Introduction
Organic aerosols are critical to human health (Dockery et al., 1993;
Krewski et al., 2003; Davidson et al., 2005), climate change (IPCC, 2007)
and visibility (Pöschl, 2005; Seinfeld and Pandis, 2006). Global
anthropogenic secondary organic aerosol (SOA) sources are underestimated by
current models (Henze et al., 2008; Matsui et al., 2009; Hallquist et
al., 2009; Farina et al., 2010) and are more likely to increase due to the
increase of known anthropogenic emissions (Heald et al., 2008). Therefore,
it is crucial to explore SOA formation mechanisms from anthropogenic
precursors.
Aromatic hydrocarbons are
major anthropogenic SOA precursors (Kanakidou et al., 2005; Henze et al.,
2008; Derwent et al., 2010). C8 (ethylbenzene, xylenes) and C9
(ethyltoluenes and trimethylbenzenes) aromatics are important aromatic
hydrocarbons in the atmosphere besides toluene and benzene (Monod et al.,
2001; Millet et al., 2005; Heald et al., 2008; Kansal, 2009; Hu et al.,
2015). The major sources of C8 and C9 aromatic hydrocarbons are
fuel evaporation (Kaiser et al., 1992; Rubin et al., 2006; Miracolo et al.,
2012), tailpipe exhaust (Singh et al., 1985; Monod et al., 2001; Lough et
al., 2005; Na et al., 2005; Correa and Arbilla, 2006) and solvent use (Zhang
et al., 2013). C8 aromatic hydrocarbons (ethylbenzene and xylenes
(ortho, meta and para) are categorized as hazardous air pollutants under the
US Clean Air Act Amendments of 1990;
http://www.epa.gov/ttnatw01/orig189.html). Toluene and C8
aromatics dominate the anthropogenic SOA precursors and SOA yield from all
C9 aromatics is currently predicted to be equal to that of toluene
(Bahreini et al., 2009). The chemical composition of aromatic SOA remains
poorly understood with less than 50 % of aromatic hydrocarbon
photooxidation products identified (Forstner et al., 1997; Fisseha et al.,
2004; Hamilton et al., 2005; Sato et al., 2007). Aromatic hydrocarbon
photooxidation mechanisms remain uncertain except for the initial step
(∼ 90 % OH-addition reaction) (Calvert et al., 2002). Hence,
understanding the atmospheric reaction mechanisms of C8 and C9
aromatic hydrocarbons and properly quantifying their SOA formation potential
presents unique challenges due to the variety in their molecular structure
and the electron density of the aromatic ring.
Volatile organic compound (VOC) structure impacts the gas-phase reaction
mechanism (Ziemman and Atkinson, 2012) and kinetic reaction rate (eg.
kOH; Atkinson, 1987), thereby influencing the resulting SOA properties and
mass yield. Molecular structure impacts on SOA formation from alkanes have
been previously studied (Lim and Ziemann, 2009; Ziemann, 2011; Lambe et al., 2012; Tkacik et al., 2012; Yee et al., 2013; Loza et al., 2014). It
is generally observed that SOA yield decreases from cyclic alkanes to linear
alkanes and to branched alkanes. The relative location of the methyl group
on the carbon chain also affects SOA yield (Tkacik et al., 2012). It is
further found that the SOA yield and structure relationship is influenced by
C=C groups (Ziemann, 2011). Understanding the SOA yield and structure
relationship of aromatic compounds in a similar way is necessary due to the
atmospheric importance of aromatic hydrocarbons.
Previously, aromatic studies categorized SOA yield solely based on substitute
number (Odum et al., 1997a, b). However, those chamber experiments were
conducted at high-NOx conditions, which are well above levels present in
the atmosphere. Song et al. (2005, 2007) found that initial
HC / NOx ratios (HC being hydrocarbon)
significantly impact SOA yields during aromatic photooxidation with yields
increasing as NOx levels decreased. Ng et al. (2007) show there is no
significant yield difference between one-substitute (toluene) and
two-substitute (m-xylene) aromatics in the absence of NOx. The current
work focuses on molecular structure impact on SOA formation at more
atmospherically relevant NOx and aerosol loadings. Li et al. (2016)
demonstrated the methyl-group-number impact on SOA formation under
low-NOx conditions. Also, aromatic compounds with para-position alkyl
groups have been observed to form less SOA under various NOx conditions
than their isomers in previous studies. Izumi and Fukuyama (1990) found that
p-xylene, p-ethyltoluene and 1,2,4-trimethylbenzene have low SOA
formation potential under high-NOx conditions. Song et al. (2007)
observed that p-xylene has the smallest SOA yield among all xylenes in the
presence of NOx. The relative methyl position to -OH in dimethyl phenols
also impacts SOA yield in the absence of NOx (Nakao et al., 2011), while
Song et al. (2007) observed no significant SOA yield difference between
o-xylene and p-xylene under NOx-free conditions. Moreover, previous
studies mainly focused on the carbon number effect on SOA formation (Lim and
Ziemann, 2009; Li et al., 2016) and seldom addressed the substitute carbon
length impact on VOC oxidation and hence SOA formation. Different percentages
of similar compounds are found when the substitute carbon length on the
aromatic ring changes (Forstner et al., 1997; Huang et al., 2007, 2011). For
example, a lower percentage of 3-methyl-2,5-furanone is observed in toluene
than that of 3-ethyl-2,5-furanone in ethylbenzene (Forstner et al., 1997).
Further, the branching structure on the aromatic substitute might impact the
reaction pathway. It is possible that fragmentation is more favored on
branched substitute alkoxy radicals than n-alkane substituents similar to
alkanes (Atkinson and Arey, 2003).
Few studies comprehensively consider the overall alkyl effect on SOA
formation from aromatic hydrocarbons, including the substitute number,
position, carbon chain length and branching structure, especially under low-NOx conditions. It is valuable to understand the relationship between
aromatic hydrocarbon molecular structures and SOA physical and chemical
characteristics. The effects of OH exposure (Lambe et al., 2012,
2015), mass loading (Shilling et al., 2009, α-pinene;
Pfaffenberger et al., 2013, α-pinene) and NO condition (Ng et al.,
2007; Eddingsaas et al., 2012, α-pinene) on SOA physical and
chemical characteristics are previously discussed. However, few studies
address the molecular structure effect of the precursor on SOA chemical
composition, especially under atmospherically relevant conditions. Sato et
al. (2012) show the chemical composition difference between ethylbenzene,
m-xylene, o-xylene, 1,2,4-trimethylbenzene and 1,3,5-trimethylbenzene
under high absolute NOx conditions and hypothesize that ketones
prevent further oxidation during aromatic photooxidation compared with
aldehydes. The SOA products detected in Sato's study are mainly small
volatile compounds which are less likely to partition into the particle
phase (Chhabra et al., 2011). Therefore, the study of Sato et al. (2012)
indicates that further oxidation or oligomerization might contribute to SOA
formation during aromatic photooxidation. Less SOA characterization data on
propylbenzene and ethyltoluene compared with trimethylbenzene are available.
However, Bahreini et al. (2009) suggest that the sum of the propylbenzene
and ethyltoluene is on average a factor of 4–10 more abundant than
trimethylbenzene.
This work examines 12 aromatic hydrocarbons, all of which are isomers
with eight or nine carbons, to investigate the impact of molecular structure
on SOA formation from aromatic hydrocarbon photooxidation under low NOx (10–138 ppb). Here, we investigate the substitute number, substitute
position, alkyl carbon chain length and alkyl branching impacts on aromatic
hydrocarbon oxidation. The effects of molecular structure impact on SOA
yield, chemical composition (H / C, O / C, OSc, f44, f43,
f57 and f71) and physical properties (density and VFR) are
demonstrated. Alkyl substitute dilution conjecture is further developed from
methyl dilution theory (Li et al., 2016).
Method
Environmental chamber
The UC Riverside/CE-CERT indoor dual 90 m3 environmental chambers were
used in this study and are described in detail elsewhere (Carter et al.,
2005). Experiments were all conducted at dry conditions (RH < 0.1 %), in the absence of inorganic seed aerosol and with temperature
controlled to 27 ± 1 ∘C. Seeded experiments to minimize wall
effects have also been conducted in our chamber experiment with no
measurable difference observed between the seeded and non-seeded experiment.
Two movable top frames were slowly lowered during each experiment to
maintain a slight positive differential pressure (∼ 0.02′′
H2O) between the reactors and enclosure to minimize dilution and/or
contamination of the reactors. Additionally, 272 115 W Sylvania 350BL blacklights are used
as light sources for photooxidation.
A known volume of high-purity liquid hydrocarbon precursors from Sigma-Aldrich (ethylbenzene, 99.8 %; n-propylbenzene, 99.8 %;
isopropylbenzene, analytical standard; m-xylene,
99 %; o-xylene, 99 %;p-xylene, 99 %;
m-ethyltoluene, 99 %; o-ethyltoluene, 99 %; p-ethyltoluene, ≥ 95 %; 1,2,3-trimethylbenzene, OEKANAL
analytical standard; 1,2,4-trimethylbenzene, 98 %; 1,3,5-trimethylbenzene, analytical standard) was injected through
a heated glass injection manifold system and flushed into the chamber with
pure N2. NO was introduced by flushing pure N2 through a
calibrated glass bulb filled to a predetermined partial pressure of pure NO.
All hydrocarbons and NO are injected and well mixed before lights are turned
on to start the experiment.
Particle and gas measurement
Particle size distribution between 27 and 686 nm was monitored by dual
custom-built scanning mobility particle sizers (SMPS) (Cocker et al., 2001).
Particle effective density was measured with an aerosol particle mass
analyzer (APM-SMPS) system (Malloy et al., 2009). Particle volatility was
measured by a Dekati® thermodenuder volatility tandem
differential mobility analyzer (VTDMA) (Rader and McMurry, 1986) with a 17 s
heating zone residence time (Qi et al., 2010a). The heating zone was
controlled to 100 ∘C in this study with volume fraction remaining
(VFR) calculated as (Dp,afterTD/Dp,before,TD)3.
Particle-phase chemical composition evolution was measured by a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS; Aerodyne
Research Inc.) (Canagaratna et al., 2007; DeCarlo et al., 2006). The sample
was vaporized by a 600 ∘C oven under vacuum, followed by a 70 eV
electron impact ionization. fx in this study is calculated as
the mass fraction of the organic signal at m/z=x. For example, f44,
f43, f57 and f71 are the ratios of the organic signal at
m/z 44, 43, 57 and 71 to the total organic signal, respectively (Chhabra et
al., 2011; Duplissy et al., 2011). Elemental ratios for total organic mass,
oxygen to carbon (O / C) and hydrogen to carbon (H / C) were determined using
the elemental analysis technique (Aiken et al., 2007, 2008). Data were
analyzed with ToF-AMS analysis toolkit Squirrel 1.56D/Pika 1.15D version.
Evolution of SOA composition (Heald et al., 2010; Jimenez et al., 2009)
refers to SOA chemical composition changes with time. f44 and
f43+57+71 evolution and H / C and O / C evolution refer to the change of
f44 and f43+57+71 with time and the change of H / C and O / C with
time, respectively.
Experiment conditions.
Precursor
ID
HC / NO
NO
HC
ΔHC
Moa
Yield
ppbC : ppb
ppb
ppb
µg m-3
µg m-3
Ethylbenzene
1142A
17.0
47.4
101
331
22.0
0.066
1142B
12.0
66.6
99.9
341
4.40
0.013
1146A
35.6
22.2
99.0
257
36.0
0.140
1146B
23.0
34.8
100
331
23.6
0.071
1147B
74.9
36.5
342
626
88.1
0.141
2084A
81.1
23.9
242
374
54.0
0.145
2084B
93.8
20.3
238
266
44.3
0.167
Propylbenzene
1245A
41.0
22.1
101
231
11.8
0.051
1246A
26.8
68.5
204
421
22.9
0.054
Isopropylbenzene
1247A
40.3
22.4
100
301
33.2
0.110
1247B
18.6
48.1
99.3
300
16.6
0.055
1253A
31.9
56.4
200.
538
53.1
0.099
1253Bb
17.6
100
196
526
16.5
0.031
o-Xylene
1315A
13.2
49.8
82.2
324
26.3
0.081
1315B
28.8
22.2
80.0
27
25.4
0.091
1320A
12.8
50.0
80.0
335
18.4
0.055
1321A
31.0
20.5
79.2
263
16.2
0.061
1321B
61.3
10.4
80.0
226
9.80
0.044
p-Xylene
1308A
15.5
55.6
78.4
279
6.80
0.024
1308B
171
22.9
78.8
274
11.3
0.041
m-Ethyltoluene
1151A
17.9
62.5
84.8
409
8.30
0.020
1151B
31.0
32.3
86.4
415
28.7
0.069
1199A
8.8
45.4
100
447
72.0
0.161
1222B
41.7
69.4
100
484
70.9
0.146
1226B
11.3
137.6
201
895
138
0.154
1232A
27.5
122.0
200
901
150
0.167
1232B
33.1
67.5
194
751
117
0.155
1421A
41.0
22.1
97.9
409
46.2
0.112
1421B
18.0
44.9
98.7
477
54.6
0.114
o-Ethyltoluene
1179A
16.3
52.9
91.7
399
86.5
0.216
1179B
15.8
52.9
93.0
415
75.3
0.181
1202A
18.5
60.3
99.7
422
69.9
0.166
1215A
29.2
107
180
637
151
0.237
1413A
12.2
21.3
100
371
64.5
0.174
1413B
24.1
45.8
98.4
455
64.4
0.141
p-Ethyltoluene
1194A
19.9
90.7
196
741
90.4
0.122
1194B
13.0
88.4
200
761
73.0
0.096
1197A
13.1
56.4
192
653
66.4
0.102
1197B
14.8
98.5
192
710
58.4
0.082
1214B
26.0
53.4
102
418
29.1
0.069
1601A
39.9
31.2
109
452
17.6
0.039
1,2,3-Trimethylbenzene
1158A
19.8
10.3
79.9
296
22.2
0.075
1158B
15.6
22.4
79.9
379
32.3
0.085
1162A
15.8
33.4
80.1
391
46.5
0.119
1162B
14.9
40.0
80.4
399
46.6
0.117
1,3,5-Trimethylbenzene
1153A
65.2
11.0
79.5
309
12.4
0.040
1153B
35.3
20.4
80.0
381
19.6
0.051
1156A
22.3
32.3
80.2
379
24.8
0.065
1156B
15.5
46.1
79.6
390
19.0
0.049
1329B
11.1
64.8
80.0
296
3.00
0.007
Note: a Mo is a wall-loss-and density-corrected particle mass concentration. b Not used in
curve fitting.
The Agilent 6890 gas chromatograph–flame ionization detector was used to
measure aromatic hydrocarbon concentrations. A Thermal Environmental
Instruments model 42C chemiluminescence NO analyzer was used to monitor NO,
NOy–NO and NOy. The gas-phase reaction model SAPRC-11 developed by
Carter and Heo (2013) was utilized to predict radical concentrations
(⋅OH, HO2⋅, RO2⋅ and NO3⋅).
Result
SOA yield
Photooxidation of 12 C8 and C9 aromatic hydrocarbons was
studied for low-NOx conditions (HC / NO ratio 11.1–171 ppbC : ppb). SOA
yields for all aromatic hydrocarbons were calculated according to Odum et al. (1996) as the mass ratio of aerosol formed to parent hydrocarbon
reacted. Experimental conditions and SOA yields are listed (Table 1) along
with additional m-xylene, o-xylene,p-xylene and 1,2,4-trimethylbenzene
experimental conditions from previous studies (Song et al., 2005, 2007; Li et al., 2016) (Table S2 in the Supplement).
The uncertainty associated with 10 replicate m-xylene and NO experiments SOA yield is < 6.65 %. SOA
yield as a function of particle mass concentration (Mo), shown in Fig. 1, includes experiments listed in both Table 1 and Table S2. It is observed
that both alkyl substitute number and position affect SOA yield. The SOA
yield of two-substitute C8 and C9 aromatic hydrocarbons depends
more on the substitute location than substitute length. This means that the
yield trend of o-xylene is analogous to that of o-ethyltoluene. Similarly, the
yield trends for meta- and para-position substituted C8 and C9
aromatic hydrocarbons will be analogous to each other. Ortho isomers
(o-xylene and o-ethyltoluene, marked as solid and hollow green circles,
respectively) have the highest SOA yield for similar aerosol concentrations
while para isomers (p-xylene and p-ethyltoluene, marked as solid and hallow
blue diamonds, respectively) have the lowest SOA yield level. Lower SOA
yields for para isomers are consistent with previous observation by Izumi and
Fukuyama (1990). Izumi and Fukuyama (1990) also suggest that 1,2,4-trimethylbenzene yields are lower than for other aromatic hydrocarbons.
The current study does not show a significant SOA yield difference between
1,2,4-trimethylbenzene and 1,3,5-trimethylbenzene. It is difficult to
compare 1,2,3-trimethylbenzene yields with the former two
trimethylbenzenes since 1,2,3-trimethylbenzene mass loading is much higher
than the former two.
Aromatic SOA yields as a function of Mo. Note: Song et
al. (2005, 2007) and Li et al. (2016) data are also included; 123TMB is 1,2,3-Trimethylbenzene; 135TMB is 1,3,5-Trimethylbenzene; 124TMB is 1,2,4-Trimethylbenzene.
![]()
Aromatic hydrocarbons having only one substitute (ethylbenzene,
n-propylbenzene and isopropylbenzene) or three substitutes
(1,2,3-trimethylbenzene, 1,2,4-trimethylbenzene and 1,3,5-trimethylbenzene)
tend to have yields similar to the meta-position two-alkyl aromatics. Odum et
al. (1997b) categorized SOA yield formation potential solely based on
substitute number and stated that aromatics with less than two methyl or
ethyl substitutes form more particulate matter than those with two or more
methyl or ethyl substitutes on the aromatic ring. However, Odum's work was
conducted for high-NOx conditions and had insufficient data to compare
isomer yield differences (e.g., only two low mass loadings for o-xylene
data). The strong low-yield (two or more substitutes) and high-yield (less
than two-methyl or ethyl substitutes) trends for high-NOx conditions
(Odum et al., 1997a, b) are not observed for low-NOx aromatic
experiments in this study. Rather, high yield is observed only for benzene
(Li et al., 2016) while low yield is seen for substituted aromatic
hydrocarbons. Similar SOA yield trends from different C8 and C9
aromatic isomers are further confirmed by comparing yields at similar radical
conditions (Table S4, Fig. S3). It is also found that molecular structure
exerts a greater impact on SOA yield than reaction kinetics (Supplement,
Table S5). A two-product model described by Odum et al. (1996) is used to fit
SOA yield curves as a function of Mo. The 12 aromatics are categorized
into five groups to demonstrate the alkyl-group number and position effect on
SOA formation. The five groups include a one-substitute group, ortho-position
two-alkyl group (ortho), meta-position two-alkyl group (meta), para-position
two-alkyl group (para) and three-substitute group. Fitting parameters
(α1, Kom,1, α2 and Kom,2;
Table 2) in the two-product model are determined by minimizing the sum of the
squared residuals. The lower-volatility partitioning parameter
(Kom,2) is the same for all yield curve fits by assuming similar
high-volatile compounds are formed during all aromatic hydrocarbon
photooxidation experiments. The ortho group is associated with a much higher
Kom,1 compared with other aromatic groups, indicating that
aromatic hydrocarbon oxidation with an ortho-position substitute forms much
lower-volatility products than other isomers. Kom,1 is also
slightly higher in the meta and one-substitute groups than in the
three-substitute and para-substitute groups.
A slight SOA yield difference remains within each group (Fig. S2 and Table S3), indicating the influence of factors other than alkyl-group position.
Generally, lower yields are found in aromatics with alkyl groups of higher carbon number
substitute, such as when comparing propylbenzene (i- and n-)
with ethylbenzene or toluene (Li et al., 2016), m-ethyltoluene with
m-xylene and p-ethyltoluene with p-xylene, respectively. These differences are
explained by the proposed alkyl-group dilution effect (Sect. 4). However,
the differences between xylenes and their corresponding ethyltoluenes are
not statistically significant.
Two-product yield curve fitting parameters for one-, two- and three-alkyl substitutes (ortho, meta
and para).
Yield curve
α1
Kom,1 (m3 µg-1)
α2
Kom,2 (m3 µg-1)
MSRE∗
One substitutes
0.144
0.039
0.137
0.005
5.38
Two substitutes (ortho)
0.158
0.249
0.024
0.005
2.03
Two substitutes (meta)
0.156
0.040
0.080
0.005
2.51
Two substitutes (para)
0.154
0.025
0.036
0.005
1.21
Three substitutes
0.180
0.025
0.052
0.005
0.84
Note: ∗ mean squared error (MSRE) = [(fitted
yield - measured yield) / measured yield]2/ (number of data
points).
Chemical composition
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mn>44</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula> vs. <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>f</mml:mi><mml:mrow><mml:mn>43</mml:mn><mml:mo>+</mml:mo><mml:mn>57</mml:mn><mml:mo>+</mml:mo><mml:mn>71</mml:mn></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>
The ratio of alkyl substitute carbon number (H : C > 1) to the
aromatic ring carbon number impacts SOA composition since the H : C ratio on
the alkyl substitute is larger than 1 and the H : C ratio on the aromatic ring
itself is no more than 1. m/z 43 (C2H3O+ and
C3H7+) combined with m/z 44 (CO2+) is critical to
characterize oxygenated compounds in organic aerosol (Ng et al., 2010, 2011). C2H3O+ is the major contributor to m/z 43 in SOA
formed from aromatic hydrocarbons, having only methyl substitute (Li et al.,
2016), while C3H7+ fragments are observed in this work for
SOA from propylbenzene and isopropylbenzene (Fig. S5, Table S6). The
CnH2n-1O+ (n= carbon number of the alkyl substitute)
fragment in SOA corresponds to a CnH2n+1- alkyl substitute to the
aromatic ring. C3H5O+ (m/z 57) and C4H7O+ (m/z 71)
are important when investigating SOA from ethyl- or propyl-substitute
aromatic precursors. While m/z 57 (C4H9+) and m/z 71
(C5H11+) are often considered as markers for hydrocarbon-like
organic aerosol in ambient studies (Zhang et al., 2005; Ng et al., 2010),
oxygenated organic aerosol C3H5O+and C4H7O+
are the major fragments at m/z 57 and m/z 71, respectively (Fig. S5, Table S6) in current chamber SOA studies, especially during the photooxidation of
ethyl- and propyl-substituted aromatics. Therefore, m/z 57 and m/z 71 are
also considered beside C2H3O+ at m/z 43 in SOA chamber
studies as oxygenated organic aerosol to compare the oxidation of different aromatic hydrocarbons.
Fig. S5 lists all fragments found at m/z 43, 44, 57 and 71 and Fig. S6 shows
the fraction of each m/z in SOA formed from all aromatic hydrocarbons studied.
m/z43+m/z44+m/z57+m/z71 accounts for 21.2–29.5 % of
the total mass fragments from all C8 and C9 aromatics studied,
suggesting similar oxidation pathways. Only a small fraction (< ∼ 0.7 %) of m/z 71 (C4H7O+) or m/z 57
(C3H5O+) was observed in ethyltoluenes and trimethylbenzenes,
respectively.
This work extends the traditional f44 vs. f43
(C2H3O+) chemical composition analysis by including oxidized
fragments (C3H5O+ m/z 57 and C4H7O+ m/z 71) of
the longer (non-methyl) alkyl substitutes. Therefore, f44 vs. f43+f57+f71 is plotted instead of f44 vs. f43. Figure S4
shows the evolution of f44 and f43+57+71 in SOA formed from the photooxidation of different aromatic hydrocarbons
at low-NOx conditions. f44 and f43+57+71 ranges are
comparable to previous chamber studies (Ng et al., 2010; Chhabra et al.,
2011; Loza et al., 2012; Sato et al., 2012). Only slight f44 and
f43+57+71 evolution during chamber photooxidation is observed for the
C8 and C9 isomers; hence only the average f44 and
f43+57+71 will be analyzed in this work .
f44+ vs. f43+57+71′ in SOA formed from different aromatic
hydrocarbon photooxidation under low-NOx colored by aromatic isomer type
and marked with individual aromatic hydrocarbon species: ethylbenzene 2084A;
propylbenzene 1245A; isopropylbenzene 1247A; m-xylene 1191A;
m-ethyltoluene 1199A; o-xylene 1320A; o-ethyltoluene 1179A; p-xylene
1308A; p-ethyltoluene 1194A; 1,2,3-trimethylbenzene (123TMB)
1162A; 1,2,4-trimethylbenzene (124TMB) 1119A; 1,3,5-trimethylbenzene
(135TMB) 1156A. Alkyl number trend is the linear fitting in (Li et al.,
2015). ∗ Error bar stands for f44+ and f43+57+71′ standard
deviation when significant particles are formed
(> 5 µg m-3).
![]()
A modification is applied to the mass-based m/z fraction in order to compare
the mole relationship between m/z 44 and m/z43+m/z57+m/z71 (Eq. 1).
f′43+57+71=4443f43+4457f57+4471f71
The average f44 vs. f43+57+71′ for all C8 and C9 isomers
(Fig. 2) are located around the trend line for methyl-group-substituted
aromatic hydrocarbons (Li et al., 2016), implying a similarity in the SOA
components formed from alkyl-substituted aromatic hydrocarbons. A decreasing
trend in oxidation from upper left to lower right is included in Fig. 2,
similar to what Ng et al. (2011) found in the f44 vs. f43 graph,
especially while comparing similar structure compounds. The methyl-group
location on the aromatic ring impacts f44 : f43+57+71′.
Decreasing f44 and increasing f43+57+71′ trends are observed from
p-xylene to o-xylene to m-xylene and from 1,2,4-trimethylbenzene to
1,2,3-trimethylbenzene to 1,3,5-trimethylbenzene. f43+57+71′ may
partially depend on the relative position between the alkyl substitute and
the peroxide oxygen of the bicyclic peroxide. For instance, allylically
stabilized five-membered bicyclic radicals are the most stable bicyclic
radical formed from aromatic hydrocarbon photooxidation (Andino et al.,
1996). Two meta-position substitutes connected to the aromatic ring carbon
with -C-O- yield higher fractions of CnH2n-1O+ fragments than
the para and ortho position, which have at most one substitute connected with
-C-O- (Fig. S7). CO2+ is generally formed during MS electrical
ionization from carbonates, cyclic anhydrides and lactones (McLafferty and
Turecek, 1993), indicating that the CO2+ is associated with -O-C-O-
structure. Within the AMS, the CO2+ is also
associated with decarboxylation of organic acids during heating followed by
electrical ionization of the CO2. We hypothesize that CO2+
formation from bicyclic peroxides is insignificant since CO2 loss is not
expected come from -C-O-O- structure during thermal decomposition. Therefore,
it is the reaction products of bicyclic peroxides that lead to the formation
of CO2+ and the difference in f44. This indicates that the alkyl
groups are more likely to contribute to SOA formation at the meta position
than the ortho and para positions. Bicyclic peroxides formed from the
OH-addition reaction pathway and their dissociation reaction products are
both used to explain the substitute location impact on the f44 and
f43+57+71′ relationship. However, the existence of longer alkyl
substitutes diminishes the alkyl substitute location impact. SOA f44 and
f43+57+71′ in ethyltoluenes are all analogous to m-xylene.
One-substitute C8 and C9 aromatic hydrocarbons have similar
f44 and f43+57+71′ with slightly lower f44 and f43+57+71′
compared to toluene (Li et al., 2016). Longer alkyl substitutes may not lower
the average oxidation per mass as further oxidation of the longer-chain
alkyls may render other oxidized components not included in Fig. 2. Their
lower total f44+f43+57+71′ (Fig. S6) further supports the possibility
of oxidation of the longer alkyl substitutes. It is also possible that
oligomerization from highly oxidized carbonyls contributes more to the SOA
formation from aromatics with long-chain alkyl substitute. Elemental ratio
(Sect. 3.2.2) and oxidation state (Sect. 3.2.3) are further used to evaluate
the impact of increasing alkyl-group size on SOA formation.
H <inline-formula><mml:math display="inline"><mml:mo>/</mml:mo></mml:math></inline-formula> C vs. O <inline-formula><mml:math display="inline"><mml:mo>/</mml:mo></mml:math></inline-formula> C
Elemental analysis (Aiken et al., 2007, 2008) serves as a valuable tool to
elucidate SOA chemical composition and SOA formation mechanisms (Heald et al., 2010; Chhabra et al., 2011). Figure S8 shows H / C and O / C evolution in
SOA formed from the photooxidation of different aromatic hydrocarbons under
low NOx (marked and colored similarly to Fig. S4). H / C and O / C ranges
are comparable to previous chamber studies (Chhabra et al., 2011,
m-xylene and toluene; Loza et al., 2012, m-xylene; Sato et al., 2012,
benzene and 1,3,5-trimethylbenzene). The SOA elemental ratio for C8
and C9 aromatic isomers are located near the alkyl number trend line
found by Li et al. (2016) for methyl substituents, indicating a similarity
between SOA from various alkyl-substituted hydrocarbons. SOA formed is among
the low-volatility oxygenated organic aerosol (LV-OOA) and semi-volatile
oxygenated organic aerosol (SV-OOA) regions (Ng et al., 2011). The
evolution trend agrees with Fig. S4 (Sect. 3.2.1), which means no
significant H / C and O / C evolution is observed in the current study.
Therefore, average H / C and O / C with standard deviation provided is used to
explore the impact of molecular structure on SOA chemical composition. The
current study concentrates on experimentally averaged H / C and O / C to explore
the impact of molecular structure on SOA chemical composition.
H / C vs. O / C in SOA formed from different aromatic
hydrocarbon photooxidation under low NOx, colored by aromatic isomer type
and marked with individual aromatic hydrocarbon species (C8 and C9 on the
lower left indicate the location of initial aromatic hydrocarbon precursor):
ethylbenzene 2084A; propylbenzene 1245A; isopropylbenzene 1247A; m-xylene
1191A; m-ethyltoluene 1199A; o-xylene 1320A; o-ethyltoluene 1179A;
p-xylene 1308A; p-ethyltoluene 1194A; 1,2,3-trimethylbenzene (123TMB)
1162A; 1,2,4-trimethylbenzene(124TMB) 1119A; 1,3,5-trimethylbenzene(135TMB) 1156A. Alkyl number trend is the linear fitting in
Li et al. (2015a). Solid black circles are SOA elemental ratios from C8
and C9 aromatic hydrocarbon predicted by SOA elemental ratio formed from
benzene. ∗ Error bar stands for H / C and O / C standard
deviation when significant particles are formed
(> 5 µg m-3).
![]()
Average H / C and O / C locations are marked with aromatic compound names in
Fig. 3. All H / C and O / C are located around the predicted values for C8
and C9 SOA (dark solid circle) based on the elemental ratio of
benzene SOA (Li et al., 2016). This confirms the presence of a carbon
dilution effect in all isomers. Ortho-position aromatic hydrocarbons
(o-xylene or o-ethyltoluene) lead to a more oxidized SOA (higher O / C and lower
H / C) than that of meta (m-xylene or m-ethyltoluene) and para (p-xylene or
p-ethyltoluene) aromatics. SOA formed from 1,2,4-trimethylbenzene and 1,2,3-trimethylbenzene is more oxidized than that from 1,3,5-trimethylbenzene.
It is noticed that 1,2,4-trimethylbenzene and 1,2,3-trimethylbenzene
both contain an ortho-position moiety on the aromatic ring. This indicates
that the ortho-position aromatic hydrocarbon is readily oxidized and this
ortho-position impact on oxidation extends to triple-substituted aromatic
hydrocarbons. Substitute length also plays an important role in aromatic
hydrocarbon oxidation. Overall, SOA from a one-substitute aromatic with more
carbon in the substitute is located at a more oxidized area of the O / C vs.
H / C chart (lower right in Fig. 3) than those multiple-substitute aromatic
isomers with the same total number of carbon as the single-substituted
aromatic. SOA from isopropylbenzene is located in a lower position of the
chart and to the right of propylbenzene, indicating that branch carbon
structure on the alkyl substitute of aromatic hydrocarbons leads to a more
oxidized SOA. Lines in Fig. S8 connect the O / C and H / C of resulting SOA to
that of the aromatic precursor. Most SOA components show a slight H / C
increase and a dramatic O / C increase from the precursor, which is consistent
with results observed for methyl-substituted aromatics (Li et al., 2016).
However, H / C barely increases (1.33 to 1.34) from the propylbenzene
precursor to its resulting SOA and there is even a decreasing trend from
isopropylbenzene to its SOA. This indicates that a high H / C component loss
reaction such as alkyl-part dissociation during photooxidation is an
important reaction to SOA formation from
aromatic hydrocarbons containing longer carbon chains. The carbon chain length of propylbenzene increases
the possibility of alkyl fragmentation. The branching structure of
isopropylbenzene facilitates fragmentation through the stability of tertiary
alkyl radicals. Elemental ratio differences between xylenes and
ethyltoluenes can be attributed to the alkyl dilution effect, similar to the
methyl dilution theory by Li et al. (2016). Prediction of elemental ratios
from toluene and xylenes is discussed later (Sect. 4) to further quantify
the carbon length and branching effect on SOA formation from aromatic
hydrocarbons.
OS<inline-formula><mml:math display="inline"><mml:msub><mml:mi/><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:math></inline-formula>
Oxidation state (OSc≈ 2O / C-H / C) was
introduced into aerosol-phase component analysis by Kroll et al. (2011). It
is considered to be a more accurate metric for describing oxidation in
atmospheric organic aerosol than H / C and O / C (Ng et al., 2011;
Canagaratna et al., 2015; Lambe et al., 2015) and therefore well correlated
with gas–particle partitioning (Aumont et al., 2012;). Average
OSc of SOA formed from C8 and C9 aromatic isomers
ranges from -0.54 to -0.17 and -0.82 to -0.22, respectively (Fig. 4),
implying that the precursor molecular structure impacts the OSc
of the resulting SOA. An OSc decrease with alkyl substitute
length is observed in one-substitute aromatic hydrocarbons from toluene
(toluene OSc=-0.049; Li et al. 2016) to propylbenzene. However,
OSc provides the average oxidation value per carbon without
considering whether these carbons start from an aromatic ring carbon or an
alkyl carbon. Alkyl carbons are associated with more hydrogen than aromatic
ring carbons, thus leading to a lower precursor OSc and therefore
lower SOA OSc. Dilution conjecture in Sect. 4 will be used to
further explore the carbon chain length effect on aromatic hydrocarbon
oxidation by considering the precursor H : C ratio. Single-substitute
aromatic hydrocarbons generally show higher OSc than
multiple-substitute ones, consistent with the yield trend of Odum et
al. (1997b). However, it is also found that ortho-position moiety containing
two- or three-substitute aromatic hydrocarbons have analogous or even higher
OSc to single-substitute aromatic hydrocarbons (o-xylene
-0.03 ± 0.098 to ethylbenzene -0.173 ± 0.033;
1,2,4-trimethylbenzene -0.425 ± 0.072 and o-ethyltoluene
-0.481 ± 0.030 to propylbenzene -0.421 ± 0.111). This
suggests that both substitute number and position are critical to aromatic
hydrocarbon oxidation and therefore SOA formation. OSc trends
also support that the meta position suppresses oxidation while the ortho
position promotes oxidation when the OSc of xylenes (o-xylene
> p-xylene > (insignificant) m-xylene),
ethyltoluenes (o-ethyltoluene > p-ethyltoluene
> (insignificant) m-ethyltoluene) and especially
trimethylbenzenes (1,2,4-trimethylbenzene (ortho moiety containing)
> (insignificant) 1,2,3-trimethylbenzene (ortho moiety containing)
> 1,3,5-trimethylbenzene (meta moiety containing)) are compared
separately. Further, SOA formed from isopropylbenzene shows the highest
OSc among all C9 isomers, nearly equivalent to that of
ethylbenzene. This demonstrates that the branching structure of the alkyl
substitute can enhance further oxidation of aromatic hydrocarbons.
Oxidation state (OSc) of SOA formed from different
aromatic hydrocarbon photooxidation under low NOx: ethylbenzene 2084A;
propylbenzene 1245A; isopropylbenzene 1247A; m-xylene 1191A;
m-ethyltoluene 1199A; o-xylene 1320A; o-ethyltoluene 1179A; p-xylene
1308A; p-ethyltoluene 1194A; 1,2,3-trimethylbenzene (123TMB) 1162A; 1,2,4-trimethylbenzene(124TMB) 1119A; 1,3,5-trimethylbenzene(135TMB) 1156A.
![]()
Correlation among SOA density, volatility (VFR) and SOA chemical
composition.
f44
f57
f71
O / C
H / C
OSc
kOH
Density
0.324
-0.056
-0.38
0.551
-0.301
0.540
-0.249
p valueb
0.304
0.862
0.223
0.063
0.341
0.070
0.435
VFRenda
0.537
0.56
0.399
0.471
-0.586
0.593
-0.937
p valueb
0.089
0.073
0.224
0.144
0.058
0.055
0.000
Note: a VFRend is the volume fraction remaining
at the end of photooxidation. b p values range from 0 to 1:
0 indicates rejection of the null hypothesis and 1 is acceptance of the null hypothesis. Alpha (α)
level used is 0.05. If the p value of a test statistic is less than α,
the null hypothesis is rejected.
Measured and predicted SOA density from different aromatic
hydrocarbon photooxidation under low NOx (colored with substitute number
and length: one substitute is red, xylenes are green, ethyltoluenes are blue and
trimethylbenzene is purple; black line is predicted density according to Kuwata
et al., 2011); 123TMB is 1,2,3-Trimethylbenzene; 135TMB is 1,3,5-Trimethylbenzene; 124TMB is 1,2,4-Trimethylbenzene.
![]()
Physical property
SOA density
SOA density is a fundamental parameter in understanding aerosol morphology,
dynamics, phase and oxidation (DeCarlo et al., 2004; Katrib et al., 2005;
Dinar et al., 2006; Cross et al., 2007). SOA density ranges from 1.29 to
1.38 g cm-3 from aromatic photooxidation under low-NOx conditions
in this study (Fig. 5). The range is comparable to previous studies under
similar conditions (Borrás and Tortajada-Genaro, 2012; Ng et al., 2007;
Sato et al., 2010). There is no significant difference in the density of SOA
formed from C8 and C9 aromatic hydrocarbon isomers and molecular
structure is not observed to be a critical parameter to determine SOA
density. The standard deviation results from differences in initial
conditions (e.g., initial HC / NO) that also determine the oxidation of
aromatic hydrocarbons (Li et al., 2015) and thus further affect density. SOA
density is correlated with the O / C ratio and OSc (0.551 and
0.540, Table 3), consistent with the observation of Pang et al. (2006) that
SOA density increases with increasing O / C ratio. The density prediction
method developed by Kuwata et al. (2011) based on O / C and H / C is
evaluated as
ρ=12+H/C+16×O/C7+5×H/C+4.15×O/C.
The black lines (Fig. 5) are predicted (Eq. 2) densities and show a good
agreement between predicted and measured SOA densities (difference between
prediction and measurement is -6.01–7.62 %). A comparatively large
negative error is found in meta-containing aromatic hydrocarbons, including
m-xylene, m-ethyltoluene and 1,3,5-trimethylbenzene. It is noted
that there should be more alkyl substitutes in SOA formed from meta-position
aromatics than other aromatics since meta-position alkyl substitutes are more
likely to participate into SOA products than other aromatics (Sect. 3.2.1 and
3.2.2). Previous work suggests that the increase of methyl groups could lead
to a change in several key organic fragments (e.g., CO+, CO2+
and H2O+), thereby altering the default fragment table for
elemental ratio analysis. This agrees with the density underestimation in SOA
formed from meta-position aromatics and supports the preference of
meta-position alkyl substitute to SOA products.
SOA volatility
SOA volatility is associated with reactions such as oxidation,
fragmentation, oligomerization and mass loading (Kalberer et al., 2004;
Salo et al., 2011; Tritscher et al., 2011; Yu et al., 2014). SOA
volatility in this study is measured as VFR. Initial (< 30 min
after new particle formation) SOA VFRs are around 0.2 for all the aromatic
precursors studied and increase up to 0.58 during photooxidation. This
suggests that aromatic hydrocarbon oxidation undergoes an evolution from
volatile compounds to semivolatile compounds. The VFR trends and ranges are
comparable to previous studies (Kalberer et al., 2004; Qi et al., 2010a, b; Nakao et al., 2012). Figure 6 shows the VFR at the end of
aromatic hydrocarbon photooxidation (VFRend). A decreasing VFRend
trend is found as the number of substitutes increase and for meta-position
(e.g., m-xylene) or meta-position-containing (e.g., 1,3,5-trimethylbenzene)
aromatic precursors. Correlations among VFRend and chemical
composition are observed in the aromatic hydrocarbons studied here (Table 3). This is consistent with recent findings that O : C ratio is correlated to
aerosol volatility (Sect. 3.3.2) (Cappa and Wilson, 2012; Yu et al., 2014),
thereby affecting the gas–particle partitioning, which in turn relates to
SOA yield. It is also observed that VFRend is strongly correlated
(-0.937) with reaction rate constant (kOH). Higher kOH is associated
with faster reaction rates of initial aromatic precursors and is therefore
expected to lead to further oxidation for a given reaction time. However,
the inverse correlation between kOH and VFRend indicates that
kOH value represents more than just the kinetic aspects. kOH increases with increasing number of substitutes on the aromatic ring.
Additionally, aromatic hydrocarbons with meta-position substitutes have
higher kOH than those with para- and ortho-position (Table S1)
substitutes. This suggests that the precursor molecular structures for
aromatics associated with kOH values determine the extent of oxidation
of the hydrocarbons and therefore impact SOA volatility more than simply the
precursor oxidation rate.
SOA volume fraction remaining (VFRend) at the end of
aromatic hydrocarbon photooxidation under low NOx (colored with
substitute number and length: one substitute is red, xylenes are green,
ethyltoluenes are blue and trimethylbenzene is purple); 123TMB is 1,2,3-Trimethylbenzene; 135TMB is 1,3,5-Trimethylbenzene; 124TMB is 1,2,4-Trimethylbenzene.
![]()
Atmospheric implication
This study elucidates molecular structure impact on a major anthropogenic SOA
source, photooxidation of aromatic hydrocarbons, under atmospherically
relevant NOx conditions by analyzing SOA yield, chemical composition and
physical properties. These observations, when taken together, indicate the
roles of alkyl substitute number, location, carbon chain length and branching
structure in aromatic hydrocarbon photooxidation. SOA yield of all C8
and C9 aromatic hydrocarbon isomers is comprehensively provided in this
study with a focus on the impact of molecular structure. It is demonstrated
that aromatic hydrocarbon oxidation and SOA formation should not be simply
explained by substitute number. The promoting of SOA formation by the ortho
position is found along with confirmation of the suppression effect by the
para position during oxidation of aromatic hydrocarbons. It is possibly due
to the alkyl substitute location impact on the further oxidation of
five-membered bicyclic radicals. Different carbonyl compounds can form as the
ring opening products from the dissociation of five-membered bicyclic
radical. It is assumed that oligomerization of these carbonyl compounds can
contribute to SOA (Li et al., 2016). Aromatic hydrocarbons with para-position
alkyl substitute tend to form more ketone-like dicarbonyl compounds than
other aromatics. Ketone might contribute less to oligomerization formation
compared with aldehyde as suggested in Li et al. (2016). Meta-position alkyl
substitutes on aromatic ring lead to a lower extent of aromatic hydrocarbon
oxidation. This might be due to a higher percentage of carbonyl with alkyl
substitute formed during the oxidation of meta-containing aromatics (e.g.,
methylglyoxal, 2-methyl-4-oxopent-2-enal), which contributes to
oligomerization and thereby SOA formation. Evidence is provided to
demonstrate aromatic oxidation increase with alkyl substitute chain length
and branching structure. Further, the carbon dilution theory developed by Li.
et al. (2016) is extended to this study. Carbon dilution theory not only
serves as a tool to explain the difference in SOA components due to the
difference in substitute alkyl carbon number but also acts as a standard to
determine the oxidation mechanism based on alkyl substitute structure.
Moreover, the five subcategories of aromatics and their two-product modeling
curve fitting parameters in this work at more realistic NOx loadings
provide a more precise prediction of SOA formation form aromatic hydrocarbons
under atmospheric conditions. Previous studies found that the humidity
insignificantly impacts SOA yield from aromatic hydrocarbons (Cocker et al.,
2001) or maintains the SOA yield relationship between isomers (Zhou et al.,
2011). Therefore, it is predicted that the observation found under dry
conditions in this study, especially the molecular structure impact on SOA
formation from different aromatic isomers, could be extended to
atmospherically relevant humidity conditions. However, recent studies observe
that the hydration of carbonyls and epoxides could lead to further
heterogeneous reaction and oligomerization (Jang et al., 2002; Liggio et al.,
2005; Minerath and Elrod, 2009; Lal et al., 2012). It is possible that
aerosol compositions and the hygroscopic properties could be altered after
the heterogeneous reactions, especially under humid conditions. The impact of
molecular structure impact on SOA formation under humidity condition needs to
be further studied to extend the findings in current the work. This study
improves the understanding of SOA formation from aromatic hydrocarbons and
contributes to more accurate SOA prediction from aromatic precursors. Further
study is warranted to reveal the detailed oxidation pathway of aromatic
hydrocarbons with longer (carbon number > 1) alkyl substitutes.