Atmospheric Research and Exposure Assessment Laboratory
° CHEMISTRY, METEOROLOGY. METHODS DEVELOPMENT. MODELING, MONITORING, PHYSICS, QUALITY ASSURANCE
MINIWORKSHOP ON ORGANIC AEROSOL
SEPTEMBER 6, 1990
Office of Modeling, Monitoring Systems, and Quality Assurance
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATMOSPHERIC RESEARCH AND
EXPOSURE ASSESSMENT LABORATORY
PEER REVIEW AND WORKSHOP MANAGEMENT SERVICES
Contract Number 68D80063
Project Officer
Ronald K. Patterson
Compiled by
Research and Evaluation Associates, Inc.
1030 15th Street, N.W., Suite 750
Washington, D.C. 20005
(202) 842-2200
100 Europa Drive, Suite 590
Chapel Hill, N.C. 27514
(919)968-4961
-------
DISCLAIMER
The use of trade names or commercial products in this document does not constitute
endorsement or recommendation for use.
-------
TABLE OF CONTENTS
SECTION PAGE
I. Executive Summary 1
A. Objectives of the Workshop 1
B. Summary of Presentations 1
1. Bruce Appel, State of California 1
2. James Huntzicker, Oregon Graduate Center 2
3. Peter McMurry, University of Minnesota 2
4. Robert Lewis, EPA/AREAL 2
5. Steve Japar, Ford Motor Company (FMC) 2
6. Peter Mueller, Electric Power Research Institute (EPRI) 3
C. Recommendations 3
II. Introduction 5
III. First Presentation: Organic Aerosols Overview, Bruce Appel 7
A. Presentation 7
B. Graphics 15
IV. Second Presentation: Sampling/Artifacts, James Huntzicker 27
A. Presentation 27
B. Graphics 33
V. Third Presentation: Smog Chamber Studies, Peter McMurry 51
A. Presentation 51
B. Graphics 61
VI. Fourth Presentation: AREAL/MRDD Activities, Robert Lewis 79
A. Presentation 79
B. Graphics 87
VII. Fifth Presentation: Organic Aerosol Research at
Ford Motor Company, Steven Japar 109
A. Presentation 109
B. Graphics 115
VIII. Sixth Presentation: Organic Aerosol Research Program at
EPRI, Peter Mueller 129
A. Presentation 129
B. Graphics 137
-------
TABLE OF CONTENTS (cont)
SECTION PAGE
IX. General Discussion 167
X. Recommendations for Future AREAL Research 183
A. Bruce Appel '. 183
B. James Huntzicker 185
C. Peter McMurry 187
D. Steve Japar 189
E. Peter Mueller 190
APPENDIX A - List of Participants 195
APPENDIX B - Agenda 199
-------
I. EXECUTIVE SUMMARY
A. Objectives of the Workshop
Ambient urban aerosol systems are a complex mixture of primary and secondary inorganic and
organic compounds with vapor pressures covering many orders of magnitude. Volatile organic
compounds (VOC) are known to play an important part in chemical reactions observed in the
atmosphere. The reactions of VOC in the atmosphere are important because (1) they are needed
for the generation of free radicals HO and HO2 and ozone and (2) they result in the transformation
of gas phase compounds to condensed organic aerosols. While organic aerosols are a major
component (with sutfates and nitrates) of secondary aerosols, little information is available on the
speciation or size distribution of organic aerosols. This paucity of data is due largely to two factors:
(1) the sheer complexity of the chemistry involved-VOC can lead to thousands of individual organic
aerosol species-and (2) quantitative, interference-free sampling of these species has always been
difficult due to the gas/paniculate partitioning phenomenon of organic aerosols.
On September 6, 1990, the Atmospheric Research and Exposure Assessment Laboratory
(AREAL) of the United States Environmental Protection Agency in Research Triangle Park, NC,
convened a miniworkshop to discuss the problems in sampling organic aerosols. Six national
experts were invited to discuss their most recent research and offer recommendations to the AREAL
as to fruitful and necessary areas of further study on this topic. This report contains a transcript of
that workshop.
B. Summary of Presentations
1. Bruce Appel, State of California
Dr. Appel discussed two general topics: (1) the problems of sampling paniculate organic
carbon and (2) the nature of the positive artifact problem in filter sampling for carbonaceous
material. He pointed out that he had worked extensively, yet unsuccessfully, in trying to develop
a true paniculate carbon sampler based on a denuder filter sorbing mechanism. He discussed
the recent work of Huntzicker/McDow and of Frtz using a denuder difference technique to study
the positive and negative artifacts involved with the gas/paniculate sampling of carbonaceous
materials.
-------
2. James Huntzicker, Oregon Graduate Center
Dr. Huntzicker discussed sampling artifacts for organic aerosols, specifically the work of
Steven McDow's Ph.D. dissertation. He concentrated on the problems of positive artifact from
vapor phase sampling on paniculate filters. He discussed the effect of filter face velocity and
its principal cause, the organic vapor adsorption.
3. Peter McMurry, University of Minnesota
Dr. McMurry spoke about measurements of size distribution made with microorifice
impactors. He emphasized the need to avoid using quartz filters due to artifacts discussed by
Appel and Huntzicker. He also discussed theoretical work on sampling efficiencies, specifically
evaporative losses. He pointed out that using a denuder to sample gas/particle organic carbon
mixtures may cause extensive evaporative losses from the filter due to shifting equilibrium. His
research efforts are directed toward new samplers that don't use quartz filters, thus avoiding
the denuder problems. He is investigating a diffusion sampler which uses laminar flow and
extracts samples out of the center of the gas stream. Separation is based on higher diffusivity
for gases.
4. Robert Lewis, EPA/AREAL
Dr. Lewis discussed the historical evolution of AREAL's program in the analytical
measurements of semivolatile organics. Collection techniques used involved Tenax cartridges
and stainless steel canisters. His group started with pesticides. He discussed the
evolution/design of the PS-1 puff sampler using polyurethane foam and the sampling of
polycyclic aromatic hydrocarbons (PAHs) using XAD-2. In recent years they have used
denuders to separate gas/particulates. Because of the low flow rate they cannot collect
short-term samples (24 hours or less).
5. Steven Japar, Ford Motor Company (FMC)
Dr. Japar discussed his ongoing organic aerosol research. Japar is interested in
determining the products and mechanisms for aerosol formation from gaseous reactions; his
laboratory is attempting to identify classes of reactions leading to organic aerosol formations.
The Analytical Sciences Department and Chemistry Department at FMC have several
-------
sophisticated analytical systems available for attacking this problem (e.g., American Petroleum
Institute triple quadruple mass spectrometer). Japar emphasized his approach of investigating
aerosol formation from simple, single-component systems. He feels that only through this
approach can any mechanistic information be obtained.
6. Peter Mueller, Electric Power Research Institute (EPRI)
Dr. Mueller discussed EPRI-funded work, both past and present, in the area of organic
aerosol sampling and analysis. The past work discussed included the design, fabrication, and
testing of MOUDI impactors; the size distribution of elemental and organic carbon in Western
U.S. rural sites; and the study of collection artifacts using denuder technology. Recent funding
involves research to characterize the nature of the particles (internally or externally mixed) and
work at California Tech to characterize primary organic emissions at a number of sources in the
Southern California Air Basin. Simultaneous research is being conducted on quantitative
evaluation of organics in ambient air. Receptor modeling would be used in an attempt to
account for the emissions.
C. Recommendations
The afternoon session was devoted to informal and formal presentations of recommended
research needs in the area of organic aerosol sampling and analysis. Speaker recommendations
are summarized below:
• Repeat the Huntzicker/McDow experiments relating organic aerosol sampling as a function
of face velocity/filter medium.
• Extend the work of Dennis Fitz using the denuder approach.
• Reexamine the denuder difference approach to sampling for specific organic compounds.
Serious questions exist as to the efficacy of this approach.
• Undertake laboratory studies to investigate adsorption phenomena on quartz fiber filters.
• Begin developing basic theory of the physical chemistry of gas/particle partitioning for
organics.
-------
Quantify aerosol volatilization problem in denuders for atmospheric organics.
Investigate alternatives to quartz fiber as filter medium for carbonaceous aerosols.
Investigate sampling/humidity interactions for organics.
Study absorption and extinction characteristics of ambient organic aerosols beginning with
simple (pure) systems.
Investigate aerosol formation potential of simple (single component) organic systems using
laboratory and smog chamber approaches.
-------
II. INTRODUCTION
WILLIAM WILSON: We've invited you here to talk with us about organic aerosols, and especially
how to sample organic aerosols. I'd like to compare this to the problem of collecting a snowflake
outside during the first winter snow, bringing it inside and aging it at 68 degrees and 50 percent
relative humidity for 24 hours and then showing it to your mother. What a lovely snowflake I found,
Mom! So, there are comparable difficulties in sampling organic aerosols. Now we've known this
for a long time.
I remember 20 years ago at Cal Tech when they were doing a study of Los Angeles smog and
they showed me that you can collect the aerosols from outside through the tube that went down
from the roof on a beta gauge and you can see the mass building up on the beta gauge, and then
if you switch from the outside air to the room air you can see this stuff evaporate, lots of organic
junk from the Los Angeles atmosphere. So we've known that was a problem for 20 years, it just
hasn't seemed necessary to deal with it.
We've also known from those early days at Cal Tech when Shel Friedlander put a second glass
fiber filter behind the first one and found a 5-10 percent correction for organic material that was
absorbed on glass fiber-we've also known there was a problem of adsorption of organic material
onto filtered material. So, we've known about that for a long time, but we haven't worried about it
too much. Although, in the last few years, more and more people have been giving consideration
to it.
I remember when I was trying to look at the composition of the organic material from a variety
of organic compounds, terpenes and aromatics, that I got all of these low molecular weight acids
and I couldn't quite understand why they should be in the condensed phase and I worried about
that a lot. And now I think maybe they were just adsorbed on the filter, although one isn't sure how
to do a phase diagram of organic vapor pressures when you have a hundred compounds mixing
in together. Certainly there would be some depression of the vapor pressure, but maybe not
enough to account for some of the things we find in organic aerosols.
In the last 5 years or so, when we've been trying to account for the visibility reduction both in
the East and the West, we've had to start worrying again about the organic aerosol and about the
artifacts, because depending on whether or not you have an artifact and whether it's positive or
negative, the organic component could be quite important in western visibility in the so-called clean
inter-mountain area, and this has become of concern to the Park Service and to EPA.
In the East, now that we are in the process of controlling SO2, with one goal being better
visibility, the power plant people are objecting to our blaming everything on sutfate and S02. So,
-------
it becomes more important to be quite definitive and correct about the contribution of organic
aerosol to visibility reduction in the East. So, this has developed some more interest due to visibility.
We also recognize whether a pollutant goes into your lung as a gas or a particle has a great
deal of influence on how much of it is retained in your lungs and what part of your lung it is retained
in and quite possibly the effect, since the concentrated pollutant in a particle might be more
damaging than the widely dispersed field from a gas. So, from the standpoint of health effects it
is important to know whether we're dealing with paniculate pollutants or gas phase pollutants.
With that introduction to the importance of what we want to do-but, maybe I should say another
word about what we want to do. As a result of this workshop, EPA hopes to plan its modest
program of research and development looking toward a better sampling technique for organic
aerosols, one that is either artifact-free or one where we can define the extent of either positive or
negative artifact formation. So, with that, I'll ask Bruce Appel to tell us something about what we've
learned in the last 20 years.
-------
III. RRST PRESENTATION: ORGANIC AEROSOLS OVERVIEW, BRUCE APPEL
A. Presentation
BRUCE APPEL: Regarding sampling of paniculate carbonaceous materials, I'm going to focus
on two general areas. One is the problem of sampling paniculate organic carbon and the second,
the problem of the chemical nature of the materials in vapor phase that contribute to the so-called
positive artifact in filter sampling for carbonaceous material-the positive artifact due either to
sorption on the filter medium and/or the particles that are collected on it.
I want to begin with what I consider to be, well, first of ail-neither of the topics I just
mentioned-but the real success story in relation to particle carbon/paniculate carbon sampling has
been the sampling of specific carbonaceous materials. And the reason why these have been the
success story, in my judgment, is because of the high degree of specificity that you're capable of
for the analysis, in contrast to the very crude measurement of carbon-either total carbon or total
organic carbon. We're dealing then with the three techniques I've listed here. (Figure 1) First, a
simple filter backed with a sorbent that's been used to provide a relatively accurate measure of the
total particle plus vapor phase concentration of specific species listed here, and some crude
measure of the particle/vapor phase ratio. In some instances this may be very inaccurate and in
some it may be relatively accurate.
But moving from here to a much more accurate technique, at least potentially, the denuder
difference methodology, adapted from its successful application in the inorganic realm for nitric acid
paniculate nitrate, but now translated into the organic realm with the same methodology. And for
this audience I won't go through the logic behind the denuder difference method. Suffice it to say
here that a technique relying on a denuder-filter-sorbent on one side, and the identical system
lacking a denuder on the other, is an approach to a measure of the true particle phase
concentration for a specific analyte and a measure simultaneously of the vapor concentration of that
material, and a measure of the degree of negative artifact. Which is to say, the degree of
volatilization from the particle phase subsequent to collection, but prior to removal of the filter from
the sampler. And this has been applied rather successfully by Bob Coutant at Battelle to the
measurement of specific PAH compounds. And lastly, annular denuders-and, of course, these can
be backed up with an appropriate filter pack to provide exactly the same information. The
advantage here is that the measurement of vapor phase materials is not a difference measurement,
which has implications with respect to the potential precision with which those measurements can
be made and this has been successfully applied to nicotine. Now, I consider this to be a goal
towards which we can aspire with a more difficult realm of paniculate organic carbon sampling.
-------
Now, let's look at the state of affairs with respect to paniculate organic carbon, and now things
become much more crude. (Figure 2) We have for example, a dual quartz filter approach, where
one measures this relatively nonvolatile carbon with some attempt to correct for the positive artifact.
But one does not address the positive artifact due to sorption on any particles, only on the filter
medium itself; one does not say anything at all about the degree of volatilization (the so-called
negative artifact). A second strategy, very similar in concept, relies on sampling with a parallel
Teflon followed by quartz filter, pioneered by Jim Huntzicker. I'll have more to say about these
strategies shortly. And here, what I consider to be, and what I'll refer to as the Holy Grail with
respect to paniculate carbon sampler, towards which I worked hard unsuccessfully. And that was
to develop a true paniculate carbon sampler relying on a denuder-fitter-sorbent mechanism which
would allow an accurate measurement of the paniculate carbon as well as the negative artifact that
exists for paniculate carbon. And lastly, I'll have more to say about Dennis Fitz's very interesting
recent report using a diffusion denuder followed by a dual quartz filter to measure relatively
nonvolatile carbon, relatively free of the positive artifact. Again, this does not address the issue of
the negative artifact. So, let's go back now and look at some of these in more detail.
Let's begin with the simple dual filter methodology. (Figure 3) Now the simplest strategy is
simply to sample with two filters in tandem, and the underlying rationale is that vapor phase
carbonaceous materials are retained with low efficiency (perhaps a percent or two at the most), and
if that is true the concentration of those materials is about the same at both filters. Which means
that the amount that would be retained (micrograms) would be about the same here and here. And,
therefore, if one measures the amount of carbon collected on a backup quartz filter in this strategy,
I'll call 'Q minus Q,1 this will provide one correction for that positive artifact due to sorption on the
filter medium. It says nothing about sorption on particles. The alternative, which I'll refer to as the
•Q minus TO* strategy, is to sample in parallel with the second system and use the quartz filter
carbon from the second sampler to subtract from this one as a measure of corrected paniculate
carbon. The rationale here is that the Teflon filter would retain lesser amounts of vapor phase
carbon, as compared to quartz, and thereby provide a better measure of gas phase organic
concentrations.
WILLIAM WILSON: [But the 'Q-Q" strategy doesn't work with strongly sorbed organics, does
it?]
That's exactly the point and I was going to dwell on that a little bit later, but since you've
introduced the topic. The rationale for this approach is~let's imagine that there is a strongly bound
carbon that contributes significantly to the apparent paniculate carbon. If that were true, then this
Teflon material presumably would retain less of it and therefore this value (the backup filter following
Teflon) would be higher in carbon in comparison with this (the backup filter following quartz). I'll
8
-------
examine that particular concept shortly to the extent that any data are available that address it (in
fact there are).
Now the rationale for what I call the "Q minus TQ* approach (Figure 4). From recent work of
McDow and Huntzicker, they did a very interesting experiment in which they looked at the two
strategies as a function of face velocity, in which face velocity change was created by successively
masking a larger and larger fraction of a filter. And to vary face velocity between about 15 and 140.
Now the rationale was this: that the correction technique which more effectively removed the face
velocity dependence was the more accurate procedure-the favored procedure to be used. And it
is quite clear from this that the 'Q minus TQ" strategy removed all but about 10 percent of the face
velocity dependence change, for the measurement in micrograms per cubic meter of organic
carbon. Whereas the *Q minus Q* strategy in this experiment will remove only about 50 percent of
that dependence. All right, well that's one of the defenses for the "Q minus TQ.'
[But where was that sample taken?]
That was taken in Portland, Oregon.
Let's consider the 'Q minus Q" strategy and what support there might be for that. (Figure 5)
If one samples with three quartz filters in tandem, and given all that I have said, one would expect
the amount of carbon here and here (Filters D and E) to be virtually identical, given a low
percentage removal by any given filter. In this particular experiment-this is typical data from a paper
of ours--we look at the ratio, carbon here to the carbon here, at two face velocities, and the mean
result for that ratio carbon E over carbon D, 0.88+0.17. The mean decreased at low face velocity.
This is a moderate face velocity. For calibration, a Hi-Vol has a face velocity close to 50 cm/s. This
is like a 47-mm filter at 20 L/min. So, one can conclude that certainly there is a substantial variance
associated with the correction "Q minus Q', and that one achieves, at moderate face velocities at
least, something approaching a 90 percent effectiveness in correction, assuming that this experiment
is meaningful. But that is a point that I have to continue to defend, and I will do so.
Now, (Figure 6) we made a direct comparison of the two correction strategies (and some of you
have seen this before), and this experiment began (which may be its fatal flaw) with the removal of
paniculate matter with a quartz prefilter. And then we have here a "QQ Train" and a Teflon-quartz
train (all sampling in parallel). The question asked is: Which provides the better measure of the
carbon here (Filter B), this or this (C or D)? And we looked at that question as a function of sample
volume, varied sampled volume, between 2 and a half and 25 cubic meters. And I show here the
carbon here and here (Filters C and D) expressed relative to the carbon on this filter (Filter B). And
this is the ratio for carbon C, following Teflon. And what these data suggest is that the correction
strategy over-corrects at low sampling volumes. At higher sampling volumes, there is effectively no
difference between the two strategies. The filter D/B ratio ranges between .86 and 1.1.
-------
Now, if in fact there was strongly bound vapor phase organics that could contribute to the
positive artifact, you can argue that: "Well, look they've all been removed here (the prefilter), and
so all you're doing is making a comparison between weakly bound things, so it's not really a valid
experiment, valid to the ambient air.* But, now let's consider some recent work by Dennis Fitz.
(Figure 7) Now, Dennis did what I think was a very nice piece of work. What we're going to do here
is sample in parallel with 2 systems, one of which has a diffusion denuder ahead of a filter pack:
quartz-quartz and quartz-quartz (both systems). Systems are identical except for the presence of
a denuder material. This is an empty housing. Now, what I show here is a measure of organic
carbon on the front filter and the back filter, elemental carbon on the front and back filter.
Now what I had intended in preparing this transparency was that this column would represent
the results here and unfortunately I 'goofed' and we're going to have to interchange columns. This
column of numbers on the left relates to what I'll refer to as the 'undenuded carbon sampler.' The
'denuded carbon samplerMhese are the results in the right column. First, I call your attention to
the elemental carbon values on the front filters. They are virtually identical. These all represent
mean results for 18 12-hour periods, sampling at Glendora, southern California, micrograms per
cubic meter, as part of the carbon intermethod comparison that took place at that time. With
regards to the elemental carbon and why it is important that these values are about the same-that
supports the idea that fine particles are not being significantly removed in passage through the
denuder, an important point.
On average, the undenuded sampler on the front fitter, as well as on the backup filter, had levels
of carbon, organic carbon, about 2 micrograms per cubic meter higher than those on the denuded
side. This (undenuded, front filter) represents the contribution both from particles, as well as
sorption on the filter medium, or the particles thereon. From this number right here (backup filter,
denuded side), one can infer that the denuder had an average efficiency of 78 percent for the
removal of vapor phase organics able to be retained by sorption on the filter medium. Now, Dennis
concluded from this, his conclusion was very interesting, he concluded that the species responsible
for sorption on filter media are initially gas phase materials, as opposed to materials that are initially
particle phase, volatilized, and readsorbed on the backup filter. Because if that were the case the
denuder would have no effect or only a secondary effect.
Now there's something else we can learn from this. Let's correct these results with a *Q minus
Q' strategy which I have done down here, and the results are virtually identical for the two samplers.
Now let's go back to this idea about the strongly-bound (the hypothetical) strongly-bound
carbonaceous material. If, indeed, such material exists, it would surely be removed by this diffusion
denuder, quite efficiently. Now, that would mean that the carbon collected on the front filter (on the
undenuded side) would have contribution then from the strongly-bound as well as the weakly
10
-------
absorbed materials. And it would follow then that the *Q minus Q" strategy on the undenuded side
would not be as accurate (it would under-correct), in comparison with the denuded side, which
should show relatively accurate correction by the 'Q minus Q' strategy, because the strongly-bound
material has already been removed. But the results are virtually identical and this would argue that
the hypothetical strongly-bound material did not exist during the time that this sampling was carried
out. So that's about the best defense that I can offer for the 'Q Minus Q' strategy, except some
practical considerations-one sampler, rather than two, and no longer having the need for very
accurate measurement of sample volumes to achieve that accurate correction.
Now, I want to switch gears and talk about a totally different topic, and that is the chemical
composition of the vapor phase materials, which contribute to the positive artifact, and I'll show you
first some old data (Figure 8) From a study of ours going back to 1975, carried out at three
locations in the Southcoast Air Basin in California. And what we were doing (the experiment) was
rather simple. We had two Hi-vol samplers, side by side, three locations. One sampler operated
for 2-hour periods (filter change every 2 hours for seven 2-hour periods). The second sampler
operated continuously for 14 hours. We then compared a calculated 14-hour value from seven
2-hour samples with a single 14-hour sample. These numbers reflect the ratio 'calculated/observed.'
We did chemical characterization by means of selective solvent extraction using solvents of
increasing polarity, (nonpolar carbon means that carbon which was extracted in cyclohexane; polar
carbon refers to that which is solubilized in methanol-chloroform). I want to call your attention to
the dramatic difference between nonpolar and polar carbon and then talk about my inference from
these data. There are at least 2 possible interpretations of these results, and I'll talk about the one
that I favor now. And, in fact in relation to nonpolar carbon, it's the conclusion I had in 1975, as
well. My interpretation of this is that the long-term sample favors the retention of nonpolar carbon.
And I infer from that, that the paniculate matter on that single 14-hour sample is sorbing nonpolar
carbon. By contrast, polar carbon, exactly the reverse, a ratio of 1.7, argues that the retention of
relatively polar organics contributes to the positive artifact due to sorption on the filter medium. Now
that doesn't seem so unreasonable when you can consider the surface of glass (by the way this
work was done with glass not quartz), covered with nice polar hydroxyl groups, it doesn't seem
unreasonable at all. And in fact Kochy Fung reported at an EPA APCA conference two years ago,
he looked with GCMS at the chemical composition of sorbed organics. And these (Figure 9) are
sorted on quartz fiber, they were initially vapor-phased so they pass through filter to reach a backup
filter, and they were polar organics that he could identify. The major species that he saw were these
(I've chosen 4 from a list of perhaps 25 compounds): tributyl phosphate, phthalate esters,
benzaldehyde, and nitrogen-containing heterocyclics. So certainly Kochy's results are consistent
with what I've been telling you.
11
-------
Now, our own work looked in a controlled environment at the sorption of nonpolar and polar
organics, nonpolar octadecane and phenanthrene, polar benzoic acid and dibutylphthalate sorbing
on quartz and glass fiber filters. And I show here (Figure 10), retention of these species in percent
on quartz and glass. We could not measure retention of nonpolar compounds on quartz and glass.
There was measurable retention of the polar materials. Now the high number 7.7, well that's hardly
fair, that's an acid-base reaction, because the glass fiber is quite alkaline. We discount this result,
although it certainly would be meaningful in the ambient air sense that William Wilson referred to-the
presence of low molecular weight dicarboxylic acids and, perhaps, monocarboxylic acids, as
well-that were observed in the particle state (i.e., on alkaline filters) when they had no reason to be
so, given their vapor pressures. And, this would argue that it is an acid-base reaction, and certainly
not too surprising. And so, polar organics sorbed on both pH-neutral and alkaline filter media.
Now for the future, given my 'druthers' of what I'd like to see: I'd like to see Dennis Fitz's
approach as a starting point to what I referred to previously as the Holy Grail, and that is the true
paniculate carbon sampler. He's part of the way along. He's gotten 78 percent efficiency for
reducing the positive artifact due to sorption on the filter medium. The denuder (in many ways) is
the hardest part. It would need to be augmented, if I am correct, with material which would remove
nonpolar materials that would otherwise be retained on the paniculate matter. And it would also
need to be backed with a sorbent. When we're dealing with carbon, that sorbent must be virtually
carbon free. We did work with a fluidized bed of aluminum oxide as that sorbent, because only with
such a sorbent able to collect the material which is volatilized from the filter can one hope to obtain
a direct measure of the negative artifact, if indeed, the negative artifact really exists.
Next point, I'd like to see a repeat of the McDow and Huntzicker experiment, where one
compared the two correction strategies as a function of face velocity, but I'd like to see that
repetition done in the way that people usually vary face velocity, which is to say, keep the filter area
constant and change the sampling volume. Change the sampling rate, which affects the total
volume, keeping total sampling time constant. I think that would be an interesting and worthwhile
experiment to do and might hope to reconcile some of the apparent differences that I have alluded
to here today.
And also in referring to differences, I believe that there is a discrepancy in conclusions about
whether in fact a negative artifact really exists. On the one hand, look at the results of Bob Coutant
using the denuder difference method-if his results, in fact, can be believed, and there are some
questions about his experimental approach. But interpreting his results literally using the denuder
difference for PAH compounds, what he found for pyrene, for example. At the moment of passage
through his denuder, he concluded based on experimental results, that 57 percent of pyrene was
in the particle phase at the time it passed through that denuder. Otherwise it wouldn't have passed
12
-------
through. But of that 57 percent that was in the particle phase, 82 percent volatilized subsequent
to collection on the quartz filter, but before removal from the sampler at the end of 24 hours. Now
that's a massive degree of volatilization loss, and you'd think that if volatilization is really so
important for at least this minor constituent in atmospheric paniculate matter, it ought to be fairly
easy to demonstrate even on a carbon basis. And yet it's very tough. A number of us have tried
this experiment, collecting, for example, overnight, paniculate matter from short-term sampling, then
passing carbon-free air through it to see if you can measure a decrease in carbon, and it's
negligible. In fact, if I interpret McDow and Huntzicker correctly, they seem to argue that the
negative artifact, in fact, may not exist. The face velocity dependent decrease in carbon, retained
on a filter with increasing face velocity, they argue, is due to diminished sorption rather than
increased volatilization. So I see an inherent conflict between, on the one hand, the results with
individual PAH compounds (and there have been other examples of this) and, on the other hand,
the results with paniculate carbon, where it becomes very difficult to demonstrate that, in fact, a
negative artifact really exists. And I consider this one of the challenges for the future. I think this
pretty well summarizes what I wanted to say.
[Thank you, Bruce, anyone have any questions? You can give us any recommendations this
afternoon about how to proceed?]
I can try.
[Good.]
[I think Bruce has pointed out a lot of stuff for discussion, but the rest of the speakers might
cover it].
[O.K. Jim Huntzicker?]
13
-------
B. Graphics
15
-------
SPECIFIC CAUON COMPOUNDS
Technioue
Measured Parameters
Application
Filtar-ftorbent
total cone., approx.
partic., vapor cone.
PAH, pesticides,
nicotine
Denuder Difference
approach true partic. &
vapor cone., measure neg.
artifact.
PAH
Annular Denude r
came
nicotine
FIGURE 1.
17
-------
FAKXICUIATC OBCAHIC CAJLBON
Technlnue
Dual quartz filcer
fcel. non-volatile C (corrected for + artifact)
Quartz v/ parallel
Teflon-quartz filter
Deaudtr- f i lter-c«rb*nt
farticulate C and neg. artifact (unsuccessful)
fienuder-Dual quartz
filter
Rel. non-volatile C, rel. free of •*• artifact
FIGURE 2.
18
-------
Quartz C
Tcflc*
Co vacuum
Quartz
Quartz
FIGURE 3.
19
-------
(ugC/m3)
2
20 40 60 80 100 120
Foce Velocity (cm/s)
_L
l
140 160
S. R. Nctev «*d J. J. Hunczicker, At«o«. Environ.
. 2563-2571 (1990)
FIGURE 4.
20
-------
•
Quarts
Face Velocity (cm/sec)
26
6.4
Quartz
Quarts
E/Carbon D
0.8810.17
0.78±0.17
Ch*ng and Salay»*h. Atao*. Environ. 21. 2147-2175 (1989)
FIGURE 5.
21
-------
to vacuum
Quartz C I
Teflon A [
10.2cm prefilter—
Vol 3
2.4
7.2
19.8
25.2
ambient air
IH C (rel. co Filter B)
Filter C Filter D
1.8 0.95
1.4 1.1
1.0 0.86
1.2 0.94
Quartz
Afp«l. Chen* «nd StUyaeh. Atmos. Environ. 21. 2167-2175 (1989)
FIGURE 6.
22
-------
Needle Valve
Quartz Rhw Slips .
(15 at 3mm spacing)
Air
Inlet
£\
\ /Cyclone
To Vacuum
ftouwnoier
Aluminum
' Oenuflcf"
Housing
Quartz
Front
Ftltef '
-Back.
Filter
Needle Va>v«
••^p
To Vacuum
Rotanwtar
Mean results
Needle Valve
r periods (ClenUora, CA)
OC Frooc
EC Frooc
OC lack
EC fcack.
16.6
1.53
2.73
0.01
14.6
1.51
0.60
0.01
Q • Q:
13.9
14.0
D. R. Fitz, Aerosol Sci. and Technol. 12, 142-148 <1990)
FIGURE 7.
23
-------
SOLVENT EXTRACTION RESULTS
1975 SCAB STUDY
Ratio CaJcd/Otw. 14-hr Results
Ct Non-Polar C Polar C Ce
1.21 0.74 1.70 1.23
FIGURE 8.
24
-------
CCKPOSITON OF FXLTOL-SOUED ATMOSPHERIC ORCAJJICS
Major identified species include: tribucyl phosphate
phthal*te esters
benzaldehyde
N-he terocyclies
K. Fung, 19*1 E?A/APCA International Synposiua: Heasureuent of
Toxic Organic* and Related Air Pollutants
FIGURE 9.
25
-------
JtETEXTlON OF VATOH-HASE COMPOUNDS ON QUARTZ AND CLASS FIBER FILTERS
Compounds Cone . (af/m*) Dosage (uf) Rt-tenrion (*)
Ou. in 7. Cl.-iss
Occadecane SO 280 0 0
Phenanthrene «sc. 50 120 K.D. <1
Benzole acid 265 1600 1.4 7.7
Dibucylphchalate SO 60 N.O. 1.2
N.D. - «et determined.
FIGURE 10.
26
-------
IV. SECOND PRESENTATION: SAMPLING/ARTIFACTS, JAMES HUNTZICKER
A. Presentation
JAMES HUNTZICKER: This morning I want to talk about our work on artifacts in the sampling
of organic aerosol. Most of what I will be talking about is the Ph.D. thesis work of Dr. Steve McDow,
who is here. It also reflects some of the work of a subsequent student, Dr. Barbara Turpin.
PETER MUELLER: Before you get started, could I make one philosophical comment here?
Sometimes this positive/negative artifact business has really bothered me a lot. What I would like
to suggest for us here, since we are starting out fresh, is that, what we're really trying to get at, is
paniculate carbon. What we're concerned about is altering paniculate carbon, somehow, in the
sampling process, which can result in over-sampling or under-sampling. I'm wondering whether we
can use the over/under sampling concept with regard to what it is we're really trying to get. That
would make life a lot easier for the people to whom we otherwise have to explain positive and
negative artifacts. Well, I didn't want to interrupt Bruce, but I wanted to say it just before you started.
BRUCE APPEL: I'm grateful you didn't interrupt me.
JAMES HUNTZICKER: Let me stan with my conclusions. The first set of conclusions (Figure
1) has two aspects. First, the apparent concentration of aerosol organic carbon as collected and
measured on a quartz fiber filter is a function of filter face velocity. Face velocity is the volumetric
flow rate divided by the exposed area of the filter. Note that I will refer to aerosol organic carbon
rather than organic aerosol, because what we measure is carbon and we separate it into organic
carbon and elemental carbon. The second aspect is that the principal cause of the face velocity
effect is the adsorption of organic vapors or gases on the filter medium.
The second set of conclusions (Figure 2) has three aspects. First, the adsorption of organic
vapors is a significant and, in fact, sometimes large positive artifact. In other words, it
over-measures the amount of carbon that you would otherwise associate with the paniculate phase.
Second, the amount of adsorbed organic carbon can be estimated by measuring the amount of
organic carbon collected on a quartz fiber filter behind a Teflon front filter (this is the TO* approach
that Bruce Appel talked about in his previous talk). It is important that both sets of filters in this
parallel sampling must be operated at the same face velocity. Finally, at organic carbon
concentrations of a few micrograms per cubic meter, the correction can be as high as 50 percent,
and I will show you results to support that
The experiment that Steve McDow and I did involved the simultaneous collection of six samples
through a common manifold (Figure 3). Each of these samplers operated at the same volumetric
flow rate, but at a different face velocity. In some experiments, each of these sampling ports had
27
-------
a quartz fiber front filter followed by a quartz fiber back filter. In some experiments, 3 of them had
a Teflon front filter followed by 2 quartz fiber backup filters, and the other 3 had quartz-quartz
combinations. The variable face velocity was obtained by masking down the area of the filters with
a top mask and a bottom mask (Figure 4), and considerable effort was expended in making sure
that all this was leak-tight. Note that back-to-back filters were in a single filter holder; they weren't
sequential filter holders. Typically, most of our samples were collected for 24 hours in downtown
Portland, Oregon, at a rooftop elevation, three floors up.
[Was it summer or winter?]
Throughout the year. The samples were taken out of the sampler, returned to the lab, and
measured in our carbon analyzer.
The carbon analyzer is a thermal-optical instrument (Figure 5) that simultaneously measures
evolved carbon with a flame ionization detector and reflectance on the filter to separate organic and
elemental carbon. Organic carbon is measured in a helium atmosphere, free of oxygen. As the
temperature is raised, the filter becomes blacker because of pyrolysis. When oxygen is added to
the helium, the elemental carbon oxidizes and the filter reflectance begins to increase. We defined
the split between organic and elemental carbon as the point when the filter reflectance returns to
its initial value. Two calibrations are performed by injecting methane and measuring the instrument
response. One calibration is done in helium and the other in helium-oxygen. A typical output is
shown in Figure 6.
Several laboratory experiments on the face velocity effect were done before we ventured into
the field. Figure 7 shows the results from an experiment in which we set up a combustion process
and sampled the effluent from that. First of all, I'd like to point out an important quality control
criterion here. Elemental carbon, being nonvolatile, should be constant as a function of face
velocity. In fact, the elemental carbon concentration is constant as a function of face velocity, but
there is a very pronounced decrease in the carbon mass collected for organic carbon with
increasing face velocity. Obviously, this is something that is very undesirable. One does not want
to have the concentration of something you're measuring depending upon a variable such as face
velocity or filter area or volumetric flow or whatever.
We then did quite a number of experiments in the field. Figure 8a is for a situation in which
there were relatively high concentrations of aerosol carbon. (TC1 is the sum of organic plus
elemental.) At relatively high concentrations, only a (relative) modest decline in organic carbon is
observed with elemental carbon staying essentially constant. At low concentrations (Figure 8b),
however, a much more pronounced relative fall-off of organic carbon with face velocity is observed.
We saw that every time we sampled in Portland. We also saw it when we went to Southern
28
-------
California and sampled in the Carbon Species Methods Compan'son Study of 1986 and again in the
SCAQS 1987.
The face velocity dependence on the backup filters for the same experiments is shown in Figure
9 in which the results are expressed as volumetric concentrations. Again, there is a very
pronounced face velocity dependence. We know that adsorption has to be an important mechanism
here. Otherwise, how else would carbon get on the backup filter? We know from seeing no
elemental carbon on the backup filters that this has to be adsorbed organic carbon.
Before proceeding, I would like to discuss in general terms (Figure 10) both adsorption on
particles in the atmosphere and adsorption on quartz fiber filters. First of all, the process is
intrinsically multicomponent. That means that there are hundreds, if not thousands, of compounds
that can potentially adsorb on atmospheric particles, on filters, but the most important adsorbing
species is water. There are data in the literature, which report adsorption of water vapor on quartz.
At about 40-50 percent relative humidity there's enough adsorbed water vapor to constitute a
mono-layer of water on quartz, and quartz should be a reasonable analog for inorganic particles in
general. Thus, in any adsorption situation water is a important player. Another important thing that
we must consider if we look at the theories for multi-component adsorption is that the amount of
adsorbed water on the filter or on the particles, regulates the amount of adsorbed organics.
Therefore, relative humidity, which is the determinant for the amount of water adsorbed on the filter
medium or on particles, is an important variable in determining how much organics can also be
adsorbed. One can also use these theories to look at the amount of adsorbed organics as a
function of relative humidity. The results are somewhat surprising, depending upon whether the
organics are miscible or immiscible with water. Essentially the same mechanism operates for the
adsorption of organics on micron- or submicron-sized fibers in filters as on submicron particles.
In my opinion, sorption is the principal mechanism by which organics are incorporated into the
aerosol, at least into the fine aerosol. Now we've talked in the literature about sorption, adsorption,
volatilization, but I want to make the point here that these are all the same fundamental
phenomenon-sorption. Moreover, what has been called blow-off, or volatilization, is simply
desorption. Note also that there is no saturation value for adsorption. There is an equilibrium value
which depends upon the vapor phase concentration of the material that is being absorbed.
Adsorption of organics is intrinsically a two-phase problem, although for some organics with very,
very low vapor pressures it can collapse into a one-phase problem.
The final point in Figure 10 concerns sorption by collected particles on the fitter. Presumably
the particles that are in the aerosol phase-i.e., suspended in the atmosphere, are at equilibrium or
close to equilibrium with the gas phase. Therefore, there is no reason why collection of these
particles on the filter should cause any more organics to be adsorbed, unless the concentration of
29
-------
the gas phase increases. Moreover, in terms of contribution to artifact, it can be shown that the
surface area of collected particles on a fibrous filter, under most conditions, is a relatively small
fraction of the surface area of the filter itself. Thus, even if the particles were adsorbing from the
atmosphere you would expect the effect to be relatively minor.
[PETER MCMURRY: I'm just bothered by your statement that adsorption is the principal
mechanism by which your organics are incorporated into the aerosols. I don't think that we know
that it's true for the bulk of organic aerosol. I think we'd agree that probably there are organic
species that are adsorbed on the particles. For example, if you were to do an experiment within a
smog chamber and react an aromatic with NOX and sunlight, you'd produce aerosol organics.
Those particles are condensed, not sorbed.]
I certainly was aware of that, Pete. I think, physically, there's not a real difference. They
condensed onto something, and the distribution between the gas and the particle phase is a
function of the surface area and the vapor phase concentration.
[PETER MCMURRY: I think the difference is how you describe the equilibrium. In one case,
using the adsorption isotherm to characterize the equilibrium.]
In the real atmosphere, there aren't pure particles, pure drops-that would be the unusual
situation. We always have a very complex type of particle. That's my hypothesis.
[PETER MCMURRY: O.K., I just think that it needs to be qualified.]
Figure 11 shows the results of another experiment. In this figure, QQ(1) refers to measurements
on the first filter in a quartz-quartz combination; TQQ(2) is for the second filter in a Teflon
quartz-quartz combination--!.e., it's the quartz fiber backup filter behind the Teflon front filter. QQ(2)
is the second filter in the quartz-quartz combination. TQQ(3) is the third filter in this
combination-i.e., it's the second quartz backup filter behind the Teflon front filter.
Figure 11 a shows that there is a significant face velocity dependence for QQ(1). The question
is what to do about it. One possibility for explaining the face velocity dependence might be that the
pressure drop in the filter increases with increasing face velocity, causing desorption or volatilization.
Based upon our understanding of theory of adsorption and measured pressure drops, we can
estimate that this is a relatively small effect compared to the amounts of decrease observed. As
explained in our paper in Atmospheric Environment 24A. 2563-2571 (1990), we can say that
volatilization or desorption cannot be the principal thing happening. Further evidence for this
conclusion can be obtained from Figure 11b. If the organic carbon on the backup filter was
absorbed material that had volatilized off the front filter, we would expect that adding the front filter
and the backup fitter results together would eliminate the face velocity dependence for the sum.
However, Figure 11b clearly shows that whether the backup filter behind a Teflon filter or a backup
filter behind a quartz filter is used, there is still a very pronounced face velocity dependence.
30
-------
Thus, the hypothesis that if adsorption of vapor phase material on the front quartz fiber filter is
the dominant process, then the backup filter might be a reasonable measure of the amount of
adsorbed organic carbon on the front filter. But which is the more appropriate measure-the backup
filter behind the quartz front filter, or the backup filter behind the Teflon front filter? If we accept
elimination of the face velocity dependence as a criterion for success here, it is clear that the TQ
combination is the preferred alternative, because this correction procedure removes virtually all of
the face velocity dependence, as shown in Figure 11 c. Note that at the very least, one criterion we
would like to see in an accurate sampling method is that there is no dependence on a parameter
such as face velocity.
Why is the concentration on the backup filter behind the Teflon front filter so much greater than
the backup filter behind quartz front filter? First, we do not think that the process has reached
equilibrium under the conditions of our experiment. Because the Teflon filter has only about
one-fourth of the specific surface area of the quartz fiber filter, what appears to be happening is that
the first quartz fiber filter in a series depletes the sample air of adsorbing organics. The process is
similar to what happens in frontal chromatography, where break-through does not occur until
equilibrium has been achieved between the gas and solid phases (Figure 12).
It is unlikely that outgassing of contaminants from the Teflon front filter is a source of adsorbable
organics on downstream quartz fiber filters. This is supported by the results of Figure 13, which
show very little difference between OC on TQQ(3) and QQ(2). Additionally, carbon analysis of Teflon
Zefluor filters at 275 °C yields only a very small amount of volatilizable carbon-certainty not enough
to account for our observations.
In conclusion then, we propose that a good measure of the adsorption artifact can be obtained
by a parallel sampling procedure involving a quartz fiber filter for aerosol collection and a
Teflon-quartz combination for measuring the adsorption artifact. The amount of organic carbon on
the backup filter in the Teflon-quartz combination is a measure of the adsorbed organic vapor on
the quartz filter used for the aerosol sampling.
When this correction procedure is used, we find that it's most important at low concentrations,
a few micrograms per cubic meter, as shown in Figure 14. At higher concentrations it is less
important.
Figure 15 lists some important questions that require further investigation. First, I don't think
that we've adequately defined the role of volatilization or desorption. We don't think it's that
important, based upon calculations we've done and some measurements Steve McDow has done,
but certainly more work needs to be done.
Secondly, the whole question of adsorption equilibrium in the sampling of organic aerosol is not
adequately understood.
31
-------
Thirdly, we need to understand the effect of such atmospheric variables as relative humidity and
temperature.
Fourth, we need to assess the validity of the backup filter correction. I think most importantly
though, we need to have a good theoretical understanding of what's going on in terms of the
physical chemistry-i.e., the adsorption thermodynamics. Such a theory would explain the face
velocity effects, would take into account the multicomponent nature of the adsorption, and would
deal with the problem of approach to equilibrium.
Finally, the problem of gas-particle partitioning for 'semi-volatile' organic species needs to be
much better understood.
32
-------
B. Graphics
33
-------
/• T/j
'* *• *"<*•**?
etftcf
/5
FIGURE 1.
35
-------
/.
•ft Hers /s a
erf aeroso/
A. T^e Q/noutt 0+ Adsorbed OC Con
fis//*i*rt*/ Ay ntaSUny thta»otoTt-eP OC
eo//ecfecj on « vuoriz. ^^ 't?*
a
of 4/t*rf toast
3^
'the, cosrc.cf/0r\ ca/7
FIGURE 2.
36
-------
-26 cm-
Rain cap
Manifold
2 m
Delivery tubes
Base
Filter holder
To pressure gage.
rotameter, and pump
FIGURE 3.
37
-------
1 1
I
1
1
'1
1 -•
1
i
i
1
I
• ' '
51 *
A — . «
^TJ
3-*-
*
Filter holder top
T
1
*S.
I
1
Top mask
Filter combination
Filter support
Bottom mask
Filter holder
bottom
Fdigrure 2.2. Aerosol filter boLder with annular Basks.
The solid black circles are 0-rines.
FIGURE 4.
38
-------
He, He/O*
or
Calibration
Gas
Combustion
Oven
Quartz
*ht
He-Ne Later
Filter MnOa,
1OOO •
Methanator
i
FID
i—
Optical
Fiber
IT Photocell
Apple II
Printer
r
Floppy Dlek
i
X-Y
Recorder
FIGURE 5.
-------
OVEN TEMPERATURE
LASER SIGNAL
FK> SIGNAL
THE DASHED LINE
MARKS THE OC/EC SPLIT
TIME (Mtn)
Figure 1. Typical output for laboratory dhecmaiL-optical
carbon analyzer. Oven temperature, optical
reflectance, and flame ionization detector.
Tbe .dashed line at about 21.3 nututes is Che
split ipotumt between organic and elemental
FIGURE 6.
40
-------
40
0)
20
TJ
CD
73
o
o 0
to
to
O
c
o
O
o
20
-H
0
5;
mJL~ • r
i I I
40
i i i I i
1 ' i r
(a)
Organic Carbon
(b)
Elemental Carbon
I I
80 120
:ace Velocity (cm/s)
160
FIGURE 7.
-------
Concentration
31
CD
•h, C
ro 33
m
00
<
o
£
8
2. <»
§
— ro o
D ro •& o> o o o c
K^ f-CM l — o — i
fOH h-CH K-O-H
101 K>1 f-CH
^ m o -H
— ^« •_/ t-f"H f— OH r~5
•
•
1 W KH ^
Si*
••
•• . ^^^^*
cr
fxj 1^11 L-^^V-J
^^^ I fT Jl f\S~ 1 |
CH KM HCH
(> JCH KW-
^ m « O H
O O fO o KM O
-
t> FO KM
Q".
i i i io i i *m i i fv"i i i i
-------
i r
i r i i
0
c 8
a>
o
I
i i
QQ(l)-fTQQ(2)
QQ(I) + QQ(2)
i i
QQ(I)-QQ(2)
QQ(I)-TQQ(2)
•L
J I
OQQ(I)
D TQQC2)
TQQ(3)
J I
0 20
40 60 80 100 120
Face Velocity (cm/s)
I
-
(b).
55%
(C)
49%
T.
i i
L
140 160
FIGURE 11.
45
-------
CM
o
o
CVJ
O I
o
f
O
ro
O
o
I
I I
(a)
- x
x
I I
(b)
X
TQQ(2) (/igC/cm2)
FIGURE 12.
46
-------
QQ(2)'.(/tgC/cm*)
FIGURE 13.
47
-------
cz
o
O
u
50
40
30
20
10
0
0
j_
4 6 8 10 12
Uncorrected OC (/igC/m3)
14
16
FIGURE 14.
48
-------
3 . Affect *fccf*o*p6tr/c
? /Aeore//ca(
*'
" / c * +*
FIGURE 15.
49
-------
V. THIRD PRESENTATION: SMOG CHAMBER STUDIES, PETER MCMURRY
A. Presentation
PETER MCMURRY: I'm not going to talk about smog chamber measurements, as in the outline,
but rather some of the work we've done on the measurement of atmospheric paniculate carbon.
The outline of my presentation is given in Figure 1.
First, I'm going to talk about measurements of carbon size distributions that we made with micro
orifice impactors. One of the critical differences between impactor measurements and the other
measurements we've heard about this morning has to do with the sampling substrate. With
impactors we use aluminum foil, and we have rather conclusive evidence (I think) that there is no
positive sampling bias with aluminum foil substrates. Indeed, I think that one of the critical problems
that we've been facing is that almost all carbon measurements, organic carbon, in particular, have
been done with quartz filter substrates. I think that in order to achieve accurate measurements in
the long term, particularly in the Southwest, where organic carbon loadings are low, it would be very
desirable to get away from quartz as a filter substrate. It would be nice if there were some
noncarbon-containing fitter other than quartz that could be used.
So, I'll talk about some of our observations regarding impactor measurements of organic
carbon, and will compare some of these measurements with filter concentrations, and then I'll talk
a bit about an experiment we did in which we investigated the effect of sampler temperature on
organic carbon size distributions. Our hypothesis was, that if we had two samplers-one cool and
one warm-sampling the same aerosol, then we should see lower loadings on the warmer impactor
if evaporative losses were significant. I'll then talk about some theoretical work that we've been
doing on evaporative losses. The purpose of this theoretical work is to provide an interpretative
framework. It's a simplistic theory in many ways that does not take account of the true
multicomponent nature of the aerosol, but it's given us a way to think about the problem that has
shaped our approach to measurements. Whether our thinking is correct, I think remains to be
demonstrated, but we found it useful. In this discussion I will talk about the difference between
evaporative losses for condensed material, absorbed, and adsorbed species. I will show that
according to our theoretical predictions, they are quite different. Finally, I'd like to talk about some
recent measurement advances. We've done quite a bit of scanning electron microscopy of
individual particles collected on beryllium substrates. This work has been done in calibration with
Suzanne Hering and Gary Casuccio at R.J. Leed Associates. And we've been really excited by our
ability to measure the carbon content of particles as small as 0. Vm with this technique. This is not
a quantitative analysis technique, but I think it's very useful for gaining information about
51
-------
characteristics of carbon-containing particles in the atmosphere. Finally, I'll talk about a new
sampler, currently in the conceptual stages, which we are planning to work on within the next year.
This work is being done in collaboration with Steve Eisenreich at the University of Minnesota, and
is supported by EPA's grant program.
First, I will talk about impactor sampling. The data in Figure 2 were acquired at the Grand
Canyon about five years ago. On the top is the residual organic carbon size distribution, measured
with a MOUDI impactor, and on the bottom is the elemental carbon size distribution. The
overwhelming feature of these data is this ugly spike sticking up at the small end of the spectrum.
These samples were collected on aluminum film impactor stages. In this case the bottom impactor
stage had a cut of .09^m, and particles smaller than that were collected with a quartz after-filter.
When we analyzed the quartz after-filter, we found a tremendous amount of organic carbon there.
Only a small amount of elemental carbon was found in the after-filter. The elemental carbon size
distribution looked reasonable, but the organic size distributions did not. We don't expect much
carbonaceous material to be smaller than .09^m. We postulate that the spike in the size distribution
below .09^m is due, at least in part, to adsorption of organic vapors on the quartz filter. On
average, roughly 60-70 percent of the organic signal was obtained on the after-filter. If you do
measurements in Los Angeles, where the paniculate organic concentrations are much higher than
at the Grand Canyon, about 30-40 percent of the fine particle organic carbon is found on the
after-fitter. On the other hand, the data in Figure 3 were obtained in Los Angeles with an impactor
that had no after-filter, but had a bottom stage with a cut of about .05^m. In this case, we find very
little signal on that bottom stage. If we had used an after-filter here, instead of an impactor stage,
I am certain that a large spike would have been found in carbon distributions at the smallest sites.
In Figure 4 the organic carbon loadings that we obtained with the impactors are compared with
data from other samplers (Hering et al., Aerosol Science and Technology. 12:200-213). These data
were obtained during the 1986 Carbonaceous Species Methods Comparison Study in Los Angeles.
Note that when the impactor after-filter is not included, the reported organic carbon concentration
obtained by a single-stage impactor is typically less than the mean organic carbon by all samplers
during that study (see bottom left plot). In this case, our bottom impactor stage had a cut of about
.06/4/77. We expect that the impactor should have collected all of the particles. If we include the
after-filter, we are much closer to the mean, on the average.
I think the most likely explanation for these observations is that virtually all of the other samplers
involved quartz filter samples. And they all over-estimated the true paniculate carbon loading.
That's my interpretation, but certainly many other people don't agree. Unfortunately, I did not bring
a plot showing a comparison between the impactor results and the OGC results. This is an
interesting comparison. Our impactor is about 10 or 20 percent below OGC, who correct for
52
-------
adsorption by subtracting the amount of carbon that is collected (presumably by vapor adsorption)
on a quartz filter located downstream of a Teflon filter. According to our measurements, even they
would over-estimate the true paniculate carbon loading.
[Peter, is that less accurate at higher concentrations?]
Is what less accurate?
[Your organic carbon loading, it looks like the difference is greater at higher concentrations.]
Yes it does. I'm really hesitant to make any conclusions about whether it's more or less
accurate at higher concentrations. Based on our measurements to date we cannot make any firm
conclusions about conditions under which organic carbon measurements are more accurate.
[It might be worthwhile pointing out that, if there was such a thing as a positive artifact, due to
sorption on paniculate matter, one might expect the discrepancy to be greater at high particle
loadings. This type of error might be applicable to all other methodologies, but less so for yours,
because of the physics of an impactor.]
Yes. Certainly if there is an error associated with gas adsorption or particles during sampling,
you would expect it to be a more significant error in the filter, because the particles are more
exposed in the filter. However, I would second Jim Huntzicker's earlier comment that if the panicles
in the atmosphere have a relatively long residence time before they're drawn into the sampler, it's
implausible that much sorption should take place on panicles within the sampler. Because of the
pressure drop within the sampler, the tendency is to have the reaction go toward the gas phase,
not onto the particles. But again, it might be that you sample some particles that had not achieved
equilibrium, and they might achieve equilibrium during sampling.
[How does the MOUDI compare with the GM cascade impactor?]
We agreed rather well. The MOUDI is normally operated as a cascade impactor, although this
particular data set was done with a single stage impactor. We also had a multistage MOUDI there.
We agreed rather well with the General Motor's multistage impactor, which had a comparable
sampling schedule and strategy to ours.
We did an experiment several years ago in which we sampled air from Los Angeles. A
schematic of this experiment is shown in Figure 5. We did these measurements at Cal Tech, and
sampled both atmospheric air and smog chamber aerosols that were produced photochemically by
reactions involving toluene, NOX, and sunlight. The sampled air stream was adjusted in heat
exchangers to temperatures of 20°C, 30eC, and 40°C into insulated sampling chambers. We also
preadjusted the temperatures of the impactors to these same temperatures at the beginning of the
sampling. In Figure 6 data that were obtained with the 20°C and 40"C MOUDIs are compared. Our
hypothesis was that if negative artifacts or evaporative losses were important, we should see lower
loadings on the warmer impactor.
53
-------
Results of these measurements are shown in Figure 6. The top plot applies to atmospheric
aerosols, and the bottom to smog chamber aerosols. The solid line is for the 20° impactor, and the
dashed line is for the 40° impactor. What you see for the atmospheric aerosol is that size
distributions of aerosol carbon are only weakly affected by sampler temperature. There is a shift
toward smaller sizes, but essentially no change in total concentration. We believe that the small shift
that is observed is due to the release of water by particles in the 40° impactor. The bottom graph
of Figure 6 shows smog chamber aerosol data. Again, the solid line applies to the 20° MOUDI, and
the dashed line to the 40° MOUDI. In this case we see a terrific difference between the samples
collected at 20° and 40°. From this we infer that significant evaporative losses occurred. We
concluded, based on these results, that the evaporative losses of secondary aerosols, such as those
produced in smog chambers, can be large. However, the data provide no evidence for the
evaporative losses of carbonaceous material in atmospheric aerosols.
[What was the smog chamber experiment, Pete?]
Toluene plus NOX.
[You did see a big difference in the amount of total carbon.]
There was a huge difference there.
[Including the total mass on the different stages?]
We only measured the organic carbon content. We did not measure the total mass, but I'm
sure there would have been. Because virtually all of the aerosol that is produced in this system is
carbonaceous.
[Did you have pre-existing particles in the smog chamber?]
No.
[So condensation was not occurring]
It was a condensation aerosol.
[Is that significant?)
Yes.
Now I'd like to talk about some theoretical work that we've been doing on sampling efficiencies.
We're focusing on evaporative losses. In the following slides, I'll be referring to the sampling
efficiency. We define the sampling efficiency as shown in Figure 7. It's equal to the mass that's
retained on the substrate, divided by the mass that's deposited, or delivered, to the substrate. In
other words, the sampling efficiency is equal to 1 minus the fraction of the delivered mass that
evaporates from the substrate during sampling. Mechanical engineers love to do mass transfer, and
so we applied mass transfer theory to evaporative losses in aerosol samplers. We did this analysis
for both the impactor and filter samplers. The approach that we use is illustrated in Figure 8. Here
is a nozzle from an impactor (impactor orifice), and this is an impaction substrate. The impactor
54
-------
deposit is shown on the substrate. The mass evaporation rate, which must be calculated, can be
determined using mass transfer theory. It depends on the mass transfer coefficient, which depends
on the radial distance from the center of the deposit. As shown in Figure 8, values for the mass
transfer coefficients are high near the center of the impactor jet, and decrease far away from the axis
of the jet. The overall rate of mass transfer also depends on the driving force for evaporation (i.e.,
the vapor density difference). This is the difference between the vapor density in the gas phase and
the vapor density at the surface of the particle deposit. The vapor density difference is lower in the
center of the jet because, due to compressional effects, the vapor density is a little bit higher there.
Studies of mass transfer to jets impacting on flat plates have been reported in the literature for
Reynolds numbers comparable to those that occur in impactors. Correlations that are available from
these studies can be used to calculate likely evaporative loss rates. When you work out the theory,
you find that the key parameter for determining evaporative losses is the ratio of the equilibrium gas
phase concentration at the surface of the deposit, divided by the paniculate concentration. If this
ratio is small, then the sampling efficiencies will be very large, and it doesn't matter how the species
is bound to the particles (whether by adsorption, as Jim was postulating, or whether it is absorbed,
perhaps dissolved in a liquid, according to Henry's law, for example, or whether it is a condensation
aerosol). If the gas particle partitioning ratio is small, then the sampling efficiency should be high
for all binding mechanisms.
The results in Figure 9 apply for a typical filter sampling.1 Theoretically predicted sampling
efficiencies are shown for condensed, adsorbed, and absorbed species. The theory assumes
steady state sampling (constant input aerosol at a fixed temperature) and accounts for evaporative
losses due to pressure drop in the sample. For species that are predominately in the gas phase
at the entrance to the sampler, it's very difficult to obtain accurate gas and particle samples if they're
condensation aerosols. A pressure drop of only 5 percent or 10 percent can lead to a complete
evaporation. The way to think about that is easy. Suppose 99 percent is in the gas phase and 1
percent is in the particle phase. If the pressure is dropped by 5 percent, that makes room for 5
percent more vapor in the gas phase. But the particle phase constitutes only 1 percent to begin
with. You could evaporate all of that and still not re-establish equilibrium. So, for condensation
aerosols, sampling is rather difficult.
For adsorption aerosols, on the other hand, theory shows that sampling efficiency should be
at least equal to 1 minus the fractional pressure drop through the sampler. Bidleman and Pankow,
et al., have argued that PAHs and PCBs are probably adsorbed on atmospheric particles.
1 For additional details, see Zhang and McMurry, Atmos. Environ.. 21,1779-1789, 1987
and Zhang and McMurry, Environ. Sci. Technol.. in press, 1991.
55
-------
Therefore, the filter sampler followed by a gas absorbing sample should give a very good measure
of gas and particle concentrations. That is in direct conflict with results from Coutant, et al., by the
way, that were discussed this morning by Bruce Appel. They concluded that evaporative losses of
PAHs from filters are substantial. There are many, many examples of discrepancies such as this
in the literature for carbon particle sampling, and I really think we don't know the right answer. I
would argue that based on theory, sampling should be very good for absorbed species.
I also want to point out the effect of using a denuder. The theoretical results in Figure 10 are
for condensation aerosol. Whether a denuder is used (right graph) or not (left) the sampling
efficiencies are going to be small for large gas/particle partitioning ratios. Of course, if a denuder
is used such that at the inlet to the filter the vapor concentration is 0, then evaporative losses from
the filter are much worse than if no denuder is used. Let's just take one case, the solid line case.
We have a gas particle distribution of 10. We would predict a sample efficiency of 50 percent if
we're using the denuder for condensation aerosol. This would apply, by the way, for typical
impactor. If we go over to this curve, the gas particle distribution of 10, we would have nearly 100
percent sampling efficiency if no denuder is used. So, keep in mind that if you use a denuder
you're going to lose a lot more material from the filter during sampling.
Figure 11 shows similar theoretical predictions for sorbed species. On the left hand side, we
have denuded, on the right hand side, nondenuded. Again, note that if you're using a denuder for
sorbed species, the sampling efficiencies can be dramatically upset, because you take all the vapor
away. However, without a denuder, you can do quite well using filters. For impactors, you tend to
have somewhat larger evaporative losses than you do for filters for adsorbed species. The reason
for that is, the pressure drops in impactors tend to exceed the pressure drops in filters. With sorbed
species the pressure drop is the key determinant for sampling efficiency. That's different from
condensation aerosols, where impactors are predicted to do better, because there's less exposed
surface area on an impactor deposit. For condensation aerosols, evaporative losses are primarily
determined by the surface area that is available for mass exchange, whereas for absorption it is the
pressure drop.
The theory appears to work well. We have not successfully applied it to any carbon paniculate
measurements. However, we have applied it to nitrate losses. Figure 12 shows data from Walter
John and his co-workers. On the left hand side, we see sampling efficiencies versus gas particle
distributions for the Berner impactor. The lines represent the range of our theoretical predictions
for different impactor stages. Particles were deposited on all impactor stages, and data were only
given for overall evaporative loss, so we were unable to do a stage-by-stage comparison. The
results in the right plot apply to laboratory ammonium nitrate aerosols deposited on filters. This is
the theoretical prediction and, again, the theory and the data are in reasonable qualitative
56
-------
agreement. In any event, John and co-workers noted that the filter losses are much, much greater
than losses from the impactor for ammonium nitrate, in agreement with our theoretical prediction.
For atmospheric aerosols ammonium nitrate appears to be sampled relatively efficiently by
impactors as predicted by theory, and relatively poorly by filters. A comparison between theory and
experiment for ammonium nitrate sampling efficiencies from field measurements is shown in Figure
13.
[Before you change, at the end of Bruce Appel's presentation, he made a comment about it
being virtually impossible to identify the negative artifact for paniculate carbon aerosols. Your theory
says it should be, under certain circumstances, quite pronounced.]
Yes, the theory says that, if the gas particle partitioning ratio is large, it's going to be very tough
to measure what's in the particle phase. This might be the case, for example, with formic acid. If
you have 1 microgram per cubic meter in the gas phase and 1 picogram per cubic meter in the
particle phase, accurate measurements of gas/particle distributions will be difficult. Any sampling
upsets the equilibrium. The tendency is always going to be to drive that material out of the particle
phase. It's going to be very difficult to accomplish.
[So, I guess the question I want to ask is, is it possible to make at least a first-hand order of
magnitude assumption that most of the stuff you're dealing with on particles in the real world is not
exhibiting this large gas to paniculate, this gas phase to panicle partition ratio?]
I second Jim Huntzicker's point earlier on. I don't think that we've paid enough attention to the
physical chemistry, to the speciation and what's in the gas phase, what's in the particle phase. The
gas particle partition question, I think, is a very critical one.
[The analogy is water, right?]
The analogy is water.
[And there the amount in the particle phase is very important at the time you look at the
atmosphere. Even though it's a small percentage of the total that's in the atmosphere.]
Exactly.
[Well, obviously, there's a problem. It just seems to me that your discussion seems to indicate
that the possibility of seeing the negative artifact in the real world is there, and it just hasn't been
looked at yet. Obviously, there are holes, there is a whole broad spectrum of things that can impact
on this, but you're predicting fairly pronounced effects for individual species that have certain
characteristics, and efforts to see this in the broader sense in the real world is not happening.]
You could also say that we are predicting that there are no effects. It depends which side of
the graph you look on. I don't think we're predicting that it's one way or the other. I think I would
certainly argue that if the gas to particle ratio is large in the atmosphere, then it's going to be very,
very difficult to get an accurate sample. I really believe that, but I don't think we have definitive
57
-------
experimental evidence to prove it at this time. We know that almost all atmospheric water is water
vapor, and there's a very small fraction in the particle phase. I don't think we have any idea what
the distribution between the gas and the particle phase is for the carbonaceous species that make
up the abundance of the paniculate organic particle.
I'd like to finish just by talking about a few of things that we're working on. Figure 14 is a
schematic of a Tandem Differential Mobility Analyzer (TDMA) which we've been using to study
properties of atmospheric aerosols. Atmospheric particles of a known size are selected with an
electrostatic classifier (DMA1). We can select the size by setting the DMA1 voltage and flow rates.
In practice, we have worked with atmospheric particles in the 0.01 to 0.5/*m diameter range. We
then adjust the humidity around those particles in the RH conditioner. We can either dry them out
or humidify them. We measure the final size of the particles with DMA2.
In Figure 15,1 am showing data for N2 versus V2. N2 is the number concentration measured
downstream of DMA2, and V2 is the DMA2 collector rod voltage. As the collector rod voltage is
increased, the size of the particles that pass through DMA2 and are detected by the CMC increases.
The data in Figure 15 were acquired in Los Angeles, and are for 0.42^n atmospheric particles.
When we adjusted the humidity to 7 percent, we got the data shown by the squares. (By the way,
the voltage at which this curve peaks is a measure of the particle size that enters DMA2.) Note that
as the humidity is increased to 28 percent, the particle size increases a little bit, and so on to 70
percent. When we got up to 91 percent humidity, we found a very interesting and surprising
response. We found that the particles split into two peaks. These particles' peak that is observed
near 4500V were smaller, and the particles at 7000V were larger. The larger particles absorbed a
lot of water upon humidification. To within experimental uncertainty (2 percent in this study), the
smaller particles did not absorb any water. They were within 2 percent of the size coming out of
DMA1. This led to the interesting question: what are those particles? In subsequent
measurements, we have been collecting these particles on beryllium substrates and examining them
by scanning electron microscopy. We used beryllium because it does not produce any interfering
X-rays when doing the elemental analysis by EDAX. It enables us to look at light elements such as
carbon. What we're finding is that the less hygroscopic particles are predominantly carbonaceous.
The more hygroscopic particles typically contain sulfur and often contain a very small carbonaceous
core, perhaps .OS^m or so. I think these observations raise very interesting possibilities regarding
studies of atmospheric aerosol properties. The data show that these aerosols are externally mixed.
These observations open new horizons in terms of understanding atmospheric aerosol dynamics
and the behavior of carbonaceous particles in the atmosphere.
Finally, one other development, which is just on the drawing boards at this stage, which we will
be pursuing during the next year, involves the development of a new sampler. I indicated my
58
-------
prejudice against using quartz fitters for measuring organic carbon loadings in clean air where the
paniculate organic carbons are low. Also, when I read the literature on diffusion denuders, it seems
to me that there are many unexplained questions and problems with use of diffusion denuders for
organic materials. I think that denuder work should go on, but I haven't been particularly convinced
that any of the denuder experiments are providing definitive results. We have an alternative
technique that we are planning to investigate, which is shown schematically in Figure 16. The
aerosol enters in the outer annular region and clean air, which had just gone through some sort of
an air purification system, enters along the axis of the sampler. A laminar flow is maintained so that
mixing of the aerosol and clean air does not occur. Gas is sampled from the core at the outlet from
the sampler, and the excess flow from the outer annulus gets thrown away. We are interested in
what's in the gas on the axis. What's the idea? If we use a laminar flow, the two flows will not mix.
On the other hand, any gaseous species that are within the sampled aerosol will diffuse because
diffusivities of gaseous species, even relatively high molecular weight organics, are typically 3-4
orders of magnitude higher than diffusivities of the smallest particles. So the organic gases will
spread out, and we can design the sampler such that the gas profile is flat at the exit, but the
particle profile is essentially unchanged. The particles, in other words, all exit with the excess
aerosol flow. We have a design in mind and are working on the development of a prototype. The
residence time in the sampler can be on the order of a second, so we anticipate that it will be
sufficiently short to prevent gases and particles from re-establishing equilibrium during sampling.
[Do you plan to use this to measure what we call the true gas phase concentrations?]
We'll be looking at various organic compounds, including PAHs and PCBs. Steve Eisenreich,
my colleague in this project, is interested in the use of supercritical fluid extraction which will enable
us to go to much smaller sample volumes, and so that will also be part of the work. That
summarizes what I had to say.
[Thank you, Pete. Our next speaker will be Bob Lewis from AREAL's analytical program. Bob
has been primarily interested in specific organics of interest, but has been involved in differentiating
the gas phase and the paniculate phase.]
59
-------
B. Graphics
61
-------
P.H. McMurry
9/6/90 EPA
Paniculate Carbon Measurement
I. Measurements of Carbon Size Distributions with MOUDI impactors
1. Atmospheric OC, EC size distributions: Grand Canyon, Los Angeles
afterfilter adsorption artifact
2. Comparisons of MOUDI and Filter OC concentrations
3. Effect of sampler temperature on OC size distributions
atmospheric samples
smog chamber samples
IL Sampling Theory for Evaporative Losses
1. Theoretical framework
2. Losses for condensed, adsorbed, and absorbed species
3. Comparison of filter and impactor evaporative losses
m. Recent Advances in carbon particle analysis
1. Scanning Electron Microscope analysis of carbon-containing particles
(Dp>0.1 Jim)
2. Diffusion sampler for artifact—free measurements of organic gases
Coworkers
Xinqui Zhang: Size distributions; evaporative loss theory
S.V. Hering, Gary Casuccio: SEM analysis
S. Eiscnreich: Diffusion sampler
Acknowledgments
EPRI, CRC, EPA
FIGURE 1.
63
-------
:;pUAl OFCAMC CAKOa* DISTRIBUTIONS
CfiANO CAWTOK. 8/B/OS-JO/U/83
0. 10
1.00
10.00
DIAMETER. w«
CAKMN otsmiounoMS
CAOTOK.
0. 15
m •
£o. 10"
Q.
O
O7
0
x 0.05'
TJ
0. DO'
0.01
0. 10
l.OO
1O. 00
DIAMETER, urn
Size distributions of residual organic carbon (ROCJ
and elemental carbon (EC) measured at the Grand Canyon
during the SCENES (Subrcgional Cooperative Electric Utility.
National Park service, and Environmental Protection Agency
Study), summer-intensive. Measurements were made with a
microorificc uniform deposit impactor (MOUDI). The
crosshatchcd data arc from the imoactor aftcrfiltcr.
FIGURE 2.
64
-------
RESIDUAL ORCANIC CARBON DISTRIOUTION:
- DUARTE. CA 0/Q/04-Q/10/04
(IB)
' E
X
07
D
CL
O
.
o
_J
TD
• 0. 01
0. 10
I. CO
10.0
O,p. cwn
distributions of R.OC and EC measured with a
MOUDI in Duartc, CA. Note tkal in nhis case ob afccriilicr was
used. Instead,' the bottom MOUDI stage Lad a cutoff diameter of
0.05.>m'.' ".'•-.
FIGURE 3.
65
-------
O
u 10
6
•5
f"'
01
c «
4
• •
u
.« 10
•Moon-Opanlc C*>t>on
(c)iCMSI:
I
O
6
TO
I
C
WAV.
IS
I
U
u 10'
• c
•
a
6
V
•
(J)UU:
MMUI OryMfc C*t»n
r
3
I"
r
I"
O
j( 10
6
(0) EPA: Ouwu n*«
IS
3
c
I
E
0 S II IS
U«an Organic CaUm
O
J) 10
0 S 10 U
M*an Orpinlc C«rt»n
Amtacnl tac particle organic carbon coaccoLralioos
as aqtmrtui ior L2-bour «mpi;n£ periods by ci&bt different
methods. The abscissa is die a*cnge of daia reported by the
AIHL..EMSI, and AV filtcre, the CM dicbolomous sampler, aad
GMiimpaclor (tec icxl). The'diagonal >on each'graph u the line of
equal results. For the GM cascade impactor only stages (or
panicles less than 1.8B./im arc included.
FIGURE 4.
66
-------
2.2 u»
Cyclones
30 1pm
LABORATORY g) OUTSIDE
T, — j
to Const.
Tenp. — -
Giro.
Bath
Schematic diagram of apparatus used in temperature effects study.
FIGURE 5.
67
-------
CA. ?/Z*
20
a. ra i. aa
. QIAMETER. UOT
10.00
.Tffr?ii_
cnznnnurmNs: FUJI za c A«a -we HOUOI.
7/za/as •.
0 :
07
0
0.01.
a. 10 ; i. oo
DIAMETER, urn
10.00
' Total carfjo'it (TQ size distributions measured on the
roof of Keck Lab, Colled^ Pasadena, CA, for ambient and smog
chamber aerosols- Samples were acquired simultaneously by two
MOUDIs that were maintained at different temperatures (20°C
and 40°C). The crosshatchcd data arc from the impaetor
aflcrfiltcr.
FIGURE 6.
68
-------
SAMPLING EFFICIENCY
= 1 -
mass lost by evaporation
mass deposited
FIGURE 7.
69
-------
EVAPORATION FROM IMPACTOR DEPOSIT
/
impactor
orifice
|Ap(r);
clrivihg force
AK(r), transfer
coefficient
Tferosol deposit
e .
Me= /K(r)Ap(r)f(t)27crdr
FIGURE 8.
70
-------
o
^c
*o
o
J"<
&
w
03
CO
Comparison of Sampling Efficiency for Condensed,
Adsorbed and Absorbed Species.
(TYPICAL FILTER)
0.8-
0.6--
0.4"
0.2"
0.1
10
100
1000'
Gas/Particle Partition (pe/Cm)
10000
FIGURE 9.
71
-------
VOLATILIZATION LOSSES TOR COHPENtATIOM AEROSOLS
(INLET SATURATION)
\\ '. » \
\\ '. \ \
\\ \ \ \
L 1.0 10 100
Gas/Particle Distribution (
10000
0.001
0.003
0.010
0.030
• .100
0.1SO
1.7 Sampling •ffleiency for eondyna^tlop aerosols
(tero InUt gas eonc«ntritlon)
1-0 10 100
Gas/Particle Distribution
1000 10000
FIGURE 10.
-------
Sampling Efficiency for Adsorbed Species r~ = 0),
Pe
10 100 1000
Oil/Particle Partition (pe/Cm)
10000
0.1
Sampling Efficiency for Adsorbed Sptoles (^ = 1).
*^^"^^"^* PC
-0.100, 8-0.100
lr«1 filter
1 10 100 1000
Oas/Partlclo Partition (pe/Cm)
1000
FIGURE 11.
-------
Figure 3-6 Sampling efficiencies of Berner impactor for nitrates
F'g"rff 3-7. Sampling efficiencies of filters for nitrates
1.0
0.8
X
O
§0.6
M
O
CM
CM
W
0.4
Q.2
O DATA 0? WANO AND JOHH
1/1140.Olx)
1/U+O.OZx)
1/tl+O.lOxJ
0.1 1.0 10.0
Gas/Particle Distribution
100.0
O D*t« of Hang and John (1988)
0.2
0.1 1 10
Gas/Particle Distribution
100
FIGURE 12.
-------
Sampling Efficiency of Filter for Ambient Nitrate
(SCAQS Data)
T70f=
i.o . 10.0
Gas/Particle Distribution
100.0
THEORY, I/(1+0.IX)
THEORY, 1/U+X)
FIGURE 13.
75
-------
TDMA SCHEMATIC
0)
T
f
DMAjf
P.BLP.I
DMA!
TO -,—IDEW PT|
PUMP
RH CONDITIONER
1
DMA2
VALVES,
t*•»•••**•••**
RH CONTROL
SATURATOR
FLOWMETER
I ^
R
B
R
0
W
N
B
0
A
R
D
FIGURE 14.
-------
CONCENTRATION AS A FUNCTION OF DMA2 VOLTAGE
AT FIVE DMA2 RELATIVE HUMIDITIES
DMA1 SIZE - 0.42 umi 7/28/87i Ili48 a.m.
0. 6
DMA2 Ralatlva Humidity
0.0
3000
4000
5000
6000
8000
10001
DMA2 VOLTAGE. V
FIGURE 15.
-------
GAS DIFFUSION SEPARATOR
Clean Air Inlet
Aerosol Inlet
I
Excess Aerosol
Flow
Gas Sample
to Adsorber
FIGURE 16.
78
-------
VI. FOURTH PRESENTATION: AREAL/MRDD ACTMT1ES, ROBERT LEWIS
A. Presentation
ROBERT LEWIS: I am not an aerosol expert. I know little or nothing about the nature and
behavior of aerosols or how you go about sampling them. I'm an organic analytical chemist and,
for more years than I would like to remember, I have been interested in measuring organics,
particularly semivolatile organics, in the ambient atmosphere and have been concerned with artifacts
that are associated with that process. I will give you a brief overview of what we have been doing
and where we are. I don't know if it really falls under the category of organic aerosols, but I'll leave
that to be determined. I should say we have supported Terry Bidleman's work over the years, off
and on, and I guess Bruce Appel has already talked about some of the work Coutant has done for
us.
Let's start off defining what I mean when I talk about semivolatile organic chemicals.
(Figure 1) These are pretty much my own definitions and classifications that are based primarily
on some years of experience in sampling organic chemicals that have vapor pressures greater than
10"1 mm Hg, or ones we can conveniently get into a canister and out again, or collect on Tenax and
thermally desorb from Tenax. If we drop down below 10"7 mm Hg, experience has shown us that
we collect pretty much everything on the filter, which implies that there is little in the way of a vapor
component for the compounds in the atmosphere. Everything in the broad range in between 10"1
and 10~7 mm Hg would presumably be distributed between the vapor phase and the
paniculate-associated phase in the atmosphere.
(Figure 2) Semivolatile organics include a broad spectrum of compounds that we've been
interested in over the years. Theoretically, I am sure that there are more semivolatile than volatile.
Pesticides are what initially got us into the semivolatile area.
(Figure 3) Pesticides cover everything from totally nonvolatile through gases, but a great
majority of them can be classified as semivolatile. So, we started our work to design a sampler to
determine pesticides and PCBs in the air.
(Figure 4) We realized we were going to have to have a filter and a vapor trap. We weren't
thinking about phase distributions or artifacts at the time. We modified the basic high-volume
sampler and replaced the 8" x 10" filter with a sampling module that encompassed both a filter and
a vapor trap and lowered the flowrate to about 250 liters per minute.
Figure 5 shows a blow-up of the sampling module. Many of you may be familiar with it. A
particle filter about 10 centimeters in diameter preceded a glass cartridge of modified Soxhlet
thimble into which we could put any kind of sorbent. We looked at polyurethane foam (PUF), at
79
-------
macroreticular polymeric sorbents, and inorganic sorbents, and settled on polyurethane foam. PDF
is a good all-around general sorbent for most pesticides and PCBs.
About 1982, General Metal Works introduced this sampler commercially. (Figure 6) They call
it PS-1 PUF sampler. I have referred to it since as the PS-1 sampler, since it doesn't have to use
polyurethane foam. It can use anything. They called it pesticide particulate/vapor collection system.
Again, you can use it for any semivolatile organic. Our first application, as I indicated, was in the
pesticide area. We did quite a bit of sampling back in the 70's. We had a network of nine sites
around the country, and we noted pretty early in the game that we rarely found anything appreciable
on the filter. The particles were almost all collected in the vapor trap. This included DDT, which has
2x10'7 torr vapor pressure. The modelers in those days were modeling the global transport of PCBs
and DDT, assuming they were particle-associated-moving as particles. We began to wonder if that
was the case. So we did a 'quick and dirty" experiment (Figure 7) published in 1976, in which we
spiked particulate-loaded filters with a broad spectrum of pesticides covering from about 10"3 to 10'7
torr-both polar and nonpolar. We spiked at levels that were comparable to what we found in the
atmosphere at that time onto filters that already had 24-hour accumulation of paniculate. We pulled
air through for another 24 hours, and, as you can see, everything stripped off. This led us to believe
that if pesticides were in the atmosphere sorbed on the surface of particles. And it is my belief that
pesticides enter the atmosphere primarily as vapors by volatilization from soils and foliar surfaces.
Anything sorbed to particles we would lose in large part in the sampling process.
We subsequently looked at other things. (Figure 8) I don't have any good slides of these, but
dioxins, tetra- and hexa-chlorodibenzodioxin, dibenzofurans are quantitatively stripped. The vapor
pressure estimates for tetra-chlorodibenzodioxin lie somewhere in 10"6 to 10'7 range. Only the
octachloro isomers are retained pretty much quantitatively on the filter.
(Figure 9) This happens to be from sampling and not from spiking, but it shows the distribution
of PAHs, according to ring size, between the vapor trap and the filter. We did spiking experiments
there, as well, and found basically the same kind of behavior.
I don't need to explain this one (Figure 10) to this audience, but we obviously became quite
concerned that the volatilization artifact, which I call "A" here, might be quite large. People were
going out and sampling with these kinds of samplers and saying this much is in the aerosol phase
in the air from the filter data, and this much is in gas phase from the vapor trap data, whereas they
may be grossly wrong. From that point on, we always combined our filter and vapor trap for
extraction and assumed that we couldn't do anything about it.
[What is the composition of the vapor trap?)
80
-------
We've used all kinds of things: for pesticides, historically, we used primarily polyurethane foam;
for PAH, historically, we have used XAD-2, but we've also used Chromosorbs to Porapack and any
number of other sorbents.
[Do you have any experience with activated carbon?]
You would not be able to recover these kinds of compounds with activated carbon. They would
be irreversibly sorbed. We have looked at that in the past, not only from the sampling standpoint,
but from binding or control purposes.
Again, this is something I don't need to bring up here. (Figure 11) There are variables that
affect, not only the vapor particle partition, but also the sampling artifact. The higher the
temperature, the greater the sampling artifact ought to be. I don't know how particle type and
concentrations are going to affect sampling artifact. One thing not shown here that has always
concerned me are the temporal considerations, the amount of time the particles have been on the
filter before the sampler was turned off. An episode that occurred 30 minutes before the end of
sampling versus one that occurred 30 minutes from the beginning of the sampling cycle will give
different degrees of volatilization artifact problems. It's going to be difficult to deal with this in terms
of interpreting the volatilization artifact.
(Figure 12) Our approach, and we didn't get into this until a few years ago, was to try to use
denuders to separate the gas phase from the particles. The first attempt we made was to use open
tube denuder system-a Gatling gun arrangement (Bob Coutant and Battelle designed this). This
denuder could be fitted directly onto the front of the PS-1 sampling cartridge. It had a limitation of
about 15 liters per minute, so we were really on the edge of what we could detect in the ambient
air. The denuder tubes were coated with silicone grease, and we demonstrated that it was at 98
percent or so effective, with any of the PAH vapors. It allowed particles down about 0.1 micrometer
to pass through to the filter. We did not test particle sizes below that. Then by doing separate
analyses, of course, of the filter and vapor trap, we could measure the volatilization artifact, albeit
with denuded air. Now I feel that the low vapor concentrations-low total gas phase concentrations
of PAHs and pesticides-found in the ambient atmosphere would not be significantly different than
denuded air in the production of volatilization artifacts-but Peter's presentation makes me wonder
about that. We've assumed we were getting pretty accurate indication of volatilization artifact with
the denuder in place.
(Figure 13) We were unable to recover the PAHs from the denuder to do a direct analysis of
them. Of course, there are very small quantities collected and the fact that they are sorbed into
nasty stuff (silicone grease) makes it very difficult to recover the gas phase PAH. Therefore, we
have to make phase distribution estimations by denuder differences, which I'm sure you're all
familiar with; that is, by collocated sampling with and without the denuder. We would combine the
81
-------
filter and the particle filter from the nondenuder sampler on the left to give us a total PAH loading.
The one on the right, with the denuder, is presumably only giving us the particle concentration of
the PAHs, and we do separate analyses of the filter and the vapor trap there. The quantities of PAH
in the vapor trap in the latter case will give us a good indication of the volatilization artifacts.
We did sampling in the summer and the winter in Columbus, Ohio, and published a paper in
Atmospheric Environment in 1988 showing these measurements.
(Figure 14) This shows the ranges and the medians. From our studies, you can see that the
volatilization artifact vary tremendously from 5 percent or 10 percent in the wintertime, up to 90
percent in the summertime, as determined from what was in the trap behind the denuder. This is
for the PAHs that have vapor pressures greater than 1 x 10~7; for everything below 10~7 torr vapor
pressure there is no measurable volatilization artifact or vapor phase. Because of low volume of air
sampled, there may have been small volatilization artifact for BaP that we couldn't measure simply
because it was below our detection limit. BaP has been reported in small proportions in vapor traps
passed behind filters in high-volume samplers.
The vapor phase as calculated by denuder differences also ranges from 10 or 15 percent to 90
percent. So, you can see, for example, for phenanthrene that on the average for a year, 45 percent
is in the vapor phase, whereas a sampler without a denuder would give you 95 percent collection
in the vapor trap. That's what everybody (Yamasaki, Cautreels, and van Cauwenberghe, Bidleman
and others) have been finding. It may well be that half of what they collected in their vapor traps,
on average, was originally in the particle-associated phase. The remainder of it was volatilization
artifact.
The primary limitation to this denuder system was flow rate. We needed a higher volume
system. We also wanted to be able to measure directly what was in the vapor phase.
Our next attempt was this high-efficiency compound annular denuder, that Coutant of Battelle
designed. (Figure 15) It consists of 12 nested cylinders of aluminum, again coated with silicone
grease and is about 8 centimeters by roughly 20 centimeters in size. It will handle a 200 liters per
minute flow rate with the same efficiency as the other one at 15 liters per minute to separate
particles and gas. We designed it also to fit in a thermal desorption unit. That's what the cylinder
on the right is, with the gas finings on the end caps, so we could purge it with inert gas while
heating it and try to recover the vapor phase from the silicone grease. This has not worked out very
well, so we're still constrained to the denuder difference measurements, but we at least have a
higher through-put.
Figure 16 is just a picture of the modified PS-1 sampler with a denuder on the front of the
sampling cartridge. We are just starting to use this system. We did some summer and winter
sampling in Columbus and combined that data with the low-volume denuder data. Starting this
82
-------
month, we are deploying the samplers at two sites as part of our Toxic Air Monitoring System
(TAMS) network-one in Houston and one in Boston. We'll be doing a broad spectrum of
semivolatiles on a 12-day sampling cycle over the next year. It will be the first time that we've done
anything other than BaP at these sites, and it will also, hopefully, give us some information not only
about the levels of other PAHs, but also their phase distributions. On a subset of the samples
collected (about 20 percent) we'll analyze the vapor trap and filter separately from the denuder
sampler to try to get a measure of the volatilization artifact and see what it's doing with respect to
temperature and concentrations.
[What does TAMS stand for?]
Toxics Air Monitoring System.
We're also doing for the first time this year polar volatile organic compounds by using a
modified canister technology.
Recently we have designed and constructed a miniaturized version of the denuder. This one
can operate at 20 liters per minute, and the reason for the downsizing is that you can't sample 200
liters indoors without upsetting the ventilation rate of the building or room. Ventilation experts like
Jim Woods tell me that 20 liters per minute is about it, and that is the rate that this is designed to
sample.
Figure 17 shows a picture of the indoor sampling cartridge, which is similar to the PS-1, with
47-mm filter covering a smaller glass thimble.
Figure 18 shows the quiet medium volume sampler that Battelle designed for us to pump air
through these cartridges. We have not used the small denuder on them yet. We have used the
samplers in a couple of small studies without any problems.
[Are these outdoors or indoors?]
Indoors. Very quiet.
[Do you have sampling programs outdoors?]
TAMS is outdoors.
[It's both indoors and outdoors?]
TAM's is only outdoors.
Figure 19 shows another recent effort that we're in the middle of now. We are working with Bob
Burton, who's in our Exposure Assessment Research Division, and his contractor, Virgil Marple, to
modify the high-volume virtual impactor (HWI) sampler. The HWI is at the top of the figure. What
we are doing here, basically, is adapting the HWI, which is designed for a 40 CFM flow rate, to a
four-channel sampler that is basically four PS-1 samplers in a package. It's not going to be much
bigger than one PS-1 sampler.
[PS-1 sampler?]
83
-------
The PS-1 is the one I showed you earlier that General Metal Works markets for pesticides and
is also used for other semivolatiles. The added bonus is that we get a size-selected inlet that we
don't have currently with the PS-1, at least commercially available. In Bob's drawing here, he's
showing that you can use various sorbents and various filters at the same time. We don't intend
to use it that way, at least initially. Our first application is going to be in the Great Lakes Air
Deposition Program. We will use the same sorbent, a combination of polyurethane and XAD-2, to
cover the different classes of semivolatiles: pesticides, PCBs, polynuclear aromatics, and perhaps
dioxins. The latter is still up in the air because of the analytical cost involved in determining dioxins.
This sampler will give us these compounds on separate filters and sorbent traps, so that we can
send them to different laboratories that have expertise in these areas. Even with the same
laboratory doing the analysis, it becomes an analytical nightmare to try to do PAHs and pesticides,
for example, in the sample. The extraction process is not going to be optimized for one or the other
class. The analytical finish may be different. So this is an attempt to get a small package to satisfy
everybody's interest in different classes of compounds and to do it in one sampler. We will be using
the XAD/polyurethane foam with or without the denuders as we go along. One combination that you
might envision here would be to have denuders on two of the channels, and no denuders on the
other two. One pair of samples could be used to do phase distributions for pesticides, the other
for PAHs.
Figure 20 is a picture of what I call a "sandwich11 trap, published in Analytical Chemistry in 1982,
which gives us the best of both worlds. The low pressure drop of polyurethane foam is a definite
advantage over other sorbents, but the shortcoming of polyurethane foam is its low capacity for
compounds with vapor pressures above 10~3 torn The sandwich trap with PUF and XAD-2 extends
the range of application and minimizes the pressure drop. We have tested to this for pesticides,
PCBs, and we're currently evaluating it for PAHs, and reevaluating it for some pesticides and PCBs.
That's all I have to say.
[You've really made a tremendous start...I'm wondering whether you try to sum up the
compounds you've recovered and compared that to the total organics that you have recovered by
combustion techniques?]
No, we have not, but I think the Great Lakes Program will be potentially a start in that area. We
will be doing all these kinds of things. Bob Stevens, who is involved with that program, will be doing
that, as well.
We're also looking at supercritical fluid extraction (everybody in the world is now) and we
actually have been since about 1983, when we first funded work by Dick Smith and Bob Wright at
Battelle Northwest. They looked at polyurethane foam, XAD, Tenax, and a number of other
adsorbents, and compared SFE with Soxhlet extraction, for PAHs primarily. The extraction
84
-------
techniques were pretty much comparable in efficiency. Of course, the advantage is that SFE is
faster. Our own shop has been trying to develop on-line SFE-GC techniques to desorb both volatile
and semivolatile compounds from sampling cartridges. The sampling traps have to be small by
necessity, so we don't want to use these big samplers I've been talking about. We might not need
these big samplers; if we can get all the sample into the GC, we'll have a 200-fold increase in
sensitivity right there.
[I've heard that some people are sampling on activated carbon by media where the amount of
activated carbon per unit of sampling medium is very constant and very small standard deviation,
so that if they analyze both blanks per sample by combustion techniques they could detect the
amount of the sample added. Have you heard anything about that?]
A little, yes. That's fine. What bothers me are people who try to use activated charcoal to
determine compounds that you can't get off activated charcoal by solvent extraction.
[But getting them off by combustion?]
Combustion, yes.
[Do you feel that is acceptable?]
Yes, I think that the problem with getting them off by combustion is interpreting the meaning
of the results. It gives a total organic content. I don't know. How do you compare chloroform or
carbon tetrachloride with ethane?
[No, it wouldn't give you any molecular resolution.]
I'm not talking about resolution, but your mass measured as total organic carbon relative to the
original molecular weight of your compound. If you're trying to use your total carbon measurement
as an indication of the total mass of organic in the air, it's a big guess. If the air is enriched in
compounds that have a lot of heavy atoms like chlorine versus those that don't, the guess might be
way off.
[You use this in TAMS, your basically urban setting. What other what you might regard as
remote settings have you used these samplers at?]
For PAHs, none. For pesticides, and we're talking about somewhat ancient history here, quite
a few, but I don't think there is any bearing now on what we're interested in today. The work that
we did best back in the 70's was directed at DDT and dieldrin and aldrin. People are still interested
in these things from air deposition standpoint in remote areas, so we'll be taking a second look at
them now in the Great Lakes Program. But we'll expand to include more current pesticides.
[Thank you, Bob. The next speaker will be Steve Japar. Steve is going to tell us about some
of the work that's going on at Ford Motor Company.)
85
-------
B. Graphics
87
-------
AIRBORNE ORGANICS
• VOLATILES (>10-a mm Hg)
• SEMIVOLATILES (101 - 107 mm Hg)
. PARTICULATES (<107 mm Hg)
FIGURE 1.
89
-------
VAPOR PRESSURES OF SEMIVOLATILE AND NON-VOLATILE ORGANICS
Vapor Pressure
(mmHg)
10 "1
10 -2
10 -3
10 -4
ID'5
10 '6
10 -7
10 -9
I
10-12
Chlorinated
p-Dichlorobenzene
Tetrachlorobenzenes
Pentachlorophenol
Aroclor 1242
Heptachlor
Aroclor 1254
Chlordane
i Lindane
Aroclor 1260
2,3,7.8-TCDD
DDT
Oxygenated or
Nitrogenous Alkanes
Aniline Dodecane
Toxaphene
Nicotine
Glycerol Hexadecane
p-Nitroaniline
Diazinon
Butyl phthalate
Parathion Eicosane
Hexacosane
DEHP
Nonacosane
Polynuclear
Aromatics
Naphthalene
Anthracene
Phenanthrene
Pyrene
BaP
Coronene
FIGURE 2.
90
-------
(O
PESTICIDE
p,p' - DDT
DIELDRIN
LINDANE
HCB
PARATHION
DIAZINON
HEPTACHLOR
RONNEL
DICHLORVOS
TOXAPHENE
METHYL BROMIDE
i
VAPOR PRESSURE*
(X 10? mm Hg)
1.9
7.8
94
109
378
1400
4000
80,000
1 20,000
2-4,000,000
7,600,000,000
REFERENCE
TEMPERATURE
(°C)
20
25
20
20
20
25
25
25
25
20
4.5
COMPILED FROM E.Y. SPENCER (1968).
FIGURE 3.
-------
FIGURE 4.
92
-------
FIGURE 5.
-------
MODEL PS-1 PUF SAMPLER
Pesticide Participate and Vapor Collection System
• Especially designed for sampling airborne particulates and
vapor contamination from pesticide compounds.
• Successfully demonstrated to efficiently collect a number
of organochlorine and organophosphate pesticides.
• Employs SURC Sampler concepts.
• By-pass blower motor design permits continuous sampling
for extended periods at rates to 280 liters per minute.
• Proven sampler components housed in aluminum shelter
anodized for outdoor service.
FIGURE 6.
-------
PESTICIDE
ALDRIN
tt-BHC
LINDANE
HEPTACHLOR
HEPTACHLOR EPOXIDE
TRIFLURALIN
DIELDRIN
ENDRIN
p,p' - DDT
o,p' - DDT
p,p' - DDD
p.p'-DDE
ENDOSULFAN 1
ENDOSULFAN II
DIAZINON
PARATHION
MALATHION
METHYL PARATHION
ETHION
INITIAL CONC.
ON FILTER
(/ng/cm2)
0.30
0.07
0.07
0.06
0.14
0.15
0.50
0.70
0.50
0.50
0.30
0.30
0.27
0.40
1.00
0.50
1.20
0.50
1.05
% RECOVERY AFTER
1 hr
STATIC
58
53
76
58
100
129
99
103
100
100
100
96
103
100
96
103
100
99
100
24 hr
AT 280 l/min
0
0
0
0
0
0
1.2
1.7
5.2**
1.7**
0
1.2
0.3
0
0
0
0.8*
0
0
^FILTERS CONTAINED A 24-hr ACCUMULATION OF PARTICULATE
MATTER (ca. 50 /^g/cm2) COLLECTED FROM AIR PRIOR TO DOSING
WITH PESTICIDE MIXTURE.
••CORRECTED FOR BACKGROUND.
FIGURE 7.
95
-------
PERFORMANCE OF PS-1 SAMPLER FOR PCDFs AND PCDDs
MEDIUM
SPIKE
LEVEL,
ng/m
AVERAGE PERCENT RECOVERY
TETRA-
CDF
TETRA-
CDD
HEXA-
CDF
HEXA-
CDD
OCTA-
CDF
OCTA-
CDD
FILTER 0.5
PUF
0.3
82
1.5
95
4.8
103
10
97
57
25
86
13
TOTAL
82.3
96.5 107.8
107
82
99
FILTER 0.01 2
PUF 79
TOTAL 81
01 3 26 41
95 93 104 71 67
95 94 107 97 108
FIGURE 8.
96
-------
PAH Phase Distributions
Vapor
90%
Filter
10%
Three-Ring PAH
Vapor
80%
Filter
20%
Four-Ring PAH
Vapor
10%
Filter
90%
Five-Ring PAH
Filter
100%
Six-Ring and Up PAH
and Nitro PAH
FIGURE 9.
-------
V
Filter
V + A
FIGURE 10.
98
-------
SEVERAL IMPORTANT VARIABLES
AFFECTING VAPOR-PARTICULATE PARTITIONING
- TEMPERATURE
- TOTAL SUSPENDED PARTICLES
• LOADING
• TYPE
- AERIAL CONCENTRATION
- TEMPORAL CONSIDERATIONS
FIGURE 11.
99
-------
DENUDER
ASSEMBLY
8
TRANSITION
SECTION
FILTER HOLDER
FIGURE 12.
-------
Approach:
Collocated Samplers
Data: Ct, Fnd, Vnd, Fd, Vnd
Unknowns: Gv, Cp, A
I n/H —
nd
V
nd
ND
Denuder
-F,
D
C,=
Cp = Fd + Vd
L/v = C/t —
A= Vd
FIGURE 13.
101
-------
PHASE DISTRIBUTIONS AND VOLATILIZATION ARTIFACT LEVELS OF POLYNUCLEAR
AROMATIC HYDROCARBONS3
COMPOUND
Anthracene
Phenanthrene
Pyrene
Fluoranthene
Percent in Vapor Phase
Before Collection
Median
57
51
43
43
Benz(a)anthracene 32
Chrysene
40
Range
14-92
25-86
5-80
27-64
8-67
15-65
Volatilization Artifact
as Percent of Total
Median
32
45
47
43
30
18
Range
13-92
13-80
16-83
7-62
8-53
6-50
Benzo(a)pyrene
Benzo(e)pyrene
Benzofluoranthene
Perylene
Benzo(g,h,i)-
perylene
Indeno(1,2,3-
cfd)pyrene
Coronene
ND
KD
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
a. Median and ranges of measurements made during summer and winter months.
b. ND «= not detected.
FIGURE 14.
102
-------
o
CO
FIGURE 15.
-------
-"• ..*»«
«w^
/ ^-
FIGURE 16.
104
-------
AIR INLET
GASKETS
STAINLESS SCREENS
QUARTZ WOOL
ALUMINUM
PYREX
STAINLESS SCREENS
GASKET
TO VACUUM
FIGURE 17.
105
-------
•51 cm
35 cm
o
o>
ACOUSTIC
INSULATION
21 cm DEEP
VIBRATION
ISOLATION
SAMPLER HOUSING FOR LOW FLOW RATE INDOOR SAMPLER
FIGURE 18.
-------
ELECTRONICS
COARSE (74-10 micron)
PART. FILTER
UNIVERSAL MULTICHANNEL
AIR SAMPLING STATION
(UMASS)
FINE (0-J.S micron)
FART. FILTERS
1. NUCLEPORE
2. TEFLON
3.PALLFLEX
i. QUARTZ
T T
SJrtn
CLAVIPS
6 S etm »J efm
J
t
0.5 cfm
i
t
0.5 cfm
FIGURE 19.
107
-------
65-mm x 125-mm
GLASS
CYLINDER
SUPPORT
SCREENS.
ii
I
i
III
liiii
00°0°00
sgsgsp
\°°l°°l°°l°°
\°°l°°l°°l°°
slsfsgs
s|s|8gs
ss°isgs
s|§|s|s
o«o2o°o
O" O ~
of§§
0°1°
II
ills
II
P
Hi
iii
Ills
III
If!
Ill
§§!§!§
m
II
I
I
!!!iH
•I
-.1*
in
in
^^.n^,
.^ ^\ ^ y*V ^^.X^
2-IN PUF
PLUG
50 cm3GRANULAR
SORBENT
1-IN PUF
PLUG
FIGURE 20.
108
-------
VII. FIFTH PRESENTATION: ORGANIC AEROSOL RESEARCH
AT FORD MOTOR COMPANY, STEVEN JAPAR
A. Presentation
STEVEN JAPAR: What I am going to do today is discuss a little bit about the research that is
currently being done or contemplated at Ford Motor Company on organic aerosols. It's really not
particularly germane to the discussion that is going here; but one aspect of it which I'll talk about
is the way we are going to generate some aerosol of studies; which may be of some interest to you.
What I would like to do, though, is indicate why Ford Motor Company is even interested in this, and
present a little bit of history to point out what we've done in the past. Most of what I'll talk about
involves urban aerosols rather than rural or remote, and the reason for this will become obvious in
a second. When you talk about urban aerosols, organic carbon urban aerosols, there are a number
of impacts people obviously realize. The visibility degradation focus has been primarily on elemental
carbon, although now the discussion is getting more into organic aerosol carbon. We have in the
past done quite a lot of work on developing techniques to measure elemental carbon in an urban
environment. We have done a lot of work using photoacoustic spectroscopy. As for other impacts,
organic aerosols are sinks for hydrocarbon and NOX, and there is an impact, fairly undefined at this
point, on urban and regional ozone questions. Just as an indication of some of the uncertainties,
if you try to do a mass balance on aromatics, for example, in the atmosphere, a considerable
amount of the material disappears. You don't know where it goes. Organic aerosol formation is
playing a part. If you look at natural hydrocarbon systems, for example, pinenes, a tremendous
amount of the mass is known to go into the organic aerosol component, again with very little
understanding of how it actually gets there. There are also potential health effects. It has been
demonstrated that primary and secondary organic aerosol is mutagenic in the Ames assay. Work
at Ford has been quite active in that area for a number of years. We have also been involved in the
area of carcinogenic animal studies with the focus being on PAHs and their oxidation and nitration
derivatives. Early work in this area had quite a bit of contribution from Ford Motor Company in the
development and the understanding of how nitro-PAHs impact some of this, especially in terms of
vehicle emissions-type considerations.
Why is Ford particularly interested in doing any research in the area? I think it has been
indicated on the previous slide-hydrocarbon and NOX emissions, for which we are well known,
participate in aerosol formation under any set of circumstances you can envision. Specifically, we
are interested in this in terms of hydrocarbon and NOX sinks, and in understanding the mechanism
involved in the formation of those organic aerosols. Aerosols are obviously a specific focus of PM-10
109
-------
regulations. The contribution of organics to PM-10 is not all that well understood. They contribute,
probably not significantly in many areas, but they're there. Light scattering-William Wilson talked
a little bit about that. There are now questions evolving about organic aerosol contributions to light
scattering in regional environments such as the Northeast. We also have the question of air toxics,
not only in the context of some of the compounds being toxics themselves, but also in terms of
these materials being vehicles to get other materials into the lungs.
If you look at that, there are a number of research opportunities that one can get involved with,
and they are characterized in three broad areas. First, there is the characterization of ambient
anthropogenic organic aerosols, including methodology (which is really the discussion today). This
is important because it will lead into questions of discerning primary versus secondary sources. If
you can identify compounds in the organic aerosol, these become starting points that allow one to
go back and look at mechanisms for the aerosol formation. Second, there is the question of
potential health effects. The area that we are mostly interested in at this point is products and the
mechanisms for their formation from gaseous reactions, involving both measurements of the physics
and chemistry. The focus of interest is the identification of classes of reactions leading to organic
aerosols. We have pretty good ideas about that, but I don't think everything is understood that
needs to be understood. There is also the question of how all of this impacts on the reactivity of
specific hydrocarbons in the atmosphere. Third, there is an area dealing with nucleation and
condensation processes which impact strongly on the efficiency of aerosol formation processes in
the atmosphere, and also is a first step toward looking at interactions among organic and inorganic
systems.
What are we doing at Ford? I'm going to talk, except for one item, in very general terms about
what is going on. We are interested, as a Laboratory, in looking at the extension of proven
techniques to difficult or novel systems in terms of organic aerosols. We are not focusing on the
collection aspect of it. What we are focusing on primarily is how you analyze an aerosol after you
get hold of it. We're looking at new mass spectrometry approaches to this. The Analytical Sciences
Department has an example: an API triple quad mass spectrometer which will eventually be used
to look at some of these organic aerosol questions. Its main focus is in other places at the moment,
but there is the intention of going back and looking at this as an analytical tool for organic aerosols.
Sample handling refers to how you work up an organic aerosol system once you have it. Do
you want to do an HPLC analysis and look at mutagenicrty, for example? There are a lot of
unanswered questions in the way things are routinely done now. You have a lot of material which
is essentially unidentified, and the idea is to improve this sample handling methodology so that you
can get more information out of those kinds of studies. In terms of the Ames assay, people are
always developing new variations of the technique, and they're not always applicable to the kind of
110
-------
systems we are interested in looking at. In this particular case, I must admit we are mostly
interested in vehicle emissions-type considerations. But the application of the Ames assay and all
its variants is not exactly straightforward, so there is work being done in the Chemistry Department
on that aspect of it. Much of this work, by the way, is done in the Analytical Sciences Department.
I am in the Chemistry Department.
One area in which I am interested is the use of infrared spectroscopy for the analysis of bulk
samples. There are reports in the literature about this, involving ambient systems, which, because
of their complexity, are very difficult to study. I would like to go back and look at more simple
systems-one component hydrocarbon systems-and see what you can do with that approach under
a more simple set of circumstances. Nothing is simple in aerosols, but if you start out with one
hydrocarbon, you can't do any better. So that's the idea-to back away from the ambient setting
to a specific situation that I can set up.
I thought it would be interesting to mention to this audience that one of the things we are trying
to do, with a certain degree of success to date, is work out cooperative research programs with
university and government laboratories. In these areas, we have been talking with a number of
university groups to try to get together on joint research. We have the opportunity to do this with
government laboratories also. We've been talking with Argonne National Laboratory, for example,
not in this area but in other areas. Ford Motor Company is actually doing some paint work with Ed
Edney over here, and that's been extremely valuable to us.
One thing that we would like to do in terms of aerosol work, in terms of organic work really, be
it gas phase or aerosol, is to look at the evolution, as I put it here, of the health impact of an
ambient aerosol as it moves along with a transported air mass. Start out at Point A. Understand
the chemistry and the physiology of what's going on there. Watch transport over a known area and
measure what's happening in terms of organic aerosol, biological activity in that aerosol, whatever,
during transport over a known distance. Again, we are trying to do this as a cooperative research
program primarily with universities, and we are trying to address the atmospheric transformation of
organics in that. Nobody has really done a very good job of that and it can't be done in a static
one-point analysis.
The one thing I want to take a little bit more time on is this program here, which is going to look
at the mechanism of organic aerosol formation in simple systems. These will be laboratory studies
which will allow time-resolved physics and chemistry to be looked at in aerosol systems. Instead
of a static reactor that most people use, we will use a flow reactor. We're going to use conventional
analytical techniques. We're not in the business at this point of trying to go beyond where the
collection techniques are at present, but we will use these techniques in whatever proportion is
appropriate, and we will go back and use GC\MS, GC\FID analyses.
111
-------
Very briefly, if you do work in a static aerosol reactor, you've got a fairly simple system. You
can bring your components together, you can irradiate them, you can do whatever you want to
them. Then, you can take a sample out and analyze it. That's fairly straight-forward, but you have
a couple of serious problems. Some people in this room have addressed some of these problems
over the years. The main problem you have is one of inhomogeneity. You're making these aerosols
and they are growing. You've got to mix the system in order to do an analysis, and mixing leads
to severe wall losses. If you're running experiments over four, five, six, or seven hours to look at
what's happening, these problems can be severe. The second problem is the long collection time
to obtain sufficient material for chemical analysis. If you want to do a conventional GC\MS analysis
or whatever, you're going to be collecting aerosol over an hour, two hours, three hours, in the
system. What you're doing is integrating your aerosol over a large dynamic change in the aerosol
formation process. While you can measure the physics of the aerosol particle; size, particle number,
etc.; essentially in real time, you can't do that for the chemistry. These are two main problems.
They are offset by what is relative experimental simplicity. I'm not trying to minimize the difficulties
with it, but it is a relatively straightforward way of doing it, at least in terms of experiment. What
we're going to do at Ford is very different. We're going to a flow reactor system. Mix all your
reactants up at the front end of the reactor. (This is the ultimate simplicity in the way I've pictured
it. There's going to be ten times more instrumentation in the front end of this than there is involved
in the tube itself, because you're going to mix everything up here.) You're going to introduce it in
the flow tube. You're going to let this flow down the tube and then you will analyze it. A sampling
probe will go to your various analyzers coming out that end. The advantage of this system is that
the reaction time starting, if this is time 0, the reaction time at the sampling port is inversely
dependent on the flowrate. If you use a specific flowrate, the reaction time is dependent on the
placement on the sampling port in the tube. We can collect a sample for as long as necessary and
remain at the same reaction time. We can then do a real time chemistry, and, of course, the real
time aerosol physics on these systems using this system. For the first time, we should be able to
look at a strict evolution of the chemistry of an organic aerosol system as it goes through the gas
phase to the point where, in a pure hydrocarbon system, you get nucleation, and then
condensation. Does the chemistry change as these systems evolve? There's every reason to
expect that it does. If you look at something simple like the ozonolysis of cyclopentene, the reaction
forms approximately 12 organic carbon-containing products, about seven of which have a sufficiently
low vapor pressure that they will homogeneously nucleate in the system or condense on other
homogeneous nuclei that are formed. For the first time I think we will be able, if we can make this
thing work, to look at those kinds of processes. The advantages with the flow reactor approach are:
minimal homogeneity problems, especially if you sample on the center line down that tube; and you
112
-------
can do the aerosol chemistry and physics in real time because of this proportionality with either
distance or flowrate, depending on how you set up the experiment. The disadvantage relative to
the static reactor approach is a great deal of experimental complexity on the front end of this thing.
The physical system, by the way, is now in place. The tube is there-a 147 liter, 3-meter tube
about 12 inches in diameter. Hopefully, in a few weeks or months, we may actually get some work
done on this.
But what would we like to look at? In terms of vehicle and manufacturing emissions, there is
the question of aromatic hydrocarbons, which I mentioned previously. If you look at the simplest
hydrocarbon, something like toluene, about 40 percent of the chemistry is unknown. Aerosol
formation is known to occur in these systems and is obviously playing a role in them. How much
of a role is something to be understood. There are also natural hydrocarbon systems that one can
look at. I talked about ozone and cyclopentene. That's the simplest system that I can conceive of
in order to prove out the system, but in the broader aspects we can look at the pinenes. There's
a tremendous amount of chemistry involved there; in gas phase chemistry and the formation of
aerosols with the ozone reaction, and the OH reaction has really not been studied. One other
system we want to look at is dimethyl sulfide, which is a biogenic emission thought to play a role
in condensation nuclei formation in marine atmospheres. All that anybody knows about it is the rate
constant for its reaction with OH, and that you can end up with SO4 at the other end. What
happens all the way through here nobody really knows. There's a lot of potential for doing things,
setting up very simple systems where we can go and probe the time evolution of the aerosol and
go back and look at it.
That's really all I have to say about that at the moment. We're not in the analysis or collection
of organic aerosol, per se. That kind of work is hard for us to justify. We can talk about
mechanisms and aerosol formation in terms of hydrocarbon and NOX because Ford Motor Company
understands that. So that is what we are doing and where we are trying to go. These things always
evolve. It's hard to set a course and maintain it, so things kind of vary back and forth in the
industrial environment.
[Thank you very much, Steve. We have one more representative of the industrial sector. Peter
Mueller represents EPRI, which derives its money from the utility industry, but operates as a relatively
independent research organization. Peter will tell us about some of the work that EPRI is doing.]
113
-------
B. Graphics
115
-------
Organic Carbon in
Urban Aerosols
Impacts:
•Visibility degradation
Focus - Elemental carbon
•Sinks for HC, NOx - Urban/regional ozone
/Aromatic HC - 50% of alkyl benzenes
disappear
/Natural HC - > 50% of pinene oxidation
leads to aerosols
•Potential health effects
/Primary/secondary organic aerosol
is mutagenic in the Ames assay
/Soots are carcinogenic in animal
studies
Focus - PAHs and their oxidation/
nitration products
S. Japar/EPA 9/6/90
FIGURE 1.
117
-------
Organic Aerosol Formation
Impact on Ford Motor Company
• HC, NOx emissions participate in
aerosol formation
• Involved in chemistry of urban/rural
ozone as sinks for HC, NOx species
• Specific focus of PM-10 regulations
• Efficient visible light scatterers
which impact visibility degradation
• Involved in air toxics problems as
vehicles for toxics to lungs
S. Japar 2/28/90
FIGURE 2.
118
-------
Organic Carbon in
Urban Aerosols
Research Opportunities:
•Characterization of ambient, anthropogenic
organic aerosol, including methodology
development
/Primary vs. secondary sources
/Presence of compounds points to
mechanism
/Potential health effects
•Products/mechanism of formation in gaseous
reactions - physics and chemistry
/Identification of classes of reactions
leading to organic aerosols
/Reactivity of specific HCs in the
atmosphere
•Nucleation and condensation processes
/Efficiency of aerosol formation process
/Interactions among organic, inorganic
systems
S. Japar/EPA 9/6/90
FIGURE 3.
119
-------
Current and Projected Research
at Ford Motor Company - 1
Mechanisms of Organic Aerosol
Formation
•Laboratory studies of the time-
resolved physics and chemistry
of aerosol systems
•Flow reactor
•Conventional collection techniques
Filters
Adsorbents
Cascade impactor
•GC/FID, GC/MS analysis
•Evolution of health impact of
ambient aerosols in transported air
masses (Cooperative research with
University, Government laboratories)
S. Japar/EPA 9/6/90
FIGURE 4.
120
-------
Current and Projected Research
at Ford Motor Company -2
Extension of Proven Techniques to
Difficult/Novel Systems
•Mass spectrometry
•Sample handling
•Ames assay
•Infrared spectroscopy of "bulk"
samples
Cooperative research with University,
Government laboratories
S. Japar/EPA 9/6/90-
FIGURE 5.
121
-------
Systems to Study
Organic Aerosol Program
Vehicle/Manufacturing Emissions:
Aromatic HCs
About 30% of NMHC emissions
from M-85 fueled vehicles
Prominent components of paint
related emissions
cj
O
25%
C*3
Aerosol
von
S. Japar February 28, 1990
FIGURE 6.
122
-------
Systems to Study
Organic Aerosol Program
Natural Hydrocarbons:
Monoterpenes - U.S. emissions rates
equal those of anthropogenic HCs
Cu * CASES, Ae«o$oLSJ CO,
oH ? ?
Dimethyl Sulfide - Biogenic emission
thought to be source of condensation
nuclei in marine environments
CrtjSCH *OM * -"2^ ???9
aeroso
S.Japar February 28, 1990
FIGURE 7.
123
-------
Static Aerosol Reactor
S. Japar/EPA 9/6/90
FIGURE 8.
124
-------
Static Reactors for
Organic Aerosol Studies
Disadvantages:
•Inhomogeneity - mixing leads to
wall losses
•Cannot follow gas-to-particle
conversion in real time because
of long sampling times required
for aerosol analysis
Advantages:
•Experimental simplicity
S. Japar/EPA 9/6/90
FIGURE 9.
125
-------
Aerosol Flow Reactor
1 1
.TV^^
S. Japar/EPA 9/6/90
FIGURE 10.
126
-------
Flow Reactors For
Organic Aerosol Studies
Advantages:
•Minimal homogeneity problems
(sample on the center line)
•Aerosol chemistry, physics
measurements can correspond in
time because reaction time =
f(flow distance) or f(flow rate)
Disadvantages:
•Experimental complexity
S.M. Japar/EPA 9/6/90
FIGURE 11.
127
-------
VIII. SIXTH PRESENTATION: ORGANIC AEROSOL RESEARCH PERSPECTIVES, PETER MUELLER
A. Presentation
PETER MUELLER: I'm glad to be here. I'm glad that the Environmental Protection Agency is
focusing in on the organics, but obviously from what we've heard from some of the EPA staff this
is really not such a new subject for many; much has already been accomplished. In addition to
visibility, they also relate to the community health issue. From the health perspective, it's a matter
of deposition with respect to eye irritation, or it's a matter of inhalation and deposition in the airways
and adsorption there. At the target site irritation would be an acute health problem, or absorption
could lead to a systemic health problem. Whatever the effect associated with the organic and other
carbonaceous material that is in the atmosphere, our science will sooner or later have to cover the
molecular composition of organics.
We first got started with looking at noncarbonate carbonaceous material in the atmosphere
because the HiVol filters we'd been looking at several decades ago were always black, especially
in urban areas. Some asked what the black material is and what fraction it is of paniculate matter.
To answer these, we tried what we thought was pretty easy, just to do a total carbon analysis.
Something we thought every microanalytical chemist knows how to do. When we tried to start doing
that, we found it wasn't that simple. In fact, one of my colleagues got his Master's degree in
chemistry at UC-Berkeley working on the problem, and had a heck of a time getting a good analysis.
We ended up with the combustion techniques that we're now talking about, and began taking some
atmospheric data. The first set was obtained in connection with ACHEX. It showed that in the Los
Angeles area as much as 60 percent (in average episodic type samples) of the fine paniculate mass
(PM2.5) could be noncarbonate carbon. That's a pretty large percentage of the total fine material.
That finding then resulted in a lot of further work. In addition to pesticide-related work discussed
by Lewis just now, relatively little molecular work has been done. Regrettably, we're still hung up
on separating elemental and organic carbon, when initially this separation was primarily an
exploration to see whether or not we should really bother looking at the molecular entities. It has
turned out that what we should do is to determine the amount of carbonaceous material that
absorbs light (i.e. soot which is a mix of elemental carbon and dark organics). Because the
molecular composition is so complex, the remaining organics have not received the attention they
deserve, not only to identify toxicants, but also to establish their hygroscopic influence on light
scattering.
The importance of noncarbonate carbon material in the atmosphere I will illustrate today relates
to the rural western atmosphere. Figure 1 shows fine sulfur mass per cubic meter versus paniculate
129
-------
organic carbon as we've determined it. The diameter of the circle is indicative of the associated light
scattering. As you can see, a great deal of data indicate that scattering increases both as a function
of paniculate organic carbon (POC) and sulfur.
Figure 2 illustrates that a large fraction of the POC is in the light scattering range. The only way
to get enough material as a function of particle size is with a MOUDI, shown in Figure 3 in one of
its latest versions, in part developed with EPRI-sponsored research.
Since I'm billed to talk about EPRI-sponsored research, Figure 4 illustrates that we consider all
of our research to be collaborative or cooperative. Somewhere in the scheme of things, we try to
keep to track of all contributors to a given effort. This figure illustrates sponsorship for the
sequential development of the MOUDI. I apologize for someone in our department who drew a box
for EPRI in somewhat excessive proportions. The University of Minnesota hardly deserves the
smallest box.
But this figure does illustrate that we like to perceive our sponsorships as important
contributions to a larger good. We get into the act somewhere in the scheme of things and support
part of the work that is collaboratively funded by a whole host of sponsors, and everyone deserves
a great deal of credit for the overall advances that result. We certainly are not in the business of
taking more credit than is proper.
[I'd like to say that EPA funded the MOUDI development from its inception. EPA supported
development of the micro orifice impactor by a Ph.D. student from about 1978 to 1984. (R.K.
Stevens)]
With this development one gets information as a function of particle size in this rural western
atmosphere. Figure 5 is a plot of the size ranges obtained in conjunction with the Knoll impactor
(which accounts for very large particles), the SCISAS (which is essentially a high volume sampler),
and the MOUDI. The SCISAS is a 120 liter per minute sampler, with size separation at 2.5 microns
using a cyclone, so you get two size cuts. The samples taken with the MOUDI are less than 2.5
microns. The after-filter provides the smallest particle fraction. So the gravimetric mass is largely
in the light scattering or larger size ranges.
Figure 6 shows analogous data for the sulfur content. Here we see that a substantial fraction
of the sulfur content is actually below the light scattering range.
[Before you go to the next figure, Peter, how did you derive size distributions from the MOUDI
data? Did you use a Walter John model or did you use a Dzubay model? (R.K. Stevens)]
[The expert is two seats next to you.]
[Those data are not inverted with either of those models. (McMurry)]
[So those are not the true size distributions? (Stevens)]
[Neither are Walter John's or Dzubay. (McMurry)]
130
-------
[That's not the true size distribution? (Stevens)]
[You can't measure the true size distribution. (McMurry)]
[You can with the algorithm and the Walter John and Dzubay work. They intercompare data
and they compare very well. Optical microscopy seems to confirm what they've been doing.
(Stevens)]
[We have inverted several hundred size distributions using Walter John's, using one that was
developed by Earl Knutson, and using one that was developed by John Seinfeld, and they would
all fall within that histogram that you see there. None of them would be exactly the same. They
would all stand up and defend their particular approach as being the more correct one. I don't think
we should argue beyond that, but I don't agree that it's been established that any one inversion
approach is the correct one. (McMurry)]
In any case, the point is there's a lot of size structure in the PM2.5 that becomes very important
to the effect of concern, whether it's inhalation or optics. One really needs to know something about
the size distribution, not only of the mass as a whole, but also as a function of the chemical
component. Figure 7 shows the same thing for organic and elemental carbon, where the organic
carbon size distribution is something like the sulfur size distribution. That was for the summer.
Figure 8 shows such data for the winter comparing sulfate and ammonium. Figure 9 shows organic
carbon versus elemental carbon. In the case of organic, we again show the problem that Pete
McMurry addressed earlier this morning with collecting gas phase organics on the quartz after-filter.
If you take the size distribution data and try to determine how much of the sulfate, etc. is in the •
scattering range, you come up with the numbers shown in Figure 10. Note that not all of the sulfate
is in the size scattering range, and a lot of the organic carbon is also in the light scattering range..
Figure 11 shows a great deal of data obtained in an 11-station WRAQS network. We sampled
for eight daytime hours in the rural western United States. These data are from Harlowton, MT.
Analyses were done by Warren White. He tried to account for the total reconstructed mass and the
relationship to the paniculate organic carbon, which was not corrected for any blanks. What he
found repeatedly is a good correlation with a positive intercept. We have analogous data for all
eleven locations.
[Can you explain those ordinates again, please?}
The point is that the data themselves indicated a positive artifact for oversampling the organic
carbon. For that reason, then, we started to devise a way of trying to account for what might be
adsorbed on a quartz filter as discussed by Bruce Appel and other speakers this morning. Figure
12 is another rendition of the apparatus developed by Fitz as explained by Bruce. Our initial
sampling in the Los Angeles area indicated that we might be able to correct for what was actually
an irreversible adsorption on the quartz. This adsorbed carbon can't be removed by heating this
131
-------
filter. The only way you get it back is by combusting the material. Several samples in series behind
the front filter indicated that the amount of carbon adsorbed per unit area on the activated quartz
was roughly constant. This morning, we were shown that this may not necessarily be true,
especially for Portland. I think a lot of it has to do with where you are doing your sampling. It
seemed to be true in LA. that the amount of carbon per square centimeter of filter surface remained
relatively constant and, hence, the idea that you could just sample on the back filter, get the amount
per square centimeter and then subtract that from the front filter. Being quite uncomfortable about
that, we came up with the idea of actually doing comparative experiments in which quartz
absorbable material was removed with denuders so that the denuded front filter would represent
the ambient paniculate organic carbon. The question of volatization of POC was addressed by also
analyzing the back filter in the denuder system.
This system, having been evaluated in the laboratory and the Los Angeles atmosphere, was
then also evaluated in the rural western United States, and I'll show you some data for both the
summer and winter seasons. By the way, this is a photo (no hard copy available) of what it looks
like. Here are the plates with glass fiber on them. This is just a head-on view of the same thing (no
hard copy available). Figure 13 illustrates some of the equalities. The mechanisms of how organics
might be distributed in the sampling apparatus are so complicated, what we hypothesize are only
plausible guesses. We are dealing with many different kinds of organic molecules, whose mix
changes from time to time (hour, day, season), and with many different kinds of physical
multicomponent phenomena that none of us have any data on. Only the kind of work that Peter
McMurry is about to do is going to uncover knowledge about them. To find a basis for comparing
data obtained with the various sampling arrays, I set up expectations for equivalence and then
analyzed their deviations for several episodic sets of 12 or 24-hour data. For the sake of processing
the data, I declared that the front plus back filter observations from the denuded system should
equal the front minus back filter observations from the undenuded system, as illustrated by the
equations in Figure 13.
Figure 14 shows two sets of data, one from Meteor Crater and one from Hillside, Arizona for
samples taken midyear. Let's compare the undenuded front filter, the undenuded front minus back
and the undenuded front plus back data. Obvious are: 1) all these move up and down more or
less together, and 2) the undenuded front minus back and the denuded front plus back seem to
be almost identical, while the undenuded front is always greater.
Figure 15 shows analogous plots for winter at Hillside. The undenuded front is always higher
than both the undenuded front minus back and the denuded front plus back, and again, the latter
two are very difficult to separate. If you now look at Figure 16, the deviations of these two are a test
of 'equivalence'. With respect to a ±.15 nQ/m3 band, which is roughly 10 to 20 percent of the
132
-------
median concentration found, you can see that most of the deviation for this particular sampling
period falls within an acceptable tolerance. The other feature, which is really interesting to me, is
that the deviations appear to be episodic in nature. Figure 17 shows an analogous plot of deviation
for the Hillside winter data. Here, we have a number of deviations that fall well outside what I would
consider being tolerable. A lot of data falls within a tolerable range. All the data again show an
apparent episodic behavior. Since these two channels are mechanistically really not equivalent, this
episodic behavior tells us something about the influence of changing organics and other materials
(like water) that are in the atmosphere on what is absorbed on the quartz fitter. Consistent with
comments by Huntzicker earlier, it could be instructive to study such deviations as a function of
variables, such as humidity. These episodically driven deviations indicate to us that we've really got
to understand the molecular nature of the organics we are sampling. I think rationalizing the merits
of various quartz, teflon, and denuder configurations based on what is known is nothing but 'hocus
pocus". It won't be productive in getting a good handle on what the actual particular organic carbon
was in the atmosphere at the time of sampling. With the current sampling technology, we could be
off by large amounts.
With respect to variations on the current sampling technology, we have obtained a lot of data
that will permit relating the results from one to the other with generally tolerable uncertainty. Figure
18 shows a pretty good regression for undenuded front minus back for the midyear samples in
Arizona. Figure 19 shows you the same sort of thing for undenuded front POC versus denuded
front minus back. Figure 20 shows the undenuded front minus back versus the denuded front plus
back during winter at Hillside. Although the denuder data show a greater variance than the
undenuded data, data sets obtained with different sampling schemes can be made comparable.
The next point that I want to make concerns the other part of the problem-the analysis of the
sample brought to the laboratory. Two different methods have been used to get a great deal of data
in the U.S. (Figure 21). I've characterized these as TMO and TOR. The TOR (Thermal Optical
Reflectance) is the method that stems originally from the Oregon Graduate Center. The results I'm
going to show you come from the modification of that method as practiced at the Desert Research
Institute. The TMO (Thermal Mn02 Oxidation) method was developed by Kochy Fung and myself
at ERT. Figure 22 shows you in principle how the TOR method works. It provides a number of
thermographic signals, all of which could be recovered. The OC4 part of the signal represents a
correction that is subtracted from the higher temperature signal indicative of elemental carbon.
Because the correction represents carbonization of organics, it could be a source of over-reporting
the elemental carbon.
Figure 23 shows a typical thermogram for the TMO method, in which no carbonization of
organics is assumed. The question is if the no-carbon assumption is satisfied. Figure 24 shows
133
-------
the result of an experiment that was done as part of the 1988 shoot-out in Los Angeles. A subset
of samples were analyzed for nonvolatile and organic elemental carbon before and after extraction
with a strong solvent. The hypothesis was that if TMO carbonizes organics, the amount of elemental
carbon reported should be a function of the organic content. This test shows that the elemental
carbon content reported is the same before and after substantial reduction in the organic content.
It supports the claim (based on tests with many reference compounds) that there is no detectable
carbonization artifact with the TMO procedure.
[Peter, one important comment. In our attempts to do a similar exercise, when you inspect the
extract and look at it under Tyndal light or something like this, you see a fine distribution of some
of that fine particle carbon that gets suspended in the extract. So, it's very difficult not to remove
some of the elemental carbon in the extraction process. For that reason, I gave up trying to do that
as a means of comparing oxidation versus extraction procedures.]
I see. In this case, this experiment supported the argument that the carbonization wasn't taking
place, but let's continue looking at results given by TMO and TOR for various kinds of samples. In
Figure 25 the TMO data are in light grey and the TOR data are in dark grey. The organic carbon
as reported by TMO is always larger than by TOR for wood smoke samples. For diesel samples,
the two give essentially the same results. The wood smoke samples also influenced the amount of
elemental carbon reported, whereas in diesel smoke, these differences are within experimental error.
Thus, the nature of the organic determines how these methods have reported the organic-element
split. Figure 26 does an analogous comparison for 60 samples taken in Page, Arizona, during
winter, where a large proportion of the carbonaceous material probably comes from wood smoke.
We see a substantial difference in the split between organic and elemental carbon, although total
carbon is similar.
[You only showed error bars on that graph. What did the paired sample look like?]
I don't have those data with me, but they should be studied.
[Each sample is different, I think.]
Figure 27 simply summarizes what I've been saying: 1) collection artifacts frequently stem from
gas phase organic absorption, which has to be somehow accounted for if you work with quartz.
Undersampling could occur as a result of revolatilization, or what's called the negative artifact.
Experiments are going on with apparatus analogous to the Fitz apparatus, where the absorbents,
instead of being quartz, are activated carbon-impregnated filters. The activated carbon is analyzed
by combustion. The filter material has a very small standard deviation I'm told, with respect to its
adsorbent carbon content, and so the differences due to sampling organics can be quantified. In
a setting like the western U.S., the results are that roughly half of the paniculate organic carbon
134
-------
reported, after you account for the positive artifact, is lost from the filter during sampling. I haven't
actually seen those data; this is what's been reported.
[Is that Delbert Eatough's work?]
Yes, that's going on right now.
[That's highly variable, Peter. I mean, he has data all over the place.]
Have you seen the most recent set?
[The most recent set I saw was about four or five months ago.]
I see. I haven't seen it. I'm concerned about it. But I wanted to mention that if we had a good
adsorber, those kinds of experiments could be done with our kind of denuder sampling apparati.
We should be able to get at it. This question is still up in the air, so to speak.
I think these are conclusions that we all agree to.
[What do you mean by "irreversibly adsorbed?']
By that I mean that the only way you can get it back from the quartz is by combustion. I don't
know how much you would get back by extraction, but you can't get it back simply by volatilization.
[At 125°C?]
Right, by heating it. Returning now to the second bullet on Figure 27, if you do sampling
without denuders, you may have this kind of error in a recorded POC. The negative artifact that we
don't know how to account for yet might be an offset. Anyway, the current sampling technique that
I would recommend is utilizing the denuder filter pack technique.
Third and fourth bullets. Using either the quartz quartz filter systems or the parallel quartz
quartz Teflon systems will also get you data that is exchangeable with the denuder techniques.
Last bullet. The analytical differences that we get seem to result from logical principles, and the
operational definitions of organic carbon versus elemental carbon. What I won't show you right now
is a scheme that DRI has developed where all the individual thermogram peaks can be quantified
in their TOR apparatus. We can then find a way of making the TMO OC-EC split equivalent to the
TOR OC-EC split in the data that we're now collecting. The other part of our interest really relates
to understanding the origins and the fate of the carbonaceous material in the atmosphere, so that
we can overcome the prediction problem. We would like to be able to simulate what would happen
to the organics in the atmosphere as a function of various kinds of emissions management options.
At the heart of that, you have to have some sort of prognosticating capability, which is a
comprehensive modeling system. (It has many, many applications.)
What we are in the process of sponsoring right now are: 1) more work to characterize the
nature of the particles themselves so that we can understand how all of the components of the
particles are mixed in the aerosol, whether they are separate particles, whether they are internally
mixed, as a function of particle size; 2) follow-on work to what has been sponsored by others at Cal
135
-------
Tech. The earlier work has characterized the primary organic emissions at a large number of
sources in the Southern California Air Basin together with the simultaneous characterization of those
types of organics in the air itself. EPRI's sponsorship is now for relating these two data sets via
exploring various receptor modeling techniques so that we can account for the emissions. If you're
going to run models, you've got to have emissions. To get emissions data in the classical way on
those kinds of materials is probably impossible, so you probably have to do it through some sort
of forensic technique. So, we're investigating what forensic techniques might be the most suitable
for doing that, using this Los Angeles Cat Tech data set; and 3) work by John Seinfeld's group to
write equations that could be incorporated into the existing modeling technology processes for
secondary organic particle formation and decay.
We have to account for the secondary organic sources that actually form in the atmosphere
through atmospheric reactions. Products evolve and decay in some sequence, which really relates
to the interests that Steve Japar was talking about as well. It's very important to account for the
secondary organics in the modeling scheme. A great deal of information exists in the literature by
at least functional group or reactivity class, which is to be coded up for modeling applications.
[Thank you very much, Peter. Thank you for getting through in just about the time you were
allotted. We'll take our break for lunch now. We're running just 10 minutes late. So let's try to be
back by 1:40 to resume our afternoon discussions. There's a cafeteria right upstairs.]
136
-------
B. Graphics
137
-------
FIGURE 1.
139
-------
Light Scattering by Particles
(Glen Canyon, AZ, Winter 1986)
Component
Sulfate
Ammonium
Organic Carbon
Elemental Carfcjon
% of Fine Mass in
hsp Range
60
60
40
50
FIGURE 2.
140
-------
FIGURE 3.
141
-------
Evolution of MOUDI
ro
1*1988
MSP
Commercialized
9/85-12/86
EPRI
Removable substrate
assemblies
Design of nozzle
geometries & welded ss
nozzle plates
7/83-8/84 . DOE
First vertical MOUDI
calibration
'&A yf*wk%wfyrfifynjesflw^&::t.^^s'*^*,^^r^i,#-^^ 4* v- •''•••"•&*.•'*•> •.'».->' »-•? "?-,~*f''' >
>> ,-'/ '^-^J
I
I
FIGURE 4.
-------
AVERAGE GRAVIMETRIC SIZE DISTRIBUTIONS
September 9 - October 9, 1985
Grand Canyon, AZ
dM/d log D (Mg/m3)
4
3
2
0
|
—
—
— —
|
i
i
1
i
i
i
•
_ 1
i !
I iih i
^•M
~~
1 1 1
1
i
i
i
[
i
i
i
ii
• i
.
i
VIUULJI
SCISAS
MDI
iNni
1 1 1 1 1 III
0.01
0.1
1.0
Diameter, \i
468
10
20 40 6080
100
58278.06
FIGURE 5.
-------
AVERAGE SULFUR SIZE DISTRIBUTION,
September 9 - October 9, 1985,
Grand Canyon, AZ
dM/d log D (Mg/m3)
.5
.4
.3
.2
.1
0.01
MOUDI
SCISAS
0.1 1.0
Diameter,
10
58278.04
FIGURE 6.
-------
AVERAGE CARBON SIZE DISTRIBUTION
September 9 - October 9, 1985,
Daytime (0800-2000 hrs) Grand Canyon, AZ
dM/d log D (fg/m3)
.7
Organic carbon
.6
.5
.4
.3
.2
.1
] Elemental carbon
rasBBasas!apBtna>IIBra^^
SCISAS
MOUDI
I-
li
0.01
0.1 1.0
Diameter,
10
58278.05
FIGURE 7.
-------
C.fit-
w
%',
AvfftAftf 6* *hb *»«'*
ca.
l.OC
JC.Gt
= S 5Cj
g !
I I
5 C.*£r
0.55-i—
e.c!
i I
s!ua *»»• :(<»:.• ;•.-.,
FIGURE 8.
146
-------
tvrutt f AU. 3*"
•!
0-—
C.01
C.Sf
£
8
o.oi
0.10 i.OC
06. US
C.1D . «.02
Co, a*>
iO.CC
:c.o:
rr|ir.i: ir.t fl.aenni e»r>iea 4litr
r.j.u.-rrf wi:>! th« yif-.'S; 4urt8( tfic Vinttr
Spec'.il E;uJ«.
FIGURE 9.
147
-------
Light Scattering by Particles
(Glen Canyon, AZ, Winter 1986)
Component
Sulfate
Ammonium
Organic Carbon
Elemental Carbon
% of Fine Mass in
bsp Range
60
60
40
50
FIGURE 10.
148
-------
"- —— •--"-" — ,....^.^ -^ ^^.,-|;;Tfrr• -,- j
Hrlovytorv
>
FIGURE 11.
149
-------
Design for Sampling
Particulate Organic Carbon
Needle Valve
Quartz Filter Strips
(15 at 3mm spacing)
Needle Valve
To Vacuum
Rotameter
To Vacuum
Aluminum
Denuder
Housing
Front
Back
Quartz Filters
Needle Valve
jo Vacuum
Rotameter
FIGURE 12.
-------
ATMOSPHERIC SCIENCES
Atmospheric Organics
(12) Total* atmospheric Organics = FUo+ BUo
(13) PM-2.5 Organics = FUo - BUo
(14) PM-2.5 Organic = FDo + BDo
(15) PM-2.5 deviation = (13) - (14)
Where: FUo= Undenuded front filter BUo = Undenuded back filter
Ul
FDo = Denuded front filter BDo = Denuded back filter
("o" denotes organic sampler)
* "Total" = particulate and quartz absorbable gaseous phases.
EPRI
FIGURE 13.
-------
ATMOSPHERIC SCIENCES
Organic Aerosol Carbon Concentrations
Via Three Sampling Methods On Quartz Filters
(Mid-1987, Meteor Crater, Hillside, AZ)
,UF
o
81 99 129 159 172 176 180 183 185 188 190 193 195 199 203 208 222 226 230
Julian Date, 1987
Underiuded o U(F-B) = Undenuded front
front filter minus back filter
D(F+B) SB Denuded front
plus back filter
EPRI
FIGURE 14.
-------
ATMOSPHERIC SCIENCES
Organic Aerosc! Carbon Concentrations
Via Three Sampling Methods On Quartz Filters
(Winter 1988-89, Hillside, AZ)
UF
/;
.0 1
336 339 342 34S 348 351 354 357 360 363 I • 4 7 10 13 16 19 22
Julian Date, 1988-89
EPRI
FIGURE 15.
-------
ATMOSPHERIC SCIENCES
Deviations Between Two Methods in 24-hr Values Due to
Particulate Organic Carbon Sampling
ng/m
O.'T -r
(Mid-1987, Rural Arizona)
-0.3 J-
Julian Date, 1987
233
EPRI
FIGURE 16.
-------
ATMOSPHERIC SCIENCES
Deviations Between Denuder and Filter Pack Sampling Methods
in Diurnal POC Concentrations
ug/m
0.8 V*
0.6 «-
0.4 --
0.2 --
(Winter 1988-89, Hillside, AZ)
-0,2
-0.4
-0.6 JL
336
Julian Date, 1988-89
EPRI
FIGURE 17.
-------
ATMOSPHERIC SCIENCES
Oversampling of Paniculate Organic Carbon
With Undenuded Quartz Filters
24-Hr Concentrations
Undenuded
front POC
qig/m3)
(Mid-1987, Rural, AZ,n=38)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Undenuded front minus back POC
1.1
1.2 1.3
EPRI
FIGURE 18.
-------
ATMOSPHrmC SCIENCES
Oversampling of Particulate Organic Carbon
V WithTUndenuded Quartz Filters
:, < 24-Hr Concentrations /
'
"~- , v Vs '-
' -; . -/'-.
d
1.8
1.6,
1.4 .-
, , 1.2
'•' *'*•
' '> ;'
, , :; ' i
0.6
,0.6 -
0,4
V K
t / f
j, *-°-2-*j
/ '*,' -
'''< ' > n
o.i
id1 987'
• *
* * «,
A
Regression Line
0.2 0.3 0.4 O.S 0.6 0.7 0.8 0,9, 1
t(ftdenuded front plus back POC (ng/m3)
1.1 1.2 1.3
EPRI
"FIQURE"19."
-------
ATMOSPHERIC SCIENCES
Relative Performance of Denuder and Filterpack POC Samplers
Undenuded (Winter 1988-89, Hillside, AZ)
front minus back POC v '
1.4
1.2 --
0.0 -•
0.6 -•
0.4 -•
0.2 -•
;•*
,«••
•"
.»•
.>•••'
.......
"
..*'
,•"'"
,.,
Xi
...«'
t,i«'
1:1 Band
0.00
0.20
0.40
060
0.60
1.00
120
Denuded front plus back POC (jig/m3)
1.40
.1.60
EPRI
FIGURE 20.
-------
ANALYTICAL METHOD
Two Thermal Analysis Methods at Two
Different Laboratories
1. Thermal MnO2 Oxidation (TMO)
2. Thermal Optical Reflectance (TOR)
FIGURE 21.
-------
o>
o
- -4-
He atmosphere
Temperature
!Li
Lighter
Reflectance
IffttHlitttKtiHilltiHillfHflTTK
He/2% Oa atmosphere
THERMAL OPTICAL REFLECTANCE METHOD
Temperature (°C)
900
800
700
600
500
400
300
200
100
Reflectance back
ir to original value
0 200
400 600 800 1000 1200 1400 1600 1800
Time (s)
36851.04
FIGURE 22.
-------
CD
THERMAL Mn02 OXIDATION METHOD
Temperature (°C)
900
800
700
600
500
400
300
200
100
FID
Output
VOC
He atmosphere
S ik
ARC
ROC \
0 100 200 300 400 500 600 700 800
Time (s) aeasios
FIGURE 23.
-------
/c>r Organic janch Nonvolatile Carbons
iV^:-i!f Selected Samples After
&'"*':.,v; -;• :V Solvent Extraction !, >
' - ° %
FIGURE 24.
162
-------
PERFORMANCE OF TMO & TOR ON SOURCE SAMPLES
/K|/cm2
300
250
200
150
- 100
8
50
0
2 wood smoke samples
228 213 225
1 diesel sample
TMO
TOR
101.3 106.7
55.3
2.7
86.7 92.4
TC
oc
EC
TC
OC
EC
Analysis of wood smoke and diesel exhaust with the thermal optical
reflectance (TOR) and thermal manganese dioxide oxidation (TMO) methods.
36851.01
RQURE 25.
-------
PERFORMANCE OF TMO & TOR AMBIENT SAMPLES
40
30
20
10
60 samples from Page*a
-\ h
6 samples from CSMCSb
30.6
31.9
TMO
TOR
6.6
9.5
6.0 6.1
0.6
3.4
24.9
22.2
9.7
5.8
0
TC OC EC TC OC EC
Analysis of ambient carbonaceous aerosols with the thermal manganese
dioxide oxidation (TMO) and thermal optical reflectance (TOR) methods.
*a Winter 86-87, page AZ; b summer '84, Glendora CA.
36851.02
FIGURE 26.
-------
SUMMARY
Significant Amount of Vapor-phase Organic
Carbon is irreversibly adsorbed by Quartz
Filter when Sampling Ambient air in the
Rural West
Samples Collected Without Denuder May
have OC Levels of 45% Higher than
Samples with Denuder
- * Sampling Artifact is largely Eliminated by
Using Denuder Constructed of the Filter
medium
Correction by Subtracting the OC on a Back
Filter seems Successful
Analytical Differences Appear to Result
from Methodological Principles and
Operational Definitions for OC/EC
FIGURE 27.
-------
IX. GENERAL DISCUSSION
WILLIAM WILSON: We want to have a general discussion of things that might be done and then
try to get into more definitive prioritization of things later on, perhaps after our afternoon break. First,
I'd like to give an opportunity to people who didn't speak this morning to make any comments they
would like to make in general, any questions they would like to address to the speakers, and any
recommendations they might want to put forward as to what sort of work needs to be done. Frank,
do you want to...?
FRANK BINKOWSKI: I was saying to William privately that I came with an operational definition
that organic aerosols are an artifact for the discussion of aerosol scientists. I was amazed at the
lack of consensus of what's going on. I'm not an aerosol specialist. I'm a meteorological air quality
type modeler. What I heard this morning was very strong disagreement about the existence of
artifacts and how to account for them. And one of the things that Peter McMurry mentioned,
perhaps we need a universally agreed upon definition of what is organic carbon and what is
elemental carbon in an analytical sense. Coming as an outsider and not an analytic chemist, but
it seemed to me that was a point of disputation. Some sort of reconciliation or reconsideration of
all the experiments-Jim, you were expressing dissatisfaction with someone's work that said there
was a large amount of loss, evaporative loss, and Bob Wilson was saying, and I think they were
quoting against the same data set, that there was a loss in the semivolatiles. In some of the work
that was presented here, it seemed to indicated that there's a possibility that you came up with that
there might not be a negative artifact. I heard you in your summary statement say that.
BRUCE APPEL: What I was alluding to was the fact that there is an apparent discrepancy. On
the one hand, if you are dealing with work involving specific analytes, specific compounds, that you
can follow and use some relatively exotic techniques like, for example, the method referred to as the
denuder difference method, which allows for specific compounds, in this case PAH compounds, to
be measured in a way that allows you to find with some confidence what is the concentration of
particle phase pyrene, for example, one particular four-ring PAH, the concentration in the particle
phase at the time of sampling, the concentration in the gas phase at the time of sampling, and the
degree of volatilization that occurs subsequent to collection. In this particular case, there were data,
with the question of how valid it is, that suggest that for the vast majority of that, pyrene is volatilized
subsequent to collection.
That's one situation. On the other extreme, rather than dealing with specific compounds, we'll
be dealing with carbon, organic carbon. It has been very difficult to demonstrate that it also is
subject to a negative artifact. We all agree conceptually that it should be, for all the reasons we've
heard. Yet, to experimentally demonstrate that in a relatively unambiguous way has been very
167
-------
difficult. So I have put the question to people: do you have unambiguous evidence of a negative
sampling artifact based on carbon, organic carbon, as an endpoint? I don't think I've heard an
unqualified "yes" as a response to that.
FRANK BINKOWSKI: The argument that Peter McMurry presented was that, in terms of
equilibrium conditions, there should be. Because you said very clearly that there's a change of
equilibrium, whether the aerosol was or was not in equilibrium with the gas phase.
PETER MCMURRY: It really depends on the gas particle partitioning ratio in our view. We have
done some experiments to look for evaporative losses to see if we can find evidence. We have not
found evidence. Our theory tells us that we might expect to have evaporative losses if the gas
particle partition ratio were large. We don't see large artifacts; we conclude that perhaps it's small.
FRANK BINKOWSKI: I have a very vested interest in this. I'm a regional modeler and we're
trying to model aerosol processes and visibility on a regional scale, and the next big area of
application is going to be the West, where concentrations are low, and presumably we're going to
be in this situation, which is unfavorable.
PETER MCMURRY: To put your comment a little bit in perspective, I don't know the numbers,
but if you look historically where the effort has gone, we've dealt with acid rain problems and sulfur
dioxide conversion problems and so on, and that work has been going on since probably the late
1960s, is that not right, William? Then in the 70's and 80's, a large amount of effort and money were
spent on those types of problems. In proportion, how much money has been spent historically on
the organic fraction? A very small amount, I believe. I think it's a much more complicated problem,
so I don't think we should be too surprised that we're where we are.
PETER MUELLER: But, you said you were struck by the disagreement... I think perhaps what
we should start with is where there is strong agreement and then analyze and identify where the
deviations are among us. I don't know that there's really a big dispersal of opinion sitting around
this table, but there are different recognitions of the problem represented by the individuals here.
That, to the outsider, can sound like a disagreement between them. I think they're actually
supplemental.
FRANK BINKOWSKI: You're saying it's the blind man and the elephant. Each of you has got
a different handle on the problem.
PETER MUELLER: We're just expressing a different handle of the problem here today. Right.
My perception of where the agreement is that noncarbonate particulate carbon is an important part
of the atmospheric fine mass. Secondly, we agree that there are two major components to that
carbonaceous material. It's elemental and it's organic. In that element-organic split, there is some
disagreement, and the disagreement is complicated. One basis of the disagreement is the analytical
168
-------
methodology, which is only operationally defined, because there is no standard reference material
against which to check it out.
Now, the operational definition of the differences between the two is not chemical; it is light
absorption. That's where a lot of fuzziness in our understanding comes in which sounds like
disagreement. The fuzziness in our understanding is just what part of the carbonaceous material
causes light absorption? If it's elemental carbon only, then we definitely disagree how much mass
is out there, based on the methods. There are two schools of thought as to which method gives
you the right amount for elemental carbon. If it's a matter of how much light-absorbing carbon is
out there, elemental plus black organic material, like stuff we have everywhere-black dyes-or blue
dyes are all absorbing organics, so there's a lot of organic material that is not elemental carbon that
absorbs light. Now the question is whether somehow the TOR method isn't actually measuring the
absorbing carbon as compared to the elemental carbon, and simply reporting it as elemental
carbon. That's where the fuzziness still exists. I don't think it's a really major hang-up anymore.
It's going to get resolved in the next year or two with work now underway.
The difficult but important problem now is what is the organic material that's actually in the
atmosphere that we're all sampling and reporting as total paniculate organic carbon. We're trying
to get at the paniculate carbon by removing it from the gas phase with porous filters. The minute
we're doing that we're changing things that simultaneously exist in both the paniculate and gas
phases. We don't know how much damage we're doing to the paniculate material by removing it
from the aerosol. What you heard is our struggle with that problem, not necessarily a difference of
opinion, about just exactly what the nature of that problem is. We've pretty much agreed that it's
anywhere from 0 to twice the amount of paniculate material that we're reporting. It is not a factor
of 10 or 100 or 1,000. It's somewhere in the order of a factor of 2, is our uncertainty on that
question. On that, I think we all agree.
BRUCE APPEL: The difference for elemental carbon/organic carbon split is less than that, as
far as the discrepancy between methodologies. On that point, I think you could say, relative to
some of the uncertainties we have addressed today in relation to sampling, there is relative
agreement. With regards to agreement, just to pursue this for a moment, I think we all agree that
the positive artifact exists. The positive artifact due to the sorption on a filter medium, particularly
quartz fiber, is real. I heard no disagreement on that point. Where we quibble in relation to the
positive artifact is whether or not it exists in relation to retention on the panicle phase material. Only
one speaker, myself, alluded to the possibility that this was a significant factor, and I presented
some data that were consistent with that hypothesis, but certainly does not prove it.
PETER MUELLER: We also are quibbling about how to correct for that.
BRUCE APPEL: Right.
169
-------
PETER MUELLER: I don't think we have any strong differences of opinion about how to do it.
BRUCE APPEL: Right. Comparing the two dual filter approaches, in some conditions they
disagree by a substantial factor, perhaps a factor of 2, but at higher face velocities they tend to
come together. Again, that may not be a major source of disagreement. Jim might feel differently.
I think there continues to be a major area of, and I wouldn't say disagreement, I would say genuine
lack of understanding, and a real need for clarification, is on this so-called negative artifact. In
relation to paniculate carbon, does it really exist? We're hearing some suggestions that support a
factor of 2. Such conclusions are so much driven by the method that was used to reach that
conclusion, and nobody is more keenly aware of the limitations of those devices than I am. I
struggled very hard to try to develop exactly that kind of technique. It's very difficult to do, and we
clearly don't know that magnitude. It can be demonstrated for individual compounds, as I said.
FRANK BINKOWSKI: I asked you during the break about the idea of determining partition
function theoretically as to Pankow's work. I think that one fundamental thing that needs to be done
is to pursue that triangulate multicomponent equilibrium approach.
JAMES HUNTZICKER: The problem there is that the physical chemistry has just not worked out.
The calculations that we made were based on a 1948 theory by Terrell Hill, which is just an
adaptation of the BET adsorption theory which has real serious limitations. Among other things, the
experimental tests of that theory are somewhat equivocal. They have never been applied to
atmospheric situations, as far as I am aware. There is a lot of fundamental understanding that just
doesn't exist right now.
FRANK BINKOWSKI: From my point of view as a modeler, I need to be able to say I have
generated so much gas phase concentration of this suite of organic compounds, which are
generally 'lumped". I'd like to know how much of that is going into the paniculate phase.
JAMES HUNTZICKER: In the work of Pankow and Bidleman and Yamasaki, their physical model
is a very simple model. It's a simplified Langmuir isotherm-type of model, which is valid only when
the adsorbed concentrations are very, very much less than a monolayer of totally absorbed material.
However, given that water is always an important adsorbing species, say, at 50 percent relative
humidity, you can almost be sure that there will be a monolayer of something adsorbed onto the
particle. Current theoretical approaches do not include the effect of water.
FRANK BINKOWSKI: Is that consistent with what Peter McMurry showed about L.A. data? I
guess those are very high relative humidities when we've got the separation of the two aerosols.
PETER MCMURRY: That was release of water. When you cook the things, the relative humidity
goes from perhaps 60 percent to 70 percent down to 20 percent in the sampler, the particles shrink..
That didn't represent a change in the organic content.
170
-------
FRANK BINKOWSKI: No, what he was saying about the 50 percent relative humidity, you're
going to get a monolayer.
JAMES HUNTZICKER: That doesn't represent much water.
PETER MCMURRY: When you look at water uptake by particles, though, it's not consistent with
adsorption in monolayers. It's so much more consistent with dissolution of water.
JAMES HUNTZICKER: Precisely.
??: OK, so that's why you said deliquescent?
JAMES HUNTZICKER: Yes.
PETER MCMURRY: Typically, a particle at 50 percent humidity is 6 percent to 10 percent bigger
than a particle at 0 percent humidity.
??: Which is a lot of water.
WILLIAM WILSON: I want to give a chance for some of the other people who didn't talk to
comment or question. Mike?
MICHAEL BARNES: Does anybody here have any information on refractive indices of organic
aerosol? If there they are important in visibility that implies (?) to know something about the
refractive indices. Or if you can dismiss it just because of its size.
PETER MCMURRY: To my knowledge, there is a need for better measurements. People have
estimated refractive indices based on likely organic composition of the particles, but I think that...
MICHAEL BARNES: Do you know if they were just taking smog chamber samples of toluene
and NOX?
PETER MCMURRY: I think you can do it with smog chamber materials.
MICHAEL BARNES: I've never seen them, I've looked for them.
PETER MCMURRY: I haven't seen it done with smog chamber materials, but Hanel collected
bulk samples and measured various physical chemical properties, including refractive indices. I
think our work has shown us, however, that there are substantial differences in refractive indices
among particles of a given size. For example, Suzanne Hering and I did an experiment where we
took monodispersed atmospheric particles and fed them into an optical counter with given collecting
optics. We found that there were typically two distinctly different types of particles, some of which
scattered a lot of light and produced a big pulse, others which scattered a small amount of light and
produced smaller pulses. That's consistent with the TDMA data that I showed this morning that
indicated that there are at least two types of particles, two distinctly different types of microscopic
particles with different hydroscopic properties. We are hoping to be able to do some measurements
of refractive index with those particles. That's one of our objectives for the next two or three years.
I think we need to know more about those refractive indices.
FRANK BINKOWSKI: You'll be able to get the real and the imaginary parts?
171
-------
PETER MCMURRY: We're hoping to do that, yes.
STEVEN JAPAR: The problem with doing it with smog chamber particles is that it might not be
representative of what's going on in the atmosphere anyway. Plus, if you do it with no base aerosol
in there, your particle is going to be one that forms through homogeneous nucleation, with
subsequent condensation on it, whereas in the atmosphere that's not going to happen. These
things are going to condense on something that's already there. If you did the measurement for
a smog chamber aerosol, you might get a number, but how applicable that is going to be to the real
world is open to question.
PETER MCMURRY: The electron microscope studies that we have been doing are very, very
interesting because you find that if you look at particles of a given size, you find that a certain
fraction appear to be mostly carbon-containing and others contain mainly sulfur and carbon,
according to what we can measure. Typically, what happens is that the particles that are primarily
sulfurous in composition have a little carbon-rich core right at the center. The carbon seems to be
very abundant and seems to be present in most particles, not always mixed homogeneously.
MICHAEL BARNES: Peter, what are you funding John Seinfeld to do? Are you doing modeling
or experimental work?
PETER MUELLER: Modeling. They have done the experimental work with other prior
sponsorship.
STEVEN JAPAR: Through the CRC, Coordinating Research Council, which is funded half by
the Motor Vehicle Manufacturers' Association, and half by the American Petroleum Institute.
PETER MUELLER: Taking into account that experimental work now, as well as the work of
others like Qrosjean's that have been done over time, they will express the processes with equations
and then code them to enable simulation of secondary organic particle formation and decay. They
have a plan for doing this, for going about it.
STEVEN JAPAR: You have to realize that Seinfeld is the only one who's doing it. The approach
is a phenomenological one. They're taking the physical data that they've seen in these various
systems and trying to explain just the physical evolution of aerosol, work up an algorithm to express
the evolution and the particle size distribution. There is virtually no chemistry that goes into that,
except at the very front end of the system: compound A reacts with OH and it starts this whole
mechanism going. They are watching the aerosol size distribution evolve and trying to represent
it. If they can do that, that's a major advance, but it's still lacking the basis in the specific chemistry
and physics that's going on in the system.
FRANK BINKOWSKI: This is all coagulation and growth, then?
PETER MUELLER: It's molecular condensation, as well. Assuming certain kinds of reactive
species are created through hydrogen abstraction and things like that, and those reactions, to the
172
-------
extent that they're known, as a function of molecular group characteristics and molecular structure,
they're going to utilize whatever is published and apply it to the kinds of compounds that have been
observed in the atmosphere. Again, they have taken an engineering approach, justifying what they
know about one or two or three members of the characteristic group, using that information
statistically to extrapolate it to other members in their group, or reactivity class.
STEVEN JAPAR: What Seinfeld has essentially is the best available data on the evolution of
aerosols in simple hydrocarbon systems. He might pick out toluene or xylene, 1-octane, or
methylcyclohexane (I think is one that he has looked at). He takes that comprehensive data,
involving the physics and, to some degree, the chemistry that is involved in that aerosol generation,
as I understand it, and he is going to apply those results in a generalized approach that Peter
Mueller was describing. So, he's going to lump everything together into groups that can be
represented by the work that he's done on specific compounds.
PETER MUELLER: Or that others have done .. .
STEVEN JAPAR: Or that others have done, and use that to feed into the mechanism.
PETER MUELLER: It's an engineering approach that says that there is a lot of knowledge about
this reactivity out there, but it needs to be integrated, and this is a first effort at integrating it and
having some kind of a model that you can then use. Once we have it, we will better understand its
deficiencies.
WILLIAM WILSON: More, Mike?
MICHAEL BARNES: One thing we don't know, and it's going to kill Seinfeld, is that we don't
know what are the organic aerosols in the atmosphere. We don't know what the ratio of primary
to secondary aerosols are.
PETER MUELLER: We're working on that, too.
MICHAEL BARNES: We know it's a very low ratio for primary to secondary sulfate.
JAMES HUNTZICKER: We've got a lot of data now on secondary organic aerosol in Los
Angeles from our work in the SCAQS study.
PETER MUELLER: The idea is for creating these codes so that you have a simulation, and then
evaluating the performance of the simulation against the SCAQS data. The primaries are to be
taken care of in another part of the same project which Cass is heading, where he is integrating the
information that is being gathered as part of work supported largely by EPA, and I don't know who
all the other sponsors were.
??: Who besides EPA is supporting Cass?
??: It might be the Competitive Grants Program.
PETER MUELLER: Was it the California Resources Board? California Air Resources Board
allowed him to create special stainless steel organic samplers that he utilized to sample sources.
173
-------
Then, he characterized the primary organics in samples that he claims represent 80 percent of the
primary organic sources in Los Angeles, from a host of different industries. During the same period,
they also sampled the Los Angeles atmosphere at a number of locations and analyzed them for
similar compounds that they find in the sources. The organics are individually characterized using
GO mass spec coupled to extraction techniques. Now he is planning to experiment with three or
four different receptor modeling techniques to associate what's observed in the atmosphere versus
what he has observed as being emitted to the atmosphere. Presumably, if one or more of those
ideas work, they could then, by analyzing the atmosphere for those key primary organics, derive the
primary emission in future studies. That's the logic behind the ultimate application of that research
as I understand it.
BRUCE APPEL: By difference, can't you get secondary?
JAMES HUNTZICKER: That's been our approach.
PETER MUELLER: Or you get it through the modeling. What you do is you couple the receptor
modeling with the deterministic modeling.
JAMES HUNTZICKER: Cass is coming out with a paper in Atmospheric Environment that is
going to say that something like 20 percent of the primary organic aerosol is from the cooking of
meat.
LEN STOCKBURGER: Meaning in the home?
JAMES HUNTZICKER: Home or fast food restaurants. The paper looked reasonable.
LEN STOCKBURGER: Peter, of the mass that he identified, how much of it did he identify? 20
percent, 50 percent of the mass by specific compounds?
PETER MUELLER: They tried to account for all of it.
LEN STOCKBURGER: How close did they come?
PETER MUELLER: I don't know.
LEN STOCKBURGER: Usually, you get 25 percent, 30 percent.
PETER MUELLER: That's typically what he came up with when he tried to do this sort of work
for us because that's all that's extractable.
LEN STOCKBURGER: The stuff that we did a long time ago back in the Ohio River Valley
Study, back in the early 80's, we could extract almost all of it off. I mean, it was over 90 percent,
was recovered by extraction. We just couldn't analyze it through a GC or a GC mass spec.
??: That's right.
LEN STOCKBURGER: And that's making the derivatives and everything else.
STEVEN JAPAR: That's the case with vehicle exhaust. If we've identified 30 percent of it in
terms of compounds, that's probably a lot. I'm probably overestimating that significantly at this
point. There's just no way to get in and deal with all of the stuff, especially the polar material.
174
-------
JAMES HUNTZICKER: How much of it do you resolve chromatographically? How much is in
the hump?
STEVEN JAPAR: I don't know. It depends on how you add up the hump and what
assumptions you make. I think if you add up the humps and make some fairly wild estimates about
what's in there, you probably come close, but what does it mean? You've got a hump of polar
organic material for fraction 7, whatever it is in the solution scheme that you have, which accounts
for 30 percent of the mass, give or take, but you don't know what it is. You know it's polar organic.
In that mass of stuff, maybe you've identified two or three compounds.
PETER MUELLER: I'm glad you brought that question up. We did have that problem when we
tried to characterize the nature of organics that we find in the rural western United States. The data
that they've gathered in Los Angeles so far I haven't seen yet, so I'm not sure how well they've done
there. I think what he's hoping for is to find marker compounds that would take care of most of the
organics.
STEVEN JAPAR: That's always been the approach. If you can find one or two markers, at least
you can put a handle on attributing mass to a specific source, but it still doesn't help you
understand the intimate nature of the problem. It's good to be able to attribute things to sources
if you're reasonably sure about what you have, but that's only half the question, maybe it's less than
half the question in the overall scheme of things. From the point of view of regulation and control,
maybe it's 95 percent of the question. In terms of understanding the science and the implications
in atmospheric chemistry, that's only the very smallest first step.
JAMES HUNTZICKER: With regard to this question of primary and secondary organic
aerosol-we made 2-hour measurements of organic and elemental carbon during the SCAQS study
using our continuous carbon analyzer. Then we looked at the ratio of organic to elemental carbon
as a function of time of day. By looking at days which one could conclude the organic aerosol was
virtually all primary, we found the OC-EC ratio to be about 2. In smog episodes, very dramatic
increases of that ratio were observed, and it was possible to estimate the amount of secondary
organic aerosol as a function of time of day.
JAMES HUNTZICKER: During summer smog episodes, about half of the organic aerosol during
the afternoon peaks was secondary. However, the highest concentrations of aerosol organic carbon
that we saw were not in Claremont in the summer, but were in Long Beach in the winter.
Concentrations up to about 50 micrograms per cubic meter were measured.
??: Primary?
JAMES HUNTZICKER: Yes. We think it was all primary, which is not unreasonable because
of the proximity of the sampling site to a strong source area.
175
-------
EDWARD EDNEY: I certainly don't know very much about this area, but I have to say that, in
listening to the discussions today, I was very impressed by the fact that you people were able to
make something that doesn't sound very interesting, sound very interesting. That was just a
comment. What I would rather do, and also I'm glad to hear that someone is doing something
about looking at the formation of aerosol that Ford tried to do. Rather than showing that I don't
know a lot about this, I'm very interested in listening to those people who gave the presentations
today. I'm very interested in their research recommendations and priorities in that area. I think the
time is better spent than talking to me.
STEVE MCDOW: I have a question about the denuder difference method. If you think about
the gas particle distribution in terms of that it's in some kind of dynamic equilibrium and you think
about the filter as a large surface area, that the air gets drawn into, you can imagine that you're
distorting that equilibrium, and even when it goes through the denuder and you've got clean air
coming across the filter, and then you subtract a backup filter used to account for the volatilization
artifact, is the distortion of that equilibrium taken into account?
BRUCE APPEL: You're confused on one point. In the case of a denuder difference type of
experiment, you're not using an after-filter to correct for volatilization. Certainly, first of all, you're
going to have enhanced volatilization for the reasons we've heard, because you've got zero
concentration, essentially of gas phase materials. In order to account for any volatilization that might
exist, consider the analog in the nitrate system. The sum of what is collected on a filter plus a
backup nylon filter in that case represents particle nitrate, and you don't try to differentiate between
that which is collected on the filter from that which is collected on the downstream nylon filter,
because everyone recognizes that volatilization from the filter is enhanced. Likewise in the case of
carbonaceous paniculate, you are not going to be able to determine total particle phase
concentration unless you have an appropriate sorbent. And that's what I was alluding to for future
development. I was suggesting more work should be done to develop what I refer to as a true
paniculate carbon sampler. If you simply use a denuder followed by two quartz filters in tandem,
for example, certainly you can expect an enhancement for volatilization from the downstream filter,
downstream of the denuder. It's also valid that the filter downstream of the backup will not collect
that volatilized material. That backup filter in all likelihood represents only the positive sampling
artifact due to absorption of initially gas phase materials. What is lost, is lost, and is a limitation of
that method until you have a sorbent able to retain it. Have I helped you?
PETER MUELLER: I can illustrate the importance of that by comparing the amounts of
ammonium nitrate lost from denuded front filters in ammonia and nitric acid denuder systems. The
problem is how to do it for organics. That gets back to the fact that we don't know what kind of
176
-------
organic molecules we were dealing with. We can't grab hold of the problem unless we understand
the molecular entities that we're dealing with.
RICHARD TROPP: There are a couple of thoughts that I had. One is that it seems as if the
magnitude of the artifact may depend in part on the types of organic compounds that you have,
whether they're polar or not, how you want to describe gas particle partition, and so forth.
Therefore, one of the questions that I have is the representativeness of the data that we have
acquired to date. I think Los Angeles may be in a class by itself. You have data from the
Southwest. I'm not sure that you have an equivalent set of data from the Northeast. If you do, it
may be very seasonal. If there is any, it may have been done in certain segments in the summer
and you don't have data for the winter, and so forth. So there is a question, then, of how do you
interpret these results if you were to try to assess transinvisibility at some point across the entire
United States, coming up with a method that will work under all the circumstances that you're likely
to have, both in terms of fraction of organic material, relative humidity, and so forth.
The other question relates to, I think, something that Peter Mueller mentioned about in the
workshop on acid aerosol, and that is, how much do you need to know in order to have confidence
in your results later on? That has to do with the uses to which you're going to put them. It may be
that you can afford a fairly large uncertainty in the Northeast because the organic matter in rural...
and another issue is that the difference between remote areas and urban areas. You may be able
to afford a larger uncertainty in the eastern U.S., because the organics is a practical sense or a
smaller fraction of the visibility reduction than they would be in the West. So I think that needs to
be balanced, that we figure out, based on what we want to do with the models, what kinds of
uncertainty we can tolerate. Then the issue is, if we can get our answers good enough, do we have
to go any further in terms of the use that we're going to put it. It may be nice to know as a scientific
issue, but if you're going to expend the money to track a number of stations, do you really want to
expend that money to do better than you really need?
PETER MUELLER: I would like to underscore that by saying that we did that for the visibility
research. If you're going to go off to do a field study because you definitely have to have
information now for some assessment purpose or another, then you have to decide, even though
you know you're limited in your measurement methodology and technology, to what trouble you
have to go to get the best kind of measurement. For instance, in the visibility situation, by making
some assumptions about humidity and the extinction efficiencies, we calculated what our tolerance
would be in measuring each one of the components as a fraction of its contribution to the humanly
perceptible change in light extinction.
177
-------
Then, we devised a method that will give us no more than that much tolerance. For instance,
for the sampling of ammonium nitrate, which has a small contribution in that particular domain 1 to
extinction, we could actually tolerate ±100 percent coefficient of variation.
That's a very practical engineering approach to a specific situation. On the other hand, if there
is something generic going on environmentally that could be crucial but we fundamentally don't
understand it, then it's worth all the money in Saudi Arabia to find out.
RICHARD TROPP: But it may be the difference between having some special samplers at a few
sites versus having a less specialized sampler at many more sites.
PETER MUELLER: My point here is you mustn't mix up the need for fundamental understanding
because of its many, many different applications, with a specific application where you can tolerate
errors and you don't need to know everything before you go ahead.
WILLIAM WILSON: Our panel of experts here have had a chance to comment, but let me give
you another chance to comment and then we'll have our break. Peter, any other comments you
want to say?
PETER MCMURRY: No, I think I said earlier what I wanted to say.
BRUCE APPEL: I have some specific recommendations, but perhaps you want to address those
after break.
STEVEN JAPAR: The same.
WILLIAM WILSON: I just want to comment that, while one of our important focuses here is a
sampling technique that will at least tell us the mass of organic aerosol that really exists, and I think
we see some ways to go about that. In order to understand what's going on, I think we need to
give some thought to work where we do not have to take the particle out of the gas phase. There's
some in situ work where we can look at individual particles by light scattering or ability analyzers
or perhaps microscopes or something-some way to get a better understanding of the dynamics of
the individual particles and how they are influenced by the concentrations of vapor around them,
water vapor, organic vapors, temperature, and so forth. We need to look at those two areas.
PETER MUELLER: The last one is really key, because water association with particles is related
very much to the chemistry of the composition of the particles. The only material in particles that
people know very much about are various sulfates and their relationship to water. Again, if we don't
have the scientific understanding of the organics, some of which are substantially more hygroscopic
than some sulfates; we don't really understand their contribution to light scattering under various
conditions of humidity. Getting the needed knowledge on organics is very important to public policy
in the United States.
PETER MCMURRY: One comment that I'd like to make-Steve Japar, you didn't say anything
about your photoacoustics spectrometry this morning, and yet you've done at Ford some really nice
178
-------
work in the last years in which you've looked at particle absorption, absorptivities and correlated
that, for example, with EC measurements done by Jim Huntzicker and his group, and gotten very
good correlations assuming a specific extinction of 10 meters squared per gram or whatever. I
wonder if more thought oughtn't to be given to the use of a technique like that. I realize it's a very
complicated technique in its present incarnation. After all, that's the reason that we're presently
looking at EC and trying to distinguish between OC and EC. Essentially, your technique is a direct
measure of the optical properties. In fact, the optical properties of elemental carbon, according to
our work, depend somewhat on the size distribution. We find that the specific absorption depends
very much on the EC size distribution. If it peaks at .2 microns, you get different specific extinctions
than if it peaks at .4 microns. That's experimental.
STEVEN JAPAR: I'd be interested to see that. I've been looking for information along those
lines.
PETER MCMURRY: This is based on measurements that we obtained in SCAQS. I really
wonder if it wouldn't make sense to focus more on the total carbon rather than OC-EC and to
develop a good technique for measuring total carbon, and then to measure directly absorption of
the aerosol itself to infer the absorptivity.
STEVEN JAPAR: We've talked about that. Peter Mueller is actually interested in some aspect
of that. One major problem we have with that-if you are in an environment where the aerosol
optical absorption is not dominated by carbon, then you can't use that technique for EC
measurements. You would like to use the spectrophone technique to separate out elemental carbon
from the total aerosol. If there are other components in there that absorb, which may be the case
as you get to into more and more remote situations, there are problems. If you're in an urban area,
probably, to a very good first approximation, aerosol absorption in the visible region is due to
elemental carbon.
??: I think Karen Adams (Ford Motor Company) makes that point very strongly in her recent
paper.
STEVEN JAPAR: It's clear. Everything that we've ever done point very clearly to that. But then
we've never done any work in a rural or remote environment, specifically because don't have the
sensitivity in the current evolution of the instrument to do that. Peter Mueller, I think, is going to try
to address some aspects of that. The other problem that you run into is one that was alluded to
here, again by Peter this morning, and that is a very specific problem involving wood smoke. We
have never used the spectrophone to look at wood smoke. The optical properties may be very
different for wood smoke, even for the supposedly elemental carbon fraction of it. People at NBS
who have essentially been involved in nuclear winter studies have done a lot of work on various
materials. They don't seem to see a lot of difference, but that's still an open issue.
179
-------
PETER MCMURRY: My point is, from a visibility point of view, if you have black organic carbon,
you'd like to measure that as an absorbing species. You don't want to use a technique that's going
to tell you that's organic carbon.
JAMES HUNTZICKER: But that point is, it's not really black. It's sort of brown or yellowish
brown. There is an important difference in that the index of refraction probably isn't constant across
wave length as it is for elemental carbon.
STEVEN JAPAR: The issue comes down to the fact that the spectrophone measures aerosol
absorption. We can set it up so you get rid of all gas phase compounds that you're not interested
in, so that you just measure the aerosol. So you measure an aerosol absorption. The question is,
what is the absorbing species? We have done all our work in an environment where the absorbing
specie is elemental carbon. If you move out of an urban area, we would have to re-prove, at the
very least, the whole thing. It's not clear that the beautiful correlations we have on all of the urban
work that we've done would remain if you took that instrument out to Colorado, or other such
places.
JAMES HUNTZICKER: Portland, Oregon in the winter.
STEVEN JAPAR: Portland, Oregon in the winter may not be bad. I don't know how much wood
smoke there is.
JAMES HUNTZICKER: A lot of wood smoke.
STEVEN JAPAR: It would be interesting to see, under a number of these circumstances, what
the situation is. I don't think it's a straightforward application. There's a lot of work that needs to
be done.
??: My point is that it does directly measure the aerosol...
STEVEN JAPAR: It measures the aerosol optical absorption, which is an important parameter
to measure.
??: At one wave length?
STEVEN JAPAR: At one wave length, in the visible, the way we have it set up. You can
measure it at other wave lengths if you wanted to do it. The way we have it set up, we do all our
measurements in the green at 514.5 nm. Peter might want to comment on other aspects of this that
I overlooked.
PETER MUELLER: I totally agree that we've got to measure aerosol absorption regardless of
what the particle phase of the aerosol absorption is. In some situations, it may be primarily
elemental carbon that's doing it. In other situations, it's not necessarily just elemental carbon in the
particles that's doing it. So we mustn't confuse absorption measurements with elemental carbon.
They are two separate measurements. Sometimes you can derive one from the other, but it's not
a given that you can. That's a very important thing to separate in people's minds, first of all.
180
-------
Secondly, the kind of work that Karen Adams and Steve Japar have done in Los Angeles really
establishes a reference method for measuring absorption that you can't do on the filters. Therefore,
it seems to me, we should have a technology that would utilize this principle as essentially a
standard reference material, or as a benchmark.
Certainly, EPRI would be interesting in co-sponsoring such a development, and if you are all
interested in doing that, that might move things along.
WILLIAM WILSON: OK, fine. I want to just mention one thing that we haven't talked much
about, that one of the primary sources of organic aerosol is forest fires. How much is known about
that, I'm just curious. Is there much information on this? How do you do a source stream for a
forest fire? Is there any information on composition that you could use, and source apportionment,
to figure out how much was due to forest fires?
STEVEN JAPAR: There are two marker compounds for wood smoke, including phenolic
material. You get a lot of phenolic-type material. If you look at what is known about the absorbed
organic component in vehicle emissions versus forest fire emissions, there are major differences.
Because of the resinous material in the trees, you get a lot of phenolic resin-type material that you
don't see in any other combustion process.
WILLIAM WILSON: Can you differentiate between a forest fire from burning wood in your stove?
??: I doubt it.
PETER MCMURRY: I think that's really a major question for the Southwest. What are the
contributions of agricultural burning, forest fires, woodburning stoves?
JAMES HUNTZICKER: The way most people burn wood, is that they take the wood, toss it in
their wood stove, then close the damper all the way down so there's minimal air in there. They
essentially distill off the organics, which then recondense as they are diluted in cool ambient
atmosphere. Thus, the organic to elemental carbon ratio is very, very high for the conditions under
which most people burn wood. Forest fires, I would expect, would be different.
WILLIAM WILSON: You certainly would tend to have a higher oxygen.
JAMES HUNTZICKER: You might have a lower organic to elemental carbon ratio, too.
WILLIAM WILSON: That's something Cass might be looking at.
PETER MUELLER: In the wintertime, if you drive at night like I sometimes have from Phoenix
up to the Grand Canyon, people are burning wood. You can characterize each community by the
nature of the forest species that they burn, by their odors. So there must be a signal there that we
could sample, but it's going to be very location-specific.
WILLIAM WILSON: Let's take our break and try to get back by 3:15 and have some
recommendations.
181
-------
X. RECOMMENDATIONS FOR FUTURE AREAL RESEARCH
WILLIAM WILSON: Let's get down to the bottom line now. What do you think EPA ought to do?
I guess you don't have to consider the fact that we probably don't have any money to do anything
with. Let's rephrase that and say, what do you recommend to the scientific community as some of
the next steps?
PETER MUELLER: In relation to what?
WILLIAM WILSON: I should mention that EPA has some special problems that we're interested
in. As a result of the anticipated SO2 reductions, EPA expects to set up a major monitoring network
throughout the eastern U.S. to see if sulfate concentrations, dry and wet, acid deposition, aerosol
acidity, and visibility decline as the SO2 emissions decline. We're interested in a sampling technique
that will give us the contribution of organic aerosol to visibility as one special problem. We are also,
of course, interested in understanding the contribution of organic aerosol to visibility in both the
West and the East, because it is possible as the result of the Clean Air Act that there will be
renewed emphasis on visibility in the West. We need to know how to sample there, and we need
to be able to figure out where the organic aerosol comes from if it does turn out to be important.
We are still interested in basis understanding of the entire process so we can tell Frank Binkowski
what equations to put into his model for the formation of precursors and how precursor gases form
particles. Bruce, we'll let you start.
PETER MUELLER: When do you need to be out sampling? Because that means that
recommendations have to be related to that.
WILLIAM WILSON: We ought to be out sampling two years before the S02 controls start.
PETER MUELLER: I have no idea how much time that is.
??: Two years from now?
WILLIAM WILSON: It depends on when Congress passes the bill and what sort of time
schedule they have.
A. Bruce Appel
BRUCE APPEL: Let me begin by repeating the two things that I had said in my previous
remarks that I would recommend. One was to repeat the McDow and Huntzicker experiment. That
was a comparison Q-Q versus what I referred to as Q-TQ as a function of face velocity, but under
conditions in which the filter area remains constant and you simply change the flowrate in order to
vary the face velocity. Under those circumstances, you'll end up looking at a range of total sample
volumes. That's one suggestion. I think that it would be useful to try to reconcile the apparent
183
-------
discrepancies in two data sets, one of which suggests that Q-TQ is clearly the way to go. Another
would suggest that under some conditions, namely, low sample volumes Q-Q seems to look better.
I think that would be a useful thing to do.
Secondly, I mentioned earlier that I thought that Dennis Fitz' approach was a useful starting
point, and his approach was a denuder consisting of a quartz filter to be used ahead of a filter pair
and, of course, he had a cyclone on the front in order to exclude the coarse particles. I would
propose that this could be a useful starting point to what I refer to as a true paniculate carbon
sampler, and that would be one in which the denuder was designed to exclude both the vapor
phase constituents able to be retained on the bare filter medium, as well as on the paniculate matter
thereon. That might be quartz fiber plus things able to retain nonpolar organics, if I am correct that
the sorption of vapor phase materials on paniculate matter is indeed a significant factor. There isn't
an unambiguous data set to really support that concept.
The other thing that you need is a sorbent downstream. There's been some suggestion, and
Peter Mueller has been alluding to it a couple of times now, that activated carbon on a filter might
be functioning in that capacity. Some work that Delbert Eatough has done, which I personally am
not acquainted with, but I think that work ought to be surveyed in some detail, perhaps by
somebody looking critically from the outside to see how valid it may or may not be. The point is,
a material that is able to retain organics that are lost by volatilization from the filter. Under those
circumstances, the sum of the filter-collected carbon plus sorbent carbon would be a measure of
true paniculate carbon, and you may or may not need this backup quartz filter.
The next point: an awful lot of work Bob Lewis has funded deals with specific compound
measurement by the denuder difference method. And now we've heard this morning that a lot of
money is being spent essentially on a scale-up for that approach, to compound annular denuders
and multiple compound annular denuders. But I question the degree to which the original data set
is valid or adequate to support this development. What I mean is, the original work that Bob
Coutant did was nice as far as it went, but I think it left some questions unanswered. I would
suggest that he take a further look at the fundamentals of that sampling approach before moving
ahead with such dispatch on an engineering development before knowing sufficiently that this is
indeed a valid approach.
I'll mention just a couple of uncertainties that I see in his approach. Others here may see other
areas that they might want to comment on. One of the uncertainties relates to the diffusion
coefficient. The efficiency of collection was measured only for the compound naphthalene. Its
diffusion coefficient was presumed to apply to all PAH compounds. That's not a valid assumption,
because with increasing molecular weight the diffusion coefficient diminishes. And with a lower
diffusion coefficient, the removal efficiency should go down. He used a correction factor of 11
184
-------
percent. All results are corrected to allow for an 11 percent penetration based on the results with
naphthalene. They obviously are not applicable to other compounds, and the higher in molecular
weight you go the larger the error.
Second point: this large degree of volatilization loss that he observed, for example, with pyrene,
82 percent of his initially collected material was volatilized. That's a very important observation if it
is, indeed, correct. I'm not satisfied that necessary control experiments have been done. Why not,
for example, do something like this? The way Bob is doing that, he's taking a denuder, a filter, and
a downstream sorbent. If one has a vapor phase PAH compound and one introduces a vapor
phase PAH compound into the front end of the denuder, making sure that it's vapor phase by
having a filter up in the front. If this denuder is, indeed, effective for the removal of that vapor phase
material, there should be zero collection on this sorbent, zero collection. I haven't seen those kinds
of control experiments, but I would suggest that they are not only useful but necessary before I
would start to feel that there is credibility in his approach as applied to things substantially heavier
than naphthalene. This can be done with a range of PAH compounds that are less nasty than
others. You don't have to pick out the most carcinogenic one that you can find. I think that pretty
well expresses my range of concerns.
WILLIAM WILSON: OK. Jim, do you have some recommendations for us?
B. James Huntzicker
JAMES HUNTZICKER: First of all, I'm happy to hear Bruce recommend that we repeat our
experiment under different conditions, and I think that's a valid suggestion. Also, I think that it's
worth understanding how general our results are by making measurements in different localities.
I think that we lack a fundamental understanding in terms of the physical chemistry of what's going
on. Pete McMurry spoke about condensation aerosols and sorption and various ways that organics
can get into the particle. Additionally, there's the whole problem of gas/particle partitioning. I think
there's a great lack of knowledge about that whole field in general. I would point out that EPA,
i
through its competitive grants program, will be funding Jim Pankow of the Oregon Graduate Institute
to study the problem of gas/particle partitioning.
PETER MUELLER: Are you also funding Biswas at the University of Cincinnati?
??: They don't talk to each other.
WILLIAM WILSON: They tell us in such a way that we don't always find out. It's not that they're
trying to keep anything from us, but there's nothing that says you are to find somebody in the
laboratories who are interested in this and make sure they know about it. Who is doing the basic
physical chemistry?
185
-------
JAMES HUNTZICKER: Jim Pankow, at Oregon Graduate Institute. He's got a grant from EPA
through the Competitive Grants Program.
WILLIAM WILSON: Who were you asking about, Peter?
PETER MUELLER: Biswas. He works at the University of Cincinnati. He's a Cal Tech graduate.
He presented a paper at the Rocky Mountain Conference on Visibility-it's in the Proceedings-on
utilizing models and our knowledge about the physical chemical thermodynamic processes to
design sampler equipment and to analyze the relative benefits between optional characteristics.
Most of us have designed by the seat of our pants-intuitively. I think it's time to put that kind of
science into designing a sampler for this problem.
JAMES HUNTZICKER: The work that Pankow is doing is on model atmospheric particles in a
laboratory situation to study the adsorption phenomenon. These studies will be done as a function
of sorbing species and other variables such as relative humidity and temperature, which we know
affect the process.
The only other thing I would recommend is that we develop the theoretical physical chemical
underpinning for all of this, which I think is in very poor shape.
RONALD BRADOW: I hope you will find this relevant. It has been, in the case of source
emissions, possible to collect sufficient particle mass to be able to work with particle mass in a
physical chemical sort of way. For example, Mark Ross and Terence Risby have actually collected
diesel particles and conducted thermodynamic studies using Gidding's techniques to measure the
heat's absorption of a large number of organic compounds on relatively stable diesel particle
extracts. What was found in those cases was, of course, these particles are dominated with
relatively high molecular weight organic material. The particles were found to contain films of
organic matter much greater than a monolayer, and in fact the thermodynamics in those cases
could be explained essentially as Henry's law solutions of the organic materials in the thick layer of
fairly high molecular weight organic material which is essentially stable in air, very low vapor
pressure.
I'm not saying that this kind of approach is employable for atmospheric aerosols in general,
because it's very likely that there are times and places when the aerosols will have very different
chemical compositions. It's certainly something one can do with model aerosols to build such
experiments. They're relatively easily done. It's not a serious complication to pack gas
chromatographic columns with model compounds in the study the degree of absorption that occurs.
In fact, with Gidding's technique to determine, you can come up with some very useful and very
fundamental physical chemical parameters and such things. I recommend Mark Ross and Terrence •
Risby in Journal of Colloid Chemistry, as I recall about 1982. I have a copy of the paper.
186
-------
PETER MUELLER: I'm glad you mentioned that group. Of course, I think they are largely
involved in the health sciences. They've been overlooked by the aerosol scientists. Risby's group
is super. They've published a lot. And it's all on carbon.
RONALD BRADOW: It's isosteric heat absorption on carbon. They haven't measured. But,
again, it's a question of having a large amount of material to work with so that Gidding's technique
can apply. I suspect that there are ways to conceive of to conduct experiments of a similar type,
at least with model compounds, and very likely also with real atmospheric aerosols. It's worth
thinking about.
PETER MUELLER: Is that at John's Hopkins?
RONALD BRADOW: Johns Hopkins.
BRUCE APPEL: Could I mention one additional thing? I knew there was something I left out.
One of the fundamental uncertainties associated with denuder techniques that we are all aware of
is the problem of induced volatilization within the denuder itself. This has been evaluated in some
detail in relation to the nitrate system, and the volatilization of ammonium nitrate has generally been
concluded there. With the residence time on the order of between .2 seconds and 1 second, it's
not a significant factor. But, the same situation potentially with organic aerosols and, given how
labile they are, it may be a much more profound problem in the area of sampling organic aerosols
by techniques that involve denuders. No one has addressed that issue. As part of the additional
QA on the Coutant approach, I would advocate that some concern be given to that issue.
C. Peter McMurry
PETER MCMURRY: My first comment has to do with the modeling efforts. I'm rather concerned
about what I consider to be an enormous gulf between what we can do computationally and what
we understand about physical and chemical processes in the atmosphere. The direction that has
been followed by EPA within the last 10 years has been to put less effort into understanding what
is going on and to put more effort into modeling what is going on. In the long run, that is probably
a mistake. Our understanding of the behavior of these species in the atmosphere is very
incomplete. I don't think we have an adequate understanding of source profiles, where the materials
come from, what are the size distributions of the particles in the atmosphere. This is a general
comment that applies to all atmospheric paniculate species, and especially to carbon-containing
particles. It's a real concern that I have. I think it applies to visibility modeling. The aim is to predict
size distributions and size resolved chemical composition. If we don't have a better handle on the
physics and chemistry of what is going on in the atmosphere, especially the chemistry, I think we're
not likely to do a very good job on the modeling side of things.
187
-------
Secondly, I think that some thought out to be put into doing various comparison studies in
places other than Los Angeles. There has been some work done in visibility research, organic
particle measurements in the Southwest and the Grand Canyon and so on. I think it's really
extremely valuable to get people together. I don't think it's realistic for us to sit here and to propose
specific ideas that are going to solve the measurement problems. The solutions evolve over time.
These intensive experiments whereby we get people together who have a range of ideas, and they
do specific experiments to compare different techniques ultimately will pay off. That should be done
in places other than Los Angeles. Los Angeles has a particular type of aerosol. We see things
which are very, very different in the East, although to my knowledge that sort of experiment has not
been done in the East for organic particles. Things are very different in the Southwest and probably
in the Northwest. I think that we should, every few years when ideas begin to evolve, try to get into
different places. As I mentioned before, I think that a lot of the frustration that we're facing with
organic measurement has to do with the fact that all filter techniques rely on the use of quartz filters.
Quartz filters inherently give problems, and it would be extremely desirable to find a different
sampling substrate. I don't know if anybody is looking, for example, into the use of silver membrane
filters. Is that viable?
??: It's viable but how's the efficiency of those?
PETER MCMURRY: I don't know the answer but we could certainly determine the efficiency if
it's not known.
??: I think one of the problems was that those weren't very efficient filters, but you're right. If
you had an efficient enough silver membrane filter, then it would be really useful.
RICHARD TROPP: They are not necessarily a problem in remote areas.
??: That's certainly true.
RONALD BRADOW: In the case we applied these to, in a source sampling back in the early
days in the early 70's, we had problems with efficiency and with capacity. In the case, as you say,
of remote areas with relatively low holdings, you don't have as big a problem.
JAMES HUNTZICKER: If you had the right efficiency, you'd probably have such low blanks on
them that your signal to blank is pretty high. You don't have to collect as much sample.
RONALD BRADOW: I've seen this once and it worked rather well for many, many of the
samples we did.
PETER MCMURRY: Aside from those points, I have ideas on specific experiments that I think
ought to be done in the future. I go along, for example, of the suggestions of Bruce and Jim. A lot
of these things would be done. If there were intercomparisons, people would try different schemes
to see if it was possible to get better correlations between various techniques. Until we understand
188
-------
the discrepancies from one sampler to another, we really don't understand measurement of
paniculate organic carbon. That ought to be our objective.
FRANK BINKOWSKI: I appreciate your comments on modeling, but one of the things about
modeling that is important is that it gives you a place to test ideas and to test them in an
environment in the model where you can vary things geographically because of the different sources
and see how various mechanisms perform and then do comparisons with field work. I agree with
you in terms of organics barely enough to even start thinking about them all.
PETER MCMURRY: I think much the same can be said about sulfates. If you want to predict
size distributions in sulfates~we shouldn't get into that probably now~but I think there are real
questions in terms of how you model that. I agree that modeling can be useful as a "what if*
scenario, but it should be done in a mode that closely couples the modeling work with well-defined
field experiments. I don't see that happening, and I think it's a mistake.
D. Steven Japar
STEVEN JAPAR: What I got out of today-and I'm really somewhat of an outsider since I'm not
directly involved-I'm sort of torn between two points. You express a clear need to measure organic
aerosol, and you need to develop a failsafe technique to do this. If you're going to do that, from
the discussion we have had here today, it seems to me that the major uncertainty is this negative
artifact. If the negative artifact doesn't exist, then we probably know how to measure organic
aerosol. You can put all of us in a room for a month and probably come up with a pretty good
approximation of how to do this if we didn't have to worry about the organic artifact. If the organic
artifact does exist, then you come back to what is in essence a definition of organic aerosol, which
is method-specific. You don't really know what you're doing. This brings me back to the
understanding of the physical chemistry of the system, which has been discussed by three or four
different people here. I think that is a basic need that has to be addressed.
The next thing that I think you folks have to address is, does it really matter? You have a
specific reason for going out and measuring organic aerosol. As Peter described it before, we can
measure organic aerosol now. His limits were plus or minus 100 percent, but it's probably
something less than that.
PETER MUELLER: Not for organics. That was for nitrates. For organics it's plus or minus 50
percent.
STEVEN JAPAR: So it's even better. It's not inconceivable that for many of the needs that you
have, that accuracy is sufficient. To go on beating your head against some aspect of this to get
another 10 percent improvement on that is something you don't need to do. That has to be a very
189
-------
important evaluation as to where you go with this. I don't feel particularly strongly about what the
degree of satisfaction with the techniques is, but that has to impact the whole thought. I'm not in
a position to recommend specific experiments and things that should be done, but I think that the
one place where a lot could be gained without getting into the area of fighting over little nitty gritty
differences about whose method is best is to actually go back and look at the physical chemistry,
and the adsorption phenomenon. We really understand virtually nothing about some of these
systems. That would be a major step forward and have a major impact on the discussion we are
having here.
WILLIAM WILSON: Thank you, Peter Mueller.
E. Peter Mueller
PETER MUELLER: The major problem is for the eastern United States where we are lacking
data. I think the Oregon Graduate Center was involved with a study for making measurements of
carbon across the United States.
JAMES HUNTZICKER: We had a project funded by Jack Durham's group where we took filters
out of the NASN data bank, so we've got yearly averages for organic and elemental carbon for some
50 or 60 cities and 30 rural sites. But, you have to consider that those filter samples had some
unknown history, and so you have to be a little uncertain as to what the numbers mean.
PETER MUELLER: You have to look at those kinds of data sets to currently understand the
scale of the carbon problem that we're talking about here, in relation to other material. That really
needs to be done-some sort of a synthesis. Then, you have to determine what your measurement
tolerance is for at least one application, your immediate application of visibility. The template for that
has been established for doing that. You need to select a measurement method that you can prove
out so that you can apply it within the next 12 to 24 months in a network for that specific application.
To that, I would encourage using the current knowledge in terms of thermodynamic and aerosol
physical chemistry calculation schemes in a diagnostic sense. I don't think that we need to spend
millions of dollars doing that. There are lots of efforts going on right now doing that, that we can
piggyback on. It should not detract from the experimental work. Then apply that kind of calculation
scheme to the analysis of various sampling designs. Certainly, I would favor a sampling design like
Bruce explained, which is a denuder. I would use a denuder that really takes up all vapor phase
organics and should be an activated carbon denuder of some kind, lets the paniculate material get
through, and then have an activated carbon absorbent after-filter which is analyzable. In other
words, it would have the carbonaceous substrate. The carbonaceous substrate should be
manufactured in such a way that its blank variability is very, very narrow compared to the amount
190
-------
that you add to it. Even though the blank could be high, if it's variability is narrow, it would be
usable. Basically it would be a denuder difference sampler for carbonaceous paniculate material.
Taking out the sampling systems to various locations, such as Pete McMurry suggested, is really
a good idea However, I believed that method intercomparison experiments should only be done
after all methods have been verified with respect to a reference material.
WILLIAM WILSON: That's what came out of our ? Aerosol Workshop, that you ought to first
do something that you understood-if you had a standard sample, do it first, if you don't have a
standard sample, do something simple that you know what it is and everybody understands-that's
a valid point.
PETER MUELLER: I think what's really urgent to do is to analyze the humidity interactions with
organics. We have one such facility of which we sponsor the development at Desert Research
Institute. We call it a benchmark facility, because it's really not a standard reference material. There
are a number of people who have samplers that indicate what happens to the aerosol as a function
of humidity, either by drying first or adding water later. The TMDA scheme is probably the most
fundamental one. It seems to me all devices ought to give the same data against the benchmark
unit. The benchmark unit now consists of a scheme where sodium chloride particles are generated
at different humidities, and the change in size distribution, which is theoretically predictable, can also
be measured at the same time with an optical counter. There are some aspects on the input side
that can't be defined experimentally, so the matching is based on calculation, on theory. That's why
we call it a benchmark. It's reproducible but it's not necessarily absolute.
The next step is to generate organic salts and repeat those experiments and to get a better
handle on the kinds of humidity factors people have been putting on the organic aerosol.
??: You said organic salts. Did you mean that?
PETER MUELLER: Salts.
??: Acetates primarily, or?
PETER MUELLER: Some pure organic compounds. I'm not quite sure how to select them at
this point. In the meantime, Fred Rogers is doing most of this work at DRI with John Watson. They
wanted to do the experiments initially with wood smoke, that was generated with some sort of
well-controlled environment. I opposed that, because we don't know what we generate. We don't
know the input generally. We don't understand it, we don't know then what we've learned, how to
relate it to the rest of the universe. As far as the nuclear winter research, I think he got a small grant
from somebody, where he did nevertheless generate wood smoke, and he has some data on that
already, which he is just working out. It's quite interesting. It definitely demonstrates the capacity
for the organic material to glom onto an awful lot of water as the humidity goes up. That is just an
191
-------
aside. The recommendation is we do need to do more experiments to study the humidity effects
on aerosols, especially on the organics.
WILLIAM WILSON: Anybody else have any comments? Parker Reist, you came in a little late
... is there anything you'd like to say? We'll give you a chance to make comments if you'd like.
PARKER REIST: Thank you. I really don't feel I've been in on the discussion long enough to
make any comments that would be worth your time, but thanks for the opportunity.
PETER MUELLER: I have one afterthought. The other important problem with the eastern
United States is the optical side of the question; not just the physical chemistry side of the aerosol,
but also the optical side of the aerosol. We've already talked about the need for an absorption
measurement technology. I think that really should be a major recommendation for you all to
participate in. The other is the extinction measurement side and the path radiance measurement.
The extinction measurement works fairly well in the western United States, but there are apparently
quite a few problems in the Eastern environment. The optical measurements for the eastern United
States, recognizing North from South as being different as well, deserve a lot of immediate attention
over the next two-year period, so that you are then in a position to do something.
WILLIAM WILSON: You don't think we can just take the transmissometer that's being used in
the West and stick it in Eastern sites?
PETER MUELLER: The experiments that have been done so far, contra-indicate that.
RICHARD TROPP: What seem to be the factors that work against that?
PETER MUELLER: I'm not expert enough to comment, but certainly the Park Service people,
or ARS, could tell you about it.
WILLIAM WILSON: Anybody else have any afterthoughts? Any comments you'd like to get into
the record?
RONALD BRADOW: Would you accept written comments at a later time?
WILLIAM WILSON: Sure.
RONALD BRADOW: Thank you.
PETER MUELLER: I have one more afterthought, and that is if you indeed go out to set up a
network at this point in time, and you do have a partial network out there now, then I would listen
very carefully to Will Richards' recommendations. They are to measure extinction, scattering, the
aerosol chemistry by particle size, and path radiance. Path radiance, so far, has been ignored in
everything we've done. It's very difficult to deal with because it doesn't have a general applicability.
It is vista-specific.
WILLIAM WILSON: What I don't know, and it is something I want us to calculate sometime,
unless somebody knows, is that, if the ratio of extinction to path radiance is a function of the size
distribution of the aerosol, it might vary from one kind of aerosol to another kind of aerosol.
192
-------
??: It sounds like a really good guess that would be the case.
WILLIAM WILSON: But it would be the same or would it vary?
PETER MUELLER: I think Will has already done that.
??: I would guess it would vary.
WILLIAM WILSON: I think one thing we need to look at for our visibility analysis is how it varies
as a function of size distribution or refractive index for different aerosol populations.
PETER MUELLER: I think he has a calculation scheme for doing that.
RONALD BRADOW: Will worked on this some years ago. I think he does.
WILLIAM WILSON: We'll take afterthoughts in writing or by comment later on. I don't know at
what point they will no longer get into the proceedings. Mike, do you want to tell us what plans you
have for any sort of an output from this? Will there be report?
MICHAEL BARNES: Yes. We're going to try to put together a summary report of 5-10 pages.
We'll try to get it out within a couple of months.
WILLIAM WILSON: I would like to thank everyone for coming, especially our invited guests from
other organizations. Thank you very much for coming and giving us the benefit of your thoughts
and considerations on these problems.
193
-------
APPENDIX A
UST OF PARTICIPANTS
195
-------
LIST OF PARTICIPANTS
MINIWORKSHOP ON ORGANIC AEROSOL
U.S. Environmental Protection Agency
Research Triangle Park, NC
September 6,1990
Dr. Bruce Appel
California Department of Hearth
2151 Berkeley Way
Berkeley, CA 94704
415/540-2477
Dr. Harold Michael Barnes
U.S. EPA - AREAL (MD-46)
Research Triangle Park, NC 27711
919/541-3086
Dr. Frank Binkowski
U.S. EPA - AREAL (MD-80)
Research Triangle Park, NC 27711
919/541-2460
Dr. Ronald Bradow
U.S. EPA - AREAL (MD-56)
Research Triangle Park, NC 27711
919/541-1394
Dr. Marcia Dodge
U.S. EPA - AREAL (MD-84)
Research Triangle Park, NC 27711
919/541-2374
Dr. Edward Edney
U.S. EPA - AREAL (MD-84)
Research Triangle Park, NC 27711
919/541-3905
Mr. Gardner Evans
U.S. EPA - AREAL (MD-56)
Research Triangle Park, NC 27711
919/541-3887
Dr. James Huntzicker
Oregon Graduate Institute
19600 NW Neumann Drive
Beaverton, OR 97006-1999
503/690-1072
Dr. Steven Japar
Chemistry Department Research Staff
Ford Motor Company (MD-E-3083)
Dearborn, Ml 48121
313/845-8072
Dr. Charles Lewis
U.S. EPA - AREAL (MD-47)
Research Triangle Park, NC 27711
919/541-3154
Dr. Steven McDow
Rosenau Hall - CB #7400
University of North Carolina
Chapel Hill, NC 27599-7400
919/966-7318
Dr. Peter McMurry
Department of Mechanical Engineering
University of Minnesota
111 Church Street, SE
Minneapolis, MN 55455
612/625-3345
Dr. Peter Mueller
EPRI
P.O. Box 10412
Palo Alto, CA 94303
415/855-2586
Dr. Parker Reist
105 Rosenau Hall - CB #7400
University of North Carolina
Chapel Hill, NC 27599-7400
919/966-7305
Mr. Eric Ringler
ABB Environmental Services
6320 Quadrangle Drive, Suite 100
Chapel Hill, NC 27514
919/493-2471
Ms. Uma Shankar
Computer Sciences Corporation
c/o U.S. EPA (MD-43)
Research Triangle Park, NC 27711
919/561-0575
Mr. Robert Stevens
U.S. EPA - AREAL (MD-47)
Research Triangle Park, NC 27711
919/541-3156
Dr. Len Stockburger
U.S. EPA - AREAL (MCM6)
Research Triangle Park, NC 27711
919/541-2561 or 1540
Dr. Richard Tropp
ABB Environmental Services
6320 Quadrangle Drive, Suite 100
Chapel Hill, NC27514
919/493-2471
Dr. Robert Lewis
U.S. EPA - AREAL (MD-44)
Research Triangle Park, NC 27711
919/541-3065
Dr. Russell Wiener
U.S. EPA - AREAL (MD-56)
Research Triangle Park, NC 27711
919/541-1910
Dr. William Wilson
U.S. EPA - AREAL (MD-75)
Research Triangle Park, NC 27711
919/541-2551
197
-------
APPENDIX B
AGENDA
199
-------
ORGANIC AEROSOL WORKSHOP
September 5, 1990
Beaunit Building Room S-23
Time
9:00
9:15
9:45
10:15
10:35
11:05
11:30
11:50
12:20
1:30
3:00
3:15
9:15
9:45
10:15
10:35
11:05
11:30
11:50
12:20
1:30
3:00pm
3:15pm
4:30pm
Speaker
H. Hi 1 son
B. Appel
J. Huntzicker
Coffee Break
P. HcHurry
B. Lewis/N. Hi 1 son
S. Japar
P. Mueller
TOPIC
Introduction
Organic Aerosols Overview
Samp 1 ing/Art ifacts
Smog Chamber Studies
AREAL/MRDD Activities
Org. Aerosol Res. at Ford Motor Co.
Org. Aerosol Res. Prog, at EPRI
Lunch
General Discussion of Presentations
Afternoon Break
Discussions of Recommendations for Future AREAL Research
201
------- |