ESPA-R3-73-036
July 1973
Ecological Research Series
I i|ill||^
I
ft-:-::-:-:-.:-
$j!!!$!^^
$j$jj^
-------
EPA-R3-73-036
Mechanisms of Formation
and Composition
of Photochemical Aerosols
by
M. Lipeles with C.S. Burton, H.H. Wang,
E.P. Parry, andG.M. Hidy
Rockwell International
P.O. Box 1085
1049 Camino Dos Rios
Thousand Oaks, California 91360
Contract No. 68-02-0562
Program Element No. 1A1008
EPA Project Officer: Dr. Marijon Bufalini
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
July 1973
-------
This report has been reviewed by the Environmental Protection Agency and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, nor does
mention of trade names or commercial products constitute endorsement
or recommendation for use.
11
-------
Science Center
Rockwell International
SC537.10FR
TABLE OF CONTENTS
Page
Abstract . f
I. Introduction 1
1.1 Goals and Method 2
1.2 Accomplishments 2
2. Observations and Simulations of Photochemical Aerosols ... 4
2.1 Field Observations 4
2.2 Laboratory Studies . 11
2.3 Goals of Aerosol Simulation 14
2.4 Approach of the Flow Reactor Experiment 16
3. The Flow Reactor System 19
3-1 Reactor Design Principles 19
3.2 Detailed Description of Experimental System 30
3.2.1 The Reactor Section 33
3.2.2 Gas Handling 33
3.2.3 Irradiation 34
3.2.4 Sampling 36
3.2.5 Measurement Techniques 43
4. Experimental Results and Discussions 45
4.1 Gas Phase Chemistry 45
4.1.1 Inorganic Gas Phase Tests 45
4.1.2 Organic Gas Phase Tests/GC Measurements .... 47
4.2 Aerosols Experiments 51
4.2.1 Ozone-Olefin Aerosols 52
4.2.2 Flow Reactor Aerosol Experiments 52
4.3 Chemical Analysis of the Aerosol Samples 55
4.3.1 Sulfate Analysis 55
4.3.2 Nitrate Analysis 59
4.3.3 Total Organic Analysis 60
4.4 Discussion of Aerosol Results 61
5. Summary and Recommendations . . 64
5.1 Aerosol Production 64
-------
Page
5.2 Chemical Analysis ................... 6k
5-3 Data Analysis and Feasibility ............. 65
S.k Recommendations ............... ..... 65
Appendix A. Modeling, Mechanism and Simulation ........ 66
A.I Scaling of Reactions ................. 66
A. 2 Extrapolation of Laboratory Experiments to
Atmospheric Transformations .............. 71
A. 3 Mathematical Modeling of Gas Phase Kinetics
Appendix B. Aerosol Forming Reactions with Ozone 6 Olefins
in a Flow Reactor ................. 81
B.I Experimental Methods ................. 81
B.2 Experimental Results & Discussions .......... 83
B.3 Chemical Analysis of Ozone-Olefin Aerosol ....... 91
References ........................... 95
-------
Science Center
Rockwell International
SC537-10FR
ABSTRACT
This report details a feasibility study of a flow reactor concept for
the study of the mechanism of formation and composition of photochemical
aerosols. The technique involves a fast flow reactor which may be operated
in a "wall free" mode. Aerosols have been produced in this reactor from a
light irradiated gas mixture including N02> 1-hexene, S02> H.O, and air.
The chemical composition of these aerosols has been analyzed. Aerosols were
also sought in a N02, 1-hexene, and dry air mixture but have not yet been
obtained.
The report includes preliminary design studies and experimental study
carried out under a joint contract with Environmental Protection Agency
(EPA) and Coordinating Research Council (CRC)* and, for completeness, a
discussion of related aspects of a Science Center funded study of aerosol
formation in ozone-olefin reactions.
"This is under the Air Pollution Research Advisory Committee (APRAC) of
CRC. APRAC represents API, MVMA, and EPA.
-------
Science Center
Rockwell International
SC537.10FR
1. Introduction
One of the major problems confronting modern civilization is the con-
tinuing degradation of air quality in major urban complexes. An important
symptom of air pollution is the increased concentration of airborne particu-
lates over cities, resulting in increased health hazards, decreased visi-
bility, and corrosive destruction of property. The formation of aerosols
by man's activity is perhaps the most easily identified manifestation of
air pollution. However, the details of the mechanisms of formation of air-
borne particles and their evolution and removal from the atmosphere, remain
poorly understood despite considerable effort devoted to this subject.
A number of years ago, workers recognized that aerosol particles could
appear in the atmosphere via two different routes. The first involves pri-
mary emissions from sources such automobile exhaust, or smokestacks. The
other is secondary in nature, where particles are produced by physicochemical
processes in the atmosphere itself. The possibility of chemical aerosol
formation was probably first demonstrated by Tyndal1 in the 19th century.
The significance of such a mechanism for removal of reactive trace gases
(2)
from the atmosphere has not been fully appreciated until recently.
There is mounting evidence that the formation of aerosols by photo-
chemical reactions in the atmosphere plays a significant local role in air
pollution over many cities. In fact, the formation of a haze of such suspended
material is an integral part of photochemical smog. It was recognized in the
classical work on smog reactions that aerosols would form in photochemically
reactive atmospheres containing hydrocarbon vapors, water vapor, and nitrogen
oxides. More recent studies in smog chambers have elucidated the relation
(4-7)
between S0_ and photochemical aerosol formation. Other work, such as that
of Bricard e_t_ £j_. and Goetz and coworkers , have indicated that sig-
nificant gas-particle interactions take place in air containing pollutant gases
irradiated by sunlight.
-------
1.1 Goals and Method
In response to APRAC request for proposals for experimental work on
the mechanism of formation and compositon of photochemical aerosols, we
proposed a novel method to produce and study aerosols in a fast flow
reactor. This report covers a feasibility study performed under APRAC contract.
The objectives of this study are to (a) determine if light scattering air-
borne particles can be produced in an air flow containing hydrocarbon vapors
and nitrogen oxides, (b) analyze such particles for their chemical com-
position, and (c) determine the feasibility of using a flow reactor for
simulation of photochemical generation of aerosols in polluted atmospheres.
The flow reactor employed maintains laminar flow in an irradiated
section with a residence time of a few seconds to a minute. The gases and
aerosols are sampled by a variety of probes and filters and analyzed by
many methods. In the case of the organic component of the aerosol we are
able to employ infrared spectroscopy and gas chromatography (GC) with a mass
spectrometer detector (GC-MS). For the inorganic components of the aerosol
techniques include spectrophotometry, pulse polarography, and standard wet
chemical methods. For gas phase analysis we employ standard chemiluminescent
instruments and GC. For physical analysis of aerosols we have used
condensation nuclei counters and optical particle counters and have available
several methods of total mass monitoring.
1.2 Accomplishments
In this program we have designed, constructed, and tested a flow reactor
for the production of aerosols in controlled experiments. The testing
carried out under the APRAC contract was in the nature of a feasibility study.
The main goal was the generation of aerosols in the reactor in an irradiated
mixture of NO , air, hydrocarbons, and possible additional trace components.
J\
We have observed aerosols in one such system: NO-, air, 1-Hexene, SO-, and
H20.
Attempts to produce aerosols with only NO-, dry air, and an olefin,
irradiated in the flow reactor, with residence times of a few seconds, have
not to date produced aerosols. Due to limitations of time, we did not
-------
Science Center
Rockwell International
SC537.10FR
completely characterize the products or obtain completely consistent gas
phase data during the unsuccessful aerosol production tests.
In general during the testing of the reactor, qualitative agreement
with expected results was accepted due to time considerations. By quali-
tative agreement, we mean that if an expected product was found within a
factor of 10 of the expected concentration, this was satisfactory. If
effects such as the peaking of a product concentration with the variation of
a parameter appeared roughly as expected, this was satisfactory. Unfortunately
in some cases, particularly the estimation of the light intensity, this led to
large uncertainties in the analysis of the data.
The aerosols produced in the SO-, 1-hexene, N0? system were collected and
analyzed for composition. These analyses included (a) SOT, which was definitely
present, (b) NO, which was detected but may have a contamination problem in
the collection method, and (c) total organics which were less than a few
micrograms in a 100 ygm sample. Techniques for chemical analysis were also
tested on organic aerosols produced in another reactor from pure ozone-olefin
reactions in dry air. Because this work bears on the flow reactor experiment,
these simpler experiments are described in Appendix B, although these experiments
with ozone and olefins were funded separately by the Science Center.
We therefore conclude that we have demonstrated the feasibility of
producing aerosols in a fast flow reactor. It should now be possible to
attempt to completely analyze the NO , air, olefin system to find the
J\
conditions, in this reactor or a similar one, for which aerosols are pro-
duced, and to determine the applicability of such data to aerosol formation
in urban atmospheres. This will be supported by our improved kinetic
modeling capabilities which are discussed in Appendix A.
-------
2. Observations and Simulations of Photochemical Aerosols
In the first two parts of this section we will review field observations
and experimental simulations of photochemical aerosols. Following that we
will draw these two together into a set of goals for aerosol simulation and
then discuss our approach to this problem.
2.1 Field Observations
Knowledge of photochemical aerosol formation in urban atmospheres remains
incomplete. However, there are considerable data accumulated by the National
Air Surveillance Network (NASN) suggesting that aerosols in a photo-
chemical ly active atmosphere differ significantly from aerosols sampled
in air where photochemical reactions are less important. For example, the
former case reveals an enrichment in nitrates and benzene soluble hydrocarbons.
A recent survey of information on the Pasadena aerosol suggested that
more than one-third of the airborne material sampled was of secondary origins.
Despite the fact that the total aerosol mass observed is only a small fraction
of the total trace material loading of the atmosphere, the suspended particles
are particularly significant in their ability to act as carriers of toxic
matter to the lower respiratory system and as the major factor in visibility
degradation.
Investigations of aerosols in the atmosphere have revealed a highly
complicated behavior that is far from being characterized in any detail. There
is comparatively little information available about the mechanism of formation
and evolution of airborne particles, making the specification and development
of a quantitative simulation model difficult. Suspended particles have been
shown to have highly complicated chemical composition that originates from a
variety of natural and anthropogenic sources. The mixture of material is
from primary and secondary sources. Some elements such as silicon and lead
are identified with primary emissions such as auto exhaust and dust rise
by wind. Other compounds are related extensively to secondary processes,
= - +
including SOi, NO,, NH., and possible hydrocarbons. Studies in photochemically
reactive and non-reactive urban atmospheres have shown both differences and
similarities in aerosol properties, some of which are listed in Table 1. In
-------
Science Center
Rockwell International
SC537.10FR
Table 1
Speculated Differences and Similarities in Physical Properties
Between Photochemically Active and Non-Active Urban Atmospheres
1. While the average number of aerosol particles is greater for a
photochemically active atmosphere, the peak number, as reflected
by Aitken nuclei concentration is much larger for non-photo-
chemical ly active atmosphere. (Data from Ref. 60)
N (Ground Average) Daily Peak
Los Angeles I.l4x]05 cm~3 ~ 2xl05 cm"3
Minneapolis (2.1 to 6.6)xlO cm ~ 2x10 cm
2. The volume arithmetic mean radii of aerosol particles in a
photochemically reactive atmosphere are smaller than in a
non-active atmosphere (Data from Ref. 60).
Dpvm (Los Angeles) - 0.086 ym
Dpvm (Minneapolis) = 0.09 to 0.11 ym
3. Light scattering and visibility reduction in photochemically
reactive atmospheres may follow the change in ozone concen-
tration (See Fig. 1).
A. The particle size spectra measured in both show broad features
that may be correlated in terms of self-preserving "similarity
variables." However, both exhibit strong diurnal fluctuations
in particle concentration in submicron size ranges and below (60)
Both photochemically active and non-active urban atmospheres
show a bi-modal distribution of mass with particle size. The
pollutant aerosols appear to be mainly in the range of submicron
size while natural particles appear to be more strongly cor-
related with larger particles (60).
-------
photochemically active atmospheres, such as that of Los Angeles (LA), there
is a larger average number concentration, but a much smaller daily peak
number of Aitken nuclei, and a smaller mean particle radius compared with a
non-photochemically dominated atmosphere like Minneapolis. Furthermore, the
LA atmosphere is enriched in nitrates and benzene soluble hydrocarbons. The
average composition of a typical sample of the Los Angeles aerosol is shown
in Table 2. From these data, it can be seen that simulation in the laboratory
of the aerosol samples in a "real," typical photochemical reactive atmosphere
would be very difficult and probably impractical.
The constituents that have been identified as dominated by atmospheric
transformations include oxidized nitrate, sulfur, and carbon compounds.
Perhaps most extensively studied of these three are the sulfate forming
reactions through oxidation of S0?, either by homogeneous gas phase reactions
or by heterogeneous transformations in the presence of liquid water and cations
such as NH. or metal ions. The possible reactions of hydrocarbon aerosols in
the atmosphere which lead to nitrate formation are poorly defined at this
time. Indeed, there remain serious questions about the nature of nitrate in
aerosols, particularly if nitrous or nitric acid are involved. The high
volatility of nitric acid compared with sulfuric acid, for example, points
to a potentially significant distinction between homogeneous nucleation of
such material vs. heterogeneous processes. The mechanisms for formation of
organic aerosols in the atmosphere remain a mystery. Presumably condensed
material of a polymeric form, or a highly oxygenated form can be derived
from reactions involving unsaturated materials of carbon number higher than
three to four. However, these must be in competition with other mechanisms
that would reduce the molecular weight of reactive intermediates such as
ozonides or hydroperoxides as a result of decomposition reactions in the
presence of water vapor. Based on current knowledge, it would appear that
only 0.1% to \% of the reactive hydrocarbons in the polluted atmosphere are
converted to condensed material on the time scale of several hours.
There is mounting evidence that water is intimately involved in different
ways in aerosol formation. Charlson and colleagues, for example, have shown
-------
Science Center
Rockwell International
SC537.10FR
Table 2
Chemical Component List for the Pasadena Aerosol from Averages
of Data Sampled in Smog During Late Summer 1969
(after Miller et al., Ref. 12)
Concentration (Wt. %)
Ba ................. 2.9><10
V ................. 9.5*10
C (non-carbonate)
C£ ................. 0.03-0.6
I ................. i».8x]0~3
Unidentified ........... *>2.2
Total 100.0%
-------
that the light scattering urban aerosol increases with relative humidity (R.H.),
particularly above 50% R.H. 3' Recent work of W. Ho* ' during the 1972
aerosol observational program in Los Angeles indicates that the 1iquid water
content of atmospheric aerosol may display a diurnal variation closely linked
with changes in aerosol mass, b , and photochemical reactants such as
NO or 0,. Unquestionably water is potentially involved in S0]j, N0_, and
oxygenated hydrocarbon stabilization in aerosols. Thus, the role of this
most universal and prevalent of all atmospheric "trace" gases must be accounted
for in aerosol formation mechanisms.
To add to the difficulties in interpretation of aerosol simulation, the
kinetics of aerosol behavior have to be included. There is little information
on the daily-evolution of aerosols in photochemical smog. Perhaps the most
detailed field investigation of photochemical aerosols, completed to date
in the Los Angeles atmosphere,was conducted in a comparative study in 19&9-
This study attempted to measure, in as detailed a way as possible, the
physical and chemical properties of photochemical aerosols as a function of
time under conditions ranging from light to moderately heavy smog in Pasadena,
California. Results of this field program revealed that there was a strong
correlation between the behavior of aerosols and the daily evolution of trace
gas concentrations associated with smog.
Results for the average diurnal variation in gas concentration and
aerosol properties for Pasadena in 1969 are shown in Figs. 1 and 2. The
pattern of behavior of the trace gases is typical of that known for photo-
chemical smog in Los Angeles. The striking feature of the aerosol data is
that the smallest particles, represented by the total number density of
Aitken nuclei concentration, , follows a pattern quite different from the
larger particle fractions, characteristic of the average surface area for
unit volume, , and the volume fraction, <<(>>. The larger particle fractions
display a mid-day peak with ozone while the Aitken nuclei peak late in the
afternoon. This may imply that much of the photochemical interaction with
particles is manifested in the larger particles. Husar, Whitby, and Liu
have attempted to interpret these physical dynamics in terms of formation and
growth mechanisms. Their interpretations suggest that aerosol growth during
8
-------
Science Center
Rockwell International
SC537.10FR
18
16
O
z
O
cc.
UJ
10
'scat
T
OXIDANT (pphm)
N ,'f
_ NON-METHANE
HC (ppm)
I
12 15
HOURS POT
l/l
O
6 »
X
O
21 2k
Fig. 1 Diurnal variation in trace gases and in light scattering by
nephelometry, in photochemical smog based on composited data
from the 1969 Pasadena study.(17)
-------
E
o
/\
26
2*4
20
16
12
. 1
I
12
10
8
6
8 10 12 ll» 16
TIME (hrs) (POT)
18 20 22 2*4
O
CM
I
O
x\
c/1
cc.
o
o
2 T
o
X
Fig. 2 Average diurnal behavior of photochemical smog aerosol properties based on the 1969
Pasadena study.U?) represents the mean Aitken nuclei concentration, the
average surface area per unit volume, and <$> the volume fraction.
-------
'1*
Science Center
Rockwell International
SC537.10FR
the daylight hours is dominated by condensation on 0.1-1 ym size particles and
that this condensation is dependent upon the photochemical activity in the
atmosphere. New nuclei primarily enter into the atmosphere directly from
combustion sources except in the early morning when the concentration of
0.1-1 ym particles is small and homogeneous nucleation may take place.
Coagulation was shown to have negligible effect except late at night (after
2300 hrs) when both sources and photochemistry are at a low level. It was also
suggested by Hidy e_t^ aj_. that the condensation process may include
reactions directly on the particles of photochemically produced materials.
2.2 Laboratory Studies
For almost two decades the chemistry of polluted atmospheres has been
simulated and measured in laboratory studies. Extensive information has
been obtained about the gas phase reactions so that we have a knowledge of
their salient features. This work continues and new observations
and clarifications are still being made. For example, although previously
neglected as a reactant, an effect of CO has been observed and a
dramatically improved nitrogen material balance has been achieved with the
(21)
identification of HNO, on the chamber walls.
The level of knowledge of the gas phase part of photochemical smog is
such that new, highly speculative mechanisms are still being proposed from
time to time. Nevertheless, the role of N02 molecules as photoabsorbers,
the resultant oxygen atoms as initiators, and some sort of free radical
chains are accepted concepts. Complex kinetic models, which incorporate
these ideas, are now being introduced to attempt to describe the overall
(22-2M
reactions and serve as input to dynamical atmospheric models. These
models give at least qualitative agreement with atmospheric and smog chamber
results, although important discrepancies still exist.
The formation of aerosols is much less well understood or even studied
from a laboratory point of view. Although Haagen-Smit demonstrated the
formation of aerosols in the laboratory in his pioneering work on photo-
(25)
chemical smog, ' progress in laboratory simulation of aerosols has been slow.
Aerosols are much more difficult to control, simulate, and measure than
11
-------
gases and appear to be more dependent on trace constituents which may not
necessarily play a significant role in the gas phase reactions. Nevertheless,
several groups have been active and there is a body of knowledge on the
simulation of aerosols.
After Haagen-Smit's work, a group at the Stanford Research Institute
(26)
did a series of studies on aerosol formation. They first demonstrated
that automobile exhaust, when irradiated, produced aerosol particles in the
0.25 to 1 pm size range. The condensation nuclei were measured and showed
initially high concentration with a lower steady state concentration. They
went on to measure the chemical composition of these particles and found
nitrate, sulfate, and both water soluble and ether soluble organics. These
accounted for about 85% by weight of the material. They then tried various
mixtures of individual hydrocarbons with SO- and NO. In conjunction with
these experiments, a group at the Franklin Institute carried out a similar
(27)
series of experiments. Generally, these observers found in the absence
of S02 many nitric oxide-hydrocarbon systems produced some aerosol. The
largest amounts were associated with branched-chain internal olefins,
cyclo-olefIns, and diolefins. The introduction of SO. at the start of the
reactions dramatically increased the aerosol formation and aerosols were
produced even for those systems that produced none without SO-. The Franklin
Institute group showed that the SO- had to be present from the start to have
its greatest effect. These experiments were all carried out in stirred or
stirred flow type reaction vessels of various sizes.
A rather extensive series of experiments were carried out by A. Goetz
(a 28)
and his colleagues using a non-turbulent flow reactor. ' In addition to
various gases, monodispersed latex aerosols were introduced into the reactor
channel. They found that when an NO- and 1-octene mixture in air is irradiated,
the latex aerosols grow in mass and that the total added mass is linearly
related to the aerosol concentration. In the absence of external nuclei
they observed the spontaneous appearance of nuclei. Then they carried out
an extended study of various permutations of reactants and conditions and
found that in the absence of SO- the largest amount of light scattering
aerosols was produced at low humidity. However, they also found that
12
-------
Science Center
Rockwell International
SC537-10FR
aerosol formation depended upon the order in which they mixed the reactants
before irradiating. They corroborated the fact that S02 increases the
aerosol formation, but made the interesting observation that at low S02 con-
centration the aerosol formation is inhibited dramatically especially for
higher humidity.
Aerosol formation in a flow reactor was studied by Stevenson, Sanderson,
and Altshuller. Their experiments corroborated many of the observations
of the Stanford research group but also included a study of the effects of
mixing two hydrocarbons. The addition of trans-2-butene to cyclohexene
increased aerosol production by a factor of 10 when the two hydrocarbon
concentrations were equal but when the trans-2-butene concentration was
increased to several times the cyclohexene concentration, the effect began
(29)
to reverse. Additional observations have been made by Harkins and Nicksic
in which SO, produced aerosols contained no organic material.
(M
More recent work of Wilson and Levy is important in the motivation of
our work. They found that in a stirred reactor the aerosol formation rate
was increased by turning off the stirring. An important aspect of this
observation is that the stirring is not removing aerosol once it is produced,
but it inhibits the production of aerosol. This was shown by producing
aerosol without stirring and looking at its subsequent behavior with stirring.
In this case once formed, the aerosol lifetime is not reduced enough by
stirring to account for the decrease in formation with stirring.
Two recent series of experiments by Groblicki and Nebel and by Cox
and Penkett are very important in that they have established that S0_ can
be oxidized, by organic material, to form sulfate aerosols. In Groblicki and
Nebel's experiments, both the dark reactions of ozone and olefins and the
photolysis of NO and olefins, oxidize S09, in presence of H_0, to form
X t, £
aerosols. In the irradiated case the presence of NO inhibited both ozone, as
is known, and aerosol formation. Cox and Penkett studied the formation of
aerosols from S02 and H20 in the presence of products of ozone-olefin reactions in
great detail, concluding that these products are efficient oxidizing agents for
S02. They also noted that the presence of molecular oxygen was not necessary.
They have extensive rate data on consumption of ozone and olefins and for-
-------
nation of aerosol for various conditions and various aerosols. The aerosols
formed In these experiments by both groups were apparently pure sulfate
material. Groblicki and Nebel also produced aerosols with organic material
from ozone-olefin reactions but only in the absence of S02 and only for
certain olefins.
Although under certain conditions organic aerosols are produced in
smog chambers, few extended efforts to analyze them have been undertaken.
One noteworthy effort was made at Battelle in conjunction with a field
program. Gas chromatography-mass spectroscopy was used to analyze aerosols
produced in the photolysis of NQ-2 and either a-pinene, cyclohexene, or
1-heptene. In each case a few of the components were tentatively identified.
Clearly .a great amount of work remains to be done in elucidating the
nature and mechanisms of formation of aerosols in the atmosphere. Even under
controlled laboratory conditions experiments are not understood completely
and the extension of these results to the atmosphere is still tenuous.
2.3 Goals of Aerosol Simulation
Putting together the material in sections 2.1 and 2.2, we obtain a very
complex picture of the behavior of aerosols as they interact with gases. In
planning an experiment for the simulation of atmospheric aerosols, one must
extract from this picture a set of goals, based on hypotheses about the
nature of chemical transformations. The ultimate goal is the complete ex-
planation of the formation and behavior of aerosol particles in polluted
urban atmospheres. This must be achieved by certain subgoals, related to
individual properties of aerosol particles and the gaseous chemical systems
from which they form. Information must be generated which will allow a
reconstruction, from individual pieces of information of the overall aerosol
behavior. As a first step, we project the properties of aerosols to be
simulated and the potential systems which must be studied. This is not
limited to the specific work of the present contract but can cover more
general experimental programs in this area.
We will now specify individual properties of aerosols in photochemical
smog which must be accounted for in a simulation experiment.
-------
Science Center
Rockwell International
SC537.10FR
1) Photochemical nucleation of new aerosol particles.
Filtered nuclei free, polluted air will on exposure to sunlight produce
new nuclei. ' There is some evidence that this is an important mechanism
for certain short times in the atmosphere, particularly in early morning.
2) Growth of 0.1 to 1 y particles under photochemical1y active conditions.
This is believed to be a dominant effect, at midday, based on studies of
aerosol size distributions.
3) Presence of organic materials in aerosols.
A major component of photochemical aerosols is non-carbonate
. (12)
carbon.
4) Presence of sulfate in aerosols.
Sulfate.is an important component of aerosols even when the S00 con-
(12)
cent rat ion is low.
5) Presence of nitrate and ammonium in aerosols.
Both oxidized and reduced nitrogen are found in aerosols collected in
(12)
photochemical smog.
6) Large increase in relative humidity causes a large increase in
1 ight scattering.( *'
7) Liquid water is present in aerosols.
Liquid water content of aerosols is positively correlated with light
MM
scattering including cases examined in photochemical smog.
Many laboratories have produced data which may be related to these
properties. Aerosol production and growth in many different chemical
systems have been observed. The photolysis of NCL, air, and many olefins
both with and without water vapor has been shown to produce aerosols, as well
as the photolysis of S02,water vapor, and "relatively clean" air. Combinations
of these two principal mixtures also show an enhancement over each separately.
What is needed, to improve on and clarify all previous experiments, is data
detailed enough to specify more exactly the mechanism of aerosol production
in each of these cases and any other which can be identified. This will
provide a means of extrapolating to a variety of atmospheric conditions.
Thus the goal is to account for the aerosol properties outlined, as
well as any new ones identified as field work progresses, by detailed care-
-------
fully controlled study of specific individual mechanisms for aerosol formation
and growth. In the next section we will discuss the Rockwell International
approach to this problem under APRAC contract.
2.^ Approach of the Flow Reactor Experiment
In the presently reported effort we have started the development of
techniques to improve the study of the formation and composition of photo-
chemical aerosols. The most important feature of these techniques is the
use of a flow reactor to eliminate the walls from affecting the reacting
gases. In this way we also gain reproducibi1ity without preconditioning the
reactor. The other important feature is to employ every practicable method
to achieve a total material balance for the aerosol particles and for the
gas phase. We intend to carry this out for simple cases with, for example,
only one hydrocarbon present.
In order for the reactor to be truly wall free, it must satisfy certain
criteria. It must be true that no material produced at the walls is col-
lected or detected as a reaction product. It is also necessary that no
material, either initial or produced in the reaction, is lost to the walls
of the reactor, including both products and intermediates. This last point
is important in that the loss or destruction of an intermediate at the
walls may change the ultimate products. This leads to the last criteria
which is that no material produced at the walls react in the reactor to
remove any key component of the system. These criteria are collected in a
slightly different form in Table 3. In section 3.1 the constraints,
brought about by these criteria, for laminar flow in a tubular reactor, are
evaluated quantitatively.
An important reason for selecting a flowing reactor involves sampling
of the reaction products. A very large completely static reaction chamber
can conceivably meet the requirements set forth but any attempt to sample
it will produce convection and turbulence and thus reintroduce the walls as
an interference. In the flowing reactor the products are sampled before
material has had time to diffuse to and from the walls and the sampling can
be adjusted so as not to disturb the flow. The results obtained in this
manner will be reproducible without any reactor pre-conditioning.
16
-------
Science Center
Rockwell International
SC537.10FR
The satisfying of these criteria for a flowing system, however, requires
a very short residence time compared with the expected overall reaction times.
For example for the small tube (5 cm dia) used in this study, only a few
seconds of residence time are allowed for meeting these criteria. The
potential solution to these problems in studying the aerosol formation and
growth involves two types of what we shall refer to as time scaling. These
will now be defined:
a) Differential Time Scaling is the stepping to a later time in a
complex reaction by setting all reactant concentrations to those which will
pertain at the selected time including all intermediates and products.
b) Total Time Seal ing is the adjusting of the initial conditions
(concentrations, light intensity, temperature, etc.) of a complex reaction
so that the overall reactions will take place at a faster rate without
changing the final products, though possibly changing their relative final
concentrations.
Our study employs both of these methods simultaneously and their achieve-
ment will now be discussed. The differential scaling involves no intrinsic
error provided that al1 of the concentrations of reactants, intermediates and
products can be reproduced and achieved; once'it has been achieved, the
reaction proceeds just as it would have if the induction period which was
skipped had actually taken place. There are two problems with this method
however. The rate of reaction for the remaining portion of the reaction may
still be too slow for the time available. If this deviates only by a small
factor, then a study of the period of time available will allow one to then
proceed differentially to another point in time. If the complete reaction
takes place over only a few such steps, the reaction mechanism may be studied
and understood over this period. However, if this is not possible, one may
further resort to total time scaling on top of the differential scaling. The
other problem of differential scaling is that it may be difficult to know the
exact conditions at a later time in a complex reaction. Toward this end we
are helped, however, by the fact that most intermediates come into equilibrium
rapidly and therefore take care of themselves, so only the principal reactants
and products need be considered.
17
-------
Total time scaling is more complex and does not hold in general. It
should be immediately clear however that any complex set of reactions which
have a rate limiting step of non-zero order in any reactant may be scaled by
increasing the concentration of that reactant. If the rate limiting step de-
pends on several reactants, then higher order scaling is possible up to the
order to the rate limiting step. For example, if the rate limiting step Is
3rd order and each reactant in this step is scaled by a factor of m, then the
rate scales by a factor of m . Of course, the reactants in the rate limiting
step are not necessarily the initial reactants, but if the rate limiting re-
actants depend in some simple way upon the initial reactants, then the scaling
can be related to the initial reactants.
Since no complete model exists for photochemical aerosol production, it
is difficult to work out possible scaling relations for aerosols. However,
it is known that aerosols do not form until after the NO has been converted
to NOp and there is an appreciable 0, concentration. Thus in our experiments
we apply the concepts of differential time scaling and total time scaling as
a first approach to the problem keeping in mind the limitations of these con-
cepts. The concept of differential time scaling is applied by starting with
N0_-hydrocarbon system rather than a NO-hydrocarbon system. Since 0, reactions
with olefins produce aerosols (See Appendix B) and 0 reactions with olefins in
the presence of 0» should have similar chemistry (see section 4.12), the concept
of total time scaling is applied by increasing the NO- concentration and light
intensity (thereby increasing the 0 atom and 0_ concentration) and by increasing
the hydrocarbon concentration from the atmospheric concentration values.
We have also set as a goal for the experiments the characterization as com-
pletely as possible of all reaction products. This includes both gas phase and
aerosol products. In this way we eventually hope to characterize conditions
under which aerosols were both produced and not produced and to have a material
balance for these cases. Further when aerosols are produced, we hope to obtain
a material balance for the aerosol as well as a good physical characterization
of it.
18
-------
Science Center
Rockwell International
SC537.10FR
3. The Flow Reactor System
We will now describe the design principles of the flow reactor followed
by the experimental details of the entire reactor system. The underlying
goals of the flow reactor design are the elimination of wall interference,
the reproducibi1ity of the experiments, and the control over and ease of
changing experimental conditions.
3.1 Reactor Design Principles
A flow reactor inherently allows rapid change in experimental conditions
since the reacting mixture is flowing through the tube continuously. Provided
the reactor gives reproducible results, one is limited only by the
residence time in the irradiated section and the gas handling and flow develop-
ment time. The reproducibi1ity of the reactor is assured if the walls play
no measurable role in the results. This is assured if any product buildup on
the walls is negligibly small and does not affect the measurements. Thus a
fundamental goal in the design and operation of the Science Center flow reactor
is to prevent the walls from playing any role in the reactions studied. In
order to achieve this, one must analyze the possible effects of having a
finite volume reactor. In a static chamber the known interferences are
removal of reactants and products by the walls and the contamination by sources
at the walls. Both these effects are limited by diffusion so one can hope
to reduce the effects of the walls by increasing the volume to surface ratio
of a chamber. However, in most experiments, samples must be physically removed
from the chamber for analysis. In this case the "static" chamber is no longer
static and mixing is required to keep it homogeneous. A flowing reactor solves
this type of sampling problem and does not have to be mixed to remain homogeneous.
Diffusion to and from the walls remains the limiting interference problem.
In a fully developed laminar flow there is no eddy diffusion so only the much
slower molecular and Brownian diffusion must be considered. We will now discuss
these diffusion problems as they affect a fully developed laminar flow reactor
in which chemical reactions are taking place and in particular in which
aerosols are formed.
In order to understand diffusional problems in a chemical flow reactor
19
-------
we must consider both loss of material and appearance of interfering sub-
stances. Each of these cases has subcases. The loss of an initial reactant
will lead to an apparent increase in its reaction rate, while the loss of a
product will lead to an apparently lower formation rate. In the latter case,
gas phase products may be lost to the walls by molecular diffusion as in the
case of loss of reactants, but aerosols may be lost by the analogous Brownian
diffusion and also by gravitational settling. For the interference case
the appearance of contaminants from the wall may produce new products not
otherwise obtained from the experimental reacting mixture, or the contaminant
may react with and remove an initial reactant, a product, or a key intermediate.
In the sub-case of removal, the result may be an apparently faster or slower
reaction rate but it may also be the insidious removal of an intermediate by
reacting it to an otherwise present species. This kind of an interference
would be almost impossible to detect but could dramatically change the experi-
mental results. These various cases are summarized in Table 3.
All of these cases, except for gravitational settling of aerosols, are
described by the following partial differential equation for diffusion in
a steady
v(r) I!- (r |f) + R(C,r,z) (l)
dz r dr dr
where C = C(r,z) is the concentration of the species in question
D is its diffusion coefficient in air
r is the radial position
z is the axial position
v(r) is the mean velocity (in z direction only for laminar flow)
R is the chemical rate of production or removal of species in
question.
In general this equation cannot be solved in closed form due to the
non-linearities introduced by the term R(C,r,z). In fact for the typical case
in question in this program, this nonlinear term couples the equation to a
large set with one equation for each species. However if R = 0 or is
linearly dependent upon C (i.e. R = ± kC) , then equation (l) separates into
20
-------
Science Center
Rockwell International
SC537.10FR
Table 3
Wall Interference in the Flow Reactor
1. Loss to the Walls*
A. Diffusional loss of initial reactant leading to apparent higher
reaction rate.
B. Loss of product leading to apparent lower formation rate.
1) Diffusional loss of gas phase molecules
2) Brownian diffusional loss of aerosols
3) Gravitational sedimentation of aerosols
2. Diffusion Contamination from the Walls*
A. Appearance of material which produces contaminant products.
B. Appearance of material which reacts to remove key reactants and
J ntermediates.
A
Note: A reaction catalyzed at the walls falls simultaneously in both 1
and 2. The catalyzed reactants are lost and the products formed
represent a contaminant source.
21
-------
an equation for the z dependence and one for the r dependence. In both
cases the z equation has a simple solution. The r equation depends upon
the factor v(r). If it is constant, v(r) = v , then the resulting r equation
(32) °
is Bessels equation of order zero. If v(r) represents laminar flow,
o
that is v(r) = v (l - 7), where a is the radius of the tube, then the r
° a (33)
equation may be reduced to the Confluent Hypergeometric Equation. It
has been shown that for a fjow reactor assuming that v(r) = v only results
(32) °
in a small change in the solution for the case of R = 0.
In order to simplify the analysis of the wall problems we will make
some extreme simplifications of the problem to keep R = 0. These will
represent "worst cases" and will not be physically realizable. However we
will then be assured that the true problem is considerably less important
than the calculated one. Since for many of the entries in Table 3 the wall
effects will be unimportant, these overstatements of the problem will be
totally satisfactory. For the cases in which the wall effects produce
limitations, the stringent nature of the "worst case" assumptions must be
kept in mind.
For case 1(A) we will assume that the reactant must be present during
the entire residence time, but that it is trapped at the walls. In other
words it will be diffusing away for the entire residence time but its presence
as a reactant is most important at the end of the period. On the other hand
for 1(B) cases we will assume that the product is formed immediately and has
the remainder of the time to decay. Of course, if such a product is other
than an intermediate posing as a case 1(A) reactant, then the residence time
would just be made shorter. For subcase 2 we will assume that the aerosol
consists of a very small nuclei which only have an effect near the end of
the time while for subcase 3 we assume the aerosol has grown to be very
large initially. For both cases 2(A and B) we will assume a source at the
wall such that the concentration at the wall is comparable to the other
important intermediates or aerosol precursors but that it does not have an
effect until it reaches the end of the residence period. Now with these
assumptions we may quantitatively explore the limitations of the flow reactor.
These are highly unrealistic cases so the solutions will be bounds on the
22
-------
Science Center
Rockwell International
SC537-10FR
true behavior in the flow reactor. In some cases the change to more
realistic assumptions will be easy; in many, however, this represents a
difficult problem.
Once R has been taken as zero and v(r) = v , equation (l) may be
(34) °
separated and solved exactly. One obtains
C - I, Vo'V)""0""21 <2>
n=l
where J (x) is the zero order Bessel function aa is the zero of J (i.e.
o. no
J (aa ) = 0 and the A 's are to be determined by the boundary conditions,
on n
and the variable z has been replaced by t such that z = v t.
For case (l) we have as boundary conditions
C = 0 for r = a and t > 0
C = f(r) for t = 0 and 0 < r < a
(34)
Now the solution is
00 « 2^ J (ra )
The solution consists of a sum of modes determined by the zeros of J . Since
°2
a increases rapidly from one zero to another and we will see that Da.t is
already appreciable, then in a reasonable time all the higher modes will
drop out and only the a. mode will remain. Thus for case 1 (A) we will assume
that when the flowing mixture enters the irradiated portion of the reactor,
its initial distribution is given by f(r) = C J (ra.). Equation (3) now
reduces to
i
2
C = C J (ro.Je"00!* (k)
o o I
The first root of J is at aa, = 2.405 and a is taken to be 2.5 cm for now.
° 2 (36)
D will range from 0.05 cm /sec for a typical small organic molecule to
23
-------
0.282 for H.0(35) all in air at STP. On the axis (r = 0) J = 1 so C/C = e'^l1
f. o o
Thus for a one second residence time the loss ranges from k.5% to 23% for the
range of D specified. The loss is sharply dependent upon the residence time
growing to 21% for D = 0.05 and 5 second residence. However for increasing
diameter the improvement is dramatic since a is inversely proportional to
a . Thus for D = 0.282, t = 1 second, and a = 5 cm, the loss drops to 6.3%.
A series of these results are summarized in Table **. Once again it must be
noted that the assumption of total adsorption at the walls is stringent for
our experiments. The solution of the equations becomes more complex for all
other cases, except that of no adsorption, in which case diffusion has little
effect.
Case l(B.l) is somewhat more complicated than 1(A). If the product is
formed in a reaction dominated by a reactant which is adsorbed at the wall,
then it will be produced with an initial distribution proportional to J (rot,)
and the solution will be the same as that above. On the other hand if the
product is formed by a group of reactants uniformly distributed (i.e. no
adsorption at the wall for reactants) then higher modes (A ^ 0, n > 1) must
be included. Since the a 's increase rapidly with n the higher modes will
damp out quickly. For example, if the first mode has a 10% loss for a given
case,the next mode will have a ^3% loss and the third mode a ~lk% loss. For
the initially uniform case the first mode will represent only 67% of the sum
and the remainder of the terms will oscillate in sign making approximate
solution difficult. Thus we conclude that for a product initially uniformly
distributed and with a moderately large diffusion coefficient the higher
modes will all be lost so a loss of 33% must be added to those estimated above.
Fortunately the types of products which will accumulate in aerosols will be much
larger molecules and have small diffusion coefficients, but it is certainly
true that such products will be adsorbed on the walls.
Although the losses to the walls for extreme cases can be large, this
is not a serious defect in a flow reactor as long as the reactions being
studied are not too much slower than the rate of loss to the walls. Even for
very accurate rate measurements such losses are regularly accounted for in
measurements of rate constants in flow reactors. '
2k
-------
Science Center
Rockwell International
SC537.10FR
Table 4
DIFFUSIONAL LOSS IN A FLOW REACTOR
(See text for symbol definitions)
D at C/Co Loss
2
cm sec
0.282
0.05
0.282
0.05
0.282
0.05
0.282
0.05
cm
2.5
2.5
2.5
2.5
5
5
5
5
sec
1
1
5
5
1
1
5
5
0.7703
0.95^8
0.2712
0.7935
0.9368
0.9885
0.7216
0.9^38
23%
k.5%
73%
21%
6.3%
1.2%
28%
5.6%
Note D = 0.282 represents H-0 in air at 1 atm and
(35)
16°C and D = 0.05 is a typical ion molecular weight
hydrocarbon in air at STP. '
25
-------
For case l(B.2) the Brownian diffusion of aerosols, the molecular
calculations can be carried over exactly. For our calculations however we
have gone to the aerosol literature directly for the required solutions.
We find that for laminar flow in a tube,
n/nQ = 1 - 2.56y2/3 + 1.2y + 0.1 77y^/3 (5)
where n is the mean number of aerosol particles at the end of a tube
n is the number at the entry to the tube
DL /,v
y = -5 (6)
' R u
where D is again the diffusion coefficient
L is the length of the tube
R is the radius of the tube
u is the mean velocity of the flow
This solution is valid for small y which is applicable to aerosol behavior.
If we now define the axial residence time T as:
T 5 - (7)
2u
so that combining equations 6 and 7:
(8)
-k
If we now take n/n = .99 and solve equation 5 for y, we find y = 2.5X10
For our reactor R = 2.5 so T = 1.95X10 vD. To find D we note that
26
-------
Science Center
Rockwell International
SC537.10FR
where k is the Boltzman constant
T is the temperature in °K
B is the mobi 1 i ty
The mobility is given by
(, + A L + Q L e-br/4}
B -- L_ - r. - (10)
where r\ is the viscosity of the medium
& is the mean free path of molecules of the medium
r is the radius of the particle
A, Q, and b are constants with values
A = 1.246, Q = 0.42, b = 0.8?
Q -L
For 0.01 y radium particles in air at STP, B = 3.4x10^ so D = 1.4lxlO giving
T = 14 sec which means in 14 seconds only 1% of the 0.01 y radius particles
wi 11 be lost.
Particles may also be lost by gravitational settling. Similarly to the
(7ft)
above a standard expression is also extracted from Fuchs. Since it is
unlikely that this would be a problem, we will just state the result here
for completeness. For a \% loss under the same conditions as in the diffusion
cases and for 0.1 y particles we will have a residence time of 650 sec.
Now we move on the cases 2(A and B) which quantitatively we can treat
together. Once again the solution is equation (2) but with new boundary
conditions.
C = C for r = a and t > 0
o
C = 0 for t = 0 and 0 < r < a
(34)
The solution satisfying these boundary conditions is
.
C a L. a J, (aa )
o n=l n 1 n
27
-------
This solution can be expressed In terms of two dimension less parameters
2 2
Dt/a and r/a. Curves of C/C versus r/a for various values of Dt/a have
(Ik) ° 2 -
been presented by Crank. ' For C/C = 0.01 and v = 0, Dt/a = 0.05 so we have
^ _ O.OSa2
t n
for our range of D from 0.05 to 0.282 and for a = 2.5 cm we find residence
times of from 6 sees down to 1 second. Again, it must be noted that this is
a restrictive calculation. Any product molecule large enough to be important
in aerosol formation probably has a smaller diffusion coefficient than that
of our present range so less of it will reach the detector. More important
is the fact that if the product diffusing from the walls is sufficiently
important in producing aerosols it will have so reacted on its way from the
wall to the detector forming an aerosol which has a much lower diffusion
rate. Thus the figures above are really for gas phase contaminants and in
the case of aerosol production, or for that matter inhibition, they are more
limiting than necessary. Nevertheless for the feasibility study described
in this report, these contamination calculations were considered to be the
limiting case, for if results can be obtained with these limitations, then
relaxing them wi 1 1 only make the experiments easier.
The above discussion assumes a non-turbulent flow in the reactor. Two
conditions are required for such a flow. The Reynolds number for the flow
must be sufficiently small and the flow must have a certain distance over
which to develop. The Reynolds number for flow in a tube is given by,
where a is the radius
28
-------
Science Center
Rockwell International
1
SC537.10FR
p Is the density
v is the average velocity
n is the viscosity.
Since for laminar flow the velocity on axis is twice the average velocity
v = £/2t where £ is the length of the reactor, t is the residence time on
the axis then:
Now for air at STP
p = 1.2x10 gms cm
n = 1.8^5 poise (gms cm sec )
and if we take a = 2.5 cm, St, = 20 cm, t = 1 sec, which represents our
reactor,
R = 163
which is well below any critical Reynolds number.
For development of laminar flow a length £ is required given by
a = 0.227 aR
= 92 cm
So the conditions of laminar flow are easily established in a laboratory
flow reactor at STP.
It is interesting to explore how these quantities vary with changes in
reactor size and residence time. We noted above that for a given diffusional
contamination from the wall
29
-------
where y 's some constant. Thus, if we want to keep this criteria the same
we may increase the time by the square of the increase in the radius.
Our expressions for R and Si now become (on substitution for t)
R =
0.277
e 2n y
so the Reynolds number becomes smaller and the flow development length remains
the same. One can also examine the quantity of gas consumed for a larger
reactor. The quantity of gas flow Q is given by
where a, &, t are as previously defined.
P is the mean pressure but for t = ya2
- 7T£P
Q =
which is independent of reactor diameter.
Thus one can quite dramatically increase the residence time without
cost in quantity of gas or length of flow development. However as the
radius becomes larger, care must be taken to keep the reactor isothermal
since thermal convection could disrupt the laminar flow.
3.2 Detailed Description of Experimental System
The experimental system is composed of the flow reactor with its
irradiated section, a gas handling system, and a sampling and measuring
system. A schematic drawing of the flow reactor, gas handling, and sampling
systems used in this study of aerosol formation is shown in Figure 3> Figure
is a photograph of the system. A high degree of flexibility, afforded by
30
-------
MANIFOLD
GAS MIXING SECTION
OPTICAL FILTER & WATER BATH
RECIRCULATING SYSTEM
VACUUM
PUMP
ROYCO
225
TSI
3200
GE
PCNC
LOENCO
70
VAC.
PRESSURE REGULATOR
PRESSURE GAUGE
0.025U
1 AEROSOL
FILTER
co
o
en
co
DDO)
O O
S-f
is
=rO
Figure 3 SCHEMATIC DIAGRAM OF THE SIMULATION REACTOR SYSTEM
0)
a
o
-------
Figure 4. PHOTOGRAPH OF THE PHOTOCHEMICAL AEROSOL REACTOR LABORATORY
32
-------
Science Center
Rockwell International
SC537.10FR
utilizing a modular concept for the overall design, allows a variety of
changes in gas handling, flow rates, sampling methods, among others. We
will now describe the design and operation of the individual components of
the system.
3.2.1 The Reactor Section. Central to our entire experiment is the reactor
section itself. The gas mixture enters the reactor through the laminar flow
development section, a 5 cm diameter, 180 cm long stainless steel tube. The
dimensions of the tube are more than adequate for the development of laminar
flow at our operating conditions, of a total flow of less than five liters
per minute (see sec. 3.1). The reactor itself is a single unit of pyrex glass
tubing, 5 cm diameter and 20 cm long, surrounded by a torroidal jacket which
contains a temperature controlled light fiItering.solution.
The spent gas that has gone through the reactor is then vented into
the exhaust system. A diaphragm indicates the pressure differential between
the reactor and the ambient and a needle valve controls the vent rate and
therefore the pressure.
3.2.2 Gas Handling. Input to the reactor system is from a series of flow
control devices and filters which prepare the gas mixture in the desired
concentrations. The dry air is reconstituted from the laboratory central
supply of oxygen and nitrogen. The oxygen is supplied in tanks from Airco.
The nitrogen is obtained from the boil off from liquid nitrogen which is
also supplied by Airco. The oxides of nitrogen and hydrocarbons are obtained
as mixtures of an individual component in nitrogen from Matheson.. Typically
these are 2000 ppm but for some of the hydrocarbons like l-hexene it was 800 ppm.
The SO- was also obtained mixed with NL at 800 ppm supplied by Lihde. Water
vapor was added by diverting a part of the nitrogen flow through a bubbler which
saturated it with water. For higher molecular weight hydrocarbons a temperature
controlled saturator has been constructed. In order to perform experiments
without water vapor present the individual gases were passed through
dessicants. For N2, 02, and NO, molecular-sieve was used and for olefins
and W^, magnesium perchlorate was used. Six sets of rotameters, pressure
regulators and valves serve to define and measure the flow-rate and
pressure. Each rotameter with its own pressure regulator and pressure gauge
33
-------
is calibrated, in the system, with a Precision Scientific wet test meter or a
soap bubble meter depending on the flow rate of interest. The flow rates of
the individual gases not only determine the concentrations but also reactor
residence time.
From the manifold, through the mixing section, the blended gas stream
is then.fi1tered to remove any residual aerosol. Two filters were used and
were tested for effectiveness: a Gelman membrane filter capsule with a mean
pore size of 0.2ym and a Millipore mixed ester of cellulose with a mean pore
size of 0.025ym. The General Electric portable condensation nuclei counter
indicated low level of background aerosol count (~ 300 counts/cc) with or
wthout these filters suggesting residual aerosol content in the blended gas
is minimal compared to reactor generated aerosol count. Most of the experi-
ments were carried out without these filters.
The whole reactor system, including the gas handling system up to the
desiccants, can be evacuated. Two high vacuum pumps are used, one upstream
of the manifold and the other downstream of the reactor section. This
arrangement provides fast pumping speeds as well as flexibility in evacuating
the flow section or the gas handling section separately. Finally, in one
sampling configuration, the sampling probes themselves can be vacuum cleaned
after each sampling cycle to remove any residual contaminants left there
from a previous run. During the set of experiments described in Sec. ^.1.2,
the system was found to be reproducible, after complete evacuation, for the
duration of the G.C. measurements.
3.2.3 Irradiation. Three Hanovia 67^-A medium pressure mercury vapor lamps,
axially arranged in an aluminum housing, provide UV radiation for the flow
reactor. Table 5 gives the approximate radiated energy in watts for several
strong mercury lines in the spectral region of interest between 2900A and
5000A.
The light from the mercury lamps is collimated by a series of 1A washer-
like disks or baffles establishing a radiation zone with abrupt edges (see
Fig. 3). Any part of the glass reactor not in the radiation zone is painted
black to prevent light leaks. The glass probes are wrapped with black tape,
-------
Science Center
Rockwell International
SC537.10FR
Table 5
SPECTRAL ENERGY DISTRIBUTION IN THE REGION OF INTEREST
Mercury Lines Radiated Energy
(angstroms) (watts)
5461 (green) 34.0
4358 (blue) 29.0
4045 (violet) 15-9
3660 (UV) 40.5
3341 3.8
3130 21.0
3025 11.3
2967 6.5
2894 2.3
35
-------
thus defining precisely the boundaries of the reaction zones and the residence
time of the reactor. Figure 5 shows a photodiode measurement of the relative
light intensity on axis from each of the three lamps.
After the radiation is collimated, it is then filtered as it passes
through the torroidal jacket around the reactor. The transmitted spectral
characteristic is determined by the optical filter used. The pyrex glass
of the reactor itself cuts off at about 2900A thus eliminating the hard
ultra-violet. Cobalt sulfate or cobalt chloride solutions can be used as
bandpass filter from about 3000A to about 5000A. Cupric sulfate solution
gives a sharp cut off at the short wavelength side of about 3200A. Aquadag
is used as a neutral density filter and its spectral characteristic is
essentially flat. The approximate light intensity, in terms of the dis-
sociation of N0?, is estimated to be a minimum of kd = 0.02 sec . The
difficulties in arriving at it are discussed in section 4.1 below.
3.2.k Samp] ing. In analyzing the chemical and physical characteristics of
the reaction products, two types of samplers were used: a) grab samplers and
probes, which are defined as devices employing relatively short accumulation
times and b) collectors or filters, which require much longer sample accumu-
lation times. Sampling of all gas phase components for chemical analysis and
aerosols for physical characterization were done with probes and devices for
obtaining grab samples. Since dark reactions occur in the short time required
for the material to travel from the tip of the sampler to the specific analyzer,
probes must either eliminate or allow measurement of this effect.
Another constraint on the design of the sampling devices is the limitation
of the specific analyzer detection range. Some of the analyzers were designed
and built as air monitoring instruments and not as laboratory instruments
required to measure trace gas concentrations much higher than ambient. They
often also require high flow rates compared with what we needed for iso-
kinetic sampling. The sampler therefore must incorporate some sort of dilution
system.
Figures 6 and 7 show the first model dilution probe. The arrows indicate
36
-------
co
90
80
70
60
50
1 I 1 I
START OF COLLIMATOR
I I I I I f
i i r i r
1 'END OF COLLIMATOR
CO
o
en
co
1234567
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
RELATIVE PROBE POSITION IN CENTIMETERS
3) CO
O O
Figure 5 PHOTODIODE MEASURES OF THE RELATIVE LIGHT INTENSITY OF THE
THREE UV LAMPS
-------
co
CO
SAMPLE GAS STREAM
NITROGEN DILUTION GAS
Figure 6 SCHEMATIC DRAWING OF SAMPLING PROBE (DILUTION PROBE #1)
-------
Science Center
Rockwell International
SC537.10FR
Figure 7. PHOTOGRAPH OF SAMPLING PROBE (DILUTION PROBE #1)
39
-------
the diluent (dry nitrogen gas) flows to the tip of the sampling probe on
the outside jacket and is injected into the sample stream diluting it and
carrying it to the analyzers. By varying the length of the teflon tubing,
joining the probe and and the analyzers, and by adjusting the dilution rate,
dark reactions during the sampling time can be minimized and measured.
Figures 8 and 9 show the second generation dilution probe operating on
the same principle except that the dilution gas flows to the probe tip
through the center glass tube and is injected into the sample gas stream
through the six holes located on the back side of the arrow shaped Teflon
tip. This probe differs from the first in that not only can grab samples be
transported to the analyzers with minimized dark reactions but dark reactions
immediately downstream of the reactor radiation zone can be measured at those
reactor flow conditions as well as the dark reactions occurring between the
probe and the analyzers. The center glass tube can be moved back thus delay-
ing the dilution process further downstream in the probe and creating a
longer dark reaction time near the tip of the probe (slower flow rate) at
a higher reactant concentration (no dilution). This variability effectively
provides us with another piece of information on the faster initial dark
reactions at relatively higher reactant concentrations. Some of the results
are discussed in section *f. 1. The residence time in the irradiated section
can be varied from zero to the maximum, as defined by the total flow rate
through the radiation zone, by simply changing the probe position within
the reactor.
The chemical characterization of the aerosol requires the use of filters
and extremely long collection times. The choice of filter is based on
several considerations: (a) reaction conditions, i.e., materials and flow;
(b) collection efficiency and mesh size for the smallest size range of
interest; and (c) the ability of the filter paper to withstand chemical
reagents during the analytical process. For examples, Whatman filters have
a particularly high affinity for water adsorption. Gelman glass fiber
filters are very fragile tending to lose weight in handling but it can
withstand elevated temperatures and strong chemical reagents. Millipore
cellulose papers have pore size ranging down to 0.025 ym but are susceptible
kO
-------
m::' f /**'''&&&&&i. :::X '* ~~ --yy*
V.:.V...... . ..y.^^ .^. .AKK..^^ ............. ^/^/fiSJ^^ ' ' ^ X" '''^^ "^-""-- ^ / - >:*X-- ,.,,,.,.,,,.,,,,,,,» ^.^
SAMPLE GAS STREAM
NITROGEN DILUTION GAS
TO INSTRUMENTS
co
o
en
Co
O
T]
9
Figure 8 SCHEMATIC DRAWINGS OF SAMPLING PROBE (DILUTION PROBE #2)
Q)
=t
O
0)
-------
Figure 9. PHOTOGRAPH OF SAMPLING PROBE (DILUTION PROBE #2)
42
-------
Science Center
Rockwell International
SC537.10FR
to damage in some chemical reagents such as Isopropyl alcohol. Nuclepore
filter seems to suit many needs: (a) it has a well defined pore size and
pore shape, (b) it is chemically inert to many solvents, strong acids, and
bases, (c) it is extremely hydrophobic, and (d) the filter itself is strong
and can withstand repeated handling. However, the Nuclepore filter cannot be
elevated to extreme temperatures. We have used primarily the Gelman glass
fiber filters and the Nuclepore filters.
3.2.5 Measurement Techniques
Gas Phase Analyzers
For the inorganic analysis standard chemi1uminescent instrumentation
was employed in order to save the time of special instrument development. As
indicated above, this created sampling problems and in addition extensive war-
ranty servicing was required to bring them up to specifications initially. Overall
these instruments save some time and effort; it was not as great as we expected.
For ozone, a REM model 612 chemi1uminescent detector was employed, and for NO ,
/\
a Bendix N0-N02-N0 Analyzer was employed.
In the gas phase analysis of organic reaction products, the Loenco
model 70 Hi-Flex Gas Chromatograph with hydrogen flame ionization detector,
helium carrier gas, was used. Retention time analysis and peak identification
are discussed further in section A.1.2.
Physical Measurements of Aerosols
Three particulate monitoring instruments were used for the detection
and characterization of aerosols generated in the simulation reactor. The
instruments are: a condensation nuclei counter, a total mass monitor, and
an optical analyzer for larger size particles.
The condensation nuclei counter used in the General Electric model
PCNC-1 which detects a total number of particles nominally in the 0.005 to
0.1 ym diameter size range. It can detect as few as ~ 300 particles per cc
and up to ~ 10 particles per cc. The instrument sampling rate is two
cycles per second with a sample flow rate of 20 cc per second. The PCNC was
not well calibrated since only an order of magnitude indication was used.
For the larger size aerosol particles, the Royco 225 optical counter measures
particles in four ranges from 0.5 to 5 ym diameter. However, this particular
-------
instrument suffers from an inherent drawback of a lack of resolution near
0.3 ym diameter associated with the geometry of the light scattering cell
and unusual high background noise level. Within the limited resolution
of this instrument, optical-size aerosol concentration was observed to be
less than a few particles per cubic centimeter. It is anticipated that the
newly acquired Royco 220 interfaced with a multi-channel analyzer will
improve size resolution as well as lower noise background.
Total aerosol mass concentration of particles of size from 0.01 to
10 ym in the range of 1 to 100,000 ygm/m can be measured on the Thermo
Systems, Inc. system 3205. Time resolution of this instrument is one second,
although best results are obtained for the lowest concentration range and
for long integration periods of 100-200 seconds. Supplemental to the mass
monitor, the weighing of filter samples over a period of collection time is
also employed. The filter samples, however, were more important in the
analysis of aerosol chemistry.
-------
Science Center
Rockwell International
SC537.10FR
4. Experimental Results and Discussion
Since the entire reactor concept was novel and we anticipated
certain limitations and potential difficulties, moderate effort was
devoted to checking the gas phase reactions in the flow reactor. Due to
limitations of time these were not as extensive as they must ultimately be
in order to fully understand the reactor and its operation. They were
carried forward to the point that we had a qualitative understanding of
the data which was sufficient to allow aerosol experiments. We will now
discuss these experimental tests and the conclusions reached.
In the second part of this section we will discuss the aerosol experi-
ments including both the aerosol production tests and the collection and
chemical analysis of the aerosols produced.
4. 1 Gas Phase Chemistry
4.1.1 Inorganic Gas Phase Tests. The first experiments performed to test
the flew reactor were the measurements of ozone production in the photolysis
of N02 in air. The choice of this system was determined by the availability
of the REM ozone monitor and the difficulties experienced with the Bendix
N0x monitor. Later we studied the simpler photolysis of NO- in N_.
Using the REM instrument, ozone was measured in the flow reactor with
an NO- concentration of 20 ppm. The ozone concentration as measured was
only approximately 5 pphm. In addition the ozone concentration was found
to decrease with increasing light intensity. At this point the dilution
probe #1 was employed to check the effect of dark reactions on the measurement
of ozone. It was found that for increased dilution the measured ozone con-
centration increased, indicating that dark reactions were important. The
dilution effect however could only be fitted quantitatively by assuming the
concentration of NO was considerably higher than the 0_, at the sampled
point, in the reactor. This is not inconsistent with the decrease in ozone
with light intensity and suggests that competing reactions of 0 atoms with
N02 were producing NO. Further tests with a telescoping hood on the probe
to increase dark reaction time, supported this possibility but at the same
45
-------
led to the design of the second dilution probe.
These data suggest that there would be a peak in the ozone concentration
as a function of the N0_ concentration. The peak was found at about 5 ppm
of NO- and was about 12 pphm of ozone (at the REM). Under these conditions
the ozone concentration increased with light intensity, but this dependence
was very nonlinear, with a 10? increase in ozone concentration for double
the 1ight intensity.
While all these observations formed a consistent picture in terms of
qualitative behavior, the quantitative measurements were inconsistent. The peaks
in ozone as a function of N02 concentration and light intensity should have
occurred at values of these parameters 10 times those observed. At these levels
the ozone concentration would be comparably higher. From the measured ozone
concentration a value of kd for NO- of about 0.02 sec could be estimated
while for the presence of the peaks in ozone concentration values from
0.06 sec to 0.5 sec would be required.
In order to clarify this, and with the availability of the Bendix instru-
ment, the dissociation of N02 in N2 was measured. In this case the concen-
tration of NO reached an equilibrium value after traversing about 1/3 of the
illuminated region (8 cm). The slope in the linear region of variation of
NO concentration with light intensity indicated a value of kd of about
0.01 sec while the occurrence of an equilibrium required a value of the
order of 0.3 sec
These discrepancies remain a problem in the interpretation of the
experiments performed in the reactor.
Although uncertainty exists about the level of the light
intensity in the reactor, it is great enough to produce appreciable
quantities of ozone. At this point due to time requirements and
the feasibility nature of the study, as well as the fact that the aerosol
production was of utmost importance, it was decided that the completed work
on the inorganic reactions was sufficient to allow some aerosol formation
experiments. In add!tion» improvements in kinetic modeling underway will, in
the future, allow improved analysis of the body of data collected in these
tests and point to new experiments which will elucidate the problem in this area.
-------
Science Center
Rockwell International
SC537.10FR
4.1.2 Organic Gas Phase Tests/GC Measurements. The novelty of our experi-
mental approach required that, prior to searching extensively for aerosol
formation, we elucidate, if only qualitatively, some of the experimental
factors related to the known chemistry of the NO-, olefin, air irradiation.
The purpose of these exploratory investigations was to obtain in the time
available an understanding of the operational aspects of a gas chromatograph
coupled to a fast flow system. Simultaneously, we would determine whether
our experimental approach was consistent with the observations of others.
The problems of sampling reproducibi15ty and memory effects were considered
important from the beginning.
The gas chromatograph and sampling equipment have been described
already (Sec. 3.2). Only two aspects of these studies require further dis-
cussion: The column choice and conditions and specific sampling conditions.
It is our ultimate hope to minimize the number of columns used in the organic
analysis; this refers to both the gaseous as well as the condensable organic
materials. To achieve this, consideration was given to selection of a
column with a suitable polarity of the liquid phase as well as its ability
to withstand injections of water laden samples and reasonably high operating
(^2)
temperatures. Fortunately, Ottenstein and Bartley have considered such
factors and it is principally based upon their work that we chose to use a
column 1.5 m in length, with \Q% SP-1200 / ]% H3PO/» on 80/100 Chromosorb W AW
as the column packing. Although a complete test of this column has not been
attempted, our preliminary tests suggest better than satisfactory separation
of aldehydes (formaldehyde not tried), monobasic acids, alcohols, alkenes
and alkanes, with excellent chances for good separation of organic nitro and
nitrate compounds. For the results discussed herein, this column was used
at 100°C with a He flow rate of 60 cc/min and flame ionization detection.
The term "sampling conditions" includes conditions used for sample
collection to achieve adequate separation and peak shapes as well as conditions
necessary to eliminate memory effects." The best peak shape and separation
*
By memory effects we mean the extent to which an analytical measurement
depends upon the past history of the sampling components.
-------
were acheived by collection of the gaseous mixture from the sample probe
by pumping the mixture through a gas loop cooled to liquid Ar temperature
for no more than k minutes. The use of liquid Ar eliminated the problems
of poor peak shape caused by too large a pressure in the sample loop due to
condensation of 0_. Prior to sample injection and still with the liquid Ar
trap in place, the loop was evacuated and the He stream then passed through
the loop. Then flash vaporization was achieved with either a hot water
bath or a heat gun. Memory effects were eliminated by careful consideration
of the valves with respect to "dead" volumes, virtual leaks, and minimization
of 0-rings which require even the smallest amount of grease. It should be
pointed out that these problems should be given consideration to every region
of the flow system, for although the impurities introduce negligible problems
in terms of affecting the chemistry (relative to the gases added) their
presence may mask the appearance of important products.
The chromatograms produced in Fig. 10 were obtained for the same
reactor conditions with regard to residence time (4 sec) NO- (33 ppm) ,
hexene-1 (2.k ppm), N_ (750 torr) , and sampling procedure (4 min. collection
time). The top chromatog ram was obtained when N0_, hexene-1, N« (no 0_) flow
was sampled by blocking the irradiation from the reactor with an aluminum
foil shutter. Peaks C-F in Fig. 10 are due to impurities in the hexene-1;
their chemical nature was not established as such information was not con-
sidered necessary for the initial tests. Peak A is hexene-1 and presumably
peak B is hexanal. The origin of the hexanal, a photoproduct, is due to
small light leaks around the light shutter. When the same mixture was ir-
radiated two significant changes in the middle chromatogram were apparent;
peak B appeared strongly as well as E,, a new peak. When 0_ was added to
the flow (20%), the lower chromatogram was obtained. Peak B is observed
to almost disappear, and the new peaks, A., B»(?), B_, and F,(?) were ob-
served. Extensive and absolute peak identification using stream splitters,
efficient collection, and IR spectroscopy were not considered practical at
this time. Tentative identification was achieved by comparing known
retention times with authentic samples along with plots of the logarithm of
the retention volume vs. number of carbon atoms in a homologous series. In
-------
I DARK REACTION
IN N, ATMOSPHERE
II RADIATED REACTION
IN N, ATMOSPHERE
III RADIATED REACTION
IN 20% OXYGEH
co
o
en
co
o
TI
73
Figure 10 N02-OLEFIN GAS CHROMATOGRAMS
-------
this manner, peaks A, and B were assigned pentanal and hexanal respectively.
Peaks B. and E. were not identified; speculation concerning their identity is
given below.
For the purposes of discussing the results of the investigations, a
brief digression into the mechanistic details of the reactions 0( P) atoms
(Z,3)
with olefins is necessary. Cvetanovic has reviewed the primary reactions
of 0( P) atoms with olefins in 0_-free atmospheres as well as discussed the
3
similarities of the products of 0( P)-olefin reaction mixtures containing
molecular oxygen with those from the corresponding ozone-olefin reaction.
Thus, the irradiation of NO^-air-olefin atmosphere leads to some products
that are similar to those formed in (non-irradiated) 0,-olefin mixtures (in
air). This similarity is accounted for by postulating the following
reactions for the case of 1-hexene:
0(3P) + CH.CH2CH2CH2CH = CH2 + CH CH2CH2CH2CH-CH2 (l)
0.
(M)
CH,CH,CH,CH0CH-CH, -> epoxide (2)
J £. i e. . | £-
0.
-> hexanal (3)
-> decomposition (k)
+02 ^ CH3CH2CH2CH2CH-CH2 (5)
0 0
.2
formaldehyde
CH CH2CH2CH2CH-CH2 -> pentanal + CH^t oxide ) (6)
0 0
.2
pentanal
-> formaldehyde + CH3CH2CH2CH2CH02( oxide )
(7)
formaldehyde oxide (+02) -» (8)
pentanal oxide (+0-) -> (9)
50
-------
Science Center
Rockwell International
SC537.10FR
It must be recognized that the above is a condensation and oversimpli-
fication of many possible elementary processes about which theoretical
arguments may be given. However, such a discussion is not our purpose;
rather the reactions given In (l)-(9) are only to provide a basis for dis-
cussing our qualitative observations with the results of others. Equations
(l)-(4) constitutes the principal reactions which occur in the hexene-l-NO_-N7
mixture, i.e., no 0_. As already mentioned, peak B is hexanal while E,
although not identified, is approximately where an epoxide would appear on
this column (This is based upon a consideration of the McReynolds numbers for
SP-1200 column and a carbowax 20 M column. Carbowax 20 M is customarily
used for ethers and epoxides.)
Addition of 0_ to the reaction mixture introduces a competition between
(2) and (5) which under our conditions leads almost exclusively to (5) since
hexanal disappears with the concomitant appearance of pentanal. No attempt
at formaldehyde detection was made; for the purpose of these experiments its
presence was accepted. The absence of hexanal in the 02-containing mixture
in these experiments is analogous to the lack of propanal formation in the
corresponding propene, NO , air irradiation, carried out by Altshuller et al.,
J\ ^^~ ^^»
and to the very small amount (~ 1/30 of propanal) of n-butyraldehyde observed
(45)
in the sunlight irradiation of NO -air-butene mixtures by Ellis.
X
No attempt at the identification of peaks B. and B_ was made. Based
upon polarity arguments of the SP 1200 column packing, they may be postulated
to be some organic nitrate. The conditions of these experiments would appear
to be particularly conducive for peroxyvaleralylnitrate, pentylnitrate, and
peroxypentylnitrate to mention only three. Nevertheless, further speculation
concerning the nature of these peaks and therefore the reactions leading
them is not warranted without some evidence.
4.2 Aerosol Experiments
The aerosol experiments will now be discussed in two parts. First section
4.2.1 will outline some Science Center-funded ozone-olefin aerosol production
experiments. These ozone-olefin experiments are discussed at greater length
in Appendix B. Then the flow reactor experiments will be detailed in section
4.2.2.
51
-------
4.2.1 Ozone-Olefin Aerosols
As detailed in Appendix B, we have employed a simpler, non-"wall free,"
reactor to study aerosol production in ozone-olefin reactions. These ex-
periments provided a variety of information, of use in the flow reactor
experiments, relating to concentrations and residence times required for
aerosol production, and relating to sampling and chemical analysis of
aerosols.
The ozone-olefin experiments have shown that a few ppm of reactants
and 20 to kO seconds of residence time are required for production of light
scattering aerosols. Whilve these conditions are possible within the flow
reactor, they were not achieved in the experiments described below. This
was a major i,nput to the decision to move ahead to the addition of other
components (S02 and H_0) to the reaction mixture in the flow reactor ex-
periments.
The aerosols produced in the ozone-olefin experiments were also used
to test the efficiency of collection of the sampling probes used with the
flow reactor and to test the IR and GC-MS methods of analysis for collected
aerosols. These chemical analysis methods proved very useful in generating
information about the aerosols produced and are discussed in detail in
Section B.3-
k.2.2 Flow Reactor Aerosol Experiments
As a program for determining the conditions for aerosol production in
the flow reactor, a series of preliminary tests were undertaken. Several
sets of conditions in the N0_-olefin-dry air system were tried including
cases with a few ppm to hundreds of ppm of olefin and a few ppm to a hundred
ppm of NOj. Table 6 shows a summary of the conditions under which no aerosol
was produced. After aerosols were detected in the NO.-olefin-SOj-H.O-air
system, no further tests were made on the N07, olefin, dry air system. It
was decided to test the sampling and chemical analysis techniques on these
aerosols produced by the 1-hexene and N02 system with SO. and H_0 present.
In the NO.-olefin-SOj-HLO-air system the necessity of each component,
for aerosol production, was checked. Figure 11 shows the strip chart response
52
-------
Science Center
Rockwell International
SC537.10FR
Table 6
Conditions Under Which No Aerosols Were Produced in the Reactor
[NO-] [1-Hexene] [CL] Residence Time
ppm ppm % sec.
21 10
20 5
10 12
21 10
21 10
20 5
10 12
21 10
10 12
20 10
20 10
20 10
21 10
20 10
20 10
20 10
20 5
21 10
20 5
21 10
20 5
20 5
21 10
21 10
6
8.7
10
10.5
13
13
17
17
17.7
20
20
20
20
20
20
20
22
30
51
54
70.6
104
189
197
50
~ 678*
4.6
50
50
~ 678*
4.4
50
8.7
4.4
10.8
26
50
61
78.1
114.1
~ 678*
50
~ 678*
50
~ 678*
~ 678*
50
50
Rotameter by-passed
53
-------
LAMP #2 OFF
LAMP #1
OFF
O.ON
02 OFF
N02 ON
N02 OFF
H20 ON
S02 ON
S02 OFF
I I
1-HEXENE ON
10 /cc F.S.
FIG. II ' CNC STRIP CHART RECORDING: 10 NUCLEI/CC FULL SCALE
(TWO MERCURY VAPOR UV LAMPS)
-------
Science Center
Rockwell International
SC537-10FR
to the output from the condensation nuclei counter at different points in
time during a reactor experiment. The sharp response of the CNC when each
gas is valved off and turned on again implies a strong dependence of the
aerosol production rate as a function of the concentration of each individual
gas component. Finally, the drop in aerosol count as the UV source was
halved and then turned off implies that the aerosols produced are photo-
chemical in nature. Reactor conditions for various experiments are shown
in Table 7'
The aerosol generated were predominately in the submicron size range.
The General Electric portable condensation counter indicated counts of 10
nuclei per cubic centimeter (maximum reading) or greater. The Royco 225
optical counter did not indicate significant particle concentration above
the noise level of a few particles per cubic centimeter. Data from weighed
filter samples (0.1 p Nuclepore) indicated mass concentrations of about
80 pg per cubic meter in the reactor under conditions shown in Table 7 for
filter collected samples.
4. 3 Chemical Analysis of the Aerosol Samples
Since the aerosols were formed in the presence of S0» and water vapor,
we first sought sulfate in the aerosol samples. In addition, as previously
planned, both nitrate and organic analyses were carried out. Since a large
amount of sulfate was found in the samples, it was expected that the organic
component would be small, so only a total organic measurement was made.
The sulfate analyses were satisfactory, but some problems arose with the
analysis for nitrate. Only an upper limit on organic content was
determined.
4.3-1 Sulfate Analysis
Typically, it was found that for a 16 hour reactor run, about 50 to
150 micrograms of aerosol were collected, and it was anticipated that the
sulfate content would be in the high nanogram to low microgram range.
An analytical method was therefore needed which has the sensitivity to
determine this amount of sulfate. The classical method for determination
of small amounts of sulfate is the turbidimetric method. There was some
55
-------
Table 7
CONDITIONS USED FOR AEROSOL PRODUCTION FROM S02> H20, N02> & AN OLEFIN IN AIR
Nuclei Production Filter Samples for
(1) (2) Chemical Analysis
[02J 18* 18% 18%
[H20j Relative Humidity 50% 50% kk%
[1-Butene] - 15 ppm
Il-HexeneJ 18 ppm - 75 ppm
[NO_J 20 ppm 20 ppm ^7 ppm
[S02] 8 ppm 8 ppm 7.7 ppm
Residence Time (average) A sec *» sec A sec
56
-------
Science Center
Rockwell International
SC537.10FR
question as to whether this method would have the sensitivity and accuracy
required. Another method which has been proposed by Scarengel1i involves
volatilizing sulfuric acid, reducing it to S02 and measuring the S0_ with a
flame photometric detector. This method is stated to have a sensitivity in
the nanogram range, but the method was not found to be workable as described.
It has been learned recently that the method can be modified and simplified
(47)
to give adequate results.
Previous experience with the method of Fritz suggested that this
method might be modified to give the sensitivity, analytical simplicity and
accuracy desired, particularly if the amount of sulfate present was 5 yg or
above. This method involves the titration of sulfate with barium perchlorate
in 80% isopropanol solvent using thorin [2(2-hydroxy, 3»6-disulfo-1-naphthyl
azo) benzene arsonic acid] as the indicator. To decrease the detection
limit and improve the titration, the final solution volume was reduced to
near 5 ml, the end point was obtained by comparison against suitable standards
with application of appropriate indicator blanks, and a micro buret was
employed which allowed adding increments of 5X10 mill-equivalents of barium
ion (equivalent to 0.5 yg sulfate). The smallest amount of sulfate which can
be determined by this technique is approximately k yg, with an uncertainty
of about ± 1.5 yg- If a spectrophotometric titration is used where the
absorption peak of the barium thorin complex at 5500A is employed for end
point detection, the detection limit can be decreased to about 2 yg with an
uncertainty of ± 0.5 yg using a 3 cm path length, and probably lower with a
longer path length cell.
The presence of cations other than hydrogen ion can interfere in this
titration. The interference is less with the spectrophotometric titration
since some ions (e.g. Na ) only react with some of the indicators to give
an interfering color. This can be compensated for spectrophotometrically.
In the aerosol samples, no cations other than hydrogen ion would be expected
except as impurities and the greatest source of these would be the filters
employed. Consequently, the papers were washed thoroughly before use. Anions
also can cause some error in this titration because of coprecipitat ion.
From the work of Fritz et al., the presence of a 6 fold excess of nitrate
57
-------
(an anticipated aerosol component) over sulfate will cause an error of
about 5% in the titration. No other interference appears likely. It is,
therefore, expected that the accuracy of the titration is well within 10%.
While complete recovery of known amounts of sulfuric acid was possible
when added directly to the solution, recovery was very poor when the acid
was placed on glass fiber filters (Gelman A) and extracted off the paper.
We found recoveries which varied from almost 0 to 8Q% depending on the paper
(46 49)
lot and level of sulfate. This is in agreement with the literature '
and has been attributed to the presence of "alkaline sites" on the paper.
We were not successful in finding a simple acid washing procedure to
neutralize these sites effectively, although we did not try some of the more
(46 49)
elaborate techniques which have been proposed. ' In order to avoid
this problem with glass fiber filter papers, we employed Nuclepore papers
as the filter media. Sulfate recovery was quantitative using these papers
and very little blank or cation impurities were found. The papers, however,
were still washed before use as a precautionary measure. In the procedure
the filters were extracted with 3~1 1/2 ml portions of 80% isopropanol
solution to remove the sulfate for subsequent determination.
Several preliminary analyses of 16 hour reactor runs showed quantities
of sulfate in the 20 to 30 yg range. Since this was considerably more
sulfate than required for analysis, the time of the run was reduced in the
interest of time economy. Two 3 hour reactor runs were made under the
following conditions. Concentrations of oxygen, 1-hexene, nitrogen dioxide,
and sulfur dioxide were 18%, 75 ppm, 47 ppm, and 7-7 ppm, respectively,
with 44% relative humidity and a residence time of about 4 seconds. In the
first run, the aerosol was found to contain 18% sulfate and in the second
it contained ]6%, A run made under the same conditions, but with the UV
light of the reactor off, showed no sulfate on the filter paper. These
limited data show the feasibility of making detailed correlations between
reactor conditions and aerosol composition.
58
-------
Science Center
Rockwell International
SC537-10FR
^.3-2 Ni trate Analysis. Since the participation of nitrate in aerosol
formation is an important aspect of this program, various methods were
considered for the analysis of microgram aerosol samples for small amounts
of nitrate. The standard method using 2,k xylenol appeared to be rather
complicated and to lack the sensitivity required. The use of chromotropic
acid and 1-amino pyrene are described as having sensitivities which
were expected to be suitable (~ 1 pg N0_) but our evaluation of these methods
showed them to be highly irreproducible and very dependent on small details
of the procedure. In using chromotropic acid the absorbance of reagent peaks
without nitrate was found to be excessive.
The wide experience of Science Center personnel with polarographic and
pulse polarographic techniques allowed evaluation of various modified
polarographic techniques for analysis of microgram quantities of nitrate.
(cM
It was found that the method of Kolthoff, Harris and Matsuyama which
involves the catalytic reduction of nitrate during the reduction of uranyl
ion could give a sensitivity to nitrate of about 2 jag NO, if a derivative
(53)
pulse polarographic procedure was used. The procedure devised was
exceedingly simple and involves extraction of sample from the paper, placing
it in a 10 ml volumetric flask, adding 2 ml of uranyl ion "activator" and
taking the derivative pulse polarogram. The activator solution pulse polaro-
graphic peak height is subtracted from the peak height of the sample polarogram
and this difference is related to the concentration of nitrate through a
previously derived linear calibration plot. Very few interferences exist in
this method. Concentrations of sulfate 50 times that of nitrate have no
effect but very large amounts can decrease the nitrate current somewhat.
(See Kolthoff and Lingane for further discussion of the classical method.)
The method is sensitive, accurate and precise and is much simpler and more
rapid than the colorimetric methods if a pulse polarograph is available.
The polarographic technique does not directly distinguish between nitrate
and nitrite. In other words both nitrate and nitrite are catalytically
reduced by uranyl ion. However, since different numbers of electrons are
involved in the two reductions, the amount of nitrate and nitrite separately
can be obtained by taking the polarogram before and after nitrite oxidation.
59
-------
The ratio of nitrate to nitrite in our samples will be determined in the
future.
Preliminary analysis of filter papers from reactor runs, showed rather
large amounts of nitrate on the filter. However, when two filters were
placed in the probe back to back, a considerable amount of nitrate was
found on the back filter. Nitrate was also found with the UV light off and
even with no N0_ in the gas mixture. Quantitative determinations of nitrate
showed that a 6 hour run without N0« give 18 yg NO, on the filter, a 6 hour
run with only N- gas flowing through the reactor gave k yg of NO- on the
filter and a filter paper put through all the handling procedures of the
above samples showed about 2 yg of nitrate. The uncertainty of all these
measurements was ± 1 yg.
The explanation for this behavior is not apparent, but should be
readily explainable with a few additional experiments. There are three pos-
sibilities which can be advanced. Perhaps most likely is that the walls of
the probe used to collect the sample on the filter papers (see reactor
design) is contaminated from previous experiments and the water vapor in the
gas mixture is desorbing or otherwise removing nitric acid. The nitric
acid is then adsorbed on the filter paper surface. The fact that nitrate is
generally found on the walls of smog chambers lends some credence to such a
hypothesis. The second hypothesis involves the possibility of HNO_ or N0«
(or other nitrate former) contamination in one of the gases used in the gas
mixture. The third possibility, that of contamination from the reactor
walls, seems less likely since calculations indicate that material from the
reactor walls should not be captured by the probe because of the laminar flow
maintai ned.
4.3-3 Total Organic Analysis
The experimental procedure in the effort to detect and measure the total
organics or the total hydrocarbons was to compare filter samples collected
from the reactor with and without UV irradiation, the non-irradiated sample
being the background signal. However, we still have the problem of dif-
ferentiating the "aerosol signal" from the organic gas phase reaction products
adsorbed on the filter substrate. Therefore, it was decided to collect
60
-------
Science Center
Rockwell International
SC537.10FR.
aerosol samples using tandem filters. The front filter sample would have
both the adsorbed gas phase reaction products and the collected aerosol. The
second filter should have only the gas phase products if we assume the front
filter is efficient in removing particulates from the gas stream.
The hydrocarbon detector used for the analysis was the hydrogen flame
ionization detector also used with the Loenco Model 70 gas chromatograph.
A filter sample was placed in a helium atmosphere inside a hermetically
sealed container which was then heated to 150°C for thirty minutes. Material
which is volatile at 150°C was driven off the filter paper and flushed into
the flame ionization detector by the helium carrier gas. A capillary tube
served as a flow restriction to prevent the volatilized gases from blowing
out the hydrogen flame but otherwise no column was used. The whole length
of the sample path from the scalable container, through the valve, to the
detector was heated to prevent sample recondensation.
Although the FID has a satisfactory sensitivity at low sample con-
centrations, in the ppb range, the overall measurement sensitivity is sig-
nificantly impaired by large background level masking the sample signal.
As the hermetically sealed container is heated to high temperatures, any
residua] contaminants from the stainless steel fittings, valves or Viton
0-rings begins to be liberated and interfere with the analysis. Adsorbed or
absorbed contaminants also are liberated from the glass fiber filters them-
selves.
A known weight of hexadecane (9 ugm) was placed on a filter and
carried through the extraction procedure. Relative to this calibration
the organic content of the aerosol samples is equal to or less than 1.5%.
k.k Discussion of Aerosol Results
We have shown that aerosols can be formed, under at least one set of
conditions in the flow reactor, which has some connection with actual
atmospheric aerosols. In addition we have tried other systems and not
observed aerosols under conditions in the flow reactor. The sulfate
aerosol formation which we have observed has already been studied ex-
tensively and is not the most interesting case for study in the flow reactor.
However, there are interesting aspects of it relating both to the gas phase
61
-------
oxidation of S02 and the growth of the nuclei which may be elucidated in the
flow reactor. The data from the ozone-olefin experiments indicate that a
longer residence time will probably be required for organic aerosol production
in the NO -olefin-dry air case. As we have seen in Section 2 the "wall free"
J\
criteria can be achieved for longer residence times by modification of the
reactor.
The data obtained for the formation of the sulfate aerosols are con-
sistent with the following mechanism:
J
c
S02 + R02. + H20 -» H2SO/t + RO-
n*H2SO^ (n*H2S04)cluster
JH
n ,m 2 A n+1 ,m
V
J2+
n ,m 2 . n ,rrri-l
J2-
Where J is the rate of formation of condensable material (HjSO, ) , n* is
the critical number of molecules for a growing nuclei, J* is the rate of
formation of such nuclei and Ji.> ^i_> ^o+» ^o_ are rates °f gain or loss
of one molecule of either KLSO; or H_0 from a particle A containing n
242 n ,m
molecules of H?SO, and m molecules of HO. Although there are potentially an
infinite number of species and, even for our cases, a very large number,
mathematical techniques are available to analyze such systems of reactions.
In addition thermodynamic data can be employed to estimate J.+ and J as
well as J.,.. The analysis of this mechanism in the flow reactor should prove
useful as a preliminary to unfolding the more complex mechanism expected
for the organic aerosols. The experiments carried out so far for filter
62
-------
Science Center
Rockwell International
SC537.10FR
collection of the sulfate aerosol have been carried out at high concen-
trations to increase the amount of material collected (see Table 7) in
order to validate the chemical analysis techniques. These must be repeated
at lower concentrations to check the variation with concentration.
63
-------
5. Summary and Recommendations
We will now summarize the material presented in this report as it
applies to three tasks set forth in the APRAC contract. Conclusions from
the data and recommendations for further work will be included.
5.1 Aerosol Production
In the flow reactor with a k second residence time and 10-20 ppm each
of N02, 1-hexene, S0_, H-0 in air aerosols were formed. These were detected
by a condensation nuclei counter and collected on various filters. Aerosols
were sought but not produced under a variety of conditions in the system
including N0_, 1-hexene, and dry air. These experiments however were not
exhaustive. ,
For the case of S0« present, it was shown that all components including
0_ and light were required for nuclei production. The aerosols were col-
lected with, among others, Nuclepore filters which had 0.1 \i pores. However
they were not detected by an optical particle counter suggesting that they
were probably much smaller than 0.1 y and were collected by diffusion in the
pores of the filter.
In a separate, Science Center funded, series of experiments, aerosols
were produced by ozone-olefin reactions in a simpler reactor.
5.2 Chemical Analysis
Both aerosols produced in the flow reactor and in the ozone-olefin
experiments were chemically analyzed. Those formed in the presence of S02
were found to contain about 20% S07 and negligible (less than 2 ygms in 50-
100 ygm sample) organic material; the analysis for NO, was not definitive.
While NO- was found in the samples, possible paths for contamination were
determined- The presence of NO, in the filter sample does not mean it was
present in the aerosol but that it may have condensed on the filter or the
filter holder directly from the gas phase or may even be left on the filter
holder from previous runs.
Aerosol from the ozone-olefin experiment was used to evaluate the
infrared and GC-MS methods. This aerosol was found to be a highly-oxygenated
oxidizing compound or compounds. The IR spectrum had strong similarities
-------
Science Center
Rockwell International
SC537.10FR
to aerosols formed in smog chambers and collected in the atmosphere. The
principal identified products in the GC-MS analysis appeared most likely to
be decomposition products. In addition iodimetry and polarography were used
to study the oxidative nature of the aerosol samples.
5.3 Data Analysis and Feasibility
The production and collection of aerosol in the irradiated N02, SO.,
HLO, hydrocarbon, and air system has shown that it is feasible to study
aerosol production and composition with a fast flow reactor. This was
carried out over a large range of conditions (Table 7) which should allow
for extrapolation to atmospheric levels. The data we have gathered are
consistent with a simple mechanism, presented in section 't.A, although it
is as yet insufficient to determine any details of such a mechanism.
The preliminary work on the NO , olefin, and dry air system while it
has not produced aerosols has shown that aerosol production should be possible
in a fast flow reactor. The residence time however as shown by the ozone-
olefin reactions may have to be somewhat longer. In addition very complete
knowledge of the inorganic chemistry will be required to achieve the proper
conditions for aerosols in the reactor. Progress has been made in modeling
the gas phase kinetics necessary for understanding this problem completely.
5.4 Recommendat i ons
Considering the accomplishments already achieved under this program,
we recommend further work on the N0_, olefin, and dry air system. This
should begin with a systematic study of the gas phase reactions in the
reactor to check the possibility of generating conditions under which
aerosols can be produced. Kinetic modeling should play an important role
in this part of the program. It is possible that modification might have
to be made in the reactor system to achieve the stated goal. The range of
possibility for this, within the "wall free" design criteria, were dis-
cussed in Section 3- Further development of aerosol models, such as that
of Section k.k, for the organic aerosol case should be undertaken as
part of this program.
65
-------
APPENDIX A. Modeling, Mechanism and Simulation
Of central importance to understanding experiments undertaken in the
flow reactor and applying the results to real atmospheric problems is our
ability to unfold the mechanism of aerosol formation and successfully model
it together with the gas phase kinetics. In this appendix we first discuss
an example of scaling of a complex reaction and then go on to details of
improved concepts in modeling both aerosol formation and gas phase kinetics.
A.1 Scaling of Reactions
As discussed in Section 3, in order to have an atmosphere simulation
reactor in which the walls are unimportant, it is necessary to go to short
residence times compared with normal atmospheric time scales. Two types of
scaling are possible. In the first, the concentration of reactants and the
light flux may be changed so that the entire part of the reaction of interest
takes place during the residence time in the reactor, thus achieving an
overall time scaling. The second is a differential approach wherein, by
changing reactant concentrations to approximate those in the atmosphere, and
possibly by adding products, one may examine a short period of time during
the overall reaction. To accomplish both of these alternate approaches for
certain gas phase mixtures and assumed mechanisms, simplifications and as-
sumptions have to be considered. These are described, and an extension of
these ideas is made to aerosol formation to show how the data taken from
scaled reactions may be interpreted.
To illustrate the basis of the gas phase scaling, the Friedlander-
(22)
Seinfeld model of the kinetics of photochemical smog is used. While this
model includes only a few reactions, and is obviously not a complete model of
photochemical smog, it is very useful for the present purpose. It concerns
itself specifically with the NO -hydrocarbon system which will be central in
X
our studies. The reactions considered may not be complete, but they represent
a plausible mechanism that is general enough to include the possibility of
some variation. The gas phase reactions included are:
66
-------
Science Center
Rockwell International
SC537.10FR
N02 + hv + NO + 0
0 + 02 + M + 0, + M
0- + NO + N02 + 02
0 + RH -> R- + products
0, + RH -* Products
NO + R. -> N02 + R'
N02 + R« -> Products
where RH is some hydrocarbon, R* Is a generalized free radical (which
includes R« , R0«, R02', etc.), and M is a third body.
By a series of approximations Friedlander and Seinfeld reduced the
kinetic equations for this system to:
dlNOj
= [N00][RH]{a[NO]-X[NO.]} (A.I)
TTZ i'^^oJ L"»liJ H-l|."VJ n. inu«
-= -a[N02J[NO][RH]
= -[NO][RHj{0+y/[NO]}
[0] = Y[N02]
[NO]
[R-] = k^'lRH 10]
where a, 3, y> ^> V, 6> and ki ' represent various groups of rate constants
of the elementary reactions; a, g, and y also depend linearly upon the light
intensity.
67
-------
An important observation about the results of the Friedlander-Seinfeld
model is that the 0, 0,, and R- concentrations were arrived at by use of steady
state approximations. It is assumed that these concentrations adjust to the
values given in a time short compared with other characteristic times in the
reactions. For example in the above model the oxygen atom concentration
reaches ylNO-] in microseconds when the time scale for significant changes
in reactants is minutes. An important advantage of this model Is that a
partially analytical solution of the system of rate equations has been found.
In this solution the N0« and RH concentrations are expressed as functions of
the NO concentration and time is expressed as an integral of a function of the
NO concentration. This integral may be easily solved numerically and has
analytical solutions for various limiting cases.
We can use the model to demonstrate how reaction times may be scaled in
a simulation experiment. The Integral for time in the solution described
above may be written in the form,
1
fi
t = TV f(x,v,6,K)dx
[NO]/[NOl
where
v = [NO,] /[NO]
i o o
T -
6, K are parameters depending upon the initial concentrations and the
rate constants, and a is a rate constant grouping which is linearly propor-
tional to light intensity. The parameters <5 and K are important only for
r*s
long times. We may define a dimensionless time t = t/i. Now the actual
***!
reaction time t in the system may be scaled by the parameter T, since t
will be approximately constant if v is kept constant. Numerical tests have
shown that K and 6 are completely negligible for times such that the reactions
have gone one half way to completion (beyond the peak in N02 concentration).
68
-------
Science Center
Rockwell International
SC537.10FR
Suppose a concentration scale factor m is introduced. Scaling is achieved
by multiplying each of the factors in T by m. Then T is changed by a
factor of m , or for m = 10 the reactions are accelerated by 1000.
If the mechanism in the Friedlander-Seinfeld model included all possible
reactions in the 09-NO -hydrocarbon photolysis, then our work would be done
^ X
any any value of m would be acceptable. Howeverr many possible reactions
are missing and some of these will be higher order in the concentrations
we are varying than the ones which have been included. Thus there are
limits on the value which m may take on and for some conditions it might
be better for us to have different multipliers for the different factors in
T. Assuming that each of the three factors in T is multiplied by m, we must
/
estimate the initial changes in concentration of the various reactants. From
2
equations (A-1) we see that [0], [0_], and [R-] will be multiplied by m ,m
3
and m respectively. Thus the most important species to consider first are
R«; the only removal mechanism included in the model is the reaction of R-
with N00; however, radical-radical recombination may also become important.
2 4
The R'+NO,, reaction will increase in rate by a factor of m but radical-radical
6
recombination will increase by m . Unfortunately, rate constants for these
reactions are not generally known, but it is not unreasonable to set them
equal. Using the numerical estimates from the paper of Friedlander and
(22) -1 -1
Seinfeld and an estimate of 10 ppm min for the recombination rate, we
get a ratio
So with respect to this reaction with an m of 10, which multiplies this
ratio by 100, the neglecting of radical-radical recombination will still be
a good approximation, but as m get larger than 10 the neglected reaction
rapidly becomes comparable to the reaction in the model.
Another similar consideration is the comparison of other oxygen atom
/,
recombination rates which increase by a factor of m with the formation of
2
0_ from 0+02 which only increases by a factor of m . In this case, however,
69
-------
several reactions of 0 atoms with oxides of nitrogen will scale as m and have
been considered. The most important of these is 0+N02 -> NO+0-, which with
m = 10 has 4^ of the rate of the reaction with 02< The ozone reactions as
well as reactions between oxides of nitrogen do not appear to create any
problem since all the excluded reactions scale with the same low powers of
m as the included reactions. This means that m of the order of a few
thousand is possible which will reduce reactions taking 15 minutes to one
hour at the 1-2 ppm level to a few seconds at higher concentrations. It is
important to observe that changing the times by such large factors does not
just multiply all the concentrations as a function of time by a fixed amount.
Rather, the ultimate products remain the same, but possibly in quite dif-
ferent relative amounts. The set of important elementary reactions, however,
remains the same and once verified may be used to extrapolate the results
back to the concentration levels desired.
Another method of "scaling" is to make a differential measurement.
That is, reactions may be carried out over short times in the flow reactor
in a sequence such that their summation generates concentration - time curves
equivalent to processes over a much longer time interval. This method has
the drawback that it requires not only introducing starting reactants but
also all intermediates and products that are expected to react further.
Although this appears to be an almost impossible task, if we look back at
the Frtedlander-Seinfeld model we employed for overall scaling, we can
make an interesting and very useful observation. The model reproduces the
important features of the photochemistry by employing steady-state approxi-
mations for all intermediate species including products such as ozone which
also takes part in further reactions. If the dominant intermediate reactions
can be approximated well by steady state assumptions, then all that Is
required Is to start the reactants in varying concentrations, and the inter-
mediates will quickly come to their appropriate steady state values In a
time short compared to the period of time being studied. If an Intermediate
does not come to equilibrium In a time short compared to the residence time,
then It must be added to the reaction mixture. This may be true for ozone
under some of the conditions used In this study. One may then step through
70
-------
Science Center
Rockwell International
SC537.10FR
a reaction system differentially by either predicting concentrations as a
function of time and checking the consistency of differential rates measured
or by proceeding with each step based upon the results of the previous step.
In the experiments described in this report a combination of the two
methods was used. The reactants are set ahead to the point where there is
no NO present and the remaining concentrations are increased to speed up
the resultant reactions. A potential problem with this procedure is that
the scaling described above works best in the early stages of the smog
reactions when NO is still present. In addition, aerosols are not included
and do not fit well into such a simple model.
A2. Extrapolation of Laboratory Experiments to Atmospheric Transformations
The modeling of physical processes historically has been a key feature
in the development of engineering methods, and it is particularly useful in
extrapolating from a model system to a prototype involving fluid dynamics.
Model-prototype dynamics also has been useful in studying atmospheric
processes involving air motion. The exact duplication of a prototype by
a model can be achieved by proper scaling of all relevant parameters required
to characterize the behavior of the system. However, many models have use-
fulness with only partial scaling of key parameters between prototype and
model. Examples of simple modeling are discussed in Bridgman's classic
lirkh
(58)
book and in Birkhoff's monograph. Atmospheric applications have been
reviewed by Hidy.
The modeling of aerosol formation by chemical reactions in the atmosphere
should be feasible, provided key dynamic parameters can be identified and
scaled such that a simulation can be realized. At this stage a detailed
modeling scheme based on knowledge of the kinetics of atmospheric aerosol
formation is difficult to project quantitiatively. This is particularly true
with the uncertainty as to whether or not atmospheric aerosol growth takes
place by formation of condensable precursors solely in the gas phase,
followed by irreversible (vapor) diffusion controlled nucleation or
deposition on existing particles or by surface catalyzed chemical reactions
to form stable products on existing particles. Despite the continuing field
71
-------
work devoted to clarification of this question, there is little more than
speculation on which to base a judgement on alternative mechanisms. However,
one can derive a hypothesis on which to base a first approach to modeling
with the following assumptions:
1. Condensed material in the atmosphere is essentially in thermodynamic
equi1ibrium with the gas phase.
2. The production of aerosol precursors by photochemical (or other
chemical) reactions is the rate controlling step in atmospheric aerosol
formation.
3. The existing aerosol in the atmosphere can be characterized by
averaged properties of volume fraction, , total surface concentration,
, a surface chemical potential (per unit gas volume) y , and a bulk
chemical potential p..
We note that essentially three kinetic processes must be considered:
precursor formation
dX
= R,(X] Xn,P)
dt
dX
£« R(X. X .P)
dt "nv"l n
IF -V*i~ V>
where X. --- X are reactants and intermediates and P is the aerosol precursor.
nucleat ion
exp (
T£n S
where J is the number of particles formed per cc per second and y is the
surface free energy of the condensing species, which depends on the temperature
T, and the product P, and S is the supersaturation ration P/pm, where p is
72
-------
Science Center
Rockwell International
the equilibrium vapor pressure of species P,
only.
SC537.10FR
is a function of temperature
^t
vapor diffusion limited deposition $ ~ v (S-l)
where x is an exponential ranging from unity to two-thirds.
Suppose the aerosol formation process is characterized in terms of
the distribution function n(v,y.,t), where n (v,y . ,t)dvdt 6y. is the number
of particles in a volume range between v and v+dv, in a time interval t and
t+dt, and with a chemical potential difference from a standard 6y. (j = s or i)
Then formally, the aerosol modeling hypothesis in simplest form is
expressed in terms of ratios in a functional form:
where p is the prototype property and m is the model property. The first
model approach is to initially simplify further by assuming that
-------
quantify the gas phase chemical kinetics sufficiently well to look for
scaling ratios between model and prototype.
A.3 Mathematical Modeling of Gas Phase Kinetics
We are now using a variety of mathematical techniques to incorporate
the pertinent rate data into kinetic models for analysis of planned experi-
ments and data. These range from simple analytical models to complete
numerical solution of the differential equations of the system of equations.
The more sophisticated numerically oriented methods allow much flexibility
in the inclusion of individual reactions. Simple numerical analysis is
achieved by conventional techniques using pseudo steady state approximations
and obtaining those integrals which are available analytically, followed by
numerical solution. The numerical solutions are carried out on an inter-
active terminal to an IBM 360 computer in APL ("A Programming Language").
Typical output from this system is shown in Figures 12 and 13. The calcu-
lations shown were used in understanding the NO in air photolysis. The
output generated could be saved and manipulated in real time. A typical
result of such a manipulation is shown in Figure 14 and represents a cross-
section of output from many graphs of the form of Figure 13. This system
of analysis is limited in the number variables, the speed of calculation,
and accuracy of approximations.
We are now nearing completion of a general reaction kinetics program
(eg)
for the CDC 6600 computer. This system employs advanced solution techniques
for stiff differential equations. Its most important feature however is
the addition of reaction by insertion of single data cards and the ability
to analyze relative rates at any point during the calculation. Many additional
features are being considered for this program but the most important factor
is that this modeling effort can be tied closely to the experimental program.
Both proposed mechanisms and experimental results will be analyzed quickly and
efficiently.
We have not employed this system yet to completely analyze a reactor
experiment but some test output is now presented to demonstrate its capability.
Figure 15 shows the input of a set of reactions as it appears on the computer
7k
-------
10/26/72
en
6 REACTIONS
INCLUDING 0 +N02
0 +N02
03+N02
Kl= 0.0056
TIME(SECS) IN02]
0
1
2
3
4
5
6
7
8
9
10
11
12
13
11
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
10
9. 942074665
9.886604424
9.835802371
9.790899868
9.752197277
9.71931611
9.691499308
9.667858021
9.647530816
9.629764456
9.613940831
9.599573459
9.586290154
9.57381169
9.561931415
9.550497828
9. 539400579
9.528559648
9.517917168
9.507431359
9.497072069
9.486817529
9.476652001
9.466564077
9.456545455
9.446590065
9.436693431
9.426852225
._.9j_H..l.IP. 63JH_1
9.407326659
-» NO + 02
* N03
* H03 + 02
[03]
0
0.05273074587
0.1015436648
0.1443868715
0.1802295949
0.2089822548
0.2312219363
0.2478790328
0.2599855087
0.2685162695
0.2743118765
0.2780561825
0.2802843626
0.2814042202
0.2817207791
0.2814592233
0.2807842639
0.279815587
0.2786397528
0.2773191499
0.2758986329
0.2744103882
0.272877477
0.2713164005
0.2697389512
0.268153545
0.2665661783
0 .2649811152
0.2634013793
0.2618291068
0.2602657988
LN01
0
0.05666404282
0.109366355
O7l~560"5~80706
0.. 1957127144
0.2282455158
0.2542382649
0.2746255665
0 .2904429287
0.3026681102
0.3121438996
0. 3195558516
0.3254404231
0 .330206375
0. 3341594432
0.3375253421
0.3404691784
0.3431109371
0.3455374054
0.3478111483
0.3499771591
0.3520677366
0.3541060349
0.3561086342
0.3580873957
0.3600507967
0.3620048898
0.363953991
0.3659011732
0.3678486192
0.3697978754
CO
o
en
CO
w
Fig. 12 APL PRINTOUT OF CALCULATED N02> NO, AND 03 CONCENTRATIONS
-------
1.501 x
1.25
1.00
0.75
0.50
-Q...25
6 REACTIONS INCLUDING 0 + NO2 * NO + 02_
0 + N02' -» N03
03+ N02 -» N03+ 02
Kl= 0.056 ~ 10 SUNS
SYMBOL
0
+
x +
V * V +
V__V.JL +
x V + "+ V
*
O
V
+
x
LN0210-PPM
_1
2
5
10
20
50
V
+
V V V V V
V V V V V
V V V V
O O O O
oooxooooo + +
O X 0000+++00
__x + + _ __
x ~ " ' " --- -----+
X + + + +
_* * * *. * __* JL__x_*_^ * _* ** * * * ** * * * * *
* * xx
OOxxxxOOOOOOOOO
o_. __*_
*
O
* O
O
OOOOOOOO
xxxx
OOO
xxxxxx
O.OOx
| |
0 5 la...
15
1
20
25
1
30
TIME (SECS)
Fig. 13 AN APL PLOT OF RESULTS OF SEVERAL SETS OF CALCULATIONS OF 03 CONCENTRATION
-------
Q.
Q.
4. 5
4. 0
3.5
3. 0
2.5
2.0
1. 5
1. 0
0.5
0. 0
0
o
o
o
*
*
*
*
o
o
O
*
o 0
*
O
O
o
1 I 1
1 10 20 30
o
Science Center
Rockwell International
SC537.10FR
k. sec
0 2.8
* .56
0 .056
50
[NO ] ppm
Fig. 14 - The calculated saturation of ozone production with initial
NO- concentration.
77
-------
Figure 15 THE INPUT SET OF REACTIONF FOR A RUN OF THE SCIENCE CENTER
GENERALIZED KINETICS PROGRAM.
KINETICS PROGRAM
RE AC T-I-0 NS-US&O- -ARE
N02*NO+0/.008
Oi»03/K»6.99E(H USE THIS CARD ONLY WHEN NOT COMPUTING 02
0*NO^N02/K=53
0*N02*NO/K=135EOO USE THIS CARD ONLY WHEN NOT COMPUTING 02
N03*N02i»N205/K=ll3
N205^N02*N03/K=,25
e3*N02i»^03/K«rOOi3EOO --------- US£~ W*S--ftftfr-Of«LY- HHEN NOT-CQMPiJT-INe 02
N03*NOi»N02*N02/K=183
78
-------
Science Center
Rockwell International
SC537.10FR
printout. Figure '6 shows actual computer printout of the solution to the
system shown In Figure 15. In addition Calcomp plots can be prepared from
the results. This program promises to be very useful in analyzing the output
data from the flow reactor.
79
-------
CONCENTRATIONS AS A KUNCFI ON OF TIME ALL 6 KEACTANTS -IN PPM
00
o
TlME(StCS)
0.
1.000
2.000
3.000
4.000
5.000
6.000
7.000
8.000
9.000
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
IB. 00
19.00
20.00
21.00
22.00
23.00
24.00
2b.OO
26.00
27.00
28.00
29.00
30.00
31.00
32.00
33.00
34.00
35.00
N02
.2SOOOE-02
.24801L-02
.24hU4E-02
.24410E-02
.24217E-G2
.2402YE-02
.23838E-02
23b52E-l)2
.2.5468E-02
2328bE-02
23105E-02
.22927E-02
.22751L-02
.2?576E-02
.22404E-02
.2?233E-02
.2?065E-02
.21898E-02
.21733E-02
.2I570E-02
.21408E-02
21249E-02
.21091E-02
.2<)935E-02
.2078JE-02
,2oft28E-02
20477E-02
20328t-u2
20181E-02
,2o035t-02
.19890E-02
.19748E-02
.19607E-02
.19467E-02
.19329E-02
19193E-02
NO
. 7s()OOt-:-02
. 7
-------
Science Center
Rockwell International
SC537.10FR
Appendix B. Aerosol Forming Reactions with Ozone and Olefins in a Flow Reactor
One of the problems that developed early in this feasibility study
involved questions of time scaling for aerosol formation in reactive gases.
Furthermore, samples of aerosols analogous to those expected to be generated
by photochemical reactions were required to test and validate sampling
methods and schemes for chemical analysis. In a complementary program
sponsored by the Science Center, work was started to investigate airborne
particle formation resulting from chemical reactions and a series of olefin
vapors. Such reactions may be important in photochemical smog. After
determining the feasibility of forming aerosols by such reactions, the for-
mation rate was investigated, and samples of the condensed material were
used to develop chemical analysis techniques for the APRAC sponsored
program.
B.1 Experimental Methods
The aerosol forming reactions of olefins and ozone were investigated
in a steady flow system made of a pyrex pipe 3 meters long and 5-1 cm in
diameter. Ozone was generated in a stream of oxygen by ultraviolet
radiation; this flow was mixed in the flow reactor with a dry nitrogen
stream (< 10 ppm H_0 vapor) containing olefins at the ppm concentration level.
Gas flows into the reactor were measured using rotameters.
The olefins studied included 1-butene, 1-hexene, and 1-octene; however,
the one of principal interest was 1-butene.
The ozone concentration was measured at the exit of the reactor with
and without the presence of olefin using a Kl titration method or a REM
chemi1uminescent ozone monitor, calibrated with a Kl titration procedure.
Reaction times, based on the mean gas flow through the reactor were
set to range between 20 sec and 1 minute. Flow was kept sufficiently low to
retain a laminar flow.
Sampling for aerosol was conducted by a glass tube centered in the pipe
axis. The sampling tube could be pushed into the pipe up to \h meters to
vary the reaction time. Shorter pipes also were used to vary the reaction
time.
81
-------
Qualitative tests for the sensitivity to water vapor or to aldehyde
vapors were performed by adding such constituents with the nitrogen stream
at the reactor entrance. Tests have been made with water vapor, acetaldehyde,
and butyaldehyde.
Aerosol production was measured by several means. The physical properties
were observed using the following instrumentation:
a. Total number concentration - GE Aitken Nuclei Counter
b. Total mass concentration - glass fiber filter
c. Particle size distribution - Portable Whitby - Liu electrical
mobi1i ty analyzer
d. Visible light scattering - laser beam and Royco 225 optical
counter
The chemical properties were derived from examination of filter collected
samples using infrared spectroscopy, gas chromatography-mass spectroscopy,
iodimetry and electrochemistry. The infrared spectra were obtained using a
Perkin Elmer Model A21 Infrared Spectrometer. This instrument is a double
grating spectrometer with resolution of 0.3 cm . Both extraction and pellet
techniques were employed. The technique used for the iodide oxidation
measurements was standard using sodium thiosulfate to titrate the ! produced
with starch as the indicator. The polarographic measurements were made with
(39).
an instrument constructed in this laboratory. For the GC-MS system a
Loenco Model 15B gas chromatograph was used with an inlet system especially
designed to allow the pyrolysis products to enter the gas chromatograph in
a "slug." The GC carrier gas is separated from the sample material using
an effusion orifice and a silver membrane enricher. The CEC 21-103C mass
spectrometer used was modified to allow a scan of the accelerating voltage
to be made within two seconds with repetitive scans every three seconds.
This gives a sweep from mass number 26 to 150. This was done by an in-house
designed and constructed scan control system which operates in conjunction
with the Dempster configuration of the mass spectrometer to give a linear
relation between mass number and time. A Honeywell Visicorder was used to
record mass spectral patterns at the appropriate times. In addition, a DEC
PDP-11 computer has been interfaced with the MS output to allow peak height
82
-------
Science Center
Rockwell International
SC537.10FR
vs. time to be obtained and stored on a :disk memory. Provision has been made
to reduce and store 500 peaks per scan with a total number of scans per
analysis of 20,000. The data are reduced by FORTRAN programs, which give
various types of information such as normalized total ionization vs. scan
number.
B.2 Experimental Results & Discussion
The study showed that aerosol can be generated in a dry nitrogen steam
in a mixture containing ppm (to tens ppm) level ozone and ppm (to tens ppm)
level olefin beginning with 1-butene over an average residence time of less
than a minute. The production of Aitken nuclei in the reactor for different
olefins is shown in Fig. 17. Here there is evidence that aerosol production
rate is enhanced, for a given residence time, using olefins of carbon number
higher than four. The increase in nuclei production rate is very sharp with
increasing olefin concentration suggesting an effective "threshold" for
nucleation of condensed phase in the reactor. This indicates that the rate
of production of new nuclei is highly non-linear and can be expected to be
very slow or negligible, for olefins of molecular weight lower than 50-80,
at the ppb level in a stream containing ppm level or lower ozone.
At concentrations of reactants higher than a few ppm, particles of size
sufficient to scatter light were easily seen using a laser beam flashed
through the pipe cross-section.
A limited number of tests of the 1-butene-ozone system were conducted
using a new portable Whitby-Liu electrical mobility analyzer as the aerosol
detector. The aerosol volume distribution as well as the total surface to
volume ratio can be derived from measurements using this device. The change
in (volume) size distribution with increasing 1-butene concentration is shown
in Fig. 18. This set of curves indicates that the early stages of nucleation
and growth by condensation of material generates an aerosol with a weak
increase in mean particle size from ~ 0.07 ym to ~ 1.0 urn, but with material
accumulating primarily in the particle sizes below 0.1 ym diameter.
Volume distributions corresponding to increases in residence time are
shown in Fig. 19. These results indicate the changes in the aerosol formed
83
-------
300
280
240
200
CO
r
o
o 160
120
80
40
1 I I I I
T
T = 30 sec
[03] « 0.3 ppm
A 1- Octene
1- Hexene
1- Butene
20 30 40
ppm Olefin
Figure 17 PRODUCTION OF AEROSOLS FROM OLEFINS, AS MEASURED BY AITKEN NUCLEI
CONCENTRATION.
84
-------
350 -
00
tn
!
ff 5
Figure 18 VOLUMETRIC SIZE DISTRIBUTION AS A FUNCTION OF 1-BUTENE
CONCENTRATION IN THE FLOW REACTOR.
a>
-------
Q.
CD
O
o
>
o
005 0.01
0.10
Dp
Figure 19 EVOLUTION OF AEROSOL SIZE DISTRIBUTION GENERATED FROM THE REACTION
OF OZONE AND 1-BUTENE.
86
-------
Science Center
Rockwell International
SC537-10FR
at constant (initial or upstream) reactant concentrations with time. There
is an increase in mean particle size with increasing time indicating quali-
tatively the complementary effects of nucleation, growth by condensation
and coagulation of particles. That is, there is a decrease in the volume of
smallest particles accompanied by a generation of larger particles and a
systematic shift in mean particle size to larger sizes.
The changes in number concentration () and mean volume concentration
() with changes in 1-butene and ozone are shown in Figs. 20 and 21. In
both cases, there is a sharp initial increase in airborne material showing
a "threshold" analogous to the increase in nuclei concentration in Fig. 10.
At higher reactant concentration the nuclei production rate tapers off, but
the surface and volume continue to increase. The data for variation in
volume production rate with reactant concentration are semi-quantitative in
nature but indicate that the generation of condensed phase is non-linear in
both olefin and ozone.
It is interesting to compare the collection of mass on a filter with
the measurements of the electrical mobility analyzer. Such a comparison is
shown in Fig. 22 for the volume concentration estimated from the analyzer.
Assuming that the condensed material has a constant density, there is a
major difference between the filter collected mass and that projected from
the electrical mobility analyzer. The filter collects considerably more
material than expected and the accumulated material appears to increase
linearly with butene concentration rather than non-1inearly. This result
points to the critical problem of sampling for aerosols in a chemically
reactive gas mixture. In this case, the filter substrate may be acting
as a reactive surface for production of condensed material. On the other
hand the filter is larger in diameter, nearly as large as the pipe, and will
be accumulating significant amounts of aerosol from slower moving regions
of gas (longer residence times) than the sampling tube used for the analyzer.
The analyzer tubing may contribute to some loss of aerosol, too, prior to
reaching the instrument. Further work is needed on this problem before
such data can be interpreted intelligently.
87
-------
13
12
350
10
100
50,
0 10
20
30 40 50 60 70 80
1-Butene Concentration (ppm)
90 ' 100 110 120
FIGURE 20 VARIATION WITH 1-BUTENE OF. TOTAL NUMBER AND VOLUME CONCENTRATION
OF AEROSOLS PRODUCED FROM A REACTION OF OZONE AND 1-BUTENE.
-------
14
00
10;
o
n i i
BUTENE 10 ppm
T = 1 min
2.0 4.0 6.0 8.0 10.0 12.0 14.0
OZONE (ppm)
?£
Q- 5"
I o
QL (D
Figure 21 CHANGE IN NUMBER AND VOLUME CONCENTRATION WITH INCREASE IN OZONE
CONCENTRATION.
§
Q>
-------
1400 -
1200 -
1 I I ' I
OZONE 1 - BUTENE REACTION
T = 54 sec
OZONE 12 ppm
O FILTER
Av
30 40 50 60
BUTENE IN TOTAL FLOW CONCENTRATION (ppm)
Figure 22 COMPARISON BETWEEN MASS CONCENTRATION FROM GLASS FIBER FILTER
COLLECTION AND ESTIMATED FROM THE WHITBY ELECTRICAL MOBILITY
ANALYZER ASSUMING A DENSITY OF UNITY.
-------
Science Center
Rockwell International
SC537.10FR
B .3 Chemical Analysis of Ozone-Olefin Aerosol
The chemical analysis of the filter collected material from the ozone-
1-butene reactions revealed that its composition was organic, with highly
oxygenated constituents.
Freshly prepared aerosol will readily oxidize iodide. If the material
stands for longer than a week, it no longer will react. It has been found
that an oxidation equivalent of approximately 0.01 meq/mg is obtained on a
fresh sample. This apparent reactivity would suggest that the sample con-
tains either a peroxide or an ozonide. Using the method of Mair and
Graupner, it has been shown that the "Method 1" or easily reduced
peroxide is responsible for iodide oxidation. The positive half wave
potential of 'the reduction wave of the material in a fresh aerosol sample
also suggests that the material falls into the class of easily reduced
peroxides [e.g. diacyl peroxides, hydro peroxides, etc.].
Examination of infrared spectra and gas chroma tography-mass spectroscopy
combination data also suggests a possible peroxide or ozonide but the
identification is still not clarified. Fig. 23 shows the infrared spectrum
of an aerosol by transfer of the material to an Irtran window. The general
similarity of this IR spectrum to those obtained from ambient aerosol and
smog chambers is interesting. Table 8 shows two GC-MS analyses of filter
collected aerosol samples. Table 9 shows a comparison of mass spectral data
obtained from one gas chromatograph peak with the known compound and with the
reference spectral pattern. The differences between the calibration pattern
and the ASTM pattern is caused by different mass discrimination resulting
from the use of different mass spectrometers and detectors.
One of the problems with identification of the aerosol composition is
associated with its potential complexity. In order to get specific identi-
fication of the components of the material, it appears to be necessary to
attempt a separation of the product mixture. The use of thin layer chroma-
(*tl)
tography and paper chromatography have been used on peroxide mixtures and
these techniques are currently being tried.
91
-------
SPECTRUM OF AEROSOL FROM 03 +1 - BUTENE REACTION ETHER EXTRACT
WAVELENGTH (microns)
2.5 3 45 6 7 8 9 10 12 15 20 25
|00| ' ' ' ' 1''''| I I ' ' I I I I I | I I I I InI1| I l I I I I I I I | I I I I I I I I I | I I I I I I III | II ll|llll|llll| I I I I | I i I III lll|llll|mii i | i | i | 111| i i i i
80
UJ
O
60
^40
<
Q:
i-
20
0
I
I
1,1,1,1,1,1,1,1,1,1,1,1,1,1,1
4000 3500 3000 2500 2000
1800 1600 1400
WAVE NO. (cm"1)
1200 1000 800 600 400
FIGURE 23 INFRARED SPECTRUM OF THE ETHER EXTRACT OF FILTER COLLECTED AEROSOL
PRODUCED FROM THE REACTION OF OZONE AND 1-BUTENE.
-------
Science Center
Rockwell International
SC537.10FR
Table 8
IDENTIFICATION BY GC-MS OF APPARENT PYROLYSIS PRODUCTS OF
AEROSOLS OBTAINED FROM OZONE 1-BUTENE GAS PHASE REACTION
GC Peak
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
% Identified
Run 1
9.8**
0.03
0.09
0.6
19.7*
10.1
0.4
3.5
44.5
19.1
1.8
0.3
14.6
0.06
2.2
0.3
0.1
1.8
0.4
0.3
94.6
Run 2
-11. 7*%
0.5
34.6*
6.7
0.6
5.2
29.2
18.5
1.7
12.7
2.2
1.0
96.8
Ident i ty
Carbon Dioxide
Ethane
Water
Formal dehyde
Methanol
Acetal dehyde
Formic Acid
Propionaldehyde
Acetic Acid
Propionic Acid
*% of total sample but excluded from other percentage calculations.
% refers to peak area percentage assuming equal thermal conduc-
tivity detector sensitivity. Run 1: 6 mg total aerosol. Run 2:
12 mg total aerosol.
93
-------
Table 9
COMPARISON OF MASS SPECTRAL PATTERNS FOR PROP I ON 1C
ACID OBTAINED FROM AEROSOL DECOMPOSITION AND FROM
CALIBRATION SAMPLE AND REFERENCE SPECTRUM
Calibration Known
M/e
26
27
28
29
30
45
55
56
57
73
74
From Aerosol
21
62
100
86
10 .
31
7.5
6.2
15
24
37
(Propionic Acid)
2k
60
100
83
11
33
9
6
18
34
53
ASTM #302
21.1
61.7
100
83.6
14.1
55.7
16.8
16.4
30.1
48.4
75.8
94
-------
Science Center
Rockwell International
SC537-10FR
REFERENCES
1. Tyndall, J., Phil. Mag. S^t 37, 38A (1969).
2. Robinson, E. and Robbins, R. C., "Gaseous Atmospheric Pollutants from
Urban and Natural Sources" in Global Effects of Environmental Pollution,
S. Fred Singer, Editor, Springer-Verlag, New York, 1970.
3. Leighton, P. A., "Photochemistry of Air Pollution," Academic Press,
N.Y., 1961.
4. Wilson, W. E., Jr. and Levy, A., JAPCA 2£, 385 (1970); Wilson, W.E., Jr.,
Merryman, E.L., Levy, A., and Tal iaferro, H. R. , JAPCA 2J_, 128 (1971).
5. Stevenson, H. J. R., Sanderson, D. E., and Altshuller, A. P., Int. J.
Air Water Poll. 9., 367 (1965).
6. Groblicki, P. J. and Nebel, G. J., "The Photochemical Formation of
Aerosols in Urban Atmospheres," in Chemical Reactions in Urban Atmos-
pheres, C. A. Tuesday, Editor, Elsevier, N.Y., 1971.
7. Cox, R. A. and Penkett, S. A., "Aerosol Formation from Sulfur Dioxide
in the Presence of Ozone and Oleflnic Hydrocarbons," J. Chem. Soc.,
Faraday Trans. 1 68_, 1735 (1972).
8. Bricard, J., Billard, F., and Madelaine, G., J. Geophys. Res. 73, M87
(1968).
9. Goetz, A. and Pueschel , R. F., JAPCA J£, 90 (1965).
10. National Air Surveillance Network (NASN), "Air Quality Data from the
National Air Surveillance Networks and Contributing State and Local
Networks," 1966 Edition, U.S. Dept. HEW, Durham, N.C., 1968.
11. Hidy, G. M. and Friedlander, S. K., "The Nature of the Los Angeles
Aerosol," in the Proc. 2nd IUAPPA Clean Air Congr., Editor, Englund,
H. M. and Beery, W. T., Academic Press, N.Y., 1971.
12. Miller, M. S., Friedlander, S. K., and Hidy, G. M., "A Chemical Element
Balance for the Pasadena Aerosol," in reference (15).
13. D. S. Covert, R. J. Charlson, N. C. Alquist, J. Applied Meteorology 11-
968 (1972).
1A. Interim Report, California Aerosol Characterization Study, Science Center,
Rockwell International, Thousand Oaks, CA, in preparation, April 1973-
15. Hidy, G. M., Editor, Aerosols and Atmospheric Chemistry, Academic Press,
N.Y., 1972.
95
-------
16. Husar, R. B., Whttby, K. T., and Liu, B. Y. H., "Physical Mechanisms
Governing the Dynamics of Los Angeles Smog Aerosol," In reference (15).
17. Hidy, G. M. , Mueller, P. K., Tokiwa, Y, and Twiss, S. , "Aerometric
Factors Affecting the Evolution of the Pasadena Aerosol," in reference (15)
18. Altshuller, A. P. and Bufalini, J. J., Env. Sci. Tech. 5., 39 (1971).
19. Bufalini, M. , Env. Sci. Tech. 5., 685 (1971).
20. Westerberg, K. , Cohen, N., and Wilson, K. W. , Science J_7_l_, 1013 (1971).
21. Gay, B. W. and Bufalini, J. J., Env. Sci. Tech. 5., ^2 0971).
22. Friedlander, S. K. and Seinfeld, J. H. , Environ. Sci. Technol. 3., 1175
(1969); Hecht, T. A. and Seinfeld, J. H. , ibid. . 6_ kj (1972).
23. Demerjian, K. L., Kerr, J. A., and Calvert, J. G., "The Mechanism of
Photochemical Smog Formation," Chemistry Department, Ohio State Univer-
sity, 1972.
2k. Niki, H., Daby, E. E., and Weinstock, B., "Mechanisms of Smog Reactions,"
in R. E. Gould, Editor, "Photochemical Smog and Ozone Reactions,"
Advances in Chemistry, Series 113, Am. Chem. Society, Washington, D.C.
(1972).
25. Haagen-Smit, A. J., Ind. Eng. Chem. kk, 13^2 (1952).
26. Doyle, G. J. and Renzetti, N. A., JAPCA £, 23 (1958; Renzetti, N. A.
and Doyle, G. J., ibid. 8^ 293 (1959); Renzetti, N. A. and Doyle, G. J.,
Int. J. Air Poll. 2_, 237 (i960); Doyle, G. J. and Jones, J. L., JAPCA
13, 365 (19 3); Endow, N., Doyle, G. J., and Jones, J. L. , ibid. 13, 1**1
TT963).
27- Prager, M. J., Stephens, E. R., and Scott, W. E., Ind. Eng. Chem. 52,
521 (I960).
28. Goetz, A. and Kallai, T., JAPCA 12, k37 (1962); Goetz, A. and Kallai, T.
J. Am. Ind. Hyg. Assoc. 2k, A53 TT963); Goetz, A. and Pueschel, R. F.,
Atm. Env. j_, 287 (1967).
29. Harkins, J. and Nicksic, S. W. , JAPCA j_5_, 218 (1965).
30. Wilson, W. E., Jr., Schwartz, W. E., and Kinzer, G. W., "Haze Formation --
Its Nature and Origin," Battelle Columbus Laboratories, Jan. 28, 1972.
31. Generalization of equation (16) from reference (32).
32. Ferguson, E. E., Fehsenfeld, F. C., and Schmeltekopf, A. L., "Flowing
Afterglow Measurements of Ion-Neutral Reactions," in Advances in Atomic
and Molecular Physics, Editor, Bates, D. R., Academic Press, N.Y., 1969.
96
-------
Science Center
Rockwell International
SC537.10FR
33. Cher, M. and Hoi 1ingsworth, C. S. , "Chemi1uminescent Reactions of
Excited Helium with Nitrogen and Oxygen," Advances in Chemistry Series
80, 118 (1969).
34. Crank, J., "The Mathematics of Diffusion," Oxford University Press, N.Y. ,
1957, pages 66-67.
35. Dushman, S. , Scientific Foundations of Vacuum Technique, 2nd ed., John
Wiley and Sons, Inc., N.Y., 1962.
36. National Research Council, International Critical Tables 5., 62 (1929).
37- Fuchs, N. A., The Mechanics of Aerosols, Pergamon Press, N.Y., 1964.
38. Chapter 4, Section 26 of reference (37).
39. Parry, E. P. and Osteryoung, R. A., Anal. Chem. 36, 1366 (1964).
40. Mair, R. D. and Grapuner, A. S., Anal. Chem. 36, 194 (1964).
41. Mair, R. D. and Hall, R. T. in Organic Peroxides, edited by Swern, D. ,
Vol. II, John Wiley and Sons, New York, 1971, see Chapter VI.
42. Ottenstein, D. M. and Bartley, D. A., "Improved Gas Chromatography
Separation of Free Acids C9-CC in Dilute Solution," Anal. Chem. 43,
952, 1971. 2 * ~
43- Cvetanovic, R. J., "Addition of Atoms to Olefins in the Gas Phase" in
Advances in Photochemistry, Vol. 1, Editors,'Noyes, W. A., Jr., Hammond,
G. S., and Pitts, J. N., Jr., Interscience Publishers, N.Y., 1963.
Sato, S. and Cvetanovic, R. J., "The Effect of Molecular Oxygen on
the Reaction of Oxygen Atoms with cis-2-Pentene," Canadian J. Chem. 37,
953, 1959.
44. Altshuller, A. P., Kopczynski, S. L., Lonneman, W. A., Becker, T. L. ,
and Stater, R. , "Chemical Aspects of the Photooxidation of the P
pylene-Nitrogen Oxide System," Env. Sci. Tech. j_, 899, 1967-
45. Ellis, C. F., "Sunlight Photooxidation of Isobutylene and Butene-1 with
NO, in Low-and High-Oxygen Concentrations," Atmospheric Environment 5_,
837 1971.
46. Scarengelli, R. P. and Rehme, K. A., Anal. Chem. 4j_, 707 (1969).
47. Huntzicker, James J., California Institute of Technology, private
communicat ion.
48. Fritz, J. S. and Yamamura, S. S. , Anal. Chem. 27., 144 (1955).
97
-------
**9. Benarce, M. and Nonet, A., Atmos. Environ. 6_ (11), kS (1972).
50. Methods of Air Sampling and Analysis published by American Public
Health Assoc. , 1972, p. 322.
51. West, P. W. and Lyles, G. L. , Anal. Chem. Acta 23_, 227 (i960).
52. Sawicki, E. , Johnson, H. , and Stanley, T. W. , Anal. Chem. j>5_, 13*» (1963).
53. Parry, E. P. and Osteryoung, R. A., Anal. Chem. 37> 1632* (1965);
Parry, E. P., Johnston, W. , and Goldner, H. J., J. Electroanal. Chem. 13,
127 (1967); Parry, E. P. and Oldham, K. B., Anal. Chem. A£, 1031 (196877
5A. Polarography by Kolthoff, I. M. and Lingane, J. J., Vol. II, 2nd ed.,
Interscience, New York, 1952, p. 536.
55. Lipeles, M. and Malmuth, N., "Analytical Smog Kinetics," Environ.
Sci. Tech. 6_, 360-365, 1972.
56. Bridgman, P. W., Dimensional Analysis, Yale University Press, New Haven,
193K
57. Birkhoff, G., Hydrodynamics, Dover, N.Y. , 1950.
58. Hidy, G. M., "On Atmospheric Simulation: A Colloquium," NCAR-TN-22
Report of National Center for Atmospheric Research, Boulder, Colorado,
1966.
59- Gear, C. W., Communications of the ACM _U, 176, 1971.
60. Whitby, K. T., Husar, R. B., Liu, B. Y. H., "The Aerosol Size Distri-
bution of Los Angeles Smog" in reference (15).
98
------- |