Highly oxygenated organic molecules produced by the oxidation of benzene and toluene in a wide range of OH exposure and NO x conditions

. Oxidation of aromatic volatile organic compounds (VOCs) leads to the formation of tropospheric ozone and secondary organic aerosol, for which gaseous oxygenated products are important intermediates. We show, herein, the experimental results of highly oxygenated organic molecules (HOMs) produced by the oxidation of benzene and toluene in a wide range of OH exposure and NO x conditions. The results suggest that multigeneration OH oxidation plays an important role in the product distribution, which likely proceeds more preferably via H subtraction than OH addition for early generation products from light aromatics. More oxygenated products present in our study than in previous ﬂow tube studies, highlighting the impact of experimental conditions on product distributions. The formation of dimeric products, however, was suppressed and might be unfavorable under conditions of high OH exposure and low NO x in toluene oxidation. Under high-NO x conditions, nitrogen-containing multifunctional products are formed, while the formation of other

Abstract. Oxidation of aromatic volatile organic compounds (VOCs) leads to the formation of tropospheric ozone and secondary organic aerosol, for which gaseous oxygenated products are important intermediates. We show, herein, the experimental results of highly oxygenated organic molecules (HOMs) produced by the oxidation of benzene and toluene in a wide range of OH exposure and NO x conditions. The results suggest that multigeneration OH oxidation plays an important role in the product distribution, which likely proceeds more preferably via H subtraction than OH addition for early generation products from light aromatics. More oxygenated products present in our study than in previous flow tube studies, highlighting the impact of experimental conditions on product distributions. The formation of dimeric products, however, was suppressed and might be unfavorable under conditions of high OH exposure and low NO x in toluene oxidation. Under high-NO x conditions, nitrogen-containing multifunctional products are formed, while the formation of other HOMs is suppressed. Products containing two nitrogen atoms become more important as the NO x level increases, and the concentrations of these compounds depend significantly on NO 2 . The highly oxygenated nitrogen-containing products might be peroxyacyl nitrates, implying a prolonged effective lifetime of RO 2 that facilitates regional pollution. Our results call for further investigation on the roles of high-NO 2 conditions in the oxidation of aromatic VOCs.

Introduction
Atmospheric oxidation of volatile organic compounds (VOCs) are crucial in the formation of tropospheric ozone (O 3 ) and secondary organic aerosol (SOA; Calvert et al., 2015;Seinfeld and Pandis, 2016). Enormous studies have been conducted for the O 3 formation potentials and SOA yields of light aromatic VOCs such as benzene and toluene (Calvert et al., 2002;Atkinson and Arey, 2003;Ziemann and Atkinson, 2012). The peroxy radicals (either RO 2 or HO 2 ) that are generated from the oxidation process convert NO to NO 2 . Ground-state oxygen atoms O( 3 P) are produced through the photolysis of NO 2 , and the reaction between the O( 3 P) and O 2 is the main source of tropospheric O 3 (Calvert et al., 2015). The oxygenated organic products, on the other hand, may form SOA through either nucleation or condensation with various mass yields, depending on the structure of the precursors and the NO x (= NO + NO 2 ) level (Ng et al., 2007;Li et al., 2016). Therefore, for both O 3 and SOA formation via VOC oxidation, NO x plays a critical role (Atkinson, 2000;Sato et al., 2012;Tsiligiannis et al., 2019;Garmash et al., 2020).
OH-initiated oxidation of light aromatics occurs mainly via OH addition, with about 90 % of preference (Calvert et al., 2002;Wu et al., 2014;Schwantes et al., 2017). As described in Sect. S1 in the Supplement, the formation of bicyclic peroxy radicals (BPRs) is central to aromatic oxidation in the absence of NO x . Significant fractions of the oxidation (e.g., 0.35 for benzene and 0.65 for toluene) may lead to the formation of BPRs (Scheme S1 in the Supplement; Birdsall et al., 2010), followed by further reactions to form highly oxygenated organic molecules (HOMs; Crounse et al., 2013;Ehn et al., 2014;Berndt et al., 2016;Bianchi et al., 2019). The fate of the BPRs has been recently investigated by using a time-of-flight chemical ionization mass spectrometer (TOF-CIMS), which is suitable for measuring HOMs. The TOF-CIMS measurements show that major products for the reactions between BPRs and HO 2 /RO 2 in the absence of NO x include carbonyls, alcohols, hydroperoxides, dimers, and alkoxy (RO) radicals (Birdsall et al., 2010;Wang et al., 2017;Molteni et al., 2018;Zaytsev et al., 2019b;Garmash et al., 2020). The BPR-derived products that still possess the bicyclic skeleton are considered as being the major ringretaining products from the oxidation of light aromatics (Zaytsev et al., 2019b). The decomposition of RO radicals may lead to fragmented products such as di-carbonyls and epoxides Arey et al., 2009;Zaytsev et al., 2019b). The formation of HOMs may involve multistep auto-oxidation and multigeneration OH reaction (Zaytsev et al., 2019b;Garmash et al., 2020;. Yet, it is still unclear whether the formation of HOMs is controlled by multistep auto-oxidation or multigeneration OH oxidation. Conditions of flow tube or smog chamber experiments have covered various conditions of OH concentrations (10 4 to 10 11 mol cm −3 ) and residence times (10 s to 1 h; Wang et al., 2017;Molteni et al., 2018;Garmash et al., 2020). Extrapolation of these results to tropospheric conditions requires further investigations under a wide range of NO x levels. NO x are rich in urban environments and compete with HO 2 and RO 2 for the termination of RO 2 radicals (Calvert et al., 2002;Seinfeld and Pandis, 2016). Early studies show a strong dependence of SOA mass yields on NO x , owing to the formation of more volatile products through the termination of RO 2 by NO (Ng et al., 2007;Sato et al., 2012). There are also a few recent studies on the gaseous oxygenated products from aromatic oxidation in the presence of NO x (Tsiligiannis et al., 2019;Garmash et al., 2020;. Garmash et al. (2020) found that nitrated phenols (NPs) contribute significantly to the gaseous nitrogen-containing HOMs produced by the benzene oxidation in the presence of NO x . Tsiligiannis et al. (2019) show the prevalent formation of organic nitrates (ONs) from 1,3,5-trimethylbenzene (TMB) oxidation, especially for atmospherically relevant [NO x ] : [VOC] ratios greater than 1. In addition to ONs, Y.  show the elevated formation of dinitrates from the oxidation of TMB isomers in the presence of NO x . In general, the formation of these nitrogen-containing products suppresses the formation of other HOMs (Tsiligiannis et al., 2019;Garmash et al., 2020;. However, the dependence of product distributions on NO x conditions (e.g., [NO x ] : [VOC] or [NO 2 ] : [NO] ratios) remains largely unclear.
In this study, we investigate the production of gaseous HOMs from the OH-initiated oxidation of benzene and toluene in an oxidation flow reactor (OFR) by using nitrate adduct TOF-CIMS (NO − 3 -TOF-CIMS). A wide range of OH exposure and NO x conditions are studied. Distributions and molar yields of key products are investigated. Kinetic analysis helps in inferring the possible formulas of nitrogencontaining HOMs.
2 Experimental section 2.1 Oxidation flow reactor The oxidation experiments are conducted in a 13.3 L Aerodyne Research Inc. potential aerosol mass (PAM) OFR reactor. The schematic of the experimental setup and the example experimental sequence are shown in Figs. S1 and S2 in the Supplement, respectively. The OFR was operated in two continuous flow modes named OFR254-5 and OFR254-5-iN 2 O (Lambe et al., 2017;Peng et al., 2018). For the OFR254-5 mode, the inside UV lamps emit photons at 254 nm to generate OH radicals in the reactor via the reaction of O( 1 D) + H 2 O → 2OH. An outside UV lamp produces O 3 and leads to about 5 ppm (parts per million) of O 3 in the OFR. For the OFR254-5-iN 2 O mode, N 2 O (99.5 %) is injected into the OFR to achieve N 2 O mixing ratios of 1.1 % (OFR254-5-iN 2 O1.1) and 4.4 % (OFR254-5-iN 2 O4.4). NO and NO 2 are then formed by the reaction O( 1 D) + N 2 O → 2NO, followed by the reaction NO + O 3 → NO 2 + O 2 (Lambe et al., 2017). The conditions for OFR254-5 are considered as being low-NO x , whereas the latter are high-NO x . Under each of these three sets of experiments, OH exposure and NO x levels were varied by ramping the voltage of the UV lamps in the OFR. The total flow rate was about 8.4 L min −1 , resulting in a mean residence time of 95 s in the OFR. Relative humidity (RH) in the OFR was about 20 % to 55 % at 25 ± 2 • C, corresponding to H 2 O mixing ratios of about 0.7 % to 1.5 %. Details about the experimental setup are provided in Sect. S2 in the Supplement.
In total, 28 experiments were conducted for initial mixing ratios of 110 ppb (parts per billion) benzene and 50 ppb toluene. Table S1 in the Supplement lists the experimental conditions and the measured and derived quantities for all experiments. The concentrations of reactive species such as OH, HO 2 , NO, and NO 2 were estimated by an OFR-based photochemical box model (PAMchem; Lambe et al., 2017). The box model simulations are described in detail in Sect. S2, for which calibration experiments with SO 2 were conducted to constrain the actinic flux at 254 nm ( Fig. S3 in the Supplement). Briefly, the model suggests OH exposure of about 1.1 × 10 11 to 2.5 × 10 12 mol cm −3 s for our experiments, corresponding to equivalent photochemical age of 0.8 to 19.3 d, with a mean OH concentration of 1.5 × 10 6 mol cm −3 . For high NO x experiments, the OFR exit NO and NO 2 concentra-tions were 0.2 to 5.1 ppb and 15.8 to 231.4 ppb, respectively, leading to [NO x ] : [VOC] ratios of 0.2 to 2.2 for benzene and 0.4 to 4.7 for toluene. The OFR-based photochemical box simulations show that the aromatic oxidation reactions in our experimental conditions were dominated by OH rather than O 3 (Lambe et al., 2011;Peng et al., 2016). Similar to other OFR studies, we think the reaction rates of O 3 with oxidation products that contain double bounds are likely slower compared to that of OH (Molteni et al., 2018;. The NO 3 concentrations in the OFR range from 0.01 to 0.09 ppb. The general reaction rate of RO 2 with OH is 1 order of magnitude greater than that with NO 3 under our experimental conditions (Jenkin et al., 2019). For such rates, the NO 3 reactions with HOM products might be minor. The NO 3 oxidation of phenols, however, may contribute efficiently to the formation of nitrated phenols in the OFR because of the high branching ratio.

Chemical characterization
HOMs were characterized by an Aerodyne Research Inc. NO − 3 -TOF-CIMS. Details about the instrument operation and data analysis are described elsewhere (Cheng et al., 2021). Briefly, mass calibration was performed on the reagent ions and selected Teflon-related ions, which covers the m/z range of 62 to 676, with mass accuracies of less than 10 ppm. The non-production ions were identified by positive matrix factorization (PMF) analysis and were removed from the analysis (Fig. S4 in the Supplement). The HNO 3 NO − 3 adduct ions were identified by the signal correlations between the NO − 3 adduct and HNO 3 NO − 3 adduct ions and were removed from the analysis (Fig. S5 in the Supplement). Only NO − 3 adduct ions are presented herein. The background signals of individual ions were determined on the basis of the measurements made without the injection of VOCs (Fig. S2). Because of the lack of standards, we applied the calibration factor of 4-nitrophenol (i.e., 0.0020 ncps ppt −1 (normalized counts per second per parts per trillion) or 1.66 × 10 10 mol cm −3 ) to HOMs, which is similar to the commonly used calibration factors of H 2 SO 4 (i.e., 1.62×10 10 mol cm −3 for our instrument herein and 1.89 × 10 10 mol cm −3 reported by Jokinen et al., 2012) and perfluoroheptanoic acid (i.e., 1.6×10 10 mol cm −3 as reported by Ehn et al., 2014). All calibration factors are corrected by wall losses. Ehn et al. (2014) reported a ±50 % of uncertainty. For our experiments, we estimate an uncertainty of 42 % on the basis of nitrated phenol calibrations (Cheng et al., 2021). In some experiments, the VOC precursors and less-oxygenated gaseous products were monitored by an IONICON proton transfer reactionquadrupole interface time-of-flight mass spectrometer (PTR-QiTOF). The instrument operation and data analysis have been described previously (Huang et al., 2019). The measurements have a total uncertainty of about 21 %. Particle size distributions were measured by a scanning mobility particle sizer (SMPS; TSI Incorporated; model no. 3938). The SOA mass concentrations are measured by an Aerodyne long timeof-flight soot particle aerosol mass spectrometer (LTOF-SP-AMS) for the calculation of condensation sink in wall loss corrections of HOMs (Zheng et al., 2020).
3 Results and discussion 3.1 Product categories Figure 1 shows the mass spectra of gaseous oxygenated products produced by benzene and toluene oxidation in typical low-and high-NO x experiments. The fitted ion peaks are categorized into fragmented products, open-shell monomeric products, closed-shell monomeric products, and dimeric products, as well as nitrogen-containing products when NO x is present. The proposed product categories are based on traditional understanding and recent mechanistic developments of aromatic oxidation (Wang et al., 2017;Mentel et al., 2015;Molteni et al., 2019). In benzene and toluene (C x H y ) oxidation by OH radicals, the addition of OH and two O 2 molecules leads to the formation of BPR C x H y+1 O 5 , which increase the molecular composition by one H atom and five oxygen atoms. The BPR C x H y+1 O 5 can undergo further autoxidation and forms radicals with the formula C x H y+1 O 7 . Autoxidation of RO 2 radicals involves intramolecular hydrogen shifts to the peroxide group from other carbon atoms and the subsequent addition of oxygen to the produced carboncentered radicals. The H shift itself does not modify the molecular composition, but O 2 addition increases the oxygen content by an even number. Besides, BPR C x H y+1 O 5 can form termination products of carbonyls (C x H y O 4 ), alcohols (C x H y+2 O 4 ), hydroperoxides (C x H y+2 O 5 ), dimers, and nitrogen-containing compounds. Notably, product distributions are affected by the numbers of the generation of the auto-oxidation pathway and the H shift reaction rates. Some product formulae may involve multiple pathways, as discussed later in Sect. 3.2, for which additional information (e.g., collision induced dissociation experiments) may be helpful to further explore possible structures of the HOM molecules (Zaytsev et al., 2019a).
Tables S2-S3 in the Supplement list the corresponding peak lists and relative signal contributions of major products in each category for the low-and high-NO x experiments in Fig. 1. The fragmented products are the most abundant category under low-NO x conditions. They have carbon numbers less than their precursors, possibly indicating ring opening or scission, presumably from RO radicals (Zaytsev et al., 2019b). In addition, fragmented products can be formed through CO elimination from an acyl radical (Rissanen et al., 2014), splitting CO 2 from an RO radical (Garmash et al., 2020), or dealkylation (Birdsall and Elrod, 2011). In this study, C 2 H 4 O 4 shows the highest signal intensity in the fragmented category under low-NO x conditions. C 4 H 4 O 5 and C 5 H 4 O 6 are the other two abundant common products in this category. Under high-NO x conditions, C 4 H 2 O 5 is the most abundant common product for benzene and toluene oxidation. Many of the fragmented products have been detected in SOA (Gallimore et al., 2011;Gowda and Kawamura, 2018).
The ring-retaining HOM monomers may have even (closed shell) or odd (open shell) hydrogen numbers (Molteni et al., 2018;Zaytsev et al., 2019b;Garmash et al., 2020). With relatively large carbon numbers, the RO 2 radicals generally have lifetimes of seconds that are much longer than the RO radicals of < 10 −4 s (Orlando et al., 2003;Seinfeld and Pandis, 2016;Zhao et al., 2018). The open-shell monomeric products observed by the NO − 3 -TOF-CIMS in this study are, therefore, more likely RO 2 radicals rather than RO radicals. Under low-NO x conditions, the BPRs have relatively low signal intensities (0.1 %). Products that are presumably formed by further auto-oxidation of BPRs, such as C 6 H 7 O 7 and C 6 H 7 O 9 for benzene (or C 7 H 9 O 7 and C 7 H 9 O 9 for toluene), have much greater signal intensities compared with the BPRs, especially for the O 9 products. Other openshell monomeric products with two fewer hydrogen atoms (e.g., H 5 for benzene and H 7 for toluene) or with an even number of oxygen atoms (e.g., O 6 or O 10 ) are also present.
Under high-NO x conditions, the main open-shell monomers are the O 9 products, which is similar to the low-NO x case.
Among the closed-shell monomeric products produced by benzene oxidation, C 6 H 6 O 5-10 and C 6 H 8 O 5-10 with one oxygen atom apart show relatively high signal intensities, and C 6 H 6 O 5 and C 6 H 6 O 6 are the most abundant ones. C 6 H 6 O 5 and C 6 H 6 O 6 might be carbonyl products from the termination of C 6 H 7 O 6 and C 6 H 7 O 7 by HO 2 or RO 2 (Zaytsev et al., 2019b;Garmash et al., 2020). C 6 H 7 O 6 is probably formed by the ring breakage of the bicyclic alkoxy radical followed by the 1,5 aldehydic alkoxy H shift reactions (Xu et al., 2020) which involve the RO pathway, as described in detail in Sect. S3 of the Supplement. For other products formed by the termination reactions of BPR with HO 2 or RO 2 , bicyclic hydroperoxide (C 6 H 8 O 5 ), carbonyl (C 6 H 6 O 4 ), and alcohol (C 6 H 8 O 4 ) have relatively small signals, whereas the products that may involve two or three steps of autooxidation (e.g., C 6 H 8 O 7 and C 6 H 8 O 9 ) have greater signal intensities. BPRs (i.e., C 6 H 7 O 5 from benzene oxidation and C 7 H 9 O 5 from toluene oxidation) can undergo unimolecular isomerization reactions to form more oxidized peroxy radicals, which compete with bimolecular reactions (Wang et al., 2017). The relative greater signal intensities of ring-retaining O 7 or O 9 HOM monomers compared with O <6 ones sug-gest that the termination of RO 2 by HO 2 and the phenol oxidation pathway are perhaps relatively less important than the auto-oxidation pathway under experimental conditions herein (Calvert et al., 2002;Schwantes et al., 2017;Garmash et al., 2020). Similarly, for toluene oxidation, the bicyclic hydroperoxide (C 7 H 10 O 5 ) shows lower signals than C 7 H 10 O 7 and C 7 H 10 O 9 that may be associated with multiple steps of auto-oxidation. C 7 H 8 O 6 and C 7 H 8 O 7 are the main closedshell monomers that might be carbonyls that are produced by the termination of C 7 H 9 O 7 and C 7 H 9 O 8 by HO 2 or RO 2 , respectively. The formation of C 7 H 8 O 7 may also be explained by the RO pathway, similar to C 6 H 6 O 5 (Sect. S3). The RO pathway, if it occurs, may potentially lead to various products with significant signals, which remains largely overlooked in studies of aromatic oxidation (Xu et al., 2020).
The other two categories are dimeric and nitrogencontaining products. In our study, the ion intensities of dimeric products with odd hydrogen atoms are low. We, therefore, focus on the dimeric products with even hydrogen atoms. Under low-NO x conditions, dimeric products with 8-14 even oxygen atoms are clearly present. C 12 H 14 O 8 and C 14 H 18 O 8 are perhaps formed via self-reactions of the BPRs for benzene and toluene, respectively. The distribution of dimeric products is in line with the monomeric openshell and closed-shell products. We observe a number of dimeric products with odd oxygen numbers that are perhaps produced by cross-reactions of odd and even oxygen RO 2 (Molteni et al., 2018;Garmash et al., 2020). The presence of NO x results in the formation of nitrogen-containing products with one or two nitrogen atoms, and the signal abundances of all other oxygenated products are much lower than the case of low-NO x experiments. Such a suppression has been reported previously (Tsiligiannis et al., 2019;Garmash et al., 2020;Mehra et al., 2020). The nitrogen-containing products are expected to be organic nitrates or nitrated phenols. For benzene oxidation, C 6 H 5 NO 3 , C 6 H 5 NO 4 , and C 6 H 4 N 2 O 6 are the most abundant, which are plausibly nitrophenol, nitrocatechol, and dinitrocatechol, respectively. For toluene, C 6 H 5 NO 3 (nitrophenol), C 6 H 5 NO 4 (nitrocatechol), C 7 H 7 NO 3 (methylnitrophenol), and C 7 H 7 NO 4 (methylnitrocatechol) are the main tentatively assigned nitrated phenols. Significant secondary production of these compounds from aromatic oxidation have been observed in urban Beijing (Cheng et al., 2021). Other products, such as C 6 H 7 NO 8 , C 6 H 7 NO 9 , and C 7 H 9 NO 8 , are most likely organic (peroxy) nitrates. Schemes S2 and S3 in the Supplement give examples of the proposed formation mechanisms for nitrogen-containing products and multistep auto-oxidation products, starting from the BPR of C 7 H 9 O 5 (Sect. S1).

Product distribution
The products that we observed herein (Tables S2-S3) are generally in agreement with those found in previous low-NO x studies (Schwantes et al., 2017;Molteni et al., 2018;Zaytsev et al., 2019b;Mehra et al., 2020;Garmash et al., 2020). Differences exist in the relative abundance of species with different oxygen contents within each product category (Molteni et al., 2018;Garmash et al., 2020). Table S4 in the Supplement lists the experimental conditions and relative signal intensities of some major oxygenated products formed by benzene oxidation. Molteni et al. (2018) observes a predominant production of C 6 H 8 O 5 (plausibly hydroperoxides) and C 12 H 14 O 8 . Garmash et al. (2020) shows relatively high signals of C 6 H 8 O 9 and C 12 H 14 O 8 in the flow tube experiments, whereas the chamber study and our study both show relatively high signals of C 6 H 8 O 7 and C 6 H 8 O 9 and lower signals of dimeric products. Moreover, the signal intensities of RO 2 (e.g., C 6 H 7 O 5 and C 6 H 7 O 7 ) are greater in our study than in other studies. Both experimental conditions and instrument detection are factors that may affect the product distribution. For example, a longer residence time might promote multistep auto-oxidation. For auto-oxidation, the H shift has rate constants of about 10 −3 to 1 s −1 and is likely the rate-determining step under atmospheric conditions, with reaction times of 1 to 10 3 s that are much longer than the subsequent O 2 addition of microseconds (hereafter µs) (Orlando et al., 2003;Bianchi et al., 2019). Garmash et al. (2020) indicated that fewer auto-oxidation steps would be expected in flow tube experiments. The flow tube study herein has a relatively longer residence time than others, which is consistent with relatively more abundant O 7 and O 9 products. Ambient environments have much lower OH concentrations but longer residence times, depending on meteorological conditions, which may suggest more oxygenated products from multisteps of auto-oxidation. Alternatively, the differences in HO 2 and RO 2 concentrations among different studies that remain unclear might affect the extent of auto-oxidation. For detection, the instrument efficiency might be different for HOMs with different clustering capabilities (e.g., numbers of functional groups as hydrogen bond donors; Hyttinen et al., 2015). Tuning might affect the transmission of product ions in different m/z ranges or affect the efficiency of dimer clustering (Heinritzi et al., 2016;Brophy and Farmer, 2015).

Effects of OH exposure
As shown in Fig. 2a, the concentrations of fragmented, monomeric closed-shell and open-shell and dimeric products formed by benzene oxidation increase with increasing OH exposure that corresponds to 2.4 to 16.1 d of atmospheric equivalent photochemical age. Elevated concentrations of monomeric open-shell products (e.g., RO 2 ) and HO 2 are ex- pected for greater OH exposure, leading to enhanced production of the monomeric closed-shell and dimeric products through the RO 2 +HO 2 and RO 2 +RO 2 reactions. These processes may be limited by the availability of RO 2 (< 0.9 ppt (parts per trillion) in our study) rather than that of HO 2 (0.5-2.4 ppb; Table S1). Consistently, the concentrations of dimeric products are much lower than those of monomeric closed-shell products. For toluene oxidation, the dependence of the product formation on OH exposure is less significant than that for benzene (Fig. 2b). The concentrations of fragmented, monomeric closed-shell and open-shell products first increase and then slightly decrease with the increasing OH exposure that corresponds to 2.4 to 19.4 d of atmospheric equivalent photochemical age. Interestingly, unlike benzene oxidation, the concentrations of dimeric products decrease as the OH exposure increases, suggesting perhaps unfavorable dimer formation under conditions of high OH exposure for toluene oxidation. One potential contributor to the difference of the dependence of dimer formation on OH exposure is the more significant elevated RO 2 concentrations (0.2 to 0.9 ppt) as the OH exposure increases more for benzene oxidation than that (0.3 to 0.5 ppt) for toluene oxidation, while the HO 2 concentrations in the two sets of experiments are similar (1.5-2.4 ppb). The enhancement of RO 2 concentrations may promote the dimer formation through the self-or crossreactions of RO 2 (Mohr et al., 2017). On the other hand, a previous study indicates that the accretion of RO 2 depends on the functional groups of the RO 2 . Whether the decreasing concentrations of dimeric products with OH exposure are related to the steric effects of the substituted methyl group of toluene requires further investigation. The overall atomic oxygen-to-carbon (O : C) ratios of HOM products increase from 1.18 to 1.26 for benzene oxidation and from 1.16 to 1.23 for toluene oxidation as the OH exposure increases, suggesting more oxygenated product distributions as detected. The overall atomic hydrogen-to-carbon (H : C) ratios of HOMs are 1.10-1.14 for benzene oxidation and 1.14-1.18 for toluene oxidation, and the changes with OH exposure are negligible. mainly C x H y+1 O 5 (BPRs) and C x H y+1 O 7 (RO 2 formed from one more step of auto-oxidation). The concentrations of C x H y+1 O 7 are approximately 1 order of magnitude higher than the concentrations of C x H y+1 O 5 . As described in Sect. 3.1, the further reactions of C x H y+1 O 5 and C x H y+1 O 7 may form closed-shell monomers, such as hydroperoxides (C x H y+2 O 5,7 ), carbonyls (C x H y O 4,6 ), and alcohols (C x H y+2 O 4,6 ; Mentel et al., 2015;Molteni et al., 2019), although these formulae may correspond to other functionalities, depending on the reaction rates and numbers of generation of autoxidation pathways. For multiple steps of auto-oxidation of BPRs or RO 2 radicals (C x H y+1 O 5 or C x H y+1 O 7 ), the products with two more hydrogen atoms than the precursor are plausibly hydroperoxides (C x H y+2 O z ), if z is an odd number, and alcohols, if it is an even number. On the other hand, products that have the same hydrogen atoms (C x H y O z ) are likely carbonyls with an even number of z. Instead, if the carbonyl formation involves the RO pathway, C x H y O z , with an odd number of z, may be formed. The concentrations of C x H y+2 O 7 and C x H y O 6 originated from C x H y+1 O 7 are 1 to 2 orders of magnitudes greater than the concentrations of C x H y+2 O 5 and C x H y O 4 , which originated from C x H y+1 O 5 . The greater signals of O 7 products suggest our experimental conditions perhaps favor the formation of more oxygenated products.
Enhanced formation of more oxygenated products was observed for elevated OH exposures. For example, the concentrations of C 6 H 7 O 7 increase first and stay relatively stable as the OH exposure increases, whereas the concentrations of C 6 H 7 O 5 decrease at high OH exposures for benzene oxidation. The concentrations of the hydroperoxide products (C x H y+2 O 5,7 ), such as C 6 H 8 O 7 , increase as OH exposure increases, whereas the concentrations of C 6 H 8 O 5 decreases. For toluene oxidation, the concentrations of C x H y+2 O 5,7 (C 7 H 10 O 5 and C 7 H 10 O 7 ) both decrease. The concentrations of carbonyl products (C x H y O 4,6 ) increase with OH exposure for benzene but not for toluene. We do not observe significant signals for C x H y+2 O 4 (alcohols) from BPR C x H y+1 O 5 , but C x H y+2 O 6 from RO 2 C x H y+1 O 7 are of high concentrations. The concentrations of C 6 H 8 O 6 (alcohol) from benzene also increase as the OH exposure increases, while the concentrations for C 7 H 10 O 6 from toluene decrease. Overall, the enhanced formation of more oxygenated products at high OH exposure is more significant for benzene than for toluene. A possible explanation is that toluene oxidation may involve more multigeneration OH oxidation because of the substituted methyl group.
Increasing OH exposure also enhances the formation of more oxygenated fragmented products (Fig. S6 in the Supplement). C 4 H 4 O 3 (perhaps epoxybutanedial) has been widely observed in aromatic oxidation (Wang et al., 2013;Wu et al., 2014;Zaytsev et al., 2019b). The concentrations of C 4 H 4 O 3 decrease monotonically as the OH exposure increases for both benzene and toluene oxidation. For toluene oxidation, C 3 H 4 O 2 (perhaps methylglyoxal), C 4 H 4 O 2 (per-haps butenedial), and C 5 H 6 O 2 (perhaps methylbutenedial) that were detected by the PTR-QiTOF also show similar monotonic decreases. By contrast, the concentrations of C 3 H 4 O 5 , C 4 H 4 O 5 , and C 5 H 4 O 6 increase with increasing OH exposures for both benzene and toluene oxidation. The concentrations of other O 5-7 products also increase as the OH exposure increases. Similar to the monomeric open-shell and closed-shell products, the increasing trends are more significant for benzene-derived products than those for toluenederived ones. This observation is in agreement with Garmash et al. (2020), who suggested that first-generation fragmented products were converted to more oxygenated ones through further oxidation as the OH exposure increases. Alternatively, these more oxygenated fragmented products might also be direct decomposition products of highly oxygenated RO radicals, which have been shown to play important roles in product formation (Noda et al., 2009;Birdsall and Elrod, 2011).

Multigeneration OH oxidation
In benzene and toluene (C x H y ) oxidation by OH radicals, we may expect that one OH addition could lead to termination products with hydrogen numbers of y and y + 2. When the second OH attack occurs, products with hydrogen numbers < y and > y + 2 may be produced by multigeneration OH reactions (Table S5 in the Supplement). For example, the closed-shell monomeric products from benzene oxidation should contain at least one double bond which can further react with OH radicals via OH addition to form products with two more hydrogen atoms (e.g., to form C 6 H 10 O z ). The formation of C 6 H 4 O z from the closed-shell monomeric products of benzene oxidation may involve multigeneration OH reactions via H abstraction and subsequent termination by a loss of OH or HO 2 (Molteni et al., 2018;Garmash et al., 2020). There are significant contributions of products with hydrogen numbers < y and > y + 2 in our experiments (Tables S2 and S3), highlighting the importance of multigeneration OH reactions in production formation. On the other hand, the formation of y and y +2 products may also involve multistep OH addition or H subtraction that make conversions between the two groups of products. For the phenolic pathway, the main products of C 6 H 6 O z and C 7 H 8 O z may be produced by dihydroxy-OH, trihydroxy-OH, and even multi-OH substituted benzene or toluene (Schwantes et al., 2017). The contributions of a third of the OH attack on the total of detected signals are often very low (Molteni et al., 2018). Figure 3 shows the concentrations and the relative contributions of the closed-shell monomeric product groups, in- For both benzene and toluene oxidation, the concentrations of these monomers increase as the OH exposure increases for the atmospherically relevant equivalent photochemical age (< 10 d). For greater OH exposures, the concentrations of y−2 products keep increasing, whereas the concentrations of other products start to decrease (Fig. 3a-b). The decreasing trends at high OH exposures are more significant for toluenederived products. Such differences might be explained by (1) the faster OH + VOC rate for toluene than for benzene and (2) the methyl group in toluene and the added −OH groups during oxidation that may increase the electron density on the aromatic ring through a resonant electron-donating effect and, thereby, activate the aromatic ring and facilitate further addition reactions to form products with multiple −OH groups (M. . Interestingly, the relative fractions of the C x H y−2 O z and C x H y+4 O z products for both benzene and toluene oxidation show opposite trends (Fig. 3c-d).
The increasing fractions of y-2 products but decreasing fractions of y + 4 products for increasing OH exposures, as well as the monotonically increasing concentration of y-2 products, suggest that the multigeneration OH oxidation may proceed preferably via H subtraction rather than OH addition, according to the current mechanistic understanding of aromatic oxidation. Garmash et al. (2020) also noted that higher OH concentrations caused a larger variety in HOM products, with H abstraction oxidation becoming possibly more significant. Compared with their precursors (benzene and toluene), the closed-shell monomeric products are less conjugated, and thus, OH addition is probably less favorable. Figure S7 in the Supplement shows the scatterplot of the concentrations of HOMs detected by NO − 3 -TOF-CIMS and the VOC oxidation rate for benzene and toluene oxidation under low-NO x conditions. As expected, the product concentrations increase with the VOC oxidation rates. To calculate the molar yields of HOM products, wall losses are corrected, including the loss in the sampling line, the loss to the OFR walls estimated by eddy diffusion, the loss to aerosol particles in the OFR (i.e., the condensation sink calculated on the basis of particle measurements), and the loss to non-condensable products due to continuous reaction with OH (Palm et al., 2016). Details are provided in Sect. S4 in the Supplement. Our estimated molar yields for the HOM products are 0.22 ± 0.10 % (mean ± 1 standard deviation) for benzene oxidation and 0.46 ± 0.20 % for toluene oxidation. These yields are much lower than the smog chamber results of 4.1 % to 14.0 % for benzene oxidation reported by Garmash et al. (2020) but slightly greater than the flow tube yields of 0.1 % to 0.2 % reported by Molteni et al. (2018). A key difference in the experimental conditions is the much longer residence time in the chamber study (Table S4), suggesting, perhaps, a long characteristic time of HOM formation from aromatic oxidation. The study of  Molteni et al. (2018) only provided the initial OH concentration (as listed in Table S4) that should decrease significantly as the reaction proceeded, for which we cannot rule out the possibility of low OH exposure that leads to fewer oxidation steps (i.e., lower yields of HOMs). Instrument sensitivity might also affect the detection of HOMs but is less likely to lead to orders of magnitude difference in the yields.

Effects of NO x level
The NO or NO 2 termination of RO 2 radicals competes with the HO 2 or RO 2 termination and forms nitrogen-containing species at the expense of other highly oxygenated products (Tsiligiannis et al., 2019;Garmash et al., 2020;Mehra et al., 2020). Yet, quantitative understanding in the effects of NO x on the formation of oxygenated products is limited for aromatic precursors compared with those for biogenic VOCs (Nah et al., 2016;Lambe et al., 2017;Sarnela et al., 2018). The high-NO x experiments herein were conducted with [NO 2 ] : [NO] ratios of 20 to 120 (Table S1), which may represent urban afternoon conditions when fresh NO x emission is mostly converted to NO 2 , and intense photochemistry fuels the oxidation of aromatics accompanying the NO x emission (Newland et al., 2021). Figure [HO 2 ] to evaluate termination pathways to form various nitrogen-containing products such as nitrated phenols, organic nitrate, and peroxynitrate. The contributions from RO 2 + RO 2 termination are probably minor because of the low concentrations of RO 2 (< 0.9 ppt). Given the low and narrow range of NO x levels in the OFR254-iN 2 O1.1 experiments, the product concentrations for all lamp voltages (i.e., a range of OH exposure) are averaged and shown as the first data point in each panel of Fig. 4. Similar to previous findings, nitrogen-containing products are the dominant species in the spectra, with concentrations up to 18.3 and 7.3 ppt for benzene and toluene oxidation, re- spectively. The concentrations of all HOMs start to decrease at high [NO x ] : [HO 2 ] ratios, except that the concentrations of dimeric products from benzene oxidation remain steady ( Fig. 4a-b). The decreasing trends for monomeric closedshell, open-shell, and fragmented products are expected, indicating significant competition of radical terminations by NO or NO 2 . The decreasing concentrations of nitrogencontaining products for increasing [NO x ] : [HO 2 ] are, however, counterintuitive. Figure 4c-d show the concentrations of main individual nitrogen-containing products. The initial increase in the concentrations of nitrogen-containing species might be explained by a decrease in HO 2 concentrations from 0.8-1.5 to 0.5-0.7 ppb as a result of the switch of 1.1 % of N 2 O injection to 4.4 % ( Table S1). The further reduction in these compounds as [NO x ] : [HO 2 ] increases is probably related to the simultaneous decrease in RO 2 or, alternatively, further reactions to products that have two nitrogen atoms. Figure S8 in the Supplement shows similar concentration trends for individual ring scission and ring-retaining products with or without nitrogen in their formulas. There seem to be optimal [NO x ] : [HO 2 ] ratios of 130 to 240 for the formation of nitrogen-containing products. The dependence of those products with only one nitrogen atom on NO x is, however, not strong. The availability of RO 2 is perhaps the key factor that limits their formation at low NO x levels and affects further reactions to form products with two nitrogen atoms at high NO x levels.
Nitrated phenols are the most abundant nitrogencontaining products under high-NO x conditions. In aromatic oxidation, these compounds are formed by the reaction of phenoxy RO radicals with NO 2 (Jenkin et al., 2003). In the Master Chemical Mechanism (MCMv3.3.1), the phe-noxy RO radicals can be formed via OH oxidation of phenols (k OH = 2.8 × 10 −11 cm 3 mol −1 s −1 ), with a low branching ratio of 0.06. They can also be formed by the NO 3 oxidation of phenols (k NO 3 = 3.8 × 10 −12 cm 3 mol −1 s −1 ), with a high branching ratio of 0.74 (IUPAC, 2008). Under high-NO x conditions, the estimated concentrations of OH and NO 3 radicals are 0.05 to 0.3 ppb and 0.01 to 0.09 ppb, respectively, suggesting that the NO 3 oxidation of phenols contributed efficiently to the formation of nitrated phenols in the OFR experiments herein. When [NO x ] : [HO 2 ] increases, the concentrations of C 6,7 H 5,7 NO 4 (perhaps nitrocatechol or methylnitrocatechol) increase first and then decrease. The concentrations of C 6,7 H 5,7 NO 3 (perhaps nitrophenol or methylnitrophenol), however, show a weak dependence on the NO x level, suggesting the availability of RO radical might be the limiting factor in controlling the formation of nitrated phenols herein. Interestingly, the concentrations of products with two nitrogen atoms such as C 6 H 4 N 2 O 6 , C 6 H 8 N 2 O 9 , and C 7 H 10 N 2 O 9 (only for toluene) steadily increase as [NO x ] : [HO 2 ] rises, suggesting a strong dependence of the formation of these species on NO 2 .
Formation of ROOH, RONO 2 , and ROONO 2 Jenkin et al. (2019) suggested that the overall rate coefficients for RO 2 + HO 2 reactions are 1.92 × 10 −11 and 1.98 × 10 −11 cm 3 mol −1 s −1 at 298 K for benzene and toluene oxidation, respectively. For the OFR conditions (Table S1), the characteristic time for the RO 2 termination by HO 2 was perhaps < 10 s, which is much shorter than the OFR residence time of 95 s. The rate coefficients of the hydroperoxide pathway (RO 2 + HO 2 → ROOH + O 2 ) may be con-strained by the concentration ratios of [ROOH] multiplied by the loss rate of ROOH to [RO 2 ][HO 2 ] (Sect. S5 in the Supplement). For example, assuming that the C 6 H 7 O 7 and C 7 H 9 O 7 are the RO 2 radicals, and C 6 H 8 O 7 and C 7 H 10 O 7 are the corresponding ROOH for benzene and toluene oxidation, respectively, the slopes in Fig. 5a indicate that the rate coefficients of hydroperoxides are 1.20 × 10 −11 and 1.26 × 10 −11 cm 3 mol −1 s −1 . These rate coefficients suggest that the branching ratios of the hydroperoxide formation under low-NO x conditions are 0.62 and 0.64 for benzene-and toluene-derived RO 2 , respectively, which are consistent with the estimated branching ratios of 0.52-1.00 in the literature (Jenkin et al., 2019).
In the presence of NO x , the reactions between C x H y+1 O 7 and nitrogen oxides lead to both chain propagation and chain termination to form RO radicals and nitrogen-containing products. Similar to the analysis of the hydroperoxide formation, the slopes in Fig. 5b suggest that the rate coefficients of RO 2 +NO(+M) → RONO 2 (+M) are 2.87×10 −11 and 6.12 × 10 −11 cm 3 mol −1 s −1 for benzene and toluene oxidation under OFR254-5-iN 2 O1.1 conditions, respectively (Sect. S5). These coefficients are more than 1 order of magnitude greater than the values reported by Jenkin et al. (2019; i.e., 8.09 × 10 −13 and 1.10-8.45 × 10 −13 cm 3 mol −1 s −1 for benzene and toluene oxidation, respectively). One explanation is that the detected C x H y+1 NO 8 contains multifunctional groups and represents not only non-peroxy organic nitrates (RONO 2 ) but also peroxy organic nitrates (ROONO 2 ). As shown in Fig. 5c Xu et al. (2020) indicate that the formation of nonperoxy organic nitrates (RONO 2 ) is minor in aromatic oxidation. The detected ROONO 2 are likely RC(O)OONO 2 because ROONO 2 are usually thermally unstable intermediates (Kirchner et al., 1999), and RC(O)OONO 2 can be detected by the NO − 3 -TOF-CIMS . Acyl peroxy radicals RC(O)OO (i.e., a type of peroxy radical) may react with NO 2 to produce peroxyacyl nitrate (RC(O)OONO 2 ; Jenkin et al., 2019). The formation of the RC(O)OONO 2 requires (1) the original RO 2 radicals to be C x H y+1 O 6 instead of C x H y+1 O 7 , and (2) an acyl (−C=O) group that is connected to the O−O bond. The formation of C x H y+1 O 6 is feasible through the RO pathway (Sect. S3). The hydroperoxides corresponding to C x H y+1 O 6 are C x H y+2 O 6 (i.e., C 6 H 8 O 6 and C 7 H 10 O 6 for benzene and toluene, respectively). Figure   , which prevents us from further investigation.

Atmospheric implications
In this study, we investigated the formation of HOMs in the OFR by the oxidation of benzene and toluene in a wide range of OH exposure and NO x conditions. Recent findings emphasize the significance of the bicyclic channel in the oxidation of light aromatics (Wang et al., 2017;Molteni et al., 2018;Tsiligiannis et al., 2019;Zaytsev et al., 2019b;Garmash et al., 2020;. The presence of NO x enhances the formation of organonitrates and even dinitrate organic compounds. Extensive autoxidation and accretion reaction has been reported for substituted aromatics (Y. . Our results show enhanced formation of more oxygenated products for elevated OH exposures. The formation of dimeric products, however, seems unfavorable under conditions of high OH exposure and low NO x for substituted aromatics. The suppression of dimeric products may affect the contribution of aromatic HOMs to new particle formation in the downwind of urban atmosphere. The changes in product distribution and concentration highlight the possibility that multigeneration OH oxidation proceed preferably via H subtraction rather than OH addition, although one HOM formula may correspond to various functionalities. For aged air masses, this may reduce the H : C ratios of HOM products from aromatic oxidation. Under high-NO x conditions, we show that the formation of products containing one nitrogen atom perhaps depends more significantly on the organic radicals (RO or RO 2 ) but less so on NO 2 , while the formation of products containing two nitrogen atoms depends significantly on NO 2 . Further investigation on the roles of high-NO 2 conditions that represent a wide range of anthropogenically influenced environments in the oxidation of aromatic VOCs are needed. Moreover, we found that non-peroxy organic nitrates might form via RO 2 + NO under low-NO 2 conditions, and RO 2 + NO 2 may dominate to produce ROONO 2 or RC(O)OONO 2 under high-NO 2 conditions. The reaction of RC(O)OO with NO 2 to produce peroxyacyl nitrates should be of particular importance with high [NO 2 ] : [NO] ratios of tens to hundreds. Both of ROONO 2 and RC(O)OONO 2 are reservoirs of RO 2 radicals. Under conditions of high [NO 2 ] : [NO] ratios (e.g., late afternoon in urban or suburban environments), the effective lifetimes of RO 2 radicals might become longer because of the formation of RC(O)OONO 2 . Subsequent slow release of RO 2 radicals with the help of NO 2 may extend the formation of HOMs from VOC oxidation in urban environments to early evening, when OH starts to decline and NO 3 has not yet built up, facilitating the development of regional SOA pollution.