EPA/600/9-85/029
Proceedings: Fifth Annual National
Symposium on Recent Advances in
Pollutant Monitoring of Ambient
Air and Stationary Sources
Radison Plaza Raleigh Hotel
Raleigh, NC
May 14-16, 1985
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Office of Research and Development
Research Triangle Park, NC 27711
-------
NOTICE
This document has been 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.
-------
FOREWORD
i f
Measurement and monitoring research efforts are designed to anticipate
potential environmental problems, to support regulatory actions by develop-
ing an 1n-depth understanding of the nature and processes that Impact health
and the ecology, to provide innovative means of monitoring compliance
with regulations and to evaluate the effectiveness of health and environ-
mental protection efforts through the monitoring of long-term trends.
The Environmental Monitoring Systems Laboratory, Research Triangle Park,
North Carolina, has the responsibility for: assessment of environmental
monitoring technology and systems; Implementation of agency-wide quality
assurance programs for air pollution measurement systems; and supplying
technical support to other groups in the Agency Including the Office of
Air, Noise and Radiation, the Office of Pesticides and Toxic Substances
and the Office of Solid Waste and Emergency Response.
Thomas R. Hauser, Ph.D.
Director
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina
m
-------
-------
INTRODUCTION
TABLE OF CONTENTS
Session I - Acid Deposition
Application of Similarity Methods to Deduce Dry Deposition Covariances
Using Imperfectly Responding Sensors
An Eddy Accumulator for Dry Deposition Measurements
A Prototype Concentration Accumulation Apparatus for Determining
Time - Averaged Dry Deposition of Gaseous Pollutants
Simple Meteorological and Chemical Filterpack Monitoring System
for Estimating Dry Deposition of Gaseous Pollutants
Evaluation of an Automated Tungstic Acid Technique for Nitric Acid
and Ammonia
Application of the Tungsten Oxide Denuder Tube Technique to the
Measurement of Nitric Acid in the Rural Troposphere and
Comparison with the Nylon Filter Method
The Use of Lichens to Measure the Transport of Airborne Heavy
Metal Pollutants
Measurement of HNOa, S02, NH,3 and Particulate Nitrate with an
Annular Denuder System
A New Analyzer for the Continuous and Unattended Measurement
of Low Levels of Sulfur Dioxide in Ambient Air
Session II - General
7
16
23
30
39
51
55
72
Alkalinity Measurements in Evaluating Hi-Vol Filter Performance
A Procedure for the Isolation and Concentration of Milligram
Quantities of Carbon Monoxide from Ambient Air
Analysis of the Effluents from Polyure^thane Foam Insulation
Quantitative Analysis of PPB Concentrations of Hydrogen Sulfide
in Air Using a Gold Film Sensor
Session III - Source Emission Monitoring Using EPA
Reference Method 25
79
81
88
92
The State of Delaware Experience with EPA Reference Method 25
Audit Samples
Utilization of EPA Method 25 for the Measurement of Volatile
Organic Carbon (VOC) Emissions from Combustion and Steam Sources
99
104
-------
Session IV - Analysis of Chemical Complex Data
An Overview of the Development of Methods of Chemometric Methods
Cluster Analysis Applied to Individual Atmospheric Particles
Evaluation of Emission Data from Hazardous Waste Incineration
Tests Using Pattern Recognition Techniques
Distributed Chemical Alarm Systems for Plant Security Based on
Ion Mobility Spectrometry (IMS)
Session V - Personal Monitoring
127
130
134
136
Personal Monitoring: Overview
Indoor Aerosol Impactor
Analysis of Nitrite in N02 Diffusion Tubes Using Ion Chromatography
Evaluation of COED-1 Portable Carbon Monoxide Personal
Exposure Monitors
Monitoring of Benzene in Ambient Air with Organic Vapor Badges
A Passive Dosimetry Method for Determining Hydrazine in Air
Session VI - Organic Pollutants
154
158
166
171
178
183
GC/FTIR Studies of Volatile Organic Compounds Collected on
Tenax Cartridges
New Redox Reaction Processes for Analysis of Atmospheric
Pollutants
Application of Multidimensional Gas Chromatography to the
Analysis of Particulate Samples
A New Microprocessor-Based Instrument for Thermal Desorption
of Adsorbent Traps
Comparison of Sample Collection Techniques for Volatile Organic
Compounds in Ambient Air
Benzene and Toluene Concentrations in Lake Charles, Louisiana
Ambient Air
Determination of 2-Diethyl aminoethanol in Air
Analysis of Semi-volatile Organic Compounds Using Supercritical
Fluid Methodologies
190
204
209
218
226
237
241
247
-------
Session VII - Source Monitoring
Validation of an Emission Method for Inorganic Arsenic
Studies on the Dynamic Impinger Sampling System
Development of a Compact Total Reduced Sulfur (TRS) Continuous
Emission Monitoring System
Measuring Volatilization Flux of Pesticides from Treated Surfaces
253
258
261
267
vii
-------
-------
INTRODUCTION
The fifth Annual National Symposium on Recent Advances in the
Measurement of Air Pollutants was held May 14-16, 1985 in Raleigh,
Morth Carolina. The symposia are sponsored by the United States
Environmental Protection Agency's (EPA's) Environmental Monitoring
Systems Laboratory in Research Triangle Park, Morth Carolina.
This fifth annual symposium was part of a continuing effort to
explore recent advances in pollutant monitoring of ambient air and
stationary sources. It served as a forum for exchange of ideas and
scientific information. In response to the Agency regulatory needs,
this symposium focused on acid deposition, personal exposure and
toxic substances. This publication is intended to assist those
researchers interested in furthering the science of air monitoring.
Because several presenters listed in the Symposium announcement
spoke extemporaneously and did not prepare a paper, the number of papers
in these proceedings are fewer than the actual numbers o.f presentations.
-------
APPLICATION OF SIMILARITY METHODS TO DEDUCE
DRY DEPOSITION COVARIANCES
USING IMPERFECTLY RESPONDING SENSORS
by
B. B. Hicks and R. T. McMillen
Atmospheric Turbulence and Diffusion Division
Air Resources Laboratory, ERL, NOAA
Post Office Box E, Oak Ridge, TN 37831
Introduction
Methods for measuring dry deposition rates are relatively few. Direct measurement of
the accunulation of material at the surface is possible mainly for particulate species to
surfaces like snow, or to surfaces that can be washed without damaging them (e.g. some
foliage). Use of artificial surfaces or collection vessels is not widely favored because
the collection characteristics of such devices are rarely similar to those of the natural
surfaces they are meant to simulate. Most modem studies make use of micro meteorological
methods, which, if properly applied, permit precise evaluation of fluxes through planes
elevated above the surface. These measured fluxes are equal to surface fluxes if
stringent site quality and temporal stationarity criteria are satisfied. However, an
overwhelming difficulty is that such methods require sensors of precision and/or time
resolution that are presently available only for very few chemical species.
Hicks (1979) addresses the matter of pollutant sensing requirements for micro-
meteorological applications. Several distinctly different factors arise to complicate the
way in which micrometeorological methods can be applied to measure dry deposition rates.
First, gradient methods require the use of sensors capable of resolving differences with
1% precision, if not better. Even if suitable chemical methods exist, the gradient method
is accurate only over surfaces with relatively short vegetation.
Eddy correlation techniques do not require an assumption regarding eddy diffusivity,
and hence tend to be preferred. However the requirement for fast response sensing limits
the application to very few chemical species. In general, ± 10% accuracy evaluations of
eddy flaxes require 1-second response time in daytime conditions, and sometimes as short
as 0.1 second at night. Most transfer usually takes place in daytime.
The similarity principles addressed here permit fluxes of some trace specie's to be
inferred from knowledge of another if appropriate gradient or turbulence intensity data
are available to serve as a linkage. The constraint imposed by the surface heat energy
balance involving (primarily) net radiation, sensible heat, latent heat of evaporation,
and heat storage provides an opportunity to bypass the need for a direct flux measurement
using micro meteorology. The method of present interest uses net radiation as a basis for
calculations, and employs similarity principles first to derive the sensible heat flux and
then to infer a dry deposition flux from knowledge of pollutant turbulence characteristics.
In practice, net radiation is easily measured, using straight-forward sensors available
from commercial sources.
Theoretical Considerations.
The so-called "similarity/heat-budget" method is based on apportionment of net
radiation Rn between sensible and latent heats, and on the measurement of turbulence
intensities of temperature and pollution concentration using sensors with matched
frequency response.
-------
The surface heat energy budget can be written as
H + LE = Rn - S
(1)
where H is the sensible heat flux, LE is the latent heat flux, and S is the rate of heat
storage beneath the level of measurement. The Bowen ratio, 3, is defined as H/LE,
so that
H(l + 1/8)
(2)
Now, suppose that a measure of pollutant concentration standard deviation ac is measured
in some selected frequency band, and that similarly constrained (i.e. defined by the same
frequencies) measures of temperature and humidity standard deviations are also obtained:
a and o_ respectively. If the flux of pollutant is F, then similarity principles dictate
that
F/H
and H/LE = 3 = Cp^/La^
From here, trivial mathematical manipulation yields that
F - ac(Rn -
Laq).
(3)
(4)
(5)
Equation (5) provides a means for evaluating a flux F from imperfect (but matched)
standard deviation data 0C, cr and aq, using measurements of net radiation as a basis.
Over many agricultural surfaces, the heat storage term S is less than 15% of Rn, and in
most circumstances can be taken to be about 10% of Rn, hence
vd = 0.9(ac/C)-Rn/(cp0T + Laq),
where the deposition velocity vd is defined as F/C.
(6)
A somewhat simpler form of (6) might be of special interest in circumstances in which
the Bowen ratio 3 can be computed, from external information, such as by use of a
combination formula derived from agricultural or agrometeorological considerations. In
this case, the need for specific humidity data to yield crq can be avoided, since (6) can
be rewritten as
vd = 0.9(Rn/C)-(ac/CpaT)/(l + 1/3).
(7)
From inspection of Equation (7), it is evident that the various sources of error
combine such that the relevant proportional errors can be compounded as in
-1,
E(vd)/vd = E(ac)/ac 4- (1+3) "E(3)/3,
(8)
where E(x) is the error on x. The Bowen ratio 3 is normally positive in daytime, but
typically less than unity over vegetation. Over sparse vegetation, or over bare soil, 3
can attain large positive values, such that the final term in (8) would then be
insignificant. In general, however, the final term of (8) cannot be disregarded. Major
interest focuses, however, on the term E(ac)/ac, since this is the factor which influences
Instrument design and application.
Inspection of Equation (7) indicates that deposition velocities can be estimated using
sensors with deficient response provided quantifies such as
-------
Branch Watershed ffpld site during May, 1983. The data were obtained during a
multUaboratory "core" research experiment conducted in collaboration with Argonne
National Laboratory and Pennsylvania State University. The tabulation Includes
information on temperature, humidity, atmospheric sulfur concentration, and particle
by smoothing records obtained with fast-response sensors to produce 1, 3, 10, and
30-second averages prior to computation. The results of Table 1 are typical of daytime,
convective conditions. The nocturnal case has yet to be addressed In as much detail.
Figure 1 Illustrates the consequences of high-frequency noise on the general utility
of standard deviations as a basis for calculations, the values plotted are geometric mean
normalized standard deviations for temperature (humidity values are Indistinguishable),
sulfur dioxide and small particles (in the two size bands used here, 0.4-0.7 um and 0.7
— 1.1 Mm diameter). As high—frequency noise is rejected by smoothing with increasing time
constant, the effect on particle standard deviations Is seen to be greatest. At the other
extreme, the effect on temperature (and humidity) standard deviations Is the least among
the presently observed quantities. This Is seen as a direct consequence of the ease with
which temperature and humidity records can be obtained using modern sensors. It should be
noted that for particles the problem Is exacerbated by effects associated with poor
counting statistics. The roll-off associated with the sulfur dioxide data Is less than
that associated with the measurement of particle concentration, but greater than that for
the meteorological variables. In this context, the use of covarLances (with w) instead of
standard deviations as Indicators of total turbulent transport seems especially
attractive, since the influence of high frequency noise Is then minimized.
Figure 2 Illustrates the ability of the similarity methods advocated here to provide
accurate estimates of vertical fluxes. For purposes of Illustration, Figure 2 plots esti-
mated latent heat fluxes, LE, derived from using the most smoothed data (30-second time
constant) used in the generation of Table 1, against the actual values measured by eddy
correlation.
It is obvious that the use of variance data as a basis for computations using the
similarity /heat-budget approach leads to potential difficulties concerning the direction
of the fhix, precisely as In the case of direct analysis of variance in conventional
meteorological applications. An alternative method Is presently being evaluated, in which
the standard deviations in Equations (3)—(7) are replaced by covarLances computed using
some other, well-measured meteorological variable. In this way, random noise should be
rejected, since by definition it will be uncorrelated with any other signal. Furthermore,
information on the sign of the vertical flux will be retained.
Acknowledgement
This work was supported by the U.S. Environmental Protection Agency as a contribution to
the National Acid Precipitation Assessment Program.
Reference
Hicks, B. B., 1979: Some Micro meteorological Methods for Measuring Dry Deposition Rates,
AICHE Symposium Series, 75, Number 188, 187 - 190.
-------
TABLE 1. Examples of data generated by a test of the theory of similarity methods for
deducing pollutant deposition from radiation data and heat budget considerations, using
slow-responding pollution sensors. All data are for 30-minute runs, above a forest canopy.
Data are quoted for 1, 3, 10, and. 30 second response time sensors. Units for global (Rg)
net radiation (R^, sensible heat flux (H) and latent heat flux (LE) are W/m~z. All other
quantities are normalized to unity for "perfect" sensors. Fluxes Fl, F2, and F3 are of
S02, small particles (0.4 -0.7 yta diameter), and larger particles (0.7 - 1.1 urn),
respectively. The remaining quantities are standard deviations, also normalized.
1500 -
1 sec
3 sec
10 sec
30 sec
1530 -
1 sec
3 sec
10 sec
30 sec
1600 -
1 sec
3 sec
10 sec
30 sec
1630.-
1 sec
3 sec
10 sec
30 sec
1700 -
1 sec
3 sec
10 sec
30 sec
1730 -
1 sec
3 sec
10 sec
30 sec
H
1530, 2 May
224
220
205
169
71
1600, 2 May
245
242
231
197
151
1630, 2 May
166
162
147
108
62
1700, 2 May
136
132
121
95
51
1730, 2 May
109
108
99
79
47
1800, 2 May
125
.123
114
98
55
LE
1983,
279
274
259
228
149
1983,
205
202
192
160
119
1983
124
120
108
78
44
1983
92
88
80
60
31
1983
82
80
74
59
39
1983
49
47
43
36
21
Fl
Rg =
1.00
1.00
.99
.97
.90
Rg =
1700
.99
.97
.92
.77
Rg =
1700
r.oo
1.01
.88
.33
Rg =
1700
.91
.71
.63
.15
Rg =
iToo
.99
.95
.73
.29
Rg =
1700
.99
.96
.82
.57
F2
775,
1.00
1.00
1.00
.94
.84
640,
1.00
1.00
.83
.83
.83
576,
1.00
.98
.84
.66
.79
473,
1.00
.89
1.20
.21
1.01
352,
1.00
.74
1.02
.38
-.32
243,
1.00
.90
.96
.19
.30
F3
Rn
1.
1.
1.
•
•
Rn
1.
1.
1.
•
•
Rn
1.
•
•
*
•
Rn
1.
•
•
•
•
Rn
1.
•
•
•
1.
*n
1.
1.
•
1.
1.
=
00
00
00
92
80
=
00
03
27
92
89
=
00
92
60
27
21
=
00
99
87
63
48
=
00
83
65
99
38
=
00
10
87
37
10
ST
540
1.00
.99
.98
.94
.85
465
1.00
.98
.96
.91
.84
385
1.00
.98
.96
.91
.82
302
1.00
.98
.96
.90
.81
196
1.00
.98
.96
.91
.67
103
1.00
.98
.96
.91
.82
Sq
1.00
.99
.98
.94
.86
1.00
.99
.96
.92
.84
1.00
.99
.96
.91
.82
1.00
.98
.96
.90
.81
1.00
.99
.96
.90
.83
1.00
.96
.94
.89
.79
SI
1.00
1.00
1.00
.99
.96
1.00
.99
.97
.95
.90
1.00
.94
.88
.78
.70
1.00
.94
.89
.81
.71
1.00
.95
.89
.78
.67
1.00
.93
.87
.75
.62
S2
1.00
.99
.99
.97
.95
1.00
.94
.85
.69
.58
1.00
.94
.84
.66
.51
1.00
.89
.84
.69
.52
1.00
.94
.85
.68
.52
1.00
.94
.84
.67
.50
S3
1.00
.98
.95
.91
.87
1.00
.94
.85
.72
. 59
1.00
.94
.84
.69
.56
1.00
.95
.85
.72
.56
1.00
.94
.83
.69
.54
1.00
.94
.85
.69
.53
-------
1.0
0.9
0.8
0.7
0.6
0.5
PARTICLES
1.0
Figure 1. The effect of
sampling response time
on standard deviation
associated with tempera-
ture (dots), sulfur
dioxide (circles), and
particle concentration
(crosses) records.
Data are normalized to
values obtained from
"raw" data obtained
from fast-response
sensors.
100
TIME CONSTANT (SEC)
30-SEC LE
200
E
3:
c
UJ
103
Figure 2. A comparison of
latent heat fluxes
derived by application
of Equation (5), with
direct-measure ments
made by eddy correla-
tion. The total heat
storage term Is assumed
to be a constant 10%
of the net radiation.
100
MEASURED FLUX (w
201
f\
-------
AN EDDY ACCUMULATOR FOR DRY DEPOSITION MEASUREMENTS
R. E. Speer, T. G. Ellestad, and J. L. Durham
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
INTRODUCTION
For applications requiring measurements of the turbulent fluxes of
trace gases to the earth's surface, the so-called eddy accumulation
variation on conventional eddy correlation continues to offer promise,
despite acknowledged technical difficulties that have so far proved
prohibitive. In essence, the method bypasses the usual requirement for a
rapidly-responding sensor of the atmospheric concentration of the trace
gas of interest. Instead, a fast-response pumping system is used to
sample air at a rate proportional to the vertical velocity. Two samples
are then accumulated, one corresponding to updrafts and the other to
downdrafts. The differences in average concentrations accumulated in
these two independent samples is directly related to the conventional
eddy flux. The method has been described in detail elsewhere (e.g. Hales
and Horst, 1974; Desjardins, 1977; Hicks and McMillan, 1985). If applied
correctly, the eddy accumulation technique would have the same limitations
as conventional eddy correlation, but would eliminate the need for two
signals to be multiplied together.
Eddy accumulation has had many strong advocates, especially in the
context of measuring dry deposition, but success has been so elusive that
there are few references in the open literature. Nevertheless, a review
of existing and potential methods for measuring dry deposition by Hicks
et al. (1980) identified eddy accumulation as a technique of special
promise for measuring dry deposition. A computer simulation of the method
(Hicks and McMillen, 1985) led to the identification of several major
problems, each technical in nature but each presently limiting the success
of the method in studies of the rates of exchange of rapidly-transferring
properties. In every case, the difficulties scale inversely with the
-------
deposition velocity (Vd, defined as the ratio of flux density to atmospheric
concentration) of the substance being studied. Thus, it should be easier
to develop an eddy accumulation system for studies of the transfer of
rapidly-deposited substances such as nitric acid vapor (Vd typically 2-5
cm/s) than it would be for a slowly-depositing substance like ozone (Vd
rarely above 1 cm/s).
THEORY
Eddy accumulation separates the flux relation, F = w'C, into an up
and a down portion,
F = (^C)u - (wTC)d.
Ideally, the sample flow rate f = Aw1, and
F = I/A [(?C)U - (fC)d]
F = AM/AT
where AM = difference in collected sample masses, Mu and Md and T is the
total sample time. In reality f = Aw' + B and is different for up and down:
F = l/Au[(f-Bu)C]u - l/Ad[(f-Bd)C]d
F - (MU/AU)(1/T - BU/VU) - (Md/Ad)(l/T - Bd/Vd)
where Vu and Vd are sample volumes and C = M/V.
SYSTEM DESCRIPTION
A general block diagram of a prototype eddy accumulation field system
that is under development is shown in Figure 1. In its initial configuration,
the vertical component of the wind velocity is sensed with a propeller
anemometer whose output is a DC voltage proportional to the wind velocity
component parallel to the propeller axis. The mean component, w, of this
signal is removed by a high-pass filter with a time constant of 100 s,
leaving only the variable component, w1. A polarity detector uses the
signal Aw1 to control a solenoid valve with a switching time of 10 ms to
select the appropriate collector for sampling.
The absolute value, |Aw'| of the signal is used to control the sample
flow rate, using three linked devices: an MKS Model 250A flow controller,
an MKS Model 251-20000SP flow control valve, and a Thermo-Systems Model
2234 mass flow sensor, with a model 67 signal linearizer, as a flowmeter
unit. The flowmeter is rated at 100 L min'l with a response time of
-------
1 ms, the control valve at 20 L min"1 with a response time of 10 ms. The
flow controller follows the varying |Aw'| signal and attempts to maintain
a sample flow rate proportional to the signal and thus proportional to
the vertical wind velocity by continuously reading the flow rate and
adjusting the control valve for zero error signal (Figure 2).
The valves, flow sensor and related hardware are mounted on an
instrument tower during field operation and are connected by signal cables
to a remote electronics package, which is kept in a controlled environment
to minimize temperature drifts. The flow rate signal is integrated to
determine the actual volume of air sampled by each sampler, permitting
corrections to be made for unequal sample volumes.
PERFORMANCE TESTS
A DC linearity bench test was performed. A range of voltages from a
precision DC voltage source equivalent to w' from 0.025 to 2.5 m s"^ was
supplied to the controller (point w, Figure 1) but with the high-pass
filter's capacitor short-circuited; the resulting flow rate signal was
measured (at point y, Figure 1). The average flow rate deviation from
linearity was equivalent to an input of 0.002 m s"1 of w. At the vertical
velocity threshold of 0.025 m s~l, this would imply a relative error of 8%
The frequency response of the combined controller electronics and
pneumatic system was, determined in laboratory and field tests. These
tests consisted of comparing the flow rate signal (point y, Figure 1)
with the input signal (point w, Fig. 1) to the flow controller. The
response of propeller anemometers as sensors of vertical wind eddies has
been dealt with previously (Hicks, 1972; Gill, 1975).
The high-frequency performance of the system operated in field
conditions was tested using spectral analysis of recorded flow rate
signals (Otnes and Enochson, 1972). Typical sections of records (102.6
duration) were digitized at 20 Hz and analyzed using a fast Fourier
transform program. Figure 3 illustrates the results, after application
of running-mean smoothing to the raw products.
-------
The frequencies of atmospheric turbulence that are associated with
the daytime vertical fluxes of materials lie mainly between normalized
frequencies, f = nz/u, of about 0.001 to about 2.0, where n = cyclic
frequency (Hz), z = height and u = mean horizontal wind velocity (Kaimal,
1973; Kaimal, et al., 1973). It appears that the eddy accumulator's linearity
and frequency response should be reasonably adequate under most atmospheric
conditions according to Figure 3.
FIELD TESTS
Table 1 gives some of the species that we have been measuring or plan to
measure with the eddy accumulator. At the present time we are concentrating
on water vapor since it has a high flux rate, can be analyzed with high
precision and can be used as a check on the eddy accumulator's performance
by comparing with flux values obtained by other independent means.
TABLE 1. SAMPLE ANALYSIS
Species
H20
S02, N02
Sulfate, nitrate
Collection Medium
Molecular sieve
TEA impregnated filter
Nylon filter
Analysis Method
Wei ghing
Ion Chrom.
Ion Chrom.
The results of water vapor flux measurements during the first field
test are given in Speer, et. al., 1985. The results did not correlate
very well with eddy correlation data from the same site and indicate that
there are problems that remain to be overcome.
Samples for S0£ have been collected during field tests but improvements
in the precision of the ion chromatographic analysis, particularly in
adequately resolving the small mass differences between the up and down
filter samples, need to be made before these are analyzed.
10
-------
FUTURE PLANS
Some changes to be made in the near future include a computer
controlled system and a sonic anemometer. The computer system's software
will replace functions that are being performed by analog circuitry in
the present system, including control of flow rate. The sonic anemometer
will improve on the propeller anemometer's slow response.
CONCLUSION
The development of the eddy accumulator should be considered at the
present time as a work in progress. Significant steps have been made
toward the goal of a working system, mainly in developing a fast response
sampling apparatus, but other obstacles need to be overcome. Effort will
be concentrated on identifying the sources of errors and on making the
planned improvements mentioned previously in this report.
REFERENCES
Desjardtns, R. L., 1977; Description and Evaluation of a Sensible Heat
Flux Detector, Boundary-Layer Meteorol., 11, 147-154.
Gill, G. C., 1975; Development and Use of the Gill UVW Anemometer, Boundary-
Layer Meteor., 8., pp. 475-495.
Hales, J. M., and T. W. Horst, 1974; A Flux Meter for Direct Field
Measurement of Deposition and Resuspension Rates, In: Pacific Northwest
Laboratory Annual Report for 1973, USAEC/DBER, Part 3, 176-178.
Hicks, B. B., 1972; Propeller Anemometers as Sensors of Atmospheric
Turbulence, Boundary-Layer Meteor., 3_, pp. 214-228.
Hicks, B. B. and R. T. McMillen, 1985; A simulation of the eddy accumulator
method for measuring pollutant fluxes, J. Climat. and Appl. Meteor., 23,
pp. 637-643.
Hicks, B. B., M. L. Wesely, and J. L. Durham, 1980; Critique of Methods
to Measure Dry Deposition: Workshop Summary, EPA Report EPA-600/9-80-
050, NTIS Publication PB81-126443.,
Kaimal, J. C., 1973; Turbulence Spectra, Length Scales and Structure
Parameters in the Stable Surface Layer, Boundary-Layer Meteor., £, pp.
289-309.
11
-------
Kaimal, J. C., J. C. Wyngaard, Y. Izumi, and 0. R. Cote, 1972; Spectral
Characteristics of Surface-Layer Turbulence, Quart. J. of the Royal
Meteor. Soc., 98_, pp. 563-589.
Otnes, R. K. and Enochson, 1972; Digital Time Series Analysis. (Wiley-
Interscience, New York).
Speer, R. E., K. A. Peterson, T. G. Ellestad, and J. L. Durham, 1985;
Test of a Prototype Eddy Accumulator for Measuring Atmospheric Vertical
Fluxes of Water Vapor and Particulate Sulfate, J_. Geophys. Res., 90, pp,
2119-2122.
12
-------
"DOWN"
COLLECTOR
Fi gure 1 .
Block diagram of the eddy accumulator system. Signals at points
w and y were the input and output signals, respectively, for
the performance tests.
13
-------
0.6
0.4
0.2
0
•0.2
-0.4
•0.6
o
5 ON
o
3|OFF
tu
cc
FLOW RATE
10
TIME, sec
3 -H
Uj"
«s
1 i
15
Figure 2. Example of eddy accumulator signals while in operation. At
the top the flow rate signal is superimposed on the anemometer
signal after mean removal. At the bottom is the signal to the
switching valve's relay.
14
-------
1.1
1.0
0.9
0.8
LU
o 0.7
LII
DC
111
0.6
u
LU"
Q
^0.5
_j
0.
< 0.4
0.3
0.2
0.1
0.01
AMPLITUDE
COHERENCE
PHASE LAG
A
LU
CO
<
o_
7T/2
0.1
1.0
10
FREQUENCY (Hz)
Figure 3. Frequency response of the eddy accumulator's flow system from
spectral analysis of flow rate vs. input signal while operating
under field conditions.
15
-------
A PROTOTYPE CONCENTRATION ACCUMULATION APPARATUS
FOR DETERMINING TIME-AVERAGED DRY DEPOSITION OF GASEOUS POLLUTANTS
B. B. Hicks, R. P. Hosker, Jr., and Ma, Guo-Jun*
Atmospheric Turbulence and Diffusion Division
NOAA/ERL/Air Resources Laboratory
Oak Ridge, TN 37831
INTRODUCTION.
Dry deposition is known to be an important component in the overall deposition of trace
gases and aerosols from the atmosphere, but no instrumentation now exists by which dry
deposition fluxes to natural surfaces can be measured easily or monitored directly.
Present practice generally requires either direct measurement of turbulent" fluxes using
complex and expensive fast-response instruments, or makes estimates of the fluxes from the
product of measured, time-varying airborne pollutant concentrations and appropriate dry
deposition velocities calculated from observed meteorological and surface conditions
(Hicks et al., 1985). The sophisticated instrumentation needed for both approaches is
generally not suitable for widespread, routine monitoring use because of its cost and
technically complex support requirements. Because dry deposition flux estimates are
needed at many locations, less expensive techniques suitable for monitoring use are pre-
ferable, if reasonable accuracy can be obtained.
A monitoring program to estimate weekly-averaged dry deposition fluxes of selected air
pollutants is described in another paper (Hosker and Womack, this symposium). That
program follows the second approach mentioned above: dry deposition velocities for the
various chemical species are estimated on the basis of recorded micro meteorological obser-
vations, and are then multiplied by the average concentrations of those species as deter-
mined from constant mass flow filterpack samplers. The method, while attractive because
of its relative simplicity, low cost, and reliability, suffers from some uncertainties.
(a) Concentrations of several critical chemical species cannot be measured with con-
fidence at all times using filterpack collection methods. In particular, techniques must
yet be perfected for determining trace concentrations of nitrogen compounds (especially
nitrogen oxides, but also ammonia and nitric acid vapor) and ozone.
(b) Understanding of the temporal and spatial variability of the dry deposition veloci-
ties of different chemical species is still limited. For some species (in particular, SO2
and 03), the concerted research effort by numerous investigators over the last few years
has succeeded in producing deposition velocity parameterizations that account for changes
in meteorological variables as well as variations in surface biology, within fairly
limited bounds.
However, knowledge of the deposition of even the better understood trace gases is still
lacking for many natural surfaces, and is unavailable for the deposition of other trace
species to nearly all vegetated surfaces. Hicks et al. (1985) describe the relevant para-
meterizations and their limitations.
(c) At any given location, both atmospheric trace species concentrations and the asso-
ciated dry deposition velocities are likely to fluctuate widely with time. This raises
Atmospheric Environment Monitoring and Research Centre, Ministry of Water Resources and
Electric Power, Nanjing, Peoples' Republic of
16
-------
some question about the accuracy of average flux estimates calculated from the product of
average airborne concentrations and average deposition velocities. In principle, at
least, the time-average ought to be computed from the product of the fluctuating con-
centration and deposition velocity.
The method described here Is intended to sidestep the errors arising from the .simple
filterpack monitoring program 's approximation of average fluxes by the product of average
concentration and average deposition velocity. The new instrumentation will be deployed
at selected locations to provide direct comparisons with the average fluxes inferred from
the product of averages, with those calculated from hourly integration of deposition
fluxes evaluated from higher time resolution data, and with fluxes determined from occa-
sional eddy correlation measurements. The new technique is designed to retain the con-
venience of a filterpack chemical sampling system without the need for expensive and
complex chemical monitors (and their required elaborate support systems), or for extensive
and costly laboratory analysis. The method is somewhat similar to "eddy accumulation", as
discussed by Hicks et al. (1980) and by Hicks and McMillen (1984). The new approach
suggested here is therefore called "concentration accumulation".
THEORY OF OPERATION.
In the prototype dry deposition monitoring program described elsewhere, suitable micro-
meteorologLcal instruments are deployed to provide the data necessary to deduce chemical
species-dependent, time-varying values of dry deposition velocities at each site. In par-
ticular, wind speed, wind direction variability, air temperature, humidity, insolation,
rainfall, and surface wetness are continuously monitored, and 15-min averages of these
quantities are recorded. Supporting information on local surface and vegetation con-
ditions is derived from weekly reports by site operators; these include observations of
vegetative species, cqndition, height, leaf area index, and water availability. The for-
mulations used to estimate dry deposition velocities for SO2 and HN03 vapor from the
meteorological and vegetation data are based on the resistance analog for mass transfer,
as described by Hicks et al. (1985). The "concentration accumulation" method makes use of
the availability of this relatively high time resolution information on dry deposition
velocity as follows.
Consider a sampling system which aspirates an air/trace-pollutant mixture. The total mass
M passing through the system in some time interval T is
1
•L
M = / (pa + C) V A dt
'0
(1)
where pa is the local air density, C is the airborne pollutant concentration, and VA is
the volume flow rate through the sampler tube. Similarly, the mass of pollutant Mp
passing through the system during that same time interval is
T
Mp = I C V A dt
-'o
(2)
Assume all (or at least a well-known fraction) of the pollutant is captured on a "filter.
The key step is to adjust the mass flow rate through the sampler so that it is directly
proportional to the pollutant's dry deposition velocity V , as calculated from the ambient
meteorological and surface conditions. That is, set
17
-------
( p_ + C ) V A = K V
D
(3)
Then, for low ambient concentrations,
« .
VD c dt.
(4)
The average dry deposition flux F Is
1? — 7^
v°cdt
(5)
so there is a direct relation between the dry deposition flux and the amount of
pollutant captured by the fUterpack sampler:
M H
P pa
¥.
(6)
The total mass of air and pollutant passing through the system is just
M
"/o
(p + C) V A dt
3.
-r
Jo
K VD dt,
or
M = K T VD,
(7)
(8)
where VD is the average dry deposition velocity over the interval T. Hence the average
dry deposition flux can be determined from Equations (6) and (8) as
F* PaVD M
r - jj M. ,
M p
(9)
The average air density and average dry deposition velocity can be evaluated from contin-
uously monitored meteorological and surface condition data, and the total mass passing
through the system can be measured. The dry deposition flux is then simply proportional
to the mass of pollutant collected on the filter. The sampler proportionality constant K
can be evaluated from Equation (8), if desired.
SYSTEM CONFIGURATION.
A simple apparatus has been developed to use this concentration accumulation concept. The
prototype system was designed to be low in cost, reliable, and compatible with commonly
available meteorological equipment. A block diagram of the system is shown in Figure 1.
Meteorological data from an array of instruments are scaled by special circuitry and
sampled by an independently scanning eight channel analog-to-digital (A to D) converter
board. An inexpensive "home" computer acquires the digital data from the A to D cir-
cuitry, converts the values into engineering units, and calculates time averages of each
quantity, along with certain important statistical quantities. The averaging period is
selectable, but is typically 15 minutes. The average values are then used in an algorithm
based on the resistance analog to mass transfer to calculate the local dry deposition
velocity for a particular air pollutant — typically HN03 or S02. The meteorological
18
-------
data and the estimated dry deposition velocity are recorded on cassette tape, together
with the date and time. A digital-to-analog (D to A) converter provides an output signal
proportional to the estimated VD. This signal is used to modulate the flow rate through
an electronic mass flow controller in a filterpack sampling system. The flow rate is also
monitored and recorded by the computer so that the total mass flow through the system is
known. At the end of the averaging period, a new dry deposition velocity is calculated on
the basis of the average meteorological and surface conditions during that period, and the
flow rate through the filterpack system is readjusted. Meteorological data collection and
concentration sampling continue for one week, just as in the constant flow rate filterpack
monitoring system described separately. An inexpensive DC to AC power inverter is
operated by a continuously charged 12V battery to protect the computer against power line
interruptions. The filterpack unit is the same as that described (Hosker and Wo mack, this
symposium) for the constant flow system. Figure 2 Is a photograph of the prototype
system.
PRESENT STATUS.
The prototype data acquisition, computational, and controller system has been completed
and tested in the laboratory, and is ready for field use. Figure 3 shows typical dry
deposition velocity estimates for HN03 over four days last spring; these are taken from
strip chart recordings of the D to A converter output that would be used in practice to
modulate the mass flow rate through the filterpack sampler. The diurnal cycle is clear.
The rather different appearance on May 21 is due to the passage of a cold front, with
strong winds and turbulence in the early evening.
It is planned to install the concentration accumulation system at the Oak Ridge "core"
research site this summer, so that comparisons can be made with flux estimates produced by
other systems. We expect to construct two more of these new systems for installation at
other "core" sites. Present plans are to collect data at three sites for six to twelve
months, and compare the results to direct eddy flux measurements, to continuous monitoring
inferential results, and to standard filterpack monitoring results. The outcome will be
presented in about a year.
REFERENCES.
Hicks, B.B., M.L. Wesely, and J.L.. Durham, 1980: Critique of Methods to Measure
Dry Deposition. U.S. EPA workshop summary. Avail. NTIS as PB81-126443.
Hicks, B.B., and R.T. McMillen, 1984: A simulation of the eddy accumulation
method for measuring pollutant fluxes. J. Glim. Appl. Meteorol. 23 (4),
637—643. '~°—
Hicks, B.B., D.D. Baldocchi, R.P. Hosker, Jr., B.A. Hutchison, D.R. Matt, R.T
McMillen, and L.C. Satterfield, 1985: On the Use of Monitored Air Concentra-
tions to Infer Dry Deposition. NOAA Tech. Memo. ERL-ARL- (in press")
ATDL contrtb. #84/24. '
19
-------
bD
•H
P-4
.20
-------
ATDL-M 84/335
1—
T
18 MAY,1984
0 In nrJh i rH Pi n
e
o
o 4
o
LU
o
Q_
LU
0
20
n ni • n
04
08 12
TIME (hours, EST )
16
Figure 3- Dry deposition velocity estimate for HNO^ over several days, using
prototype system. The mass flow rate through the filterpack would
be varied according to this signal.
21
-------
en
•H
3
o
a
o
•H
P
aJ
H
I
O
o
nJ
a
o
•H
id .
fc a
-P 0)
d -P
0) CQ
O S
fl W
o
O bD
a
0) -H
ft -d
>> ^
•p o
o o
P Q)
O JH
ft TJ
ft fi
05 -H
!-, to
bD en
O
-------
SIMPLE METEOROLOGICAL AND CHEMICAL FILTERPACK MONITORING SYSTEM
FOR ESTIMATING DRY DEPOSITION OF GASEOUS POLLUTANTS.
R. P. Hosker and J. D. Wo mack*
Atmospheric Turbulence and Diffusion Division
NQAA/ERL/Alx Resources Laboratory
Oak Ridge, TN 37831
INTRODUCTION.
The equipment and procedures described here are In use In a prototype network for long-term
monitoring of dry deposition of gaseous (and later, parttculate) pollutants. At this time,
the method Is limited to S0£ and HN03 vapor. The method Is only one of several being
tested and compared at a sub-net work of "core" research sites. The goal of this work Is to
determine dry deposition spatial distributions and temporal trends with the same
statistical confidence as for the wet deposition case.
Dry deposition fluxes can be evaluated from meteorological data and concentration
measurements by several techniques. Most of these are too complex or too operationally
demanding to be suitable for use In routine monitoring applications. However, so-called
inferential methods appear promising. Inferential methods Infer the dry deposition
velocity VD of a pollutant from meteorological and local surface condition data;
concentrations C of air pollutants are measured directly. The dry deposition flux Is then
estimated from the product of VD and C for each pollutant. Errors can arise from
Inaccuracies or Inadequacies In the procedure used to Infer VD, and from errors in
approximating the time average of products by the product of averages. However, the method
is suitable for long-term monitoring If these errors can be reduced or quantified. The
inferential method Is the basis for the system described here.
The measurement philosophy Is to measure appropriate physical, chemical, and biological
variables with enough temporal resolution to compute weekly average^ dry deposition fluxes
at a large number of sites. This monitoring network Is supported by a subset of
research-grade stations to provide details of distributions, episodes, and deposition
processes; these are used to improve the techniques applied to the less exact but more
extensive array of fUterpack monitoring stations.
It is difficult to measure low concentrations of pollutants with good time resolution using
inexpensive apparatus. One solution Is to accumulate pollutants In a fUterpack sampler
over some time Interval, an appropriate averaging time must be selected. The equations
for estimating the average dry deposition flux F can be successively approximated:
F - ^ I V^ * C dt
(1)
Affiliated with Oak Ridge Associated Universities.
23
-------
N
(2)
£ V c.
(3)
How good are these various approximations? Co-location of filterpack monitoring stations
and research-grade (core) measurement sites using continuous pollutant monitoring Is
intended to help answer this question. At present, a weekly collection schedule (average)
is used.
EQUIPMENT AND OPERATION.
Figure 1 shows a diagram of a filterpack station. The computer routine used to Infer the
dry deposition velocity requires data on wind speed, wind direction fluctuations (o@),
solar radiation, temperature, relative humidity, precipitation, and local surf ace wetness.
These data are acquired from sensors mounted at 10 m height. The sensors are sampled every
10 s by a small commercial data logger, which preprocesses the data, calculates 15 mln and
60 mln averages and statistics, and records the results on a cassette tape. A reference
voltage checks the data logger accuracy and also the AC power Hne, so that power outages,
which may cause erroneous temperature and humidity readings, are flagged. An active mass
flow controller maintains a constant flow (3 I/mln) through the filterpack, regardless of
the degree of filter loading; the data logger monitors the flow rate and controller
voltage. A trickle-charged battery powers the data logger, to preclude battery changes and
consequent reprogrammlng of the system. The Instrument enclosure Is self -dnsulated and
weathertight, so that It can be mounted on the tower without special protection. The
enclosure Is heated and fan-cooled to provide reliable Instrument performance. Figure 2
shows a typical system.
Figure 3 shows the filterpack assembly, a modification of the system commonly used In
Canada. A weakly-heated elutriator tube removes particles and helps prevent condensation
of ambient moisture on the filters. The air then flows through a Teflon particle filter
(2 ym), a nylon filter to remove HN03, and a sodium carbonate-doped cellulose filter for
collecting S02«
The data tape and the filters are replaced every Tuesday morning. The data tapes are
returned to our laboratory where a QC screening Is applied; acceptable data are then copied
into desktop computer files for analysis. The filters are sent to the U.S.G.S. laboratory
in Atlanta, GA for extraction and Ion chromatographic analysis following a protocol
designed by the EPA's ASRL/Aerosol Research Branch. Results are merged with the
meteorological data, and archived. Estimates of the weekly average dry deposition
velocities for HNO_ and SO2 are calculated using the meteorological data and observations
of surface conditions, and multiplied by the measured weekly average concentrations to
determine the average flux.
Additional system details may be found In Hosker and Womack (1985).
DRY DEPOSITION VELOCITY ESTIMATION.
Successful flux estimation via concentration monitoring requires the ablHry to Infer VD
accurately from meteorological data, pollutant properties, and observations of local
surface conditions and vegetation. Space precludes extensive discussion of the computa-
24
-------
tional methods; see Hicks et^ al. (1985) for details. Briefly, the resistance analog Is
used; the total resistance to"mass transfer, RT, Is given by the sum of three terms, an
aerodynamic resistance to turbulent transfer, R^, a diffusive boundary layer resistance,
RB, and a resistance to uptake by vegetation, soil, and other receptors, RC. Thus
+ R,
and
(RT)
-1
(4)
(5)
The aerodynamic resistance depends on the atmospheric turbulence, and may be written In
terms of the stability -dependent vdnd profile (Wesely and Hicks, 1977):
[ In(z/z0)
(u arl
or, In terms of the standard deviation of wind direction ^ 0
(neutral and stable)
RA S 9 (U o2) * for z/L < 0.1
e
(6)
(7)
(8)
(9)
(unstable)
The diffusive boundary layer resistance depends both on atmospheric conditions and on
pollutant characteristics. Mathematically, Rg depends ori the poorly known roughness
length for pollutant transfer, znf,:
Uo
ln(z0/zoc),
(10)
so an empirical expression based on numerous studies of heat transfer Is used instead:
,2/3
Rfi s 2
(Sc/Pr)
(ID
The Schmidt number Sc enters to account for the difference In heat and pollutant
dlffusivltles.
The most complex term Is the surface uptake resistance, R_, which depends on meteorological
conditions on the microscale, on pollutant characteristics, and on receptor surface
characteristics. For highly reactive gases, capture can be quite efficient, so R- =0.
Soluble gases act much Hke water vapor, but with an appropriately different dlffusivity.
Wet surfaces can be especially efficient collectors. The various biological resistances
(stomatal, cutLcular, mesophyll) must be estimated, and weighted by the amount of active
leaf area (using the leaf area index) because multiple surfaces act In parallel. Stomatal
resistance depends on available light, and hence on location in the plant canopy.
Biological uptake varies with the season, and with stresses such as heat and drought. Many
of these dependences are still poorly understood; a good deal of work Is needed to improve
this portion of the procedure to estimate V_.
THE NETWORK.
Figure 4 shows the prototype network. Additional equipment will soon_be Installed In
central Illinois and in northern Georgia; other sites are being considered.
25
-------
To assess the adequacy of the filterpack monitoring method and to allow a steady program of
improvements to the prototype computational routine for Vjj, three stations are co-located
with continuously monitoring stations at "core" research sites in Illinois, Pennsylvania,
and Tennessee. Direct eddy correlation measurements are also made periodically at these
sites, to test the other flux estimates. So-called "concentration accumulation" apparatus,
described In another paper, will be added to these core sites shortly.
RESULTS AND CONCLUSIONS.
A prototype network of meteorological and chemical fUterpack monitoring stations has been
In operation since late summer of 1984. Meteorological and surface data are collected with
15 mln time resolution and used to Infer a weekly average dry deposition velocity. Weekly
average S02 and KNO^ vapor concentrations are obtained from the filters, using the measured
flow rate. Weekly average fluxes are calculated from the product of these two terms. The
adequacy of the approximation will be tested by comparison to results from higher time
resolution data. Preliminary results for one month at one site are shown In Table 1; the
simple result Involving the product of averages Is within 35% (and usually within 20%) of
the average of the products from the much more complex and expensive continuous pollutant
monitoring system. Additional comparisons are underway, as the data are reported to
us.
ACKNOWLEDGEMENT.
This research Is funded as part of the National Add Precipitation Assessment Program by
the National Oceanic and Atmospheric Administration, as part of collaborative research by
NOAA, the EPA, and the U.S.G.S.
REFERENCES.
Hicks, B.B., D.D.Baldocchi, R.P. Hosker, Jr., B.A. Hutchison, D.R. Matt.R.T. McMillen,
and L.C. Satterfleld, 1985: On the Use of Monitored Air Concentrations to Infer Dry
Deposition. NOAA Tech. Memo. ERL-ARL- (In press). ATDL contribution #
84/24.
Hosker, R.P., Jr., and J.D. Womack, 1985: Prototype Meteorological Monitoring and
FUterpack Sampling Systems for Dry Deposit-Ion Measurement. NOAA Tech. Memo.
ERL-ARL- (in preparation).
Wesely, M.L., and B.B. Hicks, 1977: Some factors that affect the deposition rates of
sulfur dioxide and similar gases on vegetation. J. Air Poll. Cont. Assoc. 27 (11),
1110-1116. —
s !•Sample SO2 deposition flux comparisons at Oak Ridge
"core" site, for October, 1984.
Average Flux (grams/sq.m/week)
, N
Dates
Oct. 1 - Oct. 8
Oct. 8 - Oct. 15
Oct. 15 - Oct. 22
Oct. 22 - Oct. 29
w £ V*!** C(ti)
16.7
30.7
51.8
40.2
vD*c
18.8
32.6
59.7
54.4
26
-------
HEATED
Q.
s —
a.
||
U
Ul fc .
-11
il*
05 X
OTUJ
Ul
BC
2° —
1- "5
vJ
59
_i£
"•z
o
u
/
/
/•
o
^ s
OC UJ
co O
_i o E
S- u.
U. CC UJ "J < __^_
u. W H ^^^
1 = = ° K Q.
/' \ i n
1
.^ *
bk "
£ " K
Ul
0 £
S ' : " -5
Q co
w — w
g > V £ V
U O ^ 1 T 1
^J ^ ^ »^ 1 1
O CD ^3; >
£ Mw 1
£ ' ^~~^
O |
-p
cfl
-p
to
u
o
o
bD
CO
p!
O
•H
-P
oi
-p
to
bD
a
•H
M
O
-p
8
•H
-P
•H
to
O
ft
Q)
•d
UJ CQ
•P a
O c6
-p ft
O 5n
^H <1>
ft -P
O
-------
Figure 2. typical equipment enclosure, with .heater and fan, showing data
logger and recorder, signal conditioning, battery and charger, and
other accessories.
Figure 3- Filterpack assembly, with heated elutriator tube. Particle, HNOj
and SC>2 filters are shown near holders. Units are mounted in
insulated weather shield 7 m AGL.
28
-------
CO
E
01
•p
CO
t-t
O>
-P
O
o
cd
ft
JH
0)
-p
CO
o
•H
-P
Oi
-P
co
(30
C
•H
O
-P
•H
1 m
C CO
O 0)
•H -p
-t-3 'H
•H CO
CO
O fi
ft O
0) £H
>» to
?H 0)
0) -
ft ai
>> fn
-p o
o a
-p -
o
*H -P
ft cti
O '(L)
^1 Id
!n O
O O
> H
•P I
Oi O
B u
-d-
0)
bC
•H
29
-------
EVALUATION OF AN AUTOMATED TUNGSTIC ACID
TECHNIQUE FOR NITRIC ACID AND AMMONIA
B.R. Appel, Y. Tokiwa and E.L. Kothny
California Department of Health Services
2151 Berkeley Way
Berkeley, CA 94704-9980
INTRODUCTION
A laboratory and field study was performed to construct and evaluate an
automated, semi-continuous monitor for ambient air concentrations of nitric
acid and ammonia utilizing tungstic acid-coated denuder tubes (1,2).
Interference in HN03 measurements was determined with nitrogen dioxide,
nitrous acid, and particulate ammonium nitrate. Atmospheric sampling was done
in Riverside, CA with the automated tungstic acid technique (TAT), and by the
denuder difference method (DDM), considered the reference procedure for HN03.
EXPERIMENTAL
Figure 1 is a schematic of the. technique as adopted for our field trials.
The tungstic acid-coated preconcentrator functions as a diffusion denuder,
retaining both HNO 3 and NH3 with high efficiency. Following sampling, the
preconcentrator is heated, desorbing HN03 as NO and NH 3 into a stream of
carrier gas (helium). The desorbed NH3 is retained on a shorter tungstic
acid-coated transfer tube, permitting the NO to reach a chemiluminescent NO
A
analyzer for quantitation of the HN03. The transfer tube is then heated, the
evolved NH3 oxidized to NO on a heated gold catalyst in a helium-air
atmosphere to permit quantitation of the NH3.
EVALUATION OF THE TUNGSTIC ACID TECHNIQUE
The TAT is restricted to relatively low dosages. The efficiency for HN03
decreased from 98 to 91% with dosage increase from 340 to 1000 ng sampling at
1 Lpm with a 35-cm coated length of a 4 mm ID tube. The sampling time was
adjusted accordingly. Helium as carrier provided about 30% higher response
for HN03 compared to synthetic air.
The principal difficulties observed with the unit described in Figure 1
delated to memory effects with the inlet ball valve, A, and valve C. The
30'
-------
inlet problem, sorption of HN03 and NH3, was reduced by lining the Teflon
ball valve with pyrex tubing. Problems wi th valve C developed during
atmospheric sampling, apparently due to collection and subsequent slow
desorption of NH^ -containing materials. The resulting elevated and variable
system blank prevented quantisation of atmospheric NH3.
Concerning interference effects, N02 exhibited only minor interference
ranging from 0.3 to 0.5% at 50% and 0% R.H., respectively. Similarly
0,1-0.2 Mm NH^NOa particles at dosages up to 700 ng (as N03~) showed no
measurable interference. Nitrous acid, however, exhibited a substantial and
complex interference. Our inability to quantitate HONO has thusfar prevented
assessment of its interference efficiency. Thermal desorption of WO coated
tubes loaded with nitric acid-free MONO yielded two peaks, one of which
coincided with that for HN03. The peak preceding HN03 by about 12 seconds was
also observable at night and early morning during atmospheric sampling in
Riverside (peak at 1.56 minute, Figure 2). Simultaneous measurements of
atmospheric HONO using a UV absorption technique (3) were compared to the 1.56
min. peak areas (Figure 3). Quantitative correlation was not observed.
However, when HONO was detectable by the UV method, the 1.56 min. peak was
usually observed.
The precision of calibration of the TAT with HN03 from a permeation tube
source (Metronic's Part No. 110-010-0160) at 83°C was assessed by comparing
daily calibrations. For eight such cal ibrations, the calculated HN03
concentration corresponding to a typical daytime peak area exhibited a
coefficient of variation of 10.5%.
The atmospheric HN03 results by the TAT are compared to those by the
denuder difference method (4) in Figure 4. The TAT results average about 50%
higher than those by the DDM. Nitrous acid interference was not a substantial
contributor to the difference as evidenced by the lack of a substantial day-
night effect.
31
-------
ATMOSPHERIC NITRIC ACID RESULTS
Diurnal variations for two, 24-hr periods are shown in Figure 5. The
decrease in HN03 in the interval 1300-1500 hr coincided with brief periods of
rain.
STUDIES CURRENTLY IN PROGRESS
Studies continue to elucidate the cause of the difference in HN03 results
between methods, to quantitate MONO interference and measurement, and to
overcome the problems experienced in measuring NH3. We currently are using a
TAT system in which both the inlet ball valve and valve C (Figure 1) have been
eliminated. An excess flow of carrier gas is vented through the sample inlet
to prevent the intrusion of ambient air into the preconcentrator tube during
the analytical cycle (5). The modified system will be used to sample HN03 and
NH 3 in parallel with spectroscopic and other techniques for these atmospheric
species in an inter!aboratory study to be done in California.
ACKNOWLEDGEMENT
The work described herein is being sponsored by California Air Resources
Board Research Division. The nitrous acid concentration measurements were
provided by Dr. H. Biermann of the Statewide Air Pollution Research Center,
University of California (Riverside). The assistance of Dr. A. Winer and the
staff of the SAPRC is also gratefully acknowledged. Tungstic acid coated
tubes were prepared by H. K. Min.
REFERENCES
1. Braman, R.S., Shelley, T.J. and McClenny, W.A., 1982. Tungstic Acid for
Preconcentration and Determination of Gaseous and Particulate Ammonia and
Nitric Acid in Ambient Air. Anal. Chem. J54, 358-364.
2. Gailey, P.C., McClenny, W.A., Braman, R.S. and Shelley, T.J., 1983. A
Simple Design for Automation of the Tungsten VI Oxide Technique for
Measurement of NH3 and HN03. Atmos. Environ. YJ_, 1517-1519.
32
-------
3.
4.
5.
Pitts, J.N. Jr., Biermann, H.W., Winer, A.M. and Tuazon, E.G., 1984.
Spectroscopic Identification and Measurement of Gaseous Nitrous Acid in
Dilute Auto Exhaust. Atmos. Environ. 18, 847-854.
Appel, B.R., Tokiwa, Y. and Haik. M., 1980. Sampling of Nitrates in
Ambient Air. Atmos. Environ. 15, 283-289.
Roberts, J.M., Hubler, G., Norton, R.G., Goldan, P.O., Fahey, D.W.,
Albritton, D.L. and Fehsenfeld, F.C., 1984. Measurement of HN03 by the
Tungsten Oxide Denuder Tube Method: Comparison with the Nylon Filter
Method. Presented at the American Geophysical Union Meeting, San
Francisco, December.
33
-------
SAMPLE
INLET
OPEN
COMMON
MASS
FLOW
CONTROLLER
i WO3
! PRECONCENTRATOR
Au
CONVERTOR
MASS
PLOW
CONTROLLER
MASS
FLOW
CONTROLLER
PUMP
A = MOTOR-DRIVEN TEFLON BALL VALVE
B = NORMALLY CLOSED TWO-WAY SOLENOID VALVE
C = THREE-WAY TEFLON SOLENOID VALVES
Figure 1. Schematic of Automated Tungstic Acid Technique
-------
1.56
1.76
4.54
Figure 2. Possible Detection of Atmospheric Nitrous
Acid Together With HN03 (0500 hr, 9/17/84)
-------
O.Q
O O !25 &
. o a
CO CO Mv~/
i i
0 0
rH 0 0
-I
^"^-^
O
o
r
CD
O
O
O O
5 §
W W
I
O
o
oo
o
o
o
o
(M
O
o
00
o
•o
CO
O
'O
CO
o
.o
o
^^ (M
to •
1 O
to
-------
35.0-
26.4_
CO
a 17.8H
9.21
0.60
0.20
4.74
TAT = 1.24 + 1.48DDM
r = 0.94
n = 31
9.28 13.8
DOM (jug/m3)
18.3
22.9
Figure 4. Comparison of Atmospheric HN03 Concentrations
Measured by the Denuder Difference Method and
Automated Tungstic Acid Technique
.37
-------
8/18-9/19/84
45.3 _,
34.0
1
O
§
O
22.6-
11.3-
00.0
nrrrfTh
9 11 13 15 17 19 21 23 1 3 5
9/19-9/20/84
32.6-
21.7-
o
§
o
10.8-
9 11 13 15 17 19 21 23 1 3 5
TIME (PDT)
Figure 5. Diurnal Variation of HN03 at Riverside
as Measured with the TAT
38
-------
APPLICATION OF THE TUNGSTEN OXIDE DENUDER TUBE TECHNIQUE TO THE
MEASUREMENT OF NITRIC ACID IN THE RURAL TROPOSPHERE AND
COMPARISON WITH THE NYLON FILTER METHOD
J. M. Roberts*, R. B. Norton, P. D. Goldan, F. C. Fehsenfeld*,
and D. L. Albritton
Aeronomy Laboratory
National Oceanic and Atmospheric Administration
Boulder, CO 80303
*Also, Cooperative Institute for Research in the Environmental Sciences,
NOAA, University of Colorado, Boulder, CO 80309.
Nitric acid (HN03) is an important species in rural atmospheres because
it is a major end product of tropospheric odd-nitrogen (NO ) chemistry, and
because it is a major constituent of acid deposition. It is desirable to
have a relatively fast measurement (15-30 min) of HNOg in the 10-100 pptv
range for use in precipitation removal and dry deposition studies and for
comparison with measurements of other NO species such as NO and N0? which
can also be made on a short time scale. The tungsten oxide denuder tube
technique, originally described by Braman et al. (1982), was applied to the
measurement of HN03 in the rural troposphere. Laboratory studies of HNO-
- o
collection and recovery efficiency were performed along with tests for the
interference of other trace nitrogen-containing compounds. Ambient measurements
were made at a rural site in the Colorado mountains, and the results of those
measurements compared to measurements of HN03 made by the more established
Teflon/nylon filter pair method.
39
-------
The basis of the tungsten oxide denuder tube technique is the selective
chemisorption of HNO-, % (and NH3/ %) onto the inside surface of a tube coated
J3(g)
with tungsten oxide. Nitrate and NH4+, present on aerosols, are not collected
due to the greatly slower diffusion rates of aerosols relative to gas molecules.
HNO, and NH- are removed from the tungsten oxide surface by thermal desorption.
O O
Nitric acid is desorbed from the surface as NO or N02> NH3 is desorbed as NHg.
The compounds evolved from the surface are then converted to NO and analyzed by
the well-known N0-03 chemiluminescence method. HN03 and NH3 can be separated
either by recollection of NFL or by temperature programmed desorption. Due to
the preconcentration obtained by sampling through a coated tube and the inherent
sensitivity of the chemiluminescent method, this technique can yield a short
time scale measurement of HN03 at low concentrations.
Tungsten oxide coatings were prepared in a manner similar to that described
by Braman et al. (1982). Modifications to this method involved progressive zone
coating. In this technique, tungsten (IV) blue oxide was evolved from a short
length (^ 4 cm) of tungsten filament that was connected between 0.25 in. metal
rods. This arrangement could be progressively moved along the length of the
tube to produce a uniform coating of the tungsten oxide on the inside wall of
the tube. The blue oxide so deposited on quartz tubing was then oxidized to
the green-yellow oxide (WO ) at approximately 350°C under clean air flow. This
A
procedure was found to produce a more uniform coating than the single-filament
technique of Braman et al. (1982).
Collection and recovery of HN03 and the extent of interference from other
compounds were studied using the system shown in Figure 1. The system consisted
of a commercial NO monitor with a carbon monoxide-doped gold catalyst operated
at 725°C. The catalyst and NO monitor served to detect the compounds being
examined. Connected to the detector was a 4-port valve which served to switch
the WO tube in and out of line with the detector and the sample stream. The
A
40
-------
sample stream consisted of either dry or humidified zero-air with analyte
species added. Collection efficiency of the WO tube was determined directly
/»
by switching the WOX in-line with the detector and a sample stream containing
a constant level of an analyte species. The resulting decrease in the detector
signal was compared to the signal produced by the constant analyte level to
obtain a % collection. After the desired collection time, the WO tube was
A
switched out-of-line and the detector signal was observed to return to that
representing the constant analyte level. Desorption of material from the WO
X
tube was accomplished by switching the WO tube in-line with the detector and
/\
dry zero-air stream and heating the tube to 500°C. Heat was applied by a
variable transformer connected to a Ni-Cr wire wound around the outside of
the tube. Desorptions performed in this manner produced well defined peaks
in the detector. The area under the peak could be compared to the area repre-
sentina the amount of material collected to determine a % recovery. A flow
rate of 700 seem was used with a WOV coating length of 40 cm. The collection
A :
and recovery of HN03 was observed to be 95±5% in dry (<2.6 ppm HpO) air, and
100±3% for NH3 under the same conditions. The collection and recovery of HN03
decreased slightly with the addition of water to the sample stream, possibly due
to the clustering of HN03 with H20 (Eatough, 1985).
Possible interferences of other nitrogen containing compounds were also
tested at a range of HgO concentrations with the System shown in Figure 1.
The results of studies involving N02, n-propylnitrate, peroxyacetylnitrate
(PAN), and hydrogen cyanide (HCN) are shown in Figure 2. The % response plotted
on the vertical scale consists of the amount of material recovered during
desorption divided by the total amount of material that the tungsten oxide
tube was exposed to. In each case, it was apparent from the collection profile
that incomplete collection was responsible for the reduced overall response to
-------
these compounds. Nitrogen dioxide is perhaps the most important potential inter-
ferant in the atmosphere, but was not collected in the WO tube to any significant
A
extent (<1.5%). PAN was collected at 60% efficiency in dry air (<2.6 ppm H20)
but at only 6% efficiency at 18% R.H. Additional ambient air tests confirmed
that PAN was not an interferant. A representative alkyl nitrate, n-propyl-
nitrate (NPN) was found to be collected with 55% efficiency in dry air, but
at 90% R.H. the collection efficiency was only 0.2%. Although NPN was not
checked at intermediate H20 concentrations, based on the behavior of PAN it
is unlikely that significant collection occurs at lower H20 concentrations
(i.e., 20% R.H.). Hydrogen cyanide was collected with only 10% efficiency in
dry air, less than 3% efficiency for R.H. > 10%.
Ambient measurements of HN03 were made between July and November, 1984
using the tungsten oxide denuder tube method and the Teflon/nylon filter pair
method. The measurement site was located approximately 35 miles northwest of
the Denver/Boulder metropolitan area on Niwot Ridge, Colorado (10,000 ft. elev.).
The prevailing winds bring in clean continental air masses to the site from
the west. However, late morning to mid-afternoon warming of the eastern moun-
tain slopes produces an upslope circulation bringing air from the front range
urban corridor to the site.
The tungsten oxide ambient sampling system is shown schematically in
Figure 3. The WO coated tube and gold catalyst were mounted on top of the
J\
instrument enclosure. The 6-port valve and all the gas lines were made of
Teflon. The WO tube was made of quartz and had two 1/8" O.D. quartz lines
/\
teed into it for standards and zero-air desorption flow. The zero-air flow
was set so that with the 6-port valve in the "analyze" position, there was a
slight overflow out the inlet of the WOX tube. The sample flow used was 1 slpm
42
-------
and sample time was 18 min. The desorption of HN03 and NH3 was accomplished by
a linear temperature program of the WOY section from ambient to 525°C. A
*» " i
resulting desorption profile is shown in Figure 4. Standards were obtained
from permeation tubes which were calibrated by conversion to NO in the gold
catalyst and analyzed by NO-G3 chemiluminescence. Standards were introduced
to the WOX tube, either in an excess of zero-air (dry or humidified) or in a
small flow (50 seem). The detection limit was approximately 30 pptv HNO- and
• - • *J
the overall uncertainty ±20% at HNO^ mixing ratios of 100 pptv or more.
The Teflon/nylon filter pair method has been widely applied to the measure-
ment of HN03 in the rural troposphere. Details of our tests of this method
can be found in Goldan et al., 1983. These tests show that there was negligible
collection of HN03^g^ on the Teflon aerosol filter and >90% collection of HN03
on the downstream nylon filter. Additional tests using side-by-side sets of
filter pairs showed that volatilization of N03" (as NH4N03 or HN03) from the
Teflon filter onto the nylon filter was, at most, a 10% positive interference
in the HN03 measurement. Since the collected N03" is analyzed by ion chromato-
graphy, a relatively insensitive technique, large volumes of air must be sampled
over several hours with this technique.
Results of simultaneous measurements of HN03 by the WO tube technique
HN03 (WOX) and the nylon filter method are shown in Figure 5. The tungsten
oxide measurements are averages of measurements made during the 2- or 4-hour
nylon filter sample time. The results show that the tungsten oxide method
measured, on the average, 3 times more "apparent" HN03 than the nylon filter
method. The discrepancy between the two methods is sometimes as much as a
factor of ten. Only two points are at or below the 1:1 tungsten oxide-nylon
line. Thus, it appears that other species are collected on the WO surface
X
and interfere with the HN03 measurement.
43
-------
A dual tube system was employed to check for the presence of interfering
compounds. The two tubes run side-by-side agreed to within 15%. Nylon wool
was placed in the inlet of one tube to remove HN03- The tube with the nylon
wool trap gave three measurements of 270±60 pptv (average) while the untrapped
tube gave three measurements of 500 pptv. The difference, 230 pptv, compared
very well with the nylon filter measurement of 190 pptv made just subsequent
to the WOX tube measurements. Therefore, we conclude that there are other
nitrogen containing species that are collected on the WO surface and interfere
/\
with HN03.
Possible synergistic collection of PAN from a whole air sample was tested
by the addition of a small stream (50 seem) containing PAN to every-other ambient
sample collected in the WOV tube. The equivalent mixing ratio of PAN at the
A
1 slpm sample flow was 1100 to 890 pptv. No significant difference was noted
in the ambient measurements with and without added PAN, which confirmed that
PAN is collected at <5% efficiency under ambient conditions.
Correlations of HN03 (WOX) with other odd-nitrogen species measured at
this site were examined for clues as to the identity of the interfering species.
The correlation of HN03 (WOX) with NOX (NO + N02) (Singh et al., 1985) showed
that the interfering species are probably photochemical products of NO . The
X
correlation of HN03 (WOX) with PAN (Singh et al., 1985) was very good, indi-
cating that the interfering species, although not PAN, may be PAN-type com-
pounds, i.e. organic nitrates. This seems to be a feasible explanation, since
a wide variety of organic nitrates are thought to result from organic photo-
chemistry (Atkinson et al., 1982) or nitrate radical (N03) chemistry (Bandow
et al., 1980).
44
-------
ACKNOWLEDGEMENTS
We thank the staff of the University of Colorado Institute for Arctic
and Alpine Research for their logistical support at the Niwot Ridge site.
This work was supported by the National Oceanic and Atmospheric Administration
as part of the National Acid Precipitation Assessment Program.
REFERENCES
Atkinson, R., S.M. Aschmann, W.P.L. Carter, A.M. Winer, J.N. Pitts, Jr., (1982),
Alky] nitrate formation from the NO -air photooxidation of C0-C0 alkanes,
J. Phys. Chem., 8£, 4563-4569. x 28'
Bandow, H., M. Okuda, H. Akimoto, (1980), Mechanism of the gas-chase reaction of
C3H6 and N03 radicals, J. Phys. Chem., 8£, 3604-3608.
Braman, R.S., T.;J. Shelley, W.A. McClenny, (1982), Tungstic acid for preconcen-
tration and determination of gaseous and particulate ammonia and nitric acid
in ambient air, Anal. Chem., 54_, 358-364.
Eatough, D.J., V.F. White, L.D. Hansen, N.L. Eatough, E.G. Ellis, (1985),
Hydration of nitric acid and its collection in the atmosphere by diffusion
denuders, Anal. Chem., 57_, 743-748.
Singh, H.B. L.J. Salas, B.A. Ridley, J. Shetter, F.C. Fehsenfeld, D.W. Fahey,
•I' laTI!t!? E'J- Wll1iams» S.C. Liu, (1985), Relationship between peroxyacetyl
nitrate (PAN) and nitrogen oxides in the clean troposphere. Submitted to' Nature.
FIGURE CAPTIONS
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Laboratory system used for collection/desorption and
interference studies.
Results of interference tests involving N09, HCN,
n-propylnitrate, and PAN. ' ^
Tungsten oxide denuder tube system used for ambient air
measurements.
Desorption profile of an ambient air sample.
Nitric acid measured by the tungsten oxide tube, HNO.JWO ),
vs. HN03 measured by the Teflon/nylon filter pair. x
-------
Figure 1.
DRY OR HUMIDIFIED
AIR IN
NO
MONITOR
SOLID GOLD
TUBE CATALIST
WOX
COATED
TUBE
4-PORT
VALVE
46
-------
Figure 2.
IUU
o 50
T—
X
Q
LU
DC
LU
O
O
LU
DC
!i
•*••
£ 20
O
0 10
o
DC
t 5
z
< 2
•—•*
LU 1
CO
0 0.5
CO
LU
^o
0.1
A • N02
• HCN
APAN
*N-PROPYL-NITRATE
_. _
~ A
—
- . '-
-•
— • . - —
' * * -
I I I I I I I I I t
0.008 10 20 30 40 50 60 70 80 90 100
%R.H. @ 25°C
2.6 0.63% 1.56% 2.81%
47
-------
Figure 3.
CAPPED OFF-
SAMPLING PUMP
& FLOW
CONTROLLER
6-PORT TEFLON VALVE
SAMPLE
ANALYZE
ZERO AIR
700 SCCM
4 MM ID.
FUSED SILICA
TUBING
WOX, (H2W04)
COATING
GOLD CATALYST, 725 C
NO-O3
CHEM1LUMINESCENCE
DETECTOR
HUMIDIFIED ZERO AIR
DILUENT FLOW
NH3FROM PERMEATION TUBE
AND DILUTION SYSTEM
HNO3 FROM
PERMEATION TUBE
WASTE
3-WAY VALVES
POLYETHYLENE
FUNNEL
48
-------
Figure 4.
o
o
cL
.o
500
400
300
200
100
0
HNOs
1^ I I
120 240 360
Time, sec
49
-------
Figure 5.
1 I I I I I I I | p I I I I I I I I \IL I I I I J I" I A
10 10 10
HNO, (NYLON), pptv
50
-------
THE USE OF LICHENS TO MEASURE THE TRANSPORT OF AIRBORNE HEAVY METAL
POLLUTANTS
James N. Beck, Robert L. Thompson, Celinda Mueller, Dennis Casserly,
Pam Langley and John Young
Department of Chemistry
McNeese State University
INTRODUCTION Lichens have been known to be sensitive to air pollutants
since 1859 when it was observed that lichens were disappearing from
Manchester England. A variety of heavy metals are associated with air
pollutants from such industries as power generators, refineries,
smelters, etc. Lichens absorb heavy metals primarily from dissolved
ions in rain water and retain them for long periods of time. It was
thus felt that the Lake Charles, LA, area was an ideal area to observe
long range changes in concentrations of heavy metals since the
industrial complex is centrally and locally concentrated in the parish.
This study area is also isloated from other heavily industrialized
centers. The plants include two oil refineries, coal-fired power plant,
four chemical plants and an aluminum plant. Lichens were collected from
16 stations located in a grid around the industrial corridor in
Calcasieu Parish located in extreme southwest Louisiana. In addition,
two stations were also established in Cameron Parish lying south of
Calcasieu Parish and bordering on the Gulf of Mexico. These two
stations were selected to serve as control stations. The collections
were done during three seasons in 1983-1984 to determine if variations
with distance and time occur across the Parish, and the magnitude of the
51
-------
variations.
EXPERIMENTAL The lichens Parmelia praesorediosa and Ramalina stenospora
were collected using teflon coated tweezers from the trunk and lower
branches of oak trees. The lichens were placed in preweighed plastic
irradiation vials and labeled by station, date and lichen type by code
number on the vial. Collections were done on three occasions, August
1983, February and May 1984. This was done to observe seasonal
differences in metal uptake or elimination. The May sampling survey
followed an extended period of dry weather. The samples were then
irradiated for fifteen seconds at the Texas A&M University 1 MW pool
reactor at a flux of 10^ n cm~2 sec . The samples were counted after
a fixed delay time on a Harshaw Ge(Li) detector and the spectra stored
and analyzed using a Hewlett Packard Scorpio multichannel analyzer.
This method allowed the determination of vanadium, aluminum, manganese
and in a few lichens calcium, titanium and copper.
RESULTS AND DISCUSION An example of the results of the analysis of the
data are given in Figure 1 for vanadium. The stations are -given by
letters and the industries are given as circled numbers. Ths isolines
were drawn through stations having concentrations equal to the value of
the isoline. It is clear that the highest values of vanadium are
located, as expected, near the industrial corridor in central Calcasieu
Parish. The lower values were found along the Gulf of Mexico more than
80 km from the industries. However, it is clear that the concentrations
52
-------
decrease regularly from the industrial corridor. The only exception
from this trend was toward the northwest where the decrease is more
rapid. However, few stations are located in that direction and this may
have affected the results. The data plotted are the means for all
lichens and the three sampling periods. The trend shown for vanadium in
Figure 1 was very similar to those shown for manganese and aluminum.
The variation for vanadium was found to be 2.1 to 10 ppm whereas the
variation for manganese was 20-180 ppm and aluminum was 900-4500 ppm.
All high values were located near the refineries located near West Lake,
Louisiana with all low values located near the Gulf of Mexico. The
concentrations across Calcasieu Parish appear to follow a decrease given
by an equation:
C = CQe"XTD
where C is the highest concentration and C is the concentration at
distance D, X is the disappearance rate which is a function of stack
height, wind and particle size. All data were analyzed using SPSS
computer codes. The results of the analysis showed that variations with
distance were statistically significant. Discriminant analysis was also
done to see if all metal conentrations used together would show
variations between preselected groupings of stations. All groups were
found to be similar except that station I had to be put in the outer
group of stations. It would appear that the variation of metal
concentrations with distance is real and has its origin in the
industrial corridor.
53
-------
Figure 1. The variation of the concentrations of vanadium (ppm) across
Caloasieu Parish, Louisiana. The sampling stations are given as letters
and the major industries as circled numbers. The lines drawn connect
stations having similar concentrations.
I
N
VANADIUM
543210 5 10 15 20
GULF OF MEXICO
54
-------
Measurement of HNCU, SCL, NH3 and Participate Nitrate
with an Annular Denuder System
by
R. K. Stevens and R. J. Paur, U. S. Environmental Protection Agency
Research Triangle Park, NC
I. Allegrini, F. DeSantis, A. Febo, C. Perrino and M. Possanzini
CNR - Istituto Inquinamento Atmosferico, Rome, Italy
K. W. Cox, E. E. Estes, A. R. Turner and J. E. Sickles, II
Research Triangle Institute, Research Triangle Park, NC
Abstract
A ten-day study was carried out in the spring of 1985 in the Research
Triangle Park to demonstrate the performance of a system based on the annular
denuder method (ADM) for collection of a number of species of importance in
acidic dry deposition. The ADM was compared with a denuder difference
measurement (DDM) and filter pack for the measurement of gaseous nitric acid
and particulate nitrate. The ADM sulfur dioxide collection and analysis
precision was determined to be ±5% for sulfur dioxide concentrations in the
range of 1 to 6 |jg/m ; tjrie precision for HNO~ was ±4% over the concentration
range of 0.2 to 1.6 jjg/m . The sensitivity or the ADM is approximately 0.6 ug
for eitther NO- and SO,, Corresponding to a nitric acid or SOp concentration
of approximately 0.1 ug/m sampled at 15 liters/min over a 6 hour interval.
Preliminary results from the comparison of nitric acid measurements
indicate that the ADM and the DDM give the same results to ±20% at
concentrations of 1 ug/m and 12 hour sample intervals. Details of the
experimental protocol and results are discussed in this paper.
Introduction
A number of primary and secondary air pollutants contribute to the
acidity in depositions from the atmosphere onto receptors such as soils,
vegetation, bodies of water and man-made surfaces. The gaseous and
particulate pollutant species that are postulated to be among the major
contributors to acidic deposition are listed in Table 1. Typical non-urban
concentrations of those species are also presented in the table.
Commercial instrumental methods are available to monitor some of these
pollutants, such as SOp, NO, NOp and 03. However, the sensitivity of these
instruments is not adequate to measure these pollutants, except for 0~, at the
concentrations typically found in non-urban locations. In addition, the cost
and maintenance of these monitors is often prohibitive for many air quality
studies. For these reasons a number of investigators over the past five years
have been developing filter packs and denuders to measure these species by
conventional wet chemical procedures.
During the past two years a new denuder design — the annular denuder --
has been developed and tested at the Laboratory for Atmospheric Pollution of
C.N.R., Rome, (Italy) which has led to a device described by Possanzini et al.
(1983) that is suitable for the simultaneous measurement of gaseous and
particulate species, such as
contribute to deposition. The
HNO
HN0,
U.S. Environmental
NHV SOp, and SO., which
Protection Agency (US EPA)
55
-------
invited a team from C.N.R., Rome, Italy to bring the equipment needed for the
measurement of acid species and to compare the results with those obtained
with devices currently being evaluated by the Research Triangle Institute in
North Carolina for the US EPA. The principles and characteristics of the
annular denuder method (ADM) and the results from the field study are
described below.
The objective of the field study was to compare an annular denuder system
with the widely used denuder difference method (DDM) of Shaw et al. (1982) to
determine whether there are advantages over the conventional DDM system to
warrant additional research on annular denuders in the EPA methods development
program. This brief study relied on (a) a comparison of the performance of
open-tube and annular denuders, coupled with filter packs, to measure ambient
concentration of HNO-, S0? and particulate nitrates, and (2) a determination
of the precision of the annular denuder method for the measurement of S02 and
HNO~ to estimate the value of a more detailed evaluation/development program
for the annular denuders.
Principles and Characteristics of the Annular Denuder System
The C.N.R. denuder is an annular tube configuration (Fig. 1). During
sampling, air is drawn under laminar flow conditions through the annular space
between two concentric glass cylinders that have been coated with a chemical
that reacts with selected trace species. As the sample stream passes through
the annular space, the gaseous trace species travel by molecular diffusion
from the bulk gas to the reactive surface and are collected.
The reader is referred to the work of Pozzanini, et al. (1983) for a
discussion of annular denuder tube theory. The operational summary of the
theoretical comparison of the denuders described in Table 2 is:
F =30 F
where
F is the flow rate
L = length of annular denuder
Lj = length of open tube denuder
For a given flow the annular denuder can achieve equivalent collection
efficiency in ~1/30 of the length required for an open-tube denuder, or for a
given denuder length, the annular denuder can sample at ~ 30 times the flow of
an open-tube denuder. _-,
The high operating flow rate (10-30 £min ) makes the annular denuder
very useful for experiments where collection of low concentration of certain
gases are required over short term sampling (1-4 h) periods. The high flow
rate also permits more material to be collected on membrane filters downstream
of the denuder. Previously, collection of large quantities of particles while
simultaneously preserving the integrity of the particles through removal of
reactive gases (eg, HNO-, NH,, and S0p) required that the filter be preceded
by a parallel multitube denuder assembly similar to that described by Stevens,
et al (1979).
For a denuder to be effective, the system needs to ensure separation of
the gases and particles. However, diffusional and inertia! deposition at the
inlet can result in particle uptake which has been determined experimentally
(Possanzini et al., 1983) to be not larger than about 3%. The transit time of
56
-------
air through the denuder is <0.1 second, reducing the opportunity for
substantial disturbance of the atmospheric gas-particle equilibrium existing
in the atmosphere. The walls of the denuder are etched by sand blasting with
100 jjra sand particles. This feature increases the surface area available for
chemical coating and as a result the capacity of the. denuder to collect the
pollutants of interest can be increased to several milligrams.
The use of a water soluble and 1C compatible substrate to coat the walls
of the tube (e.g., Na^COO simplifies the extraction and the analysis of the
sample. Other substances deposited on the denuder may give rise to the
formation of the same ions; for example, deposition of particulate matter
containing chlorides, sulfates and nitrates interfere with the measurement of
HC1, SO z and HNO-. The absorption of NO, and PAN on Na^O, yield nitrites
which interfere with the measurement of HNOp. However, the efficiency for the
collection of these interfering species is relatively small (about 1 to 3%)
(Perm and Sjb'din, 1985, and Peak and Stevens, 1985). Thus, the amount of
relatively unreactive interferents collected in the first denuder will be
approximately equal to that found in the second denuder. This feature can be
used to correct data obtained from the analysis of the first denuder. The use
of two denuders in series will then permit the simultaneous analysis of
several acidic compounds, even though the ratio of analytes in the gas phase
and particulate matter is extremely low. For instance, the technique is
valuable for the measurement of trace levels (< 0.1 ug/m?) of S02 and HNO- in
the presence of large quantities of sulfates and nitrates in particuTate
matter and, in addition, the use of two denuders will permit the measurement
of small amounts of HNO,,. The ADM permits short and long samplings (1 hr to 1
week), and permits determination of diurnal profile of the concentration of
reactive species.
The Experiment
Experiments were conducted between March 27 and April 3, 1985. Samples
were collected at the Research Triangle Institute's acid deposition research
site located in the RTP, NC. The system based on the ADM is shown in Figure 2
and the DDM (Shaw et al 1982) is shown schematically in Figure 3. Table 2
summarizes the characteristics and operational conditions of the two types of
denuder systems that were used during the field intercomparisons. Replicate
24 h, 12 h and 4 h ADM samples were collected; DDM samples were collected for
24 h and 12 h periods.
Annular Denuder: The flow rate through the ADM (Fig. 2) was maintained with
downstream differential flow controllers. The ADM consisted of the following
components: ,
(a) A cyclone for size segregation of particles entering the sampling train.
In the present configuration a Teflon cyclone having a cut size of
approximately 2.5 urn at 15 £/min was used.
(b) Two annular denuders in series coated with a 1% Na2C03 and 1% glycerine
in a 1:1 water-methanol solution for the collection of acid species.
(c) A filter pack which accommodates a Teflon filter for,the collection of
particulate matter and a backup nylon filter to collect nitrates
evaporated from the Teflon filter.
The components of the ADM are easily assembled by means of threaded rings
and connectors. After sampling, the tubes can be recapped for safe
transportation to the laboratories where analyses will be carried out.
57
-------
Denuder Difference Method (DDM) Sampler: The sampler based on the DDM and
used in this study (shown in Fig. 3) consists of the following components:
(a) Teflon cyclone to remove particles larger than 2.5 urn when operated
at 28 Ji/min.
(b) Teflon manifold located downstream of the cyclone.
(c) A tubular denuder coated with MgO followed by a 25 mm diameter nylon
filter, 2 A/min flows through this leg of the system
(d) A Teflon tube followed by a 25 mm diameter nylon filter, 2 £/min
flows through this leg of the system.
(e) Two Teflon filter packs containing a 47 mm diameter Teflon filter
followed by a TEA-treated filter also collect aerosol from the main
sampling manifold.
One 12-hour experiment was carried out to compare the collection
efficiency of oxalic acid, phosphoric acid, and citric acid based denuder
coatings for the collection of ammonia. A 1% solution of each acid in 1:99
glycerin/methanol solvent (20 ml total volume) was poured into separate
denuders, the denuder was rotated to ensure complete coating of the surface,
the excess solution was drained out and the denuder was dried with a clean air
stream. The Na^CO- coated tubes used for acidic species were prepared in a
similar fashion.
Analysis: The nylon filters were extracted by adding 20 ml of Na^CO- ion
exchange solution to a 30 ml Nalgene bottle containing the filters. The
extract is sonicated for 30 minutes and a aliquot injected into the ion
chromotograph (1C) instrument for nitrate determination.
The annular denuders were extracted with 20 ml distilled water and
aliquots analyzed by ion chromatography, or by colorimetry (Indophenol Blue
Method) when ammonia was measured.
The 1C analysis of the extract of the Na2C03 denuders gives the
concentration of chlorides, nitrites, nitrates ancr sulfates formed on the
denuder by uptake of HC1, HN02, HNO~ and SOp, respectively.
Results and Discussion
Table 3 shows the results of replicate comparisons between two annular
denuder assemblies. Also during these replicate comparisons, a DDM system
collected samples during the 12 h and 24 h intervals in parallel with the ADM
assemblies. The nylon filters from the DDM were extracted to determine the
nitrate and nitric acid concentrations. The ambient nitrate concentration was
typically less than 0.2 ug/m?; the DDM system was unable to make this
measurement, since the flow rate through the open-tube denuders was only
2 A/min and the amount of sample collected was too small for accurate
determination.. At these flow rates, for experiments lasting 12 hours, only
1.34 m? of air is collected. If,the concentration of nitrate is 0.2 |jg/m?,
then in a 20 ml extract the concentration of nitrate per ml is less than
0.01 ug/ml which is below the detection limit of the 1C. Table 4 shows the
HNOo data from'ADM and DDM systems. We report only data for sampling periods
> 12 hrs and when the concentrations of HN03 were near or greater than
1 ug/m?. These criteria for data selection are Based on the limitation of the
DDM noted above.
One experiment was performed to test the collection efficiency of three
different acid coatings to collect NH-. The coatings were phosphoric acid
oxalic acid and citric acid. All three denuders collected, over a 16 hour
sampling period April 4-5, 1985, 1.5 |jg NHL. The differences between the 3
collection devices were nominal. Since the citric acid is essentially
58
-------
non-volatile and compatible with both 1C and colorimetric analysis, it is
preferred as the denuder coating to collect NH.. Perm (1979) used oxalic acid
coated denuders in his studies to collect NfL. However, the volatility of
oxalic acid makes its use questionable durffig hot summertime conditions.
The replicate S02 and HN03 data shown in Table 3 demonstrate that the
reproducibility of th«r collection and analysis procedures used in this study
were3typically better than ± 5% for SO. over a concentration nange of 1-6
ug/m and ±4% for HN03 over a concentration range of 0.15-1.6 ug/m .
Figure 4 shows the concentrations of HNO, and particulate nitrate
collected in 12 hr !day-night! intervals. The results are consistent with
photochemical theory predictions that HNO, concentrations would be higher
during daylight hours and tend to diminish it night. The particulate nitrate
data did not show a distinctive diurnal pattern.
An interesting feature of the annular denuder assembly used in this study
was the HN02 data obtained from the extract of the Na.CO, coated denuders.
The HNO. concentrations were often greater than 0.2 ug/m? and, as predicted
from photochemical principles, higher at night than during the daytime.
Sjodin and Perm (1985) have made this same general observation in studies
performed in Gb'tenburg, Sweden; however, for two months of their 5 month study
they measured daytime HN02 levels greater than the nighttime levels. Recent
annular denuder measurements reported by Peake and Stevens (1985) in Calgary,
Alberta also observed high daytime (<0.5 ug/m ) HNO. concentrations.
Fig. 5 shows an example of monitoring HNO. ind HNO, at 3 hr sampling
intervals in Rome, Italy in February 1985 witTi the saiffe denuder assembly
tested in the study described in this report. Nitric acid accumulates during
the day and, due to its large deposition velocity, almost disappears
overnight. Nitrous acid (HNO.) concentration is typically higher at night
because during the day it is rabidly photo!ized by the sunlight.
A typical chromatogram obtained from 1C analysis of annular denuder
extracts is shown in Fig. 6. There is no visual evidence of SOA or NO. in the
second denuder; this shows the high collection efficiency of the first denuder
for HNO, and SO. and implies that deposition of particles in the denuder is
low. Absence of N02 in the second denuder is interpreted as an indication
that the N02 in the first denuder is due to a very reactive species such as
HN02 rather than PAN or NO., which react much more slowly with the Na.C00
denoder surface (Perm and SjSdin, 1985). * d
Summary
An annular denuder system was used to collect samples for the
determination of ambient levels of HNO-, SO., NH, and particulate nitrate
during a 10-day study. The HN03 and NO- measurement! were compared to similar
measurements made with a denuder difference measurement.
Replicate air quality data obtained with ADM systems for JSO. and HNO,
demonstrate^ that the reproducibility of the ADM was ± 0.16 ug/m for SO. ana
± 0.05 ug/m for HN03> Since the volume of air sampled by the DDM was 1/15
that of the ADM, we could only reasonably compare results of the ADM and DDM
when the concentrations of HNO. were > ~/ug/m?. Based on this criteria the
DDM and ADM provided data for HNO, which was in reasonable agreement when the
HN03 concentrations were above 1 ug/m and sampling periods were equal to or
greater than 12 hours.
A sample collection system based on annular denuders followed by
particulate filters..appears to be a very promising system for measurement of
HN00, N00, S00, SO,,, NH,,, NH. and H . The svstem'.s main features are! (a)
3, 2, 4, NH3, NH. and H . The sy s tern !s main features are:
operation at a relatively high flow rate while maintaining a collection
3,
ra
59
-------
efficiency greater than 95%; (b) use of denuder coatings which are extractable
in water and compatible with conventional 1C analysis; (c) single flow train
and corresponding reduction in the number (and cost) of flow control devices;
(d) relatively easy to set up and operate; (e) all gases of interest are
removed from the sample stream by denuders prior to passage of the sample
stream through any filter medium. In this configuration the acidification or
neutralization of particles on filters is minimized.
60
-------
1.
2.
3.
4.
5.
6.
7.
8.
References
nq«R"MW" Stev?ns' R- K-> Bowermaster, J. Tesch, J., and Tew E.
(1982), ..Measurements of Atmospheric Nitrate and Nitric Acid; The Denuder
Difference Experiment,!! Atmos. Environ. 16, 845-853.
Possanzini, M., Febo, A., and Liberti, A. (1983) !!New Design of
High-Performance Denuder for the Sampling of Atmospheric Pollutants "
Atmos. Environ. 17, 2605-2610.
Gormley, P., and Kennedy, M. (1949) "Diffusion from a Stream Flowing
Through a Cylindrical Tube!! Proc. R. IR. ACAD 52A 153-169.
Ferm, M. (1979) !!Method for the Determination of Atmospheric Ammonia,!'
Atmos. Environ. 13, 1385-1393.
Sjodin A. and Ferm, M. (1985) '(Measurements of Nitrous Acid in an Urban
Area,!! Atmos. Environ., in press.
Ferm, M. and Sjodin, A. (1985) !!A Sodium Carbonate Coated Denuder for
Determination of Nitrous Acid in the Atmosphere,'! Atmos. Environ., in
|J T c o 5 •
Stevens, R.K., Dzubay, T.G., Russwurm, G.M. and Rickel, D. (1978)
Campling and Analysis of Atmospheric Sulfate and Related Species "
Atmos. Environ. 12, 55-68.
Peake, E and Stevens, R.K., '!Characterization of the Kananaskis Annular
Denuder System (KAPS) to Measure Air Pollutants,!! In Preparation (1985).
61
-------
TABLE 1
Typical Ambient Concentrations of Chemical Species Related to
Acidic Dry Deposition
Gases
Na, As, Se, Pb)
Organics (organic acids,
agricultural pesticides,
herbicides)
Concentration
so2
HN03
N02
HC1
HN02
Mil *\
3
3
Aerosols
S04
N03
H+
Trace Metals (elements, e.g.,
0.2-30
0.1 - 15
0.2-30
0.1 - 10
0.2- 2
0.1-20
30 - 150
Concentration
(ug/m?)
0.2-40
0.1 - 10
.02-10
.001 - 30
.001 - 30
62
-------
Table 2
COMPARISON OF ADM AND DDM SYSTEMS
ADM
DDM
Denuders:
Annular Space
Coating:
Flow Rate:
Filters:
Species Measured:
Analysis:
381 x 38 mm
1.5 mm
Na2C03: glycerine
15 LPM
47 mm 2 pm Teflon
47 mm 1 pm Nylon
HN03, S02, NQ~, SO"
NH3, (HN02), (H+)
1C, Colorimetry
470 x 12 mm
Not Applicable
MgO
2 LPM
25 mm 1 |jm Nylon
47 mm 2 |jm Teflon
47 mm 1 jjm Nylon
47 mm glass fiber
treated with TEA
HN03, N0~, SO^,
S02, N02
1C
63
-------
TABLE 3
ITALY - USA FIELD STUDY, RTP, NC — SPRING 1985
REPLICATE ADM TEST
so2
MEASUREMENT, ng/m«
DATE
3/27 - Day
3/27 - Night
3/29 - Day
3/29 - Night
3/30 - Day
3/30 - Night
3/31 - Day
3/31 - Night
4/1 - Day
4/1 - Night
Average
DATE
3/27 - Day
3/27 - Night
3/29 - Day
3/29 - Night
3/30 - Day
3/30 - Night
3/31 - Day
3/31 - Night
4/1 - Day
4/1 - Night
Average
1
6.44
2.84
4.69
5.00
3.87
5.01
1.11
2.61
2.45
3.90
1
1.55
0.55
1.57
0.71
1.02
0.66
1.26
0.34
0.58
0.16
2
6.44
3.13
4.49
4.40
3.68
4.56
1.43
2.30
2.85
3.42
HN03
MEASUREMENT, M<
2
1.47
0.63
1.55
0.62
1.00
0.63
1.36
0.35
0.52
0.14
%
0.0
4.8
2.2
6.4
2.5
4.7
12.6
6.3
7.5
6.5
5.4±3.7
3/m?
%
2.7
6.8
0.6
6.8
1.0
2.3
3.8
1.4
5.4
6.7
3.8±2.5
Average
6.44
2.98
4.59
4.70
3.77
4.78
1.27
2.45
2.65
.3.66
Average
1.51
0.59
1.56
0.66
1.01
0.64
1.31
0.35
0.55
0.15
64
-------
Table 4
RESULTS OF DDM VS. ADM FOR AMBIENT CONCENTRATIONS OF HNO,
DATE
DDM
ADM
3/28
3/29
4/1
4/2
3/29
3/30
4/1
12 Hour Experiment
0.78
1.31
0.29
0.84
24 Hour Experiment
1.54
0.96
0.85
0.78
1.11
0.35
0.77
1.56
1.01
0.55
65
-------
GASES AND
PARTICLES
GASES DEPOSITED ON WALLS
STRIPPED GASES
AND PARTICLES
"STAND-OFF
Figure i ANNULARDENUDER
66
-------
HN03,HCI>S02,HONO-
PNEUMATIC FLOW CONTROLLER
-NYLON FILTER
"^ TEFLON FILTER
PUMP
Na2C03- GLYCERINE
COATING
TOTAL FLOW
ADJUSTER
TEFLON CYCLONE
151/min
Figure 2. Main components for annular denuder method (ADM): a filter pack
system used to collect HNOg, HC1, SC>2, N0~, MONO, S0~, and H+. '
67
-------
co I-
1
s
n
4-1
0)
0
0)
o
J-l
0)
19
§
Q)
•H
U
rt
CJ
•H
•H P
(-
0)
!
68
-------
Figure 4
DIURNAL VARIATION IN NITRATE CONCENTRATION, RALEIGH, NC, 1985
NITRATE
D =DAY
N =NIGHT
DATE (1985)
69
-------
-PJ
CO
CM
I
z z
X X
LU
CM
'CM
•o
o
S-
-------
MARCH 27,1985
1035-1605hr,'
15 l/min
SO2 •= 6.
HN03«=1.55pg/m3
HN02 - 0.4 fjg/m3
FIRST Na2CO3 DENUDER
SECOND Na2CO3 DENUDER
Fig. 6. Typical ion chromatogram of parallel annular denuders which
have collected ambient air samples. The second denuder after
the first denuder does not contain measurable amounts of
50^ or NO^ indicating the efficiency of the first denuder.
71
-------
A NEW ANALYSER FOR THE CONTINUOUS AND UNATTENDED MEASUREMENT OF LOW
LEVELS OF SULFUR DIOXIDE IN AMBIENT AIR
G.B. Marshall
Central Electricity Research Laboratories,
Kelvin Avenue, Leatherhead, Surrey KT22 7SE, UK
D.A. Hilton
Severn Science (Instruments) Ltd.,
Thornbury, Bristol BS12 2UL, UK
1. INTRODUCTION
There is currently' a worldwide interest in the measurement of ambient and
natural background levels of sulfur dioxide in air, because this gas is a major
precursor to acid rain. Many analytical methods and instruments are available
for this determination, but all suffer from shortcomings of one sort or another
such as lack of sensitivity, lack of specificity, instrumental and procedural
complications and long response times. This paper describes the performance of
a new sulfur dioxide analyser free from most of these shortcomings and whose
principal advantage is its extreme sensitivity. It is based on the analytical
technique of Mercury Displacement Detection.
The theory of the application of mercury displacement detection to sulfur
dioxide determination has been described previously, (Marshall and Midgley,
1981 , and Marshall and Midgley, 1982). In brief, sulfur dioxide promotes the
disproportionation of mercury(I) to give mercury vapour which is then
determined with great sensitivity by atomic spectroscopy using a simple and
inexpensive mercury vapour detector.
2S02 + 2H20
Hg(S03)22~ + Hg° + 4H+
(1)
The operating principle of the continuous analyser is shown in Figs, la
and Ib. The sample air enters the reaction vessel via the sample inlet and any
sulfur dioxide present reacts with the mercury(I) ion reagent according to
reaction (1). The liberated mercury is swept out of solution by the same flow
of sample air, dried, and drawn into the mercury vapour detector for atomic
spectroscopic measurement. The reagent is pumped first to a conditioning
72
-------
vessel, where sulfur dioxide free air is continuously bubbled through to reduce
the blank, and then to the reaction vessel and finally to waste. The
instrument is calibrated by passing known concentrations of sulfur dioxide in
air from a permeation tube calibrator through the apparatus in place of the
sample air.
2. PERFORMANCE
At the start of the design work a performance specification was set for
the instrument and this is shown in the first column of Table 1. The second
column shows the actual performance of the analyser. In all cases the
specification was met. The definitions of the various parameters are those of
EPA (1979).
The instrument is programmed to operate continuously without operator
attention on a pre-set cycle of zero, calibrate and measure. The frequency of
zero and calibrate checks depends on the accuracy and precision required. When
the instrument was operating on zero for 10 minutes, calibrate for 10 minutes,
and sample for 60 minutes (which is the most frequent zero and calibrate check
envisaged), maximum zero drift was as little as 0.21 ppb between zero checks.
Over a 12 hour period the maximum zero drift (calculated as the maximum change
of continuous output response) was observed as 0.68 ppb, over a 24 hour period
the net change in response was 0.49 ppb. The stability of the instrument to
continuous inputs of sulfur dioxide is demonstrated by the span drift results,
where over two separate periods of 24 hours first 20 ppb S02 and then 80 ppb
S02 were passed continuously to the instrument. The net changes in response to
these levels of S02 over 24 hours were 0.41 ppb and 2.85 ppb respectively.
These results confirm the inherent stability of the instrument particularly
when it is remembered that the span drift results include errors introduced by
the permeation tube calibrators.
The instrument is linear over three orders of magnitude, i.e. from its
limit of detection to 100 ppb. Should higher levels of sulfur dioxide than
this be required to be measured the incoming sample should be diluted by an
appropriate amount.
The instrument proved to be of high precision - at 20 ppb and 80 ppb
sulfur dioxide relative standard deviations of 0.66% and 0.76% were calculated.
This was confirmed by a low level precision test where relative standard
deviations of 1.5% and 1.3% were recorded for 7.2 and 11.0 ppb sulfur
73
-------
dioxide. Again it should be emphasized that these values include errors
introduced by variability in the permeation tube calibrators.
The instrument responds rapidly to changes in sulfur dioxide
concentration, i.e. 1.5 minutes to 95% of final concentration. More rapid
response times for special applications can be obtained by increasing the air
flow through the analyser (at the expense of a reduced concentration range) or
by increasing the temperature of the reagents.
The length of time the analyser is able to run without attention is
determined by the volume of solution in the reagent reservoir. A 20 litre
supply enables continuous running for 22 days. The maximum time of unattended
operation is determined by the reliability of the peristaltic pump tubes.
3. DISCUSSION
The principal advantage of the instrument is its extreme sensitivity and
consequent ability to analyse true background levels. This is demonstrated by
its lower detectable limit of 0.12 ppb. Sulfur dioxide analysers based on
flame photometry or molecular fluorescence do not have this inherent
sensitivity.
The instrument has good selectivity since it does not respond to other
species present in the atmosphere (Marshall and Midgley, 1981). Of special
importance in speciation studies is its lack of response to sulphate since it
is advantageous to distinguish sulfur dioxide from particulate sulphate.
The instrument is safe to use - requiring no flammable gases and there is
no hazard from the dilute Hg(I) solution used or from the minute quantities of
mercury vapour released by the reaction.
If modified in a minor way the instrument is able to measure sulfite ion
continuously in aqueous solution. This is achieved by replacing the sample air
with a sulfur dioxide free carrier gas and feeding the sulfite solution
directly to the reagent in the reaction vessel through an additional entrance
port. This application could be extremely useful in environmental studies
where speciation is required between sulfite ion and sulfate ion in rain water,
cloud water etc.
The instrument is available commercially, being made under CEGB licence.
4. ACKNOWLEDGEMENT
This work is published by permission of the Central Electricity Generating
Board.
74
-------
TABLE 1; INSTRUMENT PERFORMANCE CHARACTERISTICS
.rerrormance parameter
Specification
Actual^
Performance
Unattended Operation
Lower Range
Lower Detectable Limit
Noise (1) Base Line
(2) 80% Upper Range
Limit (URL)
Zero Drift
Span Drift (1) 20% URL
(2) 80% URL
Response Time (1) Lag Time
(2) Rise Time
(3) Fall Time
Precision (1) 20% URL
(2) 80% URL
10 days
0 to 100 ppb
0.2 ppb
0.1 ppb
1% of concentration (0.8 ppb)
0.2 ppb between zero checks
5% of concentration (1.0 ppb)
5% of concentration (4.0 ppb)
15 s
1.5 min
2 min
5% of concentration
5% of concentration
0.12 ppb
0.06 ppb
0.30 ppb
0.21 ppb
0.41 ppb
2.85 ppb
8 s
1.5 min
2 min
0.66% r.s.d.
0.76% r.s.d.
75
-------
5. REFERENCES
1. Marshall, G.B., and Mldgley, D., 1981, Mercury Displacement Detection for
the Determination of Picogram Amounts of Sulfite Ion or Sulfur Dioxide by
Atomic Spectrometry, Analytical Chemistry, 53, 1760
2. Marshall, G.B. and Midgley, D., 1982, Continuous Atomic Spectrometric
Measurement of Ambient Levels of Sulfur Dioxide in Air by Mercury
Displacement Detection, Analytical Chemistry, 54, 1490
3. EPA, 1979, "Summary of Performance Test Results and Comparative Data for
Designated Equivalent Methods for S02; Research Triangle Park, NC,
Document No. QAD/M-79.12
76
-------
PUMP
HEAD •»- '1
PUMP
HEAD
SOLUTION IN
REACTION
VESSEL
1
PINCH
VALVE
PUMP
HEAD 2
WASTE
PUMP
HEAD 1 cv''"
/" — N x
\
^^
/
^x /
/ 1
1 •
CONDITIONING
VESSEL
FIG. la SOLUTION FLOW
77
-------
NEEDLE
VALVES FV1
MERCURY
MONITOR
—FRIT
CONDITIONING
VESSEL
r\^
r"
v
v_
1
>U ^
__^ /
/
^
SV 2 SV 2
SV 1
FILTER
•*—i XJ
SAMPLE
ZERO
AIR
CALIBRATE
'GAS
FIG. ib GAS FLOW
-------
ALKALINITY MEASUREMENTS IN EVALUATING
HI-VOL FILTER PERFORMANCE
Rita M. Harrell and John C. Holland
Northrop Services, Inc.
Environmental Sciences
P.O. Box 12313
Research Triangle Park, NC 27709
Alkalinity measurement is a useful tool in the performance evaluation of
hi-volume air filters. It is significant because alkaline sites react with
S02 and HN03 in the air forming sulfate and nitrate on the filters. Such
sulfates and nitrates are indistinguishable from particulate sulfate and ni-
trate collected by the filters. In turn, the total particulate weight, sul-
fate content, and nitrate content determined would be falsely high in propor-
tion to filter alkalinity.
Alkalinity of older filters are measured in order to evaluate appropriate
correction factors for previously obtained particulate data. New filters are
tested for compliance with EPA specifications and determination of correction
factors for the current air pollution data base.
Among the two or three laboratories, across the country, that have been
involved in alkalinity measurement, a reproducibility problem has been observed
using the standard method ASTM D202 . In part, this may be due to slight dif-
ferences in each laboratory's interpretation of ASTM D202 but the major differ-
ences observed can be attributed to factors which influence alkalinity values
obtained. Because of this problem, a study was initiated to investigate factors
having the potential of influencing observed filter alkalinity values and to
2
develop a simple, reproducible method for alkalinity measurement. Bruce Appel
79
-------
has also done extensive work in this area.
From the study it was found that many factors such as extraction volume,
extraction time, stirring method and filtration temperature influence alkali-
nity measurement. Extraction time in conjunction with extraction volume was
found to produce the most significant change in alkalinity. Magnetic stirring
with a 2" stirrer bar versus high speed mechanical agitation produced the least
significant change.
On the basis of the overall results, of the study, a modified alkalinity
measurement procedure will be recommended upon completion of interlaboratory
and intralaboratory testing of the proposed method. At present, within our
laboratory, agreement between two different analysts is good for alkalinity
evaluation of the same filters.
1,,
Standard Methods of Sampling and Testing Untreated Paper Used for Electrical
Insulation", ASTM.D202-77, in Annual Book of ASTM Standards. 39:62 (1977).
B. R. Appel, E. L. Kothny, V. Ling, and J. J. Wesolowski, "Sampling and Analy-
tical Problems in Air Pollution Monitoring Phase II", First Quarterly Progress
Report, Air and Industrial Hygiene Laboratory, California Department of Health
Services, 2151 Berkeley Way, Berkeley, California 94704-9908, EPA Cooperative
Agreement No. CR 810798-01-0, November 1983.
80
-------
A PROCEDURE FOR THE ISOLATION AND CONCENTRATION OF
MILIGRAM QUANTITIES OF CARBON MONOXIDE FROM AMBIENT AIR
by
D. A. Levaggi, R. E. England And R. V. Zerrudo
Bay Area Air Quality Management District
In evaluating the non-attainrnent status for carbon monoxide (CO)
in the San Francisco Bay Area, unexpectedly high levels in low-traffic
residential areas suggested the possibility that residential wood com-
bustion (RWC) could be contributing to the observed excesses. In this
context a C analysis of the ambient CO would indicate any CO derived
from RWC, as distinct from CO derived from fossil fuel combustion in
vehicles or buildings (1).
In order to isolate the milligram levels of CO needed for the llfC
analysis, large volumes of air, in the order 2000£, must be extracted (2).
The extracted CO must also be completely devoid of any carbon contamination
i.e., from carbon dioxide, particulate, or gaseous hydrocarbons. This
short paper will detail some of the experimental findings in the develop-
ment of a sampler to accomodate the requirements stated above. As well
some field data will also be presented.
The sampling train configuration is shown in Figure 1. Four major
components perform the necessary tasks of the sampler, namely; 1) hydro-
carbon removal, 2) carbon dioxide (C02) removal, 3) carbon monoxide (CO)
oxidation and 4) C02 (formed from CO) absorption. These will be taken up
separately, with appropriate experimental findings made for each. All the
experimental tests were performed at room temperature and a flow rate of
5£/min. This relatively high flow rate was necessary to obtain milligram
levels of CO during a sampling period of eight hours. The eight hour
sampling period was chosen to accommodate the National Ambient Air Standard
for CO (9 ppm 8-hour average).
81
-------
HYDROCARBON REMOVAL
It is well known that activated charcoal is an excellent absorber of
organic gases from air streams. This removal efficiency increases with
increasing molecular weight of the gas, e.g. methane and ethane at 5£/min.
are not absorbed and retained for any length of time. The concern, however,
was the absorption and retention of Cs and higher hydrocarbons, which had
the possibility of being oxidized by the Hopcalite located downstream in the
sampler. The Hopcalite oxidizing potential will be taken up separately
later in the text.
Experiments were performed using blends of 1.2 ppm propane and 1.0 ppm
butane in air with background methane, and 12 mesh activated coconut
charcoal. The charcoal was packed in a 120cc plastic cannister. The only
pretreatment of the charcoal was a purge using cylinder nitrogen at 5S,/nrin..
for one hour. The experiments were monitored using a Bendix Model 8201
GC-FID instrument.
Results of the experiments were not surprising and indicated the following:
• Propane was found to initially pass thru the charcoal after 1.5 hours.
At the end of 4.5 hours approximately 40% of the propane was no longer
being retained. The slope of the instrument recorder indicated that
after an additional 1.5 hours all the propane would have been passing
thru the charcoal.
« Butane was still 100% retained after 16 hours.
• After field sampling in a highway tunnel with above ambient level
hydrocarbons for 40 hours, a 120cc charge of charcoal was still
removing 100% of all the 64+ hydrocarbons.
CARBON DIOXIDE REMOVAL
Ambient C02> because of its high llfC content, must be removed from
a sample stream by a minimum of 99.95%. Soda lime of 8-12 mesh is the
primary remover of C02 in the sampling train. Indicating soda lime is
preferred only because visual inspection immediately reveals the extent
of its consumption. The soda lime is packed in two 120cc plastic
cannisters placed in series, after the charcoal and prior to the ascarite-
drierite filter. The ascarite (approximetely 60cc of material) is
merely a back-up system should the soda lime cannisters become exhausted
82
-------
prior to a completed sampling run. The C02 removal system was found to
easily accomplish the 99.95%+C02 removal criterion. Most experimental
runs were made with approximately 20 ppm C02 gas streams, derived either
from 20 ppm C0£ certified cylinders or by precise dilution of a stock 2000
ppm C02 cylinder with zero cylinder air (no C02). The C02 content of all
cylinders, generated C02 air streams, and the efficiency studies of the
C02 removal portion of the sampling train were determined with a TECO
Model 626. This NDIR instrument is multiranged; the sensitivity range
of 0-10 ppm full scale was used routinely in the removal efficiency tests.
Instrument performance was exceptional and it was felt that a lower
detectable limit of 0.07 ppm C02 could have easily been noted. The C02
removal system was found to have the following characteristics:
• An eight hour ambient sample run at 5£/min consumes approximate 75%
of the sampling train soda lime. The two cannisters are, therefore,
replenished after each 8 hour sample.
• Better than 99.95% of the incoming C02 is removed by the soda lime.
• Extensive field tests have shown that the ascarite containing
cannister need never be replaced.
CARBON MONOXIDE OXIDATION
In order to collect the required milligram levels of CO from 2000
liters of air it is necessary to convert the CO to a collectible specie.
Oxidation to C02 is the easiest and most conventient conversion for such
a large quantity of air. Hopcalite was chosen as a likely candidate to
catalize the oxidation at room temperature. This patented catalyst
available from the Mine Safety Appliance Corp, has been used for many
years to convert CO to C02 at low flow rates, and also to oxidize
hydrocarbons at elevated temperatures (3, 4). Experiments then had to
be performed to 1) determine if at room temperature any oxidation of
Ci-C3 took place (since the charcoal located upstream in the sampling
train would remove C4 and higher HC's and 2) to establish the oxidation
efficiency of the Hopcalite for CO at an elevated flow rate, i.e. 5£/min.
83
-------
The need for a selective oxidation of CO is that most all airborne
hydrocarbon gases originate from fossil fuel combustion. These gases
contain no 1IfC and would, therefore, bias any CO-llfC analysis. This
bias is exactly the opposite of that which ambient C02 would cause.
A 3.0 ppm propane, 3.1 ppm methane standard in air was used to
evaluate its oxidation by a bed of approximately 50cc of Hopcalite. The
inlet and outlet of the Hopcalite were monitored with the Bendix 8201.
Absolutely no difference could be detected in the hydrocarbon mixture
upon its passage thru the catalyst.
In another experiment a NBS traceable CO gas mixture (10.1 ppm in
air) was passed thru the Hopcalite. The exiting gas was monitored by a
Bendix Model 8501 NDIR for CO and the TECO Model 626 for C02- CO was
never detected and the C02 concentration was found to be 9.8 ppm,
indicating a quantitative conversion of the CO to C02.
Experimentation and field studies indicate that:
• Hopcalite (a 50cc charge) is capable of the quantitative conversion
of CO in air to C02» at room temperature and 5£/min.
• Hopcalite does not cause any perceptible oxidation of a Cj + C$
gas mixture in air. Based on instrument detectability we estimate
less than 1% oxidation, and feel confident that it is even less.
• Field sampling indicates that after twenty 8 hour runs the Hopca-
lite is still performing at its full CO > C02 efficiency.
• The Hopcalite should be dried at lOOoC for two hours prior to use
and desiccated. If kept dry it performs as expected.
CO COLLECTED AS C02
The final element of the sampling train is an absorber to collect
the C02 which was selectively oxidized to C02 by the Hopcalite. The
absorber contains sodium hydroxide (NaOH) which has been evenly distri-
buted on 16-18 mesh firebrick. It was determined that a 3" x 1" drying
tube packed with the treated fire brick was an effective collector of C02-
84
-------
The absorbers must be "clean" of carbonate carbon and are prepared
in a nitrogen glove box. The sodium hydroxide used to impregnate the
16-18 mesh firebrick must be carbonate free. We have used a J. T. Baker
"carbonate free" product which contains 5 eq. of NaOH in about 460 ml
of water in a sealed plastic bottle. The product, however, was not
found to be carbonate free, and requires a pretreatment with barium
chloride to render it acceptable (5). The recipe for preparation of the
absorber tubes is 60 gm of firebrick to 28cc of the 5 eq solution.
Details of the preparation are available on request. Ten tubes can be
made from the aforementioned recipe. If prepared as described the 3"
packed tubes were found to contain an acceptable 0.2 mg of carbonate
carbon. To prevent contamination from ambient C02, the tubes, after
preparation, prior to and after field sampling,are always stored in a
sealed container containing liberal quantities of soda lime.
Extensive testing of prepared tubes revealed that the collection
efficiency was strongly influenced by storage time. The collection
efficiency is satisfactory (basically 100%) up to four days after pre-
paration. This is not considered to be terribly disadvantageous since
a batch of absorption tubes can be easily prepared in less than 1.5 hours,
MISCELLANEOUS
The Drierite located in various positions in the sampling train
serves as a source of moisture removal which preserves the effectiveness
of the Ascarite and Hopcalite catalyst.
Two 47mm glass fiber filters are positioned to keep the Hopcalite
clean, and also remove any fines from the Hopcalite reaching the C02
absorber.
A detailed checkout protocol has been established for the sampling
train prior to its field use. The procedure calls for a visual inspec-
tion of the chemicals used, a leak check, and verification of an active
Hopcalite catalyst. The entire pre-sampling procedure takes about
20 minutes.
85
-------
SUMMARY
A sampling train which will selectively separate milligram levels
of ambient CO from air has been developed. It is relatively simple
and inexpensive to fabricate and has been thoroughly laboratory and
field tested. It should prove valuable in resolving the origin of
elevated CO ambient concentrations in areas with a concomitant large
percentage of homes utilizing RWC. The procedure also appears well
suited to augment and/or confirm receptor modelling studies in many
areas such as Vail, Missoula and Portland where large scale RWC has
caused intense winter air pollution problems (6).
REFERENCES
1. Cooper, J. A., Currie, L. A. and Klouda, G. H.: "Assessment of
Contemporary Carbon Combustion Source Contributions to Urban Air
Particulate Levels Using Carbon-14 Measurements", Env. Sci. Techn.
15, p. 1045, (1981).
2. Jull, A. T., Donahue, D. J. and Zabel T. H.: "Target Preparation
for Radiocarbon Dating by Tandem Accelerator Mass Spectrometry",
Nuclear Instruments and Methods in Physics Research 218 p. 509, (1983),
3. Lindsley, C. H., Yoe, J. H.: "Acidimetric Method for Deter-
mination of Carbon Monoxide In Air", Anal. Chem. 2^ p. 513 (1949).
4. McDonough, J. P.: "Detection, Reduction and Control of Gaseous
Contaminants", Mine Safety Appliances, Filter Products Div. (1978).
5. "Scotts Standard Methods of Analysis", by W. W. Scott, Fifth Ed.
1939, Edited by N. H. Furman, p. 2197, D. Van Nostrand Co., Inc.
Publishers.
6. "Proceedings 1981 International Conference on Residential Solid
Fuels, Environmental Impacts and Solutions", Edited by J. A. Cooper
and D. Malek; Oregon Graduate Center, Beaverton Oregon, Publishers.
86
-------
CL.
oo
CxJ
87
-------
Analysis of the Effluents from Polyurethane Foam Insulation
M. E. Krzymien
National Research Council, National Aeronautical Establishment
Ottawa, Ontario, Canada
Introduction
In spite of a common belief that polyurethanes are chemically
stable and do not degrade in temperatures lower than 100-200°C, there is a
growing concern that they may emit decomposition products upon heating at
temperatures as low as 80°C. In this paper the volatile degradation products
of MDI based, rigid polyurethane foam were analyzed and a number of them
were tentatively identified.
Experimental
A sample (about 2g) of rigid polyurethane foam manufactured from
PMDI, sorbitol and polyethylene glycol with the addition of Syrol 6 (diecthyl
aminoethyl phosphate) as a fire retardant and Freon 11 as a blowing agent, was
placed in an aluminum tube and maintained at 80'C in either nitrogen or Zero
Air atmosphere. These gases were also used as effluent carriers.
The off-gas samples were collected by passing the carrier gas
leaving the aluminum tube through an ice coaled adsorber. The adsorber was a
glass tube (75 mm x 63 mm O.D.) containing a 2-cm column of Tenax TA 35/60
mesh. For capillary GC analysis the samples were thermally desorbed in a
modified (1) injection port. The analysis was carried out on HP 5790A gas
chromatograph coupled with HP5970A Mass Selective Detector (MSD). The
GC was operated with DB-5 30 m x 0.32 mm I.D. fused silica column.
88
-------
The iMass spectra produced by MSD were identified using the Mass
Spectral Identification, Probabiiity Based Matching System (PBM) developed at
Cornell University and Cornell's IBM VM 370 computer.
The mass spectral identification of the off-gases was supplemented
by dual-detection gas chromatographic analysis. The analysis was performed
on Varian 4600 gas chromatograph equipped with flame ionization (FID),
electron capture (ECD) and thermoionic specific (TSD) detectors. The
detectors were operated either individually or in pairs. The Varian GC was
fitted with 30 m x 0.32 mm I.D. SPB-5 fused silica column.
Results and Discussions
Before analysis, the polyurethane foam sample was stripped of
adsorbed contaminants by heating it in a stream of high purity carrier for a
few days. The decontaminated sample was then heated in either high purity
nitrogen or Zero Air-atmosphere without flow for several days to allow the
concentration of the genuine off-gases to reach a measurable level. The
effluents from such "recharged" foam are decomposition products of the foam
and/or additives used in the production process such as catalysts, blowing
agent, residual unreacted monomers, fire retardant, etc. Their concentrations
also decrease with time at a rate depending on a carrier flow rate and their
emission rates. The dependence might be used to calculate the emission rates.
It was observed that, whereas composition of the off-gas mixture
did not change drastically with a change of carrier gas from nitrogen to zero
air, the emission rates were significantly higher in a Zero Air atmosphere.
89
-------
The individual components of the mixture were identified by
GC/MSD analysis. The obtained mass spectra were compared with the
reference spectra listed in spectral library. Table I lists the compounds
identified by the PBM system in the PUF effluents separated on an HP 5790
gas chromatograph. Columns 6 and 7 of the table list class 4 (CL 4) and class I
(CL 1) reliability of a match expressed in percentages. Class 4 reliability
means the probability that the unknown compound has a structure closely
related to that of the retrieved compound. Class 1 is the probability that the
unknown has an identical structure (or stereoisomer) to that of the retrieved
compound.
The dual-channel, dual-detector gas chromatographic analysis
provided an additional qualitative information by identifying compounds as
members of a group of compounds sensitive to a particular detector. For
example an unknown was identified as a hydrocarbon, halocarbon, amine,
nitro-compound etc.
As a result of this work over twenty compounds were tentatively
identified as probable polyurethane foam off-gases. Their identity will be
finally confirmed by comparison of their retention data and mass spectra with
those of standard compounds.
Reference
1) M. E. Krzymien, Dual Adsorber-Capillary Column System for Gas
Chromatographie Analysis of Air Samples. NRC, NAE-AN-20, 1983.
90
-------
TABLE 1
PEAKS TENTATIVELY IDENTIFIED BY PROBABILITY BASED MATCHING SYSTEM (PBM)
Spectrum
No.
31
32
33
34
35
36
37
38
39
40
41
44
45
46
47
48
49
50
51
53
54
55
56
67
68
69
70
71
75
Name
Cyclohexanone
3-Methylene pentane
Unknown
Heptanol
Cyclohexanamine (Polycat 8)
2-Phenyl propene
Octamethyl cyclotetrasiloxane (?)
Trimethyl benzene
1-Hexene
Unknown (hydrocarbon or alcohol)
Unknown
1-Octanol (?)
Phenyl methyl ketone
n-Nonyl aldehyde
2-Pentenal
4-Pentenal
1-Heptene
Butyl isocyanate (?)
Nonyl alcohol
Unknown (aldehyde or ketone)
Dodecanol (?)
Methyl methylbenzoate
n-Octanol (?)
Trichlorofluoromethane (freon 11)
Trichlorofluoromethane (freon 11)
Unknown
Methyl ethyl ketone (?)
Unknown
Cyclohexene (tetrahydrobenzene)
Atmosphere
air, N2
air, N2
air
air, N2
air, N2
air, N2
air
air
air
air
air
air
air, N2
air, N2
air
air
air, N2
air
air, N2
air
air
air
air
air, N2
air, N2
air, N2
air, N2
air, N2
air, N2
R.T.
12.50
13.12
13.66
13.84
14.04
14.12
14.22
14.34
14.42
14.86
15.04
15.58
15.62
16.16
16.34
16.54
16.60
17.16
17.24
17.62
17.84
18.10
18.22
5.66
5.70
6.76
6.94
7.04
8.00
R.A.
%
29.5
15.6
2.0
34.8
75.1
31.1
2.9
5.2
2.9
5.1
2.1
3.4
13.2
14.3
5.8
7.8
26.5
3.8
49.7
8.8
3.9
7.9
3.2
95.5
100.0
5.9
12.6
18.0
12.9
Reliability
CL4 CL1
78
52
74
55
95
33
80
60
39
66
77
53
60
66
35
76
57
60
50
41
41
43
92
47
22
40
24
69
• 13
50
27
17
32
47
23
27
32
14
44
25
27
21
17
17
18
61
91
-------
The following paper was presented at the U.S. E.P.A. Fifth Annual National
Symposium on Recent Advances in the Measurement of Air Pollutants held in April,
1985, Raleigh, North Carolina.
Richard T. WcNerney, Jerome Instrument Corporation's Sales & Marketing Department
QUANTITATIVE ANALYSIS OF PPB CONCENTRATIONS OF HYDROGEN SULFIDE IN AIR
USING A GOLD FILM SENSOR
In 1972, a new sensor was developed for the detection of nanogram quantities
of elemental mercury. This sensor, which received a U.S. patent in 1976, is a
thin gold film prepared by high vacuum evaporation on a ceramic substrate.
Previous to the development of the thin gold film, instruments for the
determination of mercury relied mainly on the atomic absorption (AA) technique.
However, the AA technique has serious limitations in terms of selectivity and
sensitivity to mercury.
Instruments using the Gold Film technique for mercury detection overcame
these limitations and now enjoy a worldwide acceptance for industrial hygiene
surveys, environmental studies, and precious metal and geothermal prospecting.
Several articles detailing the gold film and the instruments which utilize this
sensor have appeared in Science, Analytical Chemistry. American Laboratory, and
several geologic journals. Reprints of these papers are available from Jerome
Instrument Corporation upon request.
In the 1972 Science article entitled "Mercury Detection by Means of Thin
Gold Films," it was stated "...in exploratory experiments we have investigated
the applicability of measuring other gases. The gold film is at least as
sensitive to H2S (hydrogen sulfide) as it is to Hg, and so it can potentially
serve as an H2S detector."
In the development of the Gold Film Mercury Analyzers, an intake filter
consisting of mallcosorb was necessary to eliminate ambient levels of hydrogen
sulfide. And, it was the Gold Film sensor's response to these very low hydrogen
sulfide concentrations that prompted our R & D effort, beginning in 1982, and
culminating in April of this year; this effort led to the development of an
instrument capable of the part per billion determination of H2S in air.
The sensor detects hydrogen sulfide by adsorbing HLS on its surface. This
results in an increase in electrical resistance of the gold film; this resistance
change is measured by incorporating a sensor and a reference film in a Wheatstone
bridge circuit. The mechanism for the resistance change on the thin gold film
is thought to be an impedance of conduction electrons due to the adsorption of
hydrogen sulfide on the surface.
The reaction that takes place on the gold film in the presence of hydrogen
sulfide is given by the equation: Au + H?S > AuS + HpT*. It- is the formation
of this gold sulfide layer on the surface that results in the change in electrical
resistance of the film. The reaction is reversible and in practice the sensor is
regenerated periodically by means of a thin film heater that is sandwiched between
the sensor and reference. The thermal regeneration of the sensor is given by the
equation: AuS + 02 ^°^ Au + S02t. Studies have shown that sensor regeneration
in the absence of 6., wm not remove the gold sulfide layer.
92
-------
The first one and one-half years of R & D effort were directed toward the
refinement of the Gold Film sensor. As previously mentioned, the sensor used
in the mercury analyzers was sensitive to H~S, but its response to H0S would
degrade with time. ^ 2
The new sensor underwent several chemical and physical changes - changes
which cannot be addressed here because of their proprietary nature. This sensor
proved to be even more sensitive to hydrogen sulfide and exhibited excellent
stability. The sensor was also reduced in physical size to that of a larqe
postage stamp.
Since the Gold Film sensor accurately measures picogram quantities of
hydrogen sulfide, a small volume of sample (typically less than 1cc) is required
for ppb level analysis. Therefore, the next phase of the R & D effort was the
development of a sampling system that would accurately deliver a minute sample
volume of gas, yet be small and rugged enough to be contained in a portable
instrument. Also, since the instrument can be operated as a continuous, fixed
point monitor, the sensor regeneration cycle necessitated the delivery of "zero
air to the sensor during this cycle, especially when operating in an atmosphere
containing H^S.
The sample delivery system developed consists of the following major
components: dust filter, solenoid, zero air filter, and pump. The internal
pump draws the sample through the zero air filter. This filter, made of iodine
impregnated charcoal, removes any hydrogen sulfide from the sample. At a pre-
determined time, the solenoid is activated, diverting a sample directly to the
bold Film sensor. The H2$ in this sample is adsorbed and integrated by the Gold
hum sensor. The solenoid closes, resuming the sample flow through the zero air
rliter. The volume of actual sample delivered to the sensor can be varied from
1/4 cc to 25 cc.
GAS FLOW SCHEMATIC
O«MFHi»l
93
-------
There is a finite number of adsorption sites on the surface of the Gold
Film sensor. As sulfur atoms use up these sites, the response will gradually
decay due to a decrease in adsorption efficiency by the sensor. The final phase
of the R & D effort was the development of microprocessor based electronics with
support software in order to compensate for this cumulative loss of sensitivity
(non-linearity).
This electronic package handles all of the high tolerance timing and
sequencing functions demanded by the sample delivery system. The microprocessor
automatically compensates for the non-linearity of the sensor as well as the
individual characteristics of each Gold Film sensor. Also designed into the
system were non-volatile memory devices and supporting software programs that
would allow the operator to (1) automatically recalibrate the instrument with
no internal adjustments, (2) change the timing of the sample sequencing system,
and (3) operate the instrument as a fixed point, continuous monitor.
The automated calibration software is a menu-driven program that allows the
operator to calibrate the instrument precisely to a calibration gas. In this
program, the operator has the option of resetting two timing functions, called
sample wait times, in the sample delivery system. The "1st Sample Wait Time" is
the 'Pump on - sample through the zero air filter' sequence. This can be adjusted
from 2 to 10 seconds in order to allow for the delivery of a sample from a
remote location or for injection of a sample by syringe - a typical procedure for
headspace analysis. For most applications, however, the optimum time is 5
seconds. The "2nd Sample Wait Time" is the 'Solenoid open - close1 sequence; a
sample is delivered to the Gold Film sensor, bypassing the zero air filter. This
timing function can be varied from 0.1 to 9.9 seconds. It is this function that
allows the operator to adjust the sample volume to the expected hydrogen sulfide
concentration in order to maximize precision and accuracy. Studies have shown
that a 0.5 second solenoid open time gives excellent reproducibility for H2S
concentrations from 10 to 500 ppb, whereas a 5 second solenoid open time is optimum
for expected H?S concentrations less than 10 ppb. In this way, the Gold Film
sensor adsorbs the same mass of hydrogen sulfide from a sample; thus, the
electronic signal produced would be similar whether one is sampling low or high
ppb concentrations of ^S.
Once the automated calibration sequence begins, the following events occur:
(1) Sensor Regeneration ensures a gas free sensor at the beginning of calibration;
the solenoid is not activated during this cycle so that only "clean" air passes
over the sensor; (2) Linear Correction establishes a database of multipliers
stored in the instrument's memory to correct for non-linearity; sampling continues
until sensor saturation occurs; (3) Sensor Regeneration thermally desorbs all H2S
accumulated on the sensor during this portion of the calibration program; and,
(4) Comparison Measurement compares the values measured by the instrument to the
calibration gas. The correction factors are applied to maintain linearity.
94
-------
TABLE 1
cnn-J instrument is calibrated with a Metronics Dynacalibrator using hydroqen
sulfide permeation tubes. In this particular calibration sequence the calibrltion
gas had been adjusted to 50 ppb, shown as the actual value below The CailDratlon
u rniT nt^s,re!P°nse to the gas is shown undeFThJ-current value and its value
of ?he dat'f U?k °nCe the SSOr has ^acHeT-^turation (sample #6?),
^
AUTOMATED CALIBRATION SEQUENCE
Date = 04/24/85 * Time = 14:12:59
Instrument Serial Number: PPBH7
Flow Rate = 150 CC's/minute
Sample Actual Current %Error
1
2
3
4
5
6
50
50
50
50
50
50
51
51
51
50
50
49
+2.0
+2.0
+2.0
+0.0
+0.0
-2.0
31
32
33
34
35
36
37
38
39
40
• * *
59
60
61
62
63
64
65
66
67
68
69
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
50
49
50
50
50
50
50
49
50
50
50
50
50
50
50
50
50
51
49
50
50
49
-2.0
+0.0
+0.0
+0.0
+0.0
+0.0
-2.0
+0.0
+0.0
+0.0
+0.0
+0.0
+0.0
+0.0
+0.0
+0.0
+2.0
-2.0
+0.0
+0.0
-2.0
Maximum value =51
Minimum value =49
Mean value = 50
Average -error % = -0.4
Std. Dev. of Current Values = 0.6
% Relative Std. Dev. of Current Values
= 1.19
95
-------
Single point calibration is sufficient to ensure linear responses through
the range of the Hydrogen Sulfide Analyzer, 1 ppb to 500 ppb. An instrument
calibrated at 50 ppb was operated in hydrogen sulfide gas streams where the
concentration was changed from 275 ppb to 97 ppb to 58 ppb. The results are
shown in Table 2. When the calibration gas was adjusted, it took several
samples for the Hydrogen Sulfide Analyzer's readings to stabilize; this is to
be expected. Note that the response of the Gold Film sensor was not affected
by the other gases (CO, NO, and C02) in the gas cylinder.
TABLE 2
CASE STUDY: GOLD FILM PPB HYDROGEN SULFIDE ANALYZER COMPARISON MEASUREMENT
OF KNOWN CONCENTRATIONS OF HpS
The following data was obtained at a calibration laboratory for a state
environmental monitoring program. The hydrogen sulfide calibration system
included a Scott-Marin gas cylinder with 43ppm h^S. Dilution was controlled
using a Dasibi Mass Flow Controller. Sampling was done by connecting the
PPB directly to the calibration gas. No effects were noted due to any
positive pressure the calibration gas stream may have had. The calibration
gas, in addition to H2S, was made up of the following:
CO - 4990ppm
NO - 50ppm
51ppm
No interference effects resulting from these other gases were noted.
S02 -
SAMPLE
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
CAL GAS
275
97
58
PPB READING
250
261
267
271
270
268
268
269
191
116
100
97
98
98
81
57
56
58
57
96
-------
Regarding other gases, the Gold Film sensor does not respond to aromatic
hydrocarbons or water vapor. The sensor will respond to other reduced sulfurs.
The next set of test results were part of a test conducted by the Denver
Research Institute. The Gold Film sensor was used as the detector for a gas
chromatograph. Equal concentrations of a number of reduced sulfurs were analyzed
with responses shown as compared with hydrogen sulfide.
TABLE 3
SUMMARY FOR ALL REDUCED SULFURS TESTED
(In order to normalize all these values, the ml/ppm was divided into the scale
reading to give a value equal to 1 ppm.)
GAS 1 PPM
H2S - Hydrogen Sulfide
ETSH - Ethyl Mercaptan
TBM - T Butyl Mercaptan
NPM - N Propyl Mercaptan
IPM - Isopropyl Mercaptan
SBM - Secondary Butyl Mercaptan
Methyl Mercaptan
Dimethyl Disulfide
MES - Methyl Ethyl Sulfide
DES - Diethyl Sulfide
DMS - Dimethyl Sulfide
Carbon Disulfide
TMT - Tetrahydro Thiophene
* Prime Odorants
SCALE READING FOR 1 PPM
250
82
85
88
85
70
80
133
1.3
2.9
2.3
1.5
9.4
*
*
*
*
*
* (done with permeation tube)
*
(done with permeation tube)
Research is now underway for the development of selective filters that would
allow for quantification of a particular reduced sulfur group. Such an instrument
could serve as a total reduced sulfur (TRS) analyzer especially as an odorant
detector in the natural gas industry.
The Gold Film Hydrogen Sulfide Analyzer is now commercially available. It
is the only instrument of its kind - portable, battery operated, yet able to
directly determine hydrogen sulfide concentrations from 1 to 500 ppb. The
operation of the instrument is simple: Instrument ON, Press SAMPLE, and in 10
seconds the results in ppb's are displayed on the digital readout.
97
-------
The Gold Film Hydrogen Sulfide Analyzer may be operated as a fixed point,
continuous monitor when interfaced with a desktop computer and Jerome Instrument
Corporation's Single Point Continuous Monitoring Program. Sample interval is
operator selectable from every 30 seconds to 1.5 hours. An alarm threshold
value to alert personnel when hydrogen sulfide concentrations exceed a critical
value can be programmed in. Periodic summaries of the data are also operator
selectable.
Jerome Instrument Corporation has recently developed Dilution Modules which
allow the Gold Film Hydrogen Sulfide Analyzer to sample ppm concentrations of H2S.
These small modules are inserted into the intake of the analyzer. Modules are
available to expand the range of this instrument to 500 ppm H2S.
98
-------
THE STATE OF DELAWARE EXPERIENCE
WITH
EPA REFERENCE METHOD 25 AUDIT SAMPLES
CHARLES S, KRICK, P.E.
JOHN W. PERONTI
STATE OF DELAWARE
DEPARTMENT OF NATURAL RESOURCES
AND
ENVIRONMENTAL CONTROL
BACKGROUND
The State of Delaware routinely requires the analysis of audit samples
when stationary sources are tested for gaseous pollutants. EPA liquid
audit samples are required for SO2 and NOX, and EPA cylinder gases are
utilized for non-criteria organic pollutants, as well as non-methane
hydrocarbons.
We have required the use of EPA Reference Method 25 on four occasions.
Two of the tests were at bulk gasoline terminals, and the other two were at
an automobile assembly plant before and after a modification to the paint
line.
BULK GALSOINE TERMINALS
The bulk gasoline terminal tests were performed by Scott Environmental
Technology, Inc., and were accomplished one week apart. Accordingly, one
audit gas- was used for both tests.
The analytical equipment consisted of two infrared hydrocarbon
analyzers: a LIRA Model 300 (0-1.4% propane range) and an Infrared
Industries (0-70% propane range). The instruments were spanned with 1.2%
and 52.4% propane in nitrogen respectively. 100% nitrogen was used to zero
the instrument. Calibration curves were generated prior to the test for
both the LIRA and the high and low ranges of the Infrated Industries
analyzers.
The 9 point calibration curve for the LIRA Model 300 was non-linear.
The five point calibration curve for the low range of the Infrared
Industries analyzer was linear. And the seven point calibration curve for
the high range of the Infrared Industries analyzer was linear.
The test methods used were Method 2A for the direct measurement of the
gas volume. The volume meter had been factory calibrated and was checked
using a standard pitot and wind tunnel. And Method 25B was used to
determine the total gaseous organic concentration as propane.
99
-------
The propane audit sample was obtained from the EPA and was supplied by
the Research Triangle Institute. A sample was taken from the cylinder into
a bag. The bag was then connected to the sample line and analyzed directly
on the LIRA Model 300. The results were 90.5% of scale, which was 1.29%
propane on the calibration curve. The audit gas which was propane in air
had been analyzed in March, 1983, as 1.18% propane and, in May 1984, as
1.3% propane. The initial audit results were within 10% and were
considered to be acceptable.
The results of the tests were considerably different. The weighted
average for one test was 21.3 mg/1, and the other test was 5 mg/1.
In view of the test results and the analytical equipment which was
used, the following conclusions can be reached:
1) The audit gas was analyzed within acceptable
limits (10%).
2) For a non-linear detector, low and high range
audit gases would have been more meaningful.
3) The results from different sources can vary
widely. If the results can be predicted, audit
gas concentrations can be obtained which would cover
the range of concentrations to be expected.
4) The scope of the audit gas program is a function
of the purpose of the testing program. The
purpose of this test program was to establish
compliance with a State Regulation. The audit gas
used was approximately 50% of the average
concentration for compliance. More than 90% of
the concentrations were below the level of the
audit gas. The 1.18% audit gas was appropriate
for a compliance determination.
AUTOMOTIVE ASSEMBLY PLANT
The automotive assembly plant modified its existing topcoat operation
to use the base coat/clear coat coating technology. This modification
required two side-by-side spray booths. Then curing was achieved in the
existing oven. The exhaust air system is vented through 14 stacks on each
spray booth after water scrubbing for particulate removal. The oven
exhaust is vented through 5 stacks. Because of the number of exhaust
stacks, we chose to test 8 spray both stacks and an oven stack.
Method 5 testing for particulates and Method 25 for Volatile Organic
Compounds were conducted. The following analytical equipment was used for
the Method 25 testing:
100
-------
- Byron Model 401 automated G.C. for analysis of
total hydrocarbons,- non-methane hydrocarbons,
methane, carbon monoxide and carbon dioxide
- Byron Model 90 battery powered stack sampler
- Byron Model 75 digestion oven
- Beckman C02 analyzer (NDIR)
- Beckman Hydrocarbon Analyzer
- Servomex oxygen analyzer (paramagnetic)
- Metal bellows sampling pump
All instruments were calibrated with NBS traceable gases. Calibrations
were accomplished before each analyses. The Model 90 and Model 75 each
have digital mass flow meters which were calibrated against each other
daily.
The Method 25 sampling was accomplished using the Byron Model 90 which
pumps the sample through an adsorption tube and into a Tedlar Bag. The
sampling is accomplished at the stack. There were 3 to 4 liters per run.
The run times were 15 to 20 minutes in length.
There was a direct sample line which was used to continuously monitor
for C02, 02, and total hydrocarbons.-
The bags from the Model 90 were analyzed directly on the Model 401.
The adsorbent tube was purged with zero air to remove any CO2 or CO into a
second bag. The second bag was analyzed on the 401 for any VOC's removed
from the tube during the purge. The adsorbent tube was then heated in the
Model 75 oven to convert the remaining organic compounds to CO? which were
then collected in a third bag. The third bag was then analyzed for C07.
The total non-methane hydrocarbon content of the sample was calculated as
ppm carbon as the sum of the NMHC in the first two bags plus the C00 in the
third bag. -*
EPA Reference Method 25 audit gases were obtained for both the March
and the June/July test series. There were two gases for each series. None
of the audit gases were analyzed satisfactorily. But there was a marked
improvement for the second set. The results are summarized below.
101
-------
March
Component Company
CO? (%) 3.4
Total NMHC (ppm C) 194
C02 (%) 3.5
Total NMHC (ppmC) 1081
C02 (%) 4.7
Total NMHC (ppm C) 249
June/
July C02 (%) 3.9
Total NMHC (ppm C) 1240
EPA
5
107
5
775
5
205
5
1043
% Difference
-30
+80
-30
+ 40
-5
+21
-20
+19
The company consistently underestimated the C02 level while
overestimating the non-methane hydrocarbons. In conversations with Dr.
Jayanty of RTI, he indicated that the company's values for the NMHC of the
digest portion appeared to be high. This suggests that all the C02 was not
being removed from the adsorption tube, and ended up as being considered
NMHC. This is supported by the improved analysis of the audit gases for
the June/July test series. The March test series and audit gas analysis
had equal 4 liter volumes for the sample, purge, and digest. The June/July
test series had sample volumes of 3.4 liters, purge volumes of 5.1 liters
and digest volumes of 1.7 liters. It appears that the increased purge
volume resolved some of the problem with the audit gas analyses.
The C02 values in the stacks which were tested did not appear to be
high enough to affect the results. While the change in the ratio of the
sample, purge, and digest volumes could have affected the results, such
changes were not readily apparent.
The conclusions regarding the use of the audit gases at the automobile
assembly plant are that:
1) The audit gases were not analyzed satisfactorily.
2) The audit gases were not representative of the
sample gases (5% C02 vs. 0.04 - 1.5% C02)
This decreased the
analytical error.
significance of the audit gas
3) The change in methodology improved the audit
gas analytical results, but made the interpretation
of the before and after results more difficult.
102
-------
CONCLUSIONS
1. If -possible, the use of two audit gases is
desireable. One audit gas should represent the
compliance level, and one should represent the
expected range of emissions. The expected range
of emissions is not that easy to determine as
shown by difference between the two bulk gasoline
terminals.
2. The results of the audit gas analysis should be
interpreted. If there is a question regarding the
analysis which could effect the validity of the
tests, RTI should be requested to reanalyze the
gas. The bulk gasoline audit gas analysis varied
from 1.18% to 1.3%. Also, the suitability of the
audit gas for the particular test should be a
consideration when interperting the results. The
automobile assembly plant did not have the C02
levels which were in the audit gas, thus reducing
the significance of the variance.
103
-------
UTILIZATION OF EPA METHOD 25 FOR THE MEASUREMENT
OF VOLATILE ORGANIC CARBON (VOC) EMISSIONS FROM
COMBUSTION AND STEAM SOURCES
G.C. Simon, National Council for Air and Stream Improvement
ABSTRACT
A series of simple modifications to the method allows
accurate testing of combustion gases from industrial boilers
without significant interference from carbon dioxide. Similarly,
another set of minor modifications makes the sampling of steam
vents possible without trap or catalyst overloads.
The hardware alterations and procedural changes will be
fully described. Data from thirteen wood-residue fired power
boilers will be presented, including data from three units that
were tested concurrently using either a flame ionization detector
or infrared spectrometer in the field. Data from several thermo-
mechanical pulping operations will also be presented.
I INTRODUCTION
Volatile organic compounds (non-methane hydrocarbons)
participate in photochemical reactions which lead to formation of
ozone in the ambient atmosphere. Ozone has been designated as a
criteria pollutant by the U. S. Environmental Protection Agency,
and ambient air quality standards have been set to protect
against adverse health and welfare effects. To control the
formation of ambient ozone, EPA has emphasized the control of
volatile organic compound (VOC) emissions from both mobile and
stationary sources. EPA considers only methane, ethane, methyl
chloroform, methylene chloride and certain fluorocarbons and
chlorofluorocarbons to have negligible reactivity, and thus not
be categorized as VOCs (1). VOC emissions from many types of new
stationary sources, e.g. paper coating facilities, are limited by
New Source Performance Standards. In addition, obtaining con-
struction permits for new sources or for modifications to existing
sources often involves estimating VOC emission rates to determine
what types of permitting requirements and emission control
considerations apply.
Accurate VOC emission estimates are important in obtaining
permits for new or modified sources located in both ozone non-
attainment areas as well as attainment areas. In nonattainment
areas, VOC emissions from a new source must be offset by VOC
emission reductions from other existing sources in the area.
Permit applicants in ozone nonattainment areas must also demon-
strate that their VOC emissions represent the lowest achievable
emission rate. In attainment areas, permit applicants must show
that best available control technology (BACT) will be used to
104
-------
control VOC emissions from new or modified sources. However,
this BACT review need not be conducted if there is no significant
net increase_in VOC emissions resulting from the project, and
possible ambient ozone monitoring and modeling requirements can
also be avoided.
In the past few years, the National Council staff has
received numerous inquiries concerning VOC emissions from various
forest products industry sources. Most of these inquiries have
been in connection with obtaining permits for new or modified
sources. A particularly interesting situation involving instal-
lation of a new wood-fired boiler at a mill near a metropolitan
ozone nonattainment area was recently described (2). The mill
was required by the permitting agency to obtain VOC emissions
offsets for the new boiler, and VOC emission factors for the new
boiler proved to be an extremely important issue in the permitting
process. Similar situations have occurred in many other permit-
ting circumstances.
In an effort to provide the industry with reliable VOC
emission data, the National Council has been surveying major
sources at forest product manufacturing facilities in the United
States utilizing EPA Method 25. To date, VOC emission measure-
ments have been reported on kraft recovery furnaces, lime kiln's,
wood residue fired power boilers in the Northwest and Southeast,
veneer dryers and thermomechanical pulping (TMP) operations
(3,4,5,6,7,8). Modifications to the basic method were necessary
to overcome CO2 interference problems associated with combustion
source samples and high moisture levels (>90%) associated with
TMP samples. These modifications are presented in detail in the
following sections.
In addition, during the course of these studies three
opportunities arose for the National Council staff to sample
combustion sources using the modified EPA Method 25 concurrently
with contractors or mill personnel using alternate techniques.
In two instances a flame ionization detector was used on-site,
and in a third test the contractor employed an infrared spec-
trometer on-site. Results from these comparative tests are also
presented.
II METHOD 25 MODIFICATIONS
For details of Method 25, readers should refer to the
Federal Register (45 1194] October 3, 1980). Only the modifica-
tions, and reasons for them, are described below.
A. Sampling Equipment and Procedures
(1) Trap Design - To allow condensation of several milliliters
of water without trap plugging, the trap was enlarged to 1 inch
O.D. by 7.5 inches long and filled with fine quartz wool instead
105
-------
of stainless steel turnings. The trap body was low carbon 316
stainless steel and the inlet/outlet tubes, the 4 ft. x 1/8 in.
probe, and the fittings were standard 316 stainless steel (Figure
I) . It was felt that the firmly packed fine quartz wool would be
more efficient in retaining aerosols formed in the trap during
the condensation process. In addition, three No. 40 holes were
bored through the exit tube instead of one. This allowed gas
flow to continue even after more then 20 mL of water had been
collected.
Inlet Tube
Extends
^•3/8"
Into Barrel
1/4" Swagelok
Weld Socket Unions
V4" Tubing
Fine Quartz Wool
0.065" Wall
#40 Holes
FIGURE 1 NCASI METHOD 25 TRAP DESIGN
(2) Flow Controller Design - Specifications only require a flow
control system that is "capable of maintaining the sampling rate
within ±10% of the selected flow rate (50 to 100 mL/min)" (10).
Flow controllers used in our work consisted of all glass/Teflon
rotameters (Gilmont, 0 to 200 mL/min) connected via a thick
walled 1/8 in. Teflon tube to a pressure/vacuum gauge-metering
valve-quick connect assembly. All assembly components and
fittings (Swagelok) were stainless steel.
(3) Sample Tanks - Six-liter cylindrical stainless steel sampling
cylinders were used for sampling tanks (Daytron Systems, Nicholson
Division, Wilkes-Barre, PA). These tanks were thoroughly cleaned,
their volumes measured, and then equipped with stainless steel
quick connects on one end and plugs on the other end.
106
-------
(4) Particulate Filters - Filter assemblies were constructed
from 4 in. long x 1/2 in. diameter stainless steel tubes that
were firmly packed with clean borosilicate glass wool and fitted
with 1/2 in. to 1/4 in. stainless steel Swagelok reducing.unions.
A 1/4 in. tube stud to 1/8 in. Swagelok stainless steel reducing
union, also packed firmly with Pyrex wool, was used to couple the
large filter to the trap probe line. This entire assembly was
wrapped with several layers of aluminum foil to keep soot (boiler/
samples) or tar (TMP samples) off of the fittings, thus minimizing
contamination during disassembly. Any visible breakthrough of
particulate matter onto the 1/4 in. tube stud filter was indica-
tive of contamination and grounds for data rejection. In prac-
tice, this rarely occurred. Filters were placed in the flue gas
facing generally downstream to avoid particle impingement and
positioned appropriately to avoid collection of condensate from
post-scrubber aerosols or liquified tars that impinged on the
probes' and filter assemblies' outer surfaces.
(5) Pre-sampling Leak Check Procedure - Sample traps, with the
probe tips plugged, and flow controllers were assembled and then
pressurized to 25 psi with nitrogen and isolated. The pressure
was observed for 5 to 10 minutes while trap fittings were wetted
with a bead of distilled water and inspected for leaks. Flow
controller fittings were wetted with Snoop and checked for leaks
before being rinsed thoroughly with distilled water. Leaks were
either repaired on the spot or the sample train component was
replaced. A pressure drop of less than 1 psi after 5 minutes was
acceptable. This pressurized leak check allowed the use of
"leaky" equipment by providing a means of isolating and repairing
the leak without contaminating the component. After the leak
check, the pressure was released, the filter tips and evacuated
sampling tanks connected, and the filter assemblies placed
several feet into the stack facing downstream as described above.
(6) Sample Collection Procedure - Replicate boiler samples were
drawn at a nominal flow rate of 80 mL/min. The flow was
controlled by observing the rotameter and adjusting the metering
valve to compensate for the changing vacuum force in the sample
tank. A trap freeze-up, as evidenced by a very low flow rate,
was dealt with by gently warming the top of the trap near the
inlet tube with a propane torch until a normal flow was re-estab-
lished.
The NCASI trap design also allowed the collection of steam
samples from TMP vents which tended to fill the traps with water.
Because the dry gas content was so low (1 to 10% by volume) in
these sources, the flow rate into the evacuated tank was extremely
low (<10 mL/min). However, the flow rate of water vapor into the
trap was 10 to 100 times greater. A qualitative technique was
developed in the field to judge the moisture flow rate into the
trap. Steam entering the 4 ft. x 1/8 in. sampling probe
condensed to a liquid as it cooled to near ambient temperature.
The rapid temperature drop at the condensation point along the
probe length could be physically detected by touch. The dry
107
-------
gas flow rate was adjusted until the steam condensation point was
6 to 12 in. from the stack wall. This technique resulted in the
collection of 20 to 30 mL of liquid in ca. 30 min. Steam samples
were also collected in replicate.
(7) Post Sampling Leak Check Procedure - After closing the
metering valves and removing the filter/probe assemblies from the
stack, the probe tips were again plugged. The metering valves
were opened and the trap/flow controller assembles were evacuated
to the level of the respective sample tank's vacuum. The metering
valves were then closed and the vacuum gauge readings were
observed for ten minutes. A change of less than one inch of
mercury vacuum was acceptable. Larger leaks were noted and dry
gas concentrations in the replicate sample tanks were compared to
evaluate the extent of the leak. In practice, leaks were rarely
found after sampling. A relative standard deviation greater than
five percent between duplicate sample dry gas concentrations was
grounds for data correction or rejection.
B. Analysis Equipment and Procedures
(1) Condensate Sample Recovery System Modifications - Sample
trap probes were hooked directly to the recovery system's oxida-
tion catalyst and sequentially heated along with the trap and
backflush condensers to 600°C all the way up to the catalyst hot
zone using a tube furnace and a MAPP gas torch. The zero air
carrier gas flow rate was 100 to 150 mL/min. The catalyst tube
was a 14 in. long by 1 in. O.D. low carbon 316 stainless steel
tube packed with 10 inches of small grain manganese dioxide. It
was heated to 700°C by a Lindberg tube furnace in the vertical
position (Figure 2). This minimized channelling due to catalyst
settling. This catalyst was found very efficient at oxidizing
high levels of organics (i.e. 200 mg C/sampling), and resistant
to sulfur poisoning. Catalyst efficiencies remained >98% even
after a year or longer.
Both the backflush gases and the sample recovery gases (the
catalyst effluent) passed through an IR (Beckman Model 864)
before entering the appropriate collection vessels. The IR
monitor was used in the 0 to 100 ppm range for boiler samples,
and 0 to 1.5 percent for highly loaded steam samples to determine
when all of the CO9 had been removed from the system during each
cycle. Even thougn the baseline tended to drift with the changing
pressure on the system as the tanks were filled, the trap effluent
could be stopped and zero air fed directly into the IR via a
system bypass loop to re-establish the baseline level at the
higher pressure. Using this technique, very low levels (less
than 5 ppm) of C09 could be accurately observed (Figure 3). A
dry ice/water cooled double glass U-tube condensed and collected
the water in the sample trap for subsequent quantitation. The
first of these tubes was fitted with a reservoir made from a
buret and sealed to the bottom of the U.
108
-------
Pressurt
Gauge
f
Shut-
Off , „
S«p Valve <;W
r
rift"
Back flu sly f***^
~~ Trap in
Boiling
Water
Shutoff Valve
^H*
T
ShutoffHJ
Valve <
J ll
\ Wet u-tube
\ -Ice Water
\ Trap
^- Oxidation
Catalyst
Furnace
1
Cry
Ice
ARotameter
i ni
1
1 1 p^n
1 NDIR CO,
Analyze?
Pressure
Vacuum
Gauge
Sample
Tank
Backflush Mode
Flow Control Valve
Shutoff Valve
-<§>-
Shutoff
Valve
Pressure
Gauge
Backflush^
Trap (hea'tecl
with torch)
Condensate
Trap Furnace
Oxidation
Catalyst
Furnace
U-tube \
Water \
Trap >
* Dry
Ice
Burn Mode
NDIR CO-
Analyzer
Pressure
Vacuum
Gauge
Intermediate
Collection
Vessel
FIGURE 2 CONDENSATE SAMPLE RECOVERY SYSTEM
(2) Recovery System Leak Check Procedures - Recovery system leak
checks were performed by pressurizing the entire system with zero
air to 20 psi with the sample trap in place,in the backflush
mode. After waiting several minutes for the pressure to equili-
brate (the heat input from the catalyst caused some slight
pressure increase in the closed system for the first 1 to 2
minutes), the pressure was recorded and then rechecked 10 minutes
later. A drop of less than one psi was considered acceptable.
During the ten minute waiting period all trap fittings were
wetted with a bead of distilled water and observed for signs of
109
-------
leaks. All other system fittings were wetted with Snoop and
checked for leaks. Leaks were corrected when found and the 10
minute pressure drop test was restarted.
FIGURE 3 INFRARED SPECTROMETER READOUT OF
CO2 CONCENTRATION IN METHOD 25
SAMPLE RECOVERY SYSTEM EFFLUENT
WHILE PROCESSING A BOILER SAMPLE
(3) Backflush and Sample Recovery Procedures - Boiler sample
traps were subjected to a "warm purge" during the backflush
cycle. The gases in the sample trap were flushed into the sample
tank using zero grade air. In combustion source samples, the
presence of high concentrations of water vapor and carbon dioxide
causes some potential interference to hydrocarbon recovery. This
is caused by CO,, dissolving in the water in the sample trap
during cooling and subsequent entrapment in the water-ice matrix
on freezing. To eliminate this potential positive bias, the
traps were placed in boiling water and vibrated sharply during
the backflush cycle. This action effectively drove off the
entrapped CO.,, while vaporizing very little water (Figure 3) .
The backflusn gases were passed through a dry ice condenser
consisting of a 9 in. long by 3/8 in. O.D. stainless steel tube
packed with quartz wool and immersed in dry ice. This condenser
was backed up with a secondary dry ice cooled condenser con-
structed from three feet of coiled 1/8 in. stainless steel
tubing. These condensers caught any condensible hydrocarbons
110
-------
that vaporized during the warm purge and were sequentially heated
to 600°C. along-with the trap and probe during the recovery
procedure, using zero air for carrier gas (Figure 2). Traps were
heated quickly since generally organic loadings were <1 mg as
Carbon, and water contents were <3 mL. These levels represented
light loadings to our large catalyst. Steam sample traps were
subjected to the standard dry-ice temperature backflush, and then
warmed in a boiling water bath to drive off most of the organics
with very little water. Auxiliary oxygen was added at 50 to 150
mL/min. just before the catalyst when effluent CO.., concentrations
were >1.0% . When the bulk of the organics were out, the trap
was heated slowly with a torch while the system pressure was
monitored. Both backup condensers in the recovery system were
still immersed in dry ice in order to prevent any moisture and
organics from contaminating the system in case of pressure surges
due to the water vaporization. The high volumes of water (20 to
30 mL/sample) tended to cool the catalyst if put through too
quickly. Recoveries of steam samples typically took three hours.
All intermediate collection vessels were analyzed for combustion
by-products to assure efficient recoveries.
Due to the modified design of our Method 25 traps, there
were many "dead spaces" where gas flow was restricted. In order
to assure that these spaces were swept clear of CO,,, a "surge"
technique was used. This technique involved stopping the air
flow through the trap for one minute to allow diffusion of C0?
from dead spaces. The flow was then restarted and a CO2 peak was
observed on the IR. For low level samples, when the peak caused
by a surge was less than 5 ppm, and the background CO- concentra-
tion was less than 2 ppm, the system was considered cleared
(Figure 3).
After completion of the sample recovery, the trap was
removed from the tube furnace and allowed to cool in the ambient
air for about five minutes before being sealed. Eventually this
technique was changed, and the hot traps were placed in a bucket
of dry ice and allowed to cool down to -75°C while zero air was
flowing through before being sealed and stored at ambient condi-
tions until used. This technique caused the trap to be stored
under positive rather than negative pressure (trap sealed while
hot). This caused a significant improvement in blank levels
during the course of these studies (Table 1), presumably due to a
reduction in contaminate infusion which occurred when the trap
seals were broken. Although blank levels improved with investi-
gator experience, it was not until the institution of the cold
sealing technique that consistently low blank values were
achieved. This technique was used for low level samples only.
(4) Sample Analyses - A non-methane organic (NMO) analyzer from
Byron Instruments, Raleigh, NC (Model 40IS) v/as used throughout
this study for gas analyses. This unit was capable of separating
non-methane hydrocarbons, methane, carbon monoxide, and carbon
dioxide and passing each component individually through an
oxidation catalyst for conversion to CO-, and then through a
111
-------
reduction catalyst for conversion to methane and subsequent
quantitation on a flame ionization detector. All of these
actions occurred automatically once the analysis cycle was begun,
This system differed in several respects from the system recom-
mended by EPA Method 25. The same separations and measurements
were achieved, but the physical arrangement of the hardware and
the column packing materials were altered.
TABLE 1 METHOD 25 BLANK VALUES
DATE
April, 1982
October, 1982
February, 1983
March, 1983
April, 1983
TRAPS
SEALED
HOT OR COLD
Hot
Average
Hot
Average
Hot
Average
Hot
Average
Cold
Average
RESULTS
Micrograms
Carbon
177
167
168
171
115
48
85
83
42
!§_
35
66
87
38
64
20
25
20
As ppm in a
5 Liter Sample
71
67
67
69
46
19
11
33
17
14
27
35
15
26
8
10
6
8
III
LABORATORY CO,, ABSORPTION INTERFERENCE STUDY
Most combustion sources have stack moisture contents above
20% and carbon dioxide concentrations above 10%. At these
conditions the interference from CO,, absorption in the water-ice
matrix can be significant.
112
-------
Adsorption characteristics of carbon dioxide in the cryogenic
traps used in these studies were determined in the laboratory
using several different sample stream carbon dioxide and moisture
concentrations. The test covered conditions of no CO^ or moisture
to 18% CO- and 50% moisture in the gas stream. Results from the
tests are reported in a format that allowed correction of field
sample data for the interfering CO2.
(1) Procedure - Carbon dioxide in air mixtures were directed to
a heated U-tube containing water to add moisture to the:air.
stream. The water added to the U-tube was passed through acti-
vated carbon to remove organics. The quantity of moisture added
to the air stream was roughly controlled by adjusting the tempera-
ture of the U-tube. This moisturized air was passed through a
MnO,, oxidation catalyst at 600°C to insure that there were no
organics delivered to the trap. After the oxidation catalyst the
gas passed through a cryogenic trap set in dry ice, to a flow
meter and to a 17-liter evacuated tank. This apparatus is the
trap burnout system used in reverse but with a sample collected
as in the field. Gas flow was controlled to 200 cc/min until 10
liters of gas were collected.
The trap was then flushed with carbon dioxide free air
flowing at 100 cc/min until the NDIR registered less than 10 ppm
CO9. Flushing of the traps took from 20 ,min. to 1 hr,.
After flushing, the trap was heated with an acetylene torch
with the CO^-free sweep-gas passing through the oxidation furnace,
a cold trap to remove moisture, the NDIR detector, and into an
evacuated collection vessel. When all of the CO,, was removed
from the trap, as indicated by the NDIR, the collection vessel
contents were analyzed. The amount of water collected in the
cold trap was used to determine the gas stream moisture content.
(2) Results — Results of the carbon dioxide interference study
are shown in Figures 4 and 5. Figure 4 shows a constant back-
ground carbon dioxide interference for all moisture levels when
the sampled gas contained less than 5 ppm CO2 or 659 ppm CO,,.
After elimination of one data point from the 650 ppm CO2 data,
the log average CO.., background was 8.0 ppm CO2 for the 559 ppm
CO- data, and 8.1 ppm for the 0% CO2 data. Tfie log average of
all the data when no moisture was added to the traps was 7.4 ppm.
These results show that when there is no water present in the
trap, all of the carbon dioxide is successfully removed from the
trap during flushing.
Figure 5 shows the CO,., absorption interference as a function
of sampled gas moisture content at carbon dioxide concentrations
between 8 to 18%. Below 20% moisture there, appears to be no
significant interference above background. Above 20% flue
moisture there appears to be a random interference that increases
with sampled gas moisture and carbon dioxide content. The
randomness of this data makes it difficult to predict the CO2
interference encountered at high stack moistures. Figure 6 shows
the same data in terms of mL H20 caught in the trap.
113
-------
r o i i i • i i i Oi
a o prv
0^ lO UJ
_ O U)
on a
cP
1 ' 'an
CL
CO <
ct:
h-
^
«s
o
UJ
:J
CD
O
o
CM
X
CJ
E
/•-i
o o o o
ID sj- OJ
U
o
M
CO
O CO
Z
« o
U H
*£ EH
^ !^
CO «
EH
^ O
S8
O Q
Z M
W X
os o
ca H
&4 O
K;
w z
EH O
Z CQ
H
O
U
30N3d3dd31NI
114
-------
CO
W
M
fa
oo
u
-------
A closer look at all the data collected for 12% CO2 (Figure 7)
shows that the CO2 interference tended to fall into two groups,
that which grouped about a line with a slope of 18.5 ppm C02
interference per mL ll^O in the trap and that which fell below 30
ppm interference regardless of the amount of water collected in
the trap. There is no apparent reason to explain the different
groupings. Almost all of the interference data fell below the
theoretical maximum interference which occurs when the water
absorbing the CO2 is at 0°C. The 18.5 ppm CO2 interference per
mL water corresponds to the theoretical interference at a water
temperature of 3°C.
These results are consistent with other workers who reported
erratic CO^ interference levels in a similar study (9). The
"warm purge" backflush technique described above minimized this
interference by effectively driving off the CO_ from the water by
heating it to 100°C. The dry-ice immersed bacRup traps in the
recovery system prevented any loss of condensible hydrocarbons.
IV EVALUATION OF ALTERNATIVE METHODS FOR MEASURING
VOC EMISSIONS FROM INDUSTRIAL BOILERS
In an effort to identify potentially useful equivalent
alternative techniques for measuring boiler VOC emissions, three
simultaneous sampling tests were performed with either private
contractors or mill personnel. The three tests were on emissions
from a Dutch oven wood-residue fired boiler, a bagasse fired
boiler, and a combination wood fines and natural gas fired
boiler. In each test, NCASI staff sampled the flue gases with
EPA Method 25 as described earlier, while the other groups used
(a) an EPA Method 5 particulate train followed by a flame ioniza-
tion detector (FID), (b) a Byron Model 4015 hydrocarbon analyzer,
and (c) an infrared analyzer, respectively. The first two
alternate techniques were modified versions of EPA Method 25A,
while the third technique approximated Method 25B. Both of these
methods were intended for use with well characterized sample
streams, but there was some expectation of adapting them for flue
gas sampling. An equivalent method utilizing existing sampling
equipment with minor changes would offer an advantage to affected
facilities.
Descriptions of the boilers and alternate sampling techniques
and comparative results are presented below.
A. Particulate Sampling Train Followed by an FID Versus
Method 25
Mill personnel sampled Dutch oven wood-residue fired boilers
for VOCs using a sampling train consisting of an EPA Method 5
particulate sampling train followed by an FID, as shown in
Figure 8. The sampling nozzle was turned upstream in the stack
116
-------
to minimize particulate capture. Particles entering the probe
were filtered by stuffing the probe with glass fiber. No filter
was used in the sampling train oven. The sample was cooled and
water removed with three .impingers in an ice water bath. Sampling
started with 100 mL water in each of the first two impingers and
the third impinger dry. A Teflon coated diaphragm pump, drew the
sample gases through a heat traced Teflon line into a sampling
manifold in a trailer.
Back Pressure
Regulator
and Gauge
Diaphragm
Pump with
Teflon
Impingers in
Ice Water Bath
3-Way Valve
Integrator
Gas Sample
Valve and Loop
FIGURE 8 EPA METHOD 5 PLUS FID SAMPLING TRAIN
Samples were delivered to an FID via a 0.1 mL sample loop
every two minutes in a nitrogen carrier gas. The gas pressures
in the sampling loop and the sample loop temperature (100°C) were
kept constant so that an equal volume of sample was delivered to
the FID. The FID was calibrated with methane in nitrogen via the
sample loop.
A small sampling loop size was used to dilute the sample
enough so that changes in FID response due to CO_ and moisture in
the sample would be avoided. A 1/16 in. stainless steel tube led
from the valve to the FID. The GC oven and manifold were at
120°C and 200°C, respectively.
Organics that condensed in impingers were analyzed by a TOC
(total organic .carbon) analysis of the impinger water and drying
and weighing an acetone rinse of the EPA Method 5 train probe and
glassware. Impinger water TOCs were used instead of ether
117
-------
extraction and drying, as recommended for EPA Method 5 back half
analysis, because of possible volatile organic material losses
during drying. Comparison of analysis by weight to the TOC
analyses showed the drying and extraction result to be consistent-
ly much less than the TOC result. The sums of the TOC and the
dried acetone rinse weights of the EPA Method 5 train analysis
were added to the FID response to give a total gaseous organic
carbon result.
B. Results of Method 5 Plus FID Versus Method 25
Results for VOC emissions from both the EPA Method 25
duplicate average and the EPA Method 5 plus FID trains are listed'
in Table 2. Methane concentrations found in the EPA Method 25
samples were subtracted from the mill data shown in Table 2. The
EPA Method 25 data expressed as Ib methane/10 Btu were calculated
using the CO,, found in the tank during analysis,-and from mill O2
data. The mill data expressed as Ib methane/10 Btu were calcu-
lated using mill CO~ and 0,, data. These results show that the
methods were comparable, but the alternative is more complex and
expensive to run than Method 25.
TABLE 2 BOILER G - COMPARISON OF VOC EMISSIONS DETERMINED BY
EPA METHOD 25 AND METHOD 5 PLUS A FLAME IONIZATION
DETECTOR (DUTCH OVEN WOOD-RESIDUE FIRED BOILER)
TEST DATES 12/3-4/80
MILL
TEST NO.
2
3
5
NCASI-EPA METHOD 25 VOC
PPM
380
278*
341
(as methane)
n—/ -i n P
lb/10
Btu
From
0.44
0.39
0.32
lb/10
Btu
From
02
0.43
0.29
0.33
MILL-EPA METHOD 5 PLUS
FID VOC
(as me
'
lb/10"
Btu
From
ppm CO0
363
387
355
0.
0.
0.
37
34
31
lb/10'
Btu
From
0.
0.
0.
0.
42
40
34
*This sample had a leak.
C. Hydrocarbon Analyzer Field Technique Versus Method 25
A Byron 401S analyzer was housed in a small trailer on-site
at the base of a bagasse fired boiler stack. About 250 feet of
heat traced Teflon line was used to transport the sample from a
stainless steel in-stack filter holder fitted with a glass fiber
118
-------
filter. The sample line was kept at 250°F; the stack temperature
was 155°F. The analyzer was operated in the "full cycle without
CO^" mode, which gave results for a direct FID injection, non-
methane hydrocarbons, methane, and carbon monoxide. About two
liters per minute of sample gas were drawn from the stack with a
Teflon diaphragm pump to a splitter. Thirty mL/min were pulled
through the analyzer's sample loop with a second pump. All lines
and connections were wrapped with heat tape up the entrance to
the 40IS. This left about eight inches of unheated 1/8 inch
diameter stainless steel line that was warmed to about 160°F by
the instrument's heat. The sample then entered the sample loop
which was maintained at 220°F. Injections were made every eight
minutes. The limitation this imposed was recognized early on,
but it was thought that the source emissions may be steady enough
for accurate assessment despite the data point limitation.
Average non-methane organic concentrations as determined by the
instrument's integrator were reported as ppm VOC. Carbon monoxide
values were also reported in ppm concentrations.
Table 3 shows the comparative results from the two methods
in terms of concentration. The analyzer VOC and CO results are
significantly lower than the Method 25 results in all cases.
TABLE 3 BOILER H - COMPARISON OF VOC AND CO EMISSIONS
DETERMINED BY EPA METHOD 25 AND A BYRON 40IS
HYDROCARBON ANALYZER (BAGASSE FIRED BOILER)
TEST DATE 2/4/83
TEST
1
2
3
NCASI METHOD 25
CO (ppm)
1388
2268
1644
VOC (ppm)
74
137
439
CONTRACTOR (ESE)
BYRON 40IS ANALYZER
CO (ppm)
*807+
*182+
1320
VOC (ppm)
47
*67+
99
*These averages contain offscale peak values; see text.
Three problems were encountered with the use of the hydro-
carbon analyzer, which led to the low values. First, VOC and CO
concentration fluctuations within the boiler were far more
significant than anticipated. Since high concentration spikes
were short-lived by nature, it was impossible to grab a represen-
tative sample. Second, the contractor experienced some difficulty
in maintaining the sample line temperature, which contributed to
VOC loss. And third, a CO2 interference of 3.3 ppm non-methane
response per percent CO- caused a large correction factor to be
119
-------
applied to the VOC data. The concentration fluctuations severely
limited the effectiveness of the analyzer due to its eight minute
cycle time. Unfortunately, the contractor could not run the
instrument in the direct FID injection mode (sample every two
minutes), since this monitoring contract required simultaneous CO
sampling. (In the direct injection mode, the 401S analyzer fills
Method 25A requirements.)
The low sample line temperatures probably contributed to VOC
losses, but the short piece of unheated sample line inside the
instrument which was only warmed to about 160°F was a limiting
factor regardless of sample line temperature. The contractor
made no attempt to flush any condensed organics from the sampling
lines.
The CO,, interference in the instrument's non-methane response
varied with each injection, but it was not possible to measure
the CO., concentration with each injection, since the CO^ cycle
injection occurs at the conclusion of the NM-CO-CH4 cycle and
would not represent the CO level injected in that earlier cycle.
Instead, the average CO,, level, as determined by Method 25, was
used to calculate the average non-methane interference for each
run. This average value, on the order of 50 ppm, was then
subtracted from the uncorrected non-methane values, which were on
the order of 100 to 150 ppm. This technique introduced the
potential for significant unquantifiable error.
The Byron 40IS hydrocarbon analyzer did not yield results
equivalent to Method 25 when measuring VOC and CO emissions from
an industrial boiler on-site due to (a) the fluctuating nature of
the source emissions,and the instrument's eight minute cycle time
that significantly limited its ability to average high concentra-
tion spikes; (b) a variable C02 interference that could only be
estimated when operating the instrument on-site.
D. Infrared Spectrometer Method
A sample was drawn from the stack through a glass fiber
filter, an ice bath dropout condenser, a Teflon diaphragm pump,
and then pushed through an Infrared Industrial Model 711 Hydro-
carbon Analyzer. The instrument was calibrated with propane and
results were presented as both ppm propane and ppm methane. This
method was meant to approximate Method 25B, with the exception of
the ice bath condenser. It was noted that the addition of this
moisture removal device would also remove some heavier organic
compounds. However, this step was necessary to protect the IR
from the high moisture levels of the flue gas. This moisture
would otherwise condense on and damage the cooler surfaces of the
inorganic salt crystal windows in the instrument.
Comparative results for the two methods are tabulated in
Table 4. It is evident from these data that the methods are not
equivalent. The IR results ranged from 165 to 390 ppm (as CH4),
with an average of 235 ppm, while the Method 15 values ranged
120
-------
from 290 to 975 ppm (as CH ), with an average concentration of
564 ppm. The IR values ranged from 35 to 57 percent and averaged
44 percent of the Method 25 values. The moisture dropout conden-
ser was thought to be the major source of hydrocarbon loss, but
no attempt was made to analyze the condensate.
TABLE 4 BOILER J: COMPARISON OF VOC EMISSIONS DETERMINED BY
EPA METHOD 25 AND METHOD 25B (INFRARED SPECTROMETER)
RUN
1
2
3
4
5
6
STEAM PRODUCTION
RATE (Ib/hr)
90,000
75,000
50,000
60,000
90,000
40,000
VOC RESULTS, PPM AS CH
NCASI-
Method 25
*.
Contractor
IR
Average
975
631
559
415
514
290
564
390
240
195
225
195
165
235
Another point of non-equivalency Between the methods is
inherent in the technique of measuring different organic compounds
using one wavelength of infrared light. Response factors will
vary among the different compounds, and no attempt was made to
determine the extent of the variance in compounds likely to be
present in boiler emissions. The method is intended for use on
well characterized sources whose VOC emissions are primarily
alkanes, alkenes, and aromatic hydrocarbons.
V
SUMMARY OF VOC EMISSION FACTORS DETERMINED BY
METHOD 25 FOR WOOD-RESIDUE FIRED BOILERS AND
THERMOMECHANICAL PULPING OPERATIONS
A.
Northwestern Wood-Residue Fired Boiler VOC Emission
Study Summary
(1) Four wood-residue fired boilers operated on fuel derived
from Douglas fir were sampled for VOC and carbon monoxide emis-
sions using Method^25. Results are listed in Table 5, along with
boiler sizes and capacities.
121
-------
TABLE 5 NORTHWEST WOOD-RESIDUE FIRED BOILER VOC EMISSION DATA
CO
15710°
Btu
Boiler
0.06
0.19
0.22
0.18
0.10
0.14
0.08
0.05
0.21
0.04
0.06
0.06
Boiler
0.03
0.10
0.09
0.08
0.07
0.04
0.04
0.07
Boiler
0.06
0.14
0.08
0.08
0.08
0.08
Boiler
0.03
0.05
0.05
0.04
0.06
0.06
0.04
0.05
-^t
PPM
A
100
190
310
190
140
210
100
76
316
53
63
75
B
79
180
120
100
60
30
40
80
C
61
116
74
84
77
84
D
41
70
78
71
99
84
61
71
Btu
3.25
3.03
-
1.20
0.64
0.31
0.38
2.16
1.45
0.42
0.66
1.50
0.042
0.091
0.417
0
0.604
0.539
0.249
0.110
1.44
4.00
2.92
2.99
2.71
2.29
0.117
0.151
0.224
0.144
0.242
0.291
0.243
0.537
ppm
3000
1750
3050
740
640
260
300
2230
5610
350
410
1010
48
97
641
0
273
255
156
70
900
1900
1570
1460
1640
1420
87
116
217
148
230
252
212
410
STACK
PERCENT
7.5
11.2
10.5
11.5
7.3
7.8
8.4
8.0
7.0
9.0
8.6
11.5
6.0
6.8
5.4
9.5
12.5
11.6
7.8
7.8
11.0
12.1
11.6
11.3
12.0
11.3
8.9
8.9
7.4
7.2
6.6
8.8
9.3
10.2
STACK
MOISTURE,
PERCENT
12.3
25.3
17.4
11.7
.15.3
16.0
16.3
12.6
16.6
15.3
20.9
7.0
10.6
13.9
12.3
9.7
15.0
15.5
15.9
12.0
16.8
13.9
13.3
17.7
18.7
13.9
13.3
19.4
11.9
AVERAGE
STEAM
PRODUCTION,
K Ib/hr
145
75
125
130
135
100
100
130
130
140
100
105
300
350
475
350
250
250
410
420
100
80
90
100
110
100
300
300
340
350
350
340
300
275
122
-------
to organics stripped from the scrubber waters. (The low and high
after-scrubber VOC emission rates were from two side-by-side
units, Boilers C and D, respectively, that both used pre-dried
fuel from a bark dryer that operated on Boiler D's exhaust. All
of the bark dryer emissions exited through" Boiler D's scrubber,.
The average VOC emission rate from these two units, 0.13 lb/10
Btu, is in the same range as the other boilers' after-scrubber
rates.)
TABLE 6 A SUMMARY OF VOC AND CO EMISSION RATES FROM
SOUTHERN WOOD RESIDUE FIRED POWER.BOILERS
BOILER
AND STEAM
CAPACITY
(K Ib/hr)
A
(135)
Average
B
(135)
Average
C
(200)
Average
D
(400)
Average
E
(360)
Average
F
(400)
Average
AVERAGE STEAM PRODUCTION (percent, CO
Percent wet (lb/10
of Capacity (K Ib/hr) basis) Btu)
85
86
ii
79
109
78
81
89
93
89
£8
90
93
97
99
96
96
58
2!
64
101
83
77.
78
69
82
115
116
89
107
147
105
109
120
185
178
176
180
372
388
395
384
385
207
254
231
t402
333
306
313
274
326
3.4
3.2
3.J)
3.3
3.4
3.3
3.6
3.6
9.1
5.6
7.4
1.8
2.6
2.8
2.7
5.1
3.0
1.82
2.35
5.59
3.25
1.35
2.36
2.74
2.15
1.27
0.99
1.07
1.11
0.42
0.43
0.43
0.51
0.45
0.32
1.56
0.94
0.99
0.29
0.14
0.18
0.42
0.40
Before
Scrubber
0.044
0.134
0.089
0.059
0.052
0.056
0.038
0.008
0.023
Not
Available
0.059
0.046
0.053
Not
Available
VOC (Ib carbon 10° Btu)
After
Scrubber
0.075
0.078
0.232
0.128
0.085
0.109
0.133
0.109
0.051
0.035
0.029
0.-038
*0.277
*0.214
*0.139
*0.201
0.208
No
Scrubber
tO.124
0.185
0.120
0.116
0.056
0.120
Difference
+0.034
+0.098
+0.066
+0.050
+0.081
+0.066
+0.013
+0.021
+0.017
*These values include emissions from a bark dryer. See text.
tA combination of bark and oil was fired during these tests.'
(4) Average CO emission rates for Boilers A, B, C, D, E, and F,
respectively, were 3.25, 2.15, 1.11, 0.45, 0.94, and 0.40 lb/10
Btu. This wide range of CO emission rates was also noted in the
earlier NCASI study on boilers in the Northwest (0.24 to 2.63
lb/10 Btu; see Reference 5). Wet scrubbers had no significant
effect on CO emissions.
123
-------
(2) Average VOC emission rates were 0.10, 0.05, 0.07, and 0.04
lb/10 Btu for Boilers A through D, respectively.
(3) VOC emission rates appeared to be related to the percentage
of total air used as overfire air. Larger percentages of overfire
air use corresponded with lower VOC emissions.
(4) Average carbon monoxide emission rates were 0.47, 0.26,
2.63, and 0.24 lb/10 Btu for Boilers A through D, respectively.
Carbon monoxide showed considerable variation and appeared to be
linked to flue gas oxygen content when oil was burned in Boiler
B, to flue gas oxygen content for Boiler C, and was completely
random for Boiler A.
(5) An interference to the VOC sampling and analysis technique
was investigated. This interference resulted from absorption of
carbon dioxide in water condensed in the sample trap and the
subsequent incorporation into the ice matrix in the cryogenic
trap. This C09 was not removed from the trap when flushed with
zero air. The resulting interference increased in variability
and magnitude in proportion with CO_ and increased moisture
content of the sampled gas increased. For typical wood-residue
fired boiler flue gases the interference was in the range of 15
to 30 ppm or 0.012 to 0.021 lb/10 Btu. A "warm purge" backflush
technique was developed to minimize this interference. The
minimum detectable for wood-residue fired boilers was 35 ppm with
EPA Method 25 in this study.
B. Southern Wood-Residue Fired Boiler VOC Emission
Study Summary _____
(1) Six wood-residue fired boilers operated on bark fuel derived
from southern pine and mixed southern hardwoods were sampled for
VOC and CO emissions using a modified version of EPA Reference
Method 25 which effectively eliminated CO2 interference. Five of
the boilers were fitted with venturi-type wet scrubbers (for
particulate control) that used some form of recycled mill water
for makeup. Three of these units were tested before and after
the scrubbers. Results are listed in Table 6.
(2) Average before-scrubber VOC emission rates for Bcdlers A, B,
C, and E were 0.09, 0.06, 0.02, and 0.05 Ib carbon/10 Btu,
respectively. These emission rates are in the same range as
rates measured on four boilers in the Northwest and reported
earlier by NCASI (0.04 to 0.10 Ib as methane, or 0.03 to 0.08 Ib
carbon/10 Btu; see Reference 5).
(3) Average VOC
A, B, C, D, and
carbon/10 Btu,
was noted in all
were taken. The
B, and C were 0.
and 45 percent,
emission rates after a wet scrubber for Boilers
F were 0.13, 0.11, 0.04, 0.21, and 0.12 Ib
respectively. A net increase in VOC emissions
cases where before- and after-scrubber samples
average emission rate increases for Boilers A,
07, 0.07, and 0.02 Ib carbon/10 Btu, or 52, 61,
respectively. The increased rates were attributed
124
-------
(5) There was no significant correlation between CO and VOC
emission rates at the concentration levels tested, with one
notable exception during an upset on Boiler A, when both compounds
exhibited high emission rates. Based on these observations it
did not seem practical to use CO as a surrogate for VOC during
normal operation.
(6) Three additional residue fired boilers were tested individ-
ually using EPA Reference Method 25 concurrently with one of
three alternate methods in an effort to identify an equivalent
method. The first alternative method employed an EPA Reference
Method 5 train with a flame ionization detector (FID) after the
last ice bath impinger. A total organic carbon (TOC) analysis
was run on the impinger water and the results were added to the
dried acetone train rinse weight and the FID concentration
values. The final .total was within five percent of the Method 25
values, but the complexity and equipment costs of the alternate
method exceeded those of Method 25. The second alternate method
used a Byron 40IS hydrocarbon analyzer with an in-stack filter
and a heat traced sample line. Results were not comparable to
Method 25 results due to (a) the fluctuating nature of the
emissions and the instrument's eight minute cycle time limitation,
and (b) uncertainty introduced by CO2 interference when the
instrument is operated in an on-site mode. The third alternate
method used an infrared (IR) spectrometer with an in-stack
filter, ice water condensate dropout bottle, and an unheated
sample line. Results averaged 44 percent of Method 25 results.
VOC losses were attributed to the moisture dropout bottle, which
was essential to the protection, of the IR sample chamber.
(7) The VOC emission factor for wood-residue fired power boilers
listed in the January. 1984 revision of AP-42 of 0.16 Ib (as
methane) or 0.12 Ib as carbon/10 Btu fired is about twice as
high as the average factor for emissions determined prior to a
scrubber in this study. However, this value is similar to the
average after-scrubber VOC emission rate of 0.12 Ib carbon/10
Btu.
C. Thermomechanical Pulping VOC Emission Study Summary
(1) EPA-25, the reference method for the measurement of VOC, can
be adapted to the measurement of emissions from refiner pulp
processes. Due to the high moisture content of these emissions,
calculations should be based on a moisture rather than dry gas
emission rate. Either a large trap operated at -78°C or two
standard size traps in series, the first operated at 0°C and the
second at -78°C, is suggested for sample collection.
(2) VOC emissions from two TMP processes operated on western
white wood species ranged from 1.02 to 2.15 Ib carbon/ton of air
dried pulp, with a daily average of 1.42 Ib/ton. From western
pine species emissions ranged from 0.83 to 3.40 Ib/ton, with a
daily average of 1.79 Ib/ton. From southern pine species at
three facilities emissions ranged from 2.15 to 7.61 Ib/ton, and
125
-------
had a daily average of 3.19 Ib/ton at the mills employing impress-
afiners (screw presses that pretreat the chips by mixing them
with hot water and squeezing), and 6.65 Ib/ton at the mill
without an impressafiner.
(3) The VOC emission rates were proportional to moisture emission
rates for the three wood species sampled. The emission potential
for white wood species was lowest, while that for southern pine
species was highest. The use of an impressafiner to pretreat
incoming chips was found to significantly decrease VOC emission
potential.
(4) A short field study suggested that heat recovery systems
designed primarily for steam recovery would be ineffective in the
capture of volatile organic carbon. Heat recovery systems
designed to condense water would be expected to capture volatile
organics as a function of vapor pressure of the organic species
present and operational temperature of the system.
VI LITERATURE REFERENCES
(1) Federal Register 45 (194) (October 3, 1980).
(2) "Proceedings of the 1981 NCASI West Coast Regional Meeting,"
Special Report 81-10, NCASI (October 1981).
(3) "A Study of Kraft Recovery Furnace Total Gaseous Non-Methane
Organic Emissions," Atmospheric Quality Improvement Technical
Bulletin No. 112, NCASI (February 1981).
(4) "A Study of Kraft Process Lime Kiln Total Gaseous Non-Methane
Organic Emissions," Technical Bulletin No. 358, NCASI (September
1981).
(5) "A Study of Wood-Residue Fired Power Boiler Total Gaseous
Non-Methane Organic Emissions in the Northwest," Atmospheric
Quality Improvement Technical Bulletin No. 109, NCASI (September
1980).
(6) "Volatile Organic Carbon Emissions from Wood Residue Fired
Power Boilers in the Southeast," Technical Bulletin No. 455,
NCASI (April 1985).
(7) "A Study of Organic Compound Emissions from Veneer Dryers
and Means for Their Control," Technical Bulletin No. 405, NCASI
(August 1983).
(8) "EPA Method 25 TGNMO Emission Factors for the Thermomechan-
ical Pulping Process," Technical Bulletin No. 410, NCASI (October
1983).
(9) "Method 25 Evaluation of Trap Recovery Unit Design," Report
No. 82-SFS1-3-2, EPA Contract No. 68-02-3543, Pollution Control
Science (1984).
126
-------
AN OVERVIEW OF THE DEVELOPMENT OF METHODS
OF CHEMOMETRICS
W. J. DUNN III
COLLEGE OF PHARMACY
UNIVERSITY OF ILLINOIS AT CHICAGO
833 S. WOOD ST.
CHICAGO, ILLINOIS 60612
Analytical chemists have developed methods of data analysis which are
capable of detecting and accurately quantifying the components of environmental
samples. As a result it is not unusual to observe large numbers of substances in
such samples. Samples containing toxaphene, for example, may show as many as
200-300 components. The storage, display and interpretation of large numbers of
such samples then becomes impossible without data analytic methods capable of
dealing with multivariate data. This part of the symposium is designed to
introduce various applications of multivariate methods of data analysis, so-called
methods of chemometrics (1), to the analysis of chemical data.
A major objective of the application of such methods to air quality data is
catagorization or classification of samples. This may be according to site of
origin, chemical manufacturer, etc., the assumption being that a given residue
contains information of a "fingerprint" nature which can be used in the identifica-
tion of the samples. One of the first methods developed for such studies is
discussed by Fisher (2) who at the time was attempting to catagorize agricultural
crops according to varying climatic conditions. Fisher carried out monitoring
127
-------
experiments on plants as a function of rainfall, temperature, etc., and developed a
method for optimization of interclass .variance. His data were characterized as
many samples described by a few variables.
In the early 1970's a number of attempts were made to apply Fisher's methods
of discriminant analysis to chemical data. Such data were of much different
structure than that of Fisher. The data were characterized as few samples
described by many variables. The results of these applications were less than
optimal.
This deficiency was realized by Wold (3) who developed a method of
classification based on principal components modelling of multivariate data, the
SIMCA method of pattern recognition (3). This method, not being sensitive to data
characterized as few sample relative to the number of measured variables, is the
method of choice for analysis of such data. A number of the presentations which
follow are variations of this method.
There are a iew short-comings of principal components modelling if informa-
tion beyond classification or catagorization is desired. This is the case in which
calibration and quantification of the components in a complex mixture is desired.
An extension of principal components analysis, the partial least squares (PLS)
method has been developed for this application. One presentation on the program
will deal with this method.
With the advent of powerful microcomputers it is possible for such data
analytic methods to be applied to routinely so such methods of data analysis will
soon be "built" into the analytical process.
128
-------
REFERENCES
1. Kowalski, B. R., Chemistry and Industry, 882-884, 1978.
2. Fisher, R. A., Statistical Methods for Research Workers, Boyd, London, 1978.
3. Wold, S., Pattern Recognition, 8, 127-139, 1976.
129
-------
CLUSTER ANALYSIS APPLIED TO INDIVIDUAL ATMOSPHERIC PARTICLES.
T. W. Shattuck , M. S. German! , P. R. Buseck. 1. Colby Col-
lege, Waterville, Maine 04901. 2. Walter C. McCrone Assoc., 2820
S. Michigan Ave., Chicago. Illinois 60616. 3. Chemistry and
Geology Departments, Arizona State University, Tempe, Arizona
85287.
Individual aerosol particle analysis using the analytical
scanning electron microscope, with its wealth of information on
the size, shape, morphology and chemical composition of parti-
cles, is an excellent tool for source attribution and studies of
particle dynamics (J.). The elemental composition is determined
using X-ray energy dispersive analysis (EDS). The recent intro-
duction of completely automated analytical scanning electron
microscopes (ASEM) extends the usefulness of this technique to
survey studies involving many sampling sites and sampling
periods. In survey studies individual particle analysis is par-
ticularly valuable since it provides information on the distribu-
tion and speciation of elements within the sample as well as
overall elemental contributions. The vast amount of information
from the ASEM presents a challenge to the analyst, since the
characterization of about 1000 particles for 30 elements is
necessary for each sample. The problem then is to condense the
mass of information into a more concise form, without losing
information on the elemental distribution between the particles.
Cluster analysis is a useful statistical tool for identifying
the types of particles that occur in a sample. In cluster
analysis particles are grouped into clusters of similar composi-
tion (£). Each cluster identifies a type of particle occurring in
the aerosol. Every cluster is represented by a centroid, which
is the average composition of all of the particles in the clus-
ter. The number of particles assigned to each cluster may be
displayed as a particle number versus particle type histogram.
Particle number versus particle type histograms are an easy way
to monitor changes between sampling sites and sampling periods.
As such, these histograms represent a tremendous simplification
and reduction in the volume of data, without excessive loss of
information.
In order to verify a cluster analysis scheme, two things must
be shown. 1) The range of particle types determined from a
representative aerosol sample must be realistic and inclusive.
Realistic clusters will give centroids that are identifiable as
relatively homogeneous mineral or anthropogenic particle types.
The results will be inclusive if few of the particles in a sample
remain unassigned. In this study, a sample from the Phoenix
aerosol will be used for this purpose. 2) The compositions of
the centroids and the corresponding cluster assignments should
reasonably reproduce bulk analysis results. This mass balance
comparison is carried out by analyzing individual particles of a
well characterized standard sample. In this study, USGS granite
and basalt samples are used.
130
-------
Results
Aerosol Particle Types The results of cluster analysis on a
representative sample from downtown Phoenix are given in Table I.
The details of the cluster analysis are given below. The major
particle type was quartz, which accounted for 19% of the parti-
cles. Various alumino-silicate types were the next most abun-
dant. However, the abundances of these particle types vary
widely from site to site. Particles with unusual composition were
not assigned clusters. Many particles rich in heavy metals were
found in the unassigned group.
Table I. Cluster Composition for Representative
Phoenix Aerosol Sample
Elemental
Composition
Similar Mineral
% Abundance
Si K Al Fe
Si Al K Fe
Si Al Fe Ca
Si Ca Fe Al
Si Fe Al K
Si
Fe Si Al Mg
Fe
Ca Si Fe
Ca
Ca S Si
Ca Si Fe
Ti Si
Ti Fe Si
K Cl Si
Pb Cl Br
Pb Si
Fe Zn Si S
S Si Na
Unassiened
Si indicates
or may be due
tion edge).
Orthoclase
Muscovite
Albite/Montmorillonite
(Epidote)
Biotite
Quartz
Ripidolite/Chlorite
Magnetite
Pyroxene
Calcite
Gypsum
(Tremolite/Actinolite)
(Rutile)
that Si may be present in the
to a spectral artifact (carbon
7
15
14
6
4
19
2
7
4
3
1
2
2
0.5
0.5
3
3
1
1
4
particles
absorp-
the cluster.
only a possible mineral assignment for
Many clusters were well resolved, however the alumino-silicate
clusters in the Phoenix samples were probably mixtures of several
mineral types. The minerals indicated in Table I have been iden-
tified in the Phoenix aerosol in the 5 to 50 micron diameter size
range Q). They were listed not as absolute assignments but as
suggestions for the most prominent mineral type in the given
cluster. Obviously many of the particles were not necessarily
crustal in origin. For example, there were many sources of iron
and iron oxide particles other than magnetite.
131
-------
Mass Balance Elemental composition and size data were collected
for USGS standard samples of Granite, G-2 and Basalt, BHVL-1.
The data set contained particles with sizes in the 1-10 micron
range. The compositions, obtained from EDS region-of-interest
integralSt were converted to relative abundances, with the sum of
the concentrations equal to one. No other variable scaling was
used. Seven methods were tested for choosing seedpoints, and gave
comparable results. Initially, 15 seedpoints were used. The For-
gey variety of K-means cluster analysis (2.) was then used to
define the clusters. The number of clusters to retain, of the
initial 15, was determined in the following way. Cluster signifi-
cance was tested using the ratio of the between clusters sum of
squared distances to the within clusters sum of squared distances
(JJ) • The cluster with the most test failures was then rejected.
Using the remaining centroids as seedpoints, cluster analysis was
repeated. This process was continued, rejecting one cluster at a
time until the number of unassigned particles began to increase
and the number of significant clusters began to decrease. Pairs
of isolated but overlapping clusters were joined and clusters
with few members were deleted to complete the centroid set.
A few representative particles from each cluster were chosen
and reanalyzed, using a full ZAF correction scheme. These compo-
sitions, the cluster assignments and the size of each particle
were used to calculate the overall composition of the sample.
Each particle was assumed to be a sphere for volume calculations,
and the density of the particles in each cluster was approximated
by that of a similar mineral.
There were eight clusters found for the granite sample:
quartz, two plagioclase, orthoclase, biotite, muscovite, chlorite
and apatite. In addition several metal sulfide clusters where
found, but they were deleted from the centroid set because of low
abundance. For the basalt sample, seven clusters resulted:
quartz, plagioclase, two calcium pyroxenes, two titanium pyrox-
enes and a glass phase containing large amounts of potassium and
a wide mixture of other elements. The mass balance results are
given in Table II.
Table II. Mass Balance Comparisons for Granite G-2 and
Basalt BHVL-1. Weight fraction normalized to Si.
Granite
Particle EDS
Accepted
% Error
Basalt
Particle EDS
Accepted
| % Error
! Na I
! .042 I
! .059 !
I -29 I
I !
i i
|T044^
I .044 |
! !
Mg
.015
.011
36
.125
.143
-13
! Al
I .234
! .222
! 5
!
i
.283
1 .276
1 3
Si
1.0
1.0
1.0
u-1'0—.
Fe
.029
.035
-17
i
.160
.220
-27
K
.076
.064
19
i .003
.011
L_ -73
! Ca I
! .029 !
! .028 !
T 4 i
! !
I i
\ .220J
| .228 I
! -4 !
The agreement between the particle and accepted results was
132
-------
quite good considering the level of approximations involved,
except for Fe and K. In the basalt, potassium occurs predom-
inantly in a glassy phase. When crushed this glassy phase pro-
duces very small particles which fall below the cut-off limit
used in the data acquisition step. Therefore, it is not surpris-
ing that potassium is under-represented in the mass balance
results. The most important mineral phase for iron is biotite
mica. These particles were usually flat plates. Particles of this
type will show the biggest errors when using standard ZAP correc-
tion algorithms.
Summary
The particles in the Phoenix aerosol fell into relatively
homogeneous composition ranges which could be identified using
cluster analysis. The resulting clusters gave a satisfactory
range of particle types with only 4% of the particles left unas-
signed. Mass balance comparisons using the same methods on stan-
dard samples showed sufficient agreement between the individual
particle results and bulk analyses.
The usefulness of these results for a long term sampling study
was tested in the following way. Using the clusters found,
nearest neighbor discriminant analysis was used to characterize
samples from the Phoenix aerosol, taken over a two week period.
The resulting particle number versus particle type histograms
were used in factor analysis and showed strong correlation with
upper level wind directions.
The implications of the success of cluster analysis for aero-
sol particles is twofold. First, the analyst need not know or
specify in advance the kinds of particles that are expected in a
given area. The structure in aerosol particulate data is suffi-
cient to identify the underlying particle types. Secondly, the
large volume of data available is easily condensed into a more
compact form while retaining all of the advantages of individual
particle analysis.
Acknowledgments
Financial support for this work was provided by grants ATM-
8022849 and ATM-8404022 from the Atmospheric Chemistry Division
of the National Science Foundation.
Literature Cited
1. Post, J. T.; Buseck, P. R. Environ. Sci. Technol., 1984, 18,
35-42.
2. Massart, D. L.; Kaufman, L. "The Interpretation of Analytical
Chemical Data by the Use of Cluster Analysis"; Wiley: New York,
1983; p. 107.
3. Fewe, T. L.; Pewe* E. A.; Pewe, R. H.; Journaux, A.; Slatt,
R. M. Spec. Pap.—Geol. Soc. Am. 1981, No. 186.
4. Hartigan, J. A. "Clustering Algorithms"; Wiley: New York, 1975,
p. 97. ..-.-.
133
-------
ABSTRACT
EVALUATION OF EMISSION DATA FROM HAZARDOUS WASTE INCINERATION
TESTS USING PATTERN RECOGNITION TECHNIQUES
By
Karin M. Bauer and Dennis D. Wallace
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110
To be presented at the Fifth Annual Symposium on Recent Advances
in the Measurements of Air Pollutants,
Raleigh, North Carolina, May 1985
At each of eight incinerator facilities a series of test runs were con-
ducted to examine particulate and HC1 emission rates from the incinerators
and to characterize the performance of these incinerators with respect to
destruction and removal efficiency (DRE) for principal organic hazardous
constituents (POHCs).
One of the chief concerns was to characterize the incinerator inputs
and outputs of POHCs. Samples of waste feeds, control system makeup waters
(as appropriate), stack gases, and control system effluents were collected
during each run. Auxiliary fuels and incinerator ash were also sampled for
selected runs. Samples were evaluated for concentrations of Appendix VIII
organic constituents using gas chromatography/mass spectrometry (GC/MS). In
addition, waste samples were analyzed for moisture, ash and chloride con-
tent, viscosity, and heating value. Stack gas samples were analyzed to de-
termine particulate and HC1 concentrations and concentrations of 02, CO, and
C02. Total hydrocarbons (THC) in the stack gas were continuously monitored.
Incinerator operating characteristics such as combustion temperature, resi-
dence time, total heat input rate, and control device parameters were also
monitored during each test.
134
-------
One focus of the study was to analyze and quantify the impact on DREs
of the parameters listed above. The analyses examined relationships between
DREs and these various parameters and/or identified patterns among the ob-
servations within the two subsets of the data base: those observations for
which DRE was above 99.99%, and those for which DRE was below 99.99%. Given
the large amount of data, pattern recognition techniques were the most logi-
cal choice for reducing the dimensionality of the data for simpler represen-
tation and interpretation and to identify those important variables that de-
termine experimental behavior.
Correlation, factor, principal components, discriminant, and cluster
analyses were the main techniques used in the study. Engineering knowledge
of the incineration process was used to define several sets of variables on
which the principal component analyses were conducted. Typically, these
analyses considered sets of process and waste feed parameters, continuously
monitored stack gas parameters (e.g., 02, CO, C02, THC), stack gas concen-
tration of POHCs and DREs. Through the multiple analyses we examined the
impact of the removal of variables such as input and/or output waste con-
centrations or DRE on the resulting principal components, the amount of
variance different parameters explained, and within each component, the
change in hierarchy and relative importance of the individual variables.
Discriminant analysis was also used in order to find a linear combination
of significant parameters that would best discriminate between those obser-
vations that achieved 99.99% DRE and those that failed. Cluster analysis
was performed on the compounds to determine whether certain compounds or
groups of compounds tend to cluster in terms of their presence and total
concentration in the stack gas.
The SIMCA-3B Pattern Recognition software package for microcom-
puters, as well as the SAS and BMDP statistical software packages, were ex-
tensively used in the analyses and evaluation of the data. The results of
the different techniques will be presented, compared, and interpreted.
135
-------
DISTRIBUTED CHEMICAL ALARM SYSTEMS FOR PLANT SECURITY
BASED ON ION MOBILITY SPECTROMETRY (IMS)
Joseph E. Roehl
ALLIED Bendlx Aerospace
Bendix Environmental Systems Division
1400 Taylor Avenue, P.O. Box 9840
Baltimore, Maryland 21284-9840
136
-------
Abstract
This paper describes work being performed at ALLIED Bendlx Environmental
Systems Division (ESD) to develop a chemical alarm network based on smalI Ion
mobility spectrometers, which may be distributed Inside a facility or on Its
perimeter. Ion mobility spectrometers require a fair degree of signal proces-
sing and the processing burden must be divided up between local processors,
which are attached directly to the Ion mobility eel Is and a central micro-
processor which executes an alarm algorithm, displays alarm messages, and
Interfaces to other computers or display systems If desired. The paper
discusses Ion mobility spectrcmetry (IMS), chemicals which (IMS) may be used to
detect, the specifications for the alarm network, and the alarm algorithm In
general.
137
-------
INTRODUCTION
The detection of potential ly hazardous chemical leaks or spll Is Is a
difficult problem especially when the material Is highly toxic and smal I con-
centrations must be detected. ALLIED Bendlx Environmental Systems Division
(ESD) has developed an alarm technology based on Ion mobility spectrcmetry
(IMS), which Is sensitive to many hazardous chemical compounds at the parts per
billion (PPB) level and Is suitable for monitoring applications. Figure 1 Is a
list of the chemical vapors which may be detected by IMS. In a typical scenario
alarm units would be placed Indoors and out-of-doors around processing plants,
tank farms, valves, loading and unloading facilities, and the perimeter of a
plant site. The alarm units would contain a sensor and a microprocessor which
communicates with plant operators through a central alarm console. Very low
concentration vapors released from leaks or spllIs would be sensed by the alarm
units and plant operators would be alerted.
An alarm system used In this type of applIcatlon should have the following
characteristics:
Spec I f Iclty - The alarm system should only alarm to target
chemicals. It will be used In a dirty Industrial environment
and should reject other chemical compounds.
Sensitivity - The alarm should be sensitive to target
chemicals at wel I below their hazardous levels.
Logistics - The alarm should be easy to test, relI able, low
power, and required a minimum of maintenance.
- Cost - The alarm should be Inexpensive so that many alarm
stations can be Installed around a site.
138
-------
An IMS system Is extremely attractive for this type of applIcatlon because
It has the following attributes:
- High sensitivity
No wet chemicals
Minimal maintenance
Low power
Operates over a wide environmental range.
Competitive techniques rely on wet chemical systems which require frequent
maintenance, relatively large gas chromatographs, or more complex technologies.
There are two potential disadvantages to IMS;
1. The device utilizes a radioactive source (NI63-Beta emitter)
which requires a license and seme special handling, and
2. The IMS instrument Is not as selective as other chemical
Instruments and requires additional signal processing of the
sensor output to determine if a target chemical is present.
The former problem Is easily overcome and the latter problem has been
overcome In the ESD design through the development of sophisticated signal
processing techniques. Prototype IMS alarm systems have been constructed and
tested at ESD with very favorable results. They are stand-alone systems however
and must be reconfigured into sensor systems for applIcation to a fixed
facility.
THE ION MOBILITY SPECTROMETER
The heart of the alarm system described here Is an Ion mobility spectro-
meter (IMS). Ion mobility spectrometers In various forms have been around for
many years. Figure 2 is a pictorial diagram of a Bendfx ESD design. The cell
139
-------
Is made up of two contiguous cyllnderical sections - the reaction tube and the
drift tube separated by a shutter grid. Both tubes are made up of conducting
rings, separated by Insulators, which are biased using external resistors to
provide a uniform potential gradient through both cylinders. When gas molecules
enter the reaction tube they are Ionized by the NI63 radioactive source. Either
positive or negative Ions are attracted down the reaction tube toward the drift
tube depending on the direction of the electric field vector. Electric field
vector direction may be changed by reversing the polarity across the resistor
network. A shutter grid Is placed between the reaction tube and the drift tube
which functions like a camera shutter al lowing Ions to pass through It Into the
drift tube when open and annihilating ions when closed. The grid Is made up of
parallei wires which may be electrically shorted together at a potential equal
to the magnitude of the field at the appropriate point In the coaxial electric
field. This al lows most Ions to pass. Normal ly the grid Is;'biased to create a
field perpendicular to the field through the cylinders thus annihilating the
Ions. Ions that pass through the grid are detected at the end of the drift tube
when they Impact on a plate mounted In a Faraday cage. The Faraday cage sur-
rounding the plate Is necessary to reduce currents Induced in the plate circuit
by stray charge and the Ions moving down the tube. An electrometer amplifier
mounted on the back of the eel I amplifies the minute current (10"12A) at the
plate and converts It to a voltage signal.
Figure 3 Is a typical IMS positive mode signature with the trigger pulse
superimposed on It. An IMS signature Is a plot of Ion current versus time.
When the trigger pulse momentarily (10 microseconds) opens the shutter grid a
cloud of ions Is al lowed to pass Into the drift tube. The Ions In the cloud
140
-------
under go collistens and separate themselves out by molecular weight and
colllslon cross section. The result Is that the cloud of Ions entering the
drift tube separates Into a number of clouds each with a Gaussian profile. When
successive clouds Impact on the plate they produce Gaussian shaped current peaks
In the plate circuit. A Gausslem peak can be completely described by the three
parameters - time of peak maximum, ful I width half maximum (FWHM) of the peak,
and height of the maximum. In an IMS al I peaks tend to have the same FWHM, the
height of a peak Is related to the concentration of an Ionic species, and time
of the peak maximum relative to the center of the trigger pulse Is dependent on
the particular Ionic species and Is called the drift time. Drift time for a
given Ionic species Is dependent on a number of parameters such as length of the
drift tube, atmospheric pressure, and temperature but may be normalized to yield
reduced mobll Ity fJ.0. Reduced mobll Ity Is dependent only on the particular Ionic
species. Unfortunately, different Ionic species can have similar reduced
mobilities and IMS Is a low resolution technique.
In order to use an IMS as the sensor In a chemical alarm system which will
only alarm to certain chemical compounds, additional specificity must be built
Into the system. In laboratory applications IMS is used In a controlled
environment with nitrogen as a carrier gas. Sample Is Introduced Into the gas
stream for analysis. In an alarm application, however, a sampling system
"snlfs" ambient air which may contain a number of substances which produce IMS
peaks. In order to make IMS practical for an alarm application additional
chemical, electronic and software filtering must be provided In the system. It
Is also very desirable In an alarm system to eliminate the requirement for
special carrier gas. In the ALLIED Bendfx ESD design nitrogen Is replaced by
141
-------
clean air circulating behind a permeable membrane. In addition, the IMS signal
Is analyzed with a microprocessor. Again, refer to Figure 3. It Is
representative of an IMS signature from an IMS eel I embedded In the BSD alarm
system. Note that there are four peaks Illustrated In the figure. The first
two peaks are usually present In IMS signatures. They are formed fron the
lonlzatlon of ammonia and water present In the air. In clean air only the NH
and Ht peaks would appear. When a chemical compound with a higher proton
affinity permeates Into the system, charge Is transferred from the H+ Ions to
form Ions of the new material called product Ions. The total charge In the Ion
cloud and thus the total area under the IMS signature Is theoretically equal to
a constant. As material enters the IMS eel I It starts at a low concentration
producing a moncmer peak which Is a combination of the new product Ion and
water. As the concentration of the material In the system rises the moncmer
peak grows at the expense of the reactant Ion peak. Eventually a dlmer peak
appears and grows at the expense of the reactant Ion and moncmer untlI at very
high concentration only the dlmer appears. If the material Is now stopped from
entering the system, the concentration diminishes and the process Is reversed.
Figure 4 Is a pictorial diagram of a Bendlx ESD-desIgned IMS eel I which
contains the membrane holder, lonlzatlon source, reaction tube, and drift tube.
Sample gas flows Into the eel I, washes the membrane, and exhausts from the eel I.
Clean dry air Is pumped over the back of the membrane and Into the IMS reaction
tube. This flow of air Is cal led the carrier flow.
A counter flow of air passes over the Inside walI of the cyllnderical drift
tube called the drift airflow. This airflow keeps the wal Is of the drift tube
clean.
142
-------
ALARM ALGORITHM
A major problem In using IMS for alarm applications Is the specificity
required for a practical alarm system. Plant operators will fend to disregard
any alarm system with even a low false alarm rate. Two features In the Bendlx
design provide the required alarm specificity: the membrane and the alarm
algorithm Implemented In microprocessor software. Figure 5 is a block diagram
of the alarm algorithm. Analog data from the cell Is digitized and analyzed by
the microprocessor using the alarm algorithm which consists of a feature
extractor and a pattern classifier. The feature extractor filters out high
frequency noise, deconvolves partially overlapped peaks, measures the position
and height of the deconvolved peaks, and normalizes the peak position relative
to a reference peak to eliminate temperature and a-fmospherlc pressure effects on
the measurement. The normalized peak position is related to the reduced mobili-
ty by a constant. A table of normalized peak positions and peak amplitudes is
passed to the pattern classifier portion of the algorithm. The pattern clas-
sifier portion of the algorithm Is a two-step process. In the first step
normalized peak positions are checked against tolerances around the expected
reduced mobilities of the target chemicals. In the second step the set of peak
positions and amplitudes constituting the sample measurement Is treated as a
vector. The dot product between the sample vector and a set of linear discrimi-
nant functions is calculated. The set of linear discriminant functions is
predetermined by collectlng data on the target chemicals and interferents and
calculating a linear discriminant function which separates them.
143
-------
STAND-ALONE ALARM SYSTEMS DEVELOPED AT ESD
Figure 6 Is a block diagram of a prototype stand-alone alarm system
developed at Bendlx Environmental Systems Division (ESD). The system Is made up
of five basic blocks:
1. Sampling System - Draws ambient air Into the system, provides chemical
filtering by use of a membrane, and Interfaces with Internal air system.
2. IMS Gel I - Transduces the chemical signal Into an electrical signal.
The output signal Is Ion current versus time.
3. Microcomputer - Digitizes and analyzes IMS cell signal. Implements an
alarm algorithm based on pattern recognition and Interfaces with the operator.
4. Pneumatic System - Contains a dual diaphragm pump which moves air
through the system and a scrubber which scrubs the reclrculatlng air passing
from the sampling system to the IMS eel I. An Important advantage of the IMS
Instrument over other Instruments Is that It requires no external gas supply or
vacuum system.
5. Operator Interface - The operator Interface Includes a display to
alert the operator of alarms and the Instrument's status, and a three-position
toggle switch for the operator to select Instrument mode.
DISTRIBUTED ALARM SYSTEMS UNDER DEVELOPMENT AT ESD
Figure 7 Is an Illustration of the distributed IMS system under development
at ALLIED Bendix ESD. This system Is made up of remote stations containing a
sampling system, IMS cell, pneumatic system, and one chip microprocessor, and a
central computer which acts as an operator Interface. The local microprocessor
control the IMS cells, format data, and exercise bulIt-ln-test functions. A
number of local units can be tied to a single central annunciator. The central
144
-------
portion of the chemical detection and alarm algorithm. Target chemical sub-
stances can easily be added or deleted from the alarm list In the annunciator.
CONCLUSIONS
Chemical alarm systems based on IMS appear to be feasible and cost
effective. ALLIED Bendlx ESD Is actively pursuing an In-house development
program to demonstrate a distributed alarm system.
145
-------
REFERENCES
Roehl, Joseph E., "A Microprocessor-Controlled Chemical Detection and
Alarm System Based on Ion Mobility Spectrometry", IEEE Transactions on
Industrial Electronics, Vol. 32, No. 2, May 1985.
C23 Campbell, D.N., Carrlco, J. P., and Spangler, G.E., "A Compact Ion
Mobility Spectrometer System", Paper 91, Pittsburg Conference on
Analytical Chemistry and Applied Spectroscopy (Atlantic City, N.J. 1983,
146
-------
CO
CD
c
03
._ co
•a
CD
"CO
c
CD
CD
o
co
•a
3
g.
E
o
o
CO
— »
o
uo 0
to
t5
ZJ
T=3
O
•o S-
.<£ 0.
cn
CD
O "cj> ^_, *t^
•S .E ° *"" <~
*•»• ^™ «4^J <^~ C^>
o < 25 Q. S
cc o
C7> -a
— -a
ui E
O UJ
Q_
to a)
o o
Q. CO
13
cn
149
-------
(1)
o
"O
c
4)
CO
3
Ol
150
-------
N011IN9003d
1N39V
o:
<
UJ
ttt-t t
o
FEATURE EXTRACTOR 1 CLASSIFIER
01
3IAII1 1:
UJ
01
H-
CD 1
0) g
CL %
*= 4 ..,:*„.
eg 5;
U-l CO
j
k. ^
CD
1
t
IVd
k
UJ
a
Ll_ ^
1 1 i U_
0= L-.
-1
Q.
^=
dida"
k <
k
»• an^OM
CD
— 1 — 1 ^
••C -«C CO
1 — SE CO
k A
UJ
< " H
k
E
O
O)
O)
151
-------
cc
Ul
J_
to
2
cc
III
X
1-
H
lil
_J
Z
_J
o
cc
h-
z
o
o
cc
Ul
j^
Ul
X
1-
Ul
z
UJ
CO
*p
cc
UJ
X
1—
cc
Ul
f-
UJ
X
Ul
z
cc
CO
Ul
_J
cc
H
z
0
o
cc
UJ
1-
s
X
Id
z
cc
CO
Ul
w
w
^
w
cc *
1- .
to K
2 i
£ ° *
-t- ce kJ
F uj o
•^ i— CD
rr ...
rv* ^ /v*
* iTi Q-
UJ ^ H
UJ IJ /
X UJ x
a "/ *
f T
1
\
^
r
^
OPERATOR
A
CO
m
^
i
u
2
_|J
mp
II
CO
2(
53
ou.
o
z
_J
0.
CO
Id
U
K
UJ
h-
z
L
r
cc
Ul
OL
o
u
k ^.-
^v
1
Ul
u
^
2
LU
H
to
CO
o i
ft cc
1 H
Q, Ul
02 i
Ul Ul $
con <
3> A
u co ^tm^
•
-1— ^^^
/ cc \
/ UJ \
/ceo \
J <3 N
ou. \
1
t /
^
UJX O
/ * It 1 k. —"
OUJ hv H 2
^ 2 H-
3 tO
III >-
z to
a.
>< ^
>• Ul (y- 1 •
y, 00< ' 1
/\ 1 , •
N s \
-—'' \
UJ
to
ai
§1
•f
u
to
u.
o
CO
*
"n
1
1
1
I
1
1
1
J
CO
u
cc
H-
o
UJ
_l
LU
t
c
O
T3
0)
(TJ
CD
e
0)
E
TO
(U
3
Ol
UJ CC
CO <
il
< X
CO UJ
152
-------
\
t/1
>.
to
CO
T3
-------
PERSONAL MONITORING: OVERVIEW*
Alan R. Hawthorne
Measurement Applications Group
Health and Safety Research Division
Oak Ridge National Laboratory
The need for low-cost personal monitoring devices is great. These
monitors have been and will increasingly be used for numerous applications.
Examples are investigations of total exposure assessments where personal
monitors are used to measure an individual's pollutant exposure during the
complete day. Other applications include time-dependent exposure profile
studies and microenvironment apportionment studies. A final example, and
perhaps the most extensively used application, is for worker monitoring in
industrial hygiene programs.
There are several fairly obvious requirements for an ideal personal
monitor; however, meeting all of these requirements in practice is often
elusive. Following is a list of some of the important requirements.
1. Personal monitors should be portable—small, light weight, and
unobtrusive.
2. They should ideally be of low cost.
3. They should provide adequate sensitivity for the monitored pollutant.
4. They should be free of interference from other pollutants.
5. They should require minimal user attention and not distract from his
routine behavior.
6. For some applications, they should provide the capability for time-
dependent exposure profiles.
* Research sponsored by the U.S. Department of Energy, under contract DE-
AC05-840R21400 with the Martin Marietta Energy Systems, Inc.
By acceptance of this article,
publisher or recipient acknowl
the U.S. Government's right t
-------
These requirements are often hard to meet for many pollutants of interest
and trade-offs must be made the various requirements. Some of the
requirements are similar to those of fixed-station and portable monitors while
others are substantially different.
Personal monitors can be classified into general categories based on
their operation and the type of information provided. One classification
distinguishes whether the monitor is an active device using a pump and
requiring a power source or a passive device sampling by diffusion or
permeation and requiring no power. The advantages of a passive device are
obvious, and such a device is often the preferred monitor provided that
important factors such as sensitivity, accuracy, and time resolution are
adequately addressed. Depending on the particular application, monitors
capable of providing integrated or time-dependent exposures may be
preferentially desired. A second useful classification is whether the
personal monitor provides time-dependent (in some cases real-time)
information, or whether the result is an integrated or time-weighted average
response. Most often these statements lead most readers to infer 3 passive
monitors that yield time-dependent data.
There are several pollutants for which commercial or developmental
passive monitors exist. Examples include formaldehyde, nitrogen dioxide,
sulfur dioxide, radon, and volatile organic compounds. Passive monitors are
most often of the integrating type and frequently require subsequent
laboratory analysis to provide the exposure estimate.
To advance the performance of passive devices as personal monitors, there
are several areas that need attention. Perhaps the greatest need is to extend
the number of pollutants for which passive monitors are available. Improved
sensitivity is often a limiting factor in the acceptability of passive devices
and deserves continuing attention to improve performance. Another important
factor that would enhance the acceptability of passive monitors is a better
assessment of factors that affect precision and performance. Intercomparisons
of monitor performance under field conditions would help address this issue.
155
-------
Integrating samplers based on pumping air through collection devices
constitute another large class of personal monitors. A wide variety of
collection devices are used, including sorbent traps (for collecting volatile
organic compounds), filter media (for collecting particulate matter), and
impingers (usually used only as a last resort).
Items that would advance the performance of pumped collection systems
include improved pump performance, improved sorbents, and better analytical
techniques. Smaller, quieter, higher flow rate pumps (available at low 'cost)
with extended battery life would address many of the problems encountered in
this type of personal monitoring. A better selection of sorbents that have
high sorption efficiency and yet can be quantitatively desorbed would be
desirable. Finally, improvements in performance of the analytical process
would result in less material being needed from samplers, thus lessening the
requirements of the monitoring device.
Direct-reading personal monitors have both advantages and disadvantages
to offer the user. Among the advantages are the ability to provide real-time
exposure profiles and to monitor peak levels. The immediate response
capability of this type of monitor may be critical in some applications, and
in others the ability to avoid frequently long delays associated with
laboratory analysis is certainly desirable. Frequent disadvantages of real-
time personal monitors are relatively high cost, size and weight greater than
for passive monitors, and limited availability for many pollutants.
Needed advances for direct-reading personal monitors include reduction in
cost, increased miniaturization resulting in reduced size and weight, and
development of a wider selection of microsensors. Perhaps the most
challenging and difficult aspect will be development of low-cost, yet
sensitive microsensors compatible with today's microelectronics. There have
been dramatic advancements in the supporting microelectronics required for
direct-reading personal monitors and portable data loggers. Dramatic cost
reduction, increased capacity and capability, and decreased power requirements
have been achieved for microprocessors and solid-state memory. Important
advances have also been made in analog-to-digital and digital-to-analog
converters as well.
156
-------
I believe that we will see personal monitors play an increasing role in
characterizing exposure levels in ambient, indoor, and occupational
environments. The need clearly exists for such capability, and continuing
advancements in various technologies will help provide desirable personal
monitors to meet these needs.
The next several papers will provide examples of some of the recent
advances in various personal monitors. They include papers on pumped sampling
methods, passive devices, and direct-reading monitors.
157
-------
INDOOR AEROSOL IMPACTOR
William Turner and John Spengler
Department of Environmental Science & Physiology
Harvard School of Public Health
Boston, MA
Virgil Marple
University of Minnesota
Minneapolis, MN
INTRODUCTION
An aerosol impactor and constant flow pumping system designed for indoor
sampling has been developed as part of the ongoing Harvard Air Pollution
Epidemiologic Study. This impactor is designed to sample at a flow rate of 4
1pm. A double impactor was used to increase efficiency of particle separation.
Two units have been designed to provide a sharp 50 percent cutoff at 2.5 urn and
10.0 urn. Testing has also been conducted to determine the effect of flow rate
on the size cut characteristics. The system has been designed to provide
uniform deposition on 37 mm Teflon filters mounted on 2" x 2" plastic holders.
The collected material is particularly suitable for automated analysis using XRF
and other techniques.
This device was deployed in conjunction with a constant flow sampler in 300
homes during the winter of 1984-85 in Watertown, MA and will be used to make
measurements in 2000 homes in the next four years. The design criteria and
results of initial testing are presented. The system has been modified for out-
door sampling and is currently being run collocated with a PM-10 dichotomous
sampler. Geomet Technologies, Inc., is currently doing side-by-side testing of
several indoor and personal samplers including this indoor unit with results to
be available by January, 1986.
DESIGN CRITERIA
The impactor was designed to meet several criteria. The primary intended
use of the impactor is for the sampling of fine fraction (< 2.5 urn) or inhalable
fraction (< 10 urn) particle sampling in the indoor environment. The flow rate
chosen was 4 1pm. The rationale for this flow rate was that it was technically
achievable with both battery powered flow controlled pumping systems and line
power. This flow rate allows the sampling of about 5.5 m3 in a day and 40 m3
158
-------
in a week. This will facilitate improved accuracy in gravimetric measurements,
Beta-gauging operations, and when using XRF automated elemental analysis tech-
niques. Previous studies have indicated that 24 hour indoor home concentrations
can range between 5 and 300 ug/m3 (Spengler, 1981; Ju and Spengler, 1981). A
flow rate of 4 1pm and minimum concentrations will acheive a mass loading of
25 ug. This is the minimum detectable limit of either the beta-gauge or the
gravimetric method with the filter currently used.
The impactor has been designed to use dichotomous sampler type 2" x 2" PTFE
membrane filters. Both the rigid mount type (supplied by Membrana, Inc., Ghia
Div.) and removable mount type filter slides supplied by Beckman, (42 mm filter
disks supplied by Ghia) can be used in the present configuration. An option
which will also allow the "Sierra Type" dichotomous filters to be used will be
available by 12-1-85. The impactor plates are reusable and have been tested to
be adequate for one week sample periods.
UNIFORM DEPOSITION
To utilize automated weighing and analysis procedures which have previously
been developed by the US EPA for dichotomous filter samples, a uniform deposi-
tion criteria was established (Dzubay and Stevens, 1975). This criteria was
defined as the deposit density per CM2 of the filter shall not exceed 10% of the
averaged deposit of filter.
RESULTS
To date, impactor size cut and uniformity of deposition have been charac-
terized by researchers at the Harvard University School of Public Health, the
University of Minnesota and the University of Florida. The results of the 2.5
urn and 10.0 urn Minnesota size cut characteristics tests are presented in Figures
1 and 2. The Minnesota tests were performed with solid particles using fluor-
escent dyed techniques (Stober and Flachsbart; 1973). The results of the
Harvard and University of Florida testing are similar and appear in Figure 3.
Testing for uniform deposition showed the initial plate to filter distance
to be inadequate for the 10.0 urn nozzles. Lengthening the distance to its
present spacing has been shown to meet the uniform deposition criteria for both
nozzle sizes. The effect of flow rate on particle cut size has been tested for
the 2.5 urn impactor. The 50% cut point was determined at 2, 3, 4, 5, and 6 1pm.
The results are presented in Figure 4.
159
-------
Tests were conducted by researchers at Harvard to determine precision
estimates for both impactors. Thirty samplers were run in the same environment
for two separate 7 day sampling periods. A grand and associated statistics
were determined. The results of these tests are presented in Table 1. Using
the traditional two standard deviations for precision, the indoor samplers can
measure ±0.7 ug/m3 and ±1.5 ug/m3 respectively.
An experiment was also conducted by Harvard researchers to test the effects
of a "rain hat" for outdoor use. Twenty 10.0 urn samplers were run outdoors
during non-rain days. Ten heads were run with rain hats and ten without for a
total of 64.9 hours. As can be seen from these statistics (Table 2), the sampler
types agree quite well. The 95% confidence limits are in good agreement.
CONCLUSIONS
The cut points of the impactors have been determined, within experimental
error, to be at the original design criteria of 2.5 urn and 10.0 urn. Precision
tests to date have shown the samplers to be well suited for indoor sampling.
Limited outdoor testing has shown the impactor and pumping unit to be adaptable
to the outdoor environment as well. Preliminary results of the Geomet
colocated testing imply that the sampler compares well to other methods.
One hundred samplers have been built to date and deployed for a six month
sampling period. Over two hundred pumping units and four hundred impactors
will be deployed in November, 1985. To date, these impactors combined with
their programmable, flow controlled, quiet sampling unit, have allowed us to
reliably collect samples in 280 homes in Watertown, Massachusetts. They will
be deployed in St. Louis, Missouri and Kingston/Harriman, Tennessee this fall
with similar performance expected.
ACKNOWLEDGEMENTS
The authors wish to thank Ken Robow of the University of Minnesota, Robert
Vanderpool of the University of Florida, and Janet Macher, Jack Price, Anthony
Majahad and Chan MacVeagh of the Harvard School of Public Health for equipment
testing. This work was performed under National Institute of Environmental
Health Science Grant NO. ES-01108.
160
-------
REFERENCES
1. Dzubay, T.G. and Stevens, R.K. Ambient air analysis with dichotomous
sampler and x-ray fluorescence spectrometer. Environmental Science &
Technology, 9:7, July 1975.
2. Ju, C. and Spengler, O.D. Room-to-Room variations in concentration of
respirable particles in residences. Environmental Science & Technology
15(5):595-598, May 1981.
3. Marple, Virgil. Verbal communication to William Turner, April 1985.
4. Speizer, F.E., Bishop, Y. and Ferris, B.G., Jr. An epidemiological
approach to the study of the health effects of air pollution. Proceedings
of the 4th Symposium on Statistics and the Environment, March 3-5, 1976,
Washington, DC pp. 56-68.
5. Spengler, J.D.,"Dockery, D.W., Turner, W.A., Wolfson, J.M. and Ferris,
E.G., Jr. Long-term measurements of respirable sulfates and particles
inside and outside homes. Atmospheric Environment 15:23-30, 1981
6. Stober, W. and Flachsbart, H. Evaluation of nebulized ammonium
fluorescein as a laboratory aerosol. Atmospheric Environment 7, 1973.
161
-------
TABLE 1. PRECISION TESTS
N
MEAN
MEDIAN
T MEAN
STD. DEV.
SE MEAN
MAXIMUM
MINIMUM
2.5 urn
ug/m3
28
9.673
9.624
9.657
0.355
0.067
10.662
9.114
10.0 urn
ug/m3
30
14.693
14.836
14.745
0.762
0.139
16.321
12.076
162
-------
TABLE 2. OUTDOOR SAMPLER TESTS 10.0 um
(RESULTS IN ug/m3)
MEAN
MEDIAN
T MEAN
STD. DEV.
SE MEAN
MAXIMUM
MINIMUM
BOTH
20
19.39
18.91
19.33
1.72
0.38
22.96
16.99
WITH HAT
10
19.36
18.99
19.21
1.57
0.50
22.96
16.99
WITHOUT HAT
10
19.42
18.82
19.31
1.94
0.61
22.34
17.41
163
-------
Aerodynamic Particle Diameter
Figure 1. Particle collection efficiency data for fln.il 2.5 >
version (W - 0.0960 Inch) at a flow rate of i I.PH.
100
Nozzle
Diameter, Test
Nozzle Inch Aerosol
D Original 0.234 Liquid
ORecommended 0.244 Solid
5 10 20
Aerodynamic Particle Dlamecer, ura
?lguru 2. Parcicle collection, efficiency data for ^original and
final 10 pm versions at a flow rate of 4 LPM.
164
-------
o
100
90
80
70
UJ
— 60
u_
u_
UJ 50
2
® 40
I—
O
UJ 30
O
20
10 -
= 2.5
_L
1.5 2.0 2.5 3.0 4.0
AERODYNAMIC PARTICLE D I AM E T E R.,JUM
Figure 3. Collection Efficiency Characteristics of the 2.5 Micron Impactor
Flow Rate, LPM
Figure 4. Parctcle cut-point for tnul .' •* >tr. impai-toi .is n funct KM\ of 1'low rate
165
-------
ANALYSIS OF NITRITE IN N02 DIFFUSION
TUBES USING ION CHROMATOGRAPHY
D.P. Miller, K.C. McGill, H.S. Lam, T.L. Shields
Department of Chemistry
Washburn University
Topeka, KS 66621
The subject of this report is the development of an analytical method to
analyze Palmes Tubes for nitrite ion. Palmes Tubes are passive diffusion control-
led samplers for nitrogen dioxide in the atmosphere, and the N0£ collected is
analyzed as nitrite ion(l).
Passive samplers for trace components in the atmosphere are inexpensive
to fabricate and deploy, and as a result they are utilized in large numbers when
it is necessary to determine the time average concentration over a large spa-
cial area. An additional advantage is that most passive samplers are small in
size, and when carried around by people, they measure the concentration in the
particular spacial area occupied by people. As a result of large numbers of
samples obtained from large numbers of identical exposure situations, passive
devices often give the most accurate indication of human exposure to a particular
trace component.
Palmes tubes have been used for about ten years in numerous studies
by different research groups. In addition, they also have been subjected to
a number of laboratory studies. As a result, workers today are familiar with
the capabilities and limitations of the Palmes Tube, and how it behaves in a
variety of exposure situations.
The traditional method of analysis of the Palmes Tube is spectrophotometric.
Triethanolamine deposited on stainless steel screens absorb the nitrogen diox-
ide, which is converted to nitrite ion. For analysis, the nitrite is used to
make an azo dye, which is then measured spectrophotometrically. The practical
limit of detection is 7 x 10~10 moles of nitrite ion. Using the dimensions of
the diffusion path and the diffusion coefficient, this corresponds to 300 ppb
hr of collected nitrogen dioxide. Typically, Palmes Tubes are exposed for one
week to collect enough nitrogen dioxide to give reliable results.
In order to gain more information about exposure levels, it would be useful
to measure nitrogen dioxide concentrations using passive samplers for times shorter
than one week. In order to accomplish this, the analytical procedure needs more
166
-------
sensitivity and remain free from interferences. One approach, presented here,
uses a slightly modified Palmes Tube to collect the nitrogen dioxide, and ion
chromatography as the analytical method.
The traditional Palmes Tube is assembled from a cylindrical acrylic tube,
and a 1/2 inch CAPLUG which fits over one end and contains three stainless steel
screens. The tube was modified in order to minimize contamination caused by
handling, and to facilitate the analytical procedure. The modification consists
of a Teflon plug pressed into a stainless steel ring and forced inside the CAPLUG.
A stainless steel screen is pressed into a depression in the Teflon. The inside
edge of the acrylic tube is beveled to prevent the end of the acrylic tube from
touching the screen when the tube is assembled, which would cause unacceptable
chloride contamination during handling of the unassembled tube parts.
The inside of the CAPLUG with the screen is squirted with a stream of deio-
nized water, the excess water is removed by centrifugal force in a mechanical
spinner, and then set on paper towels. Using a micropippete, 20 microliters
of a solution of deionized water and triethanolamine is deposited on each screen,
and then the CAPLUG is placed immediately inside a long glass tube in which N02
free air continuously flows. The screen are coated with 3 x ID'6 mole of tri-
ethanolamine. The batch of treated CAPLUGs are allowed to dry in the tube.
If left inside the drying tube too long, the screens will pick up unacceptable
amounts of N02, since the last traces of N02 are difficult to remove from the
air stream. The CAPLUGs are removed from the drying apparatus, placed on the
beveled end of the acrylic tubes with an empty CAPLUG closing the other end of
the tube. The assembled tube is then placed inside a shipping container, which
is a friction top metal can. Before the container is closed, a polystyrene coffee
cup stuffed with a crumpled paper towel moistened with triethanolamine is placed
inside the container to scavenge any nitrogen dioxide.
The tubes appear externally identical to the traditional Palmes Tubes, and
are handled and exposed in the field in the traditional manner.
For analysis, because of the small amount of nitrite ion present on the
screens, the entire sample must be transfered and trapped in a concentrator
column. The two problems associated with this task is avoiding chloride con-
tamination and preventing air bubbles from entering the chromatograph, which
are resolved by using a transfer apparatus. The CAPLUG with the screen is re-
moved from the acrylic tube and fit over the open end of the transfer apparatus,
167
-------
and clamped tight. The appartus consists of Teflon surrounded by a metal housing.
The stainless steel ring inside the CAPLUG and the metal housing are required
to prevent the Teflon from distorting under pressure. Two Tefzel tubes enter
the bottom of the apparatus, one ending at the bottom of a cavity, the other
extending almost to the top of the cavity but not touching the screen. The cavity
is filled with deionized water from the bottom tubing until it flows out the
extended tubing, without touching the screen. The deionized water flow is then
reversed, and ten milliliters are forced through the apparatus using a Milton
Roy Mini Pump. What air remaining inside the apparatus is compressed to a tiny
bubble and does not leave. The force of the water squirting on the screen rinses
all the nitrite out and transfers it to the concentrator column. After the trans-
fer, the cavity is filled with clean air before unclamping and removing the
CAPLUG. The CAPLUG is removed, and the apparatus is ready to receive the next
sample.
The concentrator column with the trapped sample is switched into the eluent
stream of the ion chromatograph in the usual manner, to complete the analysis.
The ion chromatograph used in this work is a Dionex 2010i, wiith a TAC-1
concentrator column, an anion guard, column, an ASS analytical column, an AFS-1
fiber suppressor, a conductivity detector, and a SP4270 integrator/recorder.
The eluent was .0588 grams NaHCOs and .2333 grams Na2C03 in one liter of deionized
water.
The chromatogram takes about 12 minutes from inject to sulfate. The order
of the first peaks are: waterdip, fluoride, chloride, nitrite. Between the
nitrite and the sulfate peaks, the concentrator column is switched out and loaded
with the next sample, so that with this configuration one sample is run every
12 minutes.
The nitrite appears on the rising shoulder of the waterdip, and at the 0.3
ps level required by the small amount of nitrite in each sample, the signal out
of the detector is negative. The SP4270 integrator cannot accept negative signals,
and the detector does not have a manual offset. To observe the nitrite peaks
in the waterdip, a battery driven voltage divider was added in series to the output
of the detector which provided a positive voltage offset to the signal.
The instrument was checked for linearity by spiking the screen in a CAPLUG
with known amounts of nitrite. The blank screens would typically give no detect-
able peak or a just detectable peak, about three times the baseline noise, which
168
-------
corresponded to 1 x lO'1* mole nitrite.. Peak heights were used throughout these
preliminary runs. The instrument was linear from 10'11 through KT7 mole nitrite,
and nonlinear at 1CT6 mole. A typical trial consisting of 7 runs at 2 x ID'10
mole of nitrite gave peaks heights of 0.081 PS with an average deviation of 0.002
uS.
Replicate sets of ambient exposures were run to test precision and reveal
possible interferences. A typical run consisted of five tubes total, two blanks
and three replicate samples. All five were carried clipped to a shirt pocket
for 18 hours. The blanks gave 4 ppb hr, and the three replicates gave 80 ppb
hr with an average deviation of 3 ppb hr. None of the ambient test runs had
serious chloride contamination problems, including replicate sets left exposed
in the laboratory, representing a worst case situation due to the presence of
hydrochloric acid.
Controlled exposure in test chambers are in progress, in order to deter-
mine the accuracy, and the results will be published when this work has been
completed.
The operating costs associated with this method of analyzing Palmes Tubes
can be approximated as follows: The modified Palmes Tubes, since they contain
about twice as many components, will cost about twice as much as a conventional
Palmes Tube. The other front end cost is the ion chromatograph, tFTe transfer
apparatus, and the source of deionized water. The transfer apparatus is not
available commercially, and includes several valves, Teflon lines, a Milton Roy
MiniPump, and a part machined from Teflon and metal which fits in a clamping
device. The transfer apparatus would be straight forward to automate, if desired.
The time required to prepare the tubes is about the same as for the conven-
tional Palmes Tubes. The analysis time on a manual single channel instrument
using the present eluent and analytical column is 12 minutes per sample. The
limiting factor is the length of time to elude the sulfate peak, which if pre-
sent, would disturb the base line of a subsequent chromatogram if the next sample
were injected immediately after the nitrite peak. In addition to the time per
sample, calibration and QA samples are required.
A major consideration under operating costs is that the instrument must
be dedicated full time to the Palmes Tube analysis. Starting from scratch, it
takes the better part of a week to eliminate contamination problems at the IQ-H
mole level of detection. After an extended shut down of several days, it takes
169
-------
most of a day for the instrument to settle down, while an overnight shut down
requires a couple of hours. As a result, the hidden operating costs of start
up time greatly exceeds actual running time for the analysis of an occasional
small batch of samples.
REFERENCES
1. Palmes, E.D., Gunnison, A.F., DiMattiok J., and Tomezyk, C. (1976) Personal
Samples for Nitrogen Dioxide. Am. Ind. Hyg. Assoc. J. 37, 570-577.
170
-------
EVALUATION OF COED-1 PORTABLE
CARBON MONOXIDE PERSONAL EXPOSURE MONITORS
by
C.F. Turlington, J.K. Bostick, CW. Abel,
C.G. Weant, and J.C Holland
Northrop Services, Inc. - Environmental Sciences
P.O.Box 12313
Research Triangle Park, NC 27709
The instrument evaluations in this report were performed in support of studies of human
exposure to carbon monoxide (CO). A General Electric (GE) CO monitor was selected by EPA.
Microprocessors were purchased from Magus Group and Hewlett-Packard. Rockwell International
assembled the monitors and microprocessors into field instruments identified as CO exposure
dosimeters. Model COED-1. Northrop Services, Inc. - Environmental Sciences provided
independent acceptance and performance testing of the monitors.
Over 700 tests have been performed on 104 GE CO monitors at various times during
modification and service. The COED-1 modifications caused no changes in the performance of the
analyzers. Testing of 21 COED-1 units after field studies revealed a cell half-life of between one and
two years. Temperature sensitivity is the sign of cell degradation. Other tests at room temperature
do not indicate cell degradation. The COED-1's gas flow system limits performance if not
maintained. Out of the 21 returned units tested, 38% required attention. No change in linearity or
data storage was observed after field use. All tests of analyzer linearity or of Magus data systems
have been successful. The data systems have failed only when electrical power has failed. Some
nonlinear performance below 20 ppm was observed in the field. Since only one value below
20 ppm was measured, the acceptance linearity tests did not clearly indicate this problem.
171
-------
INTRODUCTION
The instrument evaluations reported herein were performed in support of studies of human
exposure to carbon monoxide (CO). The primary objectives of these studies were to validate a
methodology for measuring the distribution of CO exposures in a representative sample of an urban
population and to estimate the risk to the entire population. Other goals of the studies included
relating the exposure estimate to fixed site measurements (both indoor and outdoor) and
developing a method for selecting and measuring microenvironments where significant human
exposures occur. Carbon monoxide was chosen because its sources are known, its health effects are
well characterized, and instruments are available. Portable detectors for CO have been developed
by several companies; however, they are generally designed for industrial or mining applications
and lack operational features or CO concentration ranges suitable for ambient field studies.
Therefore, a commercially available system with a satisfactory range was modified to provide the
necessary operational features required for ambient monitoring.
An existing industrial CO monitor, the General Electric (GE) direct reading SPE CO detector,
Model 15ECS3CO3 (COS), was selected by EPA for these studies. This unit was readily available and
provided real time data in the desired CO concentration range, but lacked data logging, integrating,
and timing capabilities. A microprocessor-based control package was purchased from Magus Group,
Inc., to provide the data handling and control, and Rockwell International assembled the
components into a working monitor identified as the CO exposure dosimeter, Model COED-1.
Northrop Services, Inc. - Environmental Sciences' (NSI-ES) role in these studies was
coordination and testing of the monitors as they progressed through the phases of construction and
use. Performance tests of the GE units were necessary to satisfy warranty requirements, while the
completed units from Rockwell had to undergo an additional test and evaluation phase because of
the changes. Final testing was conducted after the field studies to determine any performance
degradation over the study period.
172
-------
The initial test phase consisted of a series of performance checks of the CO3 units over the
temperature range 5-40°C. Defective units were identified and shipped back to the manufacturer.
The modified units, COED-1s, were subjected to a similar series of tests to verify the correct operation
of the detector and the data handling system. When all test requirements were met, the COED-1
units were released for use in the ambient CO field studies. Primary field studies were conducted in
Denver, CO and Washington, D.C., during the winter of 1982-1983. After the units were returned
from the field, a representative sample of the units was tested. This report describes the test
procedures used, reports the results, and provides conclusions concerning improvements for future
use of the COED-1.
CONCLUSIONS and RECOMMENDATIONS
Over 700 tests were performed on 104 monitors at various times during their modification and
service life. The COED-1 modifications caused no changes in the performance of the analyzers. The
results of the COED-1 evaluations are summarized in Figure 1. Each bar consists of two to four
blocks, ranging from the best working units in the bottom block to the worst or untested units in the
top block. Each block contains the number of monitors in that classification and their performance
indicator. The performance indicators are explained by the figure caption. Figure 1A pictures the
performance of the entire population of 75 COED-1 analyzers before field testing. Figure 18,
illustrating the performance of 20 units before they were sent to the field, may be compared with
Figure 1C, showing the performance of those same monitors after they were returned from field
sites. This comparison shows no change in linearity or data storage; all 20 units did as well or better
after field use. The zero settings as-received from the field were fairly close to the correct values,
considering the time and distance of shipping since the monitors were set. Most of the span errors
indicated a systematic deviation discussed in detail below. The two areas in which performance
worsened were the flow rates and temperature ranges. The flow problems were corrected by
maintenance on the pumps and filters.
173
-------
A. All 75COED-1 Units Before Field Use.
B. Selected Units Before Field Use.
1
50 -
#0f
Units
25 -
i i
J 17 !
U
58
P-3
36
P-2
38
P-1
1
U
25
•C.999
49
>.999
2. .
U
Pass
Flow Act/ Linear- Data
Temp ity Storage
20 -
#of
Units
0
.._ — _-,
6
U
15
C
7
P-2
13
P-1
6
<.999
14
>.999
1
19
Pass
Flow Act/ Linear- Data
Temp ity Storage
C. Selected Units After Field Use.
20 -i
#of
Units
5
M
3
A
3
B
10
2
F
4
P-3
8
P-2
6
P-1
20
>.999
20
Pass
•jn <
2ppm
9 >
9ppm
11 <
4ppm
1
>4
Flow Act/ Linear- Data Zero Span
Temp ity Storage Set. Set.
The number in each block is the number of units; the letter indicates the performance of
those units according to the following key:
A- Adjusted pump voltage up and back to start, then O.K.
B- Bubble meter flow passes as-received; mass flow meter restricts.
C- Mass flow meter reading passes test.
F- Failed activation.
M- Maintenance required to pass flow test.
P- Passes activation test:
1. 5° to 40°C temperature range
2. 10° to 30°C temperature range
3. narrower than 10° to 30°C temperature range
U- Untested; but probably O.K. since 5 out of 6 initially untested units produced a B or
C when returned from the field, and no unit has ever failed the data storage test.
Figure 1. Summary Performance Charts.
174
-------
The temperature range degradation is of most concern because its correction requires
replacement of the cell, an activation period, and retesting. After the field studies, 30% of the 20
units had degraded to the point that they no longer met the acceptance criteria. Degradation with
time is more evident when the date of cell installation is considered. Three of the units had their
cells replaced and were retested shortly before field use. These units performed as well or better
upon their return, while 57% of the older units failed. If data were available for the number of
hours of use for each unit, this could be compared with the cell exposure to water and electrical
potential and with the performance to determine whether use increased the degradation. Since the
cells remain moist and have a potential constantly applied to them, their age and ambient storage
conditions probably are the dominant factors in cell degradation, but heavy use may hasten the
process. Temperature sensitivity and irreproducible behavior are the signs of cell degradation.
These indicators should be checked, or cells automatically replaced, every 9-12 months. Other tests
such as linearity, daily zero and span checks, and flow do not indicate cell degradation.
The monitor's gas flow system can also limit performance if not maintained. In addition to the
initial flow and pump voltage checks on all units, the pump motor resistance or impedance should be
measured to eliminate pumps that drain batteries too rapidly. More attention to the monitor's gas
flow system in the field is needed, as 24% of the sample of returned units required service, and 14%
required minor adjustments to achieve proper flow. A good daily and weekly maintenance program
could reduce flow failures, but the greatest improvement in flow system reliability would be, as
suggested by Jungers and Johnson, to invert the flow through the filter-scrubber tube. This design
change would improve reliability and greatly reduce flow system maintenance.
No change in the linearity and data storage performance was observed. The data systems
have failed only when electrical power or connections have failed. Generally, such failures were
obvious and prevented testing; therefore, all completed tests of linearity or data systems
performance have been successful. Only one unit out of 21 sampled from the field had a bad
electrical connection preventing it from being tested. Another unit had batteries that would not
175
-------
charge, but it passed when fresh cells were installed. Although unreliable power and electronic
problems were experienced in the field, they were not observed during acceptance testing. The field
problems may have been due to a combination of weather conditions, physical shock, and vibration.
Nonlinear performance below 20 ppm was observed in the field for some analyzers. The
acceptance linearity tests did not clearly indicate this problem since only one value below 20 ppm
was measured. To identify this nonlinearity, the data logging and linearity tests should be combined
with all measurements stored by the data system and more concentrations measured between 0 and
20 ppm. The nonlinearity problem probably was reduced for some Washington, D.C. field data by
offsetting the zero reading to 1 ppm. Nonlinearity was mitigated in Denver by balancing the
detector output to the data system input.
The performance data for the units tested after the field studies were separated into groups
by age, former field site, and performance test. Means and standard deviations were computed for
each category, and a statistical test was applied to each pair of means to determine if they differed
with better than 95% probability. The only statistically significant deviation was the high span
readings found on the units used in Denver. The high span readings were not explained by
theoretical corrections for flow deviations and the difference in mean barometric pressure between
Denver and NC. Theoretical corrections may be in error, or a systematic error in the calibration
method may exist. Closer attention must be given to this deviation if testing continues at high
altitudes.
BIBLIOGRAPHY
Bay, H.W., K.F. Blurton, J.M. Sedlak, and A.M. Valentine. 1974. Electrochemical technique for the
measurement of carbon monoxide. Anal. Chem. 46:1837-1839.
Hoel, P.G. 1966. -Introduction to mathematical statistics. John Wiley & Sons, NY.
Jungers, R.H. and T. Johnson. Quality assurance planning and implementation for the Denver
carbon monoxide study. Environmental Monitoring Systems Laboratory, Office of Research and
Development, U.S. Environmental Protection Agency, Research Triangle Park, NC.
176
-------
Kosek, J.A. and A.H. Gruber. 1981. Development of improved detection instruments for toxic gas
contaminants in mining atmospheres. [Contract #H0395132 (no report #)] Bureau of Mines,
Washington, D.C
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 In: ACS Symposium Ser
No. 149, pp. 551-573.
Turlington, C.F., J.K. Bostick, C.W. Abel, C.G. Weant, and J.C. Holland. Submitted. Evaluation of
COED-1 and COED-2 portable carbon monoxide personal exposure moitors. TIM-84-02. Research
Triangle Park, NC: Northrop Services, Inc. - Environmental Sciences.
177
-------
MONITORING OF BENZENE IN AMBIENT AIR WITH ORGANIC VAPOR BADGES
Kochy K. Fung and Barbara J. Wright
Environmental Research & Technology, Inc.
Newbury Park, CA 91320
INTRODUCTION
Benzene is a toxic substance commonly found in ambient air, especially in
the urban environment. Benzene is an industrial solvent and is also a component
of gasoline, typically present in concentrations of one to two percent.
Evaporative emissions account for some of the benzene in ambient air. However,
benzene can also be produced in combustion processes including those in
automobiles and can be emitted with the exhaust into the environment. Recently,
the California Air Resources Board (CARB) has identified benzene as th» first
pollutant to be controlled under its toxic air contaminant control program.
Thus, there is a need for an accurate, reliable, low-cost technique for the
routine monitoring of benzene.
The most common method of sampling for benzene is by adsorption on charcoal
tubes (NIOSH, 1977). Air is drawn through the tube at a known flow rate using a
pump. At the completion of sampling, the benzene adsorbed by the charcoal is
eluted with a solvent and determined by gas chromatography. Tedlar (polyvinyl
fluoride) bags and metal canisters have also been used for benzene sampling
(Singh, 1980; Grosjean and Fung, 1984). A grabbed or integrated air sample is
pumped into the bag or canister and brought back to the laboratory where the
hydrocarbons in the sample, including benzene, are determined with gas
chromatography after cryogenic sample preconcentration. Benzene has also been
monitored successfully using an on-site gas chromatograph equipped with a
photoionization detector (Hester, 1979).
While these techniques will yield reliable benzene data, their application
to large-scale field monitoring may be costly. Equipment costs, maintenance and
flow calibrations all make such programs more expensive. Availability of power,
sampler noise, and equipment security are major problems in some monitoring
situations. The advent of the passive monitor (badge) has resolved a lot of
these difficulties. This commercially-available device is inexpensive, light-
weight for easy placement and handling, and requires no associated sampling
equipment. It has been used routinely as a personnel monitor in industrial
hygiene applications in place of the charcoal tube/pump method. Thus, there has
178
-------
been a great deal of interest in determining whether the badge could be applied
to the monitoring of ambient levels of benzene at concentrations several hundred
times lower than the concentrations the device was originally meant to collect.
A study was conducted in which data from the badge samples was intercompared
with data from charcoal tube and Tedlar bag samples. Benzene levels were
measured side by side using these three techniques at the perimeter and center
of two gasoline stations, in the vicininty of a bulk terminal, and adjacent to a
busy freeway in southern California.
EXPERIMENTAL
The organic vapor badges were obtained from duPont. Type GAA contains one
strip of charcoal, and type GBB is similar with an additional backup strip of
charcoal. Type GBB was used primarily because it was not known if breakthrough
would occur, and the backup provided assurance that none had occurred during
sampling. The badges were hung on posts and allowed to sample for 24 hours
instead of shorter durations since the sampling rate of the badges for benzene
is 35.6 ccpm. The charcoal tubes were SKC sorbent tubes with 400 and 200 mg of
charcoal in a front and backup section, respectively. AC-powered sampling pumps
were used to pull air at 1 1pm through the tubes which were placed adjacent to
the badges on the posts. The Tedlar bags were 50-liter capacity. They were on
loan for the study by CARB. The bag samplers were designed to be comparable to
those used by the CARB. Filtered ambient air was pushed through a pressure
regulator, relief valve and rotameter into the bag at approximately 35 ccpm for
24 hours. The air inlet tube to the pump was also located next to the badges
and tubes so that all devices sampled the same air parcel. After sampling, the
badges were covered and sealed inside foil bags provided by the manufacturer; the
charcoal tube ends were closed with cap plugs, and the bag inlet was sealed with
a cap nut for transporting back to the laboratory.
Meteorological data such as wind speed, wind direction and temperature were
recorded on strip charts using two MRI mechanical weather stations, one located
at the upwind perimeter, and the other downwind. The percent relative humidity
was measured with a sling psychrometer whose thermometers had been compared to
an NBS thermometer.
Within 24 hours of collection, the bag samples were analyzed with high
resolution capillary gas chromatography after cryogenic sample preconcentration
179
-------
(3). All the tubes and badges were stored in a freezer upon return to the
laboratory until they were to be analyzed. They were extracted with
benzene-free carbon disulfide (CS2) and analyzed using packed column gas
chromatography with flame ionization detection.
RESULTS AND DISCUSSION
The 24-hour average benzene concentrations measured at the four locations
are summarized in Table 1. The results showed that the benzene levels
determined using the three different techniques were equivalent. The mean
benzene concentrations derived from the entire data set are 5.3 ± 1.3,
5.7 ± 1.4, and 6.1 ± 1.4 ppb, respectively for the badge, tube and bag method.
It is clear that the differences observed amongst the three are statistically
insignificant.
The CS2 was Fisher "purified" reagent grade, yet it contained considerable
quantities of benzene. For example, the CS2 contained 14 ug/ml of benzene
compared with typical benzene samples which ranged from 4 to 22 (jg/ml. This
contaminated CS2 was worthless for direct use in this study. Thus, a method for
purifying the CS2 was developed. The solvent was acid distilled, and only the
benzene-free cuts of the distillate were used in the analyses.
There has been concern regarding the effect of high humidity on sample
collection by the badge. In general, higher humidity tends to reduce the
adsorption capacity of the charcoal. For both the GAA (single strip) and GBB
(double strip) badges, duPont has rated their capacity for benzene to be greater
than 12 hours while sampling at 10 ppm and 80% relative humidity. In the
present application, the benzene levels were over a thousand times lower, and
the relative humidity ranged from 22% to 93%. High humidity was experienced
during the night and early morning hours. Analysis of the backup strips showed
that breakthrough had not occurred with any of the badges used.
Since benzene was not detectable in the field blanks, the analytical
detection limit dictated the lower quantifiable limit (LQL) of the badge. The
LQL has been estimated to be 0.3 ppbv based on a signal-to-noise ratio of two
with the present sampling and analytical conditions.
In conclusion, the badge appears to be useful for monitoring ambient levels
of benzene. Humidity was not a problem in causing breakthrough. Significant
180
-------
TABLE 1
AMBIENT BENZENE LEVELS MEASURED AT FOUR LOCATIONS
IN SOUTHERN CALIFORNIA
Site
I
II
III
IV
Post No.
1
4
5
6
9
2
4
5
6
7
2
3
4
2
6
Badge
5.8 ± 0.3*
7.6
6.0
7.0
6.2 ± 0.1*
5.7
6.5
5.3
4.9
4.5
3.7 ± 0.5*
4.1 ± 0.2*
3.6 ± 0.5*
5.6 ± 0.1*
3.5 ± 0.1*
Tube Bag
6.4 6.6
6.8
5.6 ± 0.6* 7.4
6.0
7.0
6.2 7.7
4.9
9.2
A
5.0 6.4
4.5
4.0 4.2 ± 0.1*
4.5 ± 0.4*
4.1
6.7 6.0
4-1 4.3 ± 0.3*
'Average of duplicate or triplicate samples.
181
-------
improvement in LQL can be expected if capillary gas chromatography is used for
the analysis.
REFERENCES
1. Grosjean, D. and K. Fung, 1984. "Hydrocarbons and Carbonyls in Los Angeles
Air." JAPCA, 34: 537.
2. Hester, N.E. and R.A. Meyer, 1979. " A Sensitive Technique for Measurement
of Benzene and Alkylbenzenes in Air." Env. Sci. Technol., 13: 107.
3. National Isntitute for Occupational Safety and Health P & CAM No. 127,
1977. "Organic Solvents in Air."
4. Singh, H.B., 1980. "Guidance for the Collection and Use of Ambient
Hydrocarbon Species Data in Development of Ozone Control Strategies." U.S.
Environmental Protection Agency Report. EPA-450/4-80-008, Office of Air
Quality Planning and Standards, Research Triangle Park, NC, April.
182
-------
A PASSIVE OOSIMETRY METHOD FOR DETERMINING HYDRAZINE IN AIR
Susan L. Rose and Jeffrey R. Wyatt
Naval Research Laboratory
Chemistry Division
Washington, D.C. 20375-5000
Chester M. Hawkins
Geo-Centers, Inc.
Suitland, MD 20746
INTRODUCTION
Large quantities of monomethylhydrazine (MMH) and hydrazine
are used as high energy propellants in the space shuttle.
Because MMH and hydrazine are suspected carcinogens, NASA must be
concerned about the toxicological problems that could arise-as a
result of fuel storage and handling operations. Consequently
Kennedy Space Center has established special guidelines for
maintaining personnel safety; which include routine air monitor-
ing to ensure that exposure to MMH and hydrazine does not exceed
the American Conference of Governmental Industrial Hygienists
(ACGIH) recommended threshold limit values (TLV) of 200 and 100
parts-per-billion (ppb) for MMH and hydrazine, respectively.
NASA has a need for inexpensive, quantitative personal
dosimeters that could be distributed to a large number of
personnel. The badge should be worn for at least one week and
could be monitored in the event of a hydrazine spill. In
addition, a portion of the dosimeters would be analyzed regularly
to provide average exposure levels.
The development of a dosimeter badge is complicated by the
reactivity of the hydrazines. Hydrazines have a strong tendency
to absorb and decompose on surfaces, particularly metals, making
them difficult to sample and measure. Since they are polar,
reactive compounds, they associate with water vapor, form
183
-------
aerosols, are sensitive to air oxidation and react readily with
carbon dioxide.
Few commercial detectors or acceptable techniques are
available. A passive electrochemical monitor was tested and
found to be unreliable. Liquid-sorbent badges were investigated,
but lacked the precision necessary. Strong acids on solid
supports have been used for field sampling of hydrazine, but MMH
is not stable in this medium. Antioxidants have improved the
stability of MMH solutions, therefore we decided to investigate a
solid support coated with an organic acid with antioxidant
properties, citric acid.
EXPERIMENTAL
The passive dosimeter shown in figure 1 has been developed
and is being evaluated. A plastic sample holder, 1/2" deep and
1 3/4" in diameter, supports the collection substrate. The
collector is a disc of polyester matted drafting paper that is
dipped in 20% citric acid in methanol solution and allowed to
dry. A 1/16" spacer provides an air gap between the substrate
and the top of the badge. Passive diffusion is controlled by a
one inch diameter pattern of holes.
The dosimeter badges are being tested in the glass and
Teflon constructed test apparatus shown in figure 2. Diffusion
tubes provide reproducible hydrazine vapors that are easily
varied. Hydrazine contaminated air flows into the dosimeter test
(figure 3) chamber at 10 liters/min. Teflon baffles are used to
improve laminar flow through the chamber. At least two hours are
allowed for equilibration between changes in experimental
conditions, and the concentration is verified by collecting
impinger samples from the chamber with analysis by colorimetric
and coulometric methods. Three chambers of different diameters
allow face velocity tests without changes in the gas stream
conditions.
184
-------
The badges are being analyzed using two accepted wet
chemical methods: (1) A NIOSH approved method (#5149) using
phosphomolybdic acid, and (2) a coulometric titration with
bromide and amperometric endpoint detection.
RESULTS
The precision.of the dosimeter was tested by exposing the
badges to a dry gas stream contaminated with 242 ppb MMH and
measuring the amount of MMH detected versus the exposure time.
Exposure times between 5 and 65 hours were examined over several
days. A correlation coefficient of 0.996 was obtained for 107
data points. The badge sample rate was calculated from this data
and found to be 31 ml/min. The stability was verified in two
tests. In each test, four badges were exposed to MMH, two of the
badge substrates were analyzed immediately and the other two
substrates were stored in the dark at room temperature in a
sealed container. No degraduation was observed for any of the
storage periods (3-7 days). The linearity was investigated by
exposing the badges to MMH between 70 and 240 ppb in dry air for
16 hours. Deviations from ideal performance were observed below
100 ppb. Because we suspected adsorption as the problem, the
adsorption of the hydrazine on the badge at low concentrations or
low exposure time was investigated by constructing the badge from
three different materials: (1) machined Teflon, (2) machined
polypropylene, and (3) molded Fluoroware. The badges were
exposed to MMH at the TLV in both dry and humid air from 15
minutes to 16 hours. The apparent sample rate was calculated by
comparing the moles of hydrazine collected with the actual
concentration. The apparent sample rate was reduced for all of
the badges compared to the calculated sample rate. The machined
Teflon and polypropylene badges were much worse than the molded
plastic, particularly in humid air. The apparent sample rate is
75?o of the calculated sample rate sample for the molded plastic
badges at approximately one hour. Therefore, the molded plastic
185
-------
was used for the remainder of the tests. The relative humidity
effects were tested using MMH at the TLV, a face velocity of 5
ft/min and relative humidities between 10-75?o RH. Minimal
effects on the performance were observed. The face velocity
affects on the sample rate were also tested for 2, 5 and 20
ft/min. No significant variations in the amount of MMH collected
were observed. Interference by isopropyl alcohol, ammonia, NO-2 >
and freons were examined by mixing TLV levels of each with TLV
levels of MMH. Little variation was observed in the MMH concen-
tration in the badge compared to analyzed impinger samples
from the same gas stream.
186
-------
diffuser
(section)
spacer
citric acid
layer
polyester
film
base
scale
, 1/2"
Figure 1. Dosimeter
187
-------
I 2 S
z >- <
o ?.
CO
3
-4J
rO
S-i
-P
tn
o>
EH
OJ
< g
•5 P.
188
-------
CN
Q)
J3
d
^
U
4J
03
(U
co -
% E
CD CO
(5 -^
-------
GC/FTIR STUDIES OF VOLATILE ORGANIC
COMPOUNDS COLLECTED ON TENAX CARTRIDGES
R. A. Palmer, J. W. Childers and M. J. Smith
Department of Chemistry
Duke University
Durham, North Carolina 27706
J. D. Pleil
Northrop Services, Inc.
Research Triangle Park, North Carolina 27711
W. A. McClenny
U. S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
ABSTRACT
The combination of TENAX sorbent cartridge thermal desorption with capil-
lary column GC/on-the-fly FTIR has been shown effective for the detection and
identification of volatile organics in laboratory-generated mixtures, including
the distinction between isomeric species, at the level of a few hundred nanograms
per compound per cartridge. Traces of water desorbed from the cartridges must
be reduced by the insertion of a dryer unit between the desoption chamber and
the GC column. Computer search of a standard gas phase spectral library is
used to identify components from the background corrected spectra. Methods of
lowering the detection and identification limits to less than 100 ng per compound
per cartridge are proposed.
190
-------
INTRODUCTION
Gas chromatography combined with Fourier transform infrared spectroscopy
is an efffective means of detecting and identifying volatile organics found in
ambient air samples and provides structural information that is complementary
to GC/mass spectroscopic techniques. While GC/MS is more sensitive than on-
the-fly GC/FTIR, it often lacks the ability to distinguish unambiguously
between geometrical isomers. GC/FTIR, on the other hand, can be used to make
a positive distinction between isomeric compounds based on their unique
^^T^ spectra.1.2,3 jn addition, when combined with retention time data,
GC/FTIR can also be used for independent unequivocal identification of trace
organics in many cases. However, since GC/FTIR lacks the sensitivity to
detect trace organic pollutants at ambient air concentrations, some method of
preconcentrating the samples is required. Collection on sorbent cartridges
followed by thermal desorption would appear to be a viable solution to this
problem. In this study the feasibility of the use of TENAX-GC cartridge
thermal desorption combined with subsequent cryotrapping5*6 has been evaluated
as a delivery system for on-the-fly GC/FTIR detection and identification of
selected organic compounds. Special emphasis has been given to the identifi-
cation of geometrical isomers. Instrumental problems addressed include removal
of interferences due to co-eluting water, spectral cleanup to facilitate lib-
rary based identifications, and development of an analytical method to coordi-
nate the various instrument subsystems necessary for precise analyses.
EXPERIMENTAL
CHEMICALS
Samples were prepared by the Northrop Services, Inc. (NSI) Volatile Organ-
ics Standards Laboratory. Neat compounds (5 to 10 yL amounts) were injected
into 2L static dilution bottles filled with zero grade helium. The resulting gas
mixtures were equilibrated for a minimum of one hour at 65°C. Syringe samples
of the mixture were analyzed by capillary GC and referenced to diffusion tube
standards that were certified by time dependent weight loss rates. This
established accurate concentration values (typically 500 ppmV/compound) for the
dilution bottle mixture.
The sample format consisted of Pyrex tubes (10 cm long, 1.5 cm o.d., and
1.2 cm i.d.) packed with 1.35 g of TENAX-GC and capped at both ends with glass
wool. These cartridges were initially baked out at 270°C to remove any residual
contamination and then attached to a zero grade helium purge system. Known
volumes of the gas mixture (typically 50 to 100 \il) were injected into this
purge system, which served to deposit the analytes onto the TENAX-GC. After
being loaded, the sample cartridges were individually sealed in glass vials
with Teflon-lined caps. A typical sample set of five to ten vials was placed
in a container partially filled with activated charcoal and then stored in a
freezer at the NSI facility until needed.
191
-------
INSTRUMENTATION
The analytical instrumentation incorporates three subsystems, each of
which is commercially available. They are: 1) a Nutech model 320 cartridge
desorption and sample injection system (Nutech Corp., Durham, NC); 2) a Varian
3700 gas chromatograph (Varian Associates Inc., Palo Alto, CA); and 3) an IBM
model 9195 Fourier transform infrared spectrometer (IBM Instruments, Inc.,
Danbury, CT). These have been interfaced for split-less on-column injection
and subsequent serial infrared and flame ionization detection (FID), using
the IBM Model F2752 on-the-fly (light pipe) GC/FTIR interface attachment and the
GC's FID. A block diagram of the overall system is given in Figure 1.
At the start of a GC/FTIR run, a loaded Tenax-GC cartridge is placed in
the thermal desorption chamber. The carrier gas flows through the desorption
chamber, and entrains sample desorbed from the cartridge. Since Tenax, like
other solid sorbents, has many of the properties of a chromatographic stationary
phase some compounds will be preferentially retained on the sorbent, and the
rates of thermal desorption will vary from compound to compound.
To assure that all components of the gas mixtures enter the GC at the same
time, compounds are cryogenically trapped at -150°C until the desorption pro-
cedure has been completed (typically about 30 minutes). At the end of this
time, the trap is quickly heated to 200°C, and the sample is flashed onto the
GC capillary column, where it is cryofocused at the head of the column at
-50°C.
As compounds elute from the GC column during programmed heating to 150°C,
they pass into the FTIR light pipe, which is the heart of the GC interface. IR
light from the FTIR source is focused through the 1 mm gold coated bore of the
35 cm long pyrex tube (light pipe) onto a LN2 cooled MCT detector. As compounds
pass through the light pipe, 5-6 interferograms are collected per peak and
saved for later computation. Data are collected using the Gram-Schmidt algo-
rithm. This algorithm involves the computation of a single point interferogram
based on the average of the total absorption detected by each scan relative to
the signal from the carrier gas. These single points constitute a real-time
chromatogram, referred to here as the Gram-Schmidt (G-S) trace. After passing
through the light pipe for IR detection, the eluant reenters the GC, and is
monitored by the standard flame ionization detector. The signal from the FID
follows the FTIR (G-S) peak by ca_. 3 sec.
After the GC run is completed, the interferograms corresponding to each
of the eluant peaks are co-added, and the FTIR spectra of each compound is
generated. The resulting FTIR spectra can then be compared to the Sadtler
IRVAP 8x16 library of 8000 vapor phase spectra of organic compounds, which is
stored on disk. The library search program returns the 10 best spectral matches
and an associated "Hit Quality" number for each in the range 0 to 1000 (1000
indicates a perfect match).
One of the characteristics of the combination of these three techniques,
TENAX thermal desorption, GC and FTIR, is the inevitability of the presence of
background water in the system. Under normal GC operation, the presence of small
amounts of water vapor in the column is unimportant because the FID is relatively
insensitive to it. However, because of the sensitivity of IR to water, it is
192
-------
necessary to reduce the level of water In the GC/FTIR system. For this purpose
a Perma-Pure dryer has been inserted between the desorption chamber and the
trap. The Perma-Pure dryer consists of two concentric tubes, the inner one,
which carries the GC effluent, made of Nafion, and the outer, of Teflon. Polar
compounds such as water readily permeate the Nafion and are purged by a flow of
dry nitrogen in the annular space between the tubes.
RESULTS
In order to evaluate the combined systems, preliminary separation and
spectral determinations were made using relatively concentrated gas mixtures
injected directly into the helium purge line of the Nutech desorption unit.
Two analysis runs were made on replicate samples, first without the Perma-Pure
dryer and then after insertion of the dryer. The FTIR Gram-Schmidt trace shows
dramatically the interference of H£0 with the IR detection. Without the dryer
(Figure 2), the broad double maximum band from the 1^0 extends from 13-18
minutes and virtually masks peak 9. After insertion of the dryer (Figure 3),
the water band extends only from 13 to 15 minutes. The presence of water vapor
at these target compound levels does not appear serious, even in the sample run
without the dryer. However, at the few hundreds of nanograms level, as shown
below, the removal of as much water as possible is imperative. Even at this
higher loading level the importance of the reduction in water can be demon-
strated by looking at some of the less intense G-S speaks. Peak 9 (Figures 2
and 3), for example, is barely visible in the trace run without the dryer since
it is nearly lost under the background water band. However, this peak is
clearly revealed when the dryer is employed.
The transform of the co-added interferograms collected during the interval
of peak 9 is presented in Figures 4 and 5. The spectrum in Figure 4 was obtained
without the Perma-Pure .dryer, and it is evident that much of the spectrum is
obscured by the adsorption of water. When the dryer is employed, all of the
background water disappears, and the spectrum of peak 9 is revealed, as seen in
Figure 5. When the spectrum taken without the dryer was subjected to the vapor
phase library search, it was found that the automatic peak picking routine
ignored some of the peaks which were due to the sample, and chose a peak due to
the water absorption; this resulted in a mis-identification of the spectrum.
However, when the water was eliminated, the routine properly chose peaks only
due to the sample. The spectrum of peak 9 was correctly identified as
tri chloroethylene.
The G-S trace of a mixture of several sets of isomeric compounds loaded
onto Tenax at about 380-580ng levels is shown in Figure 6. The G-S trace shows
the greater importance of the background water band for samples loaded at these
levels. The actual chromatographic peaks are barely visible, but still are
clearly above the baseline. Note that all of these compounds elute after the
broad water "band". An expanded view of the 18-24 minute portion of the G-S
trace is shown in Figure 6b. Eight compounds were loaded on the cartridge, and
seven peaks were distinctly observed (in agreement with the post-light pipe
FID chromatogram). The interferograms corresponding to each of the seven peaks
were co-added and transformed and then were corrected by subtracting a background
generated from interferograms collected adjacent to each peak. For example, the
results of co-added interferograms collected during the passage of peak 6
193
-------
(Figure 6) through the light pipe is shown in Figure 7. This figure shows: a)
the spectrum of the adjacent background, b) the unconnected spectrum and, c)
the background corrected spectrum. The spectral search results using the
Sadtler IRVAP 8X16 library and associated search program are given in Table I
for peak 6. (It should be noted that the automatic scale expansion feature oT
the plotting program gives different absorbance scales for parts (a), (b) and
(c) of Figure 7. The intensity of the background correction was adjusted so as
to give a flat baseline with no negative excursions). This procedure was
repeated for all of the remaining peaks (Figure 6). The subtracted spectra
were then submitted to the library search program for identification. The
results of the library searches are tabulated in Table II for all the peaks.
In only one case did the search fail to select the known compound in the ten
best hits. From independent evidence (supplied by NSI) peak 1 is known to
contain both m-, and p-xylene with essentially no resolution between the two
compounds. Since the bands of both isomers are present in the spectrum, the
search (which presumes the spectrum is of a single compound) could not make a
correct identification. Ortho-xylene, which elutes separately, is identified
from the corrected spectrum from peak 2 as hit 10 by the spectral search routine.
Of the other isomer sets in the mixture, all are clearly identified.
Particularly impressive is the generation of the separate isomer spectra of o-,
m-, and p-chlorotoluenes, despite the overlap of their peaks in both the FID
and FTIR chromatograms (peaks 3-5, Figure 6). The spectra were obtained by co-
adding interferograms from each of the components separately (#'s 3, 4, and
5). The ortho- isomer is hit 7, the meta-, hit 1, and the para-, hit 2 when
the background corrected spectra of peaks 3, 4 and 5 are subjected to the
standard search routine. Meta- and ortho-dichlorobenzene are both identified
as hit 1 when the library is searched for the background corrected spectra
prior to running the spectral search routine. The transform of the coadded
interferograms of peak 1 of the isomer mixture was known to be either m-xylene,
p-xylene, or a mixture of both (see Figure 8). Because the library failed to
identify it as either of the two independent isomers, and because it is well-
known that these isomers coelute, peak 1 was assumed to be the combination of
these two compounds. In order to test this hypothesis a user-generated library
spectrum of m-xylene was subtracted from the presumed combination spectrum.
The resulting spectrum shown in Figure 9 as then submitted to the library
search program. The library search successfully identified the resulting
spectrum as that of p-xylene, confirming the hypothesis.
CONCLUSIONS
The results illustrate that thermal desorption GC/FTIR, using commercially
available equipment and TENAX-6C cartridges,is capable of distinguishing geo-
metric isomers of compounds with moderately intense IR bands at loadings of
300-500 ng per compound per cartridge. Unequivocal, general identification in
conjunction with GC retention time data would be aided by the availability of a
user generated library of target compounds. Traces of water desorbed from the
cartridges create a problem with the method but can be reduced to manageable
levels by insertion of a drying element. With improvements in the method of
detecting peaks in the Gram-Schmidt trace it is likely that the detection/
identification limit for the method could be ca. 100 ng per compound per
cartridge.
194
-------
REFERENCES
(1) Shafer, K.H.; Lucas, S.V.; Jakobsen, R.J. J. Chromatog. Sci. 1979, 7, 464-
470.
(2) Gurka, D.F,; Betowski, L.D. Anal. Chem. 1982, 54, 1819-1824.
(3) Shafer, K.H.; Hayes, T.L.; Brasch, J.W.; Jakobsen, R.J. Anal. Chem. 1984,
56, 237-240.
(4) Gurka, D.F.; Umana, M.; Pellizzarl, E.D.; Moseley, A.; deHaseth, J.A. Appl.
Spectrosc. 1985, 39, 297-303.
(5) Pellizzari, E.D.; Carpenter,, B.H.; Bunch, J.E.; Sawicki, E. Environ. Sci.
Technol. 1975, 9, 556-560.
(6) Pellizzari, E.; Demian, B.; and Krost, K. Anal. Chem. 1984, 56, 793-798.
(7) Garlock, S.E.; Adams, G. E.; Smith, S. L Am. Lab. 1982, 14 (Dec), 48-55.
195
-------
TABLE I
SADTLER IR-VAP 8X16 LIBRARY SEARCH REPORT
FOR PEAK 6 IN GC-FTIR OF ISOMER MIXTURE
(M-DICHLOROBENZENE)
HIT No. HIT QUALITY SPECTRUM No. COMPOUND NAME
1
2
3
4
5
6
7
8
9
10
294
286
266
254
210
206
142
102
86
78
1597
7524
535
5856
6989
5600
3406
4579
5626
3023
BENZENE, M-DICHLORO-
BENZENESULFONiCACiD,I-METHYL-
1,2,3,3-TETRAFLUOROPROPYLESTEI
PYRIDINE/2-CHLORO-
ETHER/Bis/2-/2,4,5-TRiCHLORO-
PHENOXY/ETHYL/
ETHANOL/2-/2-/2-CHLOROETHOXY/-
ETHOXY/-
ACETALDEHYDE/2-PHENOXY-/
DIETHYL-ACETAL
ACETONITRILE,METHOXY-
METHANE,BROMOTRIFLUORO-
THIOCYANICACID/4-HYDROXY-2/3-
XYLYLESTER
S-TRIAZINE,2,4/6-TRIS/
ALLYLOXY/-
Hit No. indicates the rank of best 10 matches from the 8000 library spectra.
Hit Quality indicates "goodness of fit" of subject spectrum to the indicated
library spectrum, range 0 to 1000.
196
-------
TABLE II
RESULTS OF SADTLER LIBRARY SEARCH
FOR GC-FTIR SPECTRA FROM ISOMER MIXTURE
>EAK No.
1
2
3
5
6
7
COMPOUND
M,P-XYLENE
0-XYLENE
0-CHLOROTOLUENE
M-CHLOROTOLUENE
P-CHLOROTOLUENE
M-DICHLOROBENZENE
0-DICHLOROBENZENE
HIT NUMBER
HIT QUALITY
10
7
1
2
1
1
262A34
380/500
658/676
294/294
310/310
FROM INDEPENDENT IDENTIFICATION
Hit number indicates the rank of the match of the subject spectrum to the correct
library spectrum. . - '
Hit Quality is expressed as a ratio where the first number indicates "goodness
of fit" of the subject spectrum to the correct library spectrum, and the second
number indicates "goodness of fit" of the subject spectrum to the #1 ranked (but
possibly incorrect) match.
197
-------
VENT
CARRIER
NUTECH
CONTROLLER
NUTECH
320
- DESORPTION CHAMBER
-DRYER
- 6-PORT GC VALVE
-RELEASE VALVE
-INTERFEROMETER
LIGHT PIPE
DETECTOR
ASPECT
> 2000 <
COMPUTER
CONSOLE
> AND
MONITOR
VARIAN 3700
GC
>RECORDER
IBM 9195
FTIR
DISK
DRIVE
Figure 1. Block diagram of thermal de«orption-GC/rriR syatt
198
-------
A.9.U ~
IU
O
Z
m
cc
O
CO
S 12.5-
^
O
^"
1
0.0-
1 _f_
3
_
fCJL.» ...... I, *
1- _
4
6
E
J
7
8
'
11
9
10
1
12
14
10 20
ELAPSED TIME, min
30
Figure 2
GC/FTIR Gram- Schmidt trace of initial test eample mixture con-
taining 14 halo-organic compounds obtained without use of
Perma-Pure dryer.
f.o.\j
W
O
z
ffi
cc
0
<" 10 e
CQ 12.5 •
<
f
0
H
}
0.0
3
2
il ..
4
7
e
r^
4^
1
8
isllL
1
12
14
LJL
«UAAteU/IUMAM*JrW4
0 10 20 30
ELAPSED TIME, min
Figure 3. CG/FTIR Gram-Schmidt trace of initial test sample mixture con-
taining 14 halo-organic compounds obtained with use of perma-
Pure dryer.
199
-------
UJ
u
ABSORBA^
i
*0
1 1— ;
SPECTRUM OF ^
TRICHLOROETHYLENE ^
WITHOUT DRIER
CD
LDCD
inin
~l~\
ll1
, Mi
W^^«^^ nA*~
j
bo sobo 2000 icbo
Figure 4,
ENERGY (en"1)
Absorption «pectrum generated from transformation of interferograms
collected during passage through the light pipe of peak 9 in Figure
2 (without dryer).
UJ
0
z
ABSORB A
1 1 .-,-
SPECTRUM OF ^
TRICHLOROETHYLENE ^
WITH DRIER
00 0
ffi °
t-i f"0
^— <
i i
It /v ^
1>^C^U/,JuAJlM^A^^A^Av'<^^^ NV^UA^VJ
Tin ir rJyT 1 1 • — i —
r
I
000
2000
ENERGY (CM~ )
Figure 5,
Absorption »pectrum generated from transformation of interferograms
collected during passage through the light pipe of peak 9 in Figure
3 (vith dryer). Identified as trichloroethylene by library search.
200
-------
OS
h-
Z
UJ
u
z
o
CD ID CD CD O CD
ON Z Z Z Z Z Z
CO
m ea r^- m »— i^- un
oo r«- r- r- vo r—
% ff\ ^- -3- Jf LTv IA
oo
cr\
60
9
•O
V
u
O
a.
o
o
UJ
z
UJ
=>-
_J
UJ O
Z H-
UJ LU O
_l Z Di
>- LU O
X -J _J
I >- I
a. x o
" i i
ZOO
LU LU
Z Z
UJ UJ LU UJ
Z Z ISI M
LU LU Z Z
Z> ID UJ LU
_i _i m co
o o o o
»- t- o; a;
o o o o
a: Di _i _i
o o i i
_i _i u u
i i — —
o u a o
i— CM
UJ
a.
M
201
-------
4000
3000
2000
1000
ENERGY (CM~ )
Figure 7. Peak 6, Figure 6 (m-dichlorobenzene)
a) Absorption spectrum generated from adjacent baseline (background
correction).
b) Uncorrected spectrum from peak interval.
c) Corrected spectrum with peaks used for search identified.
202
-------
COMBINATION
SPECTRUM
M + P-XYLENE
•J.ODO aobo "aoBoToBo
ENERGY (CM )
Figure 8. Background corrected absorption spectrum generated from peak 1,
Figure 6. Combination spectrum at m, p-xylene.
-f-
SUBTRACTED
SPECTRUM
P-XYLENE
I
lobo
ENERGY (CM~
Figure 9. Spectrum of p-xylene generated from the subtraction of a library
spectrum of m-xylene from the combination spectrum in Figure 8.
203
-------
NEW REDOX REACTION PROCESSES FOR
ATMOSPHERIC POLLUTANTS
ANALYSIS OF
R. E. Slevers and S. A. Nyarady
Department of Chemistry and CIRES,
University of Colorado, Boulder CO 80309
R. S. Hutte
Sievers Research, Inc., Boulder, CO 80301
Catalyzed redox reactions of NO2 and HN03 with organic and
inorganic reducing agents can form the basis for new selective
measurements of a large number of atmospheric constituents. The
unifying feature of these reactions is that NO is formed, which
can be selectively and sensitively detected downstream by the
well-established ozone chemiluminescence technique (Fontijn, A.,
et al, 1970).
A new selective detector for gas chromatography has recently
been developed (Nyarady, S. A., et al., 1985). The redox
chemiluminescence detector (RCD) sense compounds that can serve
as reducing agents for nitrogen dioxide, in gold-catalyzed redox
reactions. These compounds include many species that are of
importance in atmospheric analyses. In the RCD, the effluent of
the gas chromatograph is mixed with a gas stream containing
NO2 prior to contacting a catalyst bed. In the catalyst bed,
oxidation/reduction reactions occur which result in the
conversion -of NO2 to NO. The NO is then detected,
downstream, by the sensitive ozone chemiluminescence technique,
in which NO is reoxidized to form NO2 in an excited
electronic state, which relaxes by emission of a photons with
energies in the visible and infra-red regions of the spectrum.
Formaldehyde + excess N02
NO + O
AU
->
400° C
-> NO.
NO + oxidation products
O2 +hv
204
-------
Compounds that can act as reducing agents for N05 in
these catalyzed redox reactions include formaldehyde and other
aldehydes, phenols, alcohols, ketones, esters, ethers,
acids, aliphatic and aromatic amines, olefins,
hydrocarbons, sulfur- and phosphorus-containing
and inorganic species including hydrogen, hydrogen
ammonia, hydrogen sulfide, sulfur dioxide, carbon
and phosgene. The sensitivity of the RCD for
carboxylic
aromatic
compounds,
peroxide,
monoxide,
compounds
that can be catalytically oxidized by NO, is
comparable to the flame ionization detector. Species which do
not act as reducing agents in the gold-catalyzed redox reactions
with NO2, and are therefore not detected by the RCD,
include the major constituents of air (N2,
water, and alkanes.
CO
2'
Ar),
The measurement of organic compounds in ambient air
presents many problems to the analytical chemist. Generally,
the concentrations of these compounds are very low (ppb-ppt) and
therefore require pre-concentration techniques to isolate
sufficient quantities of the compounds for the analysis. The
samples collected by these techniques, however, are extremely
complex. Often the compounds of interest, which are present at
very low levels, cannot be directly detected due to the presence
of other species, such as aliphatic hydrocarbons, which are
present at much higher concentrations. To overcome the
complexity of the samples, two techniques are usually employed:
fractionation of the sample by chemical and physical
techniques,prior to the analysis by a chromatographic method,
and selective detectors for gas and liquid chromatography.
Sample fractionation techniques can often be utilized to
isolate the compounds of interest from a complex sample matrix.
There are, however, many problems associated with these
fractionation techniques, including loss of the analytes,
205
-------
contamination of the sample, and incomplete resolution of the
compounds of interest from the sample matrix. These techniques
are also time-consuming, labor-intensive, and sometimes
irreproducible. Another approach is the use of selective
chromatographic detectors. By choosing a detector which
responds to the compounds of interest, but not to the major
components of the sample matrix, analysis of low level species
in complex matricies often becomes possible. Selective
detectors for gas chromatography commonly used include the
electron capture detector for halogenated species, the flame
photometric detector for sulfur- and phosphorus-containing
compounds, the thermionic flame detector for nitrogen compounds
and the photoionization detector for aromatic hydrocarbons.
Until now, there has been no detector that exhibits selectivity
for oxygen-containing compounds.
It is for these reasons that the development of a new
selective detector will enhance our capabilities to handle more
effectively the complex mixtures often encountered in
environmental analysis. The selectivity of the RCD can be
varied by changing the operating parameters such as the
temperature of the catalyst chamber, the active metal used as
the catalyst, and the reagent gas (N02 versus HNO3). For
example, at low catalyst temperature (<200°C), only those
compounds which are most easily oxidized produce a RCD response,
while at higher temperature, a greater number of compounds can
serve as reducing agents for NO2, and, therefore, are detected.
Because the RCD does not respond to the major constituents
of air or to saturated hydrocarbons, it is ideally suited for
the analysis of low levels of oxygen-, nitrogen-, sulfur-, and
phosphorus-containing compounds in ambient air and associated
with airborne particulate matter. The RCD can also detect some
gases which cannot be sensitively detected by other GC
206
-------
detectors. Potential applications of the RCD in atmospheric
analysis include the measurement of aldehydes and other
oxygenates(in the presence of much higher levels of
hydrocarbons), measurement of H2O2 in air and precipitation,
the analysis of ammonia and amines in air, and the selective
measurements of olefins and aromatic compounds in urban air.
Specific hazardous air pollutants to which the RCD has been
shown to be sensitive include acetaldehyde, formaldehyde,
benzene, toluene, phenol, o,m,p-cresol, acrolein, 1,3-butadiene,
ethylene oxide, vinyl chloride and phosgene.
When pre-concentration techniques such as cryogenic
sampling or collection on porous polymers are used for the
analysis of volatile organic compounds in urban air, the major
components are aliphatic and aromatic hydrocarbons, primarily
from the incomplete combustion of fossil fuels. Because these
compounds are present at fairly high levels, and so numerous,
they overlap with and prevent detection of minor species such as
aldehydes and other oxygenates. The RCD permits the detection of
these oxygenates without interferences from the ubiquitous
aliphatic hydrocarbons.
In the analysis of organic compounds associated with
airborne particulate matter, the RCD should permit detection of
the polycyclic aromatic hydrocarbons, and the nitro-, amino-,
and oxygenated- derivatives of the PAH's without the use of pre-
fractionation techniques, even if higher concentrations of
aliphatic hydrocarbons are also present.
The principle of catalyzed redox reaction can also be used
to detect odd nitrogen species in ambient air (Bellinger, M. J.,
et al.,). Hubler, et al,,, have described the continuous
measurement of the sum of total reactive odd nitrogen compounds
in ambient atmosphere.
207
-------
Nitric acid, nitrogen dioxide, PAN and other unidentified
nitrogen-containing species in rural tropospheric air were
converted continuously to nitric oxide by addition of excess
carbon monoxide on a catalytic gold surface. By judicious choice
of the reducing agent, selective detectors for such species as
HNO3, PAN, and organic nitrates may be possible.
Alternatively, selective pre-concentration sorption techniques
used in concert with catalysis of redox reactions may allow
measurement of species such as nitric acid at lower levels than
presently possible. For example, we have demonstrated that
nitric acid becomes adsorbed on gold surfaces, where it becomes
activated and rapidly reacts with alcohols to form NO. This may
allow pre-concentration of one or more odd nitrogen species from
ambient air on gold sorbent surfaces, followed by intermittent
injection of an excess of a reducing agent such as ethanol,
coupled with rapid heating to generate a pulse of nitric oxide
which can be measured by chemiluminescence as a peak downstream
according to the following reaction sequence:
Au:HN03 + excess C2H5OH -> NO + oxidized products + Au'
NO
O3 -> N02
hv
REFERENCES
1. Bollinger, M. J., Sievers, R. E., Fahey, D. W., Fehsenfeld,
F. C., Anal. Chem. 55,1980-1986, 1983.
2. Fontijn, A., Sabadell, A. J., Ronco, R. J., Anal. Chem. 42,
575, 1970.
3. Hubler, G., Fahey, D. W., Williams, E. J., Eubank, C. S.,
Murphy, P. c., Norton, R. B., Parish, D. D., Fehsenfeld F. C.,
Albritton, D. L., American Geophysical Union Meeting, San
Francisco, CA, Dec. 1984.
4. Nyarady, S. A. Sievers, R. E., J. Amer. Chem. Soc.,107, in
press 1985; Nyarady, S. A., Barkley, R. M., and Sievers, R. E.
Anal. Chem. 57, in press, 1985.
208
-------
APPLICATION OF MULTIDIMENSIONAL GAS CHROMATOGRAPHY TO THE ANALYSIS OF
PARTICULATE SAMPLES
Stanley L. Kopczynski
U.S-. Environmental Protection Agency
Research Triangle Park, NC 27711
INTRODUCTION
A Sichromat-2 Gas Chromatography System was employed to evaluate the
applicability of multidimensional gas Chromatography (MDGC) to the analysis
for polycyclic aromatic hydrocarbons (PAHs) extracted from particulate sam-
ples. The possibility of shortening the analysis time by using unfraction-
ated sample extracts with MDGC was also investigated.
MDGC systems typically consist of two Chromatography columns connected
in tandem and composed of stationary phases of different polarities and/or
operated at different temperatures or with different temperature programs.
Sample preparation for single column gas Chromatography (GC) generally
requires lengthy and laborious extraction and fractionation procedures to
avoid uncertainties in analyte identification and quantisation. The heart-
cutting technique used in two-stage Chromatography produces superior separ-
ations compared to a single column of comparable length. It is a means by
which analytes in complex mixtures may be isolated without prior sample
fractionation. MDGC has been used successfully with a variety of complex
mixtures including food, crude oil fractions, and water(1>2>3).
EXPERIMENTAL
The Siemens Sichromat-2 chromatograph used in this study contained two
fused silica open tubular (fsot) columns coupled in series through a T-piece
which is designed to switch eluates pneumatically from the first stage
209
-------
column (polar) to the second stage column (nonpolar) without valves and
without a sample trap (Figure 1). Connection between the two columns is
made by means of a loosely fitting capillary contained in the T-piece. The
auxiliary gas pressure at each end of the capillary is adjusted by means of
needle valves NV2 and NV3. Column 1 effluent, which is normally directed
to the column 1 detector, can be switched to column 2 by activating the
(bypass) solenoid valve, MV2. The two columns are contained in separate
ovens operated with separate temperature controls, allowing analytes cut
from the first column to be focused at a reduced temperature on the second
column. Non-vaporizing on-column injections were made by means of a Varian
on-column capillary injector.
The column of choice for the first stage was DB-210, 15 m x 0.32 mm
(J & W Scientific, Inc.). Tests with more polar columns showed excessive
column bleed and excessive tailing of the solvent and the higher molecular
weight PAHs. A DB-5 column, 20 m x 0.32 mm, (J & W Scientific, Inc.) was
employed in the second stage.
An 8.3 eV HNU photoionization-detector was employed together with a
flame ionization detector for the back (second stage) column to provide
increased confidence in the identification of PAHs.
The evaluation was conducted with selected PAHs occurring in atmos-
pheric aerosols. Test mixtures consisted of the National Bureau of Stan-
dards (NBS) Standard Reference Material (SRM) 1647 and standards prepared
on-site by Northrop Services, Inc. The toluene solvent employed was chroma-
tography grade from Burdick and Jackson. The NBS urban dust sample SRM
1649 served as a test atmospheric particulate sample. In the absence of
electronic integration, analyses were based on peak height measurements.
The extraction procedure for SRM 1649 consisted of two successive
210
-------
extractions of a 1 g sample with 5 ml toluene in an ultrasonic bath for 45
min at 60°C. Each extract was centrifuged and decanted yielding a black
mixture. They were filtered twice with a Waters Associates' sample clari-
fication kit, yielding a deep yellowish-brown solution. The filtered
extract was concentrated in a test tube to 0.5 ml by directing a gentle flow
of nitrogen gas onto the surface of the solution.
RESULTS AND DISCUSSION
A comparison of sp'litless and on-column injections showed that large
on-column injections (5yL) could be made without sacrificing chromatographic
resolution. However, this did produce a higher solvent background, which
resulted in an increased limit of detection (LOD) with the front column.
Back column LOD for PAHs was generally 0.2-0.3 ng while the front column
LOD was 5-70 times higher.
Results with SRM 1649 indicate that with an atmospheric particulate
concentration of 100 ug/m3 a high volume sampler will collect more than
enough particulate matter for quantitative FID measurements on the back
column. However, the amount collected will be generally inadequate for PID
measurements or FID measurements of front column effluents. The high limit
of detection observed with the PID may be inherent in the lamp or may be
due at least in part to aging of the lamp during the course of the study.
The concentrated toluene extract of SRM 1649 produced a strong unre-
solved envelope of response on the front column and completely obscured the
presence of trace quantities of PAHs, as shown in chromatogram A of Figure
2. Heart-cuts for particular PAHs were taken at time intervals determined
from calibration runs. The PAHs were identified by elution times in the FID
chromatogram of the back column. Confirmatory identification by the PID
211
-------
was not possible due to the low and unstable response of the 8.3 eV lamp.
As illustrated in the case of pyrene (chromatogram B, Figure 2) heart-cuts
from the raw extract contain many components in addition to the PAH.
Consequently very accurate and reproducible elution time measurements as
well as high resolution in the back column are required for identification
and measurement of the analyte of interest. Mean elution times, mean peak
heights, and standard deviations obtained for 2 yL on-column injections of
several PAHs are presented in Table 1. Elution times are very reproducible,
especially on the second-stage column.
Low recoveries were obtained for the 5 PAHs found in SRM 1649 (Table
2). Benzo(b)fluoranthene and benzo(k)fluoranthene were not found although
the NBS reports them to be present at concentrations comparable to pyrene
and B(a)P. No additional fluoranthene or pyrene could be found in a re-
extraction of SRM 1649, suggesting the possibility of volatilization losses
of these two compounds while concentrating the extract.
Grosjean(4) has reported incomplete extraction of ambient particulate
matter by ultrasonication with toluene. This possibility is supported by
the 85% recovery of B(e)P achieved in very recent extractions in our labora-
tory using a Sonifier Cell Disrupter (Heat Systems-Ultrasonics, Inc.) and
methylene chloride in place of toluene. However column degradation cannot
be ruled out as a contributor to losses of the less volatile PAHs. Column
degradation was very evident in subsequent analyses of standard mixtures
and was found to be irreversible.
CONCLUSIONS
Multidimensional GC has been found to be capable of analyzing complex
atmospheric mixtures for PAHs. Limited success was achieved in developing
212
-------
a fast and simple analysis for PAHs in atmospheric participate samples.
Some sample clean-up prior to analysis is required to protect the GC system
from "bad actors" present in the unfractionated extract. Extraction of
atmospheric particulate matter with toluene in an ultrasonic bath yields
insufficient recovery of PAHs.
The multidimensional gas chromatograph system used in this study
performed reliably and reproducibly with both splitless and on-column sample
injections. Although larger sample sizes can be employed with on-column
injections, the full potential of this technique was not realized in this
study due to the limited capacity of the front capillary column.
Multiple detectors are highly desirable for analysis of complex
mixtures containing analytes at trace concentrations. However, the 8.3 eV
PID is not recommended as an auxiliary detector for PAHs because of its
short life and low sensitivity. Better results can be expected with a lamp
of greater intensity such as the 9.5 eV PID or a mass selective detector.
REFERENCES
1. H. J. Stan and D. Mrowetz, "Residue Analysis of Organophosphorus
Pesticides in Food with Two-Dimensional Gas Chromatography Using
Capillary Columns and Flame Photometric Detection", J. of High Resolu-
tion Chromatography and Chromatography Communications, 6:255-263 (1983).
2. Anders Jonsson and Sven Berg, "Determination of Low-Molecular Weight
Oxygenated Hydrocarbons in Ambient Air by Cryogradient Sampling and Two
Dimensional Gas Chromatography", J. of Chromatography, 279:307-322 (1983).
213
-------
3. Gerhard Schomburg, "Multidimensional Gas Chromatography as Sampling
Techniques", in "Sample Introductin in Capillary Gas Chromatography";
Peter Sandra, Ed.; Heutig Publishing, Inc.: Stamford, CT., 1985; PT. 1,
pp 235-260.
4. Daniel Grosjean, "Polycyclic Aromatic Hydrocarbons in Los Angeles Air
from Samples Collected on Teflon, Glass and Quartz Filters", Atmos.
Envir., 17: 2565-2573 (1983).
214
-------
Carrier Gas
Figure 1.
COLUMN SWITCHING DEVICE
P/\> PM: Pressure Gauges
DR1, DR2: Pressure Regulators
AP: Differential Pressure Monitor
NV1, NV2, NV3: Needle values
MV1, MV2: Solenoid Valves
Drl, Dr2: Capillary Restrictors
DM: Column 1 Detector
D^: Column 2 Detector
215
-------
30
FIGURE 2.
time, min
2.0|jL raw toluene extract of SRM 1649:
Front column,and B, Back column.
Pyrene heart-cut on A,
216
-------
TABLE 1
Analytical Precision of FID Measurements with On-Column Injections of Standard
Mixtures of PAH's
First-Stage Column
Second-Stage Column
Pyrene
Chrysene
Benzo(e)pyrene
Benzo(ghi )perylene
Mean Elution
Time (min)
13.74(0.02)
16.04(0.02)
21.43(0.02)
33.39(0.01)
Mean Peak
Height (mm)
64.8(2.8)
47.5(2.1)
36.1(0.6)
Mean Elution
Time (mm)
19.58(0.01)
24.32(0.01)
36.68(0.00)
47.99(0.01)
Mean Peak
Height (mm)
65.1(5.4)
108.3(4.4)
121.6(2.6)
103.2(4.9)
TABLE 2
Analysis of SRM 1649 by Chromatographic Heart-cuts of Toluene Extracts
PAH Concentration in SRM in 1649 PAH Spike in SRM 1649 Extract
(pg/g) (ng)
Fluoranthene
Pyrene
Chyrsene
Benz(a)anthracene
Benzo(e)pyrene*
Found NBS Value % Recovery
3.1 7.1 44
2.2 6.6 33
2.1 3.6 58
1.3 2.6 50
0.64 3.3 19
Found Added % Recovery
10.6 10
106
1.7 10
17
*85% Recovery was achieved in most recent work
using a different extraction procedure.
217
-------
A NEW MICROPROCESSOR-BASED INSTRUMENT
FOR THERMAL DESORPTION OF ADSORBENT TRAPS
Robert G. Westendorf
Tekmar Company
P. O. Box 371856
Cincinnati, OH 45222-1856
218
-------
The analysis of organic compounds in air is performed for a wide variety
of reasons, including monitoring of ambient air, emissions studies, and
industrial hygiene analysis. The majority of these samples are analyzed
by collection on a solid sorbent material, which allows the collection
of organics from large volumes of air. The organic, compounds are e.luted
from the sorbent and injected into a gas chromatograph for separation
and detection. Elution of the sample from the sorbent is generally
accomplished by one of two methods; solvent extraction or thermal •
desorbtion. Thermal dosorbtion generally offers better sensitivity,
since the sample is not diluted, better precision, less handling of
traps, less cost due to both reusability of traps and elimination of
costly high purity solvents, and eliminates the nerd for hazardous
desorbtion solvents. However, thermal desorbtion has not been as
popular as solvent extraction for several reasons. The biggest of these
has been the lack of availability of a good instrument to perform the
thermal desorbtion stop. This paper details the design and performance
of an innovative now instrument, the TEKMAR Model 5000 Automatic
Desorber (Figure 1), that offers high performance, ease of operation, a
high degree of flexibility, and operates using the latest in
microprocessor technology.
SYSVEM DESCRIPTION
The heart of the now dosorber is a tube heater built on a totally now
concept. Key features of this new design are ease of sample trap
loading and unloading, all flow is positively directed through the trap,
traps are loaded when the heater is cool, and it uses a virtually
fool-proof, leak-tight sealing method. A representation of the furnace
is shown in Figure 2. The trap tube is packed so that approximately l-'g"
on one end remains empty. An o-ring is placed on the tube, and inserted
into the furnace. The cap, which has a second o-ring, is screwed down
fingertight, simultaneously compressing both o-rings. The o-ring on the
trap tube seals the exterior of the tube, ensuring that all gas used for
desorb must flow through the inside of the trap. The second o-ring
seals the furnace closure to the outside. Note that both o-rings are
located in an area where they are not subjected to heat. In addition,
the sample does not come into contact with the o-rings. An injection
port into the heater chamber is provided for introduction of standards
for calibration or troubleshooting, as well as allowing the Desorber to
be used for studies of sorbent capacities. Heaters are rapidly
interchangeable (less than 5 minutes) to permit the use of trap tube
diameters of 1/4" to 5/8". Each heater can accommodate tubes of two to
seven inches.
Loading of the trap into the instrument takes place with the. furnace
cool. This eliminates the possibility of losing sample compounds before
the furnace chamber is sealed. An exclusive Prepurge mode enables gas
to be passed through the trap before heating. This mode can serve dual
functions: it.allows any water present to be purged out of the tube
before the internal trap is cooled. This can be particularly helpful
for samples taken undor high humidity or rainy conditions. The second
benefit of prepurgo is that it enables oxygen to be completely displaced
from the tube, before heating. This can dramatically increase the
lifetime of the sorbonr used, particularly for materials like Tonnx, as
well as reduce background contamination due to the sorbent.
219
-------
After prepurge an internal trap is cooled to a preset temperature. This
trap can be an open cryogenic trap of narrow or wide bore, or it can be
a packed trap with or without cyrogenic cooling. Using liquid nitrogen
temperatures down to -150 C can be obtained. The tube heater is then
rapidly heated and simultaneously swept with helium to desorb the
sorbent. The desorb temperature can be as high as 420 C, allowing the
Desorber to be used for carbon-based sorbents such as Spherocarb or
Carbotrap. The desorbed materials are then rctrapped in the internal
trap. After the desorbtion step is complete, an eight port switching
valve is rotated to cause the GC carrier gas to b;iokflush the internal
trap, which is rapidly heated. The carrier gas is then used to sweop
the sample to the GC. For use with capillary columns a second
generation cryofocusing trap is used to focus the sample into a very
tight band for injection. This cyrof ocusirig trap can also be operated
by itself for the focusing of direct CC injections. The results
obtained for a focussed GC injection of an aromatics standard are shown
in Figure 3. After the sample is injected, the sorbent trap can be
baked to regenerate it. A representation of the gas flow scheme is
illustrated in Figure 4.
The control of the instrument is microprocessor-based, and has been
engineered to be as user-friendly as possible. The instrument features
storage for three methods, one of which enables the operation of the
cryofocus interface by itself for focusing direct GC injections of gas
standards as specified by EPA methodology (1). All parameters are
entered via a tactile response, touchtone membrane keypad. The entire
control panel is hinged to allow adjustment to the most, convenient angle
for parameter entry and to read the two-line LCD display. Building a
method can be performed quickly and easily, since all entries are
prompted. A powerful GC interface allows the Desorber to fully interact
with the chromatograph , permitting the analysis to bo fully automated.
Additionally, an RS232C data communications port permits the entry or
reading of parameters to be performed remotely. A typical use would be
allowing the printing of desorbtion parameters on the GC report.
SYSTEM PERFORMANCE
Historically, most thermal desorbtions have been run using an extremely
simple chamber. The method of construction of these furnaces allows the
flow to pass around the outside of the tube, and not actually pass
through it. Desorbtion thus relies on diffusion of the sample compound
off the sorbent and out of the tube. This is known as passive
desorbtion. The new furnace used in the Model 5000, since it positively
directs the gas flow through the trap, provides an active desorbtion.
The difference this makes can be seen by comparing the figures listed in
Table 1. The conditions under which these figures were obtained, listed
in Table 2, are those used by the USEPA for the desorbtion of Tenax
tubes used in monitoring of ambient air (1). Recovery values were
obtained by the comparison of Tenax cartridges spiked with a gas
standard of the indicated compounds versus the same standard direct, ly
By using acrtve desorbtion, the desorbtion
are greater, allowing lower detection limits to be
achieved and enabling the technique to be extended to less volatile
compounds .
injected into the GC.
efficiencies obtained
220
-------
SUMMARY
Thermal desorbtion is nn extremely useful means of injecting samples
collected on sorbent traps into a gas chromatograph for analysis. It
eliminates the use of solvents, and requires less handling, is more
sensitive, and less costly than solvent desorbtion. TEKMAR's new Model
5000 Automatic Desorber combines innovative new features with powerful
microprocessor control to produce outstanding performance for the
analysis of sorbent traps. The Desorber is applicable to monitoring of
ambient air, stack gases, and industrial hygiene analysis.
References:
1) Berkley, R.E., Bumgnrner, J.E., Driscoll, D.J., Morris, C.M.,
and Wright, L.M., "Standard Operating Procedure for the GC/MS
Determination of Volatile Organic Compounds Collected on Tenax",
U.S. Environmental Protection Agency, Environmental Monitoring
Systems Laboratory, Research Triangle Park, North Carolina,
EMSL/RTP-SOP-EMD-021, June 27, 1984
221
-------
TABLE 1: RECOVERIES OF VOLATILE ORGANIC COMPOUNDS BY ACTIVE VS PASSIVE
DESORBTION
Compound
Perflouro toluene
Toluene
Chlorobenzene
Ethy Ibenzene
o-Xylene
p-Chloro toluene
1 , 2, 4-Tri chlorobenzene
Active Desorb
104%
95%
95%
95%
94%
80%
80%
Passive Desorb
96%
76%
68%
63%
58%
46%
42%
TABLE 2: ANALYTICAL CONDITIONS
Trap:
Prepurge:
Internal Trap:
Desorb:
Transfer:
Cryofocus:
Valve:
GC Column:
Program:
Carrier:
Detector:
5/8
2 min
1/16"
8 min
1 min
-100°
270°
X 7" SS Tenax
at 10ml/min.
open bore, -100 C
at lOml/min., 190°C
at 250°C
Inject: 0.5 min © 250
Line: 210° Injector: 200°
25m x 0.32mm fused silica, 1 micron SE-54
45 for 2 min., 6 /min. to 200
Hydrogen at 50cm/s
FID, range 10 , attn. as marked
222
-------
FIGIFRE 1
223
-------
GAS IN
GAS
OUT
Figure 2: Cutaway view of tube heater
INJECTION
PORT
. - . ,
I
6 7
, ,
Figure 3: Chromatogram of focussed GC injection. Peak Identifications:
1: benzene, 2: toluene, 3: ethyl benzene, 4: chlorobenzene
5: 1,3-dichlorobenzene, 6: 1,2-dichlorobenzene
7: 1,2,4-trichlorobenzene, 6 ng each, FID attenuation 64.
224
-------
OESORB
GAS
INJECTION
PORT
IFUSED SILICA;
TRANSFER LINE
TO
DETECTOR
GC OVEN
CARRIER GAS
Figure 4: Gas Flow System
~_i~_*»JL
Figure 5:
Chromatogram of 10 liters ambient air
sampled on Tenax, FID attenuation 512. 225
Figure 6:
Chromatogram of 10 liters indoor air
sampled on Tenax, FID attenuation 512,
-------
COMPARISON OF SAMPLE COLLECTION
TECHNIQUES FOR VOLATILE ORGANIC
COMPOUNDS IN AMBIENT AIR
M. W. Holdren and R. M. Riggin
Battelle, Columbus Laboratories
Columbus, OH 43201
W. A. McClenny and J. D. Mulik
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
ABSTRACT
The objective of this study was to compare the performance of techniques
for the sampling of selected volatile organic compounds in ambient air. Specifi-
cally, three solid adsorbents (Tenax®, an experimental polyimide resin, and
Spherocarb®) were examined along with a whole air collection by SUMMA® treated
canisters.
A series of 10 experimental sampling runs were conducted at a field site
over a one-month period. During each experiment ambient air was drawn through a
sampling manifold and was continuously spiked with a mixture of fifteen volatile
organic compounds (VOCs) to give concentrations 1 to 3 ng/1 above background
air. The spiked compounds were chloroethene, 1,1-dichloroethene, dichloromethane,
3-chloropropene, l,l,2-trichloro-l,2,2-trifluoroethane, trichloromethane,
1,2-dichloroethane, 1,1,1-trichloroethane, benzene, tetrachloromethane, trichloro-
ethene, toluene, tetrachloroethene, chlorobenzene, and 1,2-dimethylbenzene. Two-
hour integrated adsorbent and canister samples were collected. Adsorbent samples
were analyzed by gas chromatographic/mass spectrometric techniques, while the
canister samples were processed with an automated gas chromatograph employing
cryogenic trapping and a multiple detection system (electron capture, flame ioni-
zation and mass selective detectors).
Compared to the three adsorbent methods, the whole air collection via canister
cryogenic trapping technique offered better precision and accuracy for the compounds
of interest. None of the three adsorbents gave optimal performance for the entire
list of compounds, although in general Tenax® gave the best results. Spherocarb®
was the best adsorbent for chloroethene (vinyl chloride), dichloromethane, and
l,l,2-trichloro-l,2,2-trifluoroethane. The polyimide material suffered from a
number of operational problems which weigh heavily against its use in ambient
air sampling.
Although the research described in this article has been funded by the United
States Environmental Protection Agency through Contract No. 68-02-3487 by Battelle's
Columbus Laboratories, it has not been subjected to the Agency's required peer
and policy review and therefore does not necessarily reflect the views of the
Agency and no official endorsement should be inferred.
226
-------
INTRODUCTION
The Methods Development and Analysis Division of the Environmental
Monitoring Systems Laboratory (EMSL) of the U.S. Environmental Protection Agency
is responsible for the development and evaluation of state-of-the-art and
emerging analytical techniques for the determination of organic compounds in
ambient air. Recently a priority listing of volatile organics has been
established and EMSL is focusing on further development of analytical
methodology for the determination of these compounds.
In general one of two approaches, whole air collection into specially treated
canisters or selective collection by solid adsorbents is used to sample volatile
organics in ambient air. By far the most widely employed adsorbent for volatile
organic compounds is Tenax®6C. Tenax®has the advantage of good thermal stability
which allows for efficient desorption of higher boiling compounds (e.g., C-12
hydrocarbons) during the analysis step. A primary limitation of Tenax®is the
low retention volume of highly volatile compounds (e.g., vinyl chloride,
1,2-dichloroethane, etc.). In order to extend the applicability of solid
adsorbent collection to more volatile compounds, EMSL has conducted development
and evaluation studies for various adsorbents which could be used in place of,
or in combination with Tenax®. The two most promising materials for this purpose
are (1) a polyimide material formed from pyromellitic anhydride and 4,4'-diamino-
diphenylsulfone, and (2) carbon molecular sieves (CMS) sold under the tradenames
Spherocarb®, Carbosieve®, or Carbosphere®.
The specific objective of the work described here was to compare the perfor-
mance of the three adsorbents (Tenax®, polyimide, and Spherocarb®) as well as
canister collection for sampling and analysis of representative volatile organic
compounds at realistic concentrations in ambient air.
PROCEDURES
Three adsorbents—Tenax®, polyimide, and Spherocarb®, as well as whole air
collection into canisters were operated in parallel to sample ambient air, using
the sampling manifold shown in Figure 1. The air stream was spiked with 1-3
ng/liter levels of each target compound in order to ensure the presence of a
detectable concentration.
227
-------
Whole air sampling was accomplished by collecting integrated samples, over
the entire two-hour sampling period, using specially treated stainless steel
canisters. A sample volume of 500 ml was taken from each canister and analyzed
by cryogenic trapping/gas chromatography using flame ionization, electron capture,
and mass selective detectors. Compound separation was facilitated with a 50
meter wide-bore fused silica capillary column (OV-1). The individual compounds
were eluted using a temperature program of -70°C to 150°C at 8°/minute.
Duplicate ten-liter samples were collected using each of the three adsorbents.
In addition five and twenty-liter samples were collected using Tenax®. All adsorbent
samples were analyzed by gas chromatography/mass spectrometry. The analytes
were thermally desorbed from the cartridges onto a liquid nitrogen-cooled trap
and subsequently transferred onto a wide-bore fused silica capillary column.
The individual compounds were eluted using a temperature program of -70°C to
150°C at 8°/minute. Components were quantified by comparing the integrated ion
intensity for a characteristic ion of each compound to that of standard injected
on the same day.
More details on the experimental portion of the program may be found else-
where (1,2).
RESULTS AND DISCUSSION
An estimate of the analytical precision of the various sampling and analytical
methods was determined from replicate analyses of canister and adsorbent samples.
Due to the limitation in software only six compounds could be monitored in real-
time with the mass selective detector. As a result, replicate analyses were
only performed on six of the fifteen compounds with samples from the canisters.
Table 1 shows the estimate of precision for these six compounds using all four
collection techniques. It is evident that, in general, the canister collection
technique offers somewhat better precision.
Data on method accuracy were obtained by flowing a known calibration mixture
through the sampling manifold and determining percent recovery of each compound
per method. The results are shown in Table 2. Again the canister sampling technique
gave the best performance. Recoveries with this technique ranged from 89 to 120
percent for the fifteen target compounds (excluding l,l,2-trichloro-l,2,2-
trifluoroethane). Tenax®, which overall showed the best compound recovery for
228
-------
the adsorbents, gave values that ranged from 8 to 104 percent (excluding
dichloromethane).
In order to compare the performance of the various methods for ambient air,
the apparent recovery of the adsorbent samples for each run was calculated,
relative to the value obtained from the analysis of the canister. Use of the
canister value was considered to be. most appropriate since this technique generally
agreed best with the concentration in the calibration cylinder, and gave better
overall precision than did any of the adsorbent techniques. Most of the values
used for the canister data were obtained using the mass selective detector, since
this detection system was less subject to potential interferences. However, the
toluene and 1,2-dimethylbenzene values were obtained using flame ionization
detection due to limitations on the number of ions which could be monitored using
the mass selective detector.
An example of the performance of Tenax® relative to canister collection for
the compound benzene is shown in Figure 2. After correcting for the outlier
data point, a relative recovery of 114 percent is obtained. Comparative data
for the various adsorbents relative to canister sampling is summarized in Table 3.
For poorly retained compounds the low volume Tenax® sample (nominally 5 liters)
was used, whereas for better retained compounds the average of duplicate 10
liter Tenax® samples was used.
As expected, none of the adsorbents performed well for all of the target
compounds. Tenax® performed best for 3-chloropropene, trichloromethane,
1,2-dichloroethane, 1,1,1-trichloroethane, benzene, tetrachloromethane,
trichloroethene, tetrachloroethene, chlorobenzene, and 1,2-dimethylbenzene.
Tenax® and polyimide gave essentially identical results for toluene. As
expected, on the basis of breakthrough volume data, Tenax® gave essentially no
recovery for chloroethene (vinyl chloride) and 1,1-dichloroethene (vinylidene
chloride).
Polyimide performed better than both Spherocarb® and Tenax® only for 1,1-
dichloroethene, although only 55 percent recovery and 28 percent standard devia-
tion were obtained. Polyimide gave essentially no recovery for chloroethene, in
spite of the good recovery obtained for this compound in high purity air
sampling (2). Polyimide also gave disappointingly low and variable recovery for
l,l,2-trichloro-l,2,2-trifluoroethane (Freon 113), 1,1,1-trichloroethane, and
tetrachloromethane. An operational problem which weighs heavily against the use
of this material for ambient air sampling is the adsorption of a significant
229
-------
amount of moisture onto the adsorbent. The presence of water in the matrix led
to the chromatographic column plugging during virtually all of the ambient air
sampling runs. A similar phenomenon is observed for Spherocarb®. However, in
that case the cartridge is prepurged with dry air at room temperature prior to
analysis, to remove adsorbed moisture. A similar approach was not successful in
eliminating the problem for the polyimide material, indicating that the collected
moisture is difficult to desorb due to kinetic or thermodynamic factors.
Spherocarb®gave the best results of the three resins only for chloroethene,
dichloromethane, and l,l,2-trichloro-l,2,2-trifluoroethane. Spherocarb®gave
extremely poor results for 1,1-dichloroethene, 3-chloropropene, 1,2-dimethylbenzene,
and tetrachloromethane. In the case of 1,1-dichloroethene artifactually high
recovery, possibly due to dehydrohalogenation of 1,1,1-trichloroethane, was a
major problem. Although low recoveries were anticipated for the other three
compounds, on the basis of earlier work, the low recovery of 1,1-dichloroethene
was not observed previously (3).
In spite of the problems discussed above, the data set is somewhat encouraging
in terms of the implications for ambient air sampling. Inspection of the raw
data reveals that, except for the major problem areas discussed above, the individual
values generally agree with the values obtained from canister sampling within a
factor of two or better (i.e., 50 to 150 percent relative recovery).
CONCLUSIONS AND RECOMMENDATIONS
Compared to the three adsorbent methods tested, whole air collection via
canisters results in better precision for the compounds of interest in this study.
An estimate of precision (% standard deviation) for the canister sampling method
ranged from 4 to 10 percent. For Tenax®, which gave the best performance for
the adsorbents, precision values ranged from 8 to 16 percent. Data on recovery
of the target compounds in the spiked manifold clearly indicate better accuracy
with the canister sampling procedure. When comparing measured recoveries with
expected concentrations, the canister sampling approach yielded values from 89
to 120 percent. Recoveries, using Tenax® adsorbent, ranged from 8 to 104 percent.
Additional studies should be undertaken under controlled conditions with
analytical uncertainties minimized as much as possible by employing the same
analysis procedure for all collected samples. Sampling techniques to be compared
230
-------
should include distributive air volume sampling with Tenax® adsorbent, passive
sampling using personal exposure devices, and whole air collection in canisters.
REFERENCES
(1) Holdren, M. VL, Smith, D. L., and R. N. Smith. Field Evaluation of An
Automated Cryogenic Preconcentration and Gas Chromatographic System.
Contract No. 68-02-3487 (WA-37). U.S. Environmental Protection Agency,
Research Triangle Park, NC, January, 1985.
(2) Riggin, R. M. and Markle, R. A. Comparison of Solid Adsorbent Sampling
Techniques for Volatile Organic Compounds In Ambient Air. Contract No.
68-02-3487 (WA-31). U.S. Environmental Protection Agency, Research Triangle
Park, NC, January, 1985.
(3) Riggin, R. M. Evaluation of Carbon Molecular Sieves As Adsorbents for the
Determination of Volatile Organic Compounds. In: Proceedings of Quality
Assurance in Air Pollution Measurements, Air Pollution Control Association,
Boulder, CO, October, October, 1984, pp. 14-18.
231
-------
TABLE 1. ANALYTICAL PRECISION DATA FROM REPLICATE
ANALYSES OF SOLID ADSORBENT AND
CANISTER SAMPLES!a;
Compound Canister Collection Tenax®
Tri chl orotri f 1 uoroethane
Trichloromethane
1,1, 1-Tr i chl oroethane
Tetrachl oromethane
Trichloroethylene
Tetrachl oroethene
10.1
7.3
8.7
6.4
7.3
4.0
11.8
7.8
8.0
16.0
14.3
11.8
Polyimide
39.0
13.5
18.1
26.4
15.6
8.9
Spherocarb®
9.8
27.8
19.1
27.9
18.3
28.4
(a) Precision = % Standard Deviation.
232
-------
TABLE 2. RESULTS FOR CALIBRATION CYLINDER SAMPLING EXPERIMENT
Compound
Expected Percent Recovery(a)
Concentration, Canister
ng/L Collection Tenax® Polyimide Spherocarb®
Chloroethene(b)
l,l-Dichloroethene(b)
Dichloromethane(b)
3-Chl oropropene (b )
l,l,2-Trichloro-l,2,2,-Tri-
fluoroethane(b)
Trichloromethane(b)
l,2-Dichloroethane(b)
l,l,l-Trichloroethane(b)
Benzene
Tetrachloromethane(b)
Trichloromethane
Toluene
Tetrachloroethene
Chlorobenzene
1,2-Dimethyl benzene
26
17
13
19
34
27
24
30
15
36
27
18
34
22
20
103
112
107
95
230
100
120
117
93
103
89
90
118
118
120
8.3
55
360
79
47
104
96
89
73
89
76
50
65
43
23
79
91
119
66
60
85
40
107
115
118
87
65
68
47
28
58
76
100
60
123
93
63
75
105
64
93
52
76
33
13
(a) Average for duplicate samples.
(b) Low volume (nominally 5 liters) Tenax® value used for these compounds.
Medium volume (nominally 10 liters) Tenax® value used for all other compounds,
233
-------
TABLE 3. PERFORMANCE DATA FOR SOLID ADSORBENTS
RELATIVE TO CANISTER SAMPLING
Tenax"9
Breakthrough
Volume(a),
Average Recovery Relative
To Canister Sampling, %
Compound
Chloroethene
1 , 1-Di chl oroethene
Dichloromethane
3-Ch1oropropene(c)
l,l,2-Trichloro-l,2,2-Tri-
fluoroethane
Trichloromethane(c)
l,2-Dichloroethane(c)
l,l,l-Trichloroethane(c)
Benzene
Tetrachloromethane(c)
Tri chl oroethene
Toluene
Tetrachl oroethene
Chlorobenzene
1,2-Dimethyl benzene
(a) Data from Reference 2 at
Liters/Cartridge
0.8
Not Given
4
6
Not Given
13
18
9
27
13
28
122
106
249
334(e)
90°F.
Tenax® P
-(b)
-(b)
83(21)
87(35)
39(25)
100(36)
100(15)
130(42)
110(18)
110(37)
112(26)
70(19)
88(27)
78(35)
55(21)
olyimide
-(b)
52(28)
86(31)
140(68)
17(14)
75(25)
65(15)
51(14)
130(34)
53(19)
100(33)
70(17)
78(30)
57(21)
40(15)
Spherocl
73(23)(
410(260)
85(12)
29(14)
69(30)
65(17)
75(14)
46(10)
140(63)
29(9)
90(33)
43(8)
72(30)
53(19)
20(7.9)
(b) No meaningful data obtained.
(c) Low volume (nominally 5
Medium volume (nominally
(d) Value in parentheses is
(e) Value for ethyl benzene.
liters) Tenax® val
10 liters Tenax®
standard deviation
ue used for these
value used for al
for all sampling
compounds.
1 other compounds.
runs.
234
-------
a:
CD
o
CXL
o.
CD
oo
Di
CD
i— t
Di
Q
UJ
CO
o
u_
oo
oo
-------
UJ
UJ
N
111
m
D
D
d
tt CO
^ d
it II
c
a
o> a
SCO
1 1 1 1 1
o o o o o
N *± T-m CO If)
CM CM CM T^ T^
1
ci
|
O
Oj
1-
D
1
C
1
1
_
D —
—
—
1 1 1
0000
en
0 |
^m ^^m
C
0 OC
CO
p
to
p
*
p
CM
O
o
O i—i
«=c
at. a
tvl UJ i—i
2: c_> ca:
UJ Z I—
co o ca
o o
0£.
o oo uj
00 O <
O I-H >
i—( h—
z oo
o z
i—i =3
h- cc:
o: a:
ca uj
o o-
UJ
O h-
O LU
O
ca
o
i—i X
CI3
I— o
o uj
_! ixi
Q. >—I
D; Q.
UJ GO
uj a_
I— s
eC
s: oo
o
a; o;
•a: o
o a;
oo u_
a;
Z3
C3
UJ 00
rD 1—1
—J z
«=c
-------
BENZENE AND TOLUENE CONCENTRATIONS IN LAKE CHARLES,
LOUISIANA AMBIENT AIR
Dennis M. Casserly
University of Houston-Clear Lake
Houston, Texas 77058
Kenneth K. O'Hara
McNeese State University
Lake Charles, Louisiana 70609
INTRODUCTION
The concentration of volatile organic compounds (VOC's) in ambient air is
primarily of interest because of their involvement in the formation of photo-
chemical smog. However,among the vapor phase components, there also exist
mutagenic and potentially carcinogenic vapors (Krost ^t aj_, 1982). The petro-
chemical industries and related activities are generally assumed to be the major
sources of VOC's in ambient air. Significant sources that have contributed to
the degradation of urban air quality include: waste lagoons (Shen, 1982);
activated sludge treatment plants (Lurker et al_, 1982); chemical waste sites
(Pellizarri, 1982); and automobiles (Pilar and Graydon, 1973). Ubiquitous back-
ground sources are animal and plant matter,, forest and grass fires and seepage
of crude oil (Brief et a]_, 1980). Atlas and Giam (1981) have found VOC's in
remote regions of the North Pacific marine atmosphere implying a global distri-
bution and transport for some refractory compounds.
A large number of VOC's have been regularly found in urban centers through-
out the world: 20 mutagenic VOC's in seven U.S. cities by Singh .et al_ (1983);
25 VOC's at three urban sites in New Jersey by Harkov et al_ (1983); 22 halo-
genated hydrocarbons in Louisiana by Pellizarri (1982); and 48 VOC's regularly
found in Sydney, Australia by Nelson and Quigley (1982), which they estimated
contributed to 90% of the non-methane hydrocarbons measured.
North of Lake Charles are pine plantations and to the south are salt
marshes both of which are seasonally burned to control undesirable plant
growth. Throughout the area are oil and gas fields. In the greater Lake
Charles area are chemical industries, refineries, hazardous waste sites and
motor vehicle activity all of which would contribute to a complex chemical
mixture in the ambient air.
The purpose of this project was to provide a baseline study of the volatile
organics benzene and toluene in ambient air at two distinct sampling locations,
237
-------
one near the center of industrial and urban activity and the other remote and
generally upwind from such areas.
METHODS
Samples were collected from January through December 1983 at existing sites
established by the Louisiana Department of Natural Resources. The stations are
part of EPA's network of State and Local Air Monitoring Stations (SLAMS).
Air samples were collected according to ANSI/ASTM (1979) procedure
D-1605-60, Standard Recommended Practices for Sampling Atmospheres for
Analysis of Gases and Vapors,and procedure D-3686-78, Standard Practice for
Sampling Atmospheres to Collect Organic Compound Vapors. A vacuum pump was
used to move air through a coconut charcoal tube (100/50 mg, 20/40 mesh). Sam-
pling occurred for a 24 hour period at a flow rate of one liter per minute,
maintained by a critical orifice. Flow rate was calibrated before and after
each sample collection using a 100 ml bubble flow meter. The collection tubes
were kept verticle to prevent channeling. After sampling, the tubes were capped
and returned to the laboratory for desorption and analysis.
Analyses were performed according to ANSI/ASTM method D-3686-78, Standard
Practices for Analysis of Organic Compound Vapors Collected by the Activated
Charcoal Method (1979). Front and back charcoal sections were placed in sep-
arate vials with teflon caps and desorbed with 0.5 ml carbon disulfide. Sample
vials were agitated for 30 minutes for desorption to occur. A gas chromatograph
with a flame ionization detector was used for analysis. The column was
20' x 1/8" stainless steel with 10% SP-1000 on 80/100 Supelcoport. Chromato-
graphic conditions were programmed for maximum separation of sample components.
Analyses were conducted in triplicate.
RESULTS AND DISCUSSION
The procedures employed were adequate to quantify low concentrations of
benzene and toluene in 53 samples analyzed. All field blanks were negative and
no sample showed breakthrough. Data were corrected for temperature and pres-
sure and reported for standard conditions (0°C and 760 mmHg). Olson et _al_ (1983)
found the charcoal technique suitable for quantitative determination of eight
organics which included benzene and toluene. According to Kring et al (1984), the
238
-------
method has a precision of 5% relative standard deviation at 95% humidity.
Levine and Schneider (1982) found sample recoveries to be greater at flow rates
of 2 liters per minute than at lower flow rates of 0.2 liters per minute, pre-
sumably due to turbulent flow that increases adsorption efficiency. Sampling
time showed no effect on recoveries but high organic loads on the charcoal
resulted in breakthrough.
In this study, both benzene and toluene concentrations were observed to
follow log-normal distributions. Although both compound concentrations were
slightly higher at the industrial site than the upwind site, paired t-test
results indicated no significant differences (p>0.05) in concentrations between
the two sites. The geometric mean and geometric standard deviation for
benzene was 6.6 +; 1.7 ug/M3 with 10% of the values exceeding 14.5 ug/M3. The
toluene concentrations averaged 5.5-+_ 1.7 ug/M3 with 10%-of the values exceeding
12.3 ug/M3. Slightly higher concentrations of both compounds were noted during
the summer months presumably due to increased volatile losses during warm
temperature periods. Toluene and benzene concentrations were positively cor-
related (r = 0.60) with an average ratio of 0.87:1.
Sievers _et al_ (1980) reported global background averages of 0.6 - 6.4 ug/M3
for benzene and 4.5 - 19 ug/M3 for toluene. Concentrations observed at indus-
trial and urban centers tend to be higher. Harkov et_ al_ (1983) found average
values of 3.4 ug/M3 for benzene and 6.8 - 17.5 ug/M3 for toluene in New Jersey.
Nelson and Quigley (1982) reports benzene concentrations of 8.2 ug/M3 and toluene
concentrations of 33.4 ug/M3 in Sydney, Australia. In Toronto, Pilar and Graydon
(1973) measured benzene concentrations at 41 ug/M3 and toluene concentrations of
113 ug/M3.
The levels of benzene and toluene in the Lake Charles area ambient air were
within the ranges found in other parts of the world. There appears to be no
significant difference between the industrial and the upwind site, indicating
no significant point sources of benzene and toluene in the area but rather
common diffuse sources.
REFERENCES
1. Atlas, E. and C. S. Giam, "Global Transport of Organic Pollutants: Ambient
Concentrations in the Remote Marine Atmosphere," Science, Vol.211,
No.9 (1981).
239
-------
2. Brief, R. S., J. Lynch, T. Bernath and R. A. Scala, "Benzene in the Work-
place," American Industrial Hygiene Association Journal, No.41, September
(1980).
3. Harkov, R., B. Kebbekus, J. W. Bozzelli and P. K. Lioy, "Measurement of
Selected Volatile Organic Compounds at Three Locations in New Jersey During
the Summer Season," Journal of the Air Pollution Control Association, Vol.
33, No.112 (1983). '
4. Kring, E. V., 6. R. Ansul, T. J. Henry, J. A. Morello, S. W. Dixon, J. F.
Vasta and R. E. Hemingway, "Evaluation of the Standard NIOSH Type Charcoal
Tube Sampling Method for Organic Vapors in Air," American Industrial Hygiene
Association Journal, Vol.45, No.4 (1984).
5. Krost, K. J., E. D. Pellizzari, 6. S. Walburn and S. A. Hubbard, "Collection
and Analysis of Hazardous Organic Emissions," Analytical Chemistry, Vol.
54, 810 (1982).
6. Lurker, P. A., C. S. Clark and V, J. Elia, "Atmosphereic Release of Chlorinated
Organic Compounds from the Activated Sludge Process," Journal of the Hater
Pollution Control Federation. Vol.54, No.12 (1982).
7. Nelson, P. F. and S. M. Quigley, "Non-Methane Hydrocarbons in the Atmosphere
of Sydney, Australia," Environmental Science and Technology. Vol.16, No.10
(1982).
8. Olson, R., D. T. Williams, and P. D. Bothwell, "Charcoal Tube Technique for
Simultaneous Determination of Selected Organics in Air," American Industrial
Hygiene Association Journal, Vol.44, No.7 (1983).
9. Pellizzari, E. D., "Analysis for Organic Vapor Emissions Near Industrial and
Chemical Waste Disposal Sites," Environ. Sci. Technol., Vol.16, No.11
(1982).
10. Pilar, S. and W. F. Graydon, "Benzene and Toluene Distribution in Toronto
Atmosphere." Environ. Sci. Techno!., Vol.7, No.7 (1973).
11. Shen, T. T., "Estimation of Organic Compound Emissions From Waste Lagoons,"
Journal of the Air Pollution Control Association, Vol.32, No.l (1982).
12. Sievers, R. E., R. M. Barkley, D. W. Denney, S. D. Harvey, J. M. Roberts,
M. J. Bellinger and A. C. Delaney, "Gas Chromatographic and Mass Spec-
trometric Analysis of Volatile Organics in the Atmosphere," Sampling and
Analysis of Toxic Organics in the Atmosphere, ASTM STP 721 (1980).
13. Singh, H. B., L. J. Salas and R. E. Stiles, "Distribution of Selected
Gaseous Organic Mutagens and Suspect Carcinogens in Ambient Air,"
Environ. Sci. Technol. Vol.16, No.12 (1982).
240
-------
DETERMINATION OF 2-DIETHYLAMINOETHANOL IN AIR
Michael R. Whitbeck
Associate Research Professor
Desert Research Institute, University of Nevada System, Reno, NV
ABSTRACT
The extensive use of 2-diethylaminoethanol (DEAE) provides the
potential for human exposure from polyurethane foams, humidifier, etc.
In this paper an ion-exchange chromatographic procedure for determining
DEAE concentration in air is presented that overcomes difficulties asso-
ciated with existing methods.
INTRODUCTION
2-Diethylaminoethanol (DEAE) is widely used in industry as a solu-
bilizer, emulsifier, anticorrosion additive to boilers, sterilizers and
food processing equipment, as a photoactivator and in epoxy resin for-
mulations. Because of this extensive use, the potential for human expo-
sure to 2-diethylarninoethanol and its degradation products as air
contaminants existsd^.S) and there is a need for a simple and rapid
method to determine its concentration in air.
Colorimetric methods for. analysis of DEAE in air have been used
based on oxidation of the aminoalcohol by dichromate^) or hypo-
chlorite^). Derivatization of the aminoalcohol to form a colored pro-
duct has also been used(3»5)> and a gas chromatography method has been
described^). These methods require very large sample volumes, suffer
from extensive sample preparation and extraction steps, or lack the
desired specificity for ambient air analysis. In this paper, an ion
chromatographic method for determining DEAE concentrations in air will
241
-------
be presented that overcomes the difficulties associated with the
existing methods.
EQUIPMENT AND MATERIALS
In this paper all water used was 18 megohm deionized water from a
Waters Associates Milli-Q water system. The DEAE was Aldrich Gold Label
(99+% pure). The chromatography was performed using a Dionex System 14
1C with a 0.5 ml sample loop and 0.005 N hydrochloric acid as the
eluent at a flow rate of 150 ml/hr. The chromatograph was equipped with
a cation concentrator column (#030830), a separation column (#030831),
and a suppressor column (#030834).
EXPERIMENTAL
A calibration curve was made by diluting DEAE with deionized water
to form solutions of 1-20 ppmm DEA-E. These solutions gave a linear
increase in peak height and peak area as a function of concentration in
the 5-20 ppmm range (Figure 1). Ammonia and amines at low con-
centrations typically give such a nonlinear response in ion chroma-
tography using a conductance detector. Although this may be corrected
by adding a chloride postcolumn, in this study the concentrations were
kept in the linear range of Figure 1.
To test the trapping efficiency of water impingers for DEAE a gas
phase sample was prepared by injecting 1.6 microliters/hr of DEAE with a
syringe pump into nitrogen flowing at 0.4 1pm to produce a DEAE con-
centration of 59 mg/m3 +/- 20%. This test stream was sampled by two
microimpingers in series. Each microimpinger contained 10 ml of
deionized water and was sampled for 9 minutes. The high solubility of
242
-------
DEAE in water combined with the short sampling times permits the use of
deionized water as the trapping media in place of dilute acid^).
Analysis of the first impinger gave a value of 17 ppmm. The 17 ppmm
corresponds to a gas phase DEAE concentration of 47 mg/m3 and is within
the expected accuracy for the syringe pump. No DEAE was detected in the
second impinger.
DISCUSSION
The determination of 2-diethylaminoethanol in air by ion chroma-
tography offers two significant advantages over existing methods, first
is the use of a simple water impinger for sampling, and second, there is
no need for derivatization or extraction steps. A typical chromatogram
is shown in Figure 2. The two small peaks after the water dip are from
sodium and ammonium cations present as trace, impurities. These ions
elute at much shorter times than DEAE- and thus do not interfere with the
method.
The principal disadvantage is the nonlinear response at low con-
centrations of DEAE. This presents a minor difficulty that may be over-
come by taking somewhat larger samples so that the aqueous aminoalcohol
concentration falls in the linear response region.
REFERENCES
(1)
(2)
(3)
P. Pryor, Health Hazard Evaluation Report, HETA 80-148-1025, NIOSH.
May (1980).
K.P. McManus, Health Hazard Evaluation Report, HETA 81-247-958,
NIOSH, Sept. (1981).
F.A. Miller, R.F. Scherberger, K.S. Tischer, and A.M. Weber,
Determination of Microgram Quantities of Diethanolamine,
2-methylaminoethanol, and 2-diethylaminoethanol in Air, Am Ind.
Hygiene Assoe. J., 28, 330-334 (1967).
243
-------
(4) o.N. Putilina, and N.T. Yarym-Agavera, Photometric Method for the
Determination of Tertiary Fatty Amines and Aminoalcohols in the
Air, Gig. Tr. Prof. Zabol., 7, 55-57 (1981).
(5) E.N. Pomozova, and N.F. Volokhova, Determination of
Diethylaminoethanol in the Air, Uch. Zap. Mosk. Nauchno Isslad.
Inst. Gig., 22, 172-173 (1975).
(6) G.O. Wood and J.W. Nickols, Development of Air Monitoring
Techniques Using Solid Sorbents, LASL Report LA-7295-PR,
NIOSH-IA-77-12.
244
-------
345
PEAK HEIGHT
8
Figure 1: Calibration curve for 2-diethylaminoethanol.
245
-------
t
"o
.c
E
H
O
ID
O
_L
JL
J_
O
4567
TIME (min)-
8
10
12
Figure 2: Typical chromatogram: (a) injection pip, (b) water dip, (c)
5.0 ppmm DEAE.
246
-------
ANALYSIS OF SEMI-VOLATILE ORGANIC
COMPOUNDS USING SUPERCRITICAL FLUID METHODOLOGIES
Bob W. Wright, Edward K. Chess, Clement R. Yonker and Richard D. Smith
Chemical Methods and Kinetics Section, Pacific Northwest Laboratory
(Operated by Battelle Memorial Institute) Richland, WA 99352
INTRODUCTION
Supercritical fluids possess favorable physical-chemical properties which
render them useful for many analytical applications. Recently, a number of
potentially important supercritical fluid based analytical methodologies have
emerged which include capillary-column supercritical fluid chromatography (SFC)
(1), capillary-column supercritical fluid chromatography-mass spectrometry
(SFC-MS) (2), direct supercritical fluid injection-mass spectrometry (DFI-MS)
(3), and direct supercritical fluid extraction-mass spectrometry (SFE-MS) (4).
Although these techniques are in an early stage of development, they have shown
considerable promise for the characterization of complex organic mixtures.
The compressibility of supercritical fluids is large above the critical
temperature and small changes in pressure result in large changes in the
density of a fluid (5). Molecular interactions increase due to shorter
intermolecular distances and solvating characteristics approaching that of a
liquid are imparted. However, the viscosity and the solute diffusivity remain
similar to that of a gas. These properties form the basis of supercritical
fluid chromatography (SFC) and result in the potential for the separation of
less volatile compounds than is possible by gas chromatography and for signifi-
cantly enhanced chromatographic efficiency per unit time relative to liquid
chromatography. These same properties are also advantageous for the efficient
removal and transport of selected organic compounds from matrix components
using supercritical fluid extraction (SFE).
This study demonstrates the applicability of SFC and analytical SFE for
the analysis of semi-volatile compounds. Mixtures of nitro-polycyclic aromatic
hydrocarbons that are not ammenable to gas chromatography were separated using
SFC with tentative compound identifications made by SFC-MS. Comparisons of
analytical SFE of XAD-2 resin and NBS Urban Dust (SRM 1649) to conventional
Soxhlet extraction are also discussed.
247
-------
SUPERCRITICAL FLUID CHROMATOGRAPHY
The instrumentation used for capillary supercritical fluid chromatography
has been previously described (6). Briefly, the apparatus used a modified
syringe pump (Varian 8500) under microcomputer control to generate the high-
pressure and pulse-free supply of mobile phase and a modified gas chromatograph
(Hewlett-Packard 5710) to provide constant temperature conditions and flame
ionization detection. A 60 nl HPLC valve (Valco C14W) and an injection splitter
were used for sample introduction onto a deactivated 15m x 50 Jim fused silica
capillary column coated with crosslinked 5% phenyl polymethylphenylsiloxane
stationary phase (7).
A capillary supercritical fluid chromatogram of a nitrated phenanthrene
mixture is shown in Figure 1. Carbon dioxide at 100°C was used for the mobile
phase. The pressure was programmed in such a way to create linear density ramps
as indicated in the time and density legends. All five of the mononitrophenan-
threne isomers were discernible and would have been completely resolved if their
concentrations were lower. The mononitrophenanthrene dimers were also eluted
with several isomers being separated. These dimers are not sufficiently volatile
to be eluted under normal gas chromatographic conditions. A mass spectrum ob-
tained by SFC-MS of one of these dimer isomers is shown in Figure 2. Ammonia
was used for the chemical ionization reagent gas and consistent fragmentation
was obtained. A similarly obtained chromatogram of a nitrated 9-hydroxyfluorene
mixture is shown in Figure 3. During nitration fluorenone was also formed.
Most notable, however, is the elution of the 9-hydroxynitrofluorene isomers
which are thermally labile and reported to be non-gas chromatographable (8).
These examples illustrate the applicability of SFC and SFC-MS for the analysis
of nonvolatile and thermally labile materials.
SUPERCRITICAL FLUID EXTRACTION
Off-line analytical supercritical fluid extraction was accomplished using
a small, flow-through high-pressure extraction cell. The high-pressure supply
was generated using either a syringe pump (Varian 8500) or a reciprocating HPLC
pump (Rainin, Rabbit HPX) that were modified to operate under pressure control.
The extraction cell was maintained at constant temperature in a gas chromato-
graphic oven. The supercritical fluid effluent was decompressed using a short
length of heated 50 \im stainless steel tubing that was crimped at the exit end.
248
-------
30 45
Time (minutes)
0.125 0.125 0.275 0.425 0.538 0.650
Density (g cm~3)
Figure 1. SFC chromatogram of nitrated phenanthrene mixture.
100
80
60
§
40
20
NH3CI
MH-30) +
(M + 18)4
0|—
300
350
400
m/z
Figure 2. NHs CI mass spectrum of a nitrophenanthrene
dimer obtained by SFC-MS.
249
-------
SFC
CO2, 100°C
l_
15
_l_
30 45
Time (minutes)
60
0.125 0.125 0.275 0.425
Density (g cm"3)
0.575
Figure 3. SFC chromatogram of nitrated 9-hydroxyfluorene mixture.
The resulting effluent was collected and trapped in a closed flask cooled to
liquid nitrogen temperatures. A more complete description of this instrumen-
tation is given elsewhere (9).
The data obtained for the supercritical fluid extraction of approximately
one gram of XAD-2 resin spiked at 50 ppm with a model compound mixture is listed
in Table 1. Carbon dioxide at 150°C and 400 atmospheres was used for the extrac-
tion. After SFE the sample was Soxhlet extracted with methylene chloride to
determine if any additinal material could be recovered. As indicated, SFE with
this fluid did not completely remove the higher molecular weight material. How-
ever, since only a small volume (s 200 cm3) of extraction fluid was utilized
and the larger components have lower solubility, it is reasonable that a more
exhaustive extraction or modified fluid phase would have recovered the remaining
material. Similar extraction data for NBS Urban Dust are compared in Table 2.
250
-------
Two samples were extracted; Soxhlet extraction followed by SFE and SFE followed
by Soxhlet extraction. The SFE recovered additional material not extracted by
Table 1. XAD-2 Resin Extraction Recovery
Compound SFEa Soxhletb After SFE
Chrysene
Benzanthrone
1-Nitropyrene
Dibenzo(a,i)carbazole
Coronene
Rubrene
75
79
83
95
81
25
0
1
0
0
17
19
aC02 at 15QQC and 400 atm
DMeCl2 for 16 hrs.
Table 2. Extraction Comparison of NBS Urban Dust
Recovery (|ig/g)
Compound
Pyrene
Benzo(e)pyrene
SFEa After
Soxhletb
1.9
Soxhletb
2.6
0.7
SFEa
5.9
1.3
Soxhletb
After SFEa
1.3
aC02 at 125°C and 400 atm
bMethylene chloride at 16 hrs.
the Soxhlet extraction and only a small amount of material was recovered by
Soxhlet extraction following SFE. An important point is the supercritical
fluid extractions required less than a hour as compared to sixteen hours for
the Soxhlet extractions. These data illustrate the potential for very rapid
and efficient extractions of adsorbents and particulates to be accomplished by
SFE. Even more efficient extractions can be anticipated using supercritical
fluid systems with higher solvating properties.
251
-------
ACKNOWLEDGEMENT
This work was supported by the U.S. Department of Energy, Office of Health
and Environmental Research and by the U.S. Environmental Protection Agency
through a Related Services agreement with the U.S. Department of Energy under
Contract DE-AC06-76RLO-1830. Although the information in this document has
been funded in part by the U.S. Environmental Protection Agency under Inter-
agency Agreement Number DW 89930650-01-1 with the U.S. Department of Energy, it
does not necessarily reflect the views of the agency and no official endorsement
should be inferred. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
REFERENCES
1. Fjeldsted, J. C. and M. L. Lee. 1984. Capillary Supercritical Fluid
Chromatography. Anal. Chem. 56: 619A-628A.
2. Smith, R. D., W. D. Felix, J. C. Fjeldsted and M. L. Lee. 1982. Capillary
Column Supercritical Fluid Chromatography/Mass Spectrometry. Anal. Chem.
54: 1883-1885.
3. Smith, R. D. and H. R. Udseth. 1983. Mass Spectrometry with Direct Super-
critical Fluid Injection. Anal. Chem. 53: 2266-2272.
4. Smith, R. D. and H. R. Udseth. 1983. New Method for the Direct Analysis
of Supercritical Fluid Coal Extraction and Liquefaction. Fuel 62: 466-468.
5. Gouw, T. H. and R. E. Jentoft. 19/5. Practical Aspects in Supercritical
Fluid Chromatography. Advan. Chromatoqr. 13:1-40.
6. Wright, B. W. and R. D. Smith. 1984. Application of Capillary Supercritical
Fluid Chromatography to the Analysis of a Middle Distillate Fuel. Chromato-
graphia 18: 542-545.
7. Wright, B. W.,P. A. Peaden, M. L. Lee and T. J. Stark. 1982. Free-Radical
Crosslinking in the Preparation of Non-Extractable Stationary Phases for
Capillary Gas Chromatography. J. Chromatogr. 248: 17-34.
8. Paputa-Peck, M. C., R. S. Marano, D. Schuetzle, T. L. Riley, C. V. Hampton,
T. J. Prater, L. M. Skewes, T. E. Jensen, P. H. Ruehle, L. C. Bosch and W.
A. Duncan. 1983. Determination of Nitrated Polynuclear Aromatic Hydro-
carbons in Particulate Extracts by Capillary Column Gas Chromatography with
Nitrogen Selective Detection. Anal. Chem. 55: 1946-1954.
9. Wright, B. W. and R. D. Smith. 1985. Analytical Supercritical Fluid Extrac-
tion of Adsorbent Materials. Submitted for publication.
252
-------
VALIDATION OF AN EMISSION MEASUREMENT METHOD
FOR INORGANIC ARSENIC
Thomas E. Ward
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
R. K. M. Jayanty
Research Triangle Institute
Research Triangle Park, NC 27709
ABSTRACT
The U.S. Environmental Protection Agency's Proposed Method 108 for measure-
ment of inorganic arsenic emissions from stationary sources has been evaluated
both in the laboratory and field. Details of the evaluations are given through
analysis of laboratory samples, preparation of filter and impinger audit samples
for field use and stability studies, and two field tests of the method using
dual and quad sampling trains at a copper smelter plant and a glass manufacturing
plant. Several conclusions and recommendations have been made regarding the
method.
INTRODUCTION
The Environmental Monitoring Systems Laboratory (EMSL) of the U.S. Environ-
mental Protection Agency has determined the industries and processes of station-
ary sources which emit significant inorganic arsenic emissions. Included among
these are the five most significant sources chosen by EPA's Office of Air
Quality Planning and Standards (OAQPS) for original consideration for regulation
of inorganic arsenic emissions: glass plants, secondary lead smelters; primary
copper smelters; cotton ginning activities; and zinc oxide manufacturing.
Presently, OAQPS is recommending only primary copper smelters and glass plants
for these regulations.
253
-------
A primary concern of EMSL has been validation of the procedure used for
sampling and analysis of the arsenic-containing emissions. EMSL determined
that the proposed EPA Method 1081, which involves collection in impinger solu-
tions and on filters and measurement using atomic absorption spectrophotometry,
was the best available method for measuring these arsenic emissions. Therefore,
a program to validate this method was designed by EMSL. The validation approach
included a multicomponent project which involved both laboratory and field
studies. Research Triangle Institute reviewed reports describing sampling and
analysis of arsenic, evaluated EPA Method 108 through analysis of laboratory
samples, prepared audit samples for field use and studied the stability of
audit samples. PEDCo Environmentalists, Inc., collected samples at a copper
smelter and a glass manufacturing plant and analyzed the collected samples.
DISCUSSION
Laboratory studies conducted to evaluate the performance of EPA Method 108
included: recovery and measurement of audit materials, stability of audit
materials, precision of the method from sample preparation to measurement, and
recommendations regarding analytical aspects of the method. Standard samples
containing arsenic with loadings approximating emissions expected at field test
sites were prepared in the laboratory. A total of 36 impinger samples and 36
filter samples were prepared from a concentration matrix of six different
impinger and filter loadings. Loadings ranged from 0.013 mg to 5.0 mg arsenic
for the impinger samples and from 1.4 mg to 52 mg arsenic for the filter samples.
Ampules were used to contain the impinger samples. Selected filter and impinger
samples from each concentration set were analyzed by atomic absorption spectro-
photometry. The precision of the analytical phase of Method 108 was determined
from triplicate analyses of standards. The relative standard deviation (RSD)
254
-------
of the analytical method was approximately 1.0 percent in the 40- to 100-micro-
grams arsenic per milliliter Ug As/ml) high concentration range; and in
the low concentration analysis range,, approximately 1.0% at the O.lb ^g As/ml
level, and 7% near the detection limit level of 0.01
-------
The standard deviations of the quad run (Plant No. 2) had a low value of
0.10 mg/dNm3, a high value of 0.51 mg/dNm3, and a pooled mean value of 0.37
mg/dNm3. The RSD values ranged from 0.95 to 5.5 percent, with the value for the
precision of the method at RSD value of 3.85 percent. The mean arsenic concen-
tration of the individual quad runs ranged from 9.18 to 10.55 mg/dNm3 with an
overall mean of 9.67 mg/dNm3, which indicated a generally consistent process
operation throughout the test period. Once again, the results indicated that
an acceptable degree of precision was achieved. The detailed results of all
tests are presented in the project report3.
CONCLUSIONS AND RECOMMENDATIONS
Several conclusions and recommendaitons have been made regarding the
proposed EPA Method 108 for the determination of arsenic emissions from station-
ary sources. These include: (1) the method is relatively straightforward and
is reasonable with respect to time and cost; (2) during its preparation, greater
care of the sample should be exercised than the proposed method requires,
including avoidance of bumping and spattering during the evaporation process;
(3) filter and impinger audit samples were found to have an acceptable period
of usefulness based upon a six-month stability study; and (4) an acceptable
degree of precision of Method 108 is achieved from the field tests. Graphite
furnace atomic absorption techniques have an improved analytical range over
hydride techniques, are more rapid and, therefore, are recommended to be an
adequate measurement substitute.
REFERENCES
1. Federal Register, "EPA Proposed Method 108 - Determination of Particulate
and Gaseous Arsenic Emissions," Vol. 48, No. 140: pp 33166-33172.
256
-------
2. Mitchell, W. J., and M. R. Midgett. "A Means to Evaluate the Performance
of Stationary source Test Methods," Env. Sci. Tech. 10: 85-88 (1976).
3. Ward, T. E., et al. Validation of an Emission Measurement Method for
Inorganic Arsenic from Stationary Sources: Proposed Method 108 - Laboratory
and Field Test Evaluations. EPA-600/4-84-080, October 1984.
257
-------
STUDIES ON THE DYNAMIC IMPINGER SAMPLING SYSTEM
0. C. Pau, .]. E. Knoll and M. R. Midgett
The dynamic impinger sampling system uses a fixed-bed, counter-current
impinger to collect gaseous samples into an absorbing solvent. The flow
rates of both the liquid solvent and the gaseous sample are controlled by
use of peristaltic pumps. The temperature of the absorbing solvent is
controlled by use of a constant-temperature bath. The dual dynamic impinger
sampling system.uses two impingers in parallel and two multi-channel peri-
staltic pumps to control the liquid and gas sample flow rates. In the
dual dynamic impinger sampling system, duplicate samples are taken, and the
precision of the sample collection can be measured directly.
Many improvements in the design of the dynamic impinger sampling
system were made last year. The impinger size was changed, and the impinyer
liquid level control was added. Different packing materials were studied
for their sampling efficiencies. Teflon turnings which were 1/8" wide and
4/1000" thick were better than either glass beads or silica chips.
The dynamic impinger sampling system was modified for the long-term
operational stability which was needed in the field test. An additional
peristaltic pump was used to control the liquid levels in the impingers. A
check valve and a carbon absorber were put in between the impinger and the
gas sample pump to reduce the ripples in the gas sample flow rate and to
prevent the solvent vapors from reaching the pump tubing on the gas sample
pump.
A peristaltic pump would not control the flow rates between the two
channels better than 1 to 2%, and in long-term operations this difference in
the flow rates could cause either the liquid level to drop significantly,
258
-------
or to overflow the system, A second pump had to be added to this system to
control the liquid levels inside the impingers. The gas sample flow rate
change was caused by the changes in the liquid levels inside the impingers
that changed the pressure drop across the system, and the erosion of the
gum rubber pump tubing on the gas sample pump by the solvent vapors which
changed the pumping efficiency of the tubing. A check valve and a carbon
absorber were added between the impinger and the gas sample pump. This
reduced the ripples in the gas flow rate and prevented the solvent vapors
from reaching the gum rubber tubing.
Two field tests were conducted last year; the first one at a vinyl
chloride manomer plant for vinyl chloride, and the second one at a degreas-
ing plant for methyl chloroform. The vinyl chloride was collected by a dual
dynamic impinger sampling system with the impinger and the absorbing solvent
(iso-octane) set at -20°C. The gas flow rate was set at 7.5 ml/min, and
the liquid flow rate was at 1 ml/min. The sample solutions were sealed in
serum bottles and brought back to the laboratory for analyses. A gas
chromatograph equipped with an electron capture detector was used for the
analyses. The analytical results were compared to EPA reference Method 106
and the continuous emission monitor installed at the test site under another
EPA evaluation program. The analytical results revealed that the precision
of the dual impingers was about 15%, and the relative accuracy of the
method compared to Method 106 was about 20%. We note here that the detec-
tion limit of this system was estimated as approximately 1 ppm, and whenever
the vinyl chloride level dropped to below 2 ppm, the precision and accuracy
values became very poor. The major problem for the vinyl chloride sample
collection was the control of the sampling temperature at below -20°C
throughout the sampling time.
259
-------
The sampling of the methylchloroform at a degreasing plant was rela-
tively simple. The EPA reference Method 23 Tedlar bag sampling technique
was used to collect the sample, which was analyzed the following, day for
comparison. Both precision and accuracy of this species was about 15%.
260
-------
DEVELOPMENT OF A COMPACT TOTAL REDUCED SULFUR (TRS)
CONTINUOUS EMISSION MONITORING SYSTEM
Herbert R. Defriez, TRS Systems, Inc.
John W. Short, Harmon Engineering & Testing
Terence A. Whitt, Harmon Engineering & Testing
261
-------
For the past several years EPA regulations have required that the
kraft pulp industry install continuous emission monitors (OEMs) at all
new or newly-modified sources of total reduced sulfur (TRS). The
regulations will soon be expanded to include all older TRS sources as
well, greatly expanding the n'eed for TRS CEMs. ^The TRS 2000 is an
instrument of innovative design that provides reliable continuous
monitoring of TRS emissions in a compact stack-mountable package.
The TRS 2000 is a single on-stack unit small enough to be carried
by two men. Since all sampling and analytical systems are on the
stack, sample lines are very short. A moderate flow rate is used
along with in-stack sample dilution which results in a very low net
sampling rate and minimal loading of filters and scrubbers. The low
sampling rate and an aerodynamically designed probe tip reduce
particulate loading to such low levels that backflushes will be
infrequent and short? and filter replacement will be reduced to
quarterly or longer intervals. One of the design objectives is to
reduce all maintenance to such intervals so that all maintenance and
repairs may be done in infrequent visits by factory representatives
with no need for mill personnel to be involved.
The TRS 2000 uses a rapidly-reporting analytical system which,
along with the short sample lines and low overall system volume,
provides nearly instantaneous response to changes in the source
concentration. The low sampling rate allows for the use of a self-
contained permeation-type calibration system. Permeation devices
provide NBS traceable calibration standards, are stable for months,
262
-------
and are inexpensively replaced. All systems in the TRS 2000 are
operated automatically by an on-board computer. Data are transmitted
to a remote data terminal by FM radio, eliminating the need for data
cables.
Sample dilution is accomplished by the loop-dilution concept
developed by Defriez. Instead of requiring a separate source of
dilution gas, a portion of the sample flow leaving the analyzer is
returned to the probe as dilution air, see Figure 1. The total intake
flow at the probe is greater than the dilution flow and the difference
is made up by stack gases entering the probe tip. The valves and
orifice governing the flow rates are enclosed in a heated chamber for
stability. These are downstream from all analytical equipment,
including the dryer. It can be seen that for a dry sample such as
calibration gas the total intake flow and the orifice flow are equal
and that the net sample intake flow equals the dump flow. When the
system is changed to wet stack gas, the orifice, dump and dilution
flows do not change; the gas flowing through' the controlling orifice
and valves remains dry. However, the removal of moisture by the dryer
increases the total intake flow, thus increasing the net sample intake
flow and reducing the dilution ratio. This action causes an automatic
adjustment of the analyzer response so that all valves are given in
the dry-gas basis of the calibration gas. Because of the low total
system volume, response to changes in stack moisture is rapid. The
TRS 2000 uses an orifice flow of about 1/2 L/min and a dilution ration
of about 20-25 to 1.
263
-------
CL
O
O
264
-------
The analyzer system operates by removal of existing sulfur
dioxide, oxidation of TRS to S02, and measurement of the produced S02.
Oxidation is effected by a catalytic device which offers several
advantages over traditional quartz tube-furnace oxidizers: rugged
all-metal construction, lower operating temperature (325°C rather than
800°C), carbonyl sulfide (COS) not converted. Laboratory and field
tests have shown the catalytic oxidizer to be efficient and accurate
in the analysis of the four TRS compounds controlled in standards of
performance for kraft pulp mills. The S02 analyzer is a Monitor Labs
pulsed-fluorescence detector of proven accuracy, ruggedness and
dependability.
Calibration gases are introduced at the probe tip, thus
calibrating the entire system. The dryer is a permeation type and is
followed by a polishing scrubber that contains activated charcoal to
remove unoxidized organics that could otherwise build up in the loop
and possibly cause interferences. A fuel-cell type oxygen monitor is
placed in the dilution dump outlet.
Probes are custom made to suit: conditions in the stacks to be
monitored. The probe sheath is Hastaloy and interval parts are
Hastaloy or Teflon. Sample lines and system tubing are of Teflon or
stainless steel. Leak-free rotary valves are used for such functions
as backflush and calibration. Built-in safeguards prevent damage from
occurrences as intrusion of liquid coater or loss of control air. The
efficiency of the S02 scrubber is automatically checked by passing the
catalytic oxidizer. Data from the S02 and 02 analyzers are converted
to digital from by the control computer. These data are then
265
-------
transmitted over the FM data link with several error checking routines
such as redundant transmission and check bits; this ensures that
accurate data are received by the output terminal.
Field trials of the TRS 2000 have been conducted at a number of
Southern pulp mills. These trials have been conducted at recovery
boilers and at calcining facilities under a variety of conditions of
stack temperature and moisture, particulate loading, TRS concen-
tration, and weather. The system has proven to be reliable, and
comparisons of data with Method 16 data or with data from previously
installed, certified CEM's have shown that accurate results are
obtained and proven that the TRS 2000 is certifiable to the standards
of PS-5.
266
-------
MEASURING VOLATILIZATION FLUX OF PESTICIDES FROM TREATED SURFACES
Michael Majewski, James Woodrow, James Seiber and Paul Sanders
Department of Environmental Toxicology
University of California
Davis, CA 95616
We have been looking at the entry of pesticides in the atmosphere through
volatilization from agricultural surfaces. Among our objectives is comparison
of volatilization rates for chemicals with wide ranges of properties using the
EXAMS computer model, a laboratory chamber, field measurements and simple
estimation methods.
The pesticides studied vary considerably in water .solubility (S) and vapor
pressure (P) and, therefore, the Henry's law constant (H = P/S) (Table 1).
According to the H constant, mevinphos and ionized MCPA, with low values, would
be expected to show low rates, while compounds such as diazinon and molinate
with higher H values should show higher rates of volatilization from water.
TABLE 1. PHYSICAL PROPERTIES (at 22°C unless otherwise specified)
Compound
Mevi nphos
MCPA DMA
Thiobencarb
Parathion
Diazinon
Molinate
3 mg/L
MW
224
245
258
291
304
187
b xlQ-6
Water3
Sol ubi 1 i ty
mi sci bl e
>300,000(25°C)
30(20°C)
12
69
800(20°C)
mmHg c xlO
Vaporb
Pressure
2200.0
0.1(25°C)
1.5(15.3)d(20°C)
6.0
162.0
3110.0(25°C)
"b (m3.atm)/mole
Henry1 s Lawc
Constant
0.052
<0.00001(25°C)
1.7(17.0)d(20°C)
18.7
94.0
96 (20°C)
remeasured value
Predicting volatilization rates from moist soil requires another physical
property, the soil binding constant (Kd). By using a volatilization factor (f)
where f = p/(S»Kd), a group of compounds can be ranked with respect to their
volatilization from soil (Thomas, 1982).
The EXAMS computer model can calculate compartmental distribution, rate
constants for individual fate processes and system purification times for
contaminated bodies of water (Burns et al., 1981). In order to perform the
267
-------
calculation of interest, environmental conditions along with the compound's
physical and chemical properties are entered. Our choice for several of the
environmental parameters were based on the dimensions and microclimate of our
laboratory chamber.
The results of EXAMS for volatilization from water (given as flux with units
of ng/cm2-hr and normalized by the water concentration) show the same trend for
this series of compounds as the H values (Table 2). This was not an unexpected
result because EXAMS uses a two film model for volatilization for which H is a
primary determinant.
The experimental laboratory chamber was as described previously (Sanders and
Seiber, 1983). The measured chamber flux, normalized for concentration, agreed
quite well (within a factor of two or three) with the EXAMS predictions for most
of the chemicals studied (Table 2). Exceptions occur with every water soluble
compounds such as mevinphos. This is because the method of estimating H from S
and P does not hold for strictly hydrophyllic compounds. The EXAMS and chamber
value for thiobencarb also did not agree. In this case we feel that the EXAMS
value was in error due to an erroneous H value. We remeasured the P and found
that it was approximately 10X larger than the reported value Herbicide Handbook
(1983). Using the new P value in the EXAMS model the resulting flux agreed
fairly well with both the chamber and field data.
TABLE 2. FLUX FROM WATER
Compound
Mevi nphos
MCPA DMA
Thi obencarb
Parathion
Diazinon
Mol i nate
a m -atm/mole
Ha
0.052
l.lxlO5
17.0
18.7
94.0
96.0
EXAMSb
0.017
0.00(pH 7)
0.45(4.5)c
4.62
22.7
51.5
K ?
ng/cm «hr'ppm
Chamberb
0.15
O.OOtpH
13.4
12.6
28.6
62.8
Rice Field5
7) 2.3
13.1
—
—
77.0
c using remeasured 1
For field flux measurements, the aerodynamic method developed by Taylor and
coworkers (1976) was used. The method requires that the concentration of the
compound of interest be measured at two or more heights above the treated
268
-------
surface along with wind speed, direction, and temperature. In one experiment
with a rice paddy, we took simultaneous high volume air samples at two
heights. The sampling station was located at the downwind edge of the field on
a pier which extended into the water. The air samples as well as field water
samples were taken at regular intervals for several days.
The day 1 flux data for this experiment agreed quite well with chamber and
EXAMS values for rnolinate and thiobencarb (when the correct P value was used in
EXAMS). MCPA, however, showed an unexpected higher flux in the field than
predicted. This is probably due to the fact that MCPA was applied as an aqueous
spray 2-3 weeks after molinate and thiobencarb (which were applied in granular
form) when the rice foliage was well out of the water. Volatilization from this
added surface area as well as the application method could contribute to the
observed MCPA flux.
The other field experiment involved measuring pesticide flux from a fallow
field. Because EXAMS does not apply to this situation, we chose to compare the
measured field flux to values predicted from the Dow method. The Dow method
uses the S, P and KQC of a compound to determine half life in soil (Thomas,
1982). By assuming a first order process, a rate constant for volatilization
can be calculated. Also, knowing the application rate and assuming the loss is
due entirely to volatilization, a flux value can be estimated.
A plot of the average flux/hr as a function of time (Figure 1) shows that
the intial field-measured flux (0.18 ng/cm2vhr) was lower than predicted (0.761
ng/cm2«hr, not shown) and that the daily flux falls off more rapidly than is
predicted. There may be several explanations for these differences. The most
obvious is the assumption that the entire loss is due to volatilization. If
other dissipative process are occurring, measured flux must be lower than
predicted. Also, the Dow method assumes that volatilization is occurring only
from the surface and that no migratory processes are involved such as
leaching. This method also does not take into account the effect of changing
soil moisture content which occurs in the field.
Diazinon was applied in granular form and the field was then immediately
irrigated. By day 4 the field surface was completely dry. On day 8, the field
was again watered. Figure 1 shows that a dramatic increase in flux coincided
with this second watering. This was as expected from previous studies (Spencer
and Cliath, 1973).
269
-------
FIGURE 1.
DIAZIMON FLUX AS A FUNCTION OF TIME
0.2
2 4 6 8 10 12
time after application Cdays5
* ~ calculated flux
* = measured f 1 ux
In summary, the EXAMS computer model is adaptable to static water situations
as occur in rice fields, giving a rapid assessment of relative volatilization
rates. The EXAMS model can also be adapted to moist soil situations. The
chamber was useful as a tool for determining volatilization from water and soil,
yielding results in good agreement with both EXAMS and field data. Collecting
field data is very expensive, labor intensive and is often of low precision.
Improvements in sampling techniques and methodology are necessary before field
flux measurements can be used to accurately provide data for model validation,
although our preliminary results have been promising. Simple mathematical
estimation methods such as the Dow equation, can determine trends in
environmental behavior, but may be too oversimplified to adequately describe
field conditions.
REFERENCES
1) Burns, L.A., D.M. Cline, and R.R. Lassiter, 1981, EPA report, EPA-600/3.82-
023, U.S. Environmental Protection Agency, Environmental Research
Laboratory, Athens, GA.
2) Herbicide Handbook of the Weed Science Society of America, 5th Edition,
1983, published by Weed Science Society of America, Champaign, IL.
3) Sanders, P.F. and J.N. Seiber, 1983, Chemosphere, 12, 999.
270
-------
4) Spencer, W.F. and M.M. Cliath, 1973, J. Environ. Quality, 2, 284-289.
5) Taylor, A.W., D.E. Glotfelty, B.L. Glass, H.P. Freeman, W.M. Edwards, 1976,
J. Agric. Food Chem., 24, 625.
6) Thomas, R.G., in "Handbook of Chemical Property Estimation Methods", Lyman,
W.J. Reehl, W.F. Rosenblatt, D.H., Eds., McGraw-Hill, New York, 1982, pp 16-
25 to 16-27.
271
-------
FIGURE 1.
(3,2
DIAZIMON FLUX AS A FUNCTIOii OF TIME
—| , , • , f —,-^-Y-L , ^u-l ,
2 4 6 8 10 12
time after appl ication Cdays}
* = calculated flux
* == measured flux
In summary, the EXAMS computer model is adaptable to static water situations
as occur in rice fields, giving a rapid assessment of relative volatilization
rates. The EXAMS model can also be adapted to moist soil situations. The
chamber was useful as a tool for determining volatilization from water and soil,
yielding results in good agreement with both EXAMS and field data. Collecting
field data is very expensive, labor intensive and is often of low precision.
Improvements in sampling techniques and methodology are necessary before field
flux measurements can be used to accurately provide data for model validation,
although our preliminary results have been promising. Simple mathematical
estimation methods such as the Dow equation, can determine trends in
environmental behavior, but may be too oversimplified to adequately describe
field conditions.
REFERENCES
1) Burns, L.A., D.M. Cline, and R.R. Lassiter, 1981, EPA report, EPA-600/3.82-
023, U.S. Environmental Protection Agency, Environmental Research
Laboratory, Athens, GA.
2) Herbicide Handbook of the Weed Science Society of America, 5th Edition,
1983, published by Weed Science Society of America, Champaign, IL.
3) Sanders, P.F. and J.N. Seiber, 1983, Chemosphere, 12, 999.
*VS. GOVERNMENT PRINTING OFFICE: 1986-646-116-20743
272
------- | |