All toluene photooxidation experiments were performed in a 4 m3 chamber. The chamber has been described in detail in our previous study (Liu et al., 2021). Briefly, the chamber was constructed with a 0.08 mm thick Teflon FEP film. Fifty UV-B lamps (TUV36W, Philips) with peak wavelengths of 254 nm were set up around the chamber and used as the light source to drive OH radical formation through hydrogen peroxide (H2O2) photolysis. Mirror-surfaced stainless steel was used as the interior wall of the enclosure to maximize and homogenize the interior light intensity. All experiments were performed at room temperature (293–298 K), and atmospheric pressure was maintained in the chamber at all times.

Before each experiment, the chamber was flushed with zero air for at least 18 h, after which the concentration of particles was less than 1 cm-3. Zero air was generated by a zero air supply (111-D3N, Thermo Scientific™, USA). The flow rate of zero air was controlled at 20 L min-1 by a mass flow controller (D088C/ZM, Beijing Sevenstar Electronics Co.) during the process of chamber reaction bag inflating. The relative humidity (RH) of zero air was about 20 %. For each experiment, measured amounts of toluene (Sigma-Aldrich, analytically pure) and H2O2 solution (Sigma-Aldrich, 30 wt % in H2O) were injected into a Teflon bulb with microsyringes. Zero air was passed through the injection tube to make sure all the liquids had evaporated to the gas phase and were blown into the chamber. Toluene concentration was measured with a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS, Ionicon Analytik, Austria). The evolution of toluene concentration for different experiments was shown in Fig. S1. The OH concentration in the chamber was calculated based on the first-order decay of toluene concentration. There was no obvious difference in OH concentrations in the different NOx and NH3 levels (Fig. S2). The average OH concentration over the entire reaction period was 5.87×107 molec. cm-3. NOx (Air Liquid Shanghai, 510 ppm NO2 in N2) and/or NH3 (Air Liquid Shanghai, 502 ppm NH3 in N2) were introduced directly into the chamber to reach the required concentrations. For experiments with NOx, although only NO2 was introduced into the chamber before photooxidation, NO could be formed through NO2 photolysis under the UV light irradiation, so NO always co-existed with NO2 in the photooxidation system (Zhao et al., 2018). NOx concentration was measured online by the NO–NO2–NOx analyzer (model 42C, Thermo Electron Corporation, USA). Each experiment was performed without seed aerosols. After all the reactants were added, the chamber stood undisturbed for 10 min without turning on the light to ensure that the reactant gases in the chamber were evenly mixed. The photooxidation started when the UV light was turn on.

The experimental conditions for the toluene photooxidation are listed in Table 1. In our work, the OH and toluene concentrations were higher than those of urban conditions. The purpose of the high OH and toluene concentrations is to obtain enough particle production samples for offline collections and accurate measurements. The toluene concentrations remained stable under the different experimental conditions, the variations in toluene-derived SOA mass concentration and yield were only affected by the different NO2 and/or NH3 concentrations in this study. Toluene was studied here as the representative of total aromatic VOCs in the urban atmosphere, thus its concentration used in this study was much higher than that in the real atmosphere. However, the concentration ratio of toluene to OH in this study is similar to that under real atmospheric conditions (Prinn et al., 1995; Zou et al., 2015).

For each experiment, a scanning mobility particle sizer (SMPS) was used to record the particle size distribution and volume concentration of the toluene-derived SOA. The SMPS was composed of a differential mobility analyzer (DMA model 3081, TSI Inc., USA) and a condensation particle counter (CPC model 3776, TSI Inc., USA), which were used for screening particles with specific electrical mobility diameter (from 14.1 to 736.5 nm) and for counting the number of the selected particles, respectively. The sheath gas velocity was 3 L min-1, and the sample gas velocity was 0.3 L min-1. The size scan was repeated every 5 min. During each scan cycle, the scan time was 240 s, and the particle sizes ranged from 13.6 to 726.5 nm. A density of 1.4 g m-3, which was measured by Ng et al. (2007b), was used for the calculation of toluene SOA mass concentration from the particle volume concentration (Ng et al., 2007b).

In this study, the toluene SOA chemical compositions were characterized with an online high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS, Aerodyne Research Inc., USA). The sample flow passed through a Nafion dryer, and the RH of the sample gas was reduced to below 20 % before entering the AMS. In the injection port, an aerodynamic lens focused particles with a vacuum aerodynamic diameter below 1 µm into a narrow beam. Particles impacted a flash vaporizer (600 ∘C) at the rear of the sizing region under high vacuum (∼ 10-7 Torr) and were subsequently ionized by electron impact ionization (70 eV). Then, the positively charged ions entered the ToF section and were separated. V-mode (m/Δm=∼2000) was used in the AMS ToF section to achieve the high signal-to-noise ratio. The separated ion fragments were analyzed by a mass spectrometer with scans from 1 to 300 m/z. The composition-dependent collection efficiency (CE) was applied to the data based on the methods established by Middlebrook et al. (2012). For mass concentration calculations, 1.1, 1.2, and 1.4 were applied as the default relative ionization efficiency (RIE) values of nitrate, sulfate, and organic compounds, respectively. The standard AMS data analysis software SQUIRREL 1.63B, coupled with PIKA 1.23B in Igor Pro (WaveMetrics, Inc., Portland, Oregon), which was retrieved from http://cires1.colorado.edu/jimenez-group/ToFAMSResources/ToFSoftware/ (last access: 26 November 2021), was used for the analysis of elemental ratios and the ion speciated compositions of toluene SOA in the chamber. Note that the elemental ratios (i.e., O/C, H/C, and N/C) and mass-to-carbon ratio (OM/OC) were all calculated using the Aiken-Ambient method for comparability with previous studies (Aiken et al., 2008). In order to further explore the changes in SOA chemical composition, a PMF of the high-resolution mass spectra was performed to determine the different organic aerosol (OA) factors during the toluene photooxidation process. We preformed the PMF analysis in the same way as Zhang et al. (2011), the details of which are provided in the Supplement.

The changes in absorption spectra and the absorbance of the toluene-derived SOA under different conditions were determined using a UV spectrophotometer (UV-3600, Shimadzu, Japan) with a 1 cm cuvette. The SOA was collected from a 3 m3 sample gas onto the 46.2 mm PTFE filter (Whatman™, UK). The collected SOA sample was dissolved in 5 mL of methanol (HPLC grade, > 99.8 %) with 30 min of sonication. As reported by Chen and Bond (2010), > 92 % of SOA is extractable by organic solvents (e.g., methanol), which means that almost all organic matter was extracted in this study. The filter extracts were filtered through 0.2 µm PTFE syringe filters to remove suspended insoluble particles. Before detection of the optical absorbance, a cuvette filled with pure methanol was scanned as a blank to provide a spectral background. The absorption was detected over the range of 200 to 800 nm with a resolution of 0.5 nm-1. The light absorption coefficient of the particles at a specific wavelength λ (Absλ, Mm-1) was calculated according to Eq. (1): 1Absλ=Aλ-A700⋅VlVa⋅L⋅ln⁡10, where A700 is the background value of light absorption intensity, calculated as the average value of light absorption intensity from 695–705 nm to reduce the limits of error in measurement; Vl and Va are the volumes of methanol with dissolved particles and sampled air, respectively; and L is the optical path length. Because Absλ was strongly dependent on the amount of SOA, all Absλ results were normalized based on the SOA mass collected on the filter. The normalized result was defined as the mass absorption coefficient (MAC, m2 g-1), calculated using Eq. (2): 2MACλ=AbsλM, where M (µg m-3) represents the concentration of methanol-soluble organic carbon.

In order to investigate the effect of NH3 and NOx on SOA formation from toluene photooxidation, a control test was carried out. SOA yield (Y) is defined as Y=ΔM0/ΔHC, where ΔM0 is the produced organic aerosol mass concentration (µg m-3), and ΔHC is the mass concentration of reacted toluene (µg m-3). The evolution of SOA mass concentrations and SOA yield at different conditions during the photooxidation process are shown in Fig. 1. Recent experiments showed that the wall loss of organic vapors to the Teflon walls should not be ignored (Zhang et al., 2014, 2015b) and represented a major challenge in investigating SOA formation with environmental chambers (Krechmer et al., 2020). The formation of SOA in laboratory chambers may be substantially suppressed due to losses of SOA-forming vapors to chamber walls, but this effect on SOA formation has not yet been quantitatively established. However, the particle wall-loss rates were detected at the end of the chamber experiment after the UV lamps were turned off, and the mass concentration was corrected in the same way as by Jiang et al. (2020) and Pathak et al. (2007). After the wall-loss correction, the particle mass concentration was almost constant; the different wall-loss effect caused by gaseous oxidation products formed in the different experiment conditions was considered remedied. There was a noticeable increase in the SOA mass concentration in the presence of NH3, which was consistent with previous studies (Qi et al., 2020; Chu et al., 2016). The mass concentration of SOA increased from 637±14.6 µg m-3 without NH3 to a maximum of 867±12.7 µg m-3 with 200 ppb NH3. Previous studies attributed the enhancement of SOA to the formation of NOA from acid–base reactions between NH3 / NH4+ and carboxyl groups or Maillard reactions of NH3 / NH4+ with carbonyl functional groups (Noziere et al., 2009; Ortiz-Montalvo et al., 2014; Liu et al., 2015; Qi et al., 2020). In contrast, SOA concentrations were lower in the presence of NOx, and the maximum mass concentration of toluene SOA was only 452±18.9 µg m-3 with 63 ppb of initial NOx. The branching of RO2 loss among different pathways has an important influence on the product distribution and SOA formation. The fate of RO2 mainly depends on the concentrations of NOx (Zhao et al., 2018; Liu et al., 2019a; Xu et al., 2020). Several studies have shown that, instead of RO2 reacting with RO2 / HO2, NO would react with RO2 to form the RO intermediate and produces more oxidation products with higher volatilities through fragmentation in the presence of NOx (Zhao et al., 2018; Liu et al., 2019a; Xu et al., 2020). Highly volatile compounds cannot readily partition into the particle phase, so this substantially suppresses the formation of SOA.

The NOx and NH3 had opposite effects on toluene SOA formation in this study. Interestingly, however, the highest SOA mass concentration (1020±10.6 µg m-3) occurred in the presence of both NOx and NH3, which was nearly 1.6 times higher than that observed in the absence of NOx or NH3. Although inorganic aerosol (i.e., NH4NO3) was formed from the interaction of NH3 and NOx in the chamber, the upper limit of the inorganic matter only accounted for 6.6 % of the total mass of particulate matter (Table S1) in the NH3 + NOx experiment. Therefore, NH4NO3 was not the main cause of the increase in particulate matter; the co-occurrence of NH3 and NOx had a synergistic effect on the SOA formation, because their combined effect on SOA formation was greater than the sum of their separate effects. Qi et al. (2020) found that the promotion of NH3 on toluene SOA formation was more obviously under high NOx concentration; SOA yield increased 3.7 % and 4.6 % for 70 and 160 ppb initial NOx concentration, respectively, when 200 ppb NH3 was added into the chamber. Li et al. (2018) showed that the presence NH3 can promote the particle size growth of SOA; at the same time, this particle growth rate was higher under low VOC / NOx (or high-NOx) conditions. All in all, the joint effect of multiple environmental factors on SOA formation is not the simple summation of the influences of various factors on SOA formation. This may at least partly explain why predictions of SOA concentrations in large-scale atmospheric models, which typically describe SOA formation from data derived from chamber experiments, are frequently lower than field observations (Volkamer et al., 2006).

The traditional SOA formation mechanism is based on the chemical compositions obtained through offline detection of the chemical composition of SOA (Jang et al., 2002; Liu et al., 2019a; Xu et al., 2020). SOA is continually evolving in the atmosphere, and the aging process of SOA co-occurs with its formation process, resulting in the transformation of SOA chemical composition continuously proceeding during the photooxidation process. Therefore, the AMS was used for online measurement of the SOA chemical composition and how the chemical composition evolved in the photooxidation process will be discussed in the following section.

Chemical composition of SOA is very complex. The average carbon oxidation state (OSC) has been shown to be an ideal conceptual framework to describe changes in the degree of oxidation undergone by SOA (Kroll et al., 2011) and has been widely applied in field and laboratory studies (Chen et al., 2018; Mandariya et al., 2019). Average OSC calculation is shown in the Supplement. Figure 2 shows the changes in the OSC of toluene SOA formed in different experiments. Notably, toluene SOA OSC values were in the range between -0.5 and 0, which is consistent with that of semi-volatile oxygenated organic aerosols (SV-OOAs). However, the different OSC values and the change trends observed for the toluene SOA formed in different conditions (with and without NH3 / NOx) in Fig. 2 indicated that there was a photooxidation mechanism active during SOA formation, which ultimately changed the SOA chemical compositions.

The OSC increased over time for all SOAs that were formed in the absence of NH3. There are several possible reasons for the increasing trend of OSC values. Firstly, a dynamic equilibrium of semi-volatile vapors may have been achieved between the particle phase and gas phase during the earlier toluene oxidation process. The increase in SOA led to a reduction in the concentration of gas-phase semi-volatile organic products. A decreasing concentration of gas-phase semi-volatile organic compound products would suppress their transformation from gas phase to particulate phase. More lower-volatility gas-phase oxidation products with higher OSC values would then be shifted to the particle phase, which would be responsible for the continuing increase in SOA and its OSC. Secondly, the formed SOA could have further been oxidized by OH radicals through heterogeneous reactions (Kourtchev et al., 2015; Liu et al., 2019b). This could be the main reason for the increase in the OSC when the SOA concentration was no longer increasing. Finally, as pointed out by Malecha and Nizkorodov (2016), even if there was no OH radical in the chamber, the photodegradation of SOA can produce small oxygenated volatile organic compounds (e.g., acetaldehyde OSC=-1, and acetone OSC≈-1.3) under UV light irradiation. The photoproduction of OVOCs (oxygenated VOCs) from SOA had a lower OSC value than that of SOA. Although the loss of SOA through photodegradation is small, the OSC value of SOA still had increased to a certain extent (Malecha and Nizkorodov, 2016).

The fact that additional photochemical processing results in the dynamic evolution of the OSC over time has been demonstrated in both field and laboratory experiments (Jimenez et al., 2009). The atmospheric oxidation of OA tends towards higher OSC regardless of the original OA source (Herndon et al., 2008). However, when NH3 was present, the OSC of total SOA was almost unchanged for the whole photooxidation period. Carboxyl and carbonyl are the main oxygen-containing functional groups responsible for the toluene photooxidation production (Ji et al., 2017). An organic ammonium salt with four H atoms can offset an increase in OSC value caused by the formation of organic acids or carboxy group with two O atoms through acid–base reactions (Kuwata and Martin, 2012; Liu et al., 2015). In addition, NH3 / NH4+ may react with carbonyl functional groups through Maillard reactions, consuming the oxygen atom in the carbonyl group and leading to the formation of species with covalent carbon-nitrogen bonds (Lee et al., 2013; Zhang et al., 2015a; Qi et al., 2020). Xu et al. (2018) showed that imidazole compounds (OSC ≈-1.3) generated through heterogeneous reaction between NH3 and carbonyl compounds might contribute to the decrease in the OSC of SOA. It is clear that an increase in OSC caused by the formation of oxygen-containing functional groups (carboxyl, carbonyl, etc.) would be counteracted through acid–base reactions or Maillard reactions in the presence of NH3. After 60 min of UV light irradiation, there was no more SOA formation; however, the OSC did decrease slightly in Exp. 2 and 3, illustrating that the NH3 could continue to react with SOA through heterogeneous processes. Huang et al. (2016) also pointed out that the portion of semi-volatile products with low OSC formed at the later stage of photooxidation also contributed to the decreased OSC.

The OSC of the toluene SOA formed with NOx was lower than that formed in the absence of NOx, no matter whether NH3 was present in the chamber or not. This indicated that an increased NOx concentration benefits the formation of high volatility oxidation products with lower OSC values (Kroll et al., 2011; Jimenez et al., 2009). However, the relationships between OSC and SOA mass concentration with and without NH3 were opposite to each other. Predictably, the SOA formation mechanism in the presence of NOx is different from that with NOx + NH3. In the absence of NH3, the RO intermediate, which is easily fragmented to produce relatively high-volatility compounds, was the dominant product of the NOx = RO2 reaction (Zhao et al., 2018; Liu et al., 2019a; Xu et al., 2020). Highly volatile compounds cannot readily partition into the particle phase, which subsequently results in a lower SOA yield in the presence of NOx (Yang et al., 2020). Thereby, both OSC and the SOA mass concentration were lower when the 60 ppb of NOx that was added into the chamber. However, when both NOx and NH3 were present, the toluene-derived SOA had the lowest OSC value, but the highest mass concentration. This result suggested that although NOx promotes the formation of higher-volatility compounds, these higher-volatility compounds (e.g., glyoxal) can react with NH3 and partition into the particle phase, which could contribute to the increase in SOA formation. Huffman et al. (2009) observed that aerosol volatility was inversely correlated with the extent of oxidation of OA components. The low value of OSC in the presence of NOx indicated that NOx would promote the formation of the relatively high-volatility compounds. However, the lower OSC value in the presence of NH3 indicated that the high-volatility compounds would promote partitioning into the particle phase when reacting with NH3.

Fragments derived from the AMS data have also been widely used to explore the bulk compositions and properties of SOA (Ng et al., 2010, 2017). The m/z 43 (f43) frequency was dominated by ion C2H3O+, which is the tracer for organic compounds with alcohol and carbonyl functional groups (Alfarra et al., 2006). Meanwhile, the m/z 44 (f44) signal was dominated by CO2+ ions, which is the tracer for organic compounds with carboxyl functional groups and an indicator of highly oxygenated organic aerosols (Ng et al., 2010). Here, we use the approach of Ng et al. (2010) by plotting the fractions of the total organic signal at m/z 43 vs. m/z 44 (f43 vs. f44). The change of f43 vs. f44, which has an inflection point during the photooxidation process, is shown in Figs. 3 and S3. In our study, the change before the inflection point was defined as the formation stage, and the linear fit of f43 vs. f44 for the formation stage is shown by the dashed lines. The change in f43 vs. f44 after the inflection point was defined as the stable stage, and the linear fit of f43 vs. f44 in this stage is shown by the solid lines. The formation and stable stages of the f43 vs. f44 relationship during the experiment are discussed separately here.

In the stable stage, the increase in f44 and decrease in f43 with increasing OH exposure indicated that the carbonyl groups in toluene SOA were oxidized to carboxyl groups by the aging process. For the experiments without NH3 and NOx, the slope ratio of f43 vs. f44 was -3.9. When there was 60 ppb initial NOx, the f43 was almost stable, while the f44 increased with the oxidation process. There was a lower slope ratio of f43 vs. f44, indicating that organic compounds with more alcohol and carbonyl functional groups had formed in the presence of NOx. But for the experiments with 200 ppb initial NH3, the slope ratios of f43 vs. f44 were only -1.1 and -1.3 in the presence and absence of NOx, respectively. According to the above results, we can see that more carbonyl groups are consumed as carboxyl groups are formed in the presence of NH3. The carbonyls can be oxidized to organic acids (Kawamura and Bikkina, 2016), but unreacted carbonyl can be nucleophilically attacked by NH3 / NH4+ to form nitrogen-containing heterocyclic compounds, e.g., imidazole (Grace et al., 2019; Lian et al., 2020). Meanwhile, the peak f44 value decreased from 0.13 to 0.10 when NH3 was added into the chamber. This suggested that the heterogeneous reaction of NH3 / NH4+ could promote the consumption of particle-phase carbonyl groups (Xu et al., 2018) and must inhibit the formation of carboxyl groups in the SOA aging process. According to the changing trend of SOA concentration over time, the photooxidation process was divided into formation stage and stable stage. As shown in Fig. 1, the first half hour of photooxidation when SOA concentration increased linearly with time was defined as SOA formation stage. After 60 min of photooxidation, SOA concentration was not changed with reaction time, and it was defined as stable stage. The differences in spectra of toluene SOA in the formation stage and stable stage are shown in Fig. 4. A lower signal intensity variation in CO2+ in the presence of NH3 also illustrated that NH3 would inhibit heterogeneous reactions that form carboxyl groups.

In the formation stage, the slope ratios of f43 vs. f44 were almost the same for both experiments without NOx. It can thus be seen that the presence or absence of NH3 does not affect the change trend of f43 vs. f44 in the SOA formation stage. Therefore, the gas-phase homogeneous reaction of NH3 on SOA formation is not important. Clearly, the particle-phase heterogeneous reaction was the main reaction pathway by which NH3 participated in the photooxidation process and toluene SOA formation. However, negative correlations were observed between f43 and f44 in the presence of NOx. Based on this, we concluded that NOx not only affects the SOA formation through the particle-phase heterogeneous reactions but also through gas-phase homogeneous reactions.

A temporal evaluation of the toluene SOA chemical composition during photooxidation is vital to the analysis of the NOA formation mechanism in the presence of NH3 and/or NOx. Therefore, this study further compared the chemical properties of the SOA generated under different experimental conditions by applying a PMF analysis to the HR-ToF-AMS data (Chen et al., 2019). A summary of the PMF results is presented in Figs. S4–S7. For the toluene OH-photooxidation experiments with NH3 and/or NOx present, two factors were identified from the PMF analysis in the same way as in Chen et al. (2019). The H/C, O/C, and N/C values of these two factors are shown in Fig. 5. The factor with higher N/C values was defined as high-nitrogen OA (Hi-NOA). In contrast, the factor with lower N/C values was defined as low-nitrogen OA (Lo-NOA). Figure 6 exhibits the evolution of Hi-NOA and Lo-NOA during the photooxidation process as resolved from the PMF analysis of different initial NOx / NH3 concentrations. While similar evolutionary trends were observed under different conditions, the relative intensities and the chemical compositions of these two factors in each experiment were not consistent.

For the toluene SOA formed under NH3 conditions, both Lo-NOA and Hi-NOA had similar O/C values, which were fully oxygenated with an average of 0.74±0.04 (Fig. 5a). These O/C values were comparable to the low-volatility oxygenated organic aerosols (LV-OOAs) with an O/C value ranging from 0.6 to 1 (Jimenez et al., 2009). The main difference between these two OA sources was the N/C ratio. The N/C ratio of Hi-NOA (N/C = 0.032) was about 3 times higher than that of Lo-NOA (N/C = 0.010) (Fig. 5a). The evolution of these two OA sources during the photooxidation process is shown in Fig. 6a. The components of toluene SOA were mostly Lo-NOA during the initial phase of SOA formation, but Hi-NOA toluene SOA started forming after 1 min and continued to increase. The Lo-NOA reached the maximum mass concentration after 30 min of the photooxidation, and then decreased. Such a declining trend of Lo-NOA at longer reaction times reflected the conversion of Lo-NOA into something else in the particle phase. As the Lo-NOA decreased, the mass concentration of Hi-NOA gradually increased. Thus, the Hi-NOA should be derived from the heterogeneous reaction of Lo-NOA with NH3 / NH4+. At the same time, it was proved that the formation pathway of Hi-NOA was not through reaction of NH3 with later-generation gas-phase products in the homogeneous gas phase. With the gradual replacement of Lo-NOA by Hi-NOA, the ratio of [Hi-NOA] / [Lo-NOA] stabilized at 5–6.

For the toluene SOA formed under NOx conditions, there was not a large difference between the N/C ratios of Hi-NOA (N/C = 0.019) and Lo-NOA (N/C = 0.014) (Fig. 5c). At the end of the NOx experiment, the ratio of [Hi-NOA] / [Lo-NOA] was only 3:2 (Fig. 6c). It follows that the contribution of the heterogeneous NOx reaction to the N/C ratio of toluene SOA was not obvious. Therefore, the formation of NOA in the presence of NOx mainly occurred through gas-phase homogeneous reactions, which was consistent with the results in Sect. 3.3.

The changing trend of N/C with time in the presence of NH3 was different to that with NOx present. The evolution trends of the N/C of SOA in different experiments are shown in Fig. 7. In the presence of NH3, the N/C value gradually increased throughout the photooxidation process. The increased N/C value in the photooxidation process was attributed to the heterogeneous NH3 reaction with SOA. But in the presence of NOx, the N/C increased rapidly to its maximum value where it was stable for the rest of the reaction. This could mean that the heterogeneous reaction of toluene SOA with NOx to form NOA was not as important as the gas-phase homogeneous reaction.

When both NOx and NH3 were added into the chamber, the N/C ratios of Hi-NOA and Lo-NOA were 0.062 and 0.029, respectively (Fig. 5b). The N/C ratio of Hi-NOA, which was comparable to the recently isolated nitrogen-enriched OA value (0.053) observed by Sun et al. (2011), was much higher than that observed in the experiments with only NH3 or NOx. It was even higher than the sum of the N/C ratios from both Exp. 2 with NH3 and Exp. 4 with NOx. In order to calculate the relative contributions of NH3 and NOx to N/C, it was assumed that the effects of NH3 or NOx on the N/C ratio in the Hi-NOA and Lo-NOA factors did not change among different experimental conditions. For Lo-NOA, the contributions of NH3 and NOx to the N/C value were 0.0126 and 0.0164, and their relative intensities were 43 % and 57 %, respectively. For Hi-NOA, the contributions of NH3 and NOx to the N/C values were 0.0404 and 0.0216, and their relative intensities were 65 % and 35 %, respectively. For the experiment with both NH3 and NOx, the contribution of NH3 to N/C was higher by 26 %, and the contribution of NOx to N/C was higher by 17 % compared to the experiments with the single pollutants. The co-existence of NH3 and NOx further enhanced the N/C value of toluene SOA, indicating that a synergetic interaction between NH3 and NOx further enhanced organic nitrogen formation.

The optical characteristics of toluene SOA formed from different NH3 and NOx conditions were investigated. The MAC of toluene-derived SOA detected over the range of 230–600 nm is displayed in Fig. 8. Over the entire UV detection range, an increase in light absorption was observed when the toluene SOA formed in the presence of NOx or NH3.

By looking at Fig. 8 in detail, one can see that the MAC of toluene SOA formed with (red line) and without (black line) NH3 overlapped at 250 nm, but when the UV wavelength exceeded 250 nm the MAC of the toluene SOA formed in the presence of NH3 was higher. The red line reflects an obvious characteristic absorption peak at 270–280 nm, which was mainly due to the absorption of the n→π* electronic transitions. The imidazole compounds were formed through the Maillard reactions between NH3 / NH4+ with carbonyl functional groups (Zhang et al., 2015a). The C=N double bonds in the organonitrogen imidazole compounds can act as effective chromophores since both π→π* and n→π* transitions are chromatically active (Nguyen et al., 2013). The UV–Vis spectra of imine and pyrrole show broad bands at 270 nm (NIST, 2020), which was consistent with the UV absorption peak of the n→π* band observed here. According to the AMS results, carbonyl was the main functional group of toluene SOA. The emergence of absorption peaks at 270–280 nm demonstrated that some organonitrogen imidazole compounds (e.g., imines and pyrrole) were formed through the heterogeneous reaction of toluene with NH3. Meanwhile, the high-molecular-weight nitrogen-containing organic species might have formed through Maillard reactions in the particle phase (Wang et al., 2010). This was also a reason for the increase in SOA mass concentration in the presence of NH3.

The green line in Fig. 8 represents the MAC value of toluene-derived SOA in the presence of NOx, which was also higher than the black line (control) throughout the UV detection range. When compared with the red line, the green line had no obvious characteristic peak at 280 nm, but it had higher absorbance in the range between 240 and 280 nm. This indicated that both NOx and NH3 increased the absorbance of toluene SOA, while the chromophores generated from the reactions between toluene-derived SOA with NH3 or NOx did not behave in the same way.

The blue line in Fig. 8 represents the absorbance of toluene SOA formed in the presence of both NOx and NH3. The MAC of toluene SOA formed in the presence of both NOx and NH3 was higher than the toluene SOA formed in the presence of either NH3 or NOx. There might have been a synergetic effect between NOx and NH3 on the absorbance of toluene SOA. Considering that the mass concentration of toluene SOA formed in the presence of both NH3 and NOx was the highest, as described in Sect. 3.1, the co-existence of NH3 and NOx may also result in the toluene SOA having stronger light absorption and atmospheric radiative forcing. We also noted a higher MAC value at 280 nm, which illustrated that the presence of NOx could promote the formation of imines and pyrrole in the photooxidation system of toluene with NH3.