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, 6
FIR MTG
2-REQD
17Vz x 93
dV4x3V4)FiR FRAME
w/1 Vi x 3VJ x
CNTR DIVIDER
CONSTRUCTION DETAILS
ALL DIMESIONS—INCHES
11V4 x 17V2
PLYWOOD
TOP & BOTTOM PANELS
1"DiA. 4 GROUPS OF
14-HOLES PER
SIDE PANEL
1 x 8 x 1/2
CEDAR SIDING
1 71/2 93
1/2 CELLULOSIC
SHEATHING
1 1/2 x 31/2 x 1 7y2
FIR CNTR DIVIDER
2-REQD
UNLINED
ALUMINUM SIDING
111/2 x 94
3/16 PLYWOOD
SIDE PANELS
i/2(Typ)
Figure 6. Urea-formaldehyde foam insulation (TJFFI) panel construction.
38
-------
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
CC
UJ
CO
UJ
UJ
± Q.
< O)
CO
UJ
CQ
o
o
o
CC
UJ
N
4-1
(N
CO
o
60
•rl
1*1
104
-------
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
-------
------- |