Role of methyl group number on SOA formation from monocyclic aromatic hydrocarbons photooxidation under low-NO x conditions

. Substitution of methyl groups onto the aromatic ring determines the secondary organic aerosol (SOA) formation from the monocyclic aromatic hydrocarbon precursor (SOA yield and chemical composition). This study links the number of methyl groups on the aromatic ring to SOA formation from monocyclic aromatic hydrocarbons photooxidation under low-NO x conditions (HC/NO > 10 ppbC : ppb). Monocyclic aromatic hydrocarbons with increasing numbers of methyl groups are systematically studied. SOA formation from pentamethylbenzene and hexamethylbenzene are reported for the ﬁrst time. A decreasing SOA yield with increasing number of methyl groups is observed. Linear trends are found in both f 44 vs. f 43 and O / C vs. H / C for SOA from monocyclic aromatic hydrocarbons with zero to six methyl groups. An SOA oxidation state predictive method based on benzene is used to examine the effect of added methyl groups on aromatic oxidation under low-NO x conditions. Further, the impact of methyl group number on density and volatility of SOA from monocyclic aromatic hydrocarbons is explored. Finally, a mechanism for methyl group impact on SOA formation is suggested. Overall, this work suggests that, as more methyl groups are attached on the aromatic ring, SOA products from these monocyclic aromatic hydrocarbons become less oxidized per mass/carbon on the basis of SOA yield or chemical composition.


Introduction
Aromatic hydrocarbons are major anthropogenic secondary organic aerosol (SOA) precursors (Kanakidou et al., 2005;Henze et al., 2008). Monocyclic aromatic hydrocarbons with fewer than four methyl groups are ubiquitous in the atmosphere (Singh et al., 1985(Singh et al., , 1992Fraser et al., 1998;Pilling and Bartle, 1999;Holzinger et al., 2001;Buczynska et al., 2009;Hu et al., 2015). Monocyclic aromatic hydrocarbons with more than four methyl groups have scarcely been investigated in previous ambient studies, possibly due to a vapor pressure decrease with carbon number (Pankow and Asher, 2008; Table S1 in the Supplement). However, a recent study observed that compounds with low vapor pressure are available to evaporate into the atmosphere (Vȏ and Morris, 2014). Monocyclic aromatic hydrocarbons with more than three methyl groups contribute to a large portion of components in products such as gasoline and crude oil (Diehl and Sanzo, 2005;Darouich et al., 2006). Moreover, hydrocarbon reactivity and OH reaction rate constant (k OH ) increase with methyl group number (k OH Table S1; Glasson and Tuesday, 1970;Calvert et al., 2002;Atkinson and Arey, 2003;Aschmann et al., 2013). OH-initiated reactions, particularly OH addition to the aromatic ring, dominate aromatic photooxidation (Calvert et al., 2002). Hence, photooxidation occurs rapidly once these low-vapor-pressure aromatic hydrocarbons evaporate into atmosphere. In addition, an increase in carbon number is associated with a decrease in vapor pressure (Pankow and Asher, 2008). Higher-carbonnumber products with a similar number of functional groups have a higher tendency to participate in the particle phase. However, aging of organic aerosol is a combination of functionalization, fragmentation and oligomerization Kroll et al., 2009). Therefore, rapid aging does not necessarily lead to the highly oxidized compounds, which serve as an important source of SOA.
Recent studies have found that SOA yields from OH initiated alkane and alkene reactions increase with carbon chain length and decrease with the increase in branched structure Matsunaga et al., 2009;Tkacik et al., 2012). However, SOA yield from monocyclic aromatics is found to decrease with carbon number by adding methyl groups to the aromatic ring (Odum et al., 1997a;Cocker III et al., 2001b;Sato et al., 2012). This indicates that the role of methyl groups on the aromatic ring is different than for alkane and alkene hydrocarbons. Previous studies show that the relative methyl group position determines the alkoxyl radical (RO q ) fragmentation ratio in alkane and alkene hydrocarbon oxidation (Atkinson, 2007;Ziemann, 2011). Therefore, it is necessary to explore the impact of methyl groups on SOA formation during monocyclic aromatic hydrocarbon oxidation.
Previous studies on SOA formation from monocyclic aromatic hydrocarbon in the presence of NO x have been conducted at high NO x levels (e.g., Odum et al., 1997b;Cocker III et al., 2001b;Sato et al., 2012). Ng et al. (2007) observed that SOA yield decreases with increasing carbon number under high-NO x conditions and no trends were observed for conditions of no NO x . Reaction mechanisms vary for different NO x conditions (e.g., Song et al., 2005;Kroll and Seinfeld, 2008) and thus impact SOA chemical composition. Therefore, it is necessary to investigate methyl group impact on urban SOA formation from monocyclic aromatic hydrocarbon under more atmospherically relevant low-NO x conditions. SOA budget underestimation of the urban environment is associated with mechanism uncertainty in aromatic hydrocarbon photooxidation and possibly missing aromatic hydrocarbon precursors (Henze et al., 2008;Hallquist et al., 2009). Previous chamber studies have seldom investigated SOA formation from monocyclic aromatic with more than three methyl groups (e.g., pentamethylbenzene and hexamethylbenzene). This study investigates SOA formation from the photooxidation of seven monocyclic aromatic hydrocarbon (ranging from benzene to hexamethylbenzene) under the low-NO x (HC/NO > 10 ppbC : ppb) condition. The impacts of methyl group number on SOA yield, chemical composition and other physical properties are demonstrated. Possible methyl group impacts on aromatic ring oxidation, decomposition and subsequent oligomerization are discussed.

Environmental chamber
All experiments were conducted in the UC Riverside/CE-CERT indoor dual 90 m 3 environmental chambers, which are described in detail elsewhere (Carter et al., 2005). All experiments were conducted under dry conditions (RH < 0.1 %), in the absence of inorganic seed aerosol and with temperature control to 27 ± 1 • C. Two movable top frames were slowly lowered during each experiment to maintain a slight positive differential pressure (0.02 inches of water) between the reactors and enclosure to minimize dilution and/or contamination of the reactors. Two hundred and seventy-two 115 W Sylvania 350BL blacklights were used as light sources for photooxidation.
A known volume of high-purity liquid hydrocarbon precursors (benzene Sigma-Aldrich, 99 %; toluene Sigma-Aldrich, 99.5 %; m-xylene Sigma-Aldrich, 99 %; 1,2,4-trimethylbenzene Sigma-Aldrich, 98 %) was injected through a heated glass injection manifold system and flushed into the chamber with pure N 2 . A glass manifold packed with glass wool inside a temperature-controlled oven (50-80 • C) was used to inject solid hydrocarbon precursors (1,2,4,5-tetramethylbenzene Sigma-Aldrich, 98 %; pentamethylbenzene Sigma-Aldrich, 98 %; hexamethylbenzene Sigma-Aldrich, 99 %). NO was introduced by flushing pure N 2 through a calibrated glass bulb filled to a predetermined partial pressure of pure NO. All hydrocarbons and NO were injected and well mixed before the lights were turned on to commence the reaction.

Particle and gas measurement
Particle size distribution between 27 and 686 nm was monitored by dual custom-built scanning mobility particle sizers (SMPS) (Cocker III et al., 2001a). Particle effective density was measured with a Kanomax aerosol particle mass analyzer (APM-SMPS) system (Malloy et al., 2009). Particle volatility was measured by a volatility tandem differential mobility analyzer (VTDMA) (Rader and Mc-Murry, 1986) with a Dekati ® thermodenuder controlled to 100 • C and a 17 s heating zone residence time (Qi et al., 2010b). Volume fraction remaining (VFR) is calculated as (D p,after TD /D p,before TD ) 3 .
Evolution of particle-phase chemical composition was measured by a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS; Aerodyne Research Inc.) DeCarlo et al., 2006). The sample was vaporized by a 600 • C oven followed by 70 eV electron impact ionization. f x in this study is calculated as the fraction of the organic signal at m/z = x. For example, f 44 and f 43 are the ratios of the organic signal at m/z 44 and 43 to the total organic signal, respectively Duplissy et al., 2011). Elemental ratios for total organic mass, oxygen Marked by the number of methyl groups in aromatic hydrocarbon to carbon (O / C), and hydrogen to carbon (H / C) were determined using the elemental analysis (EA) technique (Aiken et al., 2007(Aiken et al., , 2008. Data were analyzed with the ToF-AMS analysis toolkit Pika 1.15D with 1.56D Squirrel. An Agilent 6890 gas chromatograph-flame ionization detector was used to measure aromatic hydrocarbon concentrations. A Thermo Environmental Instruments model 42C chemiluminescence NO analyzer was used to monitor NO, NO y -NO and NO y . The gas-phase reaction model SAPRC-11 developed by Carter and Heo (2012) was utilized to predict radical concentrations ( q OH, HO 2 q , RO 2 q and NO 3 q ).

SOA yield relationship with methyl group number
SOA yields from the photooxidation of seven monocyclic aromatic hydrocarbons are calculated as the mass-based ratio of aerosol formed to hydrocarbon reacted (Odum et al., 1996). The HC / NO ratio ranged from 12.6 to 110 ppbC : ppb for all experiments used in this study. Experiment conditions and SOA yield are listed from the current work (Table 1) along with additional m-xylene experiment conditions from previous studies (Table S2)  in the UCR CE-CERT chambers. SOA yield as a function of particle mass concentration (M 0 ) for all seven monocyclic aromatic precursors (Fig. 1) includes experiments listed in both Table 1 and S2. Each individual experiment is marked and colored by the number of methyl groups on each precursor aromatic ring. It is observed that SOA yield decreases as the number of methyl groups increases (Fig. 1). A similar yield trend is also observed in previous studies on SOA formation from monocyclic aromatic hydrocarbons; however, different absolute yield values are found, presumably due to higher NO x levels (Odum et al., 1997b;Kleindienst et al., 1999;Cocker III et al., 2001b;Takekawa et al., 2003;Ng et al., 2007;Sato et al., 2012). SOA yields of benzene under comparable low-NO x conditions are higher than that in Sato et al. (2012), Borrás and Tortajada-Genaro (2012) and Martín-Reviejo and Wirtz (2005).
The two-product semi-empirical model described by Odum et al. (1996) is used to fit SOA yield as a function of M 0 . Briefly, the two-product model assumes that aerosolforming products can be lumped into lower-and highervolatility groups whose mass fraction is defined by α i and a partitioning parameter K om,i (m 3 µg −1 ), described extensively in Odum et al. (1996). Each monocyclic aromatic hydrocarbon is fitted individually except for those with methyl group number greater than or equal to 4, which are grouped as C 10+ . The experimental fitting parameters (α 1 , K om,1 , α 2 and K om,2 in Table 2) in the two-product model were determined by minimizing the sum of the squared of the residuals. The higher-volatility partitioning parameter (K om,2 ) in all yield curve fitting are assigned to a fixed value by assuming similar high-volatility compounds are formed during all monocyclic aromatic hydrocarbon photooxidation experiments. Benzene has much higher mass-based stoichiometric coefficients (α 2 ) than the other monocyclic aromatic compounds, indicating that the pathway leading to highervolatility products' formation is favored. The lower-volatility partitioning parameters (K om,1 ) vary widely for each monocyclic aromatic yield fitting curve. Benzene has the lowest K om,1 , toluene has the highest K om,1 , and the rest of monocyclic aromatics have similar mid-range K om,1 values. The extremely low K om,1 of benzene indicates that pathways associated with significant volatility decrease occur far less during benzene photooxidation than for monocyclic aromatic compounds with methyl groups. Further, K om,1 is much lower in multi-methyl group monocyclic aromatic hydrocarbons (with the exception of toluene), while α 1 decreases with methyl group number. This suggests that increasing methyl group number on the aromatic ring suppresses formation of lower-volatility products, therefore lowering the mass based aerosol yield. This suggests that monocyclic aromatics with more methyl groups are less oxidized per mass since the methyl group carbon is not well oxidized compared with the ring carbon. The aromatic SOA growth curves (particle concentration M 0 vs. hydrocarbon consumption HC) under similar HC / NO x are shown in Fig. S1 in the Supplement. The slope of the growth curve is negatively correlated with the parent aromatics reaction rate (k OH ). This observation contrasts with a previous study that observed positive correlation between SOA formation rate and hydrocarbon reaction rate for systems where initial semivolatile products dominate gasparticle phase partitioning . The reverse relationship observation in this study indicates that the effect of methyl group number on SOA yield is greater than that of the increasing k OH on SOA yield. There are two possibilities for the methyl group number effect: (1) the methyl group facilitates initial semivolatile products to react into more volatile compounds or (2) the methyl group prevents further generation semivolatile products formation by stereohindrance. Therefore, the methyl group increases hydrocarbon mass consumption more than particle mass formation.
The relationship between radical levels and SOA yield was also analyzed. Table S3 lists modeled individual average radical concentrations throughout photooxidation while Table S4 lists the correlation between SOA yields and individual average radical concentrations. None of the radical parameters (e.g., q OH/HO 2 q , HO 2 q /RO 2 q ) is strongly correlated with SOA yield. Average OH radical concentration is the only parameter investigated with a statistically significant correlation (p < 0.05), as k OH varies with aromatic species and lower average OH concentrations are present with higher k OH . Figure S2 shows the time evolution of [ q OH], [RO 2 q ] and [HO 2 q ] for different aromatic precursors under similar initial aromatic and NO x loadings. Higher [ q OH] is observed for aromatic precursors with lower k OH , while peroxide radicals ([RO 2 q ] and [HO 2 q ]), which depend on both k OH and [ q OH], are similar for all precursors. This suggests that SOA mass yield is determined by precursor structure rather than gas-phase oxidation state since radical conditions for each aromatic hydrocarbon are comparable and [RO 2 q ] and [HO 2 q ] reactions are expected to determine SOA formation .

f 44 vs. f 43
Organic peaks at m/z 43 and m/z 44 are key fragments from AMS measurement toward characterization of oxygenated compounds in organic aerosol (Ng et al., 2010. A higher f 44 and a lower f 43 indicates a higher degree of oxidation (Ng et al., 2010). The f 44 and f 43 evolution during SOA formation from different monocyclic aromatic hydrocarbon photooxidation is shown for low-NO x conditions (Fig. 2). Each marker type represents an individual monocyclic aromatic hydrocarbon with the marker colored by photooxidation time (light to dark). The f 44 and f 43 range are comparable to previous chamber studies with slight shift due to differences in initial conditions (e.g., NO x ) (Ng et al., 2010;Chhabra et al., 2011;Loza et al., 2012;Sato et al., 2012). SOA compositions from monocyclic aromatic hydrocarbon photooxidation under low NO x are in the low-volatility OOA (LV-OOA) and semi-volatile OOA (SV-OOA) range of the f 44 vs. f 43 triangle (Ng et al., 2010), with those from benzene on the left side, toluene inside and other monocyclic aromatics on the right side of the triangle, confirming that laboratory SOA f 44 vs. f 43 is precursor dependent . Evolution of SOA composition (Heald et al., 2010;Jimenez et al., 2009)  ous studies (Ng et al., 2010;Chhabra et al., 2011;Sato et al., 2012). The f 44 vs. f 43 trend is quantified by linear curve fitting (f 44 = −0.58 f 43 + 0.19, R 2 = 0.94). f 28 is assumed to be equal to f 44 in the AMS frag Table of Unit Resolution Analysis, which describes the mathematical formulation of the apportionment at each unit resolution sticks to aerosol species, based on ambient studies (Zhang et al., 2005;Takegawa et al., 2007) and CO + / CO + 2 ratio for SOA from aromatic oxidation is found around ∼ 1 (0.9-1.3) . The slope of ∼ 0.5 indicates that 2 f 44 = − f 43 or (f 28 + f 44 ) = − f 43 in SOA formed from monocyclic aromatic hydrocarbons with different numbers of methyl groups. The CO + 2 fragment ion at m/z 44 and C 2 H 3 O + fragment ion at m/z 43 are two major AMS fragmentation ions from aromatic secondary organic aerosol. No significant C 3 H + 7 is observed at m/z 43. CO + 2 represents oxidized aerosol and is associated with carboxylic acids (Alfarra et al., 2004;Aiken et al., 2007;Takegawa et al., 2007;Canagaratna et al., 2015), while C 2 H 3 O + is associated with carbonyls (McLafferty and Tureček, 1993;Ng et al., 2011). The CO + fragment ion at m/z 28 can originate from carboxylic acid or alcohol (Canagaratna et al., 2015). The (f 28 + f 44 ) = − f 43 relationship observed in this study implies that adding the methyl group to the aromatic ring changes SOA from CO + 2 to C 2 H 3 O + , implying a less oxidized SOA chemical composition in AMS mass fragments. While bicyclic hydrogen peroxides are considered to be the predominant species in aerosol phase from monocyclic aromatic photooxidation (Johnson et al., 2004(Johnson et al., , 2005Wyche et al., 2009;Birdsall et al., 2010;Birdsall and Elrod, 2011;Nakao et al., 2011), they are less likely to contribute to the CO + 2 ion fragment. Possible mechanisms to produce SOA products that form the CO + 2 fragments as well as produce C 2 H 3 O + fragments by adding methyl group are described in detail in Sect. 4.

H / C vs. O / C
Elemental analysis (Aiken et al., 2007(Aiken et al., , 2008) is used to elucidate SOA chemical composition and SOA formation mechanisms (Heald et al., 2010;Chhabra et al., 2011). Figure 3a shows the H / C and O / C time evolution of average SOA formed from hydrocarbon photooxidation of various monocyclic aromatics under low-NO x conditions (marked and colored similarly to Fig. 2). The H / C and O / C ranges are comparable to previous chamber studies with slight shift due to difference in initial conditions (e.g., NO x ) Loza et al., 2012;Sato et al., 2012). All data points are located in between slope = −1 and slope = −2 (Fig. 3a, lower left corner, zoomed-out panel). This suggests that SOA components from monocyclic aromatic photooxidation contain both carbonyl (ketone or aldehyde) and acid (carbonyl acid and hydroxycarbonyl) functional groups. These elemental ratios also confirm that SOA formed from monocyclic aromatic hydrocarbon photooxidation under low NO x are among the LV-OOA and SV-OOA regions . The change in elemental ratio (H / C and O / C) with time is referred to as elemental ratio evolution. The elemental ratio evolution agrees with the f 44 vs. f 43 evolution (significant evolution in benzene and slightly for toluene, m-xylene and 1,2,4,5-tetramethylbenzene). This study concentrates on average H / C and O / C in order to demonstrate the methyl group impact on SOA chemical composition from monocyclic aromatic hydrocarbons.
Average H / C and O / C location is marked (Fig. 3a) for each aromatic compound by name. It is observed that H / C and O / C from SOA formed from m-xylene, 1,2,4-trimethylbenzene and monocyclic aromatics with more than three methyl groups are similarly distributed in H / C vs. O / C. A general decrease in O / C and an increase in H / C are noted as the number of methyl groups on the aromatic ring increases, which is consistent with other studies Sato et al., 2012). The trend indicates that monocyclic aromatics are less oxidized per carbon as the number of methyl groups increases, which can be attributed to less oxidation of the methyl groups compared to the aromatic ring carbons. The elemental ratio trends (O / C decreases and H / C increases as the number of methyl groups increases) are also consistent with the decreasing yield trends with increasing the number of methyl groups (Sect. 3.1), suggesting that SOA yield is dependent on SOA chemical composition. It should also be noticed that the higher O / C and lower H / C observed in SOA formed from 1,2,4-trimethylbenzene (three-methylsubstitute aromatic hydrocarbon) than that from m-xylene (two-methyl-substitute aromatic hydrocarbon) is due to the isomer impact on SOA chemical composition, which is discussed in detail by Li et al. (2016). Further, the yield and O / C ratio agrees with recent findings that O / C ratio is well correlated to aerosol volatility (Sect. 3.3.2) (Cappa and Wil-son, 2012;Yu et al., 2014), thereby affecting the extent of gas to particle partitioning. The H / C vs. O / C trend linear curve (H / C = −1.34 O / C + 2.00, R 2 = 0.95) shows an approximately −1 slope with a y-axis (H / C) intercept of 2. The H / C vs. O / C trend slope observed in this work is similar to the toluene and m-xylene elemental ratio slope observed under high-NO x and H 2 O 2 -only conditions observed in Chhabra et al. (2011).
The Van Krevelen diagram can also be used to analyze the oxidation pathway from initial SOA precursor to final SOA chemical composition by comparing the initial H / C and O / C ratios from the precursor hydrocarbon to the final SOA H / C and O / C ratios. Figure 3b shows the aromatic precursor location on the left (labeled with aromatic hydrocarbon name and colored by methyl group number) and average SOA chemical composition on the right. The SOA H / C increase in the final SOA chemical composition follows the initial aromatic precursor elemental ratio trend. A large O / C increase with a slight H / C increase is observed moving from precursor to SOA composition. SOA formation from hydroperoxide bicyclic compounds contributes to O / C increases without loss of H. The slight H / C increase might result from hydrolysis of ring-opened product oligomerization (Jang and Kamens, 2001;Jang et al., 2002;Kalberer et al., 2004;Sato et al., 2012). A slight H / C decrease rather than increase is observed in the hexamethylbenzene data, suggesting that the six methyl groups sterically inhibit certain reaction mechanism (e.g., hydrolysis) to obtain H.

OS c and its prediction
O / C alone may not capture oxidative changes as a result of breaking and forming of bonds ). Oxidation state of carbon (OS c ≈ 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 Canagaratna et al., 2015;Lambe et al., 2015) and therefore better correlated with gas-particle partitioning (Aumont et al., 2012). Average SOA OS c in this study ranges from −0.9 to 0.3 for monocyclic aromatic photooxidation under low-NO x conditions (Fig. 4b) and is comparable to previous studies (Kroll et al., 2011 -toluene, m-xylene and trimethylbenzene; Sato et al., 2012 -benzene and 1,3,5trimethylbenzene). OS c observed is consistent with OS c observed in field studies (Kroll et al., 2011), especially in urban sites (e.g., −1.6-0.1, Mexico City), and supports the major role of monocyclic aromatic precursors in producing anthropogenic aerosol. Average SOA OS c values are consistent with the LV-OOA and SV-OOA regions Kroll et al., 2011). OS c only increases with oxidation time for benzene photooxidation (0.2-0.4).
The methyl group substitute (−CH 3 ) affects O / C and H / C ratios by increasing both carbon and hydrogen number as they relate to SOA OS c . It is hypothesized here that the methyl group impacts remain similar in SOA elemental ratios as they do in the aromatic precursor (−CH 3 dilution effect). This would imply that the methyl group effect on SOA elemental ratio and OS c from monocyclic aromatic hydrocarbons is predictable from benzene oxidation. Equations (1) and (2) Figure 4b compares measured (red) and predicted (green) OS c . The largest O / C and OS c overestimation is observed in m-xylene (marked as 2 in Fig. 4a, second bar in Fig. 4b). This could be explained by the isomer selected for the two-methyl-group monocyclic aromatic hydrocarbon (m-xylene). A detailed analysis on isomer structure impact on SOA chemical composition is found in Li et al. (2016). The largest O / C and OS c underestimation is observed in hexamethylbenzene (marked as 6 in Fig. 4a, sixth bar in Fig. 4b). This suggests that the methyl groups attached to every aromatic carbon exert a steric inhibition effect on certain aromatic oxidation pathways, thus leading to increased importance of aerosol formation from other reaction pathways (possibly fragmentation; Kroll et al., 2011;see Sect. 4) to form SOA. It is also noticed that O / C and OS c is slightly overestimated in SOA formed from pentamethylbenzene. This indicates that the methyl group hindrance impact on aromatic hydrocarbon oxidation should be explained by multiple pathways which have a different impact on SOA formation.
The correlation between organic mass loading and chemical composition is also analyzed. Organic mass loading is well correlated (Pearson correlation) with chemical composition parameters, including f 44 (0.907), f 43 (−0.910), H / C (−0.890) and O / C (0.923) (Fig. S3). However, previous studies show that O / C and f 44 decrease as organic mass loading increases (Shilling et al., 2009;Ng et al., 2010;Pfaffenberger et al., 2013). The findings of this study indicate that molecular species drive SOA chemical composition rather than organic mass. The positive trend between f 44 and organic mass loading is driven by benzene and toluene experiments (Fig. S3) where the high mass loading results are concurrent with high f 44 results. However, the f 44 change with mass loading increase during benzene and toluene photooxidation is less significant compared with the f 44 difference caused by number of methyl groups on aromatic ring. Moreover, no significant correlation was found between mass loading and f 44 or O / C when compared under similar mass loadings (including f 44 at low mass loading time point of toluene and benzene photooxidation). Organic nitrate accounts for less than 10 % organic in SOA components in all monocyclic aromatic hydrocarbon photooxidation experiments in this work according to AMS measurement and will not be discussed.

SOA density
SOA mass density is a fundamental parameter in understanding aerosol morphology, dynamics, phase and oxidation (De Carlo et al., 2004;Katrib et al., 2005;Dinar et al., 2006;Cross et al., 2007). SOA density ranged from 1.24 to 1.44 g cm −3 for all aromatic-NO x photooxidation experiments in this study. The range is comparable to previous studies under similar conditions Sato et al., 2010;Borrás and Tortajada-Genaro, 2012). A general decreasing density trend is found with increasing methyl group number on precursor aromatic rings (see Fig. 5a). Correlation between SOA density and chemical composition was statistically analyzed (Table S5). Besides the strong correlation with methyl group number (−0.943, Fig. 5a), SOA density was also well correlated with O / C ratio (0.873, Fig. 5b) and other measures of bulk chemical composition (Table S5) Kuwata et al. (2011) is used in this work to predict density based on elemental ratio in order to explore the methyl group impact on SOA formation. Figure 5c shows a good agreement between predicted and measured SOA densities (−6.58-10.42 %). However, SOA density difference between prediction and measurement change from positive (aromatic precursors contain 0 or 1 methyl group) to negative (aromatic precursors contain 2, 3, 4 or 5 methyl groups) with increasing methyl group number (except hexamethylbenzene), implying that the increase in methyl groups promotes a mechanism(s) leading to changes in the ratio of several key organic fragments (e.g., m/z 28 : m/z 44), thereby challenging the applicability of the default fragment table for elemental ratio analysis. It is possible that CO + /CO + 2 and H 2 O + /CO + 2 ratios are different in SOA formed from different aromatic precursors. Nakao et al. (2013) show that H 2 O + /CO + 2 increases with methyl group number due to the constant H 2 O + fraction and a decrease in CO + 2 fraction. Canagaratna et al. (2015) demonstrated that CO + /CO + 2 and H 2 O + /CO + 2 are underestimated in certain compounds (especially alcohols). Assuming that the major impact of methyl group on SOA composition is to change −COOH to −COCH 3 (or other cyclic isomers), f CO + 2 will decrease but H 2 O + and CO + fraction might not change linearly. The alcohol contribution to CO + /CO + 2 and H 2 O + /CO + 2 gradually grows as the methyl group prevents acid formation. Therefore, AMS measurements might underestimate O / C. This is consistent with the density prediction from elemental ratios where a change in error from positive to negative is seen as the number of methyl groups changes from fewer than two to two or more than two, with the exception of hexamethylbenzene. This might relate to the difference in SOA formation pathways due to steric hindrance of the six methyl groups during hexamethylbenzene oxidation.

SOA volatility
SOA volatility is a function of oxidation, fragmentation, oligomerization and SOA mass (Kalberer et al., 2004;Salo et al., 2011;Tritscher et al., 2011;Yu et al., 2014). Bulk SOA volatility can be described by the VFR after heating SOA to a fixed temperature in a thermodenuder. VFRs for SOA formed early in the experiment are around 0.2 for all monocyclic aromatic precursors and then increase as the experiment progresses. Increasing VFR indicates the gas to particle partitioning of more oxidized products, which may include oligomerization products formed during aromatic photooxidation. The VFR trends and ranges are comparable to previous studies (Kalberer et al., 2004;Qi et al., 2010a, b;Nakao et al., 2012). Figure 6a shows the relationship between SOA precursor methyl group number and SOA VFR at the end of the experiment (VFR end ). VFR shows a significant decreasing trend with increasing methyl group number from benzene to 1,2,4,5-tetramethylbenzene. This implies that volatility of SOA-forming products increases as the number of methyl groups on the aromatic ring increases. There is also a slight increase in VFR from 1,2,4,5-tetramethylbenzene to hexamethylbenzene; however, VFR in SOA formed from all C 10+ group aromatics is lower than that of 1,2,4-trimethylbenzene. The changing VFR trend suggests that chemical components contributing to SOA formation become different when more than four methyl groups are attached to a single aromatic ring. A positive correlation (0.755, p = 0.05) found between mass loading and VFR end implies that the lower the volatility in the products formed from aromatic hydrocarbons, the higher the SOA mass concentration. An opposite correlation between mass loading and VFR is found in previous studies due to the partitioning of more volatile compounds to the particle phase at high mass loading Salo et al., 2011). Therefore, mass loading does not directly lead to the VFR trend in the current study; rather, it is the methyl group number in the SOA precursor that affects the composition of SOA and therefore the monocyclic aromatic hydrocarbon yield (Sect. 3.1) and volatility. The correlation between SOA volatility (VFR) and chemical composition is statistically analyzed (Table S5). O / C (0.937, p = 0.002) and OS c (0.932, p = 0.02) have the highest correlation with VFR end . Previous studies also observed that lower aerosol volatility is correlated to higher O / C ratio (Cappa and Wilson, 2012;Yu et al., 2014) and OS c (Aumont et al., 2012;Hildebrandt Ruiz et al., 2014). Figure 6b and c illustrate the VFR end and O / C or OS c relationship among all the monocyclic aromatic precursors investigated in this study. Benzene and toluene are located in the upper right corner in both graphs, suggesting that significantly more oxidized and less volatile components are formed from monocyclic aromatic precursors with fewer than two methyl groups. The VFR end and chemical components relationship becomes less significant when only monocyclic aromatic precursors with more than two methyl groups are considered.

SOA formation pathway from monocyclic aromatic hydrocarbon
Bicyclic peroxide compounds are considered to be important SOA-forming products from monocyclic aromatic photooxidation (Johnson et al., 2004(Johnson et al., , 2005Song et al., 2005;Wyche et al., 2009;Birdsall et al., 2010;Birdsall and Elrod, 2011;Nakao et al., 2011). However, the significant CO + 2 fragment (f 44 ) observed for SOA by the AMS indicates a contribution of an additional pathway to SOA formation from monocyclic aromatic hydrocarbon photooxidation since it is unlikely that bicyclic peroxides could produce a CO + 2 in the AMS. Hydrogen abstraction from the methyl group is not further discussed here as it accounts for less than 10 % monocyclic aromatic oxidation pathway (Calvert et al., 2002). However, it is important to consider the further reaction of bicyclic peroxide ring scission products, especially in the presence of NO x (Jang and Kamens, 2001;Atkinson and Arey, 2003;Song et al., 2005;Hu et al., 2007;Birdsall and Elrod, 2011;Carter and Heo, 2013). Firstgeneration ring scission products include 1,2-dicarbonyls (glyoxal and methylglyoxal) and unsaturated 1,4-dicarbonyls (Forstner et al., 1997;Jang and Kamens, 2001;Birdsall and Elrod, 2011). These dicarbonyls are small volatile molecules that are unlikely to directly partition into the particle phase. However, these small molecules can potentially grow into low-volatility compounds through oligomerization. Previous studies have suggested that oligomerization can be an important pathway for SOA formation from monocyclic aromatic precursors (Edney et al., 2001;Baltensperger et al., 2005;Hu et al., 2007;Sato et al., 2012). While Kalberer et al. (2004) proposed an oligomerization pathway of 1,2dicarbonyls, Arey et al. (2008) found that unsaturated 1,4dicarbonyls have a higher molar yield than 1,2-dicarbonyls in OH radical-initiated reaction of monocyclic aromatic hydrocarbons. Further, OH radical reaction and photolysis rates are observed to be lower in 1,2-dicarbonyls photolysis (Plum et al., 1983;Chen et al., 2000;Salter et al., 2013;Lockhart et al., 2013) than unsaturated 1,4-dicarbonyls (Bierbach et al., 1994Xiang et al., 2007). This suggests that secondary reaction of unsaturated 1,4-dicarbonyls is more important than that of 1,2-dicarbonyls. Previous studies have found that unsaturated 1,4-dicarbonyls react to form small cyclic furanone compounds (Jang and Kamens, 2001;Bloss et al., 2005;Aschmann et al., 2011). Oligomerization is possible for these small cyclic compounds based on their similar molecular structure with glyoxal and methylglyoxal (c-2-1 and c-2-2 pathways, Fig. 7; Fig. S4). Products from further oligomerization of ring-opening compounds can also partition into the aerosol phase and contribute to SOA formation. Hydrolysis is necessary in both oligomerization pathways ( Fig. S4 and Kalberer et al., 2004), which is consistent with the slight H / C increase observed for most monocyclic aromatic hydrocarbon photooxidation results in this study. However, Nakao et al. (2012) showed that the glyoxal impact on SOA formation is majorly due to OH radical enhancement with glyoxal instead of oligomerization, especially under dry conditions. This indicates that oligomerization from small cyclic furanone is more likely to contribute more to SOA formation than 1,2-dicarbonyl in this work. Other pathways reported in previous studies are also possible to contribute to SOA formation here (Edney et al., 2001 -polyketone;Jang and Kamens, 2001 -aromatic ring retaining products, six-and fivemember non-aromatic ring products, ring-opening products; Bloss et al., 2005 -benzoquinone, epoxide, phenol;Carter and Heo, 2013 -bicyclic hydroperoxide). Our work only addresses differences in the oligomerization pathway contribution to form SOA from monocyclic aromatic hydrocarbons.
A simplified monocyclic aromatic oxidation mechanism for low-NO x conditions is shown (Figs. 7 and 8; the figures only illustrate monocyclic aromatic oxidation related to particle formation). Figure 7 illustrates the oxidation, fragmentation and oligomerization after initial OH addition to the aromatic ring, and Fig. 8 shows the kinetic scheme for SOA formation from monocyclic aromatic hydrocarbons. S 1 , S 2 and S 3 represent bicyclic hydroperoxide compounds, ring-opening compounds and oligomerization products, respectively. Table S6 summarizes the predicted vapor pressures of the benzene photooxidation products using SIM-POL (Pankow and Asher, 2008). The bicyclic hydroperoxide (S 1 , Fig. 8) is more volatile than the oligomers (S 3 in Fig. 8). The volatilities of the bicyclic hydroperoxides are sufficiently high to allow additional oxidation (e.g., add one more hydroperoxide functional group to form C 6 H 6 O 8 ). The further oxidized bicyclic hydroperoxide vapor pressure is predicted to be similar to oligomerization products from reaction of c-2-1 (Fig. S4) with glyoxal. The higher vapor pressure of oligomer products from glyoxal as compared to oligomers from other products indicates that bicyclic hydroperoxides (S 1 ) contribute more to SOA formation in benzene than oligomerization products (S 3 ), especially at higher particle mass loadings, as compared with monocyclic aromatic hydrocarbons containing methyl groups according to the two-product model fitting ( Fig. 1 and Table 2).

Methyl group number impact on SOA formation pathway from monocyclic aromatic hydrocarbon
It is observed that as the number of methyl groups on the monocyclic aromatic precursor increases, mass yield (Sect.  observed yield trend is attributed to the increasing methyl group number enhancing aromatic fragmentation and inhibiting oligomerization. First, the methyl group stabilizes the ring-opening radical (Atkinson, 2007;Ziemann, 2011), thus favoring the ring-opening pathway. Second, the methyl group hinders cyclic compound formation and oligomerization (Fig. 7). Oligomerization is unlikely to occur directly from non-cyclic dicarbonyls (Kalberer et al., 2004) or indirectly from cyclic compounds formed by unsaturated dicarbonyls ( Fig. S5) with increasing methyl group number. Methyl groups both inhibit oligomerization (c-1-3, Fig. 7) and prevent the formation of cyclic compounds from unsaturated dicarbonyls (c-2-3, Fig. 7) when methyl groups are attached to both ends of an unsaturated dicarbonyl. Oligomerization is possible for these ketones through reactions such as aldol condensation and hemiacetal formation (Jang et al., 2002) under acidic conditions. However, this is less favored for the current study in the absence of acidic seeds. Hence, less cyclic compounds are available for subsequent oligomerization, leading to more volatile products and a decrease in SOA formation. Moreover, the SOA composition trend is well explained by a −CH 3 dilution effect. Previous studies on the different gas-phase (Forstner et al., 1997;Yu et al., 1997) and particle-phase (Hamilton et al., 2005;Sato et al., 2007Sato et al., , 2012 products support this methyl group dilution theory. A typical example is that more 3-methyl-2,5-furandione is observed in m-xylene than toluene and vice versa for 2,5furandione. Sato et al. (2010) suggest that more low-reactive ketones are produced rather than aldehydes with increasing Aromatic Hydrocarbon + OH → S1 g → S2 g → S3 g ↓↑ ↓↑ S1 p S3 p number of substituents. However, most ketones or aldehydes detected are so volatile that they mostly exist in the gas phase (Forstner et al., 1997;Yu et al., 1997;Cocker et al., 2001b;Jang and Kamens, 2001). Taken collectively, this implies the importance of oligomerization and methyl substitutes on SOA formation. The observation of a slight H / C decrease from hexamethylbenzene to its SOA components in contrast with the increasing trend for monocyclic aromatic photooxidation for zero to five methyl group substitutes (Sect. 3.2.2) suggests that hydrolysis followed by oligomerization might not be significant when all aromatic ring carbons have attached methyl groups. In addition, the higher O / C and lower H / C (or the higher OS c ) than predicted in Sect. 3.2.3 indicates that SOA components from hexamethylbenzene photooxidation are more oxidized per carbon due to oxidation of the methyl groups, which is possibly related to the steric hindrance of the six methyl groups. Moreover, there is a slightly increasing trend in VFR from 1,2,4,5-tetramethylbenzene to hexam-ethylbenzene (Sect. 3.3.2). Further studies (e.g., photooxidation using isotope-labeled methyl group hexamethylbenzene) are required to probe the unique SOA aspects from hexamethylbenzene photooxidation.

Atmospheric implication
The impact of the number of methyl group substituents on SOA formation has been comprehensively studied in this work by integrating SOA yield with SOA chemical composition and SOA physical properties. A generally decreasing trend is found in the SOA mass yield and the carbon-numberaveraged oxidation level with increasing number of methyl groups. SOA physical properties agree with yield and oxidation results. Therefore, this study demonstrates that the addition of methyl group substitutes to monocyclic aromatic precursors decreases the oxidation of aromatic hydrocarbon to less volatile compounds. Offsetting the amount of CO + 2 and C 2 H 3 O + suggests a methyl group dilution effect on SOA formation from monocyclic aromatic hydrocarbons. The proposed methyl group dilution effect is then applied successfully to the predict SOA elemental ratio. Overall, this study clearly demonstrates the methyl group impact on SOA formation from monocyclic aromatic hydrocarbons.
Benzene and toluene are evaluated as the most important monocyclic aromatic precursors to SOA formation among the six compounds studied due to their high SOA yields and highly oxidized components. Hexamethylbenzene is found to be significantly more oxidized than predicted based on other monocyclic aromatic hydrocarbons studied here. This implies uniqueness in the methyl group behavior (no −H on aromatic ring) in hexamethylbenzene. Oligomerization is proposed to be an important pathway for SOA formation from monocyclic aromatic hydrocarbons. It is likely that oligomerization is even more valuable to SOA formation from monocyclic aromatic hydrocarbons in polluted areas (catalyzed effect; Jang et al., 2002;Iinuma et al., 2004;Noziere et al., 2008) and ambient humidity (Liggio et al., 2015a, b;Hastings et al., 2005).
The Supplement related to this article is available online at doi:10.5194/acp-16-2255-2016-supplement.