                              o    o
                                o '
                        J	I     I     I
J	L
              0    10   20   30    40   50   60   70   80   90    100    110
                                    CARBON LOADING, fig/cm2

Figure   2a.  Calibration  curve  relating   optical  absorptance  and   optical
             absorbance  (BL) to  sample weight loading  of soot  in yg  per  on2
             of filter area.

Figure  2b.  Calibration  curve  relating photoacoustic signal to sample weight
            loading of soot in yg per  cm2 of  filter  area.
                                       187

-------
Che  abscissa scale,  using  the  previously  established calibration  curve.
The usual ambient loadings  of  absorbing  carbon are below 10 yg/cm2,  so that
the calibration curves must be used in a region between actual  data points
and  zero loadings.    However,  as  shown  on  the   calibration  curves,  the
extrapolated  curve   is  well defined.    Imprecise   weighing  of  the  lightly
loaded samples causes  this  lack  of  data  points in  the useful range of load-
ings.

     The soot used  in loading a  reference  set of  filters  was  generated in
an oxygen-rich, propane  flame, as described  by Bennett (5).  The soot  was
mixed with  dry  air  to supply  the required  flow rate  for  a manual dichoto-
raous sampler.  Soot  loadings were weighed on  a Mettler balance  using a set
of certified weights  for  calibrating the  balance.   The  fraction  of gra-
phitic  carbon was  measured for  13  samples  as  0.80  ±  0.06  (one standard
deviation)  by combustion  analysis.
                                   RESULTS
Analysis of  Synthetic  Samples
     Previous  research  using the experimental apparatus shown in Figures la
and  Ib has  been published  (6,7),  but  is  limited  to  analyzing  synthetic
samples  that are intended  to  simulate ambient samples.   Three conclusions
are  drawn:

     1.   thermal effects,  as well as  carbon absorption,  determine the
          photoacoustic  response versus soot loading relationship;

     2.   mixing   ammonium  sulfate  with   a  given   amount   of  soot
          significantly   increases   the  apparent   absorption  values
          derived from differential  transmission measurements; and
      3.   photoacoustic  measurement  of  absorption  values
          slightly changed by mixing ammonium sulfate and soot
are
only
     The  first  conclusion is at  least  partly  due to thermal wave interfer-
ence  (6).   The  thermal effects are  evident  in the calibration curves shown
in  Figures 2a  and 2b.   In  each figure,  sample loading  for a  sample of
propane  flame-generated soot  is  plotted as the  abscissa,  with response as
the ordinate.   In Figure 2a the  response  is  given in units of absorptance,
(1-(I/I0)),  and  of absorbance,  (In I0/l) or Bl.   The response  is given
in mv  of signal per mW  laser  output power in Figure 2b.   The first of  the
above  conclusions  is .evident in  the shape  of  the  two calibration curves
near  a loading  of 80 yg/cm2.   At  this loading,  Figure 2a  shows  that no
additional light is available for absorption, while  in  Figure  2b  the photo-
acoustic  response continues to increase.   Conclusions  2 and 3 are  shown in
graphic  form in reference 7.
                                     188

-------
 Analysis  of Ambient Samples

      During the  research  on  synthetic  samples,  several  sets  of  ambient
 samples were analyzed.   These samples were the fine fraction of particulate
 matter  collected with  dichotomous samplers  in special  U.S.  Environmental
 Protection  Agency   (EPA)   field  studies.    A  set  of   routine  analytical
 procedures,   consisting  of   analysis   by  X-ray   fluorescence   and  mass
 determinations  by beta-gauge  (8),  was  performed  on the samples.   In some
 casesj  duplicate samples collected  on  quartz  fiber filters  were available
 for  analysis of total,  volatile,  and graphitic carbon (9).,

      A  summary  of the data appears  in  Table  1, as  mean  values  for several
 measured  parameters for each  of  the  five  data sets.  The  sets  are ordered
 so  that  the value  of  the  parameter  BL/PA decreases monotonically.   Three
 tentative  conclusions  seem consistent with these data:
      TABLE  1.   SUMMARY OF MEAN VALUES FOR PARAMETERS CHARACTERIZING
                 PARTICULATE MATTER COLLECTED ON TEFLON SUBSTRATES
      	(18-33  filters  per set)	
Location
EC*      PAt       BLJ       MASS§       BL/PA'
            (micrograms per square centimeter)
                                                                       CC#
Houston
Philadelphia
Shenandoah
Philadelphia
Houston
2.2
6.1
	
4.6
	
2.8
4.7
0.9
5.7
6.7
6.1
8.0
1.3
6.1
6.1
64
61
42
38
38
2.1
1.7
1.4
1.1
0.9
0.96
0.77
	
0.86
	

tAbsorbing carbon  determined  by photoacoustic  measurements
$Absorbing carbon  determined  by differential measurements.
§Mass of all particulate  collected.
//Correlation coefficient  for  BL and  PA measurements.
         BL/PA  and  MASS  are  correlated.   Since  PA values  are  not
         changed  by  adding  ammonium sulfate to a given  amount  of soot
         for  lab-generated  samples,  the .parameter  BL/PA for  ambient
         particles appears  to  be an enhancement factor  in the  differ-
         ential transmission measurement.  This enhancement  factor  is
         apparently  related to MASS.

         The values  of  EC,  PA,  and BL do  not  appear to be  related  in
         any  simple  way.    The  combustion value  for elemental  carbon
         and  the   optical  value  for  absorbing carbon  probably  vary
         depending on location.

         The value of  the  correlation  coefficient  (C.C.)  for  paired
         values of BL and PA  is  higher  for  the  filter set  (Houston,
         Texas) using the  thicker,  Type  501,  Teflon  filters.    This

                                     189

-------
      greater  value  is   consistent   with  other  observations   in   the
      laboratory.   The  effect  is  apparently  related  to  the   greater
      variability in mass for the .blank Type 501  filters, which  in  turn
      corresponds to  a variability in the  transfer  of  energy from  the
      sample to the surrounding air.

      A relationship  between BL/PA and MASS  is also evident for  individual
 filters within a single  set.  This  relationship is  shown in Figure 3 for  a
 set  of 25  filters  taken  in Philadelphia,  Pennsylvania.   The  data  were
 fitted using  a  linear,  least squares analysis.   An intercept   value of  1.0
 would correspond to  agreement of  BL  and  PA for low  loadings  of  MASS.
                                  PHILLY 235802-235827
DC
                R = BL/PA

                MASS IN MICROGRAMS
                     PER CMA2
                                                                o
                                                                o
                                                                oo
o
8
                                          MASS

  Figure 3.   Ratio, R,  of soot  estimates  by  optical  absorption  and photo-
             acoustics versus total mass loading per filter.
  Modeling of the Differential Transmission Measurement

       A mathematical  model  has  been  developed to  predict  the  results  of
  differential  transmission  measurements.    The model  simulates  the  light
  attenuation with  a mixture of absorbing  and  scattering particles,  using  a
  modification  of  a mathematical  treatment  listed  by Kortum  (10).   Kortum

                                      190

-------
assumes  that  the  sample  consists  of  layers  that  are  assigned  values  of
reflectivity,  R,  and  of  transmission,  T.    Using  the  closed  form  of  a
geometric series,  the  values  of R and T are  obtained for  a  combination of
two layers.  The resultant R  and  T  values  are assumed to be for a composite
layer, which  is  combined  with  an additional  single  layer  to  obtain values
for R and T  for  three  layers,  etc.    Thus  any  number of  layers  can  be
considered,  as  indicated  in  Figure 4.  To adapt  this  procedure  to  light
transmission  through  a sample of discrete  particles,  the  following assump-
tions were made:
Figure 4.  Schematic  representation  of  model used  in  calculating  light
           transmitted  through  and reflected from  particle  layers  in  a
           s amp1e.
     1.  The particles are divided  into  two  mutually exclusive groups:
         scattering  and  absorbing.   These correspond to soot  and  all
         particles minus soot.

     2.  All particles are spherical and  each  group has  a single diam-
         eter that is specified  at  the beginning  of the  treatment.

     3.  A scattering particle is characterized by a backscatter under
         uniform illumination that  was initially  estimated  using a Mie
         theory calculation  (11).   Contributions   of  individual parti-
         cles comprising a layer are summed  to obtain a  filled layer's
         reflectivity.  This  estimate  was modified as required to  fit

                                     191

-------
          experimental  data;  the modification appears justified, as the
          individual  particles  are  so close  that  assumptions  used to
          derive  the  Mie formulation are  not necessarily valid.

     4.   Absorbing  particles are  assigned an  absorption  coefficient;
          the  light  throughput is characterized by the average parallel
          transmission  of light  through the intercepting cross-section.
          The  transmission of  a single  layer  is  equal  to  unity minus
          the  reflectivity minus  the sum  of absorption by individual
          particles.

     5.   The  last "layer"  in the model  is the combination  of Teflon
          filter  and  opal glass, where a  reflectivity of 63 percent was
          measured at 45° and used  as the  average reflectivity of the
          last  "layer."

     The  program for calculating sample transmission  was  begun by specify-
ing  the  absorption  coefficient  of  soot,  B; diameters of  particles,  Dl for
soot  and D2  for scattering  particles;  reflectance  of  the  substrate, R2;
mass  loadings of soot and scattering particles,  Ll  and  L2,  respectively;
and  a packing  fraction,  P,  which accounted for the amount  of available area
(84  percent)  actually  occupied by particles in a  layer  composed of spheri-
cal  particles.

     The  model predictions  for  absorbance  versus  the graphitic carbon  load-
ing  for  simulated ambient  samples  are shown in Figure 5.  Individual curves
correspond  to  the indicated  loading (M in micrograms per filter)  of non-
absorbing particles.   The  curve for M =  0  is calculated by using the graph-
itic  carbon weight  to  define  the  number  of  particle  layers  in  a  sample
(i.e., spherical  particles  of a given diameter closely  packed  onto a given
surface  area,  with  a fractional void area, have a mass  which,  when divided
into  the total  mass,   gives  the  number of  layers   in  the  sample).    The
remaining curves, M  =  100,  300, 500,  and 700yg/filter, use the mass of non-
absorbing particles  to  determine the number  of  layers and  disregard the
mass  contributed by graphitic  carbon.    While this  simplification  is  of
value in  making  the  mathematical treatment  tractable, the  resulting curves
may  not  be appropriate for  sufficiently   high values of  graphitic  carbon
loadings.   In  fact,  the unexpected crossover of the M = lOOyg/filter and M
-  Oyg/filter  curves  and bending  of the  remaining  curves  for  the  higher
graphitic carbon mass  to nonabsorbing  mass ratios  may  result.   Even with
its  limitation,  the model  clearly  predicts  that:  Oscattering  from non-
absorbing particles  causes  an overestimate of graphitic carbon;  and 2) the
higher the  nonabsorbing particle loading,  the  greater the  overestimate.

     Figure  6 shows  a  comparison  of  model  predictions   and  experimental
measurements  of  absorbance  for a set of field-loaded  filters  from Houston.
Model parameters  were:  B = 13 x 106 m"1 ;  Dl  = 0.02  u;  D2 =  0.084 y; R =
0.1; R2 *= 0.63; P =  0.9.  The value of B approximately equals that measured
by Rohl and fellow researchers  (12).   The value of Dl is  an  estimate based
on existing research;  the value of D2 was  taken from  recent  work on Denver
aerosols  (13).  The  value of  P  is  an estimate; the value  of  R was adjusted
as required to obtain  a good  fit between experimental results and predicted

                                     192

-------
                      TRANSMISSION
               1.0
               0.8
               0.6
              ' 0.4
               '0.2
                                                       700
                           10
    20

SOOT fag/filter)
30
                                                          40
                        M = > MASS OF NON-ABSORBING PARTICLES
Figure 5.  Model  predictions  of absorbance  (ln(lo/l)), versus  soot  load-
           ing with  mass,  M,  of scattering particles as a  parameter.
values.    All  but  one  of  the  points  located  above  the  straight  line
correspond  to filters taken  during sequential  12-hour  sampling periods  at
the monitoring  site  (the numbers associated with individual  points  refer  to
the order  in  which  the  samples were taken).   This  subset of  filters  may  be
physically  different  from the res-t of the set.  As noted earlier,  the model
predictions appear  too  low for samples having higher values  of  soot-to-mass
ratio.  Points  corresponding  to  filters 16 to  19,  which are significantly
lower  than experimental  values,   have  an  average  soot-to-mass  ratio  of
0.125, as  compared  to 0.53 for the remainder of the set.
                           SUMMARY AND DISCUSSION

     The ratio of  absorbing carbon estimates BL/PA increases  as  sample  mass
increases  for mean values  of  different  filter sets  and  for  individual
values  within  a  set.    Based  on  similar  experiments  with   laboratory-
generated  samples, this relationship  appears to be  due to the  enhancement
of  the  BL estimate.   Some  doubt  remains,  however,   since  the  laboratory-
generated  ammonium  sulfate  particles  were  considerably  larger  in  mean
diameter (0.4 to 0.6y )  than the ambient particles.

     Average  estimates of  elemental carbon  by combustion (for a  given  fil-
ter set)  are  different  from absorbing carbon  estimates.   The  relationship
changes from  one filter  set  to another and probably depends on bulk absorp-
                                     193"

-------
           1.0
            D.8
          5 0.4
                           9 O
                                 0.4         0.6
                                   BL (EXPERIMENTAL)
                                                      0.8
                                                                1.0
Figure 6.  Model  predictions   of  absorbance,  BL,  versus  experimentally
           measured values  for  18 samples collected in Houston, Texas.


tivity of  combustion  carbon.  For  individual filters from  the  two sets of
filters  loaded  in  Philadelphia,  the  relationship  between EC  and  BL is
linear, with  reasonably low scatter  and a correlation  coefficient of 0.95
(information  from  personal  communication  with  W.J.  Courtney,  Northrop
Services,  Inc., Research Triangle Park,  NC).   If BL is being enhanced,  then
its enhancement is  linearly related to EC.

     PA estimates of  absorbing carbon are  less  precise  for Type 501  Teflon
filters.   This  fact is  apparently related to a corresponding variability in
the mass  (and I0 values)  of  individual filters within  a  set.   Given the
same amount of  soot,  thick  filters give lower signals than  thin filters.

     The  BL values for a  set  of 18  filters loaded  in Houston  have  been
predicted  with  some   success  using  a multi-layer  particle  model with
adjustable parameters  for particle size, layer reflectivity, and absorption
coefficients.   Prior  knowledge  of some  of  these parameters  is   required
before true predictive  capability can be realized.


                                 CONCLUSIONS

     The  following  conclusions  seem  justified based  on  the experiments so
far completed:
                                     194

-------
     1.  For  input  to  visibility  calculations,  use  optical measure-
         ments of absorbing carbon.

     2.  Transmission  measurements apparently  overestimate  absorbing
         carbon for higher filter  loadings.

     3.  For  PA estimates  of  absorbing  carbon,  use  thicker   filter
         substrates, such as Type  501 Teflon or Nuclepore  filters.

     4.  Reference  optical  measurements to a standard  soot  sample  of
         known absorption properties  to establish comparability  among
         different measurement  systems.
                           ACKNOWLEDGMENTS

     The author acknowledges  the  efforts  of Dr. C.A. Bennett, Jr.,  and Dr.
R.R. Patty  of  the North Carolina  State University Physics Department, and
Mr. M.A. Mason  and  Dr.  W.J.  Courtney of  Northrop  Services, Inc.,  for  pro-
viding  the  experimental measurements shown throughout this  work.    Thanks
also to Mr. R.K.  Stevens and members of his  staff  for  sustained  interest  in
this project.
1.
2.
3.
4.
5.
6.
                            REFERENCES

Rosen, H., A.D.A. Hansen, L.  Gundel,  and T.  Novakov.  1978.   Identifi-
cation of  the  optically  absorbing components in urban aerosols.  Appl.
Opt. 17:3859.                                   '                  ~~

Dzubay, T.G.,  and D.G.  Rickel.    1978.   X-ray fluorescence analysis  of
filter collected aerosol  particles.   _In P.A.  Russell  and A.E.  Hutch-
ings,  eds.,  Electron  microscopy and  X-ray  applications.    Ann Arbor
Science Publishers, Ann Arbor, MI.

Dzubay,  T.G.,   and  R.K.  Stevens.  1975.   Ambient  air  analysis   with
dichotomous  sampler  and  X-ray   fluorescence  spectrometer.   Environ.
Sci. Technol. 9:663.

Lin, C.I., M.  Baker,  and  R.J. Charlson.   1973.  Absorption coefficient
of  atmospheric  aerosol:  A  method   for  measurement.     Appl.   Opt.
12:1356.                                                    ~
Bennett,  C.A.,  Jr.    1981.    Photoacoustic  detection  of particulate
carbon.  Master's Thesis, North Carolina State University, Raleigh, NC.
60 pp.

Bennett, C.A., Jr.,  and  R.R.  Patty.   1981.   Evidence for thermal wave
interference  in thin  particulate  carbon   samples.    Submitted  to  J.
Photoacoustics.
                                    195

-------
9.
Bennett,  C.A.,  Jr., and  R.R.  Patty.   1982.   An  evaluation of  photo-
acoustic  and  transmission techniques for monitoring particulate  carbon
collected on Teflon  filters.  Appl.  Opt. 21:371.

Jaklevic,  J.M., R.C.  Gatti,  F.S.  Goulding,  and  B.W.  Loo.   1981.    A
g-gauge  method applied  to aerosol  samples.   Environ.  Sci.  Technol.
15:680.                                         	

Mason, M.A.,  and  J.W.  Tesch.   1980.   Carbon  analysis  on  the  Dohrmann
DC50    Envirotech  organic  analyzer.    Progress   report   ES-TN-81-02.
Northrop  Services,  Inc.,  Environmental  Sciences,  Research  Triangle
Park, NC.  17 pp.
10. Kortum,  G.
    York, NY.
              1969.    Reflectance  spectroscopy.    Springer-Verlag,  New
11. Bhardwaja, P.S., J. Hubert, and R.J. Charlson.   1974.  Refractive  index
    of  atmospheric  particulate matter:  An  in situ  method  for  determina-
    tion.  Appl. Opt. 13:731.

12. Rohl, R., W.A. McClenny,  and  R.A.  Palmer.   1982.  Photoacoustic deter-
    mination  of  optical  properties   of  aerosol   particles   collected  on
    filters:  Development of a method  taking into account  substrate  reflec-
    tivity.  Appl. Opt. 21:375.

13. Countess,  R.J.,  S.H.  Cadle,  P.J.  Groblicki,   and  G.T.  Wolff.    1981.
    Chemical  analysis  of  size   segregated  samples   of   Denver's   ambient
    particulate.  J. Air Pollut.  Contr. Assoc. 31:241.
                                    196'

-------
                 STATUS OF  SAMPLING  AND  ANALYSIS OF AMBIENT
                      NITRIC ACID, NITRATES  AND AMMONIA
                  Robert K'. Stevens  and Robert  W.  Shaw,  Jr.

                    U.S. Environmental Protection  Agency
                         Research Triangle Park, NC

                                Robert Braman

                         University  of South Florida
                                  Tampa, FL

                                 C.W. Spicer

                                   Batelle
                                Columbus, OH
                                INTRODUCTION

      As  investigators have  improved the methods  for  sampling and  analysis
 of  atmospheric  nitrate,  it has become evident that distribution  between the
 gaseous  and  particle phases  has  often  been masked  by experimental  arti-
 facts.   Many early particle nitrate data were based on  analyses  of  extracts
 from glass  fiber aerosol filters used in Hi-vol  samplers.   It is now  known
 that  these  filters contain  active  sites  that fix gaseous  HN03 and make  it
 appear as particle nitrate (1).

      Other  filter materials have also been  shown to  react  with  and collect
 gaseous^  HN03  and create  a positive  particle  nitrate  artifact (2).  The  use
 of  an inert  filter material,  such  as Teflon, removes  the "positive  arti-
 fact" problem,   except for the possibility  of reaction of  gases  with  the
 collected  aerosol particles.   It has been  shown, however,  that collected
 aerosol  nitrate particles (true particle nitrate) may be lost from filters
 because  of  reactions  with other materials  or evaporation.   Loss of parti-
 cles  is  known  as "negative  artifact."   Reactive loss  may occur  if,  for
 example, H2SOif  aerosol comes in contact  on  the  filter surface with nitrate
 aerosol.  Evaporative loss  may occur if, due to decreases  in ambient  gas
 concentrations,  the solid and  gaseous nitrate phases  are no longer in  equi-
 librium.   Thus   we  see  that measurements of  particle  nitrate by  means  of
 glass fiber  filters  are  expected to  be  systematically high.   On the  other
hand,  measurements of  particle  nitrate  using  inert  Teflon filters  are
 expected to be  systematically  low.   The  extent of loss  due to reaction  and

                                    197

-------
due to reaction  and  evaporation are difficult to predict.   Recent  measure-
ments indicate  that,  because  of the distribution between HNOg  and  particle
nitrate, the glass fiber-filter nitrate  overestimates  may be considerable.
                       TECHNIQUES  AND  THEIR FEATURES

     Methods  currently under  development  or  used in  research studies  to
measure gas phase nitric  acid  include  the use  of continuous (real-time) and
semi-continuous  monitors,  as  well  as integrative  collection  of  HNOs  on
adsorbing materials.   The  continuous methods (1) are:   a)  chemiluminescence
and b) Fourier   transform  long  path  infrared   spectrometry  (FTS-LPIR) .
Methods involving preconcentration (1) are:

     (a)  collection  of HN03 on nylon or  cotton,  followed  by extract-
          ion,  conversion to  nitrobenzene (CgHgNC^),  and  analysis  by
          gas chromatography;
     (b)  reduction  to ammonium ion  (NH^"*")  of fixed  inorganic  nitro-
          gen   collected   on   nylon  filter,   followed  by  iodophenol
          ammonia  test;

     (c)  collection  of HN03  on  sodium chloride-impregnated  filters,
          followed by  extraction  and  hydrazine reduction-diazotization
          analysis of  nitrate;

     (d)  denuder  difference  experiment — collection of  total  nitrate
          from  two parallel air  streams,  removal of nitric  acid  from
          one stream using a diffusion denuder,  and subsequent  deter-
          mination by  difference  (3);  and

     (e)  continuous   chemiluminescence method — a  dual  chamber  NOX
          chemiluminescence monitor is modified  so  that  both  sides  of
          the  split  sample stream pass   through molybdenum  catalyst
          converters  to reduce  the oxides of  nitrogen to NO;  one  of
          the  sample streams  passes  through  a nylon  filter  to  remove
          the  HN03;   the   difference   of  the  two  signals  from  the
          instrument  is  recorded  as nitric  acid.

     (f)  Tungsten VI  oxide technique—short-term collection of nitric
          acid  on diffusion  tubes followed  by  release  and  detection
          using chemiluminescence (4).

This last technique  is a considerable advance, permitting measurements down
to 0.25  ug/m3 of nitric  acid for  sampling intervals of 20 minutes and also
permitting  simultaneous  measurement of NH3.

     Each  technique  has  unique features.  FTS-LPIR  is  suitable  for provid-
ing benchwork measurements of HN03, because  measurement  takes  place in the
atmosphere  and  identification is  made unambiguous   by the  recognition  of
characteristic  infrared  absorption.    However,  the   equipment  is  not  por-
table,  and  the  method has  a minimum  detection level of  5  ppb.   Chemilumi-

                                    198

-------
 nescence has the  inherent  sensitivity of  a rate  sensor,  but involves  the
 measurement of small differences  in a signal  that  is  frequently  large  and
 time-varying due  to  interferences  (for example,  total  oxides of  nitrogen,
 NOX;  peroxyacetyl  nitrate,  PAN; and  organic  nitrates).   Collection  tech-
 niques  generally require relatively simple equipment; however, they  require
 documentation of  collection  and release  efficiencies,  maximum  loading,and
 possible interferences.

      In order to  compare  a number of  techniques  for measuring  atmospheric
 nitrate and nitric acid, investigators from Brookhaven National Laboratory,
 the  University of  Colorado,  the National Center  for Atmospheric  Research,
 the  University  of Michigan,  Battelle,  the  U.S.  Environmental  Protection
 Agency,  and the  University of California  at Riverside  gathered for  a  field
 study in Claremont, California during August 27  to September 3,  1979, with
 the  following objectives:   (a) to test and  compare measurement systems  for
 HNO   vapor and nitrate  aerosol;  and (b)  to  determine the  extent  to which
 measurement systems are susceptible  to artifact  nitrate  formation or loss
 and,  if possible,  to determine  the  cause of  artifacts.    The various mea-
 surement  systems  were available  at  that  time,  and we believed  that  their
 comparison could  best be. achieved by  side-by-side  evaluation under  field
 conditions,  similar to  the EPA Charleston Aerosol Sampler Comparison Study
 (5).   In  order to determine  the  absolute reliability of  the instruments,
 all  were  run side-by-side  with  FTS-LPIR,  which was  the  only  technique
 established at that time to  be interference-free.

      Detailed  results  of the Claremont study have been  published by Spicer
 and  fellow researchers  (6).   Briefly, consistent results  for nitric acid
 were  achieved using chemiluminescence,  FTS-LPIR,  and the denuder  difference
 experiment.   Of   four methods using  particle  filters  followed  by  nitric
 acid-absorbing filters,  two were  in  agreement  with  the  mean  of  the other
 methods  and two  showed  apparent  excess  nitric  acid;  hence,  these  tandem
 filter  techniques  may  not  be used  with  as  much  confidence  as  the other
 methods.  ^Collection of particulate  nitrate  on Teflon filters gave  gener-
 ally  consistent  results, although  comparison  with the denuder  difference
 experiment  suggests that some nitrate  loss  from Teflon filters  did  occur.
 Measurement  of nitrate on quartz  and  glass  fiber  filters  gave greatly vary-
 ing results, possibly depending  on filter  history.

     During  the  Claremont study,  levels  of nitric acid  ranged  from  2  to
 40yg/m3  and particulate  nitrate  from  4  to 26yg/m3.    The good  agreement
 among measurement  methods  observed in Claremont may  not be  reproducible  in
 other areas^ where nitrate  levels  are not so high as they  are in the Los
Angeles Basin.  For example,  measurements were made  during  the summer 1980
 in Research Triangle Park, NC, using  the  denuder  difference  experiment, the
Tungsten  VI  oxide  technique,  and  single  filter methods.    The  denuder
difference  experiment  and  the Tungsten  VI oxide technique  were  in  good
agreement when they were run  simultaneously.  Simultaneous  measurements for
particle nitrates  using  filter collection, compared to nitrates measured  by
denuder  difference  method,  showed   much  lower  nitrate   values,   due  to
evaporative or reactive  loss  from the filter.

     A few  reported  values  from measurements not  thought  to  be  subject  to

                                    199

-------
artifact problems are presented in Table  1.
           TABLE 1.   LEVELS OF PARTICIPATE NITRATE  AND  NITRIC ACID
.
yg/m3
Location
RTF, NC
Date
July 1980

(day)
(night)
N
1
0
O^j
.2
.9
HN03
3.4
0.7
Reference
3
Claremont,  CA
Rocky Mountains
(10.000 ft)
Aug-Sept 1979 (max)
              (min)
            26
             4
1979
(max)
38
 1

 0.01
     Recently  at  EPA, we  have modified  the  denuder  difference  nitrate-
nitric acid sampler, as shown in Figure 1.   In this  sampler,  the  aerosol  is
sampled through  an all-Teflon cyclone  at  28 liters per  minu.te  to  remove
particles > 2.Sum.  From this air stream, two  parallel samples at  3  liters
per minute  are collected simultaneously; one  aerosol  sample is  collected
(FI ) on a  nylon  filter (1 m pore size)  and  the  other  sample is  collected
(F2) after the aerosol passes through  a diffusion denuder  coated with  MgO.
This sampling  arrangement  minimizes large  particle  deposition  within the
sampler  and  reduces  HNO3  losses   within  the  cyclone  due  to  its   inert
    INLET
                                    DENUDER
                                                         ALUMINUM HOUSING
                                    TEFLON
                              ur"1—n
                           infCFO
                                                 Fi
              DEPOSITORY
                          BC  - BALLAST CAN
                          F   - NYLON FILTER HOLDER
                          CFO - -CRITICAL FLOW ORIFICE
Figure 1.  Schematic of diffusion denuder nitrate-nitric acid sampler.

                                    200

-------
 construction  (Telfon)  and  large   sample   flow.    The  nylon  filters  are
 extracted with  a 1:1 mixture  of 0.005 M  solutions of  NaHC03 and Na2C03.
 The extract is analyzed for nitrate by ion  chromatography  (8).


                                 CONCLUSION

      Work  is  continuing  to   compare   the  performance  of  the  denuder
 difference method with the tungstic acid-chemiluminescent  method  for  nitric
 acid.    Comparisons  have  been  conducted  in  Denver in  1982,  and another
 comparison is planned in North Carolina.   Braman has recently  automated  his
 tungstic acid denuder method for  NHg and  NH03  to permit measurements  of
 particle nitrate and  ammonium.   The  system typically  collects the ambient
 aerosol  for 30 minutes.  This  is  followed by analysis  of  the aerosol col-
 lected for nitrate and ammonium,  followed  by the  analysis of the HN03  and
 NH3  collected by  the denuder as  described by Braman  (4).   The   detection
 limit  for nitrates  and HN03 is 0.2yg/m3.
 1.
2.
3.
4.
5.
6.
7.
8.
                            REFERENCES

 Stevens,  R.K.,  ed.    1979.    Current methods  to  measure  atmospheric
 nitric  acid  and nitrate artifacts.   EPA-600/2-79-05, March.

 Appel,  B.R.,  S.M.  Wall, Y. Tokiwa,  and  M. Haik.   1979.   Interference
 effects  in  sampling  particulate  nitrate  in   ambient   air.    Atmos.
 Environ. 13:319-325.                                             	

 Shaw, R.W.,  R.K.  Stevens,  J.  Bowermaster,  J.  Tesch, and E.  Tew.  1981.
 Measurements  of  atmospheric  nitrate and nitric   acid;  the  denuder
 difference experiment.   Atmos.  Environ.  16:845-853.

 Braman,  R.S.,  T.J. Shelley,  and W.A. McClenny.    1982.    Tungsten VI
 oxide  for  preconcentration and  determination  of gaseous and  particu-
 late ammonia and nitric acid  in ambient  air.  Anal. Chem. 54:358-364.

 Camp, B.C.   1980.  An  intercomparison of results  from samplers used in
 the  determination  of  aerosol   composition.    Environ.  International
 4:83-100.                                       —'	

 Spicer,  C.W.,  J.E.  Howes, Jr.,  T.A. Bishop,  L.H.  Arnold,   and  R.K.
 Stevens.   1982.   Nitric acid measurement  methods—an intercomparison.
Atmos. Environ. 16:1487-1500.

 Kelly,  T.J., D.H.  Steadman,  J.A.   Ritter,  and  R.B. Harvey.    1982.
Measurement  of  oxides  of nitrogen  and nitric  acid  in  clean air.   J.
Geophys. Res. in press.

 Stevens,  R.K.,   T.G.   Dzubay,   D.   Rickle,  and  G.  Russwurm.     1978.
Sampling and analysis  of atmospheric  sulfates  and related  species.
Atmos. Environ. 17:55-68.
                                    201

-------
     A SIMPLE DESIGN FOR AUTOMATION OF THE TUNGSTEN VI OXIDE TECHNIQUE
                      FOR MEASUREMENT OF NH3 AND HN03
                       W.A. McClenny and P.G. Galley

                   U.S. Environmental Protection Agency
                        Research Triangle Park, NC

                       R.S. Braman and T.J.  Shelley

                        University of South  Florida
                                 Tampa, FL
                                 ABSTRACT

     The  tungstic  acid  technique  for  collection  and  analysis  of^ NH3  and
HN03 concentrations  in the ambient air has  been automated in a  simple  and
cost-effective  design.   The design  allows  complete separation of HN03  and
NH3 during detection.  Unattended  operation  in field trials has  been demon-
strated,  and a  three-day run sequence  with hourly  updates  is  shown.
                                INTRODUCTION

     Braman  and  fellow researchers (1) have  recently  described  a technique
for  separation and  collection of gaseous  HN03 and NH 3 using  a diffusion
tube coated  with  the selective sorbent tungstic acid,  H2 WOv   Application
of  the tungstic  acid  technique  (TAT)  to  ambient  air monitoring  has  been
reported  by  McClenny (2).   The TAT allows collection of HN03 and NH3 in the
presence  of  particulate matter without the use of in-line filters.  The TAT
appears  preferable  to  a simple  tandem  filter  arrangement  (collection  of
particulate  matter followed by collection  of gases  on a specially prepared
filter),  which is  subject  to  particle-to-gas and  gas-to-particle conver-
sions, depending  on filter type  (3).   In addition,  the TAT is sufficiently
sensitive to provide hourly  updates  of  the  ambient  concentrations.  If the
TAT  is used  with a longer-terra  integrated  measurement  technique,  such  as
th6  denuder  difference apparatus (3), a successful comparison would provide
an  extra  measure  of quality assurance.

     The  TAT uses a 4 mm  i.d.  Vycor  tube  coated  along a 35-cm  length^ with
tungsten   (VI)  oxide,   which  hydrates  to  tungstic  acid.   Ambient  air  is
pulled through the tube at 1.0 L/min.  The basis  for  a gas/particle separa-
tion is  the difference  between the  diffusion rates of HN03 and NH3 (0.122

                                     202

-------
cm Is  and  0.236  cm  /s,  respectively;  see  reference 4)  and  the ambient
particles   containing  their  corresponding  particulate  forms   (that   is,
[NH^SO^ and NH^NOg) .   The longitudinal distribution of HN03 and NH3 along
the  tube  is predicted  by Gormley  and Kennedy  (5),  assuming  that laminar
flow is maintained  in the tube,  and that every target molecule  reaching  the
tube wall sticks.   Sectional analysis  of  the longitudinal distribution  has
substantiated  the  predictions.   Under  the  established conditions  of use,
greater than 99 percent  of  the  ambient particle  mass is  expected to pass
through the tube,  while  the combined collection  and  release efficiency  of
the two gases is greater than 95 percent (1).
                           EXPERIMENTAL PROCEDURE

     A  simple  scheme for automating  the  TAT has been  developed  as a cost-
effective  alternative to  an automated- system  used  previously  (2)  and  is
shown in Figures  1  and 2.
VARIAC

V1

^
	 1




<_
* —
^
cl

                                  W03
                                 I PRECONCENTRATOR
He, 02
                                                    Au
TRANSFER CONVERTOR
^ L
JJ J J

VAHIAC
V2
J


)0 JJ

VARIAC
'
T
RO
A
0
DM
R
NOX
MONITOR

                  PUMP
Figure 1.  Schematic of  system designed  for monitoring HNOs and
           B, C, and D are  Teflon® solenoid valves.   A is a Teflon®
           ball valve.
                                     203

-------



iiov. — [py" — '

\
^



. I
• ^w 1






A-0
[
T
i ,









\
Mill 7 TDR1 I
N.C. r
L _ N.O. U\
' B 3 1
/ \

J

^N.O. N.C

SE(
)
J
'x , TDR 3
1 1 ' 	 1* +
B, C
If • "1
7
. a
"5
C;

LONG DELAY
COMPARED TO TDR 3
Figure 2.  Schematic of  electrical system  designed  for automated  sampling
           and  analysis   using  time  delay relays  (TDR).    NC  =  normally
           closed; NO = normally open.
Explanation of Operation

     The automated  sampler  operates  in  a  four-step  cycle  as   described
below:

     1.   Trapping -  Ball valve A*  opens;  solenoid D activated.  Pump
          draws  ambient  sample through the W03 diffusion tube,  which
          scrubs out NHs  and HN03.   Timing is controlled by TDR  1  and
          is typically 12 to 40 minutes.

     2.   HN03 Analysis - Ball valve A closed;  solenoid D deactivated;
          solenoids B and C activated;  power  supplied  to  Variac  Vj .
          W03 diffusion  tube is  heated,  releasing NH 3 as NH3  and HN03
          as N02-   A helium-oxygen mixture  is  directed  through  the
          diffusion tube.   This  carrier gas  entrains  desorbed  gases
          and  is  diverted  through  the   transfer  tube  and  convertor
          before  entering  the  NOX  monitor  for  measurement .     The
          transfer  tube  scrubs  out  the  NH 3  but allows NO 2  to  pass
          through  the N02~to-N0  converter to  the  NOX monitor.    The
          flow  rate  through the  converter  is  less  than  that  pulled
          into  the NOX monitor  so  that  room  air  is  sampled.    Some
          variation in baseline signal will  occur if  any NO  is present
          in thfe  room air.   TDR 2 is  involved  in timing this  and  the
          next cycle, and is typically set  for  10 to  15 minutes.
*Fluorocarbon Co., 1432  South Allen  Street, PO Box 3640,  Anaheim,  CA 92803,
 Teflon® ball valve Model EBVI-88.

                                     204

-------
      3.   NH3 Analysis  -  Conditions remain as  in  Step 2, except  that
           Variac  V2  is   activated.    The  transfer  tube  is   heated,
           releasing the NH3  for  measurement.   The NH, is converted  to
           NO in  the Au converter  and  is  sampled  by  the NOX monitor.
           TDR 3 is involved  in  timing  this  cycle and  is  typically  set
           for 3 to 5 minutes.

      4.   Cooling - Ball valve A remains closed; solenoids B, C, and D
           deactivated, Variacs V1  and  V2 deactivated.   W03 diffusion
           tube cools before  system returns  to  Step 1.  TDR 4 controls
           the length  of  this step, which may  last from  3  to  10 min-
           utes.    This TDR  can  also provide  a  time  interval   between
           sampling periods,  in  which case  longer  time  delays  may be
           used.

      Switching and timing are performed by  Potter  and Brumfield time-delay
 relays  with  variable  time  delays.   The  first  relay  (TDR 1)  completes a
 circuit,  supplying power  to  the first  step's  instrument  components, while
 its  own time  delay begins.   When  this  time  delay  ends, TDR  1 switches,
 breaking  the circuit of the  components  that operate  in  the  first  step  and
 turning on the second step  components  and the second  relay  timer  (TDR  2).
 This  proces^s of ^one  relay  turning on  the  next  continues until TDR  4 has
 completed  its  time delay.    Rather than turning  on  another   relay,   the
 switching  of TDR 4  interrupts  the power  to  the  timer in TDR  1, returning
 TDR 1 to  its normal position and breaking the circuit that powers the timer
 in TDR  2.    TDR  2  deactivates TDR 4  in  the  same manner.  However,  TDR 4's
 deactivation results  in power once again  flowing to  TDR 1,  thereby marking
 the beginning of a new cycle.  TDR 3 controls  the  transfer  tube Variac and
 operates during  part  of TDR 2's  time delay.

 Discussion of Components

      Reference  1  describes the diffusion tubes  and  the procedure for making
 them.   During thermal  desorption, a coiled heater  wire heats  the diffusion
 tube  to approximately 350°C.   The  converter  is constructed  of thin  gold
 foil,  folded and  fitted  into a  15-cm  quartz  tube.   A coiled  heater  wire
 also  heats  this tube  to  approximately  600°C.    The  NOX  analyzer  is  a
 Bendix  Model  1810B,  with  an ambient  air  ozonator  feed  and   a two-stage
 diaphragm  pump  for  evacuating the reaction chamber to  27  inches of mercury
 pressure difference.   An integrating stripchart recorder, Linear Instrument
 Co. Model  252A,  was  used to  record  and  integrate  the  data.  All  tubing used
 for transport of  gases was 0.64  cm i.d.,  thin wall  FEP Teflon®.

 Field Operations

     The system  was operated continuously over a 10-day  period  of  ambient
measurements  to  provide 20-minute  average concentrations of NH3 and HN03.
A  3-day measurement  sequence is  shown  in  Figure 3  as  a  chart  recorder
 trace.   Electronic  integration  of  peak  areas  was   also available  in  a
 length-of-line  format  on  the chart  recorder  output,  but  was   excluded  to
 improve clarity.    Peak  area values  above  a  preset,  reference   level  were
compared to  a calibration curve  relating  area  to calibrated mass  loadings

                                     205

-------
                                                                   0700 hrs
        0700 hrs
                • NIGHT
                            0700 hrs'

                          CONTINUOUS MONITORING OF AMBIENT HN03
Figure 3.  Three-day monitoring  sequence using the automated system.

of NH  from a permeation  tube  held at a constant temperature.  Because each
molecule of HN03 or NH3 is  converted to NO before  detection,  NH3 calibra-
tion was also used  to  infer  HN03 concentrations.  However, since the system
response to NH3  has been measured as slightly  higher  than the response for
HN03 (factor of  1.12;  see reference  1),  the  calibration is considered only
approximate.   The  1-ppb  HNO 3 peak  in Figure  3 indicates  the approximate
value of  the  individual  responses.   The range  of HN03  concentrations   is
estimated  as 0.1 to 1.2 ppb,  while  the NH3 concentrations  range from 0.05
to 0.3 ppb.   Detection capabilities over  these ranges  imply  a monitoring
capability in most, if not  all,  ambient monitoring sites.   Test data were
consistent  with  expectations for HN03  (that  is,  diurnal cycling, with mid-
day maximum and  low nighttime  concentrations).


                                 CONCLUSION

     A simple design to automate the tungstic acid  technique for monitoring
gaseous HN03  and NH3 has  been developed and demonstrated.   The system can
be left  unattended  to collect  and  record  integrated  average  values over
periods  adjusted  with time-delay relays.   The  system  has  been  used   to
demonstrate that  HNO 3 or NH3  concentrations  as   low  as  0.1  ppb  can   be
detected using  20-minute collection periods.
 1.
                           REFERENCES

Braman, R.S., T.J.  Shelley,  and W.A.  McClenny.   1982.   Tungstic acid
for  preconcentration  and determination  of  gaseous  and  particulate
ammonia and nitric  acid  in ambient  air.   Anal.  Chem.  54:358.

                                206

-------
McClenny, W.A.,' P.C. Galley,  R.S. Braman,  and T.J.  Shelley.    1982.
Tungstic  acid   technique  for monitoring  nitric  acid and  ammonia  in
ambient air.  Anal. Chem. 54:365.

Shaw, R.W.,  Jr., R.K. Stevens, J.  Bowermaster, J.W.  Tesch,  and E.  Tew.
1982.   Measurements  of  atmospheric  nitrate  and  nitric   acid:   The
denuder difference experiment.  Atmos. Environ.  16:845.

                              1955.   Estimation  of  diffusion coeffi-
                                      ;. Chem. 47:1253.
Wilke, C.R. ,  and C.Y. Lee.   1955.   Estimation  of d
cients for gases and vapors.  Ind. Eng. Chem. 47:1253.
Gormley, P.G., and M. Kennedy.  1949.
through a cylindrical tube.  Proc. R.
                                       Diffusion  from  a  stream flowing
                                      Ir. Acad. 52A:163.
                                207

-------
                           OZONE PRECURSOR MONITOR
                                    (0PM)
                       FOR INVESTIGATING AIR POLLUTION
                              Gordon C.  Ortman

                     Office of Research  and Development
                   U.S.  Environmental Protection Agency
                         Research Triangle Park,  NC
                                  ABSTRACT

     Ozone is  designated  by the U.S.  Environmental Protection Agency as one
of six principal air  pollutants.  Of the six,  ozone  alone  is classified as
a secondary pollutant  since it is not emitted  by  a  specific  source.   Ozone
is formed  in  the  lower atmosphere by photochemical  reactions involving the
precursors  of  ozone:  gaseous  organic  compounds  (mostly  hydrocarbons),
nitric oxide,  and nitrogen dioxide.

     This  paper  describes  a  new automated   method  for  quantifying  the
precursors  of ozone.   An ozone  analyzer is coupled  to  ,a reaction  vessel
contained  in  an irradiation chamber.   At timed intervals,  discrete samples
of  outside air  are  drawn into  the  reaction  vessel and  irradiated  with
ultraviolet light., The amount of ozone  produced is  a measure of the photo-
chemical reactivity potential  of the  precursor  blend.

     The ozone precursor monitor  (0PM)  is  designed for  urban  air  sampling
stations where analyzers  for  the principal air  pollutants  are  routinely
operated.   The monitor,  however,  has other uses.   Among  special  applica-
tions of  the   method  that  are  discussed are  its  use as an early warning
device for forecasting elevated ozone  concentrations,  a  screening  system
for assessing  the photoreactivity of  solvents,  and a procedure  for  investi-
gating the transport  of ozone  precursors from urban to rural  areas.
                                INTRODUCTION

     Ozone in the earth's  atmosphere  is  a paradox.   We want it and we don't
want it.   We want ozone  in the stratosphere,  where it forms  a protective
layer  that  encircles the  earth  and  screens out harmful ultraviolet  radia-
tion that  causes  skin cancer.   While the  depletion of this  stratospheric
ozone  layer  is  of  vital  concern, the subject  of  this paper is  the harmful
ozone  in the ambient air we breathe.  When the hourly average  concentration

                                    208

-------
 of  ozone reaches or  exceeds the National Ambient Air  Quality Standard  of
 0.12  ppm (120 ppb),  our health and welfare are imperiled  (1).

      Non-methane organic compounds and  nitrogen  oxides  are known to be  the
 precursors  of ozone that is produced  photochemically in the ambient atmos-
 phere (2,3).   In view of the pollution of the atmosphere  by organic  com-
 pounds  and  their role in the formation of ozone, it is  clear that knowledge
 of  their total concentration using non-methane organic  carbon  analyzers  (4)
 is  important.  However, knowledge  of  the total non-methane organic concen-
 tration alone makes photochemical  air pollution  modeling difficult at  best
 and poses a severe burden on  the ozone  control  strategist,  since there  is
 an  endless  possible  mix of these  organic gases  that  may number  into  the
 hundreds and  since  these gases  vary  in  their  photoreactivity  by orders  of
 magnitude.  Even with  in-depth  knowledge of the  species  of  organics  pres-
 ent,  gained by use  of state-of-the-art chromatography,  and their photoreac-
 tivity  level,  the assessment of  the ozone-forming  potential  of an air mass
 is  laborious.  The  assessment  is  also  fraught with  uncertainties  such  as
 unknown synergistic  and opposing  reactions.    The  task  is  even further
 complicated by the  fact that, in the  absence  of nitric oxide and nitrogen
 dioxide,  the  solar-  irradiated organic  vapors will  not  form  ozone  and
 because,  in  the  presence of  these nitrogen  oxides, the amount  of   ozone
 formed   and  the time  for   its  formation  will  vary  depending  on   the
 concentration and species of nitrogen oxides present.

      As early as 1956, researchers in California made an effort to circum-
 vent  the many problems associated with  estimating the  photochemical  smog
 formation potential  of the  ambient  air in  the  Los  Angeles Basin (5).   They
 developed an   empirical approach  to  the  problem  embodied  in  an pxidant-
 oxidant precusor analyzer.   The  dominant oxidant  of  interest  was  ozone.
 The basic idea  of  the analyzer  was  to   expose a flow  of  sample air  in a
 reaction  vessel   to   intense   ultraviolet   irradiation    to   generate
 photochemical oxidants (primarily ozone) from their precursors.  An oxidant
 analyzer,  coupled to  the reaction vessel, would  alternately  measure  the
 oxidant  in  the sample  air entering  and  exiting  the chamber.   The difference
 between the two  concentrations would  be  a measure of  the oxidant precursors
 in  the  air   sample.     Conceptually,   the  analytical  approach  of    the
 oxidant-oxidant  precursor analyzer  was  very meritorious, and analyzers  were
 installed in  10  sampling stations.   However, after several  years of  trial,
 the approach  was abandoned.   The bulkiness of  the equipment,  and numerous
 performance and operational problems  that included  lack of  specificity,
 sluggish  response,  frequent  negative  readings  caused by sulfur dioxide and
heat-induced  ozone  losses,  failure  of  mechanical components,  etc.,  was
 cause for shelving of  the analyzers.

     In recent years  substantial gains  have been  made  in the reliability
and performance  of non-methane organic analyzers  (6,7,8).   The promulgation
 in  the  Federal Register of  the  reference  and  equivalency program in  1975
 (9) precipitated  the  widespread  improvement of automated  analyzers used to
measure  the  principal pollutants,  including  ozone  and  nitrogen  oxides.
Concurrent  with  the  ever-improving data  base  of  the  seventies,  advanced
photochemical  models  were developed that  were  intended  to aid the.control
strategist  in  reducing ozone.  However,  the usefulness  of  these models has

                                    209

-------
been challenged.   The problems  delineated  in the second  paragraph  of this
introduction  still  exist  and  the credibility  of  the  data  base  for  the
reactivities  of organic  compounds has  been questioned.   EPA  researchers
recently  concluded  that  "Wall-contamination problems  raise  serious  ques-
t-ions as  to  the utility of  Teflon-film smog chambers in  determining reac-
tivity  of hydrocarbons. ...   With  very  low-reactivity  hydrocarbon (and
organic  compounds)  (sic),  the  reactivity  parameters  obtained  from  these
smog chambers are  of  questionable value.   Plastic  smog  chambers  are  not
reliable  for  use in multiday irradiations when  low levels  of pollutants are
present"  (10).

     Two  documents  are especially  invaluable  to  the  researcher  seeking
in-depth  knowledge of  the many  subjects  related to ozone  (11,12).   Unless
otherwise referenced,  these documents were the  source of  archived informa-
tion contained  in  this  paper.

     In October of 1979 the research  project reported here was funded.  The
research  proposal  outlined  the  development program  for a  continuous ana-
lyzer  that  would  measure  the  potential  photochemical reactivity  of  the
atmosphere  in  terms  of ozone  equivalent  units  (13).   The  design  of  the
analyzer  would  be  a modification  of the oxidant-oxidant precursor analyzer
of  the  late  1950s that would  incorporate  state-of-the-art  components  and
advanced  technology  to  eliminate  the  shortcoming  of  the  instrumental
approach  of the fifties.

     The  initial experimental approach focused  on optimization of a contin-
uous flow-through  system as used with the earlier oxidant-oxidant precursor
analyzer.   One of  the  configurations  (shared-time) used  a single  ozone
analyzer.   Another configuration  (dual channel) used two  analyzers.   How-
ever,  continuous   flow-through  methods  had several  limitations; the most
severe  was  that reproducibility and   sensitivity  were   dependent  on sample
flow rate—i.e.,  the ozone  yield  was  a function  of  residence time  in the
chamber.   A  primary  goal  of  the research  undertaking was  to  develop  a
standard  system that would  be  uniform in  construction  and operation.   It
was also  highly desirable that  any EPA-designated ozone  analyzer (14) could
be  incorporated as  a detector  for  the 0PM for  the widest  appeal.   The
flow-through  system  could   not  satisfy  this  desired   feature,   since  the
EPA-approved  analyzers  differ in sample flow rates by a factor of four.

     It was  recognition of  these  problems  that  led to  the  "discrete sam-
pling"  approach, a concept   similar to the  batch-type analysis  of an auto-
mated chromatograph.  The  discrete sampling approach renders  the perform-
ance of the  ozone  precursor  monitor  independent of sample  flow  rate; thus
one  design  can be universally  used  with  any  of  the dozen  EPA-designated
ozone  analyzers.    Moreover,  it  was  found  that the   ozone  yield  of  the
discrete  sampling   approach for various  precursor test mixtures was  more
than 50 percent greater  than  the ozone yield using continuous sampling.
The  discrete  sampling  approach also  was  found  to  have  additional opera-
tional  and  performance advantages that greatly  enhance the  appeal  of  the
0PM for a variety  of field  and  laboratory applications.

     The  discrete  sampling  approach   breakthrough  was  made in the  last

                                    210

-------
 quarter of the allotted  time  for the development of  the  0PM and all of  the
 anticipatory research  that  the  author would have liked to have accomplished
 with the system was  not possible.   However, sufficient  data were gathered
 using the discrete sampling approach to  adequately demonstrate that the  0PM
 does provide information on the presence and reactivity of ozone precursors
 in the atmosphere that has been hitherto unavailable and that the 0PM could
 be a useful tool in the  research laboratory as  well as in the field.

      The purpose  of this  paper  is  to  describe the  0PM research  and   the
 evaluation of the method.
                          DESCRIPTION  OF  INSTRUMENT

      The 0PM consists of two subsystems  (Figure 1).
                               0.25 in TEFLON FEP TUBING

                              ' AC WIRING
                                              RECORDER
                                                     3
II INLET OZONE
| || ANALYZER
i -" — ^:
1 PARTICULATE FILTERS [




-x-
ION VESSEL
<
EC
NC
-

JY-k
l| IRRADIATION
CHAMBER
PUH
b
\


	
J COM
NC/C>
V Tk
iTION VESSEL
S
2
L_r
/IP
o







PROGRAMMER
115 VAC OUTPUT
PRECURSOR PACKAGE
Figure 1.  Schematic  diagram of ozone precursor monitor  (0PM).
     The  first  subsystem,  the  "precursor  package,"  has  two sample  inlets
with in-line particulate  filters.   One of  the  lines  connects to a  reaction
vessel  that  is  contained  in  the  irradiation chamber.   The  outlet port  of
the  reaction vessel connects  to  control  valves  that  provide  for  timed
irradiation  of  the sample  in  the  reaction  vessel,  subsequent sampling  and

                                    211'

-------
analysis of  the  reaction vessel  content  for its ozone  level,  its purging,
and the introduction  of  a new sample into it.   The  other  sample inlet line
connects to  an integration vessel  in line with  the  sample/analysis valve.
An auxiliary pump and an electronic programmer that  controls the valves and
the auxiliary  pump  complete  the  system.

     FEP Teflon  1/4"  OD  tubing and  TFE  Teflon fittings were used for inter-
connecting components,  except for ground glass  ball-and-socket  fittings at
the inlet  and outlet  ports  of  the  glass vessels.   Ground  glass fittings
were also used to  couple the  inlet  port  of  the particulate  filters  to the
glass manifold sampling  systems  used in air  monitoring stations or to glass
calibration  gas  mixture delivery systems.   The assembled  system was leak-
tested by  capping  the inlet  ports  and  applying  a  vacuum with  a sensitive
flow meter in  line  to the  pump.

     The second  subsystem  is composed  of  a  chart   recorder and  an ozone
analyzer with  an integral sampling  pump.   A listing  of  the EPA-designated
ozone  analyzers  (14)  usable  for OPMs  and  a listing  of  suppliers  of  all
other 0PM components  is  available from  the  author.

     A length  of flexible  1/4" OD Teflon tubing connects the two subsystems
to make an operable 0PM.
                            INSTRUMENT OPERATION

     The electronic  programmer  is  the heart of the 0PM.  It precisely times
the  sequence  of  events  that  occur  during  each  continuously  repetitive
15-minute cycle.

     At  the  beginning of the cycle,  or time  zero,  with solenoid  valve #2
closed,  solenoid  valve #1 (Figure 1)  is  energized and directs  the flow of
the sample from the  reaction vessel  to the ozone analyzer.   For the next 2
minutes,  the ozone  concentration  of  the  irradiated  sample  is  assayed and
recorded.  Then,  at  2  minutes into the cycle,  three things occur simultane-
ously:   one—solenoid  valve #1  is de-energized and the sample  flow to the
analyzer  is  drawn (via  the  integral  sampling  pump  in the  ozone analyzer)
from the integration vessel  for analysis  and recording of  the ozone concen-
tration  of the non-irradiated ambient air,  continuing  for  13 minutes until
the beginning  of  the next cycle;  two—solenoid valve  #2 is energized to its
open position; and  three—the  purge  pump  is  powered.   The pump rapidly
draws  out  the remaining  residual irradiated  air  from the  reaction vessel
and fills  it with a fresh  parcel of ambient  air.   The time  required for
purge is 3 minutes,  so at 5  minutes  into  the cycle,  the purge pump is stop-
ped and  valve  #2  is  closed.  For  the next 10 minutes  the  newly introduced
air parcel,  with  its ozone  precursor  loading,  is  irradiated with ultravio-
let energy.  The  total ozone concentration of  this new air parcel, includ-
ing the  ozone generated  by irradiation  of the  ozone  precursors,  is  then
again  measured and  recorded for  2  minutes  at the beginning  of  the  next
cycle.   A  single continuous  recorder  trace  shows  the elevated  2-minute
total  ozone  concentration  that  includes   the  precursor ozone yield alter-
nately  with  the  13-minute  ambient   ozone  analysis  during  each 15-minute

                                     212

-------
cycle.

     The  details  for inputting  the  program into the  0PM's  electronic pro-
grammer are available  from the author.
                           INSTRUMENT CONSTRUCTION

     Two  identical  OPMs were constructed using mostly  off-the-shelf compo-
nents.   The reaction  and integration  vessels  were made  to specifications
from  standard  glass tubing  and fittings.   The irradiation chamber,  which
required  sheet  metal work and installation  of  the electrical  sockets and
ballasts  for the  8  fluorescent lamps,  was  fabricated in-house.   A drawing
of  the  irradiation  chamber  and reaction vessel is  seen in Figure  2.   The
design of the irradiation chamber  was suggested in a report written in 1959
(15).

Ventilation of Irradiation Chamber

     The  reaction vessel  is  cooled  by  an  induced  flow of room air  as  it
convects  upward   through  the chamber from  a  circular  4-inch-wide  opening
above the bottom  support  frame  seen in Figure 2.   The air carries  the heat
from the  lamps upward and out  the  open top.   The temperature of the exiting
air stream is less  than 15°C above ambient.

Portability

     The  ease  of  disassembly of the 0PM into  its  component parts  with  no
need  for tools  is  facilitated  through  the  use  of glass  ball-and-socket
clip-on  pneumatic connections  and  plug/receptacle  power  connections.   The
irradiation  chamber with reaction vessel weighs   70  Ibs,  and the  total
weight  of the  other components,   inclusive  of  the  analyzer and  recorder
(about 70 Ibs), is  less  than 100 Ibs.

Precursor Package

     Space and power  requirements, maintenance  and cost -  The  5-foot-high
irradiation  chamber takes up  slightly  more  than  a square foot of  floor
space.   Depending on layout,  the   space  requirements  for  the  other compo-
nents is 2 to 4 square  feet.   The  lamps draw 320 watts.   With  fixed compo-
nents, coupled with the  Teflon tubing  and the  glass ball-and-socket  fit-
tings, there  is   built-in flexibility,  adjustability,  and  chemical inert-
ness.  Because the  inlet  filters  keep  the  inside  surfaces  clean,  mainten-
ance-free  operation of  the  0PM's precursor  package  can  be expected  for
periods  of  six   months,   except  for  the biweekly change  of  the  Teflon
filters.   The  cost  of  all  materials  and   in-house  fabrication  (labor),
exclusive of the  ozone  analyzer, was  less than $2,000  for each  unit.

Integration Vessel

     The function of the  integration vessel is to  "average" the rapid data
fluctuations normally  observed with  EPA-designated ozone  analyzers,  which

                                    213

-------
                    B - B
                              TOP BRACKET
T*™""^
LU
CO
ID
1—
UJ
O
52
cc
O
^
u.
8
2

XT
•
]'











X

                  GASKET
o
&
                    CC(/)
                                                                      2-PL\
                                                             CYLINDER (REFLECTORIZED
                                                                ALUM. INSIDE, GALV.
                                                                SHT. MIL. OUTSIDE)
                                                       A-A
                                                                   F40BL FLOURESCENT
                                                                      TUBES (8 EA)
                                                               3 in I.D. *5\n O.D.
                                                             (GASKET ON BOTTOM)
                                                     TOP BRACKET
                                                   (FABRICATE FROM
                                                      SHT. MTL)
                                           3 in I.D. * 5 in O.D.
                                           (GASKET ON TOP)
                                       ON/OFF
                                       SWITCH
ON/OFF
SWITCH
                                        BOTTOM FRAME
                                         AND BRACKET
                                     (FABRICATE FROM GALV.
                                     SHT. MTL & ANGLE IRON)
       *3""	ADJUSTABLE FEET'


   Figure 2.   Irradiation chamber  with reaction vessel.


                                         214
                                                   CYLINDER STOP
                                                     (4 PLACES)
                                                                    BALLASTS (4 EA)

-------
 have response  times  typically  of 30  seconds or  less,  while  the  Federal
 Register allows  for  ozone  analyzers  to have a rise  and fall  time of  TIT
 minutes  and lag time of 20 minutes  (16).   An integration vessel was  deemed
 essential for the 0PM based on  the  evidence for typical ambient data,  seen
 in Figure 3.   The precursor signal  is  easily masked when there is  no  inte-
 gration  vessel in  the  sample line.   The author  recommends  an integration
 vessel volume to  be no more than necessary  to give  good  signal averaging  to
 facilitate data  reduction.   The 16.8-liter  vessel  used in this investiga-
 tion is  considered the upper  limit  of what  would  be a  satisfactory  volume
 using the Dasibi analyzer.   If ozone  analyzers  with  lower  flow rates are
 used,  the integration volume should be scaled down  proportionately.
                              IRRADIATION LAMPS
Lamp  Selection
      A wide  variety of lamp types from several manufacturers were evaluated
for  possible use  in the 0PM.  These included  fluorescent  lamps  with peak
spectral  distributions from 300  nanometers (SunLamp,  a Westinghouse prod-
uct)  to 367  nanometers (GE  F40BL) as well as mercury vapor and xenon  lamps.
The  reasons  for  final selection of the GE F40BL include:  a reasonably good
fit  with  the sun's spectral output (17) as  shown  in Figure 4 (18), consis-
tent  spectral distribution  with lamp aging, relatively long life at reason-
ably  low  light  output  decay  rate, and  good  overall  quality with long
expected  mean time before failure (MTBF).

Lamp  Output  Decay

      All  fluorescent  lamps show  appreciable  decay—40  to  50  percent—in
light output over  their  useful life.  The  life  expectancy of the 0PM lamps
used  is greater  than  20,000 hours (19).   Approximately 50 percent  of  the
total loss  of light  output occurs  in  the first  1000  hours, with 5  to  10
percent of the total output lost  in the first  100  hours.   Based on several
sources  of  information  for the  type of  lamps used  (20,21, and  personal
communication with Elton T.  Lappelmeir,  Lighting Business  Group,  General
Electric  Company,  April 15,  1981) and  empirical  tests,  the  author  recom-
mends  that the lamps be burned in  for  1000 hours  prior to 0PM  usage.   The
lamps  should then  be used for  a year (about 9000 hours) and discarded.  The
expected  percent decay during  'this  time is  about  12 percent  (Figure  5)  or
about  1 percent  per month.   By removing the lamps at one  year,  the  proba-
bility  of any lamp failing  is  about 3 percent  (Figure  6)  (21).   Lamps used
for the OPMs were  numbered  and a  record  maintained of their in-service time
(hours ON).

Temperature  Effect on  Lamps

     Light output  varies with  ambient temperature  for  all lamps,  including
the  fluorescent  type.   Exact  temperature  effect  information  for the  GE
F40BL  could  not  be  located.    In general,  however, the  F40  series  has  a
maximum output  at  20°C  and the  output energy decrease,  with  temperature
increase  up  to 40°C, is  about  10 percent  or 0.5 percent  per  degree  centi-

                                    215

-------
            12 pm
            11 pm -
            10 pm -
             9 pm -
             8 prn -
             7 pm -
             6 pm -
             5 pm
                        \   i   i    i   i    i    i
                                                      1,2,3,4 - Ozone and
                                                      ozone precursor concentration
                                                      for 9 to 10 pm
                                                     Integration vessel installed
                                                      Integration vessel removed
                           T   I   I
                                  10
20
                     7-12-81             One Division = 5 ppb Ozone

Figure  3.   Function of integration vessel.


                                       216

-------
                  100
                c
                05
                o

                CD
                Q.
               LU
               LU
               DC
                           320    340    360    380

                                WAVELENGTH, nm
400
Figure 4.   GE  F40BL lamp spectral output.



                                  217

-------
i—iti     i—i—i—i—r
                                     00
                                     CO
                                     IT)
                                         LU
                                         C3
                                         cc
                                         ID
                                         CO
                                      00
                                      cs
                           o
                           cs
                  ADH3N3
                                               0)
                                               o


                                               n)
60
i-t

0)
d
a)
                                               8
                                               OJ
                                               a)
                                               LT>



                                               0)
                  218

-------
                                                   If)
                                                   CM
                                                   O
                                                   O
                                                   if)
 c
 o
 o

 0>
 Q.

UJ
                                                   o
                                                   in
                                                   LO
                                                   CM
           +J
           •H
           iH
           nJ
           4-1
           M

           §


           I1

           43
10
                                CM
           0
           <0
           o
           CO
            'saOAIAdnS
                                                                   o
                                                                   VO

                                                                   0)
                                                                   )-l
                                                                   s
                                                                   60
                                                                   •H
                       219

-------
grade.  The  energy output decrease  is  slightly steeper as  the  temperature
goes  down from 20°C.   As  mentioned earlier,  the  temperature  inside  the
irradiation  chamber  is at  room temperature  at the  bottom  and  15°C  above
room temperature at the top.  The  lamps  therefore  are operating  at an esti-
mated 95 percent of their peak  output.   As is  true with most  analyzers,  it
is desirable  to  keep   room  temperature  within  a  set  tolerance  for  optimum
performance.  For  the  0PM, ±3°C is  suggested.
                         PREPARATION  FOR OPERATION
Precursor Package
     To  ensure  that the surfaces  making contact  with  the air  sample  were
free of  contamination,  only new materials and  components  (tubing,  particu-
late filter holders, valves,  and fittings)  were used.   All  sample  train
components  were  purged  with helium,  nitrogen,  or  zero  grade  air and condi-
tioned with a high concentration (^50 ppm)  ozone in zero air  stream.   The
glass vessels were  washed with a chromic acid solution followed by thorough
rinsing with distilled  water,  dried  with a heat gun while being purged with
zero grade  airj  and then conditioned by  a flowing stream of  ozone (^50 ppm)
in zero  grade air  for about an hour.  The ozone  in  air  was  stopped and the
vessels  sealed  for 48  hours.  The entire system was tested  by directing a
stream of  about  450 ppb of  ozone in dry air directly  to  an  ozone analyzer
and  then through the system.   A match  of  readings  confirmed  no  losses  of
ozone.

Calibration of Ozone Analyzer

     Four Dasibi ozone  analyzers were used:  two  were  incorporated as  part
of the two  OPMs;  one was a standby and one was  used as a transfer standard.
All  four analyzers were initially calibrated (22)  by the Quality Assurance
staff  of EPA's  Environmental Monitoring Systems Laboratory (EMSL).   The
transfer  analyzer was   periodically  recalibrated by  the EMSL for  use  as a
standard in the  field  (23).
                                EXPERIMENTAL

Data Reduction and Sample Integration

     Reference is  made again to Figure 3,  which shows typical ambient data
at  an  urban site.    The  chart speed is 1  inch per hour.   The 0PM's normal
operating  range is 6-.5  ppm.  This chart  scale  uses  200 divisions - there-
fore,  one chart division equals  2.5  ppb  ozone,  or  5  ppb per  percent  of
scale.   Zero   concentration is at  0.0 percent chart  for convenience  of
illustration.   The Federal Register methodology for air pollution analyzers
places  the recorder  zero at +5  percent  of  scale.    Gaseous  air pollution
concentrations are reported on an hourly average  basis.   "Eyeballing" the
average  ozone  concentration  for the hour  9:00 to  10:00 p.m.  gives 25 ppb.
Averaging  the   four  precursor readings, 9:15  to 10:00 p.m.,  gives  67 ppb.
The average precursor concentration is  therefore  67 minus  25,  or  42 ppb

                                     220

-------
ozone equivalent  units.

Test Gases

     To  carry out a multiplicity  of experiments,  zero grade  air  was pre-
pared from  outside  ambient air by drawing  it  through a particulate filter,
compressing  it,  and passing  it  through an ozonizer,  chemical and physical
scrubbers,  and catalytic  oxidizers, followed  by additional  scrubbers and
particulate  filters.   A 30 liter per minute continuous flow of treated air
was available at 60 psig.   The dry, zero  grade air was  free of  oxides of
nitrogen and had a total hydrocarbon content  of less  than  0.1 ppm carbon.
This air was  used both in its .pure  form for zeroing and as dilutent air for
making  up  of  test  gas  mixtures  using  mass   flow  controllers  to  regulate
delivery rates.

     Lecture  bottles  of  nitrogen  dioxide  and  propylene   equipped  with
specially  designed  permeation  devices  provided  a  source   of  controlled
precursor compounds  for  the  various  test mixtures (24).

Ancillary Analyzers

     Experimentation in the  research laboratory  was  facilitated by support.
equipment that included  analyzers  for  oxides  of nitrogen and organic carbon
as well  as  instruments for measuring wind  direction and velocity,  tempera-
ture, dew point,  and barometric  pressure.

Sampling, Calibration,  Sampling  Precaution

     In shifting  from  a  moisturized  ambient sample  to diluent dry zero air,
it was necessary  to  allow the 0PM to stabilize  to  the condition of dryness
of the sample.   It  has been observed using both photometric and chemilumi-
nescent  ozone analyzers  that  a -period  of  equilibration  to  a dehumidified
sample precedes  a true reflection  of the ozone concentration in the sample.
Similarly,  in reversing the  condition—i.e.,  shifting from dry  to humidi^-
fied air—a response  delay  was  encountered.   Because  of  the area  of the
sample containment  surfaces  of  the  0PM, this  equilibration period was two
or three cycles  long.   It  was  essential during this period that both sample
sources, the purge/precursor  sample and the  ambient  ozone  sample,  were at
the same humidity  level,  to  allow  the  0PM's  ozone  analyzer  to  properly
acclimate.

Lamp Decay Test

     After  289 hours of usage of eight  new GE F40BL lamps in an  0PM cham-
ber, a  propylene/nitrogen dioxide   in  air  mixture  was introduced  and the
ozone yield  for a normal cycle of  operation obtained.   The  used lamps were
replaced with new  lamps,  the same  mixture sampled,  and  the  ozone  yield
measured.   The ozone yield of the used  lamps  was 16 percent  less  than for
the new  lamps.   Examination of the earlier  referenced Energy Maintenance
Curve (Figure 5)  shows  that  the  16 percent  loss in  yield is  consistent with
the decrease  in energy output  for the lamps for the hours-in—use and points
out the need  for  an  initial  burn-in  of  1000 hours  to  achieve  a lower decay

                                     221

-------
rate.

Characterization of Response

     The  ascending nature  of  the precursor  response reflects  the continu-
ance of  irradiation of the  sample  during the 2-minute  readout  and increas-
ing ozone  yield when the end-point for the sample  is not  reached (it may be
hours before  it is).   For purposes  of  data reduction,  the  author chose the
maximum precursor reading.

Ozone Yield

     Figure 7 shows ozone yield  versus irradiation  time.   The  time  of the
analysis followed the time of the  exposure at an interval  of 5  minutes, 10
minutes  (the  normal  operating interval),  15  minutes, and, as  indicated in
the figure, at  increasing intervals up  to  420 minutes  (7 hours).   The test
gas contained 120 ppb N02 and 0.83 ppm propylene  (2.5 ppm carbon)  in air.
The maximum ozone yield, 925  ppb,  was  generated at  about 300 minutes.   It
compares well with the yield  at  10 minutes  of  387 ppb from  the standpoint
that the  latter yield, generated in  one-thirtieth the  time for  the maximum
yield, is  47  percent of the maximum  yield.  Therefore, the 10-minute read-
ing is  a meaningful  index of the  photochemical reaction potential  of the
precursor  mixture.
                        1000


                        900


                        800
                      8 400
                        100


                         0
                                            NOz = 120 ppb
                                            CaHs = 2.5 ppmC
                                 100      200     300

                                   IRRADIATION TIME, min
Figure 7.  Ozone  yield vs.  irradiation time.
                                      222

-------
Repeatability of Ozone Yield

     Table  1  shows  the results of  a 5-hour test  to  establish  the  repeat-
ability  of  the 0PM.   The test gas  contained 0.83 ppm propylene (2.5  ppm
carbon) and 120 ppb nitrogen  dioxide in pure air.  The maximum  reading  for
each of the four precursor readings  for  each hour is  given.
                  TABLE 1.  REPEATABILITY  OF  OZONE  YIELD

ppb 03
Hour 1
Hour 2
Hour 3
Hour 4
Hour 5
Average
389
392
394
390

391
391
393
386
388

390
390
389
387
387

388
392
385
386
388

388
390
386
388
390

388
     The  variation  in  values   in  Table  1  reflects  the  impact  of  all
parameters that  can  affect  the  reading,  including changes  in  line voltage,
precursor blend, barometric  pressure,  and ozone analyzer response  over  the
5-hour test.

Agreement Between OPMs

     Table 2 shows the  results  of  repetitive analysis by the  two  OPMs fol-
lowing  calibration  of  their  Dasibi ozone  analyzers  against  the  transfer
standard.  The test precursor mixture  contained 0.63  ppm  propylene (1.9  ppm
carbon) and 111  ppb  nitrogen dioxide.   The  lamps  in the irradiation  cham-
bers of  the  two OPMs  had been  on  for 1,867  hours.    Readings  were  taken
after the recommended  three  cycles  of operation  on  the dry precursor test
gas.
                     TABLE 2.  AGREEMENT BETWEEN OPMs
             Analysis cycle
                 7/3/81

                   1
                   2
                   3
                   4

             Average	
                      Ozone yield in ppb
                      0PM A        0PM B
                       348
                       350
                       354
                       352

                       351
                       350
                       355
                       355
                       350

                       352.5
                                    223

-------
     The above data  show exceptionally good agreement of  results.   On five
other occasions spanning about  800 hours of on-time  for  the  lamps,  similar
analyses  were made.   The  average ozone yield  readings  for  the two  OPMs
agreed  within 10  ppb for  different  precursor  mixes  that produced  ozone
yields of 170 ppb  to  650 ppb.

Line Voltage Effect  on Ozone Yield

     A  study  was  made of the  effect of line  voltage on ozone yield  for a
test precursor mixture.   The results of the study are seen in Table 3.
               TABLE 3.  LINE VOLTAGE EFFECT ON OZONE YIELD
              Irradiation  chamber
                 line voltage	
Ozone yield in ppb
                       100
                       110
                       117.4
                       120
                       130
         362
         377
         380
         390
         405
     Between  100 and  130 volts  the ozone yield increased 43 ppb, or 11 per-
cent of  the yield  of  the  117.4  normal  line  voltage.   While  the  yield is
obviously  voltage-dependent,  an  examination of  the  ozone  yield over  a
"normal"  voltage range  (110 to  117.4)  shows  only  a  3 ppb  or  1 percent
change,  so the  effect  is  considered  minimal except  in  situations  where
extreme  line  voltage  fluctuations may occur.  Line  voltage stabilizers are
available  for  such  circumstances.

Loss of  Ozone

     Photochemical  researchers  have  noted  a problem  of  loss  of  ozone on
surfaces  such as Tedlar,  Teflon, Mylar,  and other  plastic  films.   Tests
were conducted on  the reaction vessels  used in the  0PM  to determine ozone
loss.   The 16.8-liter reaction  vessels were  cleansed and  conditioned as
noted earlier.  A 573-ppb  concentration of ozone  in diluent air was sampled
directly and  then through the  unused vessels until  the  exiting concentra-
tion was  the  same as  the entering concentration and the vessels capped.  An
analysis  of the content of  the chamber was  performed  117  hours  later; 61
percent  of the initial concentration  remained.   An  analysis  of the second
chamber  at 168  hours  (1  week)  showed 57  percent  of the  ozone present.
After  the vessels  had  been  in use for  four months,  the test was repeated
without  cleansing  the vessels, but conditioning them to  a  500-ppb ozone in
air test mixture.   At 3 hours,  94.5 percent of the initial ozone concentra-
tion was present,  and at 93.5  hours,  56 percent of  the  initial concentra-
tion  remained.   Since  a  full  precursor  cycle   is  only  15  minutes,  the
results  indicate the  0PM's performance is not impaired by  ozone loss.  The
cleanliness of the  vessels during use was maintained by rigorous use of the

                                     224

-------
 particulate  filters  on  the  sample  inlets.    The  filters  were  changed
 biweekly,   after  ^300 hours  use.   For  areas  of  heavy  pollution,  weekly
 replacement of the filters would be necessary.

 Efficiency of Purge

      The necessity  to remove  the remnants  of  the  previous discrete  sample
 during a purge cycle  is obvious.   To  test the adequacy  of- the purge  time
 and  flow rate, a  stream of 450-ppb ozone was generated in one  of two  sample
 streams.   The second stream contained zero air.   The system was allowed  to
 acclimate  for three cycles.  Then,  in  alternately sampling the two streams
 cycle-to-cycle, the removal of ozone from  the reaction vessel was found  to
 be  greater  than   99 percent.    The introduction  of  the  ozone  into  the
 reaction vessel was also found to be greater than 99 percent.

 Other Experiments

      Now that the reader has  a familiarity with the experimental procedures
 followed,  capsulation of three critical but straightforward experiments  can
 be presented:

 1.    No  ozone  was  generated  in  the   chamber  in  irradiating zero air,
      even  when left in the chamber for  several hours.

 2.    Ozone  was generated  in  sampling  nitrogen dioxide  in air  only.   A
      trace  of  NOa  would  produce  a  perceptible  ozone  yield.   This   is
      consistent  with  the  photochemistry  of  nitrogen  dioxide and   occurs
      in  nature.

 3.    No  ozone was  generated  in  sampling  prppylene in  air  only.    This
      again  is  consistent with  the photochemistry of hydrocarbons.


                             PERFORMANCE

 Ozone Precursors  and Ozone in  a  Small City

      Figure  8 shows  one day  of  data (noon  to noon)  collected at  an air
 sampling site  on  the fringe of downtown Durham, NC.   The recording reflects
 the  typical ozone-ozone  precursor profile  commonly seen  during  the months
 of spring  and summer  at  this  location.    Attention is  drawn to  the late
 evening  precursor readings usually  observed at  the site  and  to  the sup-
 pressed ozone  readings that sometimes last  throughout the night.

      The second trace  seen on the dual pen recording is  the output  from a
 non-methane  organic carbon analyzer.   The  analyzer,  calibrated with pro-
 pane, was operating  on the standard  0 to  10 ppm carbon range.   Zero was set
at 50 percent chart.    For 18 of  the  24  hours,  there  is no  perceptible
organic  concentration.   There is  at all times,  however,  some  evidence  of
ozone  precursors,  thus  demonstrating  the  relative  ineffectiveness of  the
non-methane organic  carbon analyzer  for predicting  ozone potential.
                                     225

-------
                   11am
                                              \
                          Ozone and
                          Ozone Precursor
                          Concentration on
                          a ppb Ozone Scale
                                                     5-25-81

                                                     5-26-81
                                                   r-   04
                                                           CO
Non Methane
Hydrocarbons on a
ppm Carbon Scale
Figure 8.   Ozone precursors and  ozone in a small city.

                                        226

-------
Ozone Precursors and Ozone  in a  Large  City

     A month-long  field study in  a major metropolitan area in  the south-
eastern  United States  was  made during  the  summer of  1981.   An  0PM was
located on the campus  of  a  university  near the downtown  area.   Figure 9 is
a  reproduction of  a  single day's  run  (midnight  to  midnight)  that  shows
several  phenonema.     In   the   early  hours   of   the   morning,  the  ozone
concentration  was  suppressed;   nitric  oxide  was  scavenging  the  ozone.
Again,  as  in the  small  city,   this  usually was seen during  the  late
evening/early morning  hours.

     During  the  hours  between  3:00 and 5:00  p.m. DST,  the-  hourly  ozone
concentration  was  170  ppb,  50  ppb above  the standard.   Between 9:00 and
10:00 p.m., the ozone  precursor  hourly average was 99  ppb.   At 10:30 p.m.,
sudden winds  that  preceded  a rainstorm  and  a drop in  temperature  carried
away the stagnated  pollutants, and  a noctural ozone concentration of 48 ppb
was recorded with passage of the storm front.

     The total 0PM  down-time at  the site was  10 hours.   Power outage caused
4 hours of lost data,  and  auditing of  the ozone  analyzer required 2 hours.
Another 2 hours were  lost due to operator error.   More  than 98  percent of
both the ozone and  the  ozone precursor data were  usable.

Analytical Agreement of OPMs on  Ambient  Air

     Figure  10  has the tracings of two OPMs operating  in  tandem  off the
same ambient sample manifold.  The  zero  for the output  of the 0PM recording
on the left  was  0 percent  chart.   For  the 0PM on the right,  the  zero was
set at  50 percent  chart.    Although  the OPMs' programs  were  synchronized
with purge and sample  operations occurring at  the  same time,  the presenta-
tion on the chart shows the  tracing on the right  about  1/4 inch late due to
the offset of the two  pens  of the  recorder.

Sensitivity of 0PM

     The sensitivity  of the 0PM can  best be appreciated by  comparing its
response to the response of  continuous air pollution monitoring instruments
conventionally used for measuring  photochemically reactive compounds in the
atmosphere.  These  instruments  are nitrogen  oxide  analyzers  operating on a
0-.5 ppm range and  non-methane organic carbon analyzers operating on a 0-10
ppm range.  On a  number of  occasions, totaling more than 100  hours during
several months of sampling  ambient  air with these type  analyzers, no nitro-
gen oxides or non-methane organics  were  detected;  however, ozone precursors
measured by  the  0PM during  these  same times   ranged from 0  ppb  to  25 ppb.
On two occasions,  the  precursor  value (ozone  yield) was  0 ppb.   These two
incidents may have  been related  to the  passage of  fronts and ozone-bearing
stratospheric air movement  into  the troposphere  (stratospheric intrusion).
This hypothesis is  supported by  the fact that the  air  did contain 30  to 40
ppb of ozone, as might  be expected  with  such  an occurrence.
                                     227

-------
                     11 pm —
i    l    l
  Sudden pronounced winds
                                                  Sunny and hot
                                                      7-24-81
Figure  9.   Ozone precursors and ozone  in a large city.


                                       228

-------
12 pm-
11 pm-
10pm -
 9 pm-
 8 pm-
 7 pm-
 6 pm -
 5 pm
         I  I  I  I  i  i  i  I
                                          I  I  i  I  i  i  I  I
                                                  i  i  i  i I  I  l I
I  I  I  |  I  I  I  |

      o       o
      if)       O


Ozone and Ozone Precursor Concentration on a ppb Ozone Scale


  O.P.M. "A"              5/7/81            O.P.M. "B"
Figure 10.  Analytical agreement of 0PM's on ambient air.




                                  229

-------
Performance Specifications

     Data in  the  listing  below are based on the use  of  the  Dasibi  analyzer
(25) and  reflects the measurement  of  ozone in  the  ambient  air  on  leaving
the  integration vessel  and the  measurement  of  the ozone  yield from  the
0PM1s irradiation  chamber.

     Range:   0.000 to  0.500 or 1.000 ppm
     Incremental Sensitivity:   0.001 ppm
     Flow Rate:   2 liters per  minute nominal
     Zero Drift:   digital display 0.1  percent/day
     Span Drift:   digital display 0.1  percent/day
     Digital  Display:  0.000 to 1.000  ppm
     Rise Time:   25 sec.
     Fall Time:   25 sec.
     Noise:   ±0.002 ppm
     Interference  Sensitivity:   1 percent of  full  scale  or less

     In discussing irradiation chamber  lamp  decay,  it  was  stated that  a 1
percent per  month loss  of   energy  output  is  to be  expected.    Assuming  a
nominal linear  relationship between ozone yield and  light energy output  for
the  0PM,  the author  proposes  the  normalization  of  data,  collected  each
month following the  installation of conditioned lamps (1000 hour burn-in),
to the initial  ozone yield.  The  correction would  be 1.00 plus  .01  for each
month, times  the  recorded ozone  yield concentration.  It is believed that
this simple  procedure will  provide  an estimated ozone yield tolerance  for
the 1-year operational life of  the  lamps of ±2  percent.
                                 DISCUSSION

Maximum Precursor Concentrations

     An hourly  average  precursor reading of  187  ppb ozone was  recorded  at
Durham, NC between  10:00  and 11:00 p.m. during a period  of  air  stagnation.
Typically, the  highest  precursor readings  have  been observed between 8:00
p.m.  and  2:00  a.m.  at  the Durham Air Monitoring Demonstration  Facility
(DAMDF).

Sampling Location of  OPMs

     In general, selection of  the monitoring sites  for the 0PM1s  dual role
of measuring ozone  and ozone precursors should  be  consistent  with published
guidelines for  such an  application (26).   Briefly,  the site should  not  be
in an  area subject  to the impact of  large  sources   of emissions.   However,
there  could  well be  special cases  where an  investigator is interested  in
localized,  concentrated  pollution  sources.    In   such  cases,   additional
instrumentation  that would  include analyzers  for   oxides  of nitrogen  and
organic compounds  would be  desirable  for  a more complete  characterization
of the air mass.   Special siting would be  in order for operating  OPMs  for
forecasting ozone episode  alerts and  for transport  studies  of ozone precur-
sors in rural areas.
                                    230

-------
 Precaution

      All earlier referenced  (14)  EPA-designated acceptable chemiluminescent
 analyzers (except the Philips)  require  an ethylene support gas.  Should one
 of these  chemiluminescent analyzers be  used,  precautions must  be  taken to
 ensure that the gas system is free  of leaks.   The analyzer's exhaust gas, a
 mixture of  ethylene  and  sample  air,  should be  discharged  through  a cata-
 lytic  oxidizer,  reducing the  ethylene  to  carbon  dioxide  and water,  to
 prevent it  from  being  "recycled" into   the  air being  sampled  by  the  0PM.
 These measures are consistent with  published safety precautions  for use of
 ethylene.   They are highlighted here because ethylene  is  a reactive organic
 that, as part of the sampled air, would be an ozone precursor.   The Dasibi
 and Thermo  Electron  instruments  are  photometric  analyzers  and  require  no
 support gases.  The Philips, while  being a chemiluminescent  method,  uses an
 ozone-sensitive Rhodamine B dye and thus needs  no  gas.

 Los Angeles Basin Study

      None  of  the few  areas where  the  OPMs were  operated  are noted  for
 having an  ozone problem.  The author had included in his  research  proposal
 a   three-month  study  in the South  Coast  Air  Quality  Management  District
 (SCAQMD),  which includes  the  Los Angeles  Basin.   The  inhabitants   of  that
 area  suffer from the most severe  ozone   pollution  problem in the nation  nd
 the  location  had  been  considered  the  ideal  setting  for siting OPMs.
 Budgetary  constraints,   however, precluded the   opportunity to research  the
 0PM in the Basin.
                                APPLICATIONS
 Screening  of  Solvents
      Some  200 industrial solvents  are unclassified as  to  their photochem-
ical  reactivity.   It has  been suggested  that,  using  an 0PM  and  an appro-
priate  protocol,  these  solvents can be  rapidly  classified  for their ozone-
forming  potential.   An  initial screening  would  tell  whether  a  specific
solvent  (blended  with  nitrogen dioxide) fell  in  one of   four  or  five
classifications   from   nonreactive   to   very   reactive.      Differential
information  would  then  be  obtained   by adjusting  the  irradiation  time
(sensitivity)  for each class.   The procedure would have incorporated into
it  a zero  air purge/analysis  cycle  to  ensure  that  no  residual  solvent
remained, in  progressing  from one to  the next solvent.   The procedure would
provide  for uniform  treatment of the  samples  and minimization of artifacts
encountered   in   use  of   traditional   bag   sample-in-irradiation-chamber
procedures.

Special Air Pollution Surveys

     For  special   surveys,  programmed  methods are  available  for  simulta-
neously  collecting integrated  ambient  samples  in  plastic bags  for  pre-
scribed  time  intervals  at multiple  sites (27).   These  bags,  contained  in
light shielded black bags,  are  then  returned to  a centralized  laboratory

                                     231

-------
for an  operator-managed ozone  precursor assay  using a  single  0PM.   This
technique  has  been  successfully used  for  collecting hydrocarbon  samples
from multiple sites for subsequent GC analysis  (28).

Ozone Episode Forecasting

     Several of the EPA Regions  in the  past several years have given prior-
ity ranking to a need for a means to forecast ozone episode alerts.   Infor-
mation from strategically sited  OPMs, combined with meteorological informa-
tion, could be the basis for  predicting such episodes.  On the basis of the
prediction, the  public  could be alerted and measures could  be  implemented
to reduce  ozone precursor emissions.  Individuals with respiratory problems
or other ailments aggravated  by high ozone  levels  would  be advised  to stay
indoors.   For known areas where  concentrations exceeding the ozone standard
occur frequently, such as the Los Angeles Basin, OPMs could be particularly
beneficial.

Rural Ozone Precursor Transport  Investigations

     Economic  loss  due  to  damage to crops  and forests by  ozone amounts to
many millions  of dollars annually.   It  is thought that such losses could be
minimized  if  there were a  better understanding  of the  origin  of man-made
precursors of  ozone  transported into sensitive  areas.  Using data obtained
from several OPMs  in a network, coupled with readily available meteorolog-
ical information,  the investigator would be able to determine the source of
the  precursors.    Species  analysis  of   the organics  in  the  advancing air
could  then provide  the  detailed  information  that   could lead  to  needed
controls if the  source  were localized.    If either the abundance of species
making  up  the  precursor mix  or  the  diversity of  their origin were  to show
that  little hope  for  appreciable  improvement  in  the   situation  could be
anticipated,  then the  planting of crops less susceptible  to ozone damage
could  be  the  economical solution.   However,  it  might  be found  that the
organic fingerprint  would identify a source of  emissions  that was economi-
cally controllable.   In the latter  case, the air quality manager  could take
such action appropriate to  the situation.

Emission Control  Strategy

     Recently  the EPA Administrator told members of Congress  (29) "The more
uncertainty there  is,  the  more difficult  it  is to estimate  the proper
balance between  public  safety on the one hand, and the often  enormous  costs
of  pollution  control on the  other.  We simply can not  afford to err  badly
in  either direction.    Better  knowledge  of  the  scientific  and technical
facts almost  always  reduces the likelihood  of such errors."

     In a number  of  metropolitan areas, particularly in Southern Califor-
nia,  the  ongoing  question in  ozone  abatement  strategy  debate is, "should
multimillion   dollar  expenditures  be   made  on  source  control  of  organic
compounds  or nitrogen  oxides?"  (30).    With the development  of the 0PM,  a
research plan to resolve the above dilemma is feasible.    The study,  in its
simplest form, would involve triplicate  collection of integrated  samples  in
'v/lOO-liter Teflon bags, shielded from  sunlight, over  a fixed  time interval.

                                     232

-------
 The  untreated content of  bag #1 would  be  sampled by an 0PM for Its  ozone
 yield.   The contents of bag  #2  would be sampled  with added nitrogen  diox-
 ide;  and bag #3 sampled with  an added organic.    The  ozone yields would  be
 compared.   The procedure would  be  performed  as  often and  at such times  as
 necessary  to give  statistical meaning  to  the results  obtained.   Concept-
 ually, the  procedure could be embodied  in  an  automated  system.  The stated
 problem, as  well as the instrumented method advanced for  its solution   is
 intriguing.                                                             '
                                  CONCLUSION

      The  principal objective  of  the research  effort  that  resulted  in the
 0PM was to  provide EPA with a means  to  measure,  on a  real-time basis, the
 photochemical  reactivity potential  of   the  atmosphere  in  units  of  ozone.
 The measurement method  described allows  for  the  fact that  there  can exist
 in the atmosphere  an unlimited  number  of combinations  of  ozone precursors
 that differ  in concentration  and photoreactivity.   The method  treats any
 conglomerate of ozone precursors  in  a uniform fashion  and provides an ozone
 value for the mixture sampled.

      With deployment  of OPMs  in geographical  areas  prone to  ozone  pollu-
 tion,  it  is anticipated that a data  base will be  generated that air quality
 managers  will find  of  value.   The ultimate  goal  is enhancement  of control
 strategies and consequent  reduction  of  ozone levels  in the many  areas  of
 our nation that are not in  compliance  with  the  standards  established  for
 protection of the  health and welfare  of  the people.

      In these  times of economic  belt tightening,  the  0PM  should be  per-
 ceived  as  a money-saving device.  With  the coupling of  the  precursor pack-
 age,^ as described  in the text, to an already operating  ozone analyzer,  the
 ability of the air  quality manager  to assess problems and progress related
 to  ozone  precursor  emissions  will be enhanced  with appreciable savings  of
 time and  labor power.

     The  secondary  objective of the  research  effort was  to provide  EPA with
 a tool  for other applications of  varying  significance.   The use of the 0PM
 for  predicting the possibility  of  hazardous  ozone concentrations  is one
 such  application.    Its  use  as  a  method  for  rapidly screening  solvents  is
 another.   The method could  also be used for reassessing  the photoreactivity
 of  a broad spectrum of hydrocarbons  where uncertainties  exist because  of
 procedural  limitations  of  the past.


                              ACKNOWLEDGMEN TS

     The development of  the  0PM was  funded as an  innovative research proj-
ect  by  the  Innovative  Research  Program  of  the  Office  of  Research  and
Development  of the  U.S.  Environmental Protection Agency.   Administrative
support,  facilities,  and equipment   were provided  by  the  Environmental
Sciences  Research  Laboratory  and  the  Environmental  Monitoring  Systems
Laboratory  of  the   Environmental Protection  Agency  located  at  Research

                                    233

-------
Triangle Park, NC.  The author, a member  of  the Regional Services  Staff,  is
indebted to its Director for the opportunity to participate  in  the research
program and  to  numerous fellow employees  and colleagues in the  government
and the private sector  for their generous  assistance.
1.
2.
3.
 4.
 5.
 6.
 7.
 8.
 9.
                           REFERENCES

U.S.  Environmental Protection  Agency.    1980.    Part  50  - National
primary and secondary ambient air quality  standards.  Federal Register
40:523-525 (July 1).

Dimitriades, B.   1972.   Oxidant control strategies.  Part  1.     Urban
oxidant  control  strategy  derived  from existing  smog  chamber  data.
Environmental Science & Technology  ll(l):80-88.

U.S.  Environmental Protection  Agency.   1977.   Uses,  limitations  and
technical  basis of procedures  for  quantifying  relationships  between
photochemical oxidants  and precursors.  EPA-450/2-77-021a.  Office  of
Air Quality Planning and Standards,  Research Triangle Park,  NC.

U.S.  Environmental Protection Agency.   1980.  Guidance  for  collection
of  ambient non-methane organic compound  (NMOC) data  for use  in 1982
ozone SIP development,  and  siting criteria  for  NMOC  and  NOX  moni-
tors.   EPA 450/4-80-011.   Office  of  Air Quality Planning and  Stan-
dards, Research Triangle Park,  NC.

Romanovsky,  J.C.,  J.R.  Taylor,  R.D.  MacPhee,   and  J.E.  Dickinson.
1956.  Air monitoring  of Los  Angeles atmosphere  with automatic  instru-
ments.   Proceedings of the 49th  Annual Meeting of  APCA,  Buffalo,  NY,
May 20-24, 1956.   Air  Pollution Control Assoc.,  Pittsburgh,  PA.

Ortman,  G.C.,   and  V.L.  Thompson.   1974.    Performance  of  hydrocarbon
monitoring instrumentation.   Instrumentation  for  monitoring air qual-
ity,  ASTM STP  555. American  Society for Testing and Materials, Phila-
delphia,  PA.

Sexton,  F.W.,  F.F. McElroy,   R.M.  Mickie,  Jr.,   and  V.L.  Thompson.
1981.  Technical  assistance document for the calibration and operation
of  automated  ambient  non-methane   organic  compound analyzers.    EPA
600/4-81-015.   U.S. Environmental Protection 'Agency, Research Triangle
Park, NC.

Sexton,  F.W.,  F.F. McElroy,   R.M.  Mickie,  Jr.,   and  V.L.  Thompson.
A  comparative  evaluation  of  seven  automatic  ambient  non-methane
organic compound  analyzers.    U.S.  Environmental  Protection  Agency,
Research Triangle Park, NC.   Draft document.
 U.S.  Environmental Protection  Agency.    1981.
 monitoring reference  and  equivalent methods.
 37 (July 1).
Part  53 - Ambient  air
Federal Register  40:4-
                                     234

-------
 10.   Lonneman, W.A.,  J.J.  Bufalini,  R.L.  Kuntz,  and  S.A.  Meeks.   1981.
      Contamination  from  fluorocarbon  films.    Environmental   Science  and
      Technology 15(1):99-103.                     ~	'
 11.  National  Academy of  Science.
     oxidants.  Washington,  DC.
                                      1977.   Ozone and  other photochemical
 12.   U.S.  Environmental Protection Agency.   1978.  Air quality criteria for
      ozone and other photochemical  oxidants.   EPA-600/8-74-004.   Environ-
      mental Criteria and Assessment Office, Research Triangle Park, NC.

 13.   Ortman,  G.C.    1979.   Prognostic ozone analyzer.   Innovative research
      proposal.   Regional Services Staff, Office  of  Research & Development,
      U.S.  Environmental Protection Agency,  Research Triangle Park, NC.

 14.   U.S.  Environmental  Protection  Agency.    1981.   List of  designated
      reference  and  equivalent  methods.    Environmental Monitoring  Systems
      Laboratory, Research Triangle Park,  NC.

 15.   Austin,  R.   circa  1959.     Summary  of  development  work  on  oxidant
      precursor  irradiation chamber.  Unpublished work supported  by  a U.S.
      Public Health Research Grant.

 16.   U.S.  Environmental Protection Agency.   1980.   Part  53 - Ambient  air
      monitoring  reference and equivalent methods, Subpart B-Procedures  for
      testing  performance  characteristics   of  automated methods.    Federal
      Register 40:13-14  (July  1).                                     	"—

 17.   Demerjian,  K.L., K.L. Schere, and  J.T.  Peterson.  1980.   Theoretical
      estimates of  actinic (spherically integrated) flux and  photolytic rate
      constants of  atmospheric species in the  lower atmosphere.   Page 399 in
      J. Pitts and  R. Metcalf, eds.   Advances  in Environmental Science  and
      Technology, Vol. 10.   John Wiley &  Sons,  Inc.,  New York,  NY.

 18.   Leppelmelr,  E.T.    1980.     Spectral  energy  graph-F40BL.     Lighting
      Business Group, General Electric Company, Nela  Park,  Cleveland,  OH.
19.   General Electric  Company.    1980.    Fluorescent  lamps.
     General Electric Company, Nela Park, Cleveland, OH.
                                                                 Form  9200.
20.  Westinghouse Electric Corp.   1976.   Westinghouse "BL" and  "BLB"  black
     light fluorescent  lamps.   Lamp information  bulletin.   Lamp Division
     Bloomfield, NY.

21.  General Electric Company.  1970.  Fluorescent lamps.  Technical Publi-
     cation 111, Nela Park, Cleveland, OH.

22.  Pauer, R., and F.  McElroy.   1979.   Technical  assistance document  for
     the calibration  of ambient ozone  monitors.   EPA-600/4-79-057.   U.S.
     Environmental Protection Agency, Research Triangle Park, NC.
                                    235

-------
23.  McElroy,  F.F.    1979.    Transfer standards  for  the  calibration of
     ambient air monitoring  analyzers of ozone.   EPA-600/4-79-056.   Envi-
     ronmental Monitoring System  Laboratory,  U.S.  Environmental Protection
     Agency, Research Triangle Park, NC.

24.  O'Keeffe, A.E., and G.C.  Ortman.  1966.   Primary standards for trace
     gas analysis.   Analytical Chemistry 38(6);760-763.
25.  Ozone  monitor.   Operating  and Instruction  Manual.
     mental Corp., Glendale, CA»
                                                            Dasibi Environ-
26.  Ludwig, F.L., and E.  Shelar.   1978.   Site selection for monitoring  of
     photochemical  air  pollutants.    EPA-450/3-78-013.     Office  of  Air
     Quality Planning and Standards, Research Triangle Park, NC.

27.  Seila,  R.L.,  W.A.  Lonneman,  and S.A.  Meeks.   1976.    Evaluation  of
     polyvinyl fluoride  as  a container material for air pollution  samples.
     Journal of Environmental Science Engineering,  A 11 (2):121-130.

28.  Seila,  R.L.   1979.   Non-urban hydrocarbon  concentrations  in  ambient
     air  north  of Houston,  Texas.   EPA  600/3-79-010.   U.S. Environmental
     Protection Agency,  Research Triangle Park, NC.
29.
     Gorsuch,  A.M.   1981.    Statement  of  U.S.  Environmental  Protection
     Agency Administrator.   Presented to subcomittee on Natural  Resources,
     Agricultural  Research  and  Environment,   Committee   on  Science   and
     Technolgy, United States House  of Representatives,  October 22,  1981.

30.  Innes, W.B.   1981.  Effect of nitrogen  oxide  emissions  on ozone levels
     in   metropolitan  regions.     Environmental   Science   &   Technology
     15(8):904-912.
                                    236

-------
                     A NEW RELIABLE AMBIENT AIR CHLORINE
                              MONITORING SYSTEM
                               Eric F. Mooney

                                   Anacon
             A Division of High Voltage Engineering Corporation
                               Burlington, MA
                                  ABSTRACT                         •      • *

     This  paper describes a polarographic  probe  technique for the measure-
ment of  low ppm levels of chlorine in ambient air.

     Although t,ie range  of  applications for  such a monitoring  system are
extensive,  in many  of the installations,  the probes  are a  long distance
from the  control room.  It is  therefore  imperative that the "state-of-the-
probe" be  continually  monitored.   The system has  therefore been designed to
continually monitor  that  both  the  probe  and  the electronic   units  are
functioning  correctly  and  warning   is  immediately  given   should  any
malfunction  occur.      Consequently,   an   indication   of   zero   chlorine
concentration is  reliable.

     Additionally, because  the  probes  are often mounted in remote locations
and  a large number  of  probes are  frequently  installed  to  monitor  the
perimeter  of a site, or a particular  work  area where  chlorine  is handled,
the outputs  of  the probes are multiplexed at a local processing unit (LPU).
Data on  the chlorine concentration and probe or  machine  faults  are trans-
mitted from the LPU  as  an RS232 serial data-only  link.  Thus,  the data may
be  transmitted over normal  telephone  communication  lines and, of  course,
are ideally  suited for  connection to  a serial I/O  port  of a smart terminal
for further  processing.


                                INTRODUCTION

     The requirement for  a  good ambient air chlorine monitor  has  long been
recognized,  as  is  evident  from the amount  of  work devoted to  this  problem
by the major  chlorine manufacturing companies.

     A number  of  methods  have  been used, including amperometric  techniques
and solid state  sensors.    Neither of these  techniques   is  specific  for

                                    237

-------
chlorine  and tend to  respond to any  oxidizing gas  such as  ozone,  sulfur
dioxide and  oxides  of  nitrogen, to name  only  a few of  the  commonly  inter-
fering gases.

     There are,  of  course,  a number of colorimetric  methods available, but
none  of  these have  been used  as  a  continuous monitor to  give  immediate
warning.  A.  method  that utilizes the color developed  on a tape impregnated
with  suitable reagents  has also  been used.    This method,  again,  is  not
specific,  since  the same  color  development   is  obtained with a  range  of
oxidants and there is  some  doubt about the stability  of the tapes prior to
use.

     A further complication  arises in the measuring  of  chlorine in paper
pulp mills because  none of  the standard  methods are  capable of differenti-
ating between chlorine and  chlorine dioxide.

     The polarographic method offers many advantages  because the potential
at which the reduction occurs  is  specific for  that compound in the electro-
lyte medium  being used.   Thus,  two  variables,  the  applied potential and the
electrolyte  medium,  are capable of selecting  the  specificity of  the  reduc-
tion.  Obviously, from a practical  point  of  view,  if  reduction can occur at
zero  applied potential,  there  is  considerable  advantage in  not  having to
provide this potential.

     Jim  Young,  at  Imperial Chemical  Industries (ICI)  Limited,  Mond Divi-
sion, England, found  a  suitable  electrolyte   composition  in which chlorine
is selectively reduced at zero applied potential,  and thus  produced  a very
useful probe (information by personal  contact with Jim Young and from ICI
internal reports).   It should be noted that oxygen,  ozone,  sulfur dioxide,
and  oxides  of nitrogen  are all  reduced at  negative  applied potentials,
between -1.5 and -2.5 volts,  in  the  electrolyte  solution  being used and
hence do not interfere with the measurement  of chlorine.

     It  is  this  probe  that  Anacon has  licensed  from ICI  for  world-wide
manufacture  and  sales.
                          THE RELIABILITY OF PROBES

     There  is a general  problem with all  sensors  that are  being used for
ambient monitoring in that these monitors do not receive the same degree of
maintenance  as  other  process   instrumentation.    This is  understandable,
since  ambient air monitors  do  not  result in any  increased productivity of
the plant and are  simply often regarded as a necessary evil.

     Anacon  recognized this problem and  felt it was  most  important to have
built-in  reliability  as well  as  the  specific response   of  the  probe  to
chlorine.  If a sensor is indicating zero, owing to a malfunction, it might
be  said to  be  better not to have  the sensor  at  all  rather  than create a
sense  of false security.

     In the  new  Anacon  ambient  air  chlorine  monitoring  system,   it  is

                                    238

-------
impossible  for the probe  to be reading  zero simply  from a  malfunction of
the probe;  any malfunction is immediately  indicated.
                          CONSTRUCTION OF THE PROBE

     The probe  consists of a bobbin on which is  wound a coil of silver wire
separated  from  the  outer platinum wire  coil  by  a cellulose  insulating
layer.  A  cotton wick is  supported  between the  layers  of  cellulose insula-
tion, and  this  wick projects into a reservoir  that  contains the electrolyte
solution.   The  electrolyte,  unlike  the amperometric methods, is  not con-
sumed, and the  total volume of the reservoir is  only 1.5 ml.

     The two  coils  are connected via two separate wires  to the preamplifier
board which,  to avoid  ingression of moisture, is encapsulated;  the probes
are frequently  mounted outside and exposed to  the  elements.

     The  bobbin, the probe  sheath,  and  the   housing for  the  preamplifier
board  are  constructed   from  polypropylene   to  minimize  any   corrosion
resulting  from exposure  to chlorine  or hydrogen  chloride.   The  probe  is
shown diagrammatically in Figure 1.
                                             PREAMPLIFIER
                                             SILVER COIL

                                             PLATINUM WINDING

                                             CELLULOSE INSULATOR
                                             ELECTROLYTE
                                             RESERVOIR
Figure 1.  Cross-section of probe.
                                    239

-------
                        THE MEASUREMENT  OF  CHLORINE

     On exposure of the probe  to  chlorine,  reduction of the chlorine occurs
at the outer platinum coil and chloride  ions  are formed.   The chloride ions
are transported across  the  electrolyte,  held in  the  porous cellulose insu-
lating layers, and then react  with  the silver coil to form silver chloride;
this reaction is the driving  force  behind the total  reaction.   It is prob-
ably  almost certain  that  chlorine  is  first  absorbed  by  the  wet  outer
electrolyte  layer prior  to  reduction  at  the  platinum  electrode.   Under
steady-state conditions,  if  the chlorine is not  replenished  quickly enough
to maintain the equilibrium,  the  output  will  decrease.

     Under  steady-state  conditions,  the  output  of   the  preamplifier  is
adjusted to give  a voltage  signal  of 4.5  volts  for  full-scale  response of
the probe.  The  normalization of the output  of  the probe  to 4.5  volts for
full-scale  response is  convenient not only for  the, A  to D converter used,
but it additionally permits  probes  of differing  ranges to  be attached to a
single LPU.
                       THE  STATE  OF THE PROBE SIGNAL            .  . • •

     When the coil  is  first made  and before electrolyte is added,  the probe
capacitance is very small,  on  the, order of 50 to 60 pF.   When the probe is
saturated with electrolyte, the capacitance  increases  to  between  40 and 50
mF.   The characteristics   of  the  "cell"  formed between  the  platinum  and
silver windings  and the electrolyte,  form the basis  of  the "state-of-the-
probe signal."

     Should  the  probe dry  out, it  will not respond  to  chlorine,  and  the
capacitance of the  probe will  decrease.  This variation is monitored by the
preamplifier and reveals itself as  a "high" voltage output in the state^of-
the-probe signal train.  Normally,  a probe operating correctly will display
a state-of-the-probe  signal between 60 and 600 mV.
                                 THE SYSTEM

     In  developing  the  system,  considerable  thought  was  given  to  the
requirements  for  ambient monitoring  and  methods of dealing  with the large
amount of  data  that would be  obtained for a  comprehensive  perimeter moni-
toring system.

     The  central  intelligence  of  the system  is the Local  Processing Unit
(LPU), which  is capable  of  accepting up to ten probes.   The  outputs from
all ten probes  are  multiplexed every three seconds, and when data are taken
from each  probe,  both the voltage  from the  chlorine concentration and from
the state-of-the-probe  signals are monitored.

     If  fewer  than ten  probes are  attached,  then the  appropriate  probe
switches  in the  LPU must be  opened; otherwise, an alarm  will  be  given,
                                    240

-------
 indicating that no probe  is  attached to the system.

      Four  conductor  cables  attach  the probe  to  the LPU,  and  the  cable
 length may be 1000 feet or more;  for distances  in  excess of 1000 feet,  the
 gage of  the  wire is  decreased to  avoid any  voltage drop  along the  cable
 length.

      The output from the LPU is an  RS232 ASCII data link.  Fourteen  sets  of
 two-bit codes are given;  these  will be discussed in  detail below.   It  does
 mean that, using this  data communication system, the  LPU may be many miles
 from the  data  acquisition or  from  the  Local Control Unit  (LCU)  with  the
 associated displays.    The data  may  be  transmitted along  a twisted  pair  of
 telephone cables.  This is  useful  for  the  monitoring of  chloride  storage
 tanks that are  normally  unattended, or left  unattended  for long periods.
 Should a leak develop, an alarm may be raised at a remote security  or  fire
 department office through standard  telecommunication procedures,

      Use of the RS232  data  link  has a further  advantage in that  the  data
 may  be put directly  into a  serial  I/O  of  a  smart terminal or main frame
 computer, permitting  time-weighted  averages  to be readily calculated.

      Visual display of  the chlorine levels  may be  observed using the indi-
 vidual , display modules  mounted  in  the LCU.   The chlorine  concentration  is
 displayed as a  bar graph.   The  same  bar graph  liquid crystal  arrays are
 also used to  indicate probe  or instrument errors.

      Both the  LPU and LCU are powered  by 18 to 24V AC or  DC, hence  instal-
 lation' becomes  a relatively  simple  matter.                        '         -

      The  system is illustrated in Figure 2.
           PROBE
                                            LOCAL CONTROL UNIT
                                                (L.C.U.)
                      THERMISTER
                                                 CENTRAL PROCESSING UNIT
                                                   (C.P.U. with V.C.U.)
Figure 2.  Basic  chlorine monitoring system using local control  or  central
           processing  unit.

                                     241

-------
                               THE ERROR SIGNALS

     The error  signals  may conveniently  be  divided  into  two classes,  the
machine errors and  the probe errors.

    The four machine  errors are:

        i.   Loss of  12-volt power in the LPU
       ii.   Loss of  12-volt supply  to  the probe
     iii.   Loss of  5-volt suppty in the LPU
       iv.   Excess current being drawn by a probe

     The ASCII coded  error message and the alarm  error display shown on the
display module are  presented in Figure 3.

                      ERROR MESSAGES
             EA
             12V.
            NO-GO
                       EB
12V.
PROBE
                                EC
                                          ED
          CURRENT
          ALARM
                                                  LPU SYSTEM ERRORS

                                                  PATTERNS ARE SOLID BARS
                                                  INDICATING ON ALL DISPLAYS.
            ssasasa
             NO
            PROBE
                       E1
OPEN
PROBE
                                E2
SHORTED
 PROBE
                                          E3
                                                  PROBE ERRORS

                                                  PATTERNS ARE FLASHING BARS
                                                  INDICATING ON INDIVIDUAL
                                                  DISPLAYS.
    OVERRANGE
 OR ALARM THRESHOLD
(FLASH CURRENT READING).
Figure 3.  LCD alarm graphics on LCU.
      Naturally,  should either the  12V or 5V  supplies  in the  LPU be  lost,
the  displays associated with  all the  probes  attached  to  that LPU  will show
the  same error indication.

      The four alarms associated  with the probe  are:

        i.  No probe
       ii.  Open  probe
      iii.  Snorted probe
       iv.  Overrange - chlorine  alarm level exceeded
                                       242

-------
     Again,  the  ASCII codes  and  visual display  signals are also  shown in
Figure 3.  The overrange  (the chlorine alarm level) is obviously not really
a probe error, but  it  is  convenient to include here.   The  probe errors do,
however, require some  additional  explanation.

     The no-probe signal  appears  if one of  the. display modules is selected
to display a probe  that is  unexpected—for example, the corresponding probe
switch in the LPU is  open since no  probe  is attached.

     The open-probe signal  is  obtained if  the  probe  dries out  and needs
electrolyte  replenishing, or also  appears if a  probe is  removed  from the
connector or if  the cable between the  probe and LPU port is severed.

     The shorted probe signal is obtained if  the  two  coils become shorted
or the cellulose insulation becomes degraded.

     One further error message  is provided and indicates that communication
between the LPU and LCU is  lost and is shown by  the activation of the local
sonalert and the  red  LED "data lost"  indicator becomes  illuminated.   Data
loss could arise from complete  loss of power at  the LPU or when the twisted
pair between the LPU  and  LCU becomes detached or  severed.
                              ASCII  CODED DATA

     Fourteen  sets  of  two-bit ASCII  codes  are obtained  from  the LPU every
three seconds, and  the  line  of  coded data is terminated by a CRLF (Carriage
Return Line Feed).

     The first set  of data  are  the machine  errors EA  to  ED;  the second and
third are not  normally  used,  but  were intended to give wind speed and temp-
erature  if  needed for  correction of data.   The  next  ten sets  of  data are
for probes 1 through 10 and  show  chlorine concentration as 00 through 99, a
percentage of  scale.   In the computer  software,  provision must be  made to
identify the range  of the probe.   As mentioned earlier, the full-scale out-
put for  all  probes  is  normalized to  4.5 volts so that  probes  of different
measurement  ranges  may  be  connected  to the  same  LPU.   The final  set of
information is a  sum check  of the first thirteen sets  of data.

     Should a  probe error be  detected,  the probe  error message E0 to E2 are
obtained for the  appropriate probe  data and of  course  no chlorine concen-
tration  data will be obtained for that  probe.
                              SPEED OF RESPONSE

     As with  many electrochemical devices,  there  is  an asymptotic approach
to the maximum  output  for a given concentration  of  chlorine.  This charac-
teristic arises  because  an equilibrium is involved between the chlorine and
the  rate of  reduction.  A typical response  curve is shown  in Figure 4,  in
which a 5 ppm probe  is exposed to 5.0 ppm of chlorine.   It will be observed
that  the  normally  selected  alarm levels  of 0.5  or  1.0  ppm  are reached

                                    243

-------
                                               u^wWW0VWV^^
                                      TIME IN MINUTES
                                      TIME IN MINUT

Figure 4.  Probe  response to  5 ppm chlorine.
within  less  than  4  to 5  seconds  and  certainly 80  percent  response is
reached within  15  seconds.

     The  equilibrium  effect  may  be  readily  demonstrated  by  inserting a
probe into  a "static" atmosphere  containing 5 ppm of  chlorine.  Initially
the output  of the  probe rises very  quickly, but  then inverts and decreases
as the chlorine is  consumed.

     There  is therefore an airflow  effect  in  the probe,  and air velocities
in excess  of .5 mph  are required  to  maintain the  equilibrium conditions.
Should a probe  be  required  to be used in an enclosed space, for example, a
closed room in  which there is  little or  no  air movement,  then  the probe
should be calibrated  under  static conditions.
                                 CALIBRATION

     It  was  perhaps surprising  that,  during the  initial  field trials,  the
biggest  problem  for users was to substantiate  the calibration of the probe
supplied.  There was considerable difficulty  in obtaining cylinders of  gas
containing a low ppm of  chlorine.   We have found one company* that has been
particularly successful  in providing stable chlorine standards in,cylinders
that  are guaranteed  for  three  months.    Our  experience  has shown  that a
cylinder was stable for  six months.
*Ideal Gas Products  Inc.,  New Jersey
                                     244

-------
     A  permeation tube  device may also be  used,  but again, some users have
had  difficulty  in obtaining reliable chlorine permeation tubes.

     A  fairly simple  calibration aid is available  in which the probe to be
tested  is  inserted into a calibration chamber  and  a known concentration of
gas  passed  through the  chamber at 3 to 4 liters per minute.


                          STABILITY OF CALIBRATION

     A  probe that had been factory-calibrated over  the  range  of 0 to 5 ppm
was  checked over a  period  of three  months using  a permeation  device  in
which 1 and 2  ppm of chlorine  could be  obtained in  the  sample  gas.   These
data are shown  in Table 1.  Although  there are  differences in the absolute
calibration,  using  the same  chlorine source  shows  remarkably consistent
readings over the three months interval.
                     TABLE 1.   STABILITY OF CALIBRATION

Calibration level
Initial check
30
60
90




days
days
days




0.
0.
0.
0.
0.
0.
0.
0.
0
17
17
17
17
17
22
22
22
ppm*
ppm
ppm
ppm
ppm
ppm
ppm
ppm
ppm

1.
0.
1.
1.
1.
1.


1
0
78
0
0
0
0


ppm
ppm
ppm
ppm
ppm
ppm
ppm



1
1
1
1
1
1


2
.83
.83
.72
.83
.83
.83


ppm
ppmt
ppm
ppm
ppm
ppm
ppm



*The chlorine  readings  here are believed to arise  from the inability to be
 able to completely  isolate the flow through the oven containing the perme-
 ation tube.   When  connected to true zero air,  a  compressed air supply not
 connected with  the  permeation device, values  of  0.0 ppm  were obtained on
 every occasion.
tUnadjusted  values  compared  to original factory  calibration made  using  a
 gas cylinder.

     A more  detailed  study  of the  stability  of  calibration has been carried
out elsewhere; we are hopeful that these  results will be published indepen-
dently.
                                 CONCLUSION

     It will be clear  from  the  above account that,  at long last, a reliable
chlorine monitor has  become available that has a degree  of reliability and
specificity not  previously attained  with  a chlorine monitor.   Further, it
is a true  monitor,  being capable of  highly accurate measurements.   We can
hope that  the  days  are gone when the operator wondered if  the  monitor was
still working when  zero  chlorine concentration was  being indicated.

                                     245

-------
              THE DEVELOPMENT OF  STANDARD  REFERENCE  MATERIALS
                   CONTAINING SELECTED  ORGANIC  VAPORS  IN
                          COMPRESSED  GAS MIXTURES
                        W.P.  Schmidt and H.L.  Rook

                   Gas and Particulate  Science Division
                      Center  for  Analytical Chemistry
                       National Bureau  of  Standards
                              Washington,  DC
                                INTRODUCTION

     In the absence of absolute analytical methods  for the determination of
organic compounds  in  the atmosphere, accurate and  stable  primary standards
allow  comparative  analytical procedures  to be used  with  a high  degree of
reliability.   Unfortunately, analytical  studies to  date  on  trace  organic
vapors in  compressed  gas mixtures have  shed little  light  on  the questions
of  either  accuracy  or  stability (1).    Several  years ago,  the  National
Bureau of  Standards  (NBS) undertook  research to determine  the  feasibility
of accurately  preparing  such mixtures and,  if possible, to determine their
stability  with time.   Initially,  the project was  limited to the  study of
mixtures containing  benzene  and tetrachloroethylene in  high-pressure  gas
cylinders  at concentrations  of  0.2-10 parts per  million by mole  (ppm).

     Based on  the  results  of  the initial  study,  the program was  expanded to
include a  total of seven organic compounds in gas  mixtures  in  the concen-
tration range  mentioned above.   This expanded study  included both single-
component  and,  in various  combinations,  multi-component  mixtures.    The
organic  gas  mixtures  completed  thus far  include   vinyl   chloride  monomer
(VCM),     chloroform,     benzene,     carbon    tetrachloride,     toluene,
tetrachloroethylene,   and   chlorobenzene.    Positive   results   have  been
obtained  for  each of   these  seven  compounds.     Gas  mixtures  containing
acrylonitrile  have been  prepared,  but reliable analytical  data have not yet
been obtained.   A  list of gas  cylinder mixtures showing typical components
and concentrations is  given  in  Table  1.

     In addition to  the detailed study  of  the  gas  cylinder  mixtures  men-
tioned above,  preliminary  research  into  the preparation of lower-concentra-
tion  primary   mixtures   (0.05-0.2   ppm)   for certain  organics  has  been
initiated  and  the feasibility  of  the  calibration  of permeation  devices
using  GC-FID  has  been  ascertained.    The results   of  all  of  these  studies
will be discussed  in this  paper.

                                    246

-------




















w
f£
V.

H
X
\-
a
u_^
PS
5
h-
g
c
[2
C
P4
O

O
M
X!
O
H

•
1— 1
Cd
' J
j~j
H
























































































a
P.
a

M
a
c
g
cd
^
I
j
3
CJ
1)
~,
cd f
4J a
cu
H
Ol
c
a
'c
EH

a


^
a c

P c_
ctf n
O,J-

a
o
a
N
a


1 S
O i-t
<-l O
o o

*
1 01
O rH
>-. M
-t 4-1
CJ -H
< a
cu
01 H
cd a
j-i
a.
01
c> •
•H. O
•H a
                   00
                                                                                         o
                                                       -HOOO
                                                       ocslro
                                            COOO
                                            vooo
                                     -HOO
                                     ovo
                                     OO
                                                                                      00
                                                                                   o o
 S22S
a
S
SsS
S   IS
             CM. mo^o
                                            ^d^d

                          CO
                                                 CM    r-«cM
                     CO CO
                     CO CO
                     m CM
                      •  •
                     o o
          CM co -
          oo CM "^ in cy\ -^ t-H

          •o- CM -* CM d d d
                       CM
                       o
                       oo
OOOOOOOOOOOO'OOOOO'-HrHt-Hi—irH—<•—<.—ir-H_-I^H_I_I_I_I_   i_,
^^^^^^^^^^^M5MOTTOOOTOTOOTTOOTOTTO<»TO=°0ococooooooococo
-*-d--*-*-*cocooooooo
                                                      cococoooooCT\cyi,cTtON,cy(cr>cTtCT» cr»
                                                                                               cd
                                                                                               4-1
                                                                                               cd
                                                                                               §
                                                                                               cd
                                                                                               a,
                                                                                               01

                                                                                               a.

                                                                                               CJ
                                                                                               cd

                                                                                               00
                                                                                              T3
                                                                                              Ol
                                                                                              3
                                                                                              a)
                                                                                              -i
                                                                                              n)
                                                                                              CJ
                                                                                              cd
                                   247

-------
               PREPARATION OF PRIMARY  GAS  CYLINDER MIXTURES

     Primary gas  cylinder mixtures for  low vapor-pressure  organic  liquids
can not  be  prepared by  the  same  technique  (successive dilutions) used  to
prepare  mixtures  for high vapor-pressure  compounds.   The  problems  encoun-
tered when  using this  technique for  mixtures  containing such  organics  as
benzene  or  tetrachloroethylene  include   fractionation  due  to  cooling,
adsorption  of  the organic on the  walls of  the  cylinder, and  the  possible
condensation  of   the  organic vapor  inside the  cylinder.    (These  problems
occur  primarily  at  high concentration  levels  of  the  intermediate  gas
mixtures and cause uncertainty  in  mixtures prepared by this method.)

     To  eliminate these  problems, a  method of  preparation was  developed
that produced  the desired low concentration levels by a single dilution.   A
thin-wall glass capillary tube  was sealed at one  end and  weighed.   A small
sample (1-20 mg)  of the organic liquid  was  drawn into the  tube.   The tube
was  then sealed  and reweighed.   To aspirate  the sample  into  an evacuated
and weighed cylinder, the tube  was fitted  tightly  into a  short  length  of
Teflon tubing.  The  Teflon tubing  was  then attached to the evacuated cylin-
der.  The cylinder  valve was opened slightly and the end of the sample tube
nearest  the cylinder was cracked.   After  most  of the  liquid had vaporized
and been drawn into the cylinder,  the other end of  the  tube was carefully
broken and  atmospheric air  was  drawn  through the remaining portion  of the
sample tube to sweep  any residual vapor  into  the cylinder.   The cylinder
was  then pressurized with clean air or  nitrogen to  12.4  x  106 Pascals and
reweighed.   Results obtained by this  technique  for  vapors  of  vinyl chlor-
ide,  toluene,  and chlorobenzene are given in Tables .2, 3,  and 4,  respec-
tively.
       TABLE 2.   COMPARISON OF GRAVIMETRICALLY CALCULATED AND ANALYZED
                 CONCENTRATIONS FOR PRIMARY VCM MIXTURES

Cylinder
FF 9755
FF 9734
FF 9772
FF 9763
FF 9759
FF 9760
FF 9773
FF 6885
FF 9735
FF 9738
FF 9766
FF 9753
Date
prepared
12/5/80
12/4/80
12/5/80
8/30/80
12/5/80
12/5/80
1/27/81
8/07/81
8/30/80
1/22/81
8/30/80
1/22/81
Calculated
concentration
10.963 ppm
4.862
4.408
3.314
2.545
2.226
1.002
0.802
0.612
0.484
0.288
0.194
Analyzed
January, 1981
10.962 ppm
4.868
4.421
2.284
2.546
2.236
1.002
— *
0.586
0.485
0.239
0.200
concentration
September ,

1981
10.961 ppm
4.862
4.414
2.280
2.541
2.228
1.000
0.802
0.583
0.485
0.238
0.197












 *Cylinder FF 6885 was prepared after the intercomparison  of January,  1981.
                                     248

-------
       TABLE 3.  COMPARISON  OF GRAVIMETRICALLY  CALCULATED AND ANALYZED
                 CONCENTRATIONS FOR PRIMARY  TOLUENE MIXTURES

Cylinder
FF 9762
FF 9734
CAL-6523
FF 9738
FF 9779
FF 9772
FF 9754
FF 9759
FF 9497
Date
prepared
8/07/81
12/4/80
9/22/81
1/22/81
8/07/81
12/5/80
9/28/81
12/5/80
9/28/81
Calculated
concentration
3.952 ppm
2.747
2.276
1.581
1.381
0.947
0.336
0.325
0.287
Analyzed
January, 1981
	 ^
2.78 ppm
— *
1.58
— *
0.95
	 A
0.31
— *
concentration
September, 1981
3.960 ppm
2.737
2.279
1.571
1.381
0.948
0.342
0.326
0.285

       TABLE 4.   COMPARISON OF GRAVIMETRICALLY CALCULATED AND ANALYZED
                 CONCENTRATIONS FOR PRIMARY CHLOROBENZENE MIXTURES

Cylinder
CAL-6523
FF 9738
FF 9763
FF 12044
FF 9779
FF 9734
FF 9772
CAL-6488
FF 9747
FF 6885
FF 9754
FF 9497
FF 9735
FF 9741
Date
prepared
9/22/81
1/22/81
8/30/80
9/17/81
8/07/81
12/4/80
12/5/80
9/17/81
8/05/80
8/07/81
9/28/81
9/28/81
8/30/80
8/05/80
Calculated
concentration
9.081 ppm
3.035
2.122
2.038
1.820
1.544
1.119
0.981
0.818
0.744
0.618
0.584
0.500
0.235
Analyzed
January, 1981
— *
3.044 ppm
2.125
— *
— *
1.530
1.116
— *
0.814
— *
— *
	 A
0.365
0.233
concentration
September, 1981
9.081 ppm
3.008
2.124
2.037
1.698
1.524
1.116
0.980
0.817
0.710
0.617
0.584
0.363
0.232

     Two modifications  to  this  micro-gravimetric  technique for sample prep-
aration were made  to  improve the accuracy  of  this technique  when  used for
either  high vapor  pressure  organics  or  multi-component  mixtures.   These
modifications  were made  necessary  by a lack of  correlation between the
gravimetric and analyzed concentrations for both  VCM  and one or more compo-
nents in several of the multi-component mixtures.  In all suspect  samples,
the analyzed  concentrations  were  considerably lower  than  the  gravimetric
concentrations.
                                    249

-------
     The problems  encountered in making VCM mixtures  included  difficulties
in sealing  the microtubes  containing this  organic,  polymerization  of  the
VCM in  the  sealed microtube, and  in general,  the quantitative  addition of
the organic  to  the evacuated gas cylinder.  Occasional  lower-than-expected
concentrations  of  one or more  components  in multi-component mixtures  were
traced  to a  faulty transfer of  the  organic  compounds to  the evacuated gas
cylinder.

     To  solve  the  inconsistencies in the VCM mixtures,  only freshly filled
microtubes were used  in  the preparation of  the  mixtures, the cylinder valve
was opened  completely before the  end of the microtube  was cracked,  and a
modification was  made to  the microtube filling technique  wherein  dry ice
was applied to the  closed end  of  the microtube  after  the liquid  VCM had
been added  to  the  tube but  prior to  sealing  it.   The effect of  this modifi-
cation  was  twofold.   First,  the  extra  cooling  allowed  a  small  quantity of
air to  be drawn through  the tip, sweeping  the  organic further into the tube
and eliminating a tendency  toward  charring  and the  resultant  ashy residue
after  flame-sealing  the  tip.    Second,  the   resultant  pressure  decrease
inside  the  tube allowed  a surer seal,  since the application of  heat during
sealing had previously  caused  an increase  in  pressure,  leading  to faulty
seals.   Samples prepared by this modified  procedure showed better agreement
between gravimetrically  calculated  and  analyzed  concentrations   than did
samples  prepared  by the  original procedure (Table 5).  Mixtures prepared by
this  modified  procedure  were  compared  to mixtures  prepared using  the
successive   dilution  technique.     The  results   of  the  intercomparison
indicated  good agreement  (±0.5  percent rel.)  between mixtures  prepared by
both  techniques  but  better  precision  (±0.2 percent rel.)  among  mixtures
prepared by the modified microtube procedure (Table 6).

      TABLE 5.  COMPARISON OF GRAVIMETRICALLY CALCULATED AND ANALYZED
                CONCENTRATIONS FOR VCM MIXTURES:  MODIFIED  VS.  ORIGINAL
                PREPARATIVE PROCEDURES	
           Original procedure
                              Modified procedure
                Concentration, ppm
 Cylinder  Caleu-
    no .     lated
Analyz ed   A » ppm
                                      	Concentration ppm
                            Cylinder  Calcu-
                               no.	lated    Analyzed   A ,  ppm
 FF 9735
 FF 9766
0.612
0.288
  0.583    -0.029   FF 9738   0.484       0.485      0.001
  0.238    -0.050   FF 9753   0.194       0.197      0.003
                    FF 9773   1.002       1.000     -0.002
                                     250

-------
       TABLE 6.  COMPARISON OF GRAVIMETRICALLY  CALCULATED  AND ANALYZED
                 CONCENTRATIONS FOR PRIMARY VCM MIXTURES PREPARED  BY THE
 	SUCCESSIVE DILUTION TECHNIQUE  AND  THE MICROTUBE TECHNIQUE

 	Successive dilution	  	Micro-tube
           	Concentration	                  Concentration    ~
 Cylinder  Calcu-                     Cylinder  Calcu-
    no.     lated	Analyzed  A , %	no.    lated	Analyzed A ,  %

 000674     9.902 ppm  9.857 ppm -0.5  FF 9755   10.963 ppm  11.001 ppm  0.3
 000335     4.984      4.867     -2.3  FF 9772   4.408      4.412       0.1
 000661     0.983      0.984      0.1  FF 9759   2.545      2.547       0.1
      The modification to the  procedure used in the  addition of the micro-
 tubes in  the multi-component  mixtures  preparation  consisted  of  changing
 from a  simultaneous addition  to  a  serial  addition of  the tubes  to the
 cylinder.   In the  original procedure, all  tubes  to  be  added were wrapped
 together with Teflon* tape, and fitted  into  the end of a 1/4" O.D.  Teflon
 transfer line.   The  cylinder  valve was  opened  and the  ends  of the micro-
 tubes were broken simultaneously.  When  four or more microtubes were added
 in  this  way,  there was  insufficient  flow through  the  microtubes to ensure
 that  all of  the  organic liquid from all  of  the microtubes  was swept into
 the  cylinder.   In the modified procedure, a  1/8"  O.D.  Teflon transfer line
 was  employed and  the tubes were added  singly.   Samples prepared  by the
 modified procedure showed  no  inconsistencies between gravimetric  and ana-
 lyzed concentrations  (Table 7).


                      ANALYSIS  OF  PRIMARY GAS MIXTURES

      The mixtures  were  analyzed by GC-FID employing  a  10' x 1/8" stainless
 steel column  containing  20  percent  SP-2100 and 0.1 percent Carbowax 1500 on
 100/120  Supelcoport.  The column  temperature was 100°C and the carrier flow
 rate  was 60  cc  nitrogen/minute.   Total  time  for each  separation was  12
minutes.

      A modification to  the analytical sampling procedure was  made  during
 the past year.  In  the  analysis,  cylinder control valves  were used to sam-
ple from the  high-pressure  gas cylinders.   Originally,  a  sample flow rate
of 30 cc/min  was  established through  the sample loop of  the  automatic sam-
pling valve.  It was  necessary  to maintain this  flow  rate for several hours
prior to the analysis  to  obtain meaningful  responses  from  the analytical
*Certain commercial equipment, instruments,  and materials  are identified in
 this  paper^ in  order to  adequately  specify  the  experimental  procedure.
 Such  identification  does not imply recommendation or  endorsement by  the
 National Bureau  of  Standards,  nor does  it  imply  that  the materials  or
 equipment identified are necessarily the best available for  the  purpose.
                                    251

-------
TABLE 7.  MULTI-COMPONENT MIXTURES WITH  TOTAL  NUMBER OF  ORGANIC COMPONENTS*
          AND NUMBER OF COMPONENTS SHOWING  SIGNIFICANT (±2% REL) DEVIATION
          BETWEEN CALCULATED AND ANALYZED CONCENTRATIONS:   MODIFIED VS.
          ORIGINAL PREPARATIVE PROCEDURES

Original procedure
Cylinder
no.
FF 9735
FF 9734
FF 9772
FF 9738
FF 6885
FF 9762
Total
components*
5
4
4
4
5
4
Deviating
components
4
0
0
1
3
0
Modified procedure
Cylinder
no.
CAL-6523
FF 9754
FF 9497



Total
components*
5
5
5



Deviating
components
0
0
0




*Excluding acrylonitrile.
system.  An experiment was undertaken  to determine the  actual  time needed
to  equilibrate  the cylinder  control  valve to  the  concentration of  the
organic being  analyzed.   The results of  this experiment  indicated that the
equilibration  of  the valve was  dependent on the individual organic compound
and  the total  volume of  sample  passed  through the  cylinder  control valve,
not  on the flow  rate.    The  volume of  sample  necessary  for  equilibration
ranged  from about  0.1  L  for  VCM  to  M.O  L for chlorobenzene.    Plots  of
equilibration  volumes  and times are  shown in  Figures  1  through 5.    The
modification to  the original  procedure  consisted of using higher flow rates
(MOO cc/min)  and equilibrating the valve for one to two hours prior to the
analysis.   Once  a valve  had  been equilibrated,  the  concentration of  the
organic compound  remained constant  until the valve was depressurized,  e.g.,
as when it  was removed from the gas cylinder.


     ANALYTICAL  RESULTS  OF THE INTERCOMPARISON  OF THE  PRIMARY MIXTURES

     The  results  of the intercomparisons  of  the primary  gas  mixtures  for
three  organic solvents  are given  in Tables 2 through  4.   The  agreement
between  the calculated  concentrations  based  on gravimetric  data  and  the
concentrations determined by GC-FID  analysis   is  a  demonstration of  the
precision  and accuracy  of the  preparation of  the primary  mixtures.   As
mentioned  above,  in all cases  of  significant  inconsistencies  between the
calculated  and analyzed  concentrations,  the calculated  concentrations were
higher  than the  analyzed  concentrations,  indicating losses in the transfer
of the organic from the  microtube to the cylinder.
   PREPARATION AND ANALYSIS OF LOW-CONCENTRATION (^50 PPB) GRAVIMETRIC GAS
   MIXTURES

      Preliminary studies have been completed on the feasibility of prepara-
 tion and analysis  of  low-concentration gravimetric mixtures.   Gravimetric

                                     252

-------
CD
X

CD
to
c.
o
d.

-------
      101%
      100%
       99%
CD
CD
X

Cl>
CO
c:
o
CL
CO

£
       97%
    §
       96%
       95%
       94%
        93%
                        I
           0
                                                T
                                                3
1           23           H-

   Elapsed Time of Flow, hours
  Figure 2.   Effect of 100 cc/min sample flow rate  on equilibration time of

             cylinder control valve  for
                                    254

-------
O
CD
X

CD
CO
E
O
O.
CO
CL>
      101%
      100%
       99%
       98%
   97%
•*-•
a
       96%
       95%
       94%
       93%
                                         ®
          0
                  •
® Flow rate  =  20 cc/min
•Flow rate  = 100 cc/min
                               2           3

                      Elapsed  Time of  Flow, hours
              ~r
               4
T
 5
 Figure 3.  Effect  of  sample flow rate and flow volume on equilibration time
           of cylinder control valve for benzene.
                                  255

-------
101%
100%
99%
§ 98%
•1—4
o .a
o •-!
r— i i— <
x § 97%
CD ii 1
CO
1 °
1 §96%
Q Oi
!«•»<
»— 1
95%
93%
C

•
• • •
•
0
•
•
•




i i i i i
112345
                         Elapsed Time of Flow, hours
Figure 4.   Effect of 100 cc/min sample  flow  rate on equilibration time  of
           cylinder control valve for €20!^.

                                  256

-------
6-5
X
CO
cr
o
Q.
CO
CD
      101%
      100%
       99%
       98%
       97%
I  96%
       95%
       94%
       93%
                                         ®
                                                ®
          0
                            100 cc/min flow rate  stopped;
                            restarted  after 2 hours
                                       ® Flow  rate =  20 cc/min
                                        • Flow  rate = 100 cc/min
                  1	1	1	
                  1           23

                     Elapsed Time of  Flow, hours
 Figure  5.  Effect  of sample flow rate and flow volume on equilibration time

           of cylinder control valve for chlorobenzene.
                                  257

-------
                                                           Q
mixtures  have  been  prepared  in  both  0.85  cubic  m  (m )-size  aluminum
cylinders and,  also,  4.25 m3-size cylinders.  The  uncertainty in the accu-
racy  of  the standards prepared  in  the smaller cylinders was  determined to
be on the order of  5  to  10 percent  due to  the necessity  of  weighing micro-
gram  quantities of  the organic  reagent.   Larger quantities  of  the organic
were  used in mixtures prepared  in  the larger  cylinders,  but  the weight of
these cylinders  0\>25  Kg)  precluded   accurate  (±0.5 percent  rel.)  weight
determinations  for  the diluent gas.   Higher concentration mixtures (0.2 to
15 ppm) were also prepared in these  larger cylinders to detect  any severe
problems  in the  weight  determination  of  the  diluent  gas.    The estimated
uncertainties in the  accuracy of mixtures  prepared  in  these  4.25 m3 cylin-
ders  was about  3 percent.

      These  low-concentration  mixtures were intercompared (GC-FID)  with the
higher concentration  primary  mixtures prepared in the 0.85 m3  cylinders.  A
calibration curve was established  using the  FID response  asnd  the gravi-
metrically  calculated concentration  of the primary  mixtures.   The gravi-
metrically  calculated and the analyzed concentrations  of both the low-con-
centration  mixtures and  those  mixtures prepared  in the  4.25 m3 cylinders
are shown in Table 8.    The  uncertainties  in the  analyzed  concentrations
have  not been   completely ascertained at  this  time,  but   the  agreement
between  the  calculated  and  analytical  concentrations   demonstrates  the
feasibility of  sample  preparation at  these  levels.
DETERMINATION  OF  THE MINIMUM DETECTABLE  LEVEL USING GAS  MIXTURES  PREPARED
BY THE SUCCESSIVE DILUTION  TECHNIQUE

     To determine both  the  feasibility of the successive dilution method of
preparation  for  low concentration mixtures  and the sensitivity  of  the FID
to  low concentrations  of various  organic solvents,  two primary  mixtures
that had become depressurized during  analysis were diluted with nitrogen to
produce low  ppb  concentration mixtures.   The  mixture  in cylinder  FF  9762
was  diluted  in three  steps, each  successive  mixture   being  approximately
one-tenth  the concentration of  the  preceding mixture.    The mixture  in
cylinder FF  9779  was diluted to  one-thousandth of  its  original  concentra-
tion in a single step.   Final concentrations  in both cylinders were approx-
imately equal  and ranged from 1 to  5  ppb.   Typical GC-FID chromatograms for
each of  the mixtures  are shown  in  Figure 6 with the   retention times  and
approximate  concentrations  for  each organic  compound.   The  minimum detect-
able limits  (useful  primarily for calculations of  the concentration of  each
compound in  the diluent  gas) are  estimated to be:

                       VCM                      2 ppb

                       chloroform              5 ppb

                       benzene                  1 ppb

                       tetrachloroethylene     5 ppb

                       chlorobenzene            1 ppb

                                    258

-------



z
o
M

<
Q M
W 2
E-i
< &
iJ W
BQ
!3

!
S g
•<] 1 — 1
CJ 0
M  0
^C (—1
w PC
o
Q

o <3

Z ^N
0 «

i-i cu
->
X!
4J
01
0
}-l
O
tH
y
rt

01
EH














01
pi
01
N
PI
Ol
M




















^»


























Xi
a
PU

M
<3


a) •
N y
^1 PI
i-l O
ts y
p.
^
•a
o>
4-1 •
cs y
^H Pi
3 0
y y
T~H
c8
U

Xi
tt
ft

A
<]

01 •
N y
t>~» Pi
r-l 0
g y
<^


13
4-1 •
n) y
3 o
y y
H
cS
O

PL,
PL,

*
CJ



O ,— I CM VO CM
1 ^H
1


PL,
PL,

CO CM O CO O
O •* •* r» O
I-H t-n o r^


,Q
PL,
PL,
CO i— 1 OO ~~
•—1 O>





CM v£) 1^ 00 CM
1 1 •* 1


Xi
PL,

1 — -
CO 00 CT\ lO ^ LO *""' *~~' t^» 00
10 o*\ r**» LO vo o^ ON o^ CM CM
r^. 
-------
      CO
      CM
in

-------
      The lowest  concentrations that  can be  analyzed  using  a  5-cc sample
 loop  and a simple (non-preconcentrating)  gas  injection are estimated to  be
 approximately five times the minimum detectable levels.

      The lowest concentration  of  each organic compound  that  has been pre-
 pared and  analyzed  and  the uncertainty  (Icr) in the  average   of  the  FID
 response to replicate  5-cc samples of each is shown below:
         Compound

         VCM

         chloroform

         benzene

         tetrachloroethylene

         chlorobenzene
Concentration

   11 ppb

   70 ppb

   22 ppb

   80 ppb

   53 ppb
Uncertainty

 ±0.6 ppb

 ± 4   ppb

 ±0.4 ppb

 ± 2   ppb

 ± 1   ppb
                      STABILITY OF PRIMARY GAS MIXTURES

     The  stability  of the primary gas mixtures  for  various periods of time
was  demonstrated by  the  lack  of change in  the  relative ratios  of  the FID
response  to  the gravimetrlcally calculated  concentrations  of  mixtures pre-
pared  at  intervals  during  the  course  of  the study  (Tables 2  through 4).
For  example,  mixtures  containing chlorobenzene  were prepared  during 8/80
and  were  compared   with  freshly  prepared  mixtures  in  1/81  to  determine
short-  term  (5  months)  stability.  These  mixtures were  later  compared with
freshly prepared mixtures  in  9/81  to  determine  long-term stability  (12
months).   The  stability  of the  mixtures  was determined  by the lack  of  a
significant  change  in  the  analyzed concentrations  when compared  with the
freshly prepared mixtures at the intervals specified.  No systematic change
in  concentration with  time  was  observed  for  any  of   the seven  organic
solvents  for which  the  stability study had been completed.


                        PERMEATION TUBE  CALIBRATION

     A  set  of  100 permeation  tubes  containing benzene  has  been calibrated
by both gravimetry  and  comparison with primary mixtures  using GC-FID.   The
results  of  the  calibration of  17 representative  tubes from  the   set  are
shown in Table  9.   The  average uncertainty (2a)  in the gravimetric  calibra-
tions for these tubes was approximately ±3  percent   (relative).  The  esti-
mated total uncertainty (20) in the  calibration performed by the comparison
with  the  primary mixtures  was ±2  percent   (relative).   The  disagreement
between the  gravimetric and the  comparative calibrations was less  than ±1
percent (relative)  (Table 9).   The precision  with  which  permeation  tubes
can be  intercompared  by FID (as shown in Table 9) is  better than ±0.5 per-
cent (relative) for  all tubes  except #1-3.   The  discrepancy in the  calibra-
tion data for this  tube can not  be  satisfactorily explained at  this  time,
                                    261

-------
    TABLE  9.   CALIBRATION OF BENZENE PERMEATION TUBES AT 25.0°C.  RATES
               DETERMINED BY GRAVIMETRY AND BY COMPARISON WITH PRIMARY GAS
               CYLINDER STANDARDS

Permeation
Permeation Permeation FID* response rate, 25.0 °C
tube rate, 25.0 °C dilution flow = FID
no. (gravimetric) 100 cc/min (vs. grav. stds.
3
5
12
13
20
22
26
31
53
54
58
66
70
76
82
85
86
0.338 yg/min
0.369
0.355
0.355
0.343
0.372
0.358
0.341
0.341
0.356
0.371
0.377
0.360
0.339
0.347
0.365
0.368
2504
2710
2600
2610
2512
2711
2618
2492
2494
2595
2713
2756
2638
2494
2527
2658
2696
0.343 yg/min
0.371
0.356
0.357
0.344
0.371
0.358
0.341
0.341
0.355
0.372
0.377
0.361
0.341
0.346
0.364
0.369
A permeation
rate
, ) (FID - grav.)
1.5 % Rel.
0.5
0.3
0.6
0.2
-0.2
0.0
0.0
0.0
-0.2
0.2
0.0
0.3
0.7
-0.3
-0.4
0.3

*Flarae ionization  detector.
but  this tube  had  been subjected  to higher-than-ambient  pressure  for  a
period of several  weeks.   An effort will be made  to  quantify  the effect of
pressure on permeation tube  output in the near future.

     The demonstrated ability to  calibrate  permeation  tubes  by comparison
with primary  mixtures using GC--FID  gives more flexibility  to  the analysis
of environmental effects on the permeation rate.   A  calibration of permea^
tion  rate  may  be  performed  in  1/2  hour instead  of  several  weeks.   The
effects  of  temperature cycling,  flow  rate  of  purge  gas,  pressure (as men-
tioned above),  and storage parameters can now be  determined accurately and
quickly.


          STANDARD REFERENCE MATERIALS (SRM's) FOR TOXIC ORGANICS

    Four sets (50 cylinders  each) of gas  mixtures  containing  benzene and
tetrachloroethylene  at nominal concentrations  of  0.25 and 9.5  ppm have been
received for  the  certification process leading to  their issuance as SRM's.
The  concentrations of  these  mixtures as determined  from  their comparison
with the primary gravimetric gas mixtures are  shown in Table 10.
                                    262

-------
   TABLE 10.  COMPARISON OF NOMINAL AND ANALYZED  CONCENTRATION  FOR TOXIC
              ORGANIC SRM's

Cylinder no.
CAL 7216
GAL 6603
CAL 6522
CAL 6485
CAL 6499
CAL 6599
CAL 5718
CAL 5715

Organic
matrix
Tetrachloroethylene
Nitrogen
Tetrachloroethylene
Nitrogen
Benzene
Nitrogen
Benzene
Nitrogen

Nominal
concentration
9.5 ppm
0.25 ppm
9.5 ppm
0.25 ppm

Analyzed
concentration
9.86 ppm
9.82 ppm
9.83 ppm
0.26 ppm
0.25 ppm
9.79 ppm
0.26 ppm
0.25 ppm

                                   SUMMARY

     Primary  gas mixtures  have  been prepared  for seven  organic  solvents.
Increased  sophistication  in the  preparation technique has enabled the prep-
aration  of multi-component mixtures with both  a  large  number of components
and a  greater confidence  in the accuracy of  the  calculated concentrations.
Stability  studies  performed using these mixtures  have  shown no significant
changes  in concentration  with  time over periods of 1  to 2 years.

     Permeation  tubes  have been calibrated  by  both gravimetry  and  compar-
ison  with  primary gas  cylinder  mixtures  using  GC-FID.   No  significant
differences were observed between  the  permeation rates that were  obtained
by  these methods.   In the future,  permeation  tubes can  be calibrated  by
GC-FID comparisons  alone, thereby  decreasing the time  needed  for  calibra-
tion and increasing the flexibility of  the  calibration  procedure.

     Accurate  low-concentration  (0.05  to 0.2 ppm) gas  mixtures have  been
prepared gravimetrically  in both 0.85 m3 and  4.25 m3  cylinders.   The pro-
cedure employed  in the above preparation appears  adequate  to prepare sam-
ples at low ppb  concentration  levels.
1.
                                REFERENCES

     McGaughey, J.F.,  A.L. Sykes ,  D.E.  Wagoner,  and C.E.  Decker.    1979.
     Research Triangle Institute Report 1808/13-01.   Research Triangle
     Park, NC.
                                    263

-------
          HUMAN EXPOSURE TO VAPOR-PHASE HALOGENATED HYDROCARBONS:
                      FIXED-SITE VS_ PERSONAL EXPOSURE


  E.D. Pellizzari, T.D. Hartwell, C. Leininger, H. Zelon, and S. Williams

                        Research Triangle Institute
                        Research Triangle Park, NC

                                    and

                         J.J. Breen and L. Wallace

                   U.S. Environmental Protection Agency
                              Washington, DC
                               INTRODUCTION

     During  the  past five  years,  we have been investigating the  personal
exposure of populations to  toxic chemicals,  and assessing  methods  for their
analysis in air, water, food, blood,  urine,  and breath.  We have previously
reported  on preliminary  studies  of  human  exposure  to benzene and  other
volatile organics  (1-3).   Benzene was  monitored in lay people  in  Houston,
Texas, and Wood  River,  Illinois  (1), while volatile organics  were  measured
in student populations  in Beaumont,  Texas, and  Chapel Hill,  North  Carolina
(2), and in a nine-person group  in Elizabeth and Bayonne,  New Jersey (3).

     In a larger study,  we have investigated the major pathways  that  con-
tribute to human exposure to volatile halogenated hydrocarbons  for popula-
tions in two different geographical  areas.   Although  this  program  addresses
the  integrated  exposure and body  burden in  populations by  measuring vola-
tile halocarbons  in  the air a person breathes  and the water  he  drinks,  and
the  parent  halocarbons and important metabolites  in  the   person's  breath,
blood, and urine,  we wish to report  our findings on one route of  exposure,
air.  This paper addresses  the use of personal and  fixed-station monitoring
for  assessing exposure of populations to halocarbons.  The relative merits
of each are  discussed.
                                EXPERIMENTAL

Areas and Halocarbons  Selected for Monitoring

     Table  1 lists  the areas  and halocarbons  selected for  monitoring  in

                                    264

-------
          TABLE  1.   GEOGRAPHICAL AREAS AND HALOCARBONS MONITORED IN
         	   PERSONAL AND FIXED-SITE AIR
 Greensboro,  NC
Baton Rouge/Gelsmar, LA
Vinylidene  chloride
Chloroform
Chloroprene
1,2-Dichloroethylene
1,2-Dichloroethane
1,1,1-Trichloroethane
Carbon  tetrachloride
1,2-Dichloropropane
Trichloroethylene
Bromodichloromethane
Dichlorobutane  isomer
1,1,2-Trichloroethane
Chlorodibromomethane
Trichlorobutane isomers
Tetrarchloroethylene
Bromodichloroethane
Chlorobenzene
Bromoform
Chlorobenzotrlfluoride isomers
1,1,2,2-Tetrachloroethane
Bromobenzene
Chlorotoluene isomers
Dichlorobenzene isomers
Hexachloroethane
Trichloropentane isomers
Bis-(Chloroisopropyl)ether
Chloroni t robenz ene
Methyldichlorophenoxy acetate
Trichlorohexane isomers
Dichlorotoluene isomers
Bromopropylbenzene
Trichlorobenzene isomers
1,3-Hexachlorobutadiene
Trichlorotoluene isomers
Tetrachlorobenzene isomers
Vinylidene  chloride
Chloroform
Chloroprene
1,2-Dichloroethylene
1,2-Dichloroethane
1,1,1-Trichloroethane
Carbon tetrachloride
1,2-Dichloropropane
Trichloroethylene
Bromodichloromethane
Dichlorobutane isomer
1,1,2-Trichloroethane
Chlorodibromomethane
Trichlorobutane isomers
Tetrachloroethylene
Bromodichloroethane
Chlorobenzene
Bromoform
Chlorobenzotrifluoride isomers
1,1,2,2-Tetrachloroethane
Bromobenzene
Chlorotoluene isomers
Dichlorobenzene isomers
                                    265

-------
air.   Ambient monitoring  data were  surveyed  (4)  and  the more  prevalent
halocarbons were  then assessed as  to  biological  activity (4,5).   For  each
halocarbon,  potential  carcinogenicity  and  mutagenicity  were  ascertained
f.rom  literature   reports,  and  those  halocarbons  exhibiting activity  were
included in the monitoring effort  (4).

Sample Design

     The population  in Greensboro, North Carolina was  stratified  according
to  three  socioeconomic categories, and  Baton Rouge/Geismar, Louisiana  was
stratified relative  to suspected  point sources of halocarbons  of  interest.
The between-site  differences  in stratification occurred  because Greensboro
was considered a  reference area with no known major point  sources.   Figure
1  depicts  an example of a  stratified  area  for Baton  Rouge and  Geismar.
This approach  considered the U.S.  Census data for  each area,  the  average
wind rose  for  the season under study, and potential  point sources  in  the
areas.

     Counting and listing  of  households in these areas  was  then conducted.
Households were  screened to  identify  those  houses that  contained  at least
one  eligible individual.   People were  considered  eligible  if they  were
non-occupationally  exposed (to  the halocarbons  under  study),  non-smoking
adults between the ages  of 45 and  64 who had  lived in  the area  for at least
one year.  In Greensboro,  374 households were screened  to  yield 101 eligi-
ble  housing  units;  721 households were screened  in  Baton Rouge/Geismar,
producing  190 eligible housing units.   Participants  were then selected from
those persons  eligible within  the  eligible  households; final  sample sizes
were  28  and  66 people,  respectively,   for  Greensboro  and  for  Baton
Rouge/Geismar.  A probability sample was drawn from the screened households
for  the  stratification  variables, and  the  individuals  were solicited  to
participate  in  the   exposure monitoring  effort.   Field   interviews  were
performed with each  participant,  and a demographic questionnaire was admin-
istered.

Sampling and Analysis

     Acquisition  of  air samples was conducted during  the months of October
and November,  1980,   in Greensboro,  and during January  and  February, 1981,
in Baton Rouge/Geismar.  Two  sampling  periods for ambient air (an overnight
and  daytime period)  were used with  each participant.    Each period  was
approximately  11 to  13 hours, with  a  personal  air  and  a  fixed-site  air
sample  collected,  concurrently.    To  collect  personal  air  samples,  the
volunteer  wore a  vest equipped with a collection system as  shown  in Figure
2.   The  sampling train consisted of  a Tenax  GC cartridge  (6,7)  with  a
prefilter  for  removing particulate and  a small, personal  air  pump (DuPont
Model No.  P125 or MSA C-200).  The fixed-site sampler was  identical to the
personal air  sampler (Figure 3),   except  that  it was  placed outside in the
participant's yard for the entire time period.   In  general,  a fixed-site
sampler  represented  a cluster  of  participants (one to  three);  however,  it
always  matched a personal sample for  at least  one  participant   in  each
cluster.   A nominal  sampling  rate of 35 mL/min was used; approximately 25 L
was sampled per  time period.

                                    266

-------
Figure  1.   Stratified populations  in  Baton Rouge  and  Geismar, LA.   Solid
           circles  =  suspected sources,  light  and dark  areas  represent
           potentially high-  and low-exposure  areas (distance between sites
           not  to scale).
     table  2 lists  the  number  of  samples  obtained by  category for  each
geographical  area.   In Greensboro, 28  participants were sampled, while  66
people were  included in  Baton Rouge/Geismar.   In addition to  primary  sam-
ples, duplicate samples were also collected.

     Analysis  of  air samples  was  performed  by gas-liquid  chromatography/
mass spectrometry/computer techniques,  as  previously reported  (6,7).

Quality Control/Assurance

     A quality control and assurance program  (QC/QA) was  maintained  for the
sampling and  analysis  procedures.   Table  3 gives  the  major categories  of
the QC/QA program.   For Tenax GC  cartridges,  laboratory and  field  blanks

                                    267

-------
Figure 2.  Vest  equipped with Tenax  GC sampling cartridge,  prefilter for
           particulate and personal pump  (in pocket)  for collecting vapor-
           phase halocarbons in personal air.

                                    268

-------
Figure 3.  Sampling  system depicting  filter,  Tenax GC  cartridge  and pump
           for collecting fixed-site air samples.

                                    269

-------
           TABLE  2.  NUMBER OF  SAMPLES  OBTAINED BY CATEGORY FOR
                     EACH  GEOGRAPHICAL  AREA*
                               Greensboro
Sample type
Primary
Duplicate
Baton Rouge/Geismar
Primary   Duplicate
Personal Air
Fixed-Site Air
53
37
6
23
132
55
19
11
*Also, an additional  10  percent of the samples in  each  category were  con-
trol samples and  10 percent  were  blanks.


	TABLE 3.   QUALITY CONTROL/QUALITY ASSURANCE	


Tenax GC Cartridges

   - laboratory blanks
   - field blanks
   - laboratory controls  (spiked  with target  halocarbons)
   - field controls (spiked  with  target halocarbons)

Replicate samples

Audit of sampling  and analytical  systems

   - pump flow rates
   - battery charge
   - GC/MS performance specifications and control charts	
were  maintained.   These  were  sampling  cartridges selected  from a  batch
preparation  (^30-50)  of  cartridges  to demonstrate the  potential background,
if  any,  that  might develop  during the  period  of  sampling and  analysis.
Field blanks were  sampling cartridges transported to the field and returned
to the laboratory  unused,  and analyzed,  while laboratory blanks were stored
during the entire  period.   Thus, some indication of the potential contami-
nation that  might  occur,  not only  within a  batch of  cartridges,  but  also
from the influence of transportation was obtained.

     Laboratory  and  field controls  were also  maintained  for each  batch
production  of  Tenax GC  cartridges.    Controls   were  sampling  cartridges
spiked with  the  list  of  target  halocarbons.

     Replicate samples  (Table  2)   of  personal   and  fixed-site  air  were
collected! in general,  a  minimum of  10 percent  were acquired.    Also,  10
percent of the total samples collected  also  had  a representative  number of
blanks and controls.
                                     270

-------
     Each  sampling train was  internally audited before,  during,  and after
sampling in  each geographical area.  Flow  rates  were  checked with a bubble
meter  to  assure that  the  proper rates  were  attained.  Battery  charge was
verified on  personal  samplers to ensure that the flow rates were maintained
for  each  sampling  period.   Recharging  was instituted after  each sampling
period.

     A chain-of-custody procedure was maintained for each sample, blank and
control, throughout the period of sampling and analysis.
                           RESULTS  AND DISCUSSION

     The  recoveries  of target halocarbons from  control  sampling cartridges
for Greensboro  and Baton Rouge/Geismar  are  given in Tables  4 and 5.   The
data are  mean recoveries and percent  relative standard  deviations  for each
halocarbon,  representing the entire  time  period from preparation  to their
analysis, along with  field  samples.   In  some cases,  the  time span reached a
period of five weeks.   In  general,  acceptable recoveries were obtained, and
thus, none of the  field sample data were corrected  for  this  potential bias
in accuracy.
     TABLE 4.  RECOVERY  OF  HALOGENATED  CHEMICALS FROM CONTROL SAMPLING
               CARTRIDGES - GREENSBORO,  NC STUDY

Halogenated chemical
Chloroform
1 , 2— Dichloroethane
Trichloroethylene
Tetrachloroethylene
1,1, 1-Trichloroethane
1 ,2-Dichloroethylene
Carbon tetrachloride
1 , 1 ,2-Trichloroethane
1,1,2, 2-Tetrachloroethane
Chloroprene
Chlorobenzene
jD-Chlorotoluene
Bromodichloromethane
Hexachlorobutadiene
1,3,5-Trichlorobenzene
2 , 4-Dichlorotoluene
1 , 2-Dichloropropane
% Recovery ± S.D. (%RSD)*
79 ± 10 (13)
93 ± 8 (9)
102 ± 7 (7)
115 ± 14 (12)
111 ± 21 (19)
81 ± 12 (15)
79 ± 15 (19)
105 ± 19 (18)
92 ± 2 (2)
93 ± 5 (5)
102 ± 12 (12)
94 ± 11 (12)
88 ± 14 (16)
111 ± 33 (30)
125 ± 13 (10)
102 ± 34 (33)
112 ± 9 (8)

*N = 9.
     Replicate field  samples  for personal and-fixed-site air were  analyzed
for precision.   These  data are  presented in Figures  4 through  9.    Only
field duplicate samples  that  yielded measurable values  in  both  samples  are

                                     271

-------
     TABLE 5.  RECOVERY OF HALOGENATED  CHEMICALS  FROM CONTROL  SAMPLING
               CARTRIDGES - BATON ROUGE/GEISMAR,  LA STUDY

Halogenated chemicals
Chloroform
1 ,2-Dichloroethane
Trichloroethylene
Tetrachlonoethylene
1,1, 1-Trichloroethane
Carbon tetrachloride
1,1, 2-Trichloroethane
1,1,2, 2-Tetrachloroethane
1 , 2-Dichloropropane

% Recovery ±
63 ±
102 ±
109 ±
98 ±
92 ±
98 ±
100 ±
116 ±
81 ±

S.D. (%RSD)*
15 (23)
23 (22)
2.3 (21)
16 (16)
23 (25)
18 (18)
24 (24)
15 (13)
25 (31)

*N ^ 16.
represented; non-measurable  or trace values  were  omitted.   Thus,  for  many
chemicals,  insufficient  data were available  for performing linear  regres-
sion  analysis.    Figure  4 depicts  the  replicates  for  tetrachloroethylene
measured in Greensboro  samples.   The data  are  plotted on a log-log  scale.
Linear regression analysis of  these  data reveals a  correlation (r)  of 0.990
and 0.997  for  fixed-site and personal air  samples, respectively.   Included
are ±25 and ±50 percent  limits from  the  45° line.   f

     As indicated in  each figure  (4  through 9)  a reasonable  correlation was
obtained between  each measurement.   For  1,1,1-trichlorethane  (Figure  5),
two sets  of duplicates  were  considerably  removed  from  the  45° line;  how-
ever,  acceptable  linear correlations  were obtained  for  both personal  and
fixed-site  samples  due to the  relatively large  range  of  compound values.

     The limits  of  detection  (LOD)  and  quantifiable  limits (QL) were,  for
initial  analysis,  arbitrarily defined  as  4 a  (signal:noise) and  16 a ,
respectively.   Thus, measurements  below  the  LOD  were reported  as  non-
measurables, and  those  between 4 a  and  16 a ,  as  trace  (T)  levels.   Only
numerical values  above 16 a were  reported.

Summary Statistics

     Tables 6  and 7 give the  halocarbon levels in personal and fixed-site
air  samples for  Greensboro  and  Baton  Rouge/Geismar,  respectively.    All
measurements for  the overnight and  daytime period are presented;  however,
only measurements above  the LOD  were  used to  calculate  the mean,  standard
deviation,  and median.   Trace values  were entered as the midpoint  between
the LOD  and the  QL in  these  calculations.  These summary statistics  for
fixed-site  and personal air  samples are  preliminary data.   They are  not
necessarily accurate  estimates   of  the  study   populations,   since   the
weighting  factors for the sample selection design have  not been  included.
Similarly,  the statistical  tests must  be  based   on  assumption of  simple
random sampling and thus are tenuous, given  the stratified  cluster design.
Nevertheless,  the  general  conclusions  are  sufficiently accurate  for  an

                                     272

-------
                                                   • Fixed Site
                                                   A Personal
                                                	±25%
                                                 	±50%
                       0.1
                         0.1
                                      1.0       5.0  10.0
                                       Duplicate No. 1 (fig/m3)
0.990
0.997
 Figure  4.   Replicate samples  for Tetrachloroethylene  - Greensboro, NC study
             (—  is 45° line).
                        5.0
                        1.0


                        5.0
                         0.1
                                               0.6   1.0
                                                           4.0
                                       Duplicate No. 1 (
Figure 5.  Replicate  samples  for   1,1,1-Trichloroethane
            study (— is 45°  line).
      -  Greensboro,  NC
                                         273

-------
                       10.0


                        5.0
                     J
                     CM

                     I
                     2
1.0


0.5
                        0.1.
                                                        Fixed Site   0.987
                                                        Personal    0.962
                                                        ±25%
                                                        ±50%
                         0.1
                                   0.5
                                       1.0
                                        Duplicate No. 1 (ng/m3)
                                                              5.0  10.0
Figure 6.   Replicate  samples  for  Chloroform - Greensboro,  NC  study  (-
             is  45° line).
                      _ 10.0
                      3:
                      CM
                         5.0
                         1.0



                         0.5
                         0.1.
                          '0.1        0.5   1.0       5.0  10.0
                                         Duplicate NO. 1 I
Figure 7.   Replicate  samples  for  Carbon  tetrachloride  -  Greensboro,  NC
             study (— is  45° line).
                                           274

-------
                                 0.5
                                     1.0
                                      Duplicate No. 1 (ng/m3)
                                                          5.0  10.0
Figure 8.  Replicate samples for  1,2-Dichloroethane  - Baton Rouge/Geismar,
            LA  study (— is  45° line).
                                                              0.991
                                             0.5  1.0
                                      Duplicate No. 1 (jig/m3)
                                                          5.0  10.0
Figure  9.  Replicate samples for  1,2-Dichloroethane
            (— is  45° line).
- Greensboro,  NG study
                                        275

-------
 CO
VO
            •o
             a
                       CM

                   O CM
                     —I rH CM

                     vO vO •—I O  V
             a
                   COOOOOO-3-CM
                  o
           OO     O O r-l vO    CO
            • co  in co  •   • co   •
           •-ICMCMCMeMCT\
    cu
    C  CU
    to -a
 CU  .fi -H
 C  4-J  M  a)
 to  cu  o  c
js  o t-i  cu

 0)  O "S  "?>
 O  r-l  td  X
 P  J3  rl  4-1
    §
    o  cu
    co  S
   •H  a)
      r-l
    CU  >,

cu  cu  4J
e  N  cu
cu  e  o
            cd
            u
 cu

6
          MH  o
          O i-l
                          1   C  i-l  O  O  O
                   MO-H  O.C  Ki-H  «
                   O   1    -J3  O  O  J3  14
                  I-ICM.-H  M-Ht-l  O4J
                                                  o
                                                  MH


                                                  CO
                                                  rt
                                                           cu
                                                           o
                                                           rt
                                     I-l
                                     o
                                    IM
                                                           cu
                                                           B
                                                           o


                                                           cu
                                     o
                                     cu
                                    4-1
                                     cu
                                           c.2
                                           rt
                                           1-1  CO
                                           73  4J

                                        cu  S  cu
                                        co      a
                                        3  -a  cu
                                           C  I-i
                                           cd  3
                                               to
                                            «  rt

                                           g  5!
                                                          S
   U  'H

1-3  °
I >  n
cu 01  ni
 • T3  43

2 T3  3
                                             "g  «

                                              n)  O
                                             4-1  O
                                             •  • CM
                                                                                                CMCMCMr-ICMr-HCMCM
                                                                                                cocococococococo
                                                                                                        o                   o



                                                                                                        OrHCMrHOr-tO'-l
                                                                                                   ON         vo     co •*
                                                                                                     •  i-H CO   • CO  •
                                                                                                        +1  +1 +1  +1  +1 +1  +1  +1
                                                                   -*

                                                                   o
                                                                                                        VO  CO
                                                                                                        ocMOcomoocO'-i
                                                                                                        —IO^HOOOOO
                                                                                                        oooooooo
                                                                                                          «  «o
                                                                                                        in  oo   «-
                                                                                                        in .in in  m in in in in
                                                                                                         cu
                                                                                                        •O
                                                                                                         S-l
                                                                                                         o
                                                                                               £1
                                                                                                o
                                                                                                                           cu
    cu
    d  cu
    rt T3
 CU  ^2 -H  .
 C  4J  (J  CU
 al  cu  o  r-l
43  O i-l  >,

    O  CJ
           cu
                                                                                                     CU
                                                                                                     o •-)  rt
                                                                                                     p X!  M  O
td -H
p.
o  cu
t-<  C!
p,  cu
O  N
                                                                                                                OG4JMOO)
                                                                                                            IH  o  H
                                                                                                            O -H   I
                                                                                             >-, o  .
                                                                                             Ci r-) CM r-l
                                                                                                    £  O  M

                                                                                                     O -H  O
                                                                                                     rt O r-l
                                                                                                     H  I  43
                                                                                                 p  4-1 CM  O
                                                                                                                 -"tflCU   « -rl
                                                                                                                                         s     •§
                                                                                                                                                o
                                                                                                                                                M-l
                                                                                                                                        CU
                                                                                                                                        O
                                                                                                                                        rt
                                                                                                                          O
                                                                                                                          •H
                                                                                                                                                O
                                                                                                                                                UH
                                                                                                                                     B
                                                                                                                                     cu

                                                                                                                                     3
                                                                                                                                     co
                                                                                                                                     rt

                                                                                                                                     i
                                                                                                                                                C!
                                                                                                                                                o
                                                                                                                                                cu
                                                                                                                                               13
                                                                                                                                         co      S
                                                                                                                                         CO     1-1
                                                                                                                                         CU     iH



                                                                                                                                         11     o

                                                                                                                                         CU     i-H
                                                                                                                                T)  cu
                                                                                                                                cu  B
                                                                                                                                CO

                                                                                                                                3"§
                                                                                                                                   g  1
                                                                                                                                            to  o
                                                                                                                                         to  T3  3
                                                                                                       •§"
                                                                                                       rt Q
                                                                                                       4J O
                                                                                                       CO 1-4
                                                                          276

-------
exploratory analysis  of  this  nature.  A weighted analysis  will be reported
elsewhere.  Although  the  list  of  target  compounds  monitored was larger than
presented  here,  these  halocarbons  were the  principal  ones   that  yielded
measurable  values.    For  each  halocarbon,  the  LOD was  very close,  if  not
identical,  between personal and  fixed-site  samples  within a  geographical
area.   However,  they differed significantly between  areas and  thus  such
comparisons are not made  here.

     As  indicated by  the means and  medians, personal  air  samples appeared
to yield higher absolute  levels than fixed-site samples within a geographi-
cal  area.   These trends  were,  of course,  also  reflected in  their  ranges.
Because  of these trends, we  examined  personal versus  fixed-site  sample
values in more  detail to  elucidate  the significance of  these data.

Comparison of Personal and Fixed-Site Data

     In  order  to uncover  trends  in these  data,   an  assorted  number  of
statistical  analyses  were performed,   first  examining  data  for  overall
trends;  then,  detailed  stratifications  of  data were  tested   for  signifi-
cance .

     The frequency  distribution for  all  air values (duplicates were aver-
aged) from  personal and  fixed-site monitors were  plotted for  a few  of  the
halocarbons.   Figures 10 through  12 depict  results   for  1,1,1-trichloro-
ethane,  tetrachlorethylene,  and  trichloroethylene,  respectively.   It  is
seen clearly from these frequency distributions  that  personal  samples yield
higher values over  the major  part of the  dynamic  range  than  do  fixed-site
samples for both  the  Greensboro and  Baton Rouge/Geismar areas.

     Figures 13  and 14  show bar  diagrams  for a comparison of the  percent
detected between  personal and  fixed-site air samples in  the Greensboro  and
Baton Rouge/Geismar areas, respectively.

     In  Figure  13, the   comparisons  are for  (left  to right)  chloroform,
methylchloroform,  bromodichloromethane,  tetrachloroethylene,  chlorotoluene
isomers, dichlorotoluene  isomers, trichlorobenzene  isomers, carbon  tetra-
chloride,  trichloroethylene,  1,1,2,2-tetrachloroethane,   dichlorobenzene,
1,2-dichloroethane, and  chlorobenzene.    The percent detected  for many  of
the  halocarbons  were  statistically  significant (indicated by  asterisks)
between personal  and  fixed-site  samples.  These data represent 53 personal
and  37  fixed-site samples.  Except  for  1,2-dichloroethane and  chloroben-
zene, the personal air sample exhibited  a higher percent  detected.

     The same phenomenon  was observed with personal and  fixed-site  samples
from Baton  Rouge/Geismar (Figure  14).   For many  halocarbons, the  percent
detected in personal  air samples was higher than  fixed-site and  statisti-
cally significant at  the 0.05  level.   The  reverse  trend was   observed  for
three  halocarbons,  1,2-dichloroethane,  1,1,2-trichloroethane,  and  vinyl-
idene chloride.   In  fact,  the fixed-site  percent  detected for  vinylidene
chloride was significantly  greater (0.05 level) than for the  personal  air
samples.  A point source  for these  three halocarbons was suspected  and  may
be responsible for this association.
                                    277

-------
     1000
      100
n3

10
      0.1
              FREQUENCY DISTRIBUTION
                   OF AIR SAMPLES
                FROM FIXED STATIONS
              AND PERSONAL MONITORS:
                1,1,1-TRICHLOROETHANE
            j	I
                                                 PERSONAL
                                                 EXPOSURE'S
                                                        FIXED
                                                        STATIONS
                                         • GREENSBORO H PERSONAL
                                         O BATON ROUGE) MONITORS

                                         • GREENSBORO ~"\ FIXED
                                         D BATON ROUGEJ STATIONS
                                11  I  I  II  I
Figure 10.
            12   5  10  20 30140 50 60 70 80 90 95 9899
           PERCENT OF SAMPLES LESS THAN CONCENTRATION

     Frequency distribution for 1,1,1-Trichloroethane.

                            278
                                                           1000
                                                           100
                                                                  10
                                                           0.1

-------
       100
                                                                 1100
               FREQUENCY DISTRIBUTION
                    OF AIR SAMPLES
                 FROM FIXED STATIONS
               AND PERSONAL MONITORS:
                TETRACHLOROETHYLENE
       10
        1 -
      0.1
     0.01
                                    PERSONAL
                                    MONITORS
                                    FIXED
                                    STATIONS
                                                                 10
                                                                  >u.g/m3
                                       9 GREENSBORO"! PERSONAL
                                       O BATON ROUGEJ MONITORS
                                       II GREENSBORO~| FIXED
                                       n BATON ROUGEJ STATIONS
                                             0.1
I  I   I—I   I  I  I  I  I  I
                  12  5  10  20 30 40 50 60 70 80  90  95  98 99
                 PERCENT OF SAMPLES LESS THAN CONCENTRATION


Figure 11.   Frequency distribution for Tetrachloroethylene.


                                   279
                                                                 0.01

-------
     20
     10
     0.1
    0.01
        FREQUENCY DISTRIBUTION OF AIR SAMPLES
                  FROM FIXED STATIONS
AND PERSONAL MONITORS:
   TRICHLOROETHYLENE
                                       PERSONAL
                                       MONITOFIS
                                                        FIXED
                                                        STATIONS
                           • GREENSBORO 1  PERSONAL
                           O BATON ROUGEJ  MONITORS

                           • GREENSBORO ~1  FIXED
                           [] BATON ROUGEJ  STATIONS
                                I  I  I  I
                                              I   I
                                                     _L
                                                                 20
                                                                 yu-g/m
                                                                 0.1
                                                 0.01
                  12  5  10  20 30 40 50 60 70 80  90 95 98 99
                 PERCENT OF SAMPLES LESS THAN CONCENTRATION
Figure 12.   Frequency  distribution for Trichloroethylene (dashed lines are
            extrapolations to the LOD).
                                   280

-------
                                              * = Significant at .05 level
                                            Fixed = 37 samples
                                         Personal = 53 samples
                                              a = Fixed-site
                                              • = Personal
   100

   90
   80

"S  70
s
0)  60

Q  50
"  40
   30

   20
   10
             0)
             o
             ^
             0>
             Q.
         *   I   I










if  ^^T^^^^^»^lrf^^^*^5^^^^™T?5T^r^^?'^^^^""?S^^^™™^^^^™
                   CF  MCF BDCM PERC CT  DCT  TCB  CTC TCE  STCE DCB DCE CB
                                   Halocarbons
                               personal  air
Figure  13.   Comparison of percent  detected  for fixed-site vs.
             samples -  Greensboro, NC.
             CF =  chloroform,  MCF =  .1, 1 ,1-trichloroethane,  BDCM =  bromodi-
             chloromethane,  PERC =  tetrachloroethylene,  CT = chlorotoluene,
             DCT = dichlorotoluene  isomers,  TCB = trichlorobenzene  isomers,'
             CTC = carbon tetrachloride,  TCE  = trichloroethylene,  STCE  =
             1,1,2,2-tetrachloroethane,  DCB = dichlorobenzene isomers,  DCE =
             1,2-dichloroethane, and  CB  = chlorobenzene.

                                       281

-------
                                         * = Significant at .05 level

                                       Fixed = 55 samples
                                    Personal = 132 samples
                                         n = Fixed-site
                                         • = Personal
            •o

            £

            
            o

            o
            Q.
CTC  TCE
                              MCF PERC  DCB  CF

                                    Halocarbons
                             DCP  DCE  TCA  VEC
Figure  14.  Comparison of  percent  detected for  fixed-site  vs. personal air

             samples - Baton Rouge/Geismar, LA.   See Figure  13 for  key,  in

             addition,  DCP  =  1,2-dichloropropane,   TCA  =  1,1,2-trichloro-

             ethane, and VEC = vinylidene chloride.


                                       282

-------
      The  question  of  percent  detected  between  personal  and  fixed-site
 samples was  examined  further.    The  data were  stratified  so  that   only
 matching pairs of samples were included in the  analyses.   The data for  the
 participant whose fixed-site  sample  was in his/her  yard  were matched  with
 the same  participant's  personal sample data;  the remaining  air data  from
 the "cluster" were excluded from statistical  analysis.   The data were  also
 grouped for analysis  according to sampling period.   Table 8 lists the  per-
 cent  detected for matching pairs for Greensboro.  For the  overnight period,
 a  significant difference  was  observed for  chloroform  and tirichlorobenzene
 isomers (0.05 and  0.01  level,  respectively).   Chloroform  (0.01),  1,1,1-
 trichloroethane  (0.05),  and  tetrachloroethylene  (0.05) were significantly
 different  between  personal and  fixed-site for the daytime period.    The
 trends   corroborate  the  previous  data  (Figure  13),  where  all  data  were
 included in  the  analyses.   Fewer results are  given  in  Table  8  than in
 Figure  13, since the reduced  sample  size  yielded fewer significant differ-
 ences.
        TABLE  8.   PERCENT DETECTION IN AMBIENT AIR SAMPLES MATCHED BY
       	PARTICIPANT - GREENSBORO, NC STUDY
                               Period no.  1*
Period no. 2*
Chemical Nt =
Chloroform
1 ,2-Dichloroethane
1,1, 1-Trichloroethane
Carbon tetrachloride
Trichloroethylene
Bromodichlorome thane
Tetrachloroethylene
Dichlorobenzene isomer
Trichlorobenzene isomers
Fixed-site
18
56
33
67
44
44
6
67
39
0
Personal
18
94$
33
94
61
61
33
94
44
61§
Fixed-site
16
19
37
50
44
31
6
62
31
12
Personal
16
94 §
25
94$
50
37
19
100$
50
21

*Period no. 1 =  1800-0700  hrs;  Period no.  2  = 0700-1700 hrs.
tField samples included  in sample  size.
$Significant  differences between the  two  types  of  air samples  at  the 0.05
 level.
§Significantly different at  the 0.01  level.
     Table 9 gives  similar information for Baton Rouge/Geismar.   An inter-
esting observation  is  that vinylidene chloride was  significantly  different
(0.01) between fixed-site  and  personal  air,  for  both sampling periods.   The
trends were similar to  those in  the  data  discussed  earlier for all measure-
ments (Figure 14).

     Further  analyses   were  performed,  specifically  to  determine  whether
correlations between the overnight and  daytime periods existed for personal
and  fixed-site  air samples.   Spearman correlations  were  used,   since  the
data were highly skewed.
                                    283

-------
       TABLE 9.  PERCENT DETECTION IN AMBIENT AIR  SAMPLES  MATCHED  BY
                 PARTICIPANT - BATON ROUGE/GEISMAR, LA  STUDY	
Chemical Nt =
Vinylidene chloride
Chloroform
1 ,2-Dichloroethane
1,1, 1-Trichloroethane
Carbon tetrachloride
Trichloroethylene
Tetrachloroethylene
Dichlorobenzene isomer
Period no. 1*
Period
Fixed-site Personal Fixed-site
26 26 24
81 23
19 15
100 96
69 54
58 77
8 69:
35 731
8 881
71§
8
96
48
48
: 8
I 18
i: 12
no. 2*
Personal
24
37
25
96
67
65
71::
64:
100?
*Period no. 1 -  1800-0700 hrs;  Period  no.  2  = 0700-1700 hrs.
tField samples included  in  sample  size.
$Significantly different at  the 0.01 level.
§Significant differences between the two  types  of  air samples at  the  0.05
 level.
     Table  10  lists   the   significant   Spearman  correlations  that  were
observed for each of the halocarbons  statistically analyzed.   Remarkably,  a
number  of  halocarbons  exhibited  significant  Spearman  correlations  for
personal and fixed-site samples between the overnight  and  daytime periods,
both in Greensboro  and Baton Rouge/Geismar.   The  highest  Spearman correla-
tion  observed  was  for carbon tetrachloride  (0.70)  in  personal air  from
Greensboro.   This  suggests  that  the  levels  between overnight  and  daytime
periods were related to each other.

     The  significant  Spearman correlations  found  between fixed-site  and
personal air  samples are given in Table  11.   These correlations indicate
that,  for  certain  compounds,  the  values  for matched  fixed  and personal
samples track  reasonably  well with each other, i.e., as  one  increased,  the
other also  increased or vice-versa.   The best evidence of  this phenomenon
was for carbon tetrachloride (0.71)  in the Baton  Rouge/Geismar data.

     The  linearity  between  personal  and  fixed-site  values for  each study
participant was more evident when plotted as  shown, for example, in Figures
15  and  16  for  carbon  tetrachloride  and  1,2-dichloroethane,  respectively.
As  the  measurable  values  approach  the detection  limit  of  the analytical
method, the variation  appears to  increase.

     Analyses  of  personal and fixed-site data suggest that  personal moni-
toring  gives higher exposure  values  for  individuals.   Secondly, fixed-site
samples  for a  selected number of  halocarbons  do  track  the  personal  air
samples, while the  absolute values appear  to  be  lower.   These data suggest
that  personal  samples  may  better represent the  exposure  of  individuals to
indoor  air  pollution,  where many  of  the participants spend a higher percen-
tage  of their  time in  a 24-hour  period.   Again,   attention  should be drawn

                                     284

-------
            TABLE 10.   SIGNIFICANT* SPEARMAN CORRELATIONS BETWEEN
                       PERIOD NO. 1 AND PERIOD NO. 2 MEASUREMENTS

Sample type
Fixed-air




Personal air









Greensboro
Chemical
1 , 2-Dichloroethylene
Carbon tetrachloride
1 , 2-Dichloropropane
Trichloroethylene
Vinylidene chloride
Chloroform
Carbon tetrachloride
Trichloroethylene
Tetrachloroethylene
Dichlorotoluene isomers
Vinylidene chloride
1 , 2-Dichloroethylene
1,1, 1-Trichloroethane
1 , 2-Dichloropropane
Dichlorobenzene isomer
Corr.
0.58
0.53
0.52
0.53
— — °!°
0.46
0.70
0.42
0.41
0.41
—
—
—
—
—
NT
2.0
20
20
20
—
24
24
24
24
24
—
—
—
—
—
Baton Rouge/
Geismar
Corr.
0.44
0.51
0.53
—
0.48
_.—
0.29
0.47
0.37
—
0.31
0.35
0.43
0.41
0.35
N
27
27
27
—
27
——
59
59
58
—
59
59
58
59
59

TNumber of measurements.
$Not significant  at  0.05  level.
      TABLE  11.  SIGNIFICANT  SPEARMAN  CORRELATIONS  FOR FIXED-SITE VS.
                 PERSONAL AIR SAMPLES*

Geographical area
Baton
Period no.t
1



2



Chemical
Carbon tetrachloride
Trichloroethylene
1 , 2-Dichloroethane
Dichlorobenzene isomer
1 , 2-Dichloroethane
Trichloroethylene
Chloroform
Carbon tetrachloride
Greensboro
0
0

0
0
0


.68
.52
— §
.49
.60
.49
—
—
(20)$
(20)

(20)
(18)
(18)


Rouge/
Geismar
0.

0.

0.
0.
0.
0.
71
—
69
—
67
41
45
47
(27)

(27)

(25)
(25)
(25)
(25)

tPeriod no. 1 = 1800-0700 hrs; Period no. 2 =  0700-1700 hrs.
^Number of samples given in parenthesis.
§Not significant at 0.05 level.
                                    285

-------
                               O.S  1.0
                                     Fixed Site (M/m3)
                                                       5.0  10.0
Figure 15.   Spearman Correlation for personal vs.  fixed-site air  levels of
             carbon  tetrachloride  (overnight   sampling  period)   -  Baton
             Rouge/Geismar, LA study (45° line depicted).
                     0.1.
                               0.5   1.0
                                                       5.0  10.0
                                     Fixed Site (jig/m3)
Figure  16.   Spearman Correlation for personal  vs.  fixed-site air  levels of
             1,2-dichloroethane    (overnight   sampling   period)   -   Baton
             Rouge/Geismar, LA study (45° line  depicted).

                                      286

-------
to  the situation  that the  overnight  personal  sample  can,  for  practical
purposes, be considered as an indoor  fixed-site  sample  and  thus  indoor/out-
door air quality can be directly compared  for matched  samples.

     Some caution, however,  should be noted in evaluating  these  data.   For
example,  the  comparison  of  overnight  and  daytime   period  samples   for
personal samples  may  be confounded  by the inclusion  of housewives  in  the
statistical analyses,  and  thus,  significant correlations between overnight
and  daytime samples  are  heavily  reflecting of  exposure  near  and  inside
their home.  Because the sample participant  size  is  not  sufficiently large,
further post-stratification  of the data could not  be performed.
                                  SUMMARY

     Personal  and  fixed-site  air  monitoring  was   conducted   to   assess
exposure  of  lay  populations  to  halocarbons.     Twenty-eight   people  in
Greensboro,  North   Carolina,  and  66  people   in  Baton   Rouge/Geismar,
Louisiana, participated in a  24-hour sampling program.  A QC/QA  program was
maintained for the sampling and analysis procedures.   This  included  blanks,
controls, and replicate field samples.

     The significant findings were:

     1.   mean recovery of halocarbons  from Tenax GC cartridges  were  79 to
          125 percent (except chloroform, 63 percent);

     2.   for  personal  and  fixed-site  air  samples,   between  replicate
          variability  of  sampling  and  analysis  was  ±25  percent   for
          1,1,1-methylchloroform  to ±50 percent for chloroform;

     3.   the  levels  of halocarbons  in personal  air  samples  were  higher
          than fixed-site samples;

     4.   a  majority   of  the   halocarbons  exhibited   a  statistically
          significant higher  percent  detected in  personal than  fixed-site
          air samples within  each geographical area;

     5.   statistically  significant  correlations  (at   0.05  level)   were
          observed    for    carbon    tetrachloride,     trichloroethylene,
          1,2-dichloroethane,   dichlorobenzene   isomer,    and    chloroform
          between personal and fixed-site air levels,

     6.   statistically significant correlations (at 0.05 level)  were  found
          between overnight and daytime period  halocarbon levels  (five  for
          fixed-site and 10 for personal).
                             ACKNOWLEDGEMENTS

     The authors wish to thank  the  many people who have contributed to  the
success of this study,  and  especially the participants who graciously gave

                                    287

-------
their time.   This research was  supported by  EPA Contract Nos.  68-01-4731
and 68-01-3849.
                                REFERENCES

1.   Zweidinger, R.A., S.D. Cooper, B.S.H. Harris,  III,  T.D.  Hartwell,  R.E.
     Folsom, Jr., E.D. Pellizzari, A.W.  Sherdon,  T.K.  Wong,  and  H.S.  Zelon.
     1980.  Measurement  of  benzene body-burden for populations  potentially
     exposed to benzene  in  the  environment.  EPA-560/13-80-028,  December.

2.   Wallace, L.,  R.  Zweidinger,  M. Erickson,  S.  Cooper, D. Whitaker,  and
     E.D. Pellizzari.  In press.   Monitoring individual  exposure—measure-
     ments  of  volatile  organic  compounds in  breathing-zone air,  drinking
     water, and exhaled  breath.  Environ. Intl.

3.   Pellizzari, E.D.,  T.  Hartwell, H.  Zelon, C.  Leininger, M.  Erickson,
     and  C. Sparacino.    1982.   Total   exposure  assessment:  methodology
     (TEAM):  prepilot  study -  Northern New  Jersey.   U.S.  Environmental
     Protection Agency Final Report.   Contract No.  68-01-3849.

4.   Pellizzari, E.D., M.D. Erickson,  and R.A. Zweidinger.   1979.   Formula-
     tion of a preliminary assessment of halogenated organic compounds  in
     man and environmental  media.  EPA-560/13-79-006,  July.

5.   Huffman,  R.D.,  C.M. Latanick,  T.K. Collins, J.A.  Caldwell,  and  J.D.
     Wiese.   1979.   Metabolism summaries of  selected halogenated  organic
     compounds  in  human and  environmental  media,  a  literature  survey.
     EPA-560/6-79-008, April.

6.   Pellizzari, E.D., M.D. Erickson,  and R.A. Zweidinger.   1979.   Analyti-
     cal  protocols  for   making a  preliminary  assessment  of  halogenated
     organic compounds in man and environmental media.   EPA-560/13-79-010,
     September.

7.   Krost, K.J.,  E.D.  Pellizzari, S.G.  Walburn,  and S.A. Hubbard.   1982.
     Collection  and  analysis  of hazardous organic  emissions.   Anal.  Chem.
     54:810.
                                     288

-------
       A PERSONNEL  OR AREA DOSIMETER FOR POLYNUCLEAR AROMATIC VAPORS
                                 T.  Vo-Pinh

                    Health  and  Safety Research Division
                        Oak  Ridge National Laboratory
                                Oak  Ridge, TN
                                  ABSTRACT

     A  new passive  dosimeter has  been developed  for monitoring  airborne
vapors  and liquid aerosols  of potentially  hazardous polynuclear  aromatic
compounds.  The  device is a self-contained,  badge-size  unit  that passively
collects on filter paper the compounds to be monitored  at  a  diffusion-con-
trolled  rate.   Collection  is followed by  in-situ, room temperature phos-
phorescence analysis  of  the  compounds  adsorbed  on the dosimeter.   The dosi-
meter can  detect pyrene, phenanthrene,  and quinoline at  sub-part-per-bil-
lion for an 8-hour exposure.
                                INTRODUCTION

     Polynuclear  aromatic  (PNA)  compounds  are  produced  primarily  as  a
result of incomplete  combustion of  organic  matter.   They are believed to be
present  in  the atmosphere in many  industrial and  residential  environments
as both  vapor  and  aerosol.   With the advent  of increasing  public awareness
of the potentially long-term health hazards  associated  with  PNA compounds,
great emphasis has  been  placed  on  the development  of new instrumental moni-
toring techniques  to detect  those compounds  in  the atmosphere.   However,
because  of  their  low equilibrium  vapor concentration,  vapors  from  multi-
ring PNA compounds  are difficult to  detect  by simple methods.  Conventional
procedures  to  monitor these  species involve  1) collection  of PNAs  by draw-
ing large volumes  of air  through a  sorbent  material, 2)  thermal or  chemical
desorption  of  the PNAs,  3)  chromatographic  fractionation  of  the  samples,
and 4) identification and quantification of the extracted materials.   These
procedures  are usually   elaborate,  time-consuming,  and not  suitable  for
routine  applications or  field measurements.

     This paper reports  on the  development  of a new type of personnel dosi-
meter that has been recently developed  for  monitoring select  PNAs in  vapors
and liquid aerosols.  This dosimeter collects the PNA compounds via molecu-
lar diffusion  and  sorption.   Collection is followed by  direct  identifica-
tion  and quantification of  the  analyte   compounds via  room  temperature

                                    289

-------
phosphorimetry (RTF).  There  is  no  requirement  for a prior chemical extrac-
tion or chemical desorption protocol.   The results of the evaluation of the
dosimeter for various homocyclic as  well as heterocyclic PNAs during labor-
atory-controlled  experiments  and  field-test  measurements  are  presented
here.
                   A. PASSIVE  DOSIMETER FOR PNA COMPOUNDS

     The  PNA dosimeter  is a self-contained,  badge-size passive  monitor.
The device is lightweight  (^200g)  and can be  conveniently worn  by  a person
or placed at a  stationary location.   Figure 1 shows  a photograph  of  the
pen-size  dosimeter  worn on the  pocket of a  person  at a synfuel  facility.
The monitor  basically  consists of a holder,  a filter paper  substrate,  and
an interchangeable  diffusion chamber.   The heart of  the dosimeter  is  the
sample collection substrate of filter  paper  that  is treated with  a heavy-
atom chemical (1-3).   The  monitoring procedure is based  on  the  measurement
of the quantity  of the PNAs transferred to the substrate surface via molec-
ular diffusion  in  air.  The  unique  feature  of this  dosimeter  is  the  dual
use  of the  heavy-atom  chemical both  as a  sorbent agent  and  as  an  RTF
inducer.   The sorbent  material (paper/heavy-atom  chemical) maintains  the
concentration of the PNA  compounds  at  the collection  surface  at  zero  or
near-zero concentrations while the air outside the  dosimeter is at ambient
concentration.   This sets  up  a concentration  gradient for diffusion of  the
compounds from the outside of the  dosimeter to the  inside.   Interchangeable
diffusion chambers  of  different  sizes can be used  to  select  the  rate  of
molecular diffusion.  The  concentration gradient  provides the driving force
to move the PNA  molecules  onto the paper, eliminating the need  for a pump.
The transfer of  the PNA  compounds  by diffusion is  described  by Pick's First
Law:
                              T  -  n dc
                              J  ~ ~D dl
(1)
where

    J « diffusion flux

    D = coefficient  of  diffusion of  the  PNA compounds

   dc s concentration gradient  dc along  diffusion path dl.
   dl

     If the  concentration gradient  is  constant,  (dc/dl = C/L  = constant),
the mass of PNA compounds  collected  at  the sorbent surface  is given by:
                           M = D •   ' c '
(2)
                                     290

-------
Figure 1.  Photograph of  the  PNA dosimeter worn  by a worker  at a synfuel
           facility.
                                    291

-------
where

    A - collection area  of  the  dosimeter

    L - diffusion path length of  the  dosimeter

    C = ambient concentration

    t = exposure time.

     After  exposure,  the dosimeter  is inserted  into  a  luminescence  spec-
trometer for identification and quantification of the  PNAs collected on the
monitor using the RTF technique (4).


                                RTF DETECTION

     The RTF method  is  a relatively  new approach in  phosphorescence  based
on  the  emission of organic  compounds adsorbed on various  solid substrates
such as  filter  paper, silica gel,  sodium acetate,  etc.   At  room tempera-
ture, phosphorescence is normally a very  weak emission that is difficult to
detect  in  liquid  solutions  or  in the gas phase.  This is due  to the fact
that the phosphorescence is almost totally  quenched by collisions in  solu-
tions or air, or is deactivated by intramolecular vibrations  arid rotations.
The conventional phosphorescence  technique,  therefore, requires low-temper-
ature  equipment and  frozen solvents to  reduce  the  probability  of  these
quenching  processes  so   that the  phosphorescence  signal may  be  more easily
detected.   Unlike  conventional phosphorimetry,  the  RTF technique does  not
use cryogenic technology.    This  feature  is  one  of  the main  attributes of
this method for routine  applications  and  field measurements.

     The  paper  substrate  of the dosimeter  is  treated  with  a heavy-atom
chemical  that  is  used   to  increase  the  adsorption  characteristics of  the
collection  surface and  to enhance the RTF emissions  of  the  PNA compounds.
This latter process, known  as external heavy-atom perturbation, provides an
invaluable aid to RTF detection (4).   The presence  of  heavy-atom species in
the vicinity of a molecule  can enhance  a photophysical process,  known as
spin-orbit  (S-0) coupling,  which is  responsible  for  phosphorescence  emis-
sion.   Qualitatively,  the S-0  coupling  arises from the  interaction of  two
magnetic fields  resulting  from the  nuclear  and  electron spin  motion (5).
Since the  magnitude  of  the nuclear magnetic  field is  directly proportional
to  the  nuclear charge   and  hence to  the atomic number,  the  S-0 coupling
increases with increasing atomic  number.   The method  based on the S-0 cou-
pling enhancement  by  the heavy-atom effect has been developed into a  prac-
tical and  simple  tool of considerable analytical interest in the areas of
air pollution analysis   (6).  Table  1 .shows  that the   detection limits  for
several homocyclic and  heterocyclic PNA  compounds  by  RTF are  in the  pico-
gram range.
                                     292

-------
        TABLE  1.   LIMITS  OF DETECTION (LOD)  FOR SEVERAL PNA COMPOUNDS
                  BY  ROOM TEMPERATURE PHOSPHORESCENCE

Compound
Xex*
(nm)
Aemt
(nm)
LOD
(ng)
Homocyclics

   Benzo[a]pyrene
   Benz o[e]py rene
   2,3-Benzofluorene
   Benzo[ghi]perylene
   Chrysene
   1,2,3,4-Dibenzanthracene
   1,2,5,6-Dibenzanthracene
   Fluoranthene
   Fluorene
   Phenanthrene
   Pyrene

N—Heterocycllcs

   Acridine
   5,6-Benzoquinoline
   7,8-Benzoquinoline
   Carbazole
   Dibenzocarbazole
   Quinoline	
395
335
343
398
330
295
305
365
270
295
343
360
355
353
296
295
315
688
543
505
626
518
567
555
545
428
474
595
660
502
509
415
475
505
0.07
0.001
0.025
0.6
0.03
0.08
0.005
0.05
0.2
0.007
0.1
0.4
0.06
0.04
0.005
0.002
0.1
*Aex = excitation wavelength.
tAem = emission wavelength.
                 MONITORING PNA VAPORS  AND  LIQUID  AEROSOLS

     The  detection of a  typical  PNA compound,  phenanthrene,  by  the  dosi-
meter  is  illustrated  in Figure  2.   This  figure shows  the  RTP  spectrum
obtained after exposing the dosimeter for  two  hours  in a chamber containing
phenanthrene vapor  at  40°C.   The RTP response due to the  background  emis-
sion  of  the paper  substrate of  an unexposed  dosimeter is also shown  in
Figure 2  (dashed  curve).   A typical response  of the dosimeter  exposed  for
various time periods to vapors of another PNA  compound,  pyrene,  is given in
Figure 3.

     The  dosimeter is able  to detect  homocyclic as  well  as heterocyclic
compounds,  such  as  the   important  family  of  aza-arenes.   Aza-arenes  are
widely distributed in the  atmosphere  and  studies  have shown  significant
levels in marine  sediments,  tobacco smoke,  urban particulates,  automobile
exhausts, synfuel plants, and  effluents from many industrial sources.   The
dosimeter can detect  various  aza-arenes such as quinoline,  phenanthridine,
and  acridine  in  the  vapor  phase.    From  the  standpoint  of  human health
                                    293

-------
        12
        10
      >:  8
      o>
      
-------
                                            CO
                                                       i-l
                                                       O
                                                       ft
                                                       CD
                                                       d
                                                       tu
      O

      CO

      O
     •H

      <1)
      ft

      CU


•s   i
 £   o

 o    B-
 O    a)

 S    g
 Q-    o
 f n   'H
                                                 CO
                                                      0)
                                                      4J
                                                      W
                                                      O
                                                      CO

                                                      g
                                                      ft
                                                      CO
                                                      cu
                                                      t-i
                                                      cd
                                                      a
                                                     •H
  1VN9IS dlH
                                                     CO


                                                      0)
                                                      Mi
                                                     •H
295

-------
consideration, quinoline  is  a very  important  compound.   This  compound  was
found to induce hepatoma  in  rats and to increase  the  carcinogenic  activity
of  benzo[a]pyrene (7).   Current  methods  for  detection  of  quinoline  are
elaborate,  involving  active  pump  filtration,  desorption,  fractionation of
collected  samples,  and  analysis  by  chromatographic  techniques.    Since
quinoline has  a  relatively high vapor  pressure and  is  easily  desorbed by
forced-air  flow,  the  values  reported  for  this compound may  well be  far
below the true ambient  concentrations.

     The  dosimeter can effectively  monitor quinoline in the  vapor  phase.
The detection  limit for quinoline  is 0.75 ppb  x hr.   Figure 4 shows the RTF
spectra of  quinoline  and  isoquinoline detected by  the dosimeter.  The 15-nm
spectral  shift in the  RTF spectra shows that  it  is possible  to  differen-
tiate quinoline from  its  isomer.  When  these  two compounds are both present
in  a mixture,  the specificity  to  characterize  them  individually  can be
further enhanced  using the synchronous scanning method  by which both emis-
sion  and  excitation   wavelengths   are  simultaneously  varied  (8).    The
synchronous  scanning  method  results  in much  sharper  emission bands  and
allows  greater selectivity.   The  band-narrowing effect  of  the synchronous
excitation  method is  illustrated  in Figure  5 for quinoline  and  isoquino-
line.

     The PNA dosimeter has  been used to monitor a single component as well
as  various  compounds  under  actual  field-monitoring  situations.   Figure 6
illustrates the  capability of  the dosimeter  for  multi-component  detection
during  field test measurements.   The figure  shows the  RTF  signals of  four
sets of dosimeters exposed at four  different  areas  at a synfuel production
plant.  Three  dosimeters  were used  per  set, and  the variation of dosimeter
responses within  one  set  was  typically  15 percent.  The RTF response of  the
dosimeters  placed in a clean room showed a broad  emission.   This emission
was  similar to that  of a background RTF signal of  an  unexposed dosimeter
and  indicated  that the location had no  detectable levels of PNA compounds.
In  contrast,  the dosimeters  placed at  the  location near  a  fractionator
(FTR)  revealed the predominant  presence of phenanthrene.   Other compounds
detected  at  this  location  included  fluorene  and  pyrene.    Fluorene  and
phenanthrene were  also  detected  by the  dosimeter  exposed near  a vacuum
tower  (VAC TWR).   Finally,  the dosimeters  exposed at  a  location near  the
bottom of  a  fractionator (FTR-BOT)  detected  fluorene  and  quinoline.    The
detailed  results  of  these  field  measurements  will be  reported  elsewhere
(currently  in preparation by the author and G.H. Miller).
                                 CONCLUSION

      The development of personal monitors is an essential  step  in  the  quan-
 tification of  human exposure  for  health protection,  epidemiological,  and
 regulatory purposes.  Whereas  active samplers  can provide real-time detec-
 tion of hazardous pollutants,  these  devices  often are not the  most practi-
 cal and  economical  choice  for  monitoring individual  exposure to airborne
 pollutants.   Passive  dosimeters  offer  the  advantages  of  lower capital
 expense, simplicity, and compactness.   The dosimeter  described in  this work
 is a valuable tool for monitoring personal exposures  to PNA  pollutants that

                                     296

-------
          11
          10
      c
      05
      CD
      >
      JD

      0)
                                                                  N       —
                    J	I	I	L
                                                              J	I     I
               400
                                    500



                                   WAVELENGTH (nm)
                                             600
Figure 4.
Room  temperature   phosphorescence   spectra  of  quinoline  and
isoquinoline.
                                      297

-------
            12
           10
            _
         §  8
         
         ta
         
-------
    11
    10
C/3
c
0)
0>

'
JO
0)
     5
p.   4
DC
    0
              PNA DOSIMETER
                Aex = 290 nm
                                                        SYNFUEL PLANT
                                                        FTR
                                                      —  FTR-BOT
                                                	—  VACTWR
                                                         CLEAN ROOM
          400
500                  600


  WAVELENGTH (nm)
                                                                        700
 Figure 6.,  RTF  response  of  various dosimeters  exposed  at different  loca-
            tions inside a synfuel plant.
                                     299

-------
can  provide critical  information  to  assess   the  health  implications  of
exposure to these compounds.


                              ACKNOWLEDGMENTS

     Research  presented here  was  sponsored  by  the  Office  of  Health  and
Environmental Research,  U.S.  Department of Energy,  under contract  W-7405-
eng-26 with the Union Carbide Corporation.


                                REFERENCES

1.   Vo-Dinh, T.  1981.  InTech 5:45;  also, 1982,  U.S.  patent pending.

2.   Vo-Dinh, T., and  G.H.  Miller.   1982.   Proceedings of  the  1982 Pitts-
     burgh Conference,  Atlantic City,  NJ.

3.   Vo-Dinh,  T.    1981.    Proceedings of  the  1981  Instrument  Society  of
     America, St. Louis, MO.

4.   Vo-Dinh, T., and J.R.  Hooyman.  1979.   Anal.  Chem.  51:1915.

5.   McGlynn,  R.,  T.  Azumi,  and Kinoshita.   1969.    The   triplet  state.
     Academic Press, New York, NY.

6.   Vo-Dinh,  T.,  R.B.  Gammage,  and  P.R.  Martinez.    1981.    Anal.  Chem.
     53:253.

7.   Dong, M.,  I. Schmeltz, E.  Lavoie, and D.  Hoffmann.   1978.   Page 97 in
     Carcinogenesis.   P.W.  Jones  and R.I.  Freudenthal,  eds., Raven Press.

8.   Vo-Dinh,  T.   1981.   Page  167  in  Modern Fluorescence Spectroscopy.
     E.L. Wehry,  ed.,  Plenum Press.
                                     300

-------
                  THE NBS PORTABLE AMBIENT AEROSOL SAMPLER


                R.A.  Fletcher,  D.S.  Bright, and R.L. McKenzie



                                  ABSTRACT

     A  portable,  lightweight, battery-powered  particle  sampler  for col-
lecting ambient level concentrations of inhalable particles has been  devel-
oped  at  NBS.   The  unit has  a flow  rate of 6  liters  per minute  and  is
capable of  operating for longer  than  24 hours on a  single battery charge.
It  separates  and collects the  ambient  inhalable  and  respirable particulate
size  fractions  by series  filtration.    The  first filtration  stage,  an 8ym
pore size Nuclepore* filter  collects the "coarse" fraction of the inhalable
particles  (>_ S.Sym aerodynamic particle diameter).   The  smaller particles
«  3.5ym or  respirable fraction), which pass  through the first filter, are
collected  by  a  high-efficiency Zeflour  filter.   Both filtration stages are
weighable  to  ±10yg  certainty   and  are  amenable  to   subsequent  chemical  or
physical  analysis  of particulate  material.    The  sampler  inlet  removes
particles  larger  than  15ym  aerodynamic  diameter   (or  other  preselected
sizes) by  impaction.   The sampling efficiency of  the inlet has been  tested
in  a wind  tunnel  using a series of  monodisperse  test  aerosols.  Wind  tunnel
tests  showed that  the  sampling  efficiency  of  the  inlet  has  some wind
velocity  dependence.    Design .and  use  of   the  wind  tunnel  testing  is
discussed.

     The  sampling protocol, which is  well  suited  for  indoor  particulate
monitoring, is  also  useful for N02  pollutant  monitoring.   A common  indoor
gas pollutant is  N02, which  results  from gas-fired  flame sources.   Test
results for  a  passive N02  sampler that  could  be fitted  to  the  portable
particle sampler  is  presented.


                                INTRODUCTION

     The increasing  concern  for measuring personal exposure  to  particulate
and gaseous pollutants (1) and the  recognition  that such  exposures  cannot
*Certain commercial  equipment,  instruments, or materials are  identified in
 this report  to  specify  adequately the experimental procedure.   Such iden-
 tification  does  not imply  recommendation or endorsement  by the  National
 Bureau  of  Standards, nor  does  it  imply  that the  materials or  equipment
 identified are necessarily  the  best  available for  the purpose.

                                    301

-------
always  be  extrapolated  from outdoor  monitoring  measurements  (2,3)  has
resulted in the  need  to develop personal exposure  samplers  for indoor use.
The NBS portable ambient  aerosol  sampler  (Figure 1) was developed with this
application in mind.   It  is a small, quiet,  unobtrusive  sampler capable of
collecting analytically significant  samples  in sampling periods as short as
8 hours (depending  on average ambient  concentration levels).   The samples
collected are  quantified gravimetrically and can  subsequently  be  analyzed
by other  chemical  or  physical analytical  techniques  (for  example,  x-ray
fluorescence, microscopy,  ion chromatography).
                                   -Inlet
                             • Air Flow
                                                   Exhaust
Figure 1.  Schematic  of  the  sampler.
     The following  are  some of  the characteristics of the sampler:

a.   6 liter  per  minute (L/min) flow rate  stable  to  within 10 percent
     (up  to 350-400ug  mass loading) for  in excess  of  24-hour opera-
     tional time  period;

b.   Particle  size  separation by means  of tandem filtration, with sam-
     ple  collection in the  range of 15-3.5ym  (or  10-3.5)  aerodynamic
     diameter  and <3.5ym  aerodynamic diameter;

c.   An  inlet  that  samples  only  particles  <15ym  (or  optionally,
     .<10um);
                                     302

-------
d.   Collection  of an  analyzable  sample  during  an 8-hour  integrated
     sampling  period under normal  ambient particulate  concentrations
     (4-6).

The  sampler components  have  been  extensively tested  to characterize  the
sampler  performance,  including the  pump-battery  system, the  tandem filter
unit, and  the  inlet using a  wind  tunnel to determine  sampling efficiency.
These tests  are  discussed in detail.   Additional  information is available
in  references  4,  5,  and 6,  and  a  components list  is  available  from  the
authors.

     Because pollutant  gases  such as N02,  are  frequently present (1) in the
indoor environment,  we  evaluated  the feasibility of  incorporating  N02 sam-
pling  capabilities  into  the  portable  aerosol  sampler  for  use in indoor
pollution  monitoring situations.    A  small passive N02 sampler with high
sensitivity  can  be used with the aerosol  sampling  unit  that  is  capable of
collecting  measureable  quantities   of  N02  during  an  8-hour exposure  to
ambient  level  concentrations.

     The following discussion describes the performance characteristics of
these various  components  of the NBS  portable ambient aerosol sampler.


  DESCRIPTION  AND  EVALUATION  OF THE  NBS PORTABLE  AMBIENT AEROSOL SAMPLER

Tandem Filter

     The NBS portable  sampler achieves a particle  size  separation  (50 per-
cent cut)  at approximately 3 to 3.5ym  aerodynamic diameter using  a tandem
filter unit (7,8)  to collect  the respirable particle fraction « 3.5ym)  (9)
of the inhalable particles (<_ 15ym).   The  filter  unit  is composed of an 8ym
pore-sized  37-mm Nuclepore filter  in  series  with  a highly  efficient (10)
3ym  pore-sized 37-mm Zeflour filter.    The Nuclepore  filter  is  precoated
with a  small  amount  of  Apiezon L   grease  to  reduce  particle bounce  (8).
Particles  in the 15-3.5ym aerodynamic  diameter size range are collected by
the Nuclepore, while the  <_ 35ym aerodynamic  diameter  particles  penetrate
that filter and are  subsequently  collected  by  the Zeflour  filter.   The
particles  are  uniformly  distributed on both  filter  surfaces,' making  the
samples  amenable  to  analysis by  x-ray fluorescence.    The  tandem filter
arrangement has  the  practical advantage of being convenient,  low cost,  and
good for gravimetric analysis  due to low water retention.  This filter unit
has high flow  conductance at  6L/min (drawing  only  15  cm H20), which is an
important  factor in determining battery power and pumping  energy  require-
ments.   Two disadvantages are that  the  50  percent  cut-point  at 3.5pm is  not
sharp,  and  that  filter  clogging  for heavy  particle  loadings may  change  the
sampler  flow rate  and  collection  efficiency.   These two disadvantages will
be discussed later.

     Cut-point tests  on the Nuclepore  filter  were made  using monodisperse
solid  (ammonium fluorescein  and  ammonium sulfate)  and  liquid  (dioctyl-
phthalate-DOP)  particles  generated  by  a vibrating  orifice  particle gener-
ator (VOPG)  (11).   An  optical particle counter  was used  as  the  particle
                                     303

-------
detector downstream of a 37-mm filter  holder  (12,13).   The number of parti-
cles of known aerodynamic diameter  penetrating the filter was  compared to
the number of test particles  of  the  same  aerodynamic diameter measured with
the filter  removed from  the holder.   (Particle  aerodynamic diameter  was
determined by particle  sedimentation measurements in air.)   The collection
efficiency of the filter is  defined  as

          _ number of particles  through holder with filter in place
            number of particles  through holder without  filter
     The  results of  the filter  collection efficiency  tests  are  shown in
Figure 2 for clean and  preloaded  (with >150yg  of particulate matter) Nucle-
pore filters.   Liquid (sticky) particles are  more  efficiently collected by
the Nuclepore  filter  than  the  solid (resiliant) particles because the solid
particles "bounce" off  the filter surface and  thus  have  a better chance of
flow reintrainment  that penetrates  the filter  (12,13).    Liquid particles
are collected  at  50 percent  efficiency for  2.3pm-diameter particles, where-
as only 35  percent  of the solid  particles  of  that  diameter are collected.
The  collection efficiency  curve for  preloaded  filters  is shifted toward
smaller particle  size for  liquids,  probably because the effective pore size
of the filter  is reduced by the  material deposited on  the filter  surface.
The  shift  is  not  so  pronounced  for  solid  particles,  perhaps  because the
deposited  particles  enhance  the  bounce  effect  by  covering   the grease
layer.

     The  cut-point  differences  reported above  will be  somewhat moderated
when  sampling  ambient  particles,  as  they  will  often be  characterized by
properties  intermediate to those of  liquid and  solid test aerosols.   This
will minimize, to some  extent, the importance  of the  variability in the 50
percent  cut-point.   A  second  factor  that  will minimize  the  importance of
this variability is  that  there  is  a minimum in most  naturally occurring
bimodal  ambient  particle  concentration  distributions  at about  3pm  (14).
The  effects of  filter  clogging  are  not significant  for  mass  loadings of
less  than 400yg  (on the  Zeflour  filter),  and  only  in cases  of  sampling
high-concentration  environments  would  the  filter  unit  need  replacing in
time intervals less  than 8 hours.

Pump and Power Supply

     The flow  rate  of the sampler  is  an important  parameter in  setting the
sampling  time required  to  collect   a  sufficient  sample  for   gravimetric
analysis.   There is  a  ±10yg uncertainty  in weighing a clean filter; there-
fore,  the  sample must weigh at least  lOOyg  if  it is to  be determined  to an
uncertainty of 10 percent or  less.   Thus,   for  normal  ambient aerosol  con-
centrations  (coarse + fine modes) of  50  to  lOOyg/m3 (15,16),  J>4 m  must be
sampled.   The  portable sampler  has  a flow rate of 6 L/min  or .36 m3/hr.
The  minimum sampling time,  in hours,  to  collect  sufficient  sample on the
filter (lOOug) for <10  percent uncertainty in the gravimetric  analysis is
2.8  x  102/c,  where ~c is the  concentration of the  particulate  in  yg/m3 of
the  fraction (inhalable or respirable) of concern.

                                     304'

-------
               1.0
            I
           it
           LU
            o
            O
               0.5
               0.0
 'lid particles
liquid particles
                              Particle Aerodynamic Diameter—
Figure  2.   Filter collection efficiency  as  a function of  aerodynamic diam-
            eter.   The symbols  (1) and  (s)  denote  liquid and  solid parti-
            cles.   The Atomic Energy Commission's  respirable curve  is  also
            shown.
     Changes  in  the flow rate of  the  sampler can introduce  two errors:  an
uncertainty in the  total volume of air sampled, and  shifts  in the inlet and
coarse  filter cut-points.   The  flow rate  decreases  if the  voltage driving
the  pump decreases or  the pressure  drop  of  the  system  increases.   The
batteries selected  will maintain  constant  voltage under the  load presented
by the  pump for  over 24 hours.  The pressure drop of  the system is primar-
ily that of the  filters and filter loading.   As the filters  load,  the flow
becomes  restricted.   A loading of 400yg on the  filter  unit  reduces  the flow
rate by  12 percent.   Our studies  show,  however, that  a flow rate reduction
of 33 percent,  i.e., from 6 L/min to  4 L/min  changes  the cut^-point  on the
Nuclepore filter by only 7  percent.   Calculations  indicate  that a similar
reduction in  flow rate leads  to  only  a 7 percent  change  in  the  inlet cut-
point  (4).    Thus,  filter  loadings  normally  encountered will  not   cause  a
large enough  decrease  in flow rate to  have  any  significant effect  on the
sampler  performance.   The  uncertainty  in  the  calculation  of  the  particle
mass  concentration, pg/m3,  is  a  result of  uncertainties  in  weighing  the
mass  and obtaining  the total volume  of air  sampled.    The  uncertainty  in
total sample  volume resulting  from flow rate changes will in most instances
be less  than  the weighing uncertainties.   However, for extreme cases, flow
rate  monitoring  may  be necessary.   It may  also  be possible to count  the
piston  strokes  of  the  pump  to obtain  a  more precise  measure  of the total
volume  sampled.
                                     305

-------
Inlet

     The  sampler  inlet  should  perform  two  major  functions—first,  it
defines  the  upper aerodynamic  diameter limit of  the particles  to  be sam-
pled, and second,  it  minimizes  wind velocity and orientation biasing of the
sampler  (17,18).   The inlet used in  this  sampler was made  by  scaling down
an  inlet designed by Walter John  and Steve  Wall  (19).   The  sampled air
enters  the  entrance  slits  of  the  inlet (see  Figure 3) and  passes  between
the  inverted "funnel" and  the  annular impaction surface  directly beneath.
Particles  larger  than  the cut-point  are  removed  from the  air  stream  by
impaction on the  oil-soaked frit used as  the impaction surface.  The dimen-
sions  of the "funnel"  determine   the cut-point  of  the  inlet.   We have
designed  inserts  that  have  15,  10,  or  7ym  cut-points.     The  oil-soaked
sintered annular  ring provides a  long-lived sticky  surface  that minimizes
particle bounce (19).   The inlet  is connected directly  to the  37-mm tandem
filter cassette by an o-ring  seal.

     The  cut-point of  the  impactor  was  determined  by  measuring  percent
penetration  as a  function of  particle diameter,  using the  vibrating orifice
generator to produce  monodisperse  oleic acid droplets and  using the optical
particle counter  as the  detector.   The results of the cut  test  are shown in
Figure 4 for the  15ym,  lOym, and  7pm inserts.    The two boxes  define the
region  of unacceptable  performance  for  a  15ym inlet  designed  to  sample
inhalable  particulates  according  to  the  criteria  by Smith, et  al.,  (20).
These measurements were  made  under  quasi-static  wind velocity conditions  to
define the  cut-point  characteristics  of  the inlet  impactor.   To determine
the  sampling efficiency of  the  inlet,  it  is  necessary  to  conduct  wind
tunnel tests.

Inlet Testing

     The sampling efficiency  of inlets is  difficult to predict  on the basis
of  theoretical  models (17).   The  existing  theoretical treatments  (21-24)
deal  mostly  with  calm  (zero wind  velocity)   conditions.    Wind  tunnel
testing,  therefore,  has  been necessary to  characterize  inlet performance
(17,18,25-29).  The wind tunnel (Figure 5) used  at NBS is  made  of corrosion
resistant stainless steel,  has a 0.46 m x 0.46  m cross-section,  and  has  a
working  length  of 2.4 m.   Air is  drawn through  the tunnel by  a variable
speed fan-type  blower located on  the exit  end.   Over  the  entire velocity
range (0.3-2.4 m/s),  the turbulent  levels  do not  exceed  4  percent  of the
wind velocity values.   The  turbulence value is  defined as   the RMS  of the
fluctuating  longitudal  wind  velocity component.   A turbulent mixing grid
and an airfoil are used  to stabilize  the aerosol plume  (Figure 5).   Narrow
(Figure  6),   but  stable   (with  respect to particle  concentration)  aerosol
plumes are  employed  to  provide a  more concentrated  particle stream,  which
reduces sampling  time (4).

     The inlet sampling  efficiency is defined as the ratio  of  the particle
concentration measured  by sampling  through  the  inlet to  the particle con-
centration  measured   by  sampling  with an  isokinetic  probe   (an  ideal sam-
pler).  The  monodisperse test aerosol was  oleic  acid-flucrescien dye parti-
cles generated by the VOPG.  The aerosol  was  sampled through  the inlet  or

                                     306

-------
                                               Inlet Slits
                                               -Oil Soaked FRIT
                                                Impaction Surface
                                                 O-Ring Fitting
                                                 to Filter Casette
Figure  3.   Schematic of  the  .inlet.   The values  of  A and  B determine  the
           cut-point.

                  Critical  dimensions for funnel inserts
                       Cut-point
                         (um)

                          15
                          10
                           7
  A
(cm)

3.696
3.863
3.871
  B
(cm)

2.372
2.568
2.675
an isokinetic probe,  and the particles collected on 0.3pm  pore (37-mm diam-
eter) Nuclepore  filters.  A particle-laden filter was  then washed with 3 mL
of  0.1  N  ammonia  solution  and  the  fluorescien  concentration  determined
spectrophotometrically.    The sampling  efficiency  was  calculated  from the
measured fluorescien concentration,  the known sampling  flow rate,  and the
sampling duration.

     The results of the  inlet sampling efficiency tests, shown in Figure 7,
show  that  the inlet  is  somewhat wind  velocity-  and orientation-sensitive.
The cap  or hood  present on  the full  size model (19)  has  not  been used on
the  scaled-down  inlet,  and  considerable  influence by  wind  on  the  inlet
without the  hood is expected (information from personal communication with
J.B.  Wedding, Colorado  State  University,  1981).    The  relative  standard
deviation  of the  inlet  data  is  about 10  percent.   In  some  instances the
                                     307

-------
            997
         5
         s
                              4    6  8  10      20   30  40   60  80 99


                                 Aerodynamic Diameter, fjm              '
Figure 4.
Inlet cut curves  for 15ym,  lOym, and 7ym cut-points.   Collection
efficiency is plotted versus aerodynamic particle  diameter.


           Exhaust Fan & Hood
                                                                 VOPG
                                                                 with Air Foil


r
•^x^u
fopc
f
•* Air Flow
&
-— 	 	 - 	 9 Am

	 ^—
^
                                                              Mixing
                                                               Grid
                                                            Absolute
                                                              Filter
                 Test Section
                      with
                   Traverser
                                             Aerosol
                                             Injection
                                             Section
Figure  5.   NBS  wind tunnel test facility.

                                      308

-------
                                  Horizontal Particle
                                Concentration Profile
             1  I—
      en
      _o>
      o
      CD
      Q_
      0)
      .Q
      E
      0)
      >
      '-t-*
      _CO
      0)
      OC
                            Normalized Horizontal Distance

Figure  6.   Horizontal particle concentration profile  in  the wind  tunnel.


inlet has  a sampling  efficiency greater  than 100 percent.   Such  oversam-
pling is  predicted for  large  particles at  low  wind  velocities (21-23) and
has  been  observed for full-scale inlets (18).  At high wind velocities, the
inlet does  not  sample large particles  efficiently.   At  low wind  velocities
the  inlet  sampling efficiency is essentially independent of wind velocity,
making  the  portable sampler better  suited for indoor sampling applications
where the  wind  velocity averages 0.15  m/s and  rarely exceeds  0.5 m/s  than
for  outdoor environments  where  the wind velocity  can reach  much  higher-
ranges .
                  INCORPORATION OF N02 SAMPLING CAPABILITY

     The NBS portable  ambient  aerosol sampler monitors the particle  concen-
tration  of the  sampled environment.   In  many  instances,  there  is  equal
concern  for health  and/or  safety reasons  over the  concentration of  other
airborne  pollutants,  such  as  S02,  organics,  radon,  and  N02,  which are
common indoor  pollutants (1).   The sampling  capability of portable  aerosol
samplers  can,   in principle,  be  easily  expanded  to include  gas sampling
without  increasing the bulk of  the  sampler  or its  energy  requirements by
means of passive  gas sampling  techniques.   Passive gas sampling employs the
kinetic  energy  of the  gas to transport (diffusional transport) the  species
of interest to  a trapping agent.   The Palmes tube  (30) sampler  for  N02 is
an example.  For indoor ambient  monitoring,  the gas monitor  must be  small

                                    309       "  -•''.--'

-------
 1.0
 o.o
0
 1.0
 0.5
 0.0
                       a.
          2 pim diameter
                 	o-
                 1            2
               Wind Velocity m/s

                      C.
                10^m diameter
                 1            2

               Wind Velocity m/s
                                          1.0
                                       CD
                                      "o
                                           LU  0.5
                                          0.0
                                                             b.
                                                       4 to 7\jm diameter
                                          1.0
                                           o
                                           CD
                                           'O
                                      £  0.5
                                          0.0
                                             0
  1            2
Wind Velocity m/s

       Cl.
   15 ^m diameter
                                                                         45°
                                                         75
  1             2

Wind Velocity m/s
Figure  7.   Inlet  sampling  efficiency  as a function of  wind velocity in the
            wind tunnel.  Inlet orientation is shown where (o)  is  upright 0°
            angle,  (•) is  for 45° angle tilt  forward  into wind,  and  * is  a
            75° angle tilt  forward.
                                       310

-------
 and have  a high  collection rate.   The  Palmes tube  sampler is  small but
 requires about 24  hours  at normal ambient  N02  levels to  collect  a detect-
 able sample of N02.   Because  the  portable sampler  can  potentially collect
 sufficient particle  concentrations in as short as  8-hour  sampling periods,
 it is useful to have an N02 sampler  that  can  collect a measurable sample in
 the same time period.  The development of a small  passive N02 monitor that
 has a high  collection rate has  become possible  by modifying  the patented
 Dupont Pro-Tek organic vapor  badge (31,32).  The  collection medium  is a 1
 cm x 6  cm  Gelman A  filter strip soaked  in a  1/10  (by volume)  solution of
 triethanolamine (TEA)  in acetone  (information  from personal communication
 with B.  Cadoff, NBS, 1981).  TEA has been used  as  an N02  trapping agent for
 many applications (33-37).

      The sampling  rate by the Pro-Tek badge  is  defined  by  diffusion from
 the outer surface  through 280,  1-mm-diameter holes to the  collection sur-
 face (TEA-soaked  filter  paper).    The glass  fiber  filter provides  a high
 surface  area collection medium for the N02.  The  sampling  rate  (Figure 8)
 of the  device  was  determined  from  the  quantity  of N02  collected by the
 sampler  for a known N02  exposure.   The N02 collected, which was  generated
 using  permeation  tubes,   was  measured  by  treating  the filter  strip  with
 Saltzman reagent  and analyzed spectrophotometrically.  The sampling rate is
 5.3)jg  N02  collected per ppmxh exposure, which is approximately 50  times the
 sampling rate  of  the Palmes tube sampler.
                               0.4            0.8

                              No2 Concentration X Time (ppmxh)
Figure 8.  Summary of  the N02  collection rate for the Dupont Pro-Tek badge.
           The  collected N02  mass  in yg  is  plotted as  a function  of N02
           concentration in parts  per million  multiplied  by  the  time in
           hours.
                                    311

-------
                                  SUMMARY

     The NBS  portable  sampler  fills  the  need for  an ambient:  atmospheric
pollutant personal  exposure monitor.   It  is capable of monitoring  ambient
concentrations  of  inhalable particulates and N02 in order to  assess  indi-
vidual  exposures  to these  pollutants, which may  not  always 'be  accurately
determined by conventional  stationary  outdoor monitoring  systems.   The sam-
pler  separates  and  collects  the coarse  (3.5ym  to  15ym-diameter)  and  the
respirable  (<3.Sum-diameter)  fractions of  the  inhalable  particulates  by
series  filtration.   The collected fractions are  gravimetrically quantified
and can be subsequently chemically  analyzed if desired.  The high sampling
rate  of 6  L/min  makes  it possible  to  collect  analytically  significant
(analyzable  with 10  percent  precision)  samples  at ambient  environmental
concentration levels in as  short  as  an 8-hour  sampling period.   The sampler
includes a specially designed inlet that transmits  only  the  less-than 15vm
(or  alternatively  10pm) aerodynamic   diameter  particulates.    The  larger
particles are removed  by impaction in  the  inlet.   The  sampling performance
and  characteristics of the  various  components  of the  sampler  have been
extensively  laboratory- and/or wind tunnel-tested,  and the entire  unit  is
now   being   field-tested.    A   small  passive  N02   sampler   capable  of
incorporation into  the portable particulate sampler extends  the monitoring
capability of the  sampler to  simultaneously measure N02  exposure.  The N02
sampler has a high  sampling rate (50  times  that  of  the Palmes  tube sampler
commonly used in occupational monitoring  situations).
                                 REFERENCES

 1.    Committee on Indoor Pollutants, Board  of  Toxicology and Environmental
      Health Hazards.   1981.  Indoor pollution.   Assembly of Life Sciences,
      National Research Council.   National Academy Press, Washington, DC.

 2.    Dockery,  D.W.,  and J.D.  Spengler.   1981.   J.  Air  Pollut.  Control
      Assoc. 31:153.

 3.    Repace,  J.L., and A.H. Lowrey.  1980.  Science 208;464.

 4.    Bright,  D.S., and  R.A.  Fletcher.   In  press.   Amer.  Ind.  Hyg. Assoc.
 5.    Fletcher,  R.A., and  D.S.  Bright.   1981.   International Conference  on
      Powder and Bulk Solids,.Cahners Exposition Group, Chicago,  IL.

 6.    McKenzie,   R.L.,  D.S.  Bright,  R.A.  Fletcher,  and  J.  Hodgeson.    In
      press.  Environ.  International.

 7.    Parker, R.D., G.H. Buzzard, T.G. Dzubay, and J.P. Bell.   1977.  Atmos.
      Environ. 11:617.

 8.    Cahill, T.A.  1977.  J. Air Pollut. Contr. Assoc. 27:675.
                                      312

-------
 9.
 10.
 11.
 12.
      Miller,  F.,  and  D.E.  Gardner.   1979.   J.  Air Pollut.  Contr. Assoc.
      29:610.                                                           	
13.


14.


15.

16.

17.



18.

19.



20.
      Liu,  B.Y.H., D.Y.H.  Pui, K.L. Rubow,  and G.A.  Kuhlmey.    1978.   EPA
      progress  report (May).   EPA Grant R 804600.

      Berglund,  R.M.,  and  B.Y.H.  Liu.   1973.   Environ. Sci. ' Tech.  7:147-
      153.                                        	

      John,  W.,  S. Hering, G.  Reischl, and J.J.  Wesolowski.   1980.   Final
      report,   Interagency Agreement  ARB  A7-139-30,  Air  and  Industrial
      Hygiene    Laboratory   Section.      Lab   Services,   Berkeley,   CA.
      CA/DOH/AIHL/SO-21.

      John,  W.,  G.   Reischl,   S.  Goren,  and  D.  Plotkin.    1978.    Atmos.
      Environ.  12:1555.
21.

22.

23.


24.
     Whitby, K.T.,  R.B.  Husar,  and B.Y.H.  Liu.   1972.  J. Colloid Interface
     Sci. 29:177.                                                  :

     Macias, E.S.,  and R.B.  Husar.   1976.   Environ.  Sci. & Tech. 10:904.

     Lewis, C.W.,  and E.S. Macias.   1980.   Atmos.  Environ. 14:185.

     Liu, B.Y.H.,  and D.Y.H. Pui.   1980.   Page 383  in Proceedings  of  EPA
     conference  on advances  in  particle  sampling  and  measurement,  October
     1979, Daytona  Beach, FL.   EPA-600-9-80-004, January.

     Wedding, J.B.   1981.  Environ.  Sci. & Tech. 16:154.

     John, W.,  S.M.  Wall,  and J.J.  Wesolowski.   1981.  Air  and Industrial
     Hygiene Laboratory,  California Department of  Health  Services.   Report
     CA/DOH/AIHL/SP-27.

     Smith, W.B., K.U.  Gushing,  M.C. Thomas, R.P.  Wilson,  and  D.B.  Harris.
     1980.  Page 316 in Proceedings  of  EPA conference on advances in parti-
     cle  sampling  and  measurement,   October   1979,  Daytona   Beach,   FL.
     EPA-600-9-80-004, January.

     Davies, D.N.   1968.  Brit.  J. Appl. Phys.  (J. Phys. D.)  1:921.

     Fuchs, N.A.  1975.  Atmos.  Environ. 9:697.

     Fuchs, N.A.   1964.   The mechanics of  aerosols.   Pergammon  Press,  New
     York, NY.

     Agarwal,  J.K.,  and  B.Y.H.  Liu.   1980.   Amer.  Ind. Hyg.  Assoc.  J.
     41:191.                                                   '	
25.  May, K.R., N.P. Pomeroy, and S. Hibbs.   1976.  J. Aerosol  Sci.  7:53.
                                    313

-------
26.  Ogden, T.I., and J.L.  Birkett.   1978.   Ann.  Occup.  Hyg. 21:41.

27.  Pattenden, N.J., and R.D.  Wiffen.   1977.   Atmos.  Environ. 11:677.

28.  Caplan, K.J., L.J.  Doemeny,  and S.D.  Sorenson.   1977.   Amer. Ind. Hyg.
     Assoc. J. 38:83.

29.  Wedding, J.B.,  A.R. McFarland,  and J.E. Cermak.  1977.   Environ. Sci.
     & Tech. 11:387.

30.  Palms, E.D., A.F. Gunnison,  J.  Dimattio,  and C.  Tomczyk.   1976.  Amer.
     Ind. Hyg. Assoc. J. 37:870.

31.  Pro-Tek Organic Vapor Air Monitoring  Badges.   Laboratory  validation
     protocol for diffusion-type  air monitoring badges with solid sorbents.
     Copyright  1981  by  E.I.  duPont  de Nemours  and  Co., Inc.,  Wilmington,
     DE.

32.  Lautenberger, W.J., E.V.  Kring, and J.A.  Morello.   1980,.   Amer. Ind.
     Hyg. Assoc. J.  41:737.

33.  Levaggi, D.A.,  W. Siu, E.L.  Kothny,  and M.  Feldstein.   1972.  Environ.
     Sci. & Tech. 6:250.

34.  Levaggi,  D.A.,  W.  Siu,   and M. Feldstein.    1973.    J.  Air.  Pollut.
     Contr. Assoc. 25:30.
35.  Willey, M.A., C.S. McCammon, Jr.,  and L.J.  Doemeny.   1977.   Amer. Ind.
     Hyg. Assoc. J. 38:358.

36.  Fletcher,  R.A.,  D.S. Bright,  B.C. Cadoff,  and  J.A.  Hodgeson.   1980.
     EPA/NBS annual report.

37.  Blacker, J.H.  1973.  Amer. Ind. Hyg.  Assoc.  J.  34:390.
                                     314

-------
            DEVELOPMENT OF  A PROTOTYPE ACTIVE PERSONAL MONITOR
                   FOR S02,  N02, AND  AIRBORNE PARTICLES


              Tahir R. Khan,  Jean  C.  Meranger, and Belinda Lo

                     Environmental Health Directorate
                         Health Protection Branch
                         Health and Welfare Canada
                              Ottawa,  Ontario


                               INTRODUCTION

     Traditionally, air pollution  monitoring has  been carried out at fixed-
station networks.   During  the past several  years,  however, it  has  become
apparent  that a  significant portion  of  the  air pollution  burden  that  a
person is exposed to in the  course of  daily activities occurs in the indoor
rather than  the outdoor  environment  (1,2,3).   Typically, people  in North
America  spend  about  90  percent  of  their  time indoors  (4,5).    Indoor
pollutants may  originate  from  specific  sources  such  as  home  cooking  and
smoking,  and  may have  levels different  from  ambient air  monitored by  a
fixed-station network.

     Indoor/outdoor air pollution  comparisons are being made with increas-
ing  frequency  in  an  attempt to   determine  relationships  between  fixed-
station  community monitoring results  (ambient  air)  and  actual  community
exposures.  Ambient air  quality  results  would then be used  to  predict  the
frequency distribution curve for the  exposure of  a community.

     A potentially more  serious  problem arises, however,  when  the criteria
upon which air  quality  objectives  are based  are  derived from dose-response
data that incorporate ambient pollutant measurements  rather than integrated
indoor/outdoor  exposures  and doses.    Thus, health  effects criteria  have
been based on exposure data  that may  have underestimated  (or overestimated)
the response due to a presumed level  of exposure.

     The variability of  individual human response to  air  pollutants  and of
human  exposure  to them  complicates  the  situation.   Even  if   ambient  air
quality, with respect  to the pollutants  of interest  and  their  indoor/out-
door air  quality  relationships,  are known reasonably accurately,  the  very
different  daily patterns  of  individual  human  activity  may  confound  an
attempt  to  relate  exposure  estimates  to  the likelihood  of human  health
effects.  In this sense, personal  dosimetry of air  pollutants may be essen-
tial to determine whether  current  indoor/outdoor (daily activity)  doses of
criteria pollutants are causing detectable  health effects.  More precisely,

                                    315

-------
personal  dosimetry  may be  necessary to determine  the sub-population  that
may be experiencing air pollution-related morbidity.

     Earlier we reported that no  dosimeter  for monitoring  personal  exposure
was currently  available or  is  likely  to  be produced  commercially in  the
near  future  (6).    Certain  promising  devices  were found,  however.   These
either measured  only  one of  the  pollutants, Harvard/EPRI  (7),  or did  not
have  the  required  sensitivity or operational  life  required,  GAGE  (8).   We
initiated a project with the aim of assembling  a prototype  personal  dosi-
meter  from the  most  promising  commercially  available air  pump,  a  solid
sorbent media  of  the  desired collection properties for N02 and S02 and  an
H&H cyclone with a  Teflon filter  for collecting  size-selected particles.

     Five  candidate pumps  were   chosen  primarily on  the  basis of  results
from  our  earlier  study on personal  dosimeters  for  SC>2,  NOX,   63  and
particulates (6).   The Geomet respiration  controlled  sampler  was not  chosen
because of its inherent  flow variability.   The Gage Research  Institute pump
was excluded  from further  testing,  since  it failed  to deliver the  target
sampling  rate  of  2 liters  per  minute  (L/min).   Two  sorbents,  triethanol-
amine  (TEA)-impregnated silica  gel and molecular  sieve,  were tested  for
collecting  N02 and S02«   The  results of  these investigations,  and  the
prototype finally assembled are  described  in this paper.
                                EXPERIMENTAL
NO/N02  Sampling System
The  sampling   train   used  in  evaluating  the  collection  efficiency  of
triethanolamine (TEA)-impregnated silica gel (Merck,  35-70 mesh) and molec-
ular sieve  (Linde Company, 13X, 1/16" size)  for  N02  is shown  in Figure 1.
Matheson  (Whitby, Ontario)-certified  N02 (49.7 ppm)  gas  was  diluted  with
zero air  (Matheson-certified) to generate the  desired  concentrations.   The
diluted  gas was  standardized  routinely  against  NBS  cylinder  gases  every
week.   The calibration  curve  was constructed  by  plotting NOX  analyzer
(Monitor  Labs  Model  8840) response  against several  dilutions  of  the  NBS
Standard.   The flow  rate  of  each  gas,  including zero  air,  was  measured
precisely with Matheson-calibrated mass  flowmeters before  these gases  were
allowed  to  mix in the precision  calibrator (Thermo Electron,  Model 102).
The  calibrator  was  used  only  for  mixing  the  test  gas  and  zero  air  at
atmospheric pressure.    Stainless  steel  regulator,   fittings,   and  Teflon
tubings were  used to  assemble the N02  and N02/S02 sampling system.

N02/S02 Sampling System

     This system, shown  in  Figure 2, was  used to evaluate  the collection
efficiencies  of TEA-coated 13X molecular sieve, 45/60 mesh (Chromatographic
Specialties),  and operational  parameters  of  the  prototype personal monitor
for  the  collection  of mixed S02 and NO2 gases.   It incorporates  water-
filled gas  bubblers   maintained  at  55°C   to  humidify  the   test  gases.
Scrubbed  laboratory   air supplied  by   the  Cast  Manufacturing's   (Benton
Harbor, Michigan) double diaphragm pump was used as the diluent.  The first

                                     316

-------
Figure 1.  Schematic of N0/N02 flow  system for  testing solid sorbent.
three scrubbers  removed  moisture, the fourth  removed SC>2 and N02,  and the
fifth scrubbed the  remaining  S02«  The final scrubber was  to  remove organ-
ics and  to  act  as  a final back-up  scrubber.    The  silica gel  and  back-up
charcoal scrubber were  replaced daily.   Scrubbed air was  checked  for con-
tamination daily by the  use  of  a sampling pump/TEA-sorbent tube  set-up at
the outlet  of  the   last  scrubber.  Standard NBS traceable gases—50.6 ppm
SOa and  41.4 ppm N(>2 in air—used in  this  assembly were  supplied  by Scott
Environmental  Technology of  Pennsylvania.   After  humidification,  the gas
stream enters  an all-Teflon  32 cm x  7.5 cm I.D. mixing chamber and then
goes to  a Teflon sampling manifold (Figure  3).   The  manifold  consisted of a
cylinder 32 cm x 7.5 cm  I.D.  x  9.5 cm  O.D.  with friction-fit  caps  and three
equally  spaced holes on  the sides to allow the  cyclones  to be  placed within
the manifold.

     The flow  to the  manifold  was  10 L/min,  with  a maximum of 6 liters
being drawn off  for test purposes.  The  remaining 4  liters were allowed to
vent.  Of that,  200 cm3/min was pulled through a liquid  impinger containing
25 mL of 1 percent  hydrogen peroxide oxidant.   The  impinger was subsequent-
ly analyzed for  SO^2", and provides  a  reference point for the  amount of S02
expected to be collected by the sorbent  tubes.

     The cyclone (H&H Custom Work,  West  Hill,  Ontario)  used  was  identical
to  those currently employed  by  the  U.S.  Environmental  Protection Agency
(information from personal communication  with  R.K.  Stevens of  the EPA), and
was designed  to  operate  between 1.9  and 2.1  L/min.   The  particle  cut-off
diameter  is 5u'  when operated  at 1.7  L/min (NBS).   Filters  used  in the
cyclone  were 25  mm  diameter,  l.Oy pore size, FALP  Teflon (Millipore Corpo-
ration)  mounted  on  a stainless  steel  support  screen.   The  filter material

                                     317

-------
                                             60
                                             PI
                                      *• r
                                             f-i
                                             O
                                             M-l
                                             O
                                             •H
                                             4-J
                                             •H
                                             •«

                                             O
                                             O
                                             0)
                                             •H
                                             a

                                             fi

                                             +

                                              CM
                                             O
                                             a
                                             cf
                                              00
                                              d
                                             •H
                                              &i
                                              s a
*^
O  CD
i-l  0)
4-1  ,0

IH  4->
O
   4-)
o  d
                                              S  o
                                              0)  CO

                                              O 4-1
                                              w  o
                                              CN
                                              0)

                                              s,
318

-------
                                            To vent
                  Sorbent
                  tube
            I
Needle
orifice
High flow pump
                                                                            Teflon
                   From gas mixer
                   (SO2+NO2+H2O(V) in air)
                                            10cm
                                                                 Low flow
                                                                 Dupont pump
Figure  3.   Sampling arrangement  at the Teflon manifold.

                                           319

-------
chosen was  reported  to  be  ideal for N02/S02 sampling (9)  and exhibits high
(99.9%) particle  collection efficiencies (10).'  The  high-flow pump sampled
2 L/min through the cyclone and  a restrictor by-pass,  achieved  by placing
an 18-gauge syringe  needle immediately behind the  sorbent tube, was used to
set  the flow through the sorbent  tubes to the desired level (Figure 3).

Preparation of  Sorbent  Tubes

     Both substrates, silica gel  and molecular sieve, were thoroughly wash-
ed until free of  Cl~, NOs",  and PO^3"".  These were  then coated with TEA by
using the procedure  described  by Vinjamoori  (11) with  some minor modifica-
tions.   The Chromatographic  Specialties'  molecular sieve  used with  the
N02/S02 sampling  system was dried,  after  coating,  in an  oven  at 95°C with
the  nitrogen purged  every  half-hour.

     A Dionex System 10 ion chromatograph (1C) was  used for analysis using
the  following operating conditions:  a  100-mm  pre-column, 250-mm  plastic
separator,  250-mm suppressor, .003-M  NaHC03/.0024  Na2C03  standard  eluent,
flow rate 30 percent, loop 300 or 500yL and  sensitivity 3 or 10 yMHO.

Procedure

     The procedure followed for  the  extraction  of  N02  from  the  silica gel
used in  the NO 2  sampling  system was  the same  as  described  by  Vinjamoori
(11), except that the use  of H202  was eliminated.    For  extracting  N02 and
S02  from  the molecular sieve,  the exposed  sorbent  was  placed  in  a  1-oz
(28-mL) "Nalgene" bottle  containing 10 mL standard  eluent, swirled  for  a
few  minutes, and  allowed  to  stand for  5 minutes.   Exactly  8  mL  of  the
extract was placed in  another  1-oz (28-mL)  bottle  containing 2  mL  of 0.1
percent H202  and  was  analyzed  after a brief mixing period.   The  concentra-
tions of N02  and  S02 were  calculated by subtracting the blank and by refer-
ring to the standard peak  height-to-concentration  ratio determined prior to
each sample injection.
                          RESULTS  AND  DISCUSSION

Nitrogen Dioxide Measurements

     During  the  analysis it  was  observed  that  the molecular  sieve  (Linde
Company)  extracts  contain  suspended  material   (most  likely alumina),  and
cannot  be  removed completely either  by  filtration  or by  centrifugation.
This  resulted in  clogging  of the  pre-column  of  the ion  chromatograph.
Silica gel, on the other hand,  produced clear extracts.  It  was  also noted
that the ratio of N02~ to NOs" peak  heights varied over a wide  range  in the
molecular sieve experiments  as  compared to the  silica  gel.    An estimate of
peak height ratios in ten experiments  showed a  variance of 3.6.with  a mean
of 4.5  for  molecular sieve  and a  variance of 0.98  with a  mean  of 8.3  for
silica gel.   On the basis  of  these observations,  silica gel was used  for
further testing.

     For  calculating the  N02  concentrations from  spectrophotometric  and
                                    320

-------
other  measurements,   a  stoichiometric  factor is  used  to  account  for  the
theoretical  formation of half  a  mole of N02~ for  each mole  of  N02 (gas).
The variations  in  the peak height ratios indicate  that  the  factor can vary
even  for  a given  sorbent  and  depends  upon the  type  of the  sorbent  used.
Different values varying between  0.72 to 1.0 have been reported in the past
(12).  We estimated  the  stoichiometric  factor from the mean value of silica
gel results  and found it to be 0.64, identical to  the  Vinjamoori estimate
(11).   However,  a  factor  estimated from  molecular  sieve  results  would
differ  significantly from  this  value.    Two inferences  can be  drawn from
this discussion:

     1.  The methods in  which N02 recovery  is unaffected by the variations
         in  stoichiometric factor should be preferred for  the  analysis  of
         N02.

     2.  For  the  methods   determining  N02- only  (e.g.,  colorimetric),  a
         careful  measurement  of  the  stoichiometric  factor for  the  given
         sorbent should  be  made in  order to  achieve an accurate quantifica-
         tion of NO2 (gas).

     Ion  chromatography enables  simultaneous  quantification  of N02~  and
N03~ and  eliminates  the use  of a stoichiometic  factor.   We  measured both
NP2~ and N0g~ concentrations  independently  from the chromatograms and added
•them together to compute the  recovery of N02 (gas).  A typical chromatogram
is shown in Figure 4.  Nitrogen dioxide was collected on one  sampling tube
freshly filled  with  0.4 gm TEA-impregnated silica  gel and analyzed.   The
results are  shown in Table  1.   The  set-up  for  the first set  of exposures
was such  that  the main  stream was  divided  into  two streams  through  a  tee
for simultaneous collection of  N02.gas  in duplicate sorbent  tubes.  For Set
2, only a  single  tube was  used.  The values in Set 2  represent  an average
of at  least  two independent measurements with a  maximum variation of   ±4
percent.   Similar variation  is found in the duplicate  exposure of  Set  1
(Table 1).  Recoveries in  both cases average 98 percent.  Since the recov-
eries were  good,  no  further  experiments were performed  with  back-up  tubes
to study the collection  efficiency.   Percent recovery is viewed as the col-
lection efficiency of the  sorbent  tested.   The  effect  of  humidity  on  the
quantitative collection  and analysis  of  NOa  and  the breakthrough studies  of
the sorbent are now  being  conducted.   The work presented above is intended
to show  that  silica  gel coupled  with 1C is very  effective for collection
and analysis of N02  and  involves  a simpler  procedure  because  filtration  of
the extract is  eliminated  from the  process.

Prototype Nitrogen Dioxide and Sulfur  Dioxide Personal  Monitor Components
Tests

Solid Sorbent

     Two sorbent tubes,  (1.6-g  sieve  in the front  and 0.8 g in the back-up
tube) were sampled at 250  and  325 cm3/min using higher flow Dupont pumps  to
sample 2 L/min  through the  cyclone  and,  by means  of the restricter arrange-
ment, the desired  flow through the  sorbent  tubes.   The  sample results with
the test  conditions  given  in  Table  2 show  that  S02  collection efficiency

                                    321

-------
                                            WUj
                                     .®AJ\
                                        ..A...
                                a. Blank
                                b. 1 ppm standard
                                c. Duplicate Injection of Sample


Figure 4.  A typical  chromatogram of  blank  (a),  1  ppm standard (b),  and
           duplicate  injection  of sample (c).


exceeded 100 percent,  while  that  of N02 was approximately 40  percent.   The
level of N02  in the scrubbed air, as detected  by  a sorbent tube  placed  at
the end of  the  scrubbing train, was negligible.   High  concentration levels
(25 percent),  however,  were detected  for  S02.   The  high  S02  collection
efficiencies probably arise from sampling  blanks  not accounted for  in the
corrections.

     To  study  the  effects  of  relative humidity  (RH)  on  the  collection
efficiency, tests were conducted  near 25, 50, 70,  and  100 percent RH  at  1
ppm N02  and S02 concentration  levels.    Triplicate sorbent tubes  (Teflon)
were  exposed  at about  200 cm3/min  flow rate.   The results  are shown  in
Table 3.  Generally,  the S02 collection efficiencies  did not appear  to  vary
significantly with humidity, although for N02  there was a distinct increase
above 50 percent RH.   At 75 RH, for  example,  these average 91 percent.and
56 percent for N02  and S02,  respectively.  Impingers show, on the average,
85 percent efficiencies  (ignoring the outlying  65  percent).

     As an indication of whether  the Teflon filter absorbed any N02 or S02
with change in humidity, the filters were removed  and analyzed for N02~ and
S02~.  For the most part,  the filters did not absorb more than 0.2  percent
of the exposed concentrations,  as would be  expected (9,10).
                                    322

-------
     TABLE 1.  COLLECTION EFFICIENCY OF TEA-SILICA GEL SORBENT FOR N02
Flow through
sorbent tube
  cm3/min
Exposure
 hours
           ppm
Expected
Found
                                                                  Percent
                                                                  recovery
  .01
  .51
90.0
91
86.5
85.0
93.21
88. Of
89.0)
91.0/
                     Set 1 (.3-.5-ppm exposure levels)
                       6
                       6
                       6
                       6
                       4
                       4
                       4
                       4
                   .50
                   .50
                   .30
                   .30
                   .49
                   .49
                   .31
                   .31
                 .51
                 .48
                 .27
                 .26
                 .52
                 .54
                 .28
                 .29
                                                               Av.
                102
                 96
                 90
                 87
                106
                110
                 90
                 94

                 97%
39.5
21.0
59.5
31.5
28.5
31.6
50.5
                       Set 2 (1—ppm exposure level)
                       5
                       5
                       2.5
                       2.5
                       2.5
                       1.5
                       1
                   .94
                   .96
                   .96
                   .94
                   .96
                  1.00
                   .94
                 .90
                 .90
                 .92
                 .94
                 L.01
                 .92
                 .94
                                                               Av.
                 96
                 94
                 96
                100
                105
                 92
                100

                 98%
*Bracketed values  are from  the  simultaneous exposures  of  the main  stream
 branched off into two equivalent sub-streams.
           TABLE 2.  SORBENT TUBE BREAKTHROUGH* AT HIGH LOADINGS

Flow through
sorbent tube
250
325
3
Cm3
cm
/min_ ,
/min
Sampled
volume
90.0 L
117 L
Total
yg expected
NO 2 SOa
340
442
463
600
yg Found t
Front
NO 2
138
174
S02
469
719
Back
NO 2
0.4
0.6
S02
6.2
12.6
•Front tube
collection
efficiency,,%
NO 2 S02
41
39
101
120

Conditions:  1.99 ppm N02;  1.97  ppm S02;  51% relative humidity; 24°C room
 temperature; glass  sorbent  tubes;  flow of 2  L/min-1 through cyclone with
 filters,  high-flow  Duponts used;  1.6 g  TEA-coated 13x  molecular  sieve,
 front, 0.8 g, back;  6-hour test.
 tSorbent blank corrected.
                                    323

-------
£,
EC
^
PC
h-
,
I
PM
O
r-

H
<

|
PC
C
l_
to
fc

e
0
M
E-
ft
>-
£
a
o

n:
M
^^
E-
c
H
i
§
K
£




H
'0
•H
**,^
r>
§
£*
o
H
4-1
3
3
5*
•••*
^



*^
3
4



•^l"r~-cMO voooom VOCTICOCM voin'Oco
*si" <}• i
,_ t , 	 i
i!

m m co o o
•— 1 CT> CO in CT>VOvOv£> COOOCO CMI^^OvC
*~* ' — ' • — it — it— , ^^ QLJ OO VO "••^* P-
o^ r*** r^"* p*^ ^«. vo t^**! b^ ^— i CNJ c\j ^z , i , i rr\ ^
^-« ^^ ^— • ' v ^ * ^* ^-1 * ' *^^ • ^ (t-l
CNJ i— 1 >— 1 CM i-H i— t

gggg Igsjjq SS§I gSgS
^1



COS
CT^ CM »^ r* l O"* vO VD Ct4 CM *ct* CO CU ^t" CD O\ QJ
<— 1 "-H >— 1 t— li—  03 • CJ 0
O f-J 03 0) •
,Q 0) 4-> P AJ 4-J
n) 4J to M cd s
iH CU O (U
CO v-| 4-J O M C

03 ' »H ^1 iH Pl4 Pl4
Pi & 1 rO 0
O VD -H 4-)
HO) 4-1 O

H O 03 CU
T3 i— 1 , 3 0 0 p.,
O CJ 4-1 CO M SI
+-+*«»
324

-------
Sampling Pumps

     The  five  sampling  pumps  chosen were  based primarily  on an  earlier
study (6).  Life  tests were performed on thes^e pumps  in  order to  determine
the total time that  the  units  would be able to sample  the  2  L/min required
for the prototype based  on the use  of the particulate cyclone  and  filter.
The pumps were tested under conditions simulating the back pressure (=50 cm
H20) of sampling  according  to  the  full sampling train depicted in Figure 5.
The pressure was  simulated by  constricting  the sampling "line  to  the pump.
All pumps  tested were fully charged  prior  to use,  and  pumps  were  tested
twice each.   Results  are  shown  in  Table  4.   Data  for  the  Harvard/EPRI
system  are  not  reported  because  the  batteries   in  the  unit  tested  were
five years old.

     The results  indicated  that,  based on the ability  of  the  pump to main-
tain constant flow for  a 12-hour period,  the  Dupont P4000 was  clearly the
best overall  performer  when  compared to  the  Dupont  P2500 and the  Gilian
HFS-113.  All pumps  were generally good for 6 hours,  but  at  8  hours,  only
the Dupont P4000  and the Gilian HFS-113 were performing satisfactorily.

     The Gage Research Institute  pump (with liquid  impinger  removed) could
not deliver  2  L/min; therefore,  it was excluded from further  testing.  It
should  be  noted  that  the  Gage pump  used  was  the  first-generation pump.
There is  at  present a  second-generation unit,  but  unfortunately  it could
not be made available  at the  time  of testing.   The  second-generation  pump
unit is  equipped  with   two  sampling  pumps  and  also,  provision for a 1.7
L/min cyclone has  been incorporated.

     Replicate  testing  of  the Gilian,  Harvard/EPRI,  Dupont  P4000,   and
Dupont P2500 was  then performed at  a relative humidity  of 75  percent and
S(>2 and N(>2  concentrations   =1.0  ppm.    The  sampling  rate  through the
cyclone  (with  filters)  was set  at  2 L/min,  while the  flow  through the
sorbent tube was  set at  200 cm 3/min.

     Prior to  each test,  all  pump flows  were set  and were  identical (±5
percent) to each  other.  Sorbent tubes (Teflon)  containing 1.6 g sieve  were
used for  all  tests.   The  results  given in Table 5  indicate that,  for the
test performed, the Dupont  4000 again performed best.   On inspection of the
actual sorbent flows, the  following  remarks  may be made:

     1.   The Gilian  and  P2500  pumps  always  showed lower  than nominal
         flows.

     2.   The Harvard pump  always showed more than nominal flow.

     3.   The P4000 was closest to  the set  flow.

     4.   The mean collection  efficiency  for N02  is  twice as much  as
         for SO2,  an observation  that was made  in the earlier  set  of
         exposures  (Table  3).    Impinger  efficiencies  again  average
         close to  100 percent.
                                     325

-------
Figure 5.  Photograph of assembled prototype.




                                    326

-------















to
PM
S
o
1—
t—
PM

5
PL

e
r-
!2
CJ

o
CO
H
CO
w


w

M

j.

w

m
tn















rH
cfl S •
4-1 O
O rH ^
H M-i i—

ro
t— j
—1
1 4-1
CO C! T
*IJc>
a o m
cfl CO O
•H
•H
O

-

•P




•H £
Cfl ^ *r-
4J O E
O i — 1 ^^


o
0
0 4-J C
-* d -r
PM CU S 0
,JQ O ""^
a o m 0
O CO C.
X

Q



CU
e







cfl 5 -H
4-1 o e
p_l tj_l . "1

^3
•^ jj rt
2| g|
4-1 M rHco
£3 O "4-1 0
' CO O
I





^
^
H

ocNcx)co«d-ooin o o "~> 01 CNJ — i
oc^oooooooooo Ocor-~\ocTicr»
^M1 '• '' — (i— it — ii — r CNi — ii — ii — ii — ii — i



CU cu
<-~i ^* pt
CN" '' — '' — '« — '' — (i — ICO -CN t — 1 i — 1 i — I < — I i — i CO




"t^.HWM^jW 4J ?-i^i^i^4
J"j H H — * rC rC ,A p0 M J-1 M p0 pd jll ,0
4-1 O »— I LO O ' — I 4J OCNc^ £> OO —1 — 1 — 1 — 1






OOOOCT.OOO t-HCT\CTi
•H d
ctf O
3 ft
T3 3
•£°
•H CU
T"O ft~!
d 4J
•H
•a
^[ cj
3 cfl
IH TJ
3
m o
O rH
*
327

-------

s

^

Kj
M
O

»4
S
P3
£4
>
M
E-J

1-3
H
= *•**
R

W
O
5
W


m
Q
>y>
^2

.J
«
w
3
M
PH
PH
{—}
•
r-4
E_|
^J

co

g
PH

O
§
^
pj
p?
^C
CO

pri
o
H

C/2
8
13
1
H

PH
P:

^
(—,

lyi
H-
£

E-

















































































>,
0
CO
•H
O
•r^
lj_l
UH
CO

fHI
o
•H
CJ
CO
rH
O
O



o
CO








CM
o
55






•M-
TJ
c
3
O
PH

00
CM
o
co



CM
i


n*j
Q]
4->
u
cu
ft

w
60
CM
o
CO



CM
o

_^
t-l
-a 3
CU ~ 4->
iH 0) CJ
ft S cd
g 3 x_x
Cd r-l
CO O -U
> cu
CO

4 —
S X"S
O i-l
r-t C Cd

*" *s S
AJ **^- CJ
pJCO (fl
co 0 ^
rQ CJ
M 4->
O CO
CO CO


ft
6
PH
J.
CO
P
4-

,•_
c
o
CJ

,
u
a
H



COsfvOOO i— ii— irHCO O •— 1 r^ CX> "-HCMOO v£> 60/"^
r-H r-l I-H -rl CU
13 60
3 0
• iH -i-l
> O ft
*"» 1"^ f^» CO
r-Hr-Ht-HCM i-H r-l I— 1 CM r-Hr-Hr— ICM i— 1 i— 1 i— 1 I— 1



^- ^J- ^t p* i-Hi-H^-IPH -3-O-vTPH r-H— Ir-HpH
i— 1 i— 1 i— 1 i-H i-H i— < •— 1 i— 1 i-H •— 1 i— 1 i— 1
X~X X1— \ S~ N
X-N S~** X-% X"N X"N X"N . X— V X"N CO CO *-O CO
•^* vo »-H CM i^* CM r^» vo vo I-H vo • * • *
.... .... . X"N . . i-H CM P^ i — 1
ONVOcO» X~N X-N X*^ X™\ X™N X™S X"V X"^ X"N X"N
•4* in oo CM co vo co *** N co *>J* !**• CM **i" t ^3 < _>
O\ OO •— 1 r-H ONOONi-H •— IONVD OO OO O — 1
i-Hr-HCMCM i-HCMi-HCM CMi-Hr-H i-Hr-HCMCM
^^X *^S \^f \-S N— X ^-X S-X ^*^ N«X >^X V_X x^ X
ooo ooo ooo ooo
OOO OOO OOO OOO
CMCMCM CMCMCM CMCMCM CMCMCM


5-1 JH M fH
T3 Q) ^ CU ^ CO CU
(3M60 JH6Q5H0600 60
OcflctJC O03O0 CdOtd0 COOO0
SS^'ft OMS"ft SoS'ft r^So'ft
cMi-tcdd -vtn)cM0 cO-sT-HS 1-1 CM •* 0
PnUCuiH pHpCPHi-l ECPnO-H OP-iP-iT-l

CMCM CMCM CMCM CMCM
OO OO OO OO
cots cos cos coS
PCJ0E3 &j00 p^00 r^t fi H
ftft ftft ftft ftft

oo o oo f--
.mm .coco • in m .CMCM
moo vooo moo -*oo
r^« . . r^« . • r>> • • [**• * •
r-Hi-H i— 1 " — I ' — !• — t « — li— 1

r-H CM CO -*
=fe =Sfe =H= =S=















































































CO
>
CU
•H
CO

1-1
cd
i— 1
3
O
Q)
i-H
O
e

CO
I— 1

CU •
4-1 CO
cd iH
o cu
CJ >
l cu
 CM
CO O
cu S
3 cu
•H *3
S 4J 4-1
CO
OO CU 4J
CO 4J 0
co cu
CU rO
CO nfl M
Cd 4J O
|3 CO
MM
,0 O x-\
CJ rfj
•H CU CU
,0 CO CO
|2 r-l 0
3 ft
« OX
^. ° s
"**" CU 3
4-1 JC2 x_x
ft 4J
cu ^
O 60 0
K 0 rd
CU •!-( r-l
fH rO
CO • 13 CO
4-1 13 CO
co cu S cu
cu co o i-i
4J 3 r-l
MH 13
M CO 0
3 cU cu 3
O rQ 60 O
r0 3 Cd M-l
1 4-1 rH
VO CO r-l 4J
4-1 > Cd 0
iH 0 Cd 3 CU
rH CU 4-10
4H r-l O S
•H CO Cd <4H
•d H 3 1
0 4-1 •«
O 0 CJ 60 PH
C_? *H -^ p. S
•X -r- H-i-000
328

-------
     The  Dupont P4000 pump  has several  features that  make it  useful and
attractive  for  deployment in  a prototype sampler, most  notably  the autom-
atic timer  shut-off and the visual indication  of cumulative sampling time.
The final testing  incorporated several changes based  on the results of the
sampling phases  completed.   An impinger filled with 25 mL of 1  percent H202
was  placed   immediately  after  the  charcoal  scrubber  to  make  an  accurate
assessment  of the  amount  of  S02 escaping the scrubber system.  The mean 100
percent collection efficiency of the impinger  at  the  manifold  for  S02 com-
pared to the much  lower efficiencies  of the solid sorbent led us  to suspect
the quality  of  the latter.   A fresh batch  of 13X molecular  sieve was wash-
ed,  dried,  and  coated with  an increase  in the  concentration  of  TEA and
ethylene  glycol (50 g TEA + 8 g glycol + 25 mL acetone + water  diluted to
75 mL)  according  to  the  procedure given earlier.   The sieve was  removed
from the  oven at the first sign of free-flow.   In order  to accurately set
flows through the  sorbent tube  while  passing 2 L/min  through  the  cyclone,
the restricter-bypass arrangement  was not  employed.   The  Dupont  P4000 pump
was used for high  flows (1.8 to 1.9 L/min) through the  cyclone,  and a low-
flow pump was employed to pump  the  desired flow  through  the sorbent tube.
The total flow  through the cyclone  would therefore be 2 L/min.

     Sampling blanks were  determined with  the scrubbed  air only  at  flow
rates of  100,  150, and 200  cm3/min at 75  percent relative  humidity.   The
results of  these tests are shown in Table 6.   The recoveries for  N02 at the
concentration  levels  investigated  are  nearly  100 percent, with  relative
standard deviation of 6.5 percent.  For S02,  recoveries average 96  percent,
with a relatively  large  standard deviation of  32  percent.   The  correspond-
ing  impinger recoveries  including  those  obtained  in the earlier  tests
(Tables 3 and  5) exceed 85  percent and  do  not show  such variations.   The
variance  is  believed  to  be  due to  inefficient  collection  of  the  sorbent
under the  test  conditions.   S02 losses  in the  60 to 90 percent  humidity
range have been  reported by Vinjamoori  (11).

The Prototype

     The prototype personal  monitor (Figure 5) was assembled with  the  fol-
lowing components  (numerals  in parenthesis  refer  to Figure 5) :    ,

     (a)  a Dupont  P4000 sampling pump

     (b)  an 18-gauge syringe needle restrictor/bypass  to  allow low flows
          through  the sorbent  tube  (I)

     (c)  a 6-mm OD Teflon  tube containing 1.6 gm TEA-coated 13X molecular
          sieve (II)

     (d)  an H&H particulate  cyclone  with  25-mm-diameter  FALP  (Millipore
          Corporation) ly Teflon filter  (III)

     (e)  a side-arm  restrictor  (Teflon  needle  valve)  (IV)

The operational parameters of  the device  can  be summarized as follows:
                                    329

-------




























o
23

CO

^
gj
M

PN
o
CO
CO
S

*
vO
M
S
»2
iw
^^



































•








































































^N
ft
a
o
iH
U
•H
PJ
CO
_4
o
•H
4-1
O
CU

rH
o


,
^
c
3

•y.

60
P.


0)
4J
o
0)
p

^q

60
p.

•«
p.
0
CO
CO


ij
0
1

o
CO
CU
6t
rrj
S
OJ

<5
^*4






















— .
Q
CO





CM
0
K







CM
o
CO




CM
g


c\
O
CO




CM
O

ft
OJ
0

O
t>


rH

*H

*^
CO"
0
{J
)

K
c

It
*fl


jj
CO
c

4-
*r*

c
t

4.

<2







coo &••$
CM r*"* vo oo P"* P** LO ON vo CM ON ON CM
\O f^. \o ON OO O% ON ^.5 «— ( LO O] ^D CO
t— 1 Ol »— < «— 1 r— 1



cex> t>9

i-f ^H ^H rH i— 1 VD

Q
r^



CO ^^ ON ON CO ^^ C^ r~'
*^* co i^ co co c^i r*** co co *sj* co '•^
mooiON •— ' CM CM co co co .co co
r-4 r_| r_|


vD ^O ^^ *^J* ""^ 1~™' ON
• • • • • • •
 r-HCMCMCO rHCMCMCO
i— 1 i— 1 r-l


CP> i— 1 O\ CM CM vO rH
» i^iip • • • _ • • •
COCOt^-P-l -^•Ol~»Pq - CO
Cfl n^J 4-1
4-1 CO CO
a 4-i n
O r-l CU
O CO p,
CO O
CO Pi
CO -H CM
r° 0
3 J-( CM
4-1 CO pd
4J rH 60
C -iH C
OJ f| l "iH
o C3
J-l rCj "tH
O 4J CO
CO -H 4-1
S d
cd o
o •• o
rH co
l)-l CO rJ
Q) C CU
Hr? £
.». cj «H
U >-, P.
o O 0
sf 'H
CM fl
60 CO
030
4-1 O O
r-l
• 4-1 Cfl"
CO P.
CU -H P.
H CO
3 CO 4-1
4-1 CO 0 •
CO P< O CO
r-l P. r^
cu 3 0
P. I o . cd
0 C3 CO rH
CU lH & 4-1 rQ
4J 0 O CD
<. rH CU 4-1
B r4 <4H 4J 0
0 1 CU
O CM 13 rJ rP
I-l O 3 r-l
IW rH O O
O r0 CO
•• &*•» 1
Tj te Jb vD Tj
05 0
CO rH TJ rH CO
m cu rH
•> rH Cd 60
CU ">H 0
> CU O ••> -H
O > r-l rJ rH •
,a cu 4-1 T-I Pi 4J
cd TH 0 CO 0 0
CO O CO CU
CO O ,P CO 0
<3 rrj cO 'H
(U CO rH r-l 4-1
4-1 CO OP
• • CO rP Tj t-l CU
CO O 3 CU P.
PJ CJ 4-1 rP Tj
O rP CO 4-1 •
•H >, 4J 3 4-> O r-l
4-1 rH 0 r-l O S CO
•rH ,0 CO CJ CU -iH
TJ CO > CO r-l 1 rH
p! CU rH ,-1 4J
O J-i O 4-1 O P4 3
u m co o o a o
* +- -M-GOS
330

-------
      1.   The  pump  operates  unattended at a sampling  rate  of  2 L/min over a
          12-hour period.

      2.   The  FALP  Teflon  filter in  the  cyclone poses  no  flow restriction
          problems  and appears  to  absorb a  maximum  of  only 0.5yg  S02  and
          negligible  N02.

      3.   The  bypass/restrictor  arrangements  set  the desired  flow through
          the  sorbent tubes  while simultaneously passing 2 L/min through the
          cyclone.  Flow  stability  of ± 5-10  percent are readily achieved at
          a  constant  pump  flow.

      4.   The  TEA-coated  13X molecular sieve  effectively collects  0.2  and
          1.0  ppm concentration N02 exposures at  75 percent  relative humid-
          ity,  at flow rates between  100 and  200 cm3/min for 6  hours.   For
          S02,  collection  efficiency  is about   the  same,  although  there
          appears to  be a  lesser dependence  on relative humidity.

      5.   The  capacity of  1.6 g of the  sorbent  in a Teflon  tube appears to
          be  80yg   S02 and  200pg N02.   Sampling  air at  the  rate of  100
          cm3 /min through  the sorbent  tube at  a concentration of  1  ppm SOo
          and N02 is  well within the  capacity  of  the sorbent  tubes.    .

The photograph  of  the assembled prototype  is  shown  in  Figure 5.   The  ion
chromatographic analysis  is well suited for the  quantification  of the tar-
get in the  extracts.  Below about  1  to 2yg S02  or N02 collected, the analy-
sis is performed  close to  the detection limit.   Higher  injection volumes
would increase the sensitivity.  Arrangements  are underway to carry out the
validation  tests of  the prototype  in an exposure chamber.   The results will
be reported later.


                               ACKNOWLEDGMENTS

     Part  of  this  work  was  conducted  by  Concord  Scientific  Corporation
under Contract Reference No. 808.
1.
2.
3.
                           REFERENCES

Godin, G., G. Wright, and R.J.  Shepard.   1972.
bon monoxide.  Arch. Environ. Health 25:305.
                                                      Urban exposure to car-
Wright, G.R.,  J.  Jewczyk,  J.  Onrot, P.  Tomlinson,  and R.J.  Shepard.
1975.  Carbon monoxide in  the urban  atmosphere.   Arch.  Environ.  Health
30:123.	

Ott, W.R.,  and D.T. Mage.   1975.   A method  of  stimulating the  true
human exposure of  critical population groups to  air pollutants.   Page
2097 in Recent advances  in the  assessment  of the  health effects  of
environmental pollution.   Commission of the European Communities,  EUR
5300, Luxembourg.
                                    331

-------
Chapin, F.S.    1974.   Human  activity patterns  in the  city.
Intersciences, New York, NY.
                                                                      Wiley-
5.   Szalai, A.   1972.   The use  of  time:  Daily activities  of urban  and
     suburban populations in twelve countries.  Mounton,  The Hague.

6.   Meranger, J*C., T.R. Khan, and R.B.  Caton.   1981.   State-of-the-art of
     commercially  available personal  monitors fbr  NOX,  SOa  and  particu-
     late matter  in ambient air.   Proceedings  of  the  15th Conference  on
     Trace  Substances  in  Environmental  Health,  University  of Missouri,
     Columbia, MO.

7.   Dockery, D.W., and J.D. Spengler*  1981.  Personal  exposure to respir-
     able particulates and  sulfates.  JAPA 31:153.

8.   Mintz, S., R.H. Hosein, B. Batten, and F. Silverman.   n»d.   A  personal
     sampler for three respiratory irritants.  The Gage  Research Institute,
     Toronto, Ontario, Canada.

9.   Appel,  B.R.,  Y;  Tokiwa,  S.M. Wall,  E.M* Hoffer,  M.  Hailc,  and  J.J.
     Wesolowski.    1978.    Effect  of  environmental  variable  and  sampling
     media  on the  collection of atmospheric sulfate and  nitrate.   Califor-
     nia Air Resources Board, Contract ARB5-1032, January.

10.  Appel, B.R.,  S.M. Wall, Y. Tokiwa,  and M. Haik.    1979.   Interference
     effects in sampling  particulate  in ambient air.   Atmospheric  Environ-
     ment 13:319-325.

11.  Vinjamoori, D.V., and  C«  Ling.   1981.  Personal  monitoring method for
     N02 and S02 with  solid sorbent sampling and  ion chromatographic deter-
     mination.  Anal.  Chem. 53:1689-1691.

12.  Blacker, J.H.   1973.  Triethanolamine for collecting  nitrogen dioxide
     in the TLV range.  Am.  Ind. Hyg. Assoc. J.;394.
                                     332

-------
              DEVELOPMENT OF SPE DIFFUSION HEAD INSTRUMENTATION
                 J.A.  Kosek, J.P. Giordano, and A.B. LaConti

                          General Electric Company
                      Direct Energy Conversion Programs
                               Wilmington, MA
                                INTRODUCTION

     General Electric has developed  a line of  electrochemical  sensors for
monitoring  such  gases  as CO,  NO, and  N02 in  mine and  industrial ;atmos-
pheres,  using a unique  solid polymer electrolyte electrochemical cell.tech-
nology  (1-3).  The  sensor development  has  been carried  out  largely under
contract to  the  U.S. Bureau  of Mines,  supplemented  by  an  in-house IR&D
program.  Several models of carbon monoxide instruments  are now in commer-
cial  production.    These instruments  include  a direct-reading  CO  detector
and a CO dosimeter, both of which use an air-sampling pump to  bring a gas
sample  to the electochemical  sensor  cell.   A feature common to  all these
instruments  is  the  use  of a solid polymer  electrolyte membrane  as  the sole
electrolyte  in the electrochemical  sensor  cell.  Catalytic  electrodes are
integrally bonded to  the membrane to  form a unique  unitary structure..  Use
of the  SPE  sensor cell  eliminates  problems such as  corrosion and  contain-
ment  associated  with  caustic or  acidic electrolytes  and leads  to highly
invariant sensor  cell response  and  long-life  operation with  minimal main^
tenance  and  calibration.

     Recently,  the  family of  instrumentation  has  been expanded  to  include
diffusion head  sensor cells.  In  these instruments,  the  air  sampling pump
has been removed,  and  the  electrochemical sensor  cell  has  been  modified
such that an  air  sample  reaches  the  sensor  cell by means  of natural gaseous
diffusion.

     A  diffusion-based  instrument  offers several advantages  over  conven-
tional,   pumped instrumentation.   First,  the  pump  is usually the  least
reliable  component  in the instrument.   Therefore,  elimination  of  the pump
increases  the  reliability  of  the instrument.    Secondly, in  a  battery-
powered  instrument  or in one  that  has a backup battery  system  for power,
the  pump  usually  requires   the  most   power  of   any  component   in  the
instrument.    Elimination  of the  pump decreases the current draw of  the
instrument and  increases  usable battery life.   Also,  elimination of  the
pump and its  associated  tubing and rotameters  results in a  size and weight
reduction that is important  for  hand-held portable instrumentation.

                                    333

-------
      The  SPE gas  detectors  are intended  for use by  military,  government,
and  industrial  personnel involved  in air quality measurements.  The commer-
cial SPE CO dosimeter  and direct-reading  detection instruments  are  being
widely  used  by  steel mills,  fire departments,  and various  city,  state, and
federal regulatory agencies.
                                EXPERIMENTAL

     The  membrane and  electrode  assembly  used  in the  CO sensor  cell  has
three  electrodes of  identical composition:   sensing,  counter,  and  refer-
ence.   These  electrodes  are fabricated  from a  platinoid black  catalyst
composition  blended with a Teflon (T.M.—E.I.  DuPont) binder.   A transition
metal  screen is  embedded into the electrodes  to  obtain improved mechanical
integrity  and  current collection.   The  counter electrode is  the same size
(2.4 cm2)  and  configuration as the  sensing electrode.   The  reference elec-
trode  has  an area of  approximately 0.32 cm2.   The  electrodes  are bonded to
the membrane by a proprietary method developed  by General Electric.   The
spatial configuration of the resultant membrane and  electrode assembly  has
been developed to achieve good vapor phase transport and  to  provide  a high
output signal.

     A schematic  of the hydrated  solid polymer electrolyte sensor cell used
in a pumped  instrument  is depicted in Figure  1. . The membrane and electrode
assembly  is  housed  in Lexan  (T.M.—General Electric  Company)  polycarbonate
hardware.    Lexan was  selected  because of  its   good  physical  properties
(transparency,  shock  resistance), and chemical inertness/minimal  elution.
The catalytic  sensing and  reference electrodes  are positioned  on  one side
of  the cation exchange membrane; a catalytic counter electrode  is  posi-
tioned on the other  side of the  membrane  opposite the sensing electrode.
The  counter electrode  compartment  is flooded  with water.    Electrolytic
contact between the sensing electrode and the  platinoid metal/air reference
is achieved  through a hydrated solid polymer electrolyte membrane bridge.
The performance characteristics and  other properties  of the  cell are  highly
dependent  on the  morphological structure of the membrane  and  the method of
hydrating  the membrane  to achieve a  fixed water  content.

     All work  has been accomplished using perfluorosulfonate  ion  exchange
membranes  manufactured  by E.I. DuPont and sold under the  trade name Nafion.
Nafion is  a copolymer  of  polytetrafluorethylene  (PTFE)  and  polysulfonyl-
fluoride vinyl ether  containing pendant  sulfonic  acid groups.   The sulfonic
acid groups  are chemically bound  to  the  perfluorocarbon backbone.

     An  example   of  the  composition is  shown below.   EW,   or  equivalent
weight, is defined  as the weight  of  XR  resin that neutralizes  one equiva-
lent of base.
                                     334

-------
      CN   •=  O
CO
3     LU


S     5
CC     D_
O     J-
"^     cc
cc     O
LU     0-
>-

m
S
LU LU
"> £
CO <
                   z
                   0_
                LU
             co
CC O CO
CO Z
< O
O O
      cc  Lu
      O  cc
      I-  O
      CO  CO


      tr  O
                             LU
                                             LU

                                       LU    Q
                                       Q LU <-)
                                       X Q CC
                                       cc O
                                       I- CC
                                       O h-
                                       LU O
CJ
UJ

LU

LU
O
                       I— H-
                       LU LU
                       ^^
                    ft* CO CO
                                       LU cc fr, ^
                                                 O
                                                 H-
                                                 CO
                LU
                cc  ^
                UJ  CC
   ^ ^<- ^. — u-  "J
LU < < O LU LU  X
   O O O CO CC  h-
                                                                          •i
                                                                          0)
                                                                          co
                                                                          CO
                                                                          tfl
                                             O
                                             CO
                                             C3
                                             0)
                                             CO
                                                                                           o>
                                                                                           i1
                                                                                           s
                                                                                           00
                                                                                          •H
                           335

-------
                                 1240 EW
                           •(CF2CF2)8	CFCF2	


                                        1
                                        CF2

                                        FC '—*	CF3
                                        I
                                        0

                                        CF2

                                        CF2

                                        S03H
     Membranes  are  also characterized by their ion  exchange  capacity (IEC)
- milli-equivalents (meq)  of sulfonic  acid/dry  weight  of  membrane.   The
relationship between IEC and  EW  can be  expressed  as:
     IEC
1000
 EW
(1)
     Generally, IEC is  determined  by  acid/base titration methods and gravi-
metric weight analysis.

     Three  electrode  potentiostatic  systems  utilizing  the hydrated  solid
polymer electrolyte as  the electrolyte for the electrochemical  sensor cell
were used in all cases.  Among  the unique  design  characteristics of the SPE
sensor are:

      • The electrolyte is embedded in the solid polymer  material,  provid-
ing longer life, greater stability, and  improved  reliability.

      • Electrodes are  firmly attached  to  and embedded in  the  electrolyte
sheet, also contributing to  enhanced  life  and  reliability.

      • The sensor  contains  no  corrosive  liquids—the  only liquid in the
sensor is distilled water.

     A detailed  description of pumped SPE CO  sensor cell response charac-
teristics may be found  in  reference 3.
                                THEORETICAL

     Prior to  experimentation,  a rudimentary theory of  sensor  response  was

                                    336

-------
 derived  in order  to assess  the  predicted  order  of  magnitude of  sensor
 response  to  carbon  monoxide diffusion  conditions  (that  is,  yA/ppm  CO).
 Pure  diffusion in  unsteady-state  conditions  is  given  by  the  following
 expression.             '       •  •     .
      D
32C
"9X2"
              3t
                                                                          (2)
      D = diffusion coefficient
          (0.175 cm2/sec for CO)

      C = CO concentration (moles/cm3)

      X = distance up the tube

      t = time,  sec (from time of admission of sample  to  base  of  tube)

      The general solution to the equation is:
      C  =  (a/t1/2)  exp (-X2/4Dt)
                                                                  (3)
where  a is  a parameter to be determined.  Next, at X - Q (base  of  tube)  the
CO  concentration  is  the  ambient  value,  C^.   With  this  substitution  we
arrived at
     C =  C*  exp  (-X2/4Dt)

The flux  of  CO molecules  at the electrode surface is as follows:
  DA
               ,  (at  X = L)
                                                                  (4)
                                                                 (5)
A  =  -rrd2/4 is  the cross-sectional  area  of  the  tube.   With  the reasonable
assumption that  all  CO molecules that hit  the  electrode  surface react, the
electrochemical  current  generated is:
     i = zFJ =? -zFDA

( 3C/ 3X)  is  obtained  by differentiation of  Equation  4,
arrived at an expression for  current,  i:
         zFALC..
         —-•  i i.y
           2t
           exp (-If/4Dt)
The steady-state current  occurs  when di/dt  = 0 i.e.,  at

     t = L2 /4D

                                    337
         (6)

Finally,  we



         (7)




        (8)

-------
 Substitution of Equation 8 into 7 yielded
      .    2zFAC*D       , 1N
      i  - 	=	   exp (-1)
(9)
First,  an estimate can be  made  of expected  response  time from Equation  8.
Tube  length is L = 10  cm and D =  0.175 cm2/sec.  Using  these values, t =
143 seconds was computed.  Secondly, the steady-state current  was  estimated
for 1 ppm CO by means of  Equation 9.  Note  that  tube  area A = 2.85 cm2,  zF
*  2  x  96,500  coul/mole,  and  1  ppm  CO  is  equivalent  to  4.08  x   10"11
moles/cm3.   Using these  numbers,  a sensor  response  of 0.29  yA/ppm CO was
calculated.   The simplified diffusion model provided predictions  of  opera-
ting  characteristics that  were  sufficiently promising  to  proceed with  an
experimental study.
                           RESULTS AND DISCUSSION

Wall-Mounted  Transducer Module

     A  simplified schematic  diagram of a typical  diffusion sensor cell  is
shown in Figure 2.   With this device,  ambient air enters  the  open end  of
the  diffusion tube, diffuses  the length of  the tube to  the sensing  elec-
trode through an integral interference filter  (not pictured), and  reacts  at
the  sensing electrode.   As will  be described  later,  the  geometry of  the
diffusion   tube  greatly  influences  the  observed  diffusion   sensor   cell
response.
                                  SPE
                                  SENSING
                                  ELECTRODE
                                                DIFFUSION
                                                TUBE
Figure 2.  Schematic  of  SPE Diffusion Head Gas Sensor.
     A  photograph  of  a prototype  instrument  incorporating  a  diffusion
sensor cell  is shown in  Figure 3.  The  diffusion head  transducer module,
designed  for  long-term  unattended use,  is a  wall-mounted  instrument for
                                    338

-------
Figure 3.  Remote diffusion head  transducer  monitor.

monitoring ambient levels  of  CO.   The  output of  the transducer module, when
used in conjunction with  a dedicated control module shown  in Figure 4, can
be displayed directly  in  parts  per million  (ppm).  The  control module also
supplies power  to the  transducer  module.    The  transducer module  can also
operate  in  conjunction with  specified  external  power and  data acquisition
systems.  The present  system is designed to operate  over a +7.2 to 21 VDC
range  of  input  voltage and  to produce  an  output  voltage of  0 to  5 VDC.
When used with  the control module,  CO concentration  readings  are  obtained
over the range  from 0  to  100  ppm  CO.  If an external  power and data acqui-
sition system is  used, the instrument range is  0 to  50 ppm CO.   In opera-
tion, total current draw  of the transducer module is  less than 5 ma.  Cali-
bration  of  the unit  is  readily  accomplished  directly  at  the transducer
module.

     Also visible in Figure  3 is  the removal tube  that  houses the integral
interference filter.   In  use,  the interference  filter  is  located directly
beneath  the sensing  electrode.   Using a dry gas stream, this  filter had a
lifetime  of  23,300 ppm-hr before breakthrough  was observed.   Use  of the
removable tube  facilitates changing  the interference filter without  removal
of  the sensor  cell  from  the transducer  module  assembly.   Not  visible in
Figure  3 is the  porous metal  disk  incorporated as  part  of  the diffusion
tube   assembly   to  improve  the   flow/response   characteristics   of  the

                                     .339

-------
                                        CARBON MONOXIDE MONITOR
                                           CONTROL MODULE
Figure  4.   Transducer and Control Modules Carbon Monoxide Monitor System.


transducer  module.   The only routine maintenance  for the transducer module
is periodic (biannual) filling of  the water  reservoir with distilled water
and  checking the  interference gas  filter.   The  unit  contains  no moving
parts.  Features  of  the transducer module are summarized in Table 1.

     Through proper  choice of  the geometry of the diffusion tube, especial-
ly by  varying the  length to  diameter  ratio  (L/D  ratio) of  the diffusion
tube, a diffusion  sensor cell  response  that was independent of the external
air  flow  rate could  be  obtained.   One problem associated with  the  use of
such a  diffusion sensor cell was  an increase  in the  flow dependence of the
diffusion   sensor   cell  response  when   the   diffusion  tube  length  was
decreased.   A porous metal  disk,  incorporated as  an integral part  of the
diffusion tube assembly, minimized this  flow dependency.

     Figure  5 shows  the  effect  of  external  air  flow  rate on diffusion
sensor  cell response  for  a diffusion sensor  cell having  an L/D ratio of
0.91.   In this particular example, the diffusion  tube was  completely open,
with no interference  filter present.   As  is  clearly  visible,  the diffusion
sensor  cell  response  was highly dependent on external air flow rate.

     Also shown  in Figure 5 are  results  obtained  with  the  same diffusion
sensor  cell, but this  time with  a porous metal disk  across  the  entrance to
                                     340

-------
                   TABLE 1.  TRANSDUCER MODULE FEATURES
VISUAL DISPLAY - CASE OPEN

  •  Water level, Purafil color

VISUAL DISPLAY - CASE CLOSED

  •  None

FRONT PANEL (INTERNAL)
CONTROLS AND ADJUSTMENTS

  •  Zero adjustment

  •  Span adjustment

POWER INPUT VOLTAGE

  •  7.2-21 VDC

OUTPUT VOLTAGE

  •  0-5 VDC (0.5-4.5 VDC linear)

REMOTE OUTPUT

  •  0-100 mVDC, linear (for cali-
     bration - within case, banana
     jack fittings)

ENCLOSURE

  •  Fiberglass case, NEMA 4 or
     equivalent, 7 1/4" (H) x
     5 1/2" (W) x 4 3/4" (D)
SAFETY
     Intrinsic safety for methane-air
     mixtures (Class 1-D) to pass
     MSHA approval and certification
CO SENSOR CELL

  •  Diffusion tube, operate
     at mine air velocities
     of 50-400 fpm

  •  Noise:   less than 1 ppm
     CO

  •  Zero drift (30 days):
     ±1 ppm CO

  •  Precision (30 days):
     ±5% of  reading

  •  Relative humidity:
     0-99+%  RH

  •  Operating temperature:
     1-40°C

  •  Accuracy:  ±1 ppm to 10
     ppm CO, ±10% of reading
     over range of 10-100
     ppm CO

  •  Response time:  2 .min-
     utes to 90%

  •  Interferents:  highly
     selective to CO in pre-
     sence of CH^, NO, N02,
     C02, H2, NH3, H20 vapor
     with Purafil filter.
     Filter  removal where
     early fire warning re-
     quired.
                                    341

-------
z '-D
Ul
(0
Ul
DC
a.
rf 1-4
Ul
S
Z
g 1.2
O
Q.
1
g
•f 1.0

Ul
Ul
-l 0.8
UJ
W

?
I I I

0^^ 	 a U ^°
nX^^

f
.J
-
	 . 	
^__ 	 — '
^-~cr~~~ — """""""
	 : — 	
o 	 " —

O POROUS METAL ABSENT

D POROUS METAL PRESENT



L/D = 0.91
g Or , . . ,
1-
Z
Ul


14 2j
Ul
5
'3
12 g
O
a.
"i
Q.
a
10 .-•»
3.
Ul
Ul
8 -1
Ul

* CO
z
O
a.
CO
Ul
0 100 200 300 400
FLOW RATE (fpm)
Figure   5.   Effect   of   a  porous  metal  disk  on  diffusion  sensor  cell
           response  L/D =0.91.


the diffusion  tube.

     Without the  porous  metal disk,  a 12 percent rise in response was noted
as the external air flow rate was increased from 175 to 375 feet per minute
(fpm).   However,  with the porous  metal disk present, only a 2 percent vari-
ation in signal was  noted over the range 60 to 300 fpm.

     It  should also  be noted in Figure 5 that the response of the diffusion
sensor cell  without  the  porous metal disk present was an order of magnitude
higher than  the  response observed when the porous metal  disk  was  present.
The very high  response  level in  the absence  of  the porous metal  disk  was
due to air flow  into the diffusion  tube.   The response  of  an electrochem-
ical sensor  cell  is  dependent on the sample  flow rate  striking the sensing
electrode surface.   The  presence of  a  porous metal disk  across  the diffu-
sion tube  entrance  significantly  decreased the amount of  direct  flow into
the diffusion  tube  and  up  to the sensing electrode, greatly  improving  the
diffusion contribution  and minimizing  the  flow contribution of  the diffu-
sion sensor  cell  response.

     A diffusion  sensor  cell response level  of  1.53 ya/ppm was  predicted
for  the  diffusion  sensor  cell utilized  in  this  phase  of  testing.   The
average  diffusion sensor cell response level  of  1.45 ya/ppm over the range
from 60  to 300 fpm was in excellent  agreement with the  predicted response.
                                     342

-------
       Figure 6  shows  the  effect  of external  air flow  rate on  a diffusion
  sensor cell having an L/D ratio  of  2.39.  This  diffusion sensor  cell was
  identical to that located inside the  transducer module, but  had an inter-
  ference .filter located inside  the diffusion tube.  A porous  metal disk was
  threaded  into  the  bottom of  the diffusion  tube.  Results are  shown both
  with and without  the presence of  the  porous metal.  Over the  range  40 to
  250 fpm,  a  17 percent  rise  in  signal was noted for  the  diffusion sensor
  cell system without   the porous metal,  while almost no  change  was noted for
  the same  diffusion sensor cell over a  wider  flow rate range after the addi-
  tion -of  porous  metal.  Equation 9 predicts  that  a  diffusion  sensor cell
  having an L/D ratio  of  2.39  will have a  response level of 0.45 ya/ppm CO.
  Diffusion sensor cells placed  in prototype wall-mounted  transducer modules
  had L/D ratios of 2.39, with observed  response levels  on  the  order of 0.45
  ya/ppm.
  1.0
  0.8


I
13
3- 0.6
  0.4
O
a.
ca
  0.2 _
                                  _L
               O POROUS METAL ABSENT

               D POROUS METAL PRESENT
                                                _L
                   100
200             300
   FLOW RATE (fpm)
                                                               400
                                                                             500
 Figure  6.   Effect  of porous metal on diffusion cell response    L/D = 2.39.


      Equation  9  also predicts that, for a given sensing  electrode  area,  the
 response  of a  diffusion sensor cell is inversely proportional  to the length
 of  the  diffusion  tube.   This  is  demonstrated  in Figure  5  and  6,  where
 increasing  the  length of  the  diffusion  tube  (decreasing the L/D  ratio)
 resulted  in a  decrease in diffusion sensor  cell  response level.   Figures  5
 and  6 also  demonstrate that,  without  a porous metal  disk,  the smaller  the
 L/D  ratio,  the greater  the chance for  direct-air flow  into   the  diffusion
 tube, resulting  in a more flow-dependent response.

      The  response  of  the  tranducer module was  tested as a  function  of  CO
                                      343

-------
concentration.  These  results  are shown in Figure 7.  A linear relationship
was  observed  between the diffusion  sensor cell  response  and CO concentra-
tion.   This  line has  a slope  of 0.58 ya/ppm,  in good agreement  with the
predicted response  level  of  0.45  ya/ppm,  and an intercept of -0.19 ya(-0.33
ppm).   The correlation coefficient of  the  data  points  is  0.997, demonstra-
ting excellent  linearity.   The prototype  transducer module shown in Figure
3 was  designed  for  operation over the  range  from 0 to 50  ppm  CO;  the data
displayed  in  Figure  7 demonstrate  that  the  transducer  module  has  the
required linearity  over this range of CO  concentrations.
   60
   50
5. 40
111
to
z
o
o_
in
DC
_J
IU
   30
   20
o
CO
ID
ul 10
5
                          J_
_L
                                         J_
_L
            10     20      30     40      50     60
                            CO CONCENTRATION (ppm)
                                                      70
                                                              80
                                                                     90
                                                                            100
Figure 7.  Response  vs.  concentration,  diffusion sensor cell.
     The  transducer  module  was  also  tested  to  determine  response  at
extremely high  CO  concentrations as  might be observed during an underground
mine fire.   These  results  are shown in Figure 8.   Pumped sample CO sensors
may show  severe signal non-linearity at  extremely high (i.e., 1-5 percent)
CO levels as a consequence of diminishing  availability of surface reaction
sites or interference  by desorption  of  C02  reaction product (4,5).

     Increased  linearity is observed with a diffusion instrument in compar-
ison with a pumped instrument,  especially at  high concentrations  of CO.
                                     344

-------
100,000 -
                     100
                                   1,000
                                CONCENTRATION OF CO (ppm)
10,000
50,000
Figure  8.  Application of the SPE  Diffusion  Cell for the detection  of  high
           concentrations of CO.
This  is  due to  the  concentration  gradient  established  over the length  of
the  diffusion tube.    The  sensing  electrode  experiences  a CO concentration
lower  than that  actually  present  at  the entrance to the  diffusion  tube.
The  concentration gradient  across  the  diffusion  tube  is  given by  Pick's
first  law  of  diffusion:
     J  =  -D  (3C/9X)
                  (10)
where J  is  the  flux of the diffusing species per unit area, D  the  diffusion
coefficient  of  the  species  of  interest,  and  C  the concentration of  the
diffusing species  at any point X along the diffusion path.

     The response  of  a  diffusion sensor cell  having an L/D  ratio of  2.39
was tested  at  three concentrations of CO/air and 5 percent CO/N2»   A linear
response was  obtained  up  to  1000 ppm  (0.1  percent   CO/air).    After  10
minutes  of  sampling 5 percent CO/N2, the diffusion sensor cell was  respond-
ing at  70 percent  of  its  expected value  as shown in Figure  8.  These  data
clearly  show that  the  diffusion CO sensor can function  at  extremely  high
concentrations.

     The transducer modules were also tested to determine response  over the
range 0-40°C.   A background change of 1 to 2 ppm  CO was measured over  this
temperature  range, with  only a  minimal  change in  span response observed.
It  was  further  verified  that   the  transducer  module  assembly,   when

                                     3'45

-------
thoroughly dried, may  be  subjected to storage temperatures  as  low as  -55°C
with no harm to  the module.

     Long-term calibration stability testing has revealed the  rate  of loss
of sensitivity to  be  1 percent per month,  indicating  that monthly calibra-
tion checks are  sufficient.

Personal Dosimeters

SPE Diffusion Cell CO  Dosimeter

     A  prototype personal  dosimeter  utilizing a CO  diffusion  sensor cell
was  also  designed,   constructed,   and   tested.    Unlike  the  wall-mounted
transducer module,  the diffusion dosimeter may be worn  either  clipped onto
a  belt  or in  a shirt  pocket.   A  photograph of  the prototype  diffusion
dosimeter is shown in  Figure 9.   The diffusion tube in  this instrument has
an L/D  ratio of 0.91.   A replaceable  interference filter  is  an integral
part  of  the  diffusion   tube  assembly;  a  photograph  of   the  diffusion
dosimeter with the filter  removed  is  shown in Figure  10.
Figure 9.  External view,  CO diffusion dosimeter.
                                     346

-------
Figure  10.   Exterior view, CO diffusion  dosimeter with interference  filter
     
-------
                                     o
                                     o
                                     CO
                                     o

                                     .8
                                  •
                                      o
                                      o
                                      00
                                         E
                                         a.
                                         EC
                                      O
                                      o
                                      CM
                                               I
                                               •H
                                               CO

                                               -§

                                               rt
                                               o
                                               •H
                                               CO
                                                         01
                                                         CJ
                                       O
                                       O
                                                         CD

                                                         Pi
CO

O

-------
            TABLE 2.

Flow rate Angle of incidence
200 fpm 0°
+45°
-45°
+90°
-90°
0°
400 fpm Oo

+45°
-45°
+90°
-90°
0°
600 fpm 0°
+45°
-45°
+90°
-90°

O.LUIN uuaiMETER
• 	 	 	 _
Response ratio
1.10
1.11
1.13
1.11
1.13
1.11

1. 13
1.14
1.14
1.18
1.20
1.14
1.15
1.21
1.22
1.23
1.24
1.18

Further  studies  are  being performed to minimize this dependence.


and  10oTnrVnVtUdi^ ^ performed  over the range  from  0 to  80  ppm CO
               °

between the CO
                         dosimeter and the  standard  pumped  dosimeter.
   m        second test, sPan g^s was introduced  to  the  diffusion  dosimeter
   means of the calibration device  for  a two-minute period.  The  diffusion

        '
                                  to the


    Results  of the  linearity  testing presented  in Table  3  demonstrate a

                                   349

-------
                            °UMPED DIFFUSION DOSIMETER RESPONSE Ippm)
Figure 12.  CO diffusion  dosimeter linearity data.
_ 	 _ 	 """" — — — ~ "
	 — — ~~ Response to test gas*
Diffu


*A11
ision dosimeter
R0462
P0588
P0624
P0622
R0056

values expressed as
100 ppm
100
100
100
100
100

a percentage
491 ppm
101
99
101
104
98

of the tes t
750 ppm
101
101
99
103
98

1010 ppm
100
98
101
99
97

gas concentration.
 linear response for each dosimeter  over  the range 100 to  1010  ppm CO.   The
 values in  Table 3  are the  diffusion dosimeter responses  expressed  as  a
 percentage of the test gas concentration.

      The CO diffusion  dosimeter  responses  were also tested as a function of
 temperature over the  range from  1  to 40 °C.   Typical results  are shown in
 Figure 13.   Thermistor circuitry was included to compensate  very closely
 the span response over the temperature range  of  interest.   Figure 13 shows
 both  the uncompensated diffusion  sensor  cell  response  (in  ya) and the
 compensated diffusion  dosimeter  output (in ppm).

      The dosimeter  function of  the diffusion  dosimeter was  tested against
 that  of  a standard dosimeter.   Typical  data are  shown  in  Table 4.   To
 obtain  these data,  the diffusion  dosimeter  was  placed  in an  air  stream

                                      350

-------
  120

  110

  100

  90

  80

W 70


I 6°
UJ
CC 50
CC
UJ
!D 40

I »
O
D 20

  10

   0

  -10
                   1  I   I   I
                    o-
                                  I   I   1  I   1  i   I   I  I   1   I  i   I   r
                             O Sensor cell current (/ja)
                             D Compensated dosimeter response (ppm)
                               Test gas = 100 ppm CO/air
                                                      Background
                        I   I  I   I
                                  J	I
                                         J	I
                          8  10 12  14 16 18  20 22  24 26 28  30 32  34 36 38  40 42
                                    TEMPERATURE (°C)
Figure  13.  Response  of a  typical  diffusion dosimeter  as  a  function  of
             temperature.
                      TABLE 4.  DOSIMETER COULOMETER DATA

Time
(min)
0
5
10
15
20
25
30
Average
Diffusion response
(ppm)
60
58
60
60
59
60
61
59
Pumped response
(ppm)
61
58
59
59
58
59
60
59
Calculated
Coulometer output

Measured  output
                           29.5 ppm-hr.

                           30   ppm—hr.
29.5 ppm-hr.

30   ppm—hr.
flowing  at 400 fpm.   CO concentrations were  measured with both a  diffusion
dosimeter and a standard dosimeter.   After  the response  of both dosimeters
had  stabilized,  their  coulometers  were  discharged.    CO levels  were  then
integrated for  one-half  hour,  and  the coulometer values read  after  this
period.    During  this  test  period,  instantaneous CO  concentrations  were
recorded at five-minute intervals.
                                       351

-------
 The data  shown in Table  4 include  the  instantaneous dosimeter  response,
the average value,  the calculated dosimeter value, and the  dosimeter value;
as  read out  oh the  support  console.   Very  good agreement  was  obtained
between the two dosimeter values  in all cases  tested.   This  showed that the
diffusion  dosimeter will function as efficiently as  a  standard  (pumped)
dosimeter.

     Long-term  stability testing  of  the  diffusion dosimeter response indi-
cated that  the  calibration  stability  dropped  less than  10  percent  over  a
three-month period.    This  indicates  that  monthly calibration checks  are
sufficient for  accurate instrument usage.

SPE Diffusion Cell NO  Dosimeter

     The diffusion  cell technology described  previously  for the  CO  diffu-
sion dosimeter  was applied  to  the development  of  an NO diffusion dosimeter.
The diffusion sensor cell utiized for  detection  of NO was identical to that
for  CO, except  the sensing electrode catalyst   was  changed  to  a  graph-
ite/Teflon mixture.   To avoid interferences from such  gases  as  NOa,  H2S,
and  S02,  an interference  filter   consisting of  triethanolamine (TEA)  on  a
molecular  sieve was  placed  in front  of the sensing electrode;  CO does not
react on the graphite  sensing  electrode.

     The NO  diffusion dosimeters were packaged  in cases similar  to those
for  the CO diffusion  dosimeter;  a photograph  of  an NO diffusion  dosimeter
is  shown  in Figure  14.  As  with  the  CO  diffusion  dosimeter,  the  interfer-
ence filter may be  removed  easily to  facilitate  changing the  filter mater-
ial.
Figure  14.   Exterior view,  NO diffusion dosimeter.

                                     352

-------
     Figure  15  shows the results of  flow testing for three diffusion.dosi-
meters.   Each  data  point is  the average  of  'six values,  corresponding  to
various  orientations of the diffusion  dosimeters in the moving air  stream.
Diffusion  dosimeter  responses  were compared with NO levels as measured  by a
prototype  pumped direct-reading NO detector.  The response ratio plotted  in
Figure  15  is   defined  as  the  ratio of  diffusion  dosimeter  response  to
direct-reading  detector response, and  is  plotted as a function of external
air  flow rate.   NO  concentrations were  held  at  ^50 ppm.  The  three dosi-
meters exhibited less  than a 10 percent variation in response over the  wide
range of air flows indicated.





o

cc
LU
C/5
2
O
0_
C/5
LU
CC.







\.&
1 1
i . i
1.0

0.9
0.8

0.7

0.6


0.5


0.4

0.3

0.2
0.1

1 1 1 | 1 	 T 	 — 1 	 ~
0
n n -n 	 °
D 	 D 	 -~°
— . . _
- -.''"-_

— •«

— _


- '_


O DNO-01
O PNO-03
* ' M
D DNO-05
~ Test gas = -50 ppm NO/air ~
-
i i i it i i
>




















100 200 300 400 500 600 700 800
                               FLOW RATE (fpm)
Figure 15.  NO diffusion  dosimeter  flow studies.
     A  plot  of  diffusion  dosimeter  response  versus  the  direct-reading
detector response is shown  in  Figure 16.   Data points  were  obtained for NO
concentrations  from 0 to  100  ppm.  A least-squares  analysis  of  the  data
plotted in Figure 16 yielded a slope of 0.99  ya/ppm and an intercept of 0.1
ya.  This  shows the response of the diffusion dosimeter is  as  linear  over
the  range  0  to 100 ppm  NO as  is the response  of  the direct-reading  NO
detector.
                                    353

-------
   110


   100


   90

1
LU 80
to
2
a!. 70
CO
LU
cc
-j 60
LU
O
2 50
g
CO
LL.
LL
Q
O 30
             40
             20
             10
                                                     I
                   10   20   30    40   50   60   70    80   90   100  110

                     DIRECT-READING NO DETECTOR RESPONSE (ppm)

Figure 16.  NO diffusion  cell response  as  a function of NO detector
            response.
     The variation  in diffusion dosimeter output  as  a function of tempera-
ture is  plotted in Figure  17.   Both  the  diffusion sensor  cell  output (in
ya) and  the  diffusion dosimeter response  (in  ppm)  are shown for background
and  span  response   to 28  ppm  NO.   Thermistor   temperature  compensation
circuitry was used  to compensate the span response.

     The  compensated  diffusion dosimeter  response  varies  by ^10  percent
over the temperature  range  from 0  to 40°C.  For this experiment, the diffu-
sion dosimeters  were  placed inside an environmental  chamber,  and 28 ppm NO
was introduced  through the  calibration device.

     Table  5 lists results  of  coulometer  tests for  three  diffusion  dosi-
meters.  The  coulometer calibration was tested  by  introducing 28 ppm NO to
the diffusion dosimeters  for  30 minutes at a  flow  of  120 cc/min.  Column 2
of  Table  5  lists  the  calculated  coulometer  response,  in  ppm-hr,   while
Column  3 gives  the  actual  observed  values.    As  can be  seen, excellent
agreement was obtained in all cases.
                                     354

-------
- o
                                                    CO

                                                    0)
                                                    I
                                                    o

                                                    •H

                                                    CO
                                                    <4-t

                                                    O
                                                    1
                                                    a)
                                                    &

                                                    4i

                                                    CD
                                                    Vl
                                                    2
                                                    a)
                                                    0)

                                                    H
erf) iN3aano Tiao aosNas
              a3i3iAiisoa
          355

-------
             TABLE 5.   NO DIFFUSION DOSIMETERS. COULOMETER DATA

Dosimeter
no.
DNO-01
DNO-03
DNO-05

Calculated
reading
14.5 ppm-hr
14.5 ppm-hr
14.0 ppm-hr

Actual
reading
14 ppm-hr
14 ppm-hr
14 ppm-hr

Background
0 ppm-hr
0 ppm— hr
0 ppm-hr

Time
4 1/4 hr
4 1/4 hr
7 1/3 hr

     The  fourth column lists the results  of  background coulometer studies.
For  this  experiment,  the  coulometers  were  discharged  and   the  diffusion
dosimeters  placed inside  a glove box.   The units  were removed  after the
time periods  indicated,  and the  coulometers  were read  out.   No observable
drift was  detected in the  coulometer output over  the time  frame  of these
experiments.
                                   SUMMARY

     Gas  detection  instruments  highly  specific for  CO  and NO  utilizing
diffusion  sensor cells  have been  developed.    All  diffusion  sensor  cells
utilize a  solid  polymer.electrolyte with integrally bonded  fuel-cell  elec-
trodes.  Linear  responses were obtained  over  wide  ranges of  test  gas  con-
centrations.  Responses  independent of external air flow rate,  over a wide
range of  air flow  rate,  were also observed.   The  instruments are  fully
temperature-compensated  over the  range from 1 to 40 °C.   Excellent  calibra-
tion stability has  also  been observed.
                              ACKNOWLEDGMENTS

     This work was  partially supported by the U.S. Department  of  Interior,
Bureau of Mines.  The authors would  like  to  thank Drs.  J.  Emery Chilton and
George Schnakenberg  of  the Bureau of Mines, whose many helpful suggestions
and comments  contributed   significantly   to  the  technology  developments
described here.
                                REFERENCES

1.   Gruber, A.H., A.G.  Goldstein,  and A.B. LaConti.   1979.   A new family
     of miniaturized self-contained CO dosimeters  and  direct  reading detec-
     tors.  Paper presented at U.S. Environmental  Protection  Agency Sympos-
     ium  on Personal  Air  Pollution  Monitors,  Chapel  Hill,  NC,  January
     11-13.
                                    356

-------
2.   Kosek, J.A., A.B.  LaConti,  and A.G. Goldstein.   1979.   Development of
     fuel  cell  gas  detection instruments  for  use in  a mine  atmosphere.
     Final report, U.S. Bureau of Mines  Contract HO 357078,  March.

3.   LaConti, A.B., M.E. Nolan, J.A. Kosek,  and J.M. .Sedlak.   1980.   Recent
     developments in  electrochemical SPE sensor cells for measuring  CO and
     oxides of  nitrogen.   pages  551-573 in ACS  Symposium Series No.  149,
     paper  no.   CHAS  40,   Second  Chemical Congress of  the  North  American
     Continent,  Las Vegas, NE, August.

4.   Kosek, J.A., and A.H. Gruber.    1981.   Development  of  improved  detec-
     tion  instruments  for  toxic gas  contaminants  in mining  atmospheres.
     Interim report, U.S.  Bureau  of  Mines Contract  HO  395132,  February.

5.   Kosek, J.A., and  A.B.  LaConti.   1982.   Development  of  improved  detec-
     tion  instruments  for  toxic gas  contaminants  in mining  atmospheres.
     Final report, U.S. Bureau of Mines  Contract HO 395132,  January.
                                    357

-------
              LABORATORY  STUDIES  OF A PASSIVE ELECTROCHEMICAL
                 INSTRUMENT  FOR MEASURING CARBON MONOXIDE
                             Vincent  A.  Forlenza

                             Energetics  Science
                 Division  of Becton  Dickinson and Company
                                  ABSTRACT

     Diffusion  or passive electrochemical  sensors offer manufacturers  and
end  users  many  advantages,  such  as   lower  cost  and  less  maintenance.
Despite these advantages,  many professional hygienists have  been skeptical
of the accuracy of  these sensors  when they  are exposed to variable bulk air
movements.  This  study quantifies the effect  of "convection  on the ECOLYZER
210 passive carbon  monoxide monitor.   The  instrument was operated  at con-
stant concentration, while the face  velocity of  the  sample  impinging on the
sensing  system was  varied  from  approximately  5  ft/min  to  2,426  ft/min.
Tests were run at 25 ppm and 197  ppm.  The  errors in concentration reading
registered by the instrument were less than 6.5 percent.  The ECOLYZER 210
convection barrier  attenuates  most of the unwanted effects  of convection on
the sensor, eliminating any  strong dependency  on face velocity.
                                INTRODUCTION

     Health and  safety professionals require small,  lightweight,  portable,
and reliable monitors  for many chemical and physical agents.   Devices that
rely  on  sample  pumps  in many  cases are  too  bulky  to  be  effectively  and
comfortably  used as  personal  monitors.    Recent  developments in  passive
monitoring devices  have enabled manufacturers  to market  small instruments
and badges that  are ideal for  personal  monitoring.   Easily worn on a collar
or a belt, these devices  minimize  the distraction and discomfort of wearing
a  monitor.   They  often  offer  the  user  the additional  benefits   of  lower
cost, and less maintenance,  because they have fewer parts.  When  the moni-
tor is a passive type  instrument,  it generally  requires  less power than the
sample-draw type, so that the  need for  large battery packs is  eliminated.

     In the early 1970's,  electrochemical  sample-draw devices  for  measuring
ambient carbon  monoxide, such  as  the ECOLYZER 2000 series  from Energetics
Science, provided  the health  and  safety  professional  with a  portable  and

                                     358

-------
highly accurate survey  tool,  but one that was  too  large  for personal moni-
toring.  The ECOLYZER 210  is  designed as a personal  monitor.   The ECOLYZER
210 has a digital display  for survey work and  alarm  functions  for use as a
safety tool.   It weighs only thirteen ounces, and  can be worn  on  a belt.
This paper presents  laboratory test  data on this  passive monitor.

The Theory of Passive CO Monitoring  Devices

     The ideal  passive  carbon monoxide  monitor is not affected  by changes
in temperature, pressure,  humidity,  the  presence  of other chemical species,
or  convection,  but  only  by  changes  in  ambient  concentration   of  carbon
monoxide.  The effect of convection  on active monitors is eliminated by the
use of a constant-volume  pump.  Passive  devices  use  other  means to elimi-
nate the effect  of  bulk air  movement.   William J. Lautenberger  and fellow
researchers (1) list four  means of eliminating the  effect of  convection on
a passive monitor.   They are:

1.   The use of wire screens  followed  by a stagnant air space.

2.   The use of thin attenuating sheets.

3.   The use of a tube-like  cavity,  such as the Palmes tube ,or a series of
     cavities that have a  length-to-diameter  ratio greater than three.

4.   The use of  a  permeation  membrane directly in front  of the  collection
     medium.

The ECOLYZER 210 employs  three  of  these four means  of  attenuating convec-
tion,  the  thin attenuating  sheet,  the  tube-like  cavity, and  a  permeation
membrane.

The ECOLYZER 210 Sensing System

     The ECOLYZER  210  carbon  monoxide  sensing  system  consists of  three
parts:  the electrochemical  sensor, an  interference  filter, and a convec-
tion barrier, or attenuating sheet.   Figure  1  is  a  diagram of the system.
The sensor is an Energetics  Science patented  three-electrode  electrochemi-
cal cell.  The cell  functions according  to the following mechanism.  Carbon
monoxide diffuses through  a semipermeable  membrane into the cell.  The cell
contains a sulfuric acid electrolyte.    Inside  the cell  carbon monoxide
molecules are oxidized  to  C02. according  to the following equation:
CO + H20 •*• C02 + 2H+ + 2E~
                                                                         (1)
     This  reaction  generates an electrical current  that  is  proportional to
the ambient  concentration of carbon monoxide.  The  current  flows between a
working  and  counter electrode.   A third electrode  is  used as  a reference
electrode  to maintain the  cell at  a  constant bias.   The current  is  then
amplified  and used  to  drive  a digital  display, showing the concentration in
parts per  million.

     A selective  absorbent  is placed  between the ambient air  and the cell

                                     359

-------
                POLYMERIC DISC
             (DIFFUSION BARRIER)
            INTERFERENCE FILTER
      SEMIPERMEABLE MEMBRANE
            WORKING ELECTRODE
           REFERENCE ELECTRODE
                                                       CAVITY WITH LENGTH TO
                                                       DIAMETER RATIO 3/1
                                                       3 ELECTRODE
                                                       ELECTROCHEMICAL
                                                       SENSOR
                                                       COUNTER ELECTRODE
                                                       AMPLIFIER CIRCUITRY
Figure  1.   ECOLYZER 210 CO sensing system.
to  remove any  interfering  gases  that  might give  spurious  readings.    In
sample-draw  instruments,  the  sample  is  pumped  through  the  interference
filter and over the semipermeable membrane,  where the CO diffuses  into the
cell.  In a  passive instrument, the  sample  must diffuse through  the selec-
tive filter,  so sizing of the filter and  its media is critical.   In appli-
cations  where selectivity is  not  critical,  the  interference filter  can be
eliminated.

     In  order to eliminate  the effects  of  bulk air  movement,  a  polymeric
disc is  used  in conjunction with  the  interference filter to  eliminate  most
convection through the filter.  In  instances where the interference  filter
is not employed,  the disc is used  with a tube that has a length-to-diameter
ratio greater than  3:1.   The polymeric  disc was chosen so  that pore  size
and thickness  allowed maximum  diffusion of  carbon  monoxide, while  attenu-
ating most of  the convection.   If  the  effects of  convection can be  elimi-
nated, then Pick's  Law of Diffusion applies:
                                                                          (2)
                                 J - -DAG
                                      A X

where J s mass flux  of  migrating carbon monoxide
      D » diffusion  coefficient of the carbon monoxide
     AC - concentration gradient that exists in space
     AX

     In order  to  verify this, a  constant  gas  concentration was passed  over
the sensing  system at various ? velocities.   Data from  tests  run at 197 vppm
                                     360

-------
 CO  and 25 ppm CO in  air  are presented  in  Figures 2 and,  3.   Table 1 gives
 the error due to changing the  face  velocity for the test  at  197 ppm.   The
 data from both  tests  show only a very  slight dependence  on face  velocity
 over an extremely wide range.   Below 20.6  ft/min,  the  errors are  slightly
 negative.   At these low air  velocities,  the sensor  is  being starved.   The
 concentration gradient is changing  as  the sensor oxidizes  CO to C02; how-
 ever,  the errors are  small,  only 3.1 percent  at  5.1  ft/min for the 197  ppm
 test.   It is  safe  to assume that these velocities   are  outside  the  normal
 use conditions for the instrument.   On  the other hand.,  the positive error
 due to face velocity  is  only 6.3 percent  at  27.5 mi/hr  (2,425  ft/min)  in
 the 197 ppm test.  Data  from  the 30  ppm test show  a  similar trend, indi-
 cating that  the convection barrier  is effective  over a range of concentra-
 tions.
       TABLE  1.   INSTRUMENT ERROR DUE TO THE EFFECT OF FACE VELOCITY
Face velocity (ft/min)
5.1
10.3
20.6
51.6
103.2
154.9
206.5
258.0
310.0
361.0
413.0 : • - • .
464.5
516.0
731.0
970.0
1218.0
1311.0
2426.0
Percent error*
-3.1
-2.1
0
1.6
2.6
3.6
3.6
4.2 '
4.2 ; .'-•••••
4.7
4.7
4.7 : .-..-,
5.2 •' "'• "''• ' "
. 5.2
5.7
5.7
' 5,7 -••••
6.3

*Percent error =[(Instrument  - actual)X 100]/actual.
     To highlight  the  effectiveness  of  the barrier, a test was run with and
without the  convection barrier in place.   Figure 4  illustrates  the strong
flow dependency  of the system once the  barrier  has been  removed.   This is
as expected, and is  supported  by  results from many sample-draw systems.

Instrument Response  Time

     The  porosity  of  the  diffusion  barrier  and  the diffusion  path length
will certainly affect  response time.   There  are  actually three resistances
in  series through which  the  gas must  diffuse:   the  polymeric  disc,  the
interference filter, and  the semipermeable membrane  of  the  cell.   Figure 5

                                    361

-------
    o
 U.-Z
 tu O
 oc o
V
   .\
                                                                         -I P x
                                                                         111 [I -i-
                                                                                            0
                                                                                            O
                                                                                           •H
                                                                                           4-1
                                                                                            n)
                                                                                            lj
                                                                                           4-1
                                                                                            a
                                                                                            8
                                                                                            CO
                                                                                            rt
                                                                                            oo
                                                                                           4-1
                                                                                           4-1
                                                                                           CO
                                                 O
                                                 O
                                                                                           o
                                                                                           o
                                                                                           o
                                                                                           cd
                                                                                           M-H
                                                                                           00
                                                 Cfl
                                                 01
                                                                                           C

                                                                                           I
                                                                                           CM


                                                                                           0)
                      O
                      O5
O
00
                  00
                          362

-------
       40
      30
 O
 O
 CL
 Q.
20
      10--
            X  X
                     100
                                                                         REF.
                                                                        CONIC.
                                  200
                                               300
                                                           400
                                                                        500
                                 FT/MIN  VELOCITY
Figure 3.  Ambient  readings  from  the  ECOLYZER  210 vs.  face  velocity  at
           constant CO concentration.
                                      363

-------
     500
     400
     3OO
Q.
Q.
     200'
     100
                      1	1
20
                     40
                                             1	1	1
                            60
80
  100


VELOCITY

 (FT/MIN)
                                               110
                                                     120
                                            140
                                                                  1(30
                                                                        1 80
                                                                               200
 Figure 4.   ECOL^ZER  210  reading  at  constant  CO  concentration with  and

             without  convection barrier.


                                       364

-------
                     TIME TO ALARM FOR VARIOUS CO CONCENTRATIONS
                     AND ALARM SET POINTS
      LU
 o
: LLJ
             60
             50
             40
             30
            20
             10
                     CO CONC.
                       PPM

                         396
                         588
                       1502
                       1826
TIME TO ALARM
  AT 250 PPM

  21.4 ±.8 SEC.
  14  ±1.3 SEC.
  8.4 ± .8 SEC.
  7.6 ± 1.0 SEC.
                            25
                                         50
                                              75
                        100
                                      % SIGNAL
Figure 5.  Percent  signal vs. time for ECOLYZER 210-averaged data  from  five
           instruments.
                                     365

-------
shows that 90 percent  of  signal  is  achieved in less than 60 seconds.  Typi-
cal alarm data are  also  shown  to  illustrate  that instrument  response  is
quite rapid.

Linearity

     The  instrument  reading should be proportional  to  the  partial pressure
of the gas to be measured in  the ambient air.  Figure .6  shows  that this  is
indeed the  case over  the range  of  the  instrument,  which is 0-1999  ppm  of
carbon monoxide for  the ECOLYZER 210.

Specificity

     The ECOLYZER  210  was  exposed  to  a series  of  gases commonly  found  in
the workplace.   Table 2  shows  the interference  equivalent defined  as  the
concentration  necessary  to  register  as  1 ppm  on  the instrument.    For
example, 270 ppm N02 will register  as  1  ppm on the  instrument.   No response
differs from no  interference  in this  case.   No response indicates  that  up
to the  concentration  tested, the  instrument did  not   respond  to  the  gas
tested.   It is  possible  that, above that  concentration,  the instrument  may
respond to  the  gas.   Ethane  illustrates this phenomenon.   No  interference
indicates that  the  instrument will not  respond to  any  concentration.   The
data in Table  2 are  with the interference filter  in place.   Inspection  of
the data shows that  the ECOLYZER 210  exhibits excellent specificity towards
carbon monoxide.
        TABLE 2.  INTERFERENCE EQUIVALENTS  FOR CO DIFFUSION SENSOR
Gas tested
Concentration
   tested
   Interference
equivalent (ppm)
CH^ (methane)
C0£ (carbon dioxide)
NH3 (ammonia)
NO (nitric oxide)
NO, (nitrogen dioxide)
S02 (sulfur dioxide)
H2S (hydrogen sulfide)
C^2 (acetylene)
C-jHij (ethylene)
C2Hg (ethane)

CaHg (propane)
CH3OH (methanol)
C2H5OH (ethanol)
2-Propanol
99%
99.8%
29.4 ppm
48.2 ppm
387 ppm
21.2 ppm
27.2 ppm
100 ppm
19.4 ppm
50 ppm
500 ppm
105 ppm
500 ppm
500 ppm
500 ppm
No interference
No interference
135
No response
270
145
130
170
135
No response
1200
425
No response
140
750
                                     366

-------
Q
O
     22 •
     20 •
     18
     16'
     14
g-    12
O    10
                         CO PPM   LCD READOUT
  49.8
 101.0
 587.6
 985.0
1499.6
1835.6
45
97
593
984
1507
1850
                                   8     10
                                                12
                                                       14
                                                              16
                                                                    18
                                                                           •20
                                                                                  22
     (BOTH SCALES X 100)
                CO PPM
   Figure  6.   CO vs.  LCD readout (ppm ) .

                                          367

-------
                                CONCLUSIONS

     1.   The ECOLYZER 210 convection  barrier sensing  system  is a  highly
          effective attenuating system, which  all  but  eliminates  the  effect
          of face velocity on the instrument.

     2.   Below a velocity of 20 linear ft/min,  the  sensor  becomes  starved.
          However, the errors  are not significant,  even for velocities  as
          low as 5 linear ft/min.

     3.   The  instrument  shows  excellent  linearity,  response  time,  and
          specificity.
                                REFERENCES

1.   Lautenberger, W.J.,  E.V.  Kring, and  J.A.  Morello.   1981.
     passive monitors.  Ann. Am. Conf. Gov.  Ind.  Hyg.  1:91-99.
Theory of
                                     368

-------
                RESULTS OF TESTING DIFFUSION-TYPE NITROGEN
              DIOXIDE PERSONAL MONITORS AT LOW CONCENTRATION
                             James B. Flanagan

    Rockwell International Environmental Monitoring and  Services  Center
                              Chapel Hill, NC

                                    and

                                Joseph Ryan

                   U.S. Environmental Protection Agency
                        Research Triangle Park, NC
                               INTRODUCTION

     Facilities of  the  U.S.  Environmental Protection Agency's  (EPA)  Clini-
cal  Environmental  Laboratories  (GEL)  were  used  to  test  a  commercially
available  diffusion-type personal  NO2  monitor  based  on  a  design by  E.D.
Palmes (1), under  conditions simulating sub-industrial exposure  levels  and
while  being  worn  by  human  subjects.   Like  many  other   personal  monitors
commercially available,  the  Palmes  tube design is  optimized  for  the  higher
levels of  industrial  exposure.   -The monitors  were tested with low ambient
pollutant  concentrations under  a variety  of  conditions  to establish their
accuracy,  precision,  collection  efficiency,  and the effects  of  orientation
and human  wearers.

     The GEL provides chambers  large enough for human subjects  to exercise
and  move  around  normally  while  being  exposed   to  accurately  controlled
levels  of  pollutant  gases  with controlled  conditions  of temperature  and
humidity.   Delivery  of controlled  amounts  of  the inorganic  criteria  air
pollutants  has .been  engineered  into  the system,  so  that levels  approxi-
mating  the amounts seen in  the  ambient  environment  can  be  delivered  with
high precision and accuracy.  In addition to  air  pollutant  delivery, temp-
erature and relative  humidity can  be varied  and controlled.

     The Palmes tube  system tested.is manufactured by MDA Corporation (2).
The  monitor  consists  of a  tube approximately 4  inches  long,  open  at  one
end,  and  with a wire  mesh  coated  with  triethanolamine  absorbent  at  the
other.  Mass  transfer is accomplished  by passive diffusion  of N02 in air.
Assuming  no  convective  mixing  in  the  tube,  diffusion provides  a constant
sampling   rate almost  independent  of  temperature and   pressure.    After

                                     369

-------
exposure  of  several hours,  the N02 is analyzed by standard chemical, color-
imetric techniques.

     The  colorimetric-microprocessor supplied by MDA in the Palmes tube kit
has  simple  lighted switches controlling its  timing  and analysis functions.
A yellow  filter over a small incandescent  bulb provides the "monochromated"
light  source for the analyzer.  Percent transmittance and calculated ppm-hr
dosage is read out via LED  display.  The  internal  microprocessor controls
the  timer (for timing the color reaction),  provides percent  transmittance,
and  calculates  the  ppm-hr result using internal calibration curves.  Accum-
ulated dosages  in the range  from 1 to 50 ppm-hr can be measured.  The read-
out  device  has  a resolution of  0.1  ppm-hr; for a  typical  8-hour exposure,
this would yield a  maximum resolution of 0.012 ppm in mean concentration.
                                EXPERIMENTAL

     Before  use in  the  controlling mode  with the  Palmes  tubes,  GEL  ana-
lyzers  were  calibrated  no  more  than  three days  before  the  exposure.
Follow-up  calibrations within  two weeks  of  the  initial  calibration  were
also routinely  used  to check for drift.   Follow-up  calibrations  made after
the exposures showed no  cases  of  excessive calibration drift.

     Nitrogen dioxide standard  was  a  dilute NO tank that was NBS-traceable.
Bendix  model 8101B  analyzers  were calibrated  using this  gas  directly for
NO.  Titration  with  ozone was  used to calculate  conversion  efficiency for
the NO/NOX instruments.   Exposures in  the chamber used the  analyzers  as  a
transfer standard.

     Additional  exposures  were  made  utilizing   the  GEL  gas  calibration
system.   The source of  calibration  gas  was a tank of  109  ppm  N02 in  N£
(Airco Industrial Gases),  which was traceable to  NBS.   Partially  humidified
zero air was mixed using Tylan  mass flow controllers with controlling volt-
age set manually with potentiometers.   Data acquisition was under automatic
control  of  the GEL  gas  computer system.   The concentrations  of  N02  thus
obtained  were   monitored   by   two  recently  calibrated   Bendix  NO-NOX
monitors.    A  3-liter  flask  was  used  as  the  exposure  chamber  when
calibrations  were   done.    Flow  rates  through  the  3-liter  flask  were
typically between  2  to 5 liters per minute.   A positive gas flow  from the
exhaust  tube was checked  before  and  after each  session  to assure  proper
operation with no  leakage.   Calibration gas ran through the  flask  at least
15 minutes prior to  insertion  of  tubes.

     For all concentration measurements, the  recently  calibrated analyzers
were taken  as  the  standard.   Through  careful quality control,  quarterly
audits,  and  reference to  NBS  permeation  tubes  at each calibration,  these
analyzers are thought to be  accurate  within ±5 percent.
                                    370

-------
                                  RESULTS

     Manual  checks  of  the  microprocessor's  calculations  of  dosage  were
carried  out  using  samples  spiked  with  standard  sodium  nitrite  solution.
The results of this  calibration  are shown in Table 1.   The  expected ppm-hr
doses are  calculated for the  spikes  using  the  formula in  literature  pro-
vided by the manufacturer (2):
     Q = 2.3 X (ppm-hr)
(1)
In the  present  case, Q (quantity  of N02) is known,  and (ppm-hr) is  to  be
derived.  The conversion  factor is obtained from  the  assumed stoichiometry
of the  color  reaction (3) and  from diffusion theory, assuming  a diffusion
coefficient of 0.154  and  the  geometric constants of the  device.   Units for
Q in the above equation are nanoequivalents  nitrite in a 2—cm3 volume.
                  TABLE  1.  RESULTS  OF  NaN02  CALIBRATION

Q
blank
4 neq.
8
12
20
40
%T
98.9
80.5
64.7
52.5
33.8
12.9
Abs.
0.005
0.094
0.189
0.280
0.471
0.889
Colorimeter
	
2.3 ppm-hr
4.7
7.0
11.9
22.4
Eqn.(l)
	
1.74 ppm-hr
3.48
5.22
8.70
17.4

     Linear regression  of  the  data  shown in Table 1 and Figure 1 revealed a
slope  different  from unity (m=1.28,  b=0.28).    When the manufacturer  was
contacted,  it was  found  that  the microprocessor  algorithm included  an
undocumented  "efficiency  factor" of 0.72 in the  calculation  of  ppm-hr from
a given percent T.  Equation  (1)  then  becomes
     Q = 2.3 X  (0.72) X  (ppm-hr)
(2)
Use  of  this  factor  corrects  the  slope  of  the regression  line  to  within
about 8  percent of theoretical,  so  that the  algorithm used by  the  micro-
processor-colorimeter  appears to  be valid.   The  need for  the  efficiency
factor in  the wet impinger  techniques  utilizing reagents similar to those
used  in  the Palmes  tubes  is  thought  to  arise  from  a  combination  of
stoichiometric  factors  in the color  reaction  and the  uptake  efficiency of
the absorbent during sampling.   It is a simple matter  to  convert the final
ppm-hr  reading  from  the  colorimeter  to a  different  assumed  efficiency
factor:
     (new ppm-hr) =  (old  ppm-hr) X 0.72  /  (new factor)
(3)
     Table  2  summarizes  the  testing  carried  out  in  the  chamber  and
calibration systems.   Table  2 includes only exposures made  in a stationary

                                    371

-------
                                  QNOi/2.3 (ppm-hr)
Figure 1.  Calibration  of Palmes  tubes  analysis system with standard NaN02<
position, either  in the GEL or  in  the calibration flask.
of subject wear are  given later.
Tabular results
     The  first  set of data  taken (18 sample  tubes),  as shown  in Table 2,
illustrates  the  difficulty  of  obtaining measurements at  extremely  low
levels  of net exposure:   the  ratio  of  the analyzed  dosage to  the dosage
supplied  by  the GEL system  is  in error  by a  large percentage,  though the
scatter  as  reflected  in the  relative standard  deviation is  fairly good.
The next  two  entries  in  Table 2, (four sample  tubes)  are blanks exposed to
atmospheres free  of N02  for  different lengths  of time.  This shows  that the
high values  at  low dosage were not  due  to zero  offset.   The  next set of
twelve  replicate  samples  was  exposed to  approximately 2 ppm-hr dosage at
0.5 ppm concentration.   The precision of  estimate was  excellent,  but the
measured/actual  ratio was   still  high.    The   next group  of  samplers  (18
tubes)  was  exposed to high  levels of N02  approximating 10 ppm.   Due  to a
design  limitation of  the GEL, control of  concentrations at these levels is
very poor, so that an accurate  estimate  for concentration and true  net dose
is  not  available.    The  reproducibility  of  the  ensemble  of  tubes  thus
exposed  is  excellent, however.   The  next groups  of  exposures  in  Table 2
were done in the calibration  system rather than  in the GEL.   The purpose
was to  obtain results over  a  wider variety of  conditions  of concentration
and duration  of exposure than is possible in the GEL.   There was no signif-
icant  influence  of   concentration on  accuracy  (as  expresed  by  the  mea-
sured/actual  ratio),   although  there was  an apparent  positive  correlation
when the  measured/actual ratio  was fit to a linear function of net  dose.

     Figure  2  illustrates  results  of   virtually  all  of   the  individual

                                     372

-------
                TABLE  2.   PALMES  TUBES N02 EXPOSURE RESULTS
                           STATIONARY EXPOSURE

Palmes
Exposure Exposure
time cone
Calculated
dosage
dosage
Mean
S.D.
(ppm-hr)
N
R.S.D.
Meas. /actual
(T = 85° F, RH = 60 % in GEL)
1.5 hr 0.5 ppm
0.75
1.5
5.0
20

4.08

0.5
1.07
1.35

2.65
4.03
0.60
8.45
24.27
15.68

2.33

2.33
1.0
1.0
0*
Ot
(T = 72°, RH =
0.5
(T = 70° F, RH
' I
$
t
(T = 72°, RH =
1.7
0.88
3.00
0.87
0.116
0.407
(horizontally
0.946
(vertically in
0.946
0.75 ppm-hr
0.75
1.5
0*
OT
40% in GEL)
2.04
= 50% in GEL)
t
* 1
1.65
1.72
2.52
0.2
0.1

3.0

4.42
1.25
$ 11.7
40% (appx.),
4.5
3.55
1.8
7.35
2.82
6.38
in flask)
2.21
flask)
2.21
0.18
0.25
0.13
—
—

0.15

0.37
0.31
0.56
Calibration
5.4
4,1
2.0
9.3
2.9
8.6

2.6

2.8
0.28
0.21
0.21
2.72
0.28
1.08

0.18

0.12
6
6
6
2
2

12

6
6
6
system)
2
3
3
3
5
3

6

6
1.0%
14%
5%
—
—

5%

8%
3%
5%

5%
5%
10%
29%
10%
13%

7%

4%
2.2
2.3
1.68
—
—

1.47

—
—


1.2
1.15
1.11
1.27
1.03
1.35

1.17

1.26

*0zone + NHt^NOs in chamber.
tSealed tubes kept approximately  24  hours  before  analysis.
fBeyond capability of  CEL  system to control/analyze - manually  set  to appx
 5-10  ppm.    Included to  illustrate  R.S.D.  of  identically exposed  tubes
 only.
                                     373

-------
Ratio  Palmes' Result/CEL vs.  Dose
              r-
              00

              d

                  5
                 o_
                                        m EO

                                        —i _r _r
                                        < LU UJ
                                        o o o
                                        • o <
                        <0



                        o
mdd
dnSOdX3
                                                                          
-------
exposures  made during  this  testing.    Data are  plotted  as  the  ratio  of
measured/actual ratio  versus net exposure.  The actual  concentrations used
are indicated  by  each  group of points.   It can  be  seen  that  the scatter
between  samples  exposed  simultaneously is  significant, and  that relative
accuracy  becomes  poorer  as net  exposure  decreases.   At  lower concentra-
tions,  blank  errors  become more  significant,  with  clearly  discernible
effects on accuracy  at  the  left side  of the  graph  (low net dosage).

     An  orientation  effect  test was  carried out  in  the flask  to  check if
the positioning of tubes  in the flask  contributed  any  bias in the analysis
(final  entries in Table  2).   The  means  of  six  tubes  in each  group were
found to be within 7 percent (0.2 ppm-hr).  Student's  t-test analysis gave
a borderline significant  difference.   Horizontal  position  of the tubes was
used in all other testing in the calibration system.   Examination of Figure
2 also  reveals a tendency  for  the analysis results  to  be higher  than the
actual exposure levels.

     Investigation  of  the   literature  related   to  the  method  of  nitrogen
dioxide analysis  used  by the Palmes tubes has  revealed considerable uncer-
tainty  in  the  efficiency  of uptake of  the  pollutant by  various absorbing
media that have been used in the  standard methods of analysis.   The stan-
dard  method  proposed  by EPA  (4)  uses  a  factor  of 0.82  in  its  arsenite
bubbler.   Other  workers  have  proposed values for  the efficiency factor
between 0.62 and  1.00  (5).   Palmes  himself recommended an efficiency factor
of 1.00 (1).   Our present work  seems  to indicate an efficiency between 0.72
and 1.00,  although the scatter in the data prevents assignment  of  a more
specific value.   If  a factor of unity,  rather  than  0.72,  were assumed for
the data of Figure 2,  the expectation  value for the  ratio  of Palmes/actual
dosage would be raised to 1.39.   There is also evidence  in  the literature
for a concentration  dependence  of  the  efficiency factor (5, 6, 7), although
our data do not show clear evidence of this in the range of  concentrations
employed.

     Figure 3  illustrates the  same  data set as Figure 2, but  with actual
observations rather  than  ratios on the vertical axis.  The two darker lines
are the  expectation lines  for efficiency  factors of 1.00 and 0.72.   The
lighter  line is the  actual regression line through the points shown.

     Because  of  their  size and  sensitivity range,  the Palmes  tubes were
ideally  suited to in-chamber testing  on subjects  of  the  biomedical proto-
cols.   Exposures  were made  during two scheduled sessions,  with monitors
both on  subjects  and in a stationary  location in the chamber.  One exposure
was to  0.5  ppm N02 gas,  while  the  second  was   to  ozone  with  NH^NOg aerosol
as a  control  exposure.   Table  3  illustrates   the  results  of  these tests.
Each tube was  on  a separate subject.

     In  order  to  gain  a larger statistical  basis  in  the chamber, a further
experiment was done.   A  total  of 24  tubes  were exposed.  Twelve, of these
were  on  a single  subject,  and  twelve  were  stationary.     The  following
exposures were made:
                                     375

-------
                                                                            10
Figure 3.  Palmes  results  versus dosage.




                                     376

-------
                 TABLE 3.  ON-SUBJECT EXPOSURES  IN CHAMBER
                            (T =  85° F, RH =  60%)
                                           Palmes
Exposure  Exposure  Calculated   	dosage (ppm-hr)	
  time      cone      dosage	Mean    S.D.   N    R. S.D.   Meas./actual
(on subjects)
  4 hr   .   0.5

(stationary)
5 hr 20 min 0.5
2.0
2.7
3.5     0.55   3     16%


4.5     02—
1.75
1.66
     (a)  6 samplers worn  on  upper  body,  around first button of shirt.

     (b)  6 samplers worn  on  lower  body,  clipped to belt.

     (c)  6 samplers  held in vertical  position in the  chamber,  clipped to
          an equipment  rack.

     (d)  6  samplers  in  horizontal position,  in  a  wicker chair  to  allow
          free flow of  gas past  the tubes.

     Subject  activity  during  all   exposures   consisted of  the  following
approximate proportions:

     30 minutes  -  treadmill,  3.5 mph
     30 minutes  (approximate) - miscellaneous  activity
       3 hours -  sitting in chair
      4 hours  - total exposure

     As  can  be  seen from  Table 4,  no  important  differences  were  seen
between  any groups  except  the  lower body  group,  which  is  slightly  low.
Since much of  the exposure  was  done  in  a sitting  position,  it is possible
that  occlusion  of  the  openings  by  contact  with  the subject's  clothing
caused  the  difference  observed.    Statistical  tests  of  the  four  group
results  indicate  that  the  mean from the  six tubes  clipped  to  the  lower
body is  significantly different  (p<0.05) than  the  means  of  the other three
groups.  No other statistically significant differences were found.


                                   SUMMARY

     It  was  found  that  dosages  as  low as  2.0  ppm-hr  could  be  reliably
measured with  the commercially available Palmes tube kit, although for best
precision, higher  levels  of  exposure are desirable.   The design  of   the
sampler  is presently  optimized  for  "industrial hygiene"  levels,  and could
be  modified   to  improve  sensitivity  in  the  lower  limits  of  exposure.
                                     377

-------
              TABLE 4.   EXPOSURE OF PALMES TUBES IN GEL CHAMBER
                            (T = 70° F, RH = 40%)
                                            Palmes
Exposure  Exposure  Calculated   	dosage (ppm-hr)
time
On subject,
4
On subject,
4
Stationary,
4
Stationary,
4
cone. dosage Mean S.D. N R. S.D. Meas. /actual
upper body
0.5 2.0 3.04 0.144 6 4.7%
lower body*
0.5 2.0 2.75 0.217 6 7.9%
vertical
0.5 2.0 2.98 0.133 6 4.5%
horizontal
0.5 2.0 3.06 0.186 6 6.1%

1.52

1.38

1.49

1.53

*Statistically  significant  difference  of mean  from means  of  other  three
 groups  (p<0.05).
Specific modifications  to  achieve this  goal might include:

     1.  Shorter  diffusion length to  increase capture of pollutant gas
         because  of  steeper  diffusion  gradient in  tube.    Use  of  a
         membrane or  multi-cavity diffuser could be  considered  if the
         diffusion permeability is  greater than the present design.

     2.  Larger  absorption cross-section  to  increase  the amount  of
         material for analysis.   In combination with a shorter  dif-
         fuser, a badge  type  configuration might be achieved.
                              ACKNOWLEDGMENTS

     The work presented here was  sponsored  by the U.S.  Environmental  Pro-
tection Agency.  The kit  used  for testing was loaned by  the U.S.  Bureau of
Mines.  George  Schnakenberg and Emory Chilton  of  the U.S. Bureau  of  Mines
and Roberta McMahon of MDA  Corporation provided  helpful discussions.
                                REFERENCES

1.   Palmes, E.D.   1976.   Personal sampler for nitrogen dixoide.   Am.  Ind.
     Hyg. Assoc. J. 37:570.

2.   McMahon, R.   1980.   New  technology for personal  sampling of N02  and
     NOX in  the workplace.   American Chemical  Society Exposition  Sympos-
     ium, 1980.  MDA-M-FS-4, MDA  Corporation, Glenview,  IL.
                                    378

-------
Saltzman,  B.E.   1954.  Colorimetric  determination of nitrogen  dioxide
in the atmosphere.   Anal.  Chem.  26:1949-1955.

U.S. Environmental  Protection Agency.   1976.   Guidelines for  develop-
ment of a  quality assurance program:   Vol. XVI - Method  for  the deter-
mination  of  nitrogen dioxide  in the atmosphere  (sodium arsenite  pro-
cedure).   EPA-650/4-74-005-p.

Crecilius, H.J., and W.  Forwerg.  1970.  Investigations  of  the  "Saltz-
man Factor."  Staub-Reinhalt.   Luft.  30:23-25.

Blacker, J.H.   1973.  Triethanolamine  for collecting nitrogen  dioxide
in the TLV range.   Am.  Ind. Hyg. Assoc. J. 34:390-395.

Huygens,  Ir.C.   1970.  Reaction of  nitrogen  dioxide with griess  type
reagents.  Anal. Chem.  42:407-409.
                                379
                                               *U.S. GOVERNMENT PRINTING OFFICE 1983 - 659-095/1942

-------

-------