Evaluation of the UNC toluene-SOA mechanism with respect to other chamber studies and key model...
Transcript of Evaluation of the UNC toluene-SOA mechanism with respect to other chamber studies and key model...
ARTICLE IN PRESS
1352-2310/$ - se
doi:10.1016/j.at
�Correspondfax: +919 966 7
E-mail addr
Atmospheric Environment 41 (2007) 6465–6477
www.elsevier.com/locate/atmosenv
Evaluation of the UNC toluene-SOA mechanism with respect toother chamber studies and key model parameters
Di Hu, Richard M. Kamens�
Department of Environmental Sciences and Engineering, the University of North Carolina, Chapel Hill, NC 27599, USA
Received 11 September 2006; received in revised form 15 April 2007; accepted 16 April 2007
Abstract
In a companion paper by Hu et al. [2007. A kinetic mechanism for predicting secondary organic aerosol formation from
toluene oxidation in the presence of NOx and natural sunlight. Atmospheric Environment, doi:10.1016/j.atmosenv.
2007.04.025], a kinetic mechanism was developed from data generated in the University of North Carolina’s (UNC) 270m3
dual outdoor aerosol smog chamber, to predict secondary organic aerosol (SOA) formation from toluene oxidation in the
atmosphere. In this paper, experimental data sets from European Photoreactor (EUPHORE), smog chambers at the
California Institute of Technology (Caltech), and the UNC 300m3 dual-outdoor gas phase chamber were used to
evaluate the toluene mechanism. The model simulates SOA formation for the ‘low-NOx’ and ‘mid-NOx’ experiments
from EUPHORE chambers reasonably well, but over-predicts SOA mass concentrations for the ‘high-NOx’ run.
The model well simulates the SOA mass concentrations observed from the Caltech chambers. Experiments with the three
key toluene products, 1,4-butenedial, 4-oxo-2-pentenal and o-cresol in the presence of oxides of nitrogen (NOx) are also
simulated by the developed mechanism. The model well predicts the NOx time–concentration profiles and the decay of
these two carbonyls, but underestimates ozone (O3) formation for 4-oxo-2-pentenal. It well simulates SOA formation from
1,4-butenedial but overestimates (possibly due to experimental problems) the measured aerosol mass concentrations from
4-oxo-2-pentenal. The model underestimates SOA production from o-cresol, mostly due to its under-prediction of o-cresol
decay. The effects of varying temperature, relative humidity, glyoxal uptake, organic nitrate yields, and background seed
aerosol concentrations, were also investigated.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Toluene; Kinetic mechanism; Aerosol modelling; Secondary organic aerosol (SOA)
1. Introduction
A number of kinetic models for predicting ozone(O3) and secondary organic aerosol (SOA) formationfrom the photo-oxidation of aromatic hydrocarbonshave been developed over the past 5 years. Stroud
e front matter r 2007 Elsevier Ltd. All rights reserved
mosenv.2007.04.026
ing author. Tel.: +919 966 5452;
911.
ess: [email protected] (R.M. Kamens).
et al. (2004) used the Master Chemical Mechanismversion 3.0 (MCMv3.0) and implemented more than20 partitioning species that participated in SOAaccumulation. They observed that nitro-aromaticsand multifunctional organic nitrates are the mostsignificant contributors to SOA. Griffin et al. (2002)and Pun et al. (2002) developed a gas phasemechanism and shunted both terpene and aromaticSOA accumulations into 10 surrogate species.Johnson and Jenkin et al. modified the master
.
ARTICLE IN PRESSD. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776466
mechanism (Master Chemical Mechanism Version3.1 (MCMv3.1) (Bloss et al., 2005)) to includeaerosol phase reactions of aldehydes and hydroper-oxides (Johnson et al., 2004). With a different scalingfactor applied to aerosol mass concentrationsobserved in each experiment, they were able tomodel toluene SOA formation in the 195m3
European Photoreactor (EUPHORE) in the presenceof natural sunlight and oxides of nitrogen (NOx). Tofit the toluene decay, an artificial source of OH wasadded and ‘‘tuned’’ to each experiment. Althoughmany advances have been made, all of the abovemechanisms do not address the recent discovery ofparticle phase reactions that may lead to verysignificant SOA formation from aromatic systems(Jang and Kamens, 2001; Kalberer et al., 2004).
Most recently, a kinetics mechanism developed atthe University of North Carolina was designed toinclude particle phase oligomerization reactions (Huet al., 2007). It also employs an alkyl peroxy radicalreaction with gas phase toluene (Broyles and Carpen-ter, 2005) to remove toluene after OH has peaked inthe system. Very good to reasonable simulations ofNOx, O3 and particle time–concentration behaviourswere observed for eight different experiments in theUniversity of North Carolina’s (UNC) 270m3 dualoutdoor aerosol smog chambers, operated undernatural sunlight. The most important particle phasecompounds are glyoxal oligomers (GLYPOLY),ketene oligomers (SEED1) from the photolysis ofthe toluene OH reaction product, 2-methyl-2,4-hexadienedial, organic nitrates (RONO2), acylper-oxy nitrates (RCO-OONO2), methyl nitro-phenol
Table 1
Initial conditions of experiments conducted in EUPHORE chambers t
Experiment ID Initial concentration (ppm)
[Tol]0a [NO]0 [Tol]0
‘low-NOx’ 092401 0.514 0.047 10.9
‘mid-NOx’ 092701 0.496 0.122 4.07
‘high-NOx’ 092501 0.512 0.422 1.21
1,4-Butenedial [Bud]0a [NO]0 [Bud]
0.681 0.082 8.30
4-Oxo-2-pentenal [Oxo]0a [NO]0 [Oxo]
0.302 0.186 1.62
o-Cresol [o-cresol]0 [NO]0 [o-cre
0.298 0.024 12.42
aTol, Bud and Oxo stand for toluene, 1,4-butenedial and 4-oxo-2-pebNo temperature and relative humidity data were available for 1,4-b
run at 298–303K and 10% relative humidity.
analogues, C7 organic peroxides, and for the lowconcentration experiments, unsaturated hydroxylnitro-acids. A glossary of terms is given in details ina companion paper (Hu et al., 2007). The distributionof products depends on different initial conditions.The starting conditions, depending on the experiment,had toluene that ranged from 0.1 to 1.0 ppmV andNOx that ranged from 0.09 to 0.45ppm. Themaximum O3 concentration ranged from 0.21 to0.95ppm. In this paper, the model’s ability to simulatetoluene SOA data reported by other investigators willbe described and model sensitivity to differentparameters will be discussed.
2. Discussion
2.1. Model simulations of toluene experimental data
from the EUPHORE outdoor chamber system
Three toluene experiments were carried out in the195m3 EUPHORE chamber A (Johnson et al.,2004; Bloss et al., 2005) and were simulated with theUNC toluene mechanism. These experiments wereconducted on 24 September 2001, 25 September2001 and 27 September 2001, and experimentalconditions are listed in Table 1. The initialconcentration ratio of toluene to NO ranged from1.21 to 10.9 and temperatures ranged from 298 to304K. Aerosol mass concentrations were calculatedfrom reported scanning mobility particle sizer(SMPS) aerosol volume concentrations by assuminga particle density of 1.4 g cm�3(Bahreini et al.,2005). The mean particle diameter was 130 nm,
hat are used to evaluate the UNC toluene mechanism
Temp. (K) Relative humidity (%)
/[NO]0
298–303 0
298–303 0
298–304 0
0/[NO]0NAb NAb
0/[NO]0NAb NAb
sol]0/[NO]0NAb NAb
ntenal, respectively.
utenedial, 4-oxo-2-pentenal and o-cresol. Model simulations were
ARTICLE IN PRESSD. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6467
and no background seed aerosol was used in theseexperiments (Johnson et al., 2004). The relativehumidity and the aerosol phase water content usedfor model simulations were 10% and 0.5%. Thefirst-order rate constants for dilution, O3 walldeposition and particle wall deposition were0.03 h�1, 1� 10�6 s�1 and 5� 10�5 s�1, respectively(Johnson et al., 2004). Because the latitude of theEUPHORE chamber is almost identical to theUNC chambers, the UNC 23 September 1996 lightfile was used to run all the simulations.
Comparisons between experimental data andmodel simulations of NOx, O3 and SOA massconcentrations are shown in Fig. 1. Fits that werewithin 75% of the data are called ‘‘very good’’ or‘‘simulates very well’’, fits that were within 715%of the data are called good (or ‘‘simulates well’’),and fits within 730% are called ‘‘reasonable’’ (orreasonably fits or tracks). The model properlypredicts NOx conversion for all three tolueneexperiments. It also well fits the toluene profile forthe ‘low-NOx’ and ‘mid-NOx’ systems, but under-estimates toluene decay for the ‘high-NOx’ system.The model tends to over-predict measured O3
concentrations after it peaks for both ‘low-NOx’and ‘mid-NOx’ systems, but it under-predicts O3
accumulation rates for the ‘high-NOx’ run. Sincesimulations were run with a 0.5% aerosol watercontent (Jang and Kamens, 1998), the amount ofNO2 converted to aerosol phase HNO3 was verylimited. The simulated SOA mass concentrationsare in very good agreement with calculated SMPSparticle data for the ‘mid-NOx’ system, but are 24%lower and three times higher than the SMPS data inthe ‘low-NOx’ and the ‘high-NOx’ systems.Although the UNC toluene model greatly over-predicts the SOA formation in the EUPHORE‘high-NOx’ system, it should be noted that ourmodel did a very good job of simulating particlegeneration for a similar experiment on 16 October2004 (101604S) in the south chamber of the UNC270m3 dual-outdoor aerosol smog chambers (Huet al., 2007). This experiment had almost the sameinitial reactant concentrations as the EUPHORE‘high-NOx’ experiment. Since the UNC toluenemechanism can simulate the formation of SOAfrom the other two EUPHORE toluene runsreasonably well, it could be that the large particlesin the particle size distribution for the ‘high-NOx’system may be out of SMPS range (as was observedfor similar UNC experiments). Model fits to severalgas phase products measured in the ‘mid-NOx’ run
are shown in Fig. 2. The model tends to predictformaldehyde generation within 5% of the experi-mental data. The model also properly tracks thetiming of glyoxal and cresol formations, although itunder-estimates the observed peak glyoxal concen-tration by 24% and reported cresol concentrationsby 17%.
2.2. Model simulations of the photo-oxidation of
o-cresol and unsaturated dicarbonyl intermediate
products from the UNC 300 m3 dual-outdoor gas
phase chambers and the EUPHORE outdoor
chamber
Since unsaturated dicarbonyl chemistry is animportant component of the toluene mechanism,the model was used to simulate experimental datafor the photo-oxidation of 4-oxo-2-pentenal and1,4-butenedial. Two UNC experiments were used toevaluate the mechanism and were carried out in theoriginal UNC 300m3 dual-outdoor gas phasechambers located near Pittsboro, NC (Liu et al.,1999). One experiment was initiated with 0.82 ppmV4-oxo-2-pentenal and 0.61 ppm NOx and the otherstarted with 1.08 ppmV 1,4-butenedial and 0.64 ppmNOx. The detailed experimental conditions, cham-ber descriptions, experiment procedures and sampleanalysis techniques are available elsewhere (Liuet al., 1999).
Comparisons between experimental data andmodel simulations of the NO–NO2 conversion, O3
formation and the decay of parent compounds areshown in Fig. 3. The decay of both 4-oxo-2-pentenaland 1,4-butenedial are well simulated by the toluenemechanism. The model reasonably simulates theconversion of NO to NO2 for 4-oxo-2-pentenal butseems to react NO a little too fast for the 1,4-butenedial/NOx system. The model accurately pre-dicts the maximum O3 concentration measured inthe 1,4-butenedial experiment despite an over-pre-diction of morning O3 formation. The model,however, under-predicts the maximum O3 concen-tration in the 4-oxo-2-pentenal system. Themeasured O3 formation time profile for 4-oxo-2-pentenal showed an abrupt cease in ozone accumu-lation from 8:00 to 9:30 a.m. after an initial sharprise, and was then followed by a steady increase inO3 until the end of the experiment. For O3
accumulation to cease from 8:00 to 9:30 a.m.,suggests that all of the gas phase NO2 must havebeen stored in acylperoxy nitrate compounds, whichthen released back to gas phase as temperature
ARTICLE IN PRESS
EUPHORE 'low-NOx' 092401
LDT (hours)
To
luen
e, O
3, N
O*5
, N
O2*5
co
nc (
pp
mv)
0
Part
icle
mass c
on
c
Toluene
NO
O3NO2
LDT (hours)T
olu
en
e, O
3, N
O*2
, N
O2*2
co
nc (
pp
mv)
0
50
100
Part
icle
mass c
on
c (
g
/m3)
O3
Toluene
NO NO2
To
luen
e, O
3, N
Ox c
on
c (
pp
mv)
0
Part
icle
mass c
on
c
O3
NO
NO2
TSP
0.80
0.60
0.40
0.20
0.00
0.80
0.60
0.40
0.20
0.00
10:00 11:00 12:00 13:00 14:00 15:00
60
40
20
EUPHORE 'mid-NOx' 092701
16:0014:0012:0010:00
EUPHORE 'high-NOx' 092501
TSP
1.00
0.80
0.60
0.40
0.20
0.00
Toluene
TSP
17:0015:00
LDT (hours)
13:0011:009:00
160
120
80
40
( g
/m3)
( g
/m3)
Fig. 1. Comparison of model simulated and measured concentrations of toluene, NOx, ozone and aerosol from EUPHORE chamber data
sets. The solid lines (—) represent model simulations. The pluses (+) are measured NO: asterisks (�) are the sum of total measured gas
phase nitrates, diamonds (B) are measured toluene, and squares (&) are calculated SMPS particle mass concentrations.
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776468
increased. In spite of this atypical O3 behaviour, asshown in Fig. 3, the model reasonably tracked thetrend of the O3 profile.
The comparison of measured and simulatedconcentrations of glyoxal and methyl glyoxal from4-oxo-2-pentenal and 1,4-butenedial is shown inFig. 4. The model underestimates measured glyoxalconcentrations by 64–79% and methyl glyoxal by44%. Unfortunately, no error range was given for
the product quantification. Since glyoxal andmethyl glyoxal were calibrated by using theircommercially available 40wt% water solution, thereal measured concentrations might be lower thanthose calculated from the calibration curves ifthe dicarbonyls had already polymerized to someextent in the solution. Unfortunately, no particlemass data was collected for these two earlierunsaturated dicarbonyl UNC experiments because
ARTICLE IN PRESS
0
10
20
30
40
50
Gly
oxal an
d f
orm
ald
eh
yd
e c
on
c (
pp
bv)
Cre
so
l co
nc (
pp
bv)
60
EUPHORE 'mid-NOx' gas phase products
8.0
6.0
4.0
2.0
0.0
cresol
formaldehyde
glyoxal
16:0014:0012:0010:00
LDT (hours)
Fig. 2. Comparison of model simulated and measured concen-
trations of gas phase formaldehyde, glyoxal and cresol from
EUPHORE ‘mid-NOx’ toluene experiment. The solid lines (—)
represent model simulations. The connected triangles (n) are
measured formaldehyde, connected solid diamonds (~) are
measured glyoxal, and connected solid circles (K) are measured
cresol.
4-o
xo
-2-p
en
ten
al, N
Ox-O
3 c
on
c (
pp
mV
)
O3NO
NO2
1,4
-bu
ten
ed
ial, N
Ox-O
3 c
on
c (
pp
mV
)
O3NO
NO2
070298 UNC 4-oxo-2-pentenal
4-oxo-2-pentenal
1.0
0.8
0.6
0.4
0.2
0.014:0012:0010:008:006:00
LDT (Hours)
061898 UNC 1,4-Butenedial1.20
0.90
0.60
0.30
0.00
1,4-Butenedial
7:00 9:00
LDT (Hours)
11:00 13:00 15:00
Fig. 3. Comparison of model simulated and measured concen-
trations of 1,4-butenedial, 4-oxo-2-pentenal, NOx, and ozone.
The solid lines (—) represent model simulations. The pluses (+)
are measured NO; asterisks (�) are the sum of total measured gas
phase nitrates; triangles (n) are measured O3; and diamonds (B)
are measured 1,4-butenedial and 4-oxo-2-pentenal. TotNO3
represents the sum of gas phase PAN, organic nitrates and
acylperoxy nitrates.
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6469
they were conducted primarily for gas phaseproduct analysis.
Particle data, however, exist for one 1,4-butene-dial experiment and one 4-oxo-2-pentenal experi-ment that were carried out in chamber A of the195m3 EUPHORE chambers (Johnson et al., 2004;Bloss et al., 2005). Initial reactant concentrationsare listed in Table 1. Experiment conditions used formodel simulation, such as the particle wall deposi-tion rate and the relative humidity, etc., were thesame as used for the three EUPHORE tolueneexperiments. Because no measured NOx and O3
concentrations were available for these two experi-ments, only the decay of parent compounds and theformation of SOA are simulated with the toluenemechanism. As is illustrated in Fig. 5, the modelwell predicts particle formation from the photo-oxidation of 1,4-butenedial but underestimates the1,4-butenedial decay. The model for the experimentof 4-oxo-2-pentenal, accurately tracks the decay ofthe 4-oxo-2-pentenal, but greatly over-predicts themeasured SMPS particle mass concentrations. Itshould be noted that very low aerosol massconcentrations were observed by SMPS for 4-oxo-2-pentenal, which were less than 8 mgm�3, while the
maximum SMPS particle mass concentration ob-served for 1,4-butenedial was 97 mgm�3.
It is hard to understand why the aerosol massesformed in these two systems differ that much, sincethese two unsaturated dicarbonyl compounds havesimilar structures and also form similar productsfrom their gas phase degradations. Johnson et al.
ARTICLE IN PRESS
Gas phase glyoxal and methylgloxal from 070298
0.00
0.07
0.14
0.21
0.28
0.35
6:00 8:00 10:00
LDT (Hours)
Gly
oxal
& M
eth
yl
gly
oxal
co
nc (
pp
mV
) MGLY_m
Mgly
GLY_m
Glyoxal
Gas phase glyoxal from 061898
0.0
0.1
0.1
0.2
0.2
6:00 8:00 10:00 12:00 14:00
LDT (Hours)
Gly
oxal
co
nc (
pp
mV
)
Gly_m
Glyoxal
Fig. 4. Simulation (line) vs. experiment data (connected symbols)
of gas phase glyoxal and methyl glyoxal from 1,4-butenedial and
4-oxo-2-pentenal experiments.
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776470
tried to justify this difference of SOA mass by thedifference between NO concentrations in these twosystems, where the ratio of 4-oxo-2-pentenal to NOwas 1.6 and the ratio of 1,4-butenedial to NO was8.3. As we discussed later in Section 2.4.1, the initialexperimental ratio of hydrocarbon to NO servesas an indicator of the balance between GLYPOLYand other oxygenated products in the aerosolphase. Speciation of simulated aerosol material
was determined for the two unsaturated dicarbonylssystems. The simulated aerosol phase products weresimilar in both systems. The major SOA masscontributors were GLYPOLY and a C4 carbonyl-carboxylic acid-nitrate (BUDACIDONO3) for 1,4-butenedial, and were GLYPOLY and a C5 hydroxycarbonyl nitrate (ROHALDNO3) for 4-oxo-2-pen-tenal. At the end of the model simulation, morethan 70% of the aerosol mass in both systemscame from the two major aerosol phase products.Therefore, the difference in NO concentrationsseems to be an inadequate explanation for theobserved phenomenon.
One EUPHORE o-cresol experiment was alsosimulated by the UNC toluene mechanism (Johnsonet al., 2004). Cresol showed a very high potential forSOA formation; 430 mgm�3 of suspended particlemass (measured SMPS volume data� 1.4) wasgenerated after the system reacted for 3 h. Theinitial concentrations for o-cresol and NO were298 ppbV and 24 ppb as listed in Table 1. Again nomeasured NOx and O3 concentrations were avail-able, therefore only the decay of o-cresol and theformation of SOA were simulated. The modelgreatly under-predicts the o-cresol decay (Fig. 5).This indicates there may be other unknown reactionpathways that remove o-cresol in the system. Themodel under-estimates maximum corrected SMPSmass by 53%, and this may be due to the under-prediction of o-cresol decay.
2.3. Model simulations of experimental data from
California Institute of Technology
Four sets of experimental results from the 60m3
outdoor Teflon bag reactor located at the CaliforniaInstitute of Technology (Caltech) (Odum et al., 1997)were simulated. An aerosol density of 1.4 g cm�3 wasassumed, and this factor was applied to correct thereported SMPS particle mass concentrations. Theinitial experimental conditions, the reported andcorrected SOA mass concentrations and yields, andthe simulation results are shown in Table 2. Modelsimulations were run with 5mgm�3 injectedinorganic seed aerosol and 250 ppbV propylene(Odum et al., 1997). The typical mean diameter ofseed aerosol used for all Caltech experiments was100 nm and the particle diameters increased withmore aerosol mass (Bowman et al., 1997). A meanparticle diameter of 150 nm was applied to all modelsimulations. The relative humidity and the aerosolphase water content used for model simulation were
ARTICLE IN PRESS1,4
-bu
ten
ed
ial co
nc (
pp
mV
)
0
20
40
60
80
Part
icle
mass c
on
c
LDT (hours)4-o
xo
-2-p
en
ten
al co
nc (
pp
mV
)
0
160
Part
icle
mass c
on
c
o-c
reso
l co
nc (
pp
mV
)
0
100
200
300
400
500
Part
icle
co
nc
o-cresol
TSP
EUPHORE 1,4-Butenedial
0.8
0.6
0.4
0.2
0.0
1,4-ButenedialTSP
9:00 11:00 13:00 15:00
LDT (hours)
140
120
100
0.4
0.3
0.2
0.1
0.0
EUPHORE 4-oxo-2-pentenal200
120
80
40
15:0013:0011:009:00
EUPHORE o-Cresol0.35
0.28
4-oxo-2-pentenalTSP
0.21
0.14
0.07
0.0010:00 11:00 12:00 13:00 14:00 15:00 16:00
LDT (hours) (
g
/m3)
( g
/m3)
( g
/m3)
Fig. 5. Comparison of model simulated and measured concentrations of 1,4-butenedial, 4-oxo-2-pentenal, o-cresol and aerosol from
EUPHORE chamber data sets. The solid lines (—) represent model simulations. The diamonds (B) are measured hydrocarbons and
squares (&) are calculated SMPS particle mass concentrations.
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6471
5% and 0.5% (Jang and Kamens, 1998). The first-order rate constants for dilution, O3 wall depositionand particle wall deposition were 0.005 h�1,1� 10�6 s�1 and 5� 10�5 s�1, respectively (Bowmanet al., 1997). The latitude of the Caltech chamber isalso almost identical to the UNC chamber, with thedifference of solar zenith angles for these twochambers in early July at 10 a.m. being just about0.51. Thus the UNC 29 June 1996 light file startingfrom 10 a.m. was used to run all the simulations. As
shown in Fig. 6, the maximum aerosol massconcentrations (M0_max) simulated by the UNCtoluene mechanism agree well with the correctedCaltech SMPS particle mass concentrations (M0)computed with a 1.4 g cm�1 density.
2.4. Model uncertainties
The greatest uncertainties in the model are due toorganic nitrate formation and the uptake of glyoxal
ARTICLE IN PRESS
Table 2
Initial conditions of experiments conducted in Caltech chambers that are used to evaluate the UNC toluene mechanism
Experiment
ID
Initial concentration (ppm) Temp.
(K)
Reporteda Correctedc Simulated
[Tol]0 [NO]0 [NO2]0 DM0
(mgm�3)Yb (%) DM0
(mgm�3)Yb(%) DM0_max
(mgm�3)Yb (%)
06/28/96a 0.680 0.443 0.242 303 133 0.094 186 0.132 195 0.121
07/01/96a 0.642 0.200 0.145 315 111 0.088 155 0.123 122 0.078
07/05/96a 0.706 0.488 0.305 312 171 0.100 239 0.140 213 0.121
07/10/96b 0.815 0.149 0.104 307 68 0.074 95 0.103 102 0.067
aReported aerosol mass concentrations were calculated from measured SMPS particle volume concentration by assuming a particle
density of 1.0 g cm�3.bY ¼ DM0/DHC, where DHC is reacted toluene (mgm�3).cA factor of 1.4 is applied to reported aerosol masses and yields.
50
100
150
200
250
300
corrected M0
Pre
dic
ted
M0_m
ax
1:1 line
30025020015010050
( g
/m3)
( g/m3)
Fig. 6. Comparisons of predicted maximum aerosol mass
concentrations with measured aerosol mass concentrations from
Caltech chamber by assuming particle density of 1.4 g cm�3 (solid
diamonds).
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776472
to aerosol phase. Glyoxal, organic nitrate, and otherproduct accumulations in the particle phase weresubsequently impacted by temperature and humid-ity. In addition, the different experiments startedwith different amounts of chamber backgroundaerosols that were present as seed aerosols forgas-particle partitioning of semi-volatile organicproducts. The sensitivity of the model to variousphysical and chemical parameters was determinedby varying the values of the parameters. By doingso, we tried to answer the following questions: (1)How can temperature and relative humidity affectaerosol formation? (2) What changes in the con-centration—time profiles of O3 and aerosol are
expected by varying the yields of organic nitratesand the uptake coefficient of glyoxal? and (3) Do thebackground aerosols (SEED) in the chambers havean important impact on the particle generation?
2.4.1. Effect of temperature on volatile vs. non-
volatile SOA
Experiments, from the UNC database (Table 3)and used for model development by Hu et al.(2007), were simulated by varying the diurnaltemperature profile over a range of 710 1C. Witha model temperature change of 710 1C, the max-imum simulated aerosol mass (M0_max) changed byno more than 710% to 18% over all of theexperiments (Table 3); increasing temperaturedecreases aerosol production and vice versa. Pos-sible temperature effects on SOA formation wouldbe impacted by: how temperature influences simplegas-particle partitioning of semi-volatile products,the impact of temperature on the rate of gas phaseformation of these products, and the extent to whichthe resulting particle phase is composed of non-volatile oxygenated products.
As was described in the companion paper (Huet al., 2007), the uptake rate of glyoxal ontoparticles is directly related to organic aerosol mass.The balance between GLYPOLY and other oxyge-nated products in the particle phase, such as organicnitrates, will be influenced by the availability oforganic aerosol mass for the uptake of glyoxal, theabsorption of semi-volatile products, and the abilityto form organic nitrates, acylperoxy nitrates andmethyl nitro-phenol analogues. In the UNC toluenedata sets, two possible indicators of this balance arethe initial toluene concentration and the initialexperimental ratio of toluene to NO. High and midtoluene/NO concentration ratios were defined as
ARTICLE IN PRESS
Table 3
The model sensitivity to temperature and relative humidity
Experiment date Initial concentration (ppm) [Tol]0/[NO]0 Temp. (K) M0_max/M0_max_basea M0_max_80%/M0_max_10%
c
[Tol]0 [NO]0 Temp+10 Kb Temp�10 Kb RH effect
092304N 0.500 0.204 2.45 296–307 �8% +8% +13%
092304S 0.980 0.280 3.50 296–307 �1% +1% +12%
101604N 1.023 0.096 10.66 288–300 +2% +6% +3%
101604S 0.540 0.370 1.46 288–300 �7% +5% +8%
111504N 0.864 0.272 3.18 284–294 �8% +18% +2%
072705N 0.186 0.091 2.04 300–310 �14% +18% +19%
072705S 0.095 0.091 1.04 300–310 �14% +16% +12%
aM0_max_base is the maximum simulated aerosol mass concentration at the base case. M0_max is the maximum simulated aerosol mass
concentration when temperature changed by 710K.bThe diurnal temperature was changed by 710K with relative humidity unchanged.cM0_max_80% is the maximum simulated aerosol mass concentration at a relative humidity of 80%. M0_max_10% is the maximum
simulated aerosol mass concentration at a relative humidity of 10%.
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6473
42 and low toluene/NO ratios as p2. High tolueneconcentrations were defined as 40.5 ppmV and lowas p0.1 ppmV. Given the simulated speciationof aerosol material at the end of experiments listedin Table 3, more oligomers, i.e. the sum ofGLYPOLY, MGLYPOLY and SEED1, and lessorganic nitrates and acylperoxy nitrates were pre-dicted in the systems with high toluene concentra-tions and high to mid toluene/NO ratios than in thesystems with low concentrations and low toluene/NO ratios. As aerosol formation is dominated byparticle phase reactions to form polymers andSEED1, the model is less sensitive to temperaturechanges despite the important temperature-depen-dent relationship on the gas-particle partitioningprocess. This is because resulting particle phaseoligomers are highly non-volatile and phasere-distribution will not be overly influenced bytemperature changes. When the temperature de-creases, more acylperoxy nitrates and methyl nitro-phenol analogues are formed. Since the particledesorption rate, koff, decreases with decreasingtemperature, a bit more particle mass will begenerated through gas-particle partitioning pro-cesses. But with more acylperoxy radicals reactingto acylperoxy nitrates, less glyoxal, methyl glyoxaland their oligomers are formed due to lower OHand HO2 production.
2.4.2. Effect of relative humidity
The effect of relative humidity on both O3 andaerosol formation was evaluated for all experimen-tal conditions. It was found that M0_max increaseswhen relative humidity was increased from 10% to
80%, but this change was usually less than 13%except for the low concentration run 072705N,where the change was 19%. This is consistent withCocker et al. (2001), who did not observe a dramatichumidity impact on SOA formation from aromaticsystems. Toluene decays faster at higher relativehumidity. The maximum O3 concentration de-creases by less than 10% with relative humiditychanging from 10% to 80%. For systems with lowtoluene/NO ratios, such as 072705S (Table 3), O3
concentrations first accumulated faster at higherrelative humidity, but slowed down later in theexperiment; hence lower maximum O3 concentra-tions were observed when relative humidity wasvaried from 10% to 80%.
We have previously reported (Jang and Kamens,1998) that the water content in the aerosol phaseincreases from 0.5% to 4% over a 10–80% relativehumidity change. In our mechanism, particle phasewater reacts with NO2 and converts it into particlephase HNO3. Therefore, with increasing relativehumidity, more gas phase NO2 is shunted intoaerosol phase HNO3 and less O3 will be generated.For systems with low toluene/NOx ratios, theaerosol mass formed in the early stages of thesimulation was not high enough to remove all excessNO2. Therefore, at higher relative humidity O3
first appeared faster and then slowed down withparticle accumulation. As has been discussed earlier,GLYPOLY, MGLYPOLY and organic peroxidesare the major aerosol mass contributors (Hu et al.,2007). With more NO converted to NO2 at higherrelative humidities, less NO reacts with RO2 andmore RO2 reacts with toluene to form RO without
ARTICLE IN PRESS
O3 c
on
c (
pp
m)
O3 c
on
c (
pp
m)
0
O3
0
5
20
O3
092304S1.0
0.8
0.5
0.3
0.0
TSP
LDT (hours)
16:0015:0014:0013:0012:0011:0010:009:00
Part
icle
co
nc
Part
icle
co
nc
200
300
100
072705S 25
15
10TSP
0.6
0.5
0.4
0.3
0.2
0.1
0.017:0015:00
LDT (hours)
7:00 9:00 11:00 13:00
( g
/m3)
( g
/m3)
Fig. 7. Comparison of simulated and measured concentrations of
O3 and aerosol by changing nitrate yields by 750% on 092304S
and 072705S: base case (—), +50% (– –), and �50% (- -).
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776474
consuming NO. More oxygenated products andHO2 from RO decomposition are formed, leading tomore GLYPOLY, MGLYPOLY and organic per-oxides in the aerosol phase.
2.4.3. Effect of organic nitrates and acylperoxy
nitrates
As with the a-pinene and d-limonene mechanisms(Kamens and Jaoui, 2001; Leungsakul et al., 2005),two of the major uncertainties in the current toluenemechanism are the actual yields of organic nitrates(RONO2) that are formed from RO2+NO reac-tions, and the decomposition rate of acylperoxynitrates (RCO-OONO2). The effect of yields oforganic nitrates on both O3 production and aerosolformation were evaluated for all experimentalconditions. An integrated rate analysis (sum of allof the mass flowing through a given reactionpathway) showed that more than 60% of theintegrated mass of the organic nitrates for all theRO2+NO reactions went through the reactionbetween TOLO2 and NO. About 15% of theintegrated mass went through the reactions of NOwith RO2 radicals formed from cresol. Based on ourmodel fits to experimental SOA data, the yields ofall the organic nitrates formed in the toluene systemare assumed to be the same, which is 0.14.Speciation of the simulated SOA material showsthat up to 85% of the particle phase organic nitratescame from two bicylic organic nitrates with multihydroxyl groups formed from cresol reactions(RgDOHNO3 and RgTOHNO3). By varying theRONO2 yields by 750%, similar effects on O3
were observed for all eight experiments, whereas theeffects on SOA formation were related to initialreactant concentrations. In general, with smallervalues of RONO2 yields, O3 accumulated faster andhad higher maximum concentration; predictedaerosol mass concentrations decreased in the lowtoluene concentration systems but increased in thehigh concentration systems (Fig. 7).
As mentioned earlier, the uptake rate of glyoxalincreases with increasing organic aerosol mass. In thehigh toluene concentration systems such as 092304S,where high aerosol mass was formed, GLYPOLYwould contributed 61% of the particle mass at theend of experiment and organic nitrates contributed7%. In the low toluene concentration systems such as072705S, much less aerosol was generated andGLYPOLY only account for 17% of the aerosolmass. Particle phase organic nitrates became moresignificant than in high volume systems, and
contributed 26% to the total particle mass. Withincreasing nitrated yields, fewer TOLO radicals andless gas phase glyoxal and particle phase GLYPOLYwere generated. Since GLYPOLY is a much moresignificant contributor to SOA mass than organicnitrates in the high toluene concentration systems,the total aerosol mass decreased even thoughthere was an increased organic nitrate production.For the low toluene concentration system, organic
ARTICLE IN PRESS
O3 c
on
c (
pp
m)
O3 c
on
c (
pp
m)
0
50
100
150
200
Part
icle
co
nc
Part
icle
co
nc
O3
0
5
10
15
20
25
O3
092304N1.0
0.8
0.5
0.3
0.0
TSP
LDT (hours)
16:0015:0014:0013:0012:0011:0010:009:00
072705S 0.6
0.5
0.4
0.3
0.2
0.1
0.017:0015:0013:0011:009:007:00
LDT (hours)
TSP
( g
/m3)
( g
/m3)
Fig. 8. Comparison of simulated and measured concentrations of
O3 and aerosol by changing the uptake coefficient of glyoxal by
750% on 092304N and 072705S: base case (—), +50% (– –),
and �50% (- -).
D. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6475
nitrates are as important as GLYPOLY in theparticle phase. Overall, however, when yields oforganic nitrates were varied by 750%, the changesof maximum O3 and SOA mass concentrations werewithin 715%.
Model simulations were not very sensitive tochanges in acylperoxy organic nitrate thermaldecomposition rate constants. When the decom-position rate was varied by 750%, almost identicalO3 and aerosol mass concentration profiles wereobtained. This is because excess NO2 released backby the decomposition of PAN-type compoundscould be removed by particle phase water.
2.4.4. Effect of the uptake of glyoxal
The uptake coefficient of glyoxal, q, is a principalkinetic parameter that controls particle phasereactions in the toluene mechanism as discussed byLiggio et al. (2005) and Hu et al. (2007). The uptakecoefficient for a given gas phase species is the ratioof the absorption rate by the particle phase to thetotal collision rate with the particle phase. Modelsensitivity analyses were performed to demonstratehow aerosol growth and O3 generation wereaffected by the uptake coefficient of glyoxal. Modelsimulations were run with q of 5.7E-4 as a base case,where the rate was chosen based on best model fits(Hu et al., 2007). Again, since the glyoxal uptakerate is a function of aerosol mass, the effect of theuptake coefficient of glyoxal on aerosol formationbecomes greater with increasing particle mass andwith the increasing mass fraction of GLYPOLY inthe particle. When the uptake coefficient was variedby 750%, for a low toluene concentration experi-ment (072705S), M0_max changed by 710%; for ahigh toluene concentration experiment (092304N)M0_max changed by +23% and �43% (Fig. 8). Thischange had almost no perceptible influence on theO3 time–concentration curves for all of the experi-mental conditions.
2.4.5. Effect of background seed
Background aerosol concentrations in the cham-ber at the beginning of the experiments usuallyranged from 1 to 9 mgm�3. These background seedsare assumed to be largely inorganic aerosol and theyprovide the initial particle surfaces for gas-particlepartitioning in the mechanism. Rapid particlegeneration by homogeneous nucleation was ob-served in the presence of these background aerosolsafter toluene was added to the chamber and thesenewly formed nuclei provide additional aerosol
surfaces for initial gas-particle partitioning andparticle phase reactions in the system. Modelsimulations were run without initial backgroundseed, and it was found that the background aerosolhad almost no discernable influence on O3 and SOAformation in high concentration experiments. Butwith decreasing toluene/NO ratios, the mass frac-tion of total oligomers in particle phase willdecrease and gas-particle partitioning becomes moreimportant for aerosol mass accumulation. As forexperiments of 072705N, 072705S and 101604S,
ARTICLE IN PRESSD. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–64776476
simulated SOA mass decreased without backgroundaerosol, but just by less than 10%.
3. Summary
Experimental data sets from other investigatorswere used to evaluate the toluene mechanism. Themodel well simulates SOA formation from the ‘low-NOx’ and ‘mid-NOx’ experiments conducted in theEUPHORE chambers, but over-predictes SOAmass concentrations from the ‘high-NOx’ run. Thismay be due to not experimentally considering thetotal size distribution of particles in the EUPHOREchamber measurements. The model well simulatesthe SOA mass concentrations observed from theCaltech chambers. Simulations of 1,4-butenedial,4-oxo-2-pentenal and o-cresol in the presence ofnatural sunlight and NOx were also undertakenwith the developed mechanism. The model wellpredicts the NOx time–concentration profiles andthe decay of these two carbonyls, but under-estimates O3 formation for 4-oxo-2-pentenal. It wellsimulates SOA formation from 1,4-butenedial butoverestimates (possibly due to experimental pro-blems) the measured aerosol mass concentrationsfrom 4-oxo-2-pentenal. The model underestimatesSOA production from o-cresol, mostly due to itsunder-prediction of o-cresol decay (Fig. 5).
What impacts model SOA generation the mostare processes that lead to glyoxal oligomers andorganic nitrates formation, and future work shouldquantify these processes. Changing the temperatureby 710 1C affected SOA formation by no morethan 10–18%. Increasing the humidity from 10% to80% increased SOA production by less than 13%for almost all conditions. Changing organic nitrateyields by 750% increased SOA production in thelow concentration systems but decreased it in thehigh concentration runs; the change of SOA mass,however, was within 715%. The model is sensitiveto the uptake coefficient of glyoxal for the highconcentration runs. By varying its value by 750%,SOA product would be changed by up to 43%.Model simulations were not sensitive to the decom-position rates of acylperoxynitrites, and back-ground seed aerosol.
Acknowledgement
This work is supported by a STAR grant from theUS EPA (RD-83108401) to the University of NorthCarolina at Chapel Hill.
References
Bahreini, R., Keywood, M.D., Ng, N.L., Varutbangkul, V., Gas,
S., Flagan, R.C., Seinfeld, J.H., Worsnop, D.R., Jimenez,
J.L., 2005. Measurements of secondary organic aerosol from
oxidation of cycloalkenes, terpenes, and m-xylene using an
aerodyne aerosol mass spectrometer. Environmental Science
and Technology 39 (15), 5674–5688.
Bloss, C., Wagner, V., Jenkin, M.E., Volkamer, R., Bloss, W.J.,
Lee, J.D., Heard, D.E., Wirtz, K., Martin-Reviejo, M., Rea,
G., Wenger, J.C., Pilling, M.J., 2005. Development of a
detailed chemical mechanism (MCMv3.1) for the atmospheric
oxidation of aromatic hydrocarbons. Atmospheric Chemistry
and Physics 5, 641–664.
Bowman, F.M., Odum, J.R., Seinfeld, J.H., 1997. Mathematical
model for gas-particle partitioning of secondary organic
aerosols. Atmospheric Environment 31 (23), 3921–3931.
Broyles, D.A., Carpenter, B.K., 2005. Experimental detection of
one case of benzene epoxidation by a peroxy radical and
computational prediction of another. Journal of Organic
Chemistry 70 (21), 8642–8644.
Cocker, D.R.I., Mader, B.T., Kalberer, M., Flagan, R.C.,
Seinfeld, J.H., 2001. The effect of water on gas-particle
partitioning of secondary organic aerosol: II. m-xylene and
1,3,5-trimethylbenzene photooxidation systems. Atmospheric
Environment 35 (35), 6073–6085.
Griffin, R.J., Dabdub, D., Seinfeld, J.H., 2002. Secondary
organic aerosol 1. Atmospheric chemical mechanism for
production of molecular constituents, Journal of Geophysical
Research, [Atmosphere] 107(D17), AAC3–1-AAC3/26.
Hu, D., Tolocka, M., Li, Q., Kamens, R.M., 2007. A kinetic
mechanism for predicting secondary organic aerosol forma-
tion from toluene oxidation in the presence of NOx and
natural Sunlight. Atmospheric Environment, doi:10.1016/
j.atmosenv.2007.04.025.
Jang, M., Kamens, R.M., 1998. A thermodynamic approach for
modeling partitioning of semivolatile organic compounds on
atmospheric particulate matter: humidity effects. Environ-
mental Science and Technology 32 (9), 1237–1243.
Jang, M., Kamens, R.M., 2001. Characterization of secondary
aerosol from the photooxidation of toluene in the presence of
NOx and 1-propene. Environmental Science and Technology
35 (18), 3626–3639.
Johnson, D., Jenkin, M.E., Wirtz, K., Martin-Reviejo, M., 2004.
Simulating the formation of secondary organic aerosol from
the photooxidation of toluene. Environmental Chemistry 1
(3), 150–165.
Kalberer, M., Paulsen, D., Sax, M., Steinbacher, M., dommen, J.,
Prevot, A.S.H., Fisseha, R., Weingartner, E., Frankevich, V.,
Zenobi, R., Baltensperger, U., 2004. Identification of poly-
mers as major components of atmospheric organic aerosols.
Science 303 (5664), 1659–1662.
Kamens, R.M., Jaoui, M., 2001. Modeling aerosol formation
from a-pinene+NOx in the presence of natural sunlight using
gas-phase kinetics and gas-particle partitioning theory.
Environmental Science and Technology 35 (7), 1394–1405.
Leungsakul, S., Jeffries, H.E., Kamens, R.M., 2005. A kinetic
mechanism for predicting secondary aerosol formation from
the reactions of d-limonene in the presence of oxides of
nitrogen and natural sunlight. Atmospheric Environment 39
(37), 7063–7082.
ARTICLE IN PRESSD. Hu, R.M. Kamens / Atmospheric Environment 41 (2007) 6465–6477 6477
Liggio, J., Li, S.-M., McLaren, R., 2005. Reactive uptake of
glyoxal by particulate matter. Journal of Geophysical
Research 110, D10304.
Liu, X., Jeffries, H.E., Sexton, K.G., 1999. Atmospheric
photochemical degradation of 1,4-unsaturated dicarbonyls.
Environmental Science and Technology 33 (23), 4212–4220.
Odum, J.R., Jungkamp, T.P.W., Griffin, R.J., Forstner, H.J.L.,
Flagan, R.C., Seinfeld, J.H., 1997. Aromatics, reformulated
gasoline, and atmospheric organic aerosol formation. Envir-
onmental Science and Technology 31 (7), 1890–1897.
Pun, B.K., Griffin, R.J., Seigneur, C., Seinfeld, J.H., 2002. Secondary
organic aerosol 2, thermodynamic model for gas/particle
partitioning of molecular constituents. Journal of Geophysical
Research, [Atmosphere] 107(D17), AAC4–1–AAC4/15.
Stroud, C.A., Makar, P.A., Michelangeli, D.V., Mozurkewich,
M., Hastie, D.R., Bardu, A., Humble, J., 2004. Simulating
organic aerosol formation during the photooxidation of
toluene/NOx mixtures: comparing the equilibrium and kinetic
assumption. Environmental Science and Technology 38 (5),
1471–1479.