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 I—n—I—1—|  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

-------