Significant source of secondary aerosol: formation from gasoline evaporative emissions in the presence of SO₂ and NH₃

Gasoline evaporative emissions have become an important anthropogenic source of urban atmospheric volatile organic compounds (VOCs) and secondary organic aerosol (SOA). These emissions have a significant impact on regional air quality, especially in China where car ownership is growing rapidly. However, the contribution of evaporative emissions to secondary aerosol (SA) is not clear in an air pollution complex in which a high concentration of SO₂ and NH₃ was present. In this study, the effects of SO₂ and NH₃ on SA formation from unburned gasoline vapor were investigated in a 30 m³ indoor smog chamber. It was found that an increase in SO₂ and NH₃ concentrations (0–151 and 0–200 ppb, respectively) could linearly promote the formation of SA, which could be enhanced by a factor of 1.6–2.6 and 2.0–2.5, respectively. Sulfate was most sensitive to the SO₂ concentration, followed by organic aerosol, which was due not only to the acid catalytic effect, but was also related to the formation of organic sulfur-containing compounds. In the case of an increasing NH₃ concentration, ammonium nitrate increased more significantly than organic aerosol, and nitrogen-containing organics were also enhanced, as revealed by the results of positive matrix factorization (PMF) analysis. New particle formation (NPF) and particle size growth were also significantly enhanced in the presence of SO₂ and NH₃. This work indicates that gasoline evaporative emissions will be a significant source of SA, especially in the presence of high concentrations of SO₂ and NH₃. Meanwhile, these emissions might also be a potential source of sulfur- and nitrogen-containing organics. Our work provides a scientific basis for the synergistic emission reduction of secondary aerosol precursors, including NO_x, SO₂, NH₃, and particularly VOCs, to mitigate particulate matter (PM) pollution in China.

1 Introduction

Many areas in China such as the Beijing–Tianjin–Hebei region (BTH), Yangtze River Delta (YRD), Sichuan Basin, and Pearl River Delta (PRD) are suffering from severe haze events (Li et al., 2017; Sun et al., 2016; Shen et al., 2015; He et al., 2014; Huang et al., 2014; Guo et al., 2014; Tan et al., 2009). Haze pollution has attracted widespread attention in recent years because of its adverse effects on human health, climate change, and visibility (Thalman et al., 2017; Davidson et al., 2005; Pöschl, 2005).

During haze events, high concentrations of SO₂, NH₃, and volatile organic compounds (VOCs) have always been observed (Zou et al., 2015; Liu et al., 2013; Meng et al., 2011; Yang et al., 2009), which are the precursors of secondary aerosol (SA). Although the emission of SO₂ has decreased continuously since 2005 (Lu et al., 2010), China is still the largest contributor of SO₂ emissions in the world, mainly owing to the great demand for coal combustion (Bauduin et al., 2016). Also, high concentrations of SO₂ of more than 100 ppb (parts per billion) have been observed in northern China, especially during the heating period (Hou et al., 2016; Tong et al., 2016; Yang et al., 2009). As for atmospheric NH₃, as an alkaline inorganic gas, its main emission source is agricultural practices in China (Zhang et al., 2018; Fu et al., 2015). Vehicles equipped with three-way catalytic converters also contribute to NH₃ emissions in urban areas (Sun et al., 2017). Sometimes, high concentrations of NH₃ of up to 100 ppb have been observed in Beijing, China (Ianniello et al., 2010), which are mainly derived from regional transportation for agricultural activity and fertilizer use, but the influence of local traffic emissions in Beijing cannot be excluded (Pan et al., 2016; Kang et al., 2016). With respect to VOCs, aromatics from anthropogenic sources (especially vehicle-related sources in urban areas) are critical secondary organic aerosol (SOA) precursors (T. Liu et al., 2015a; Gordon et al., 2014; Platt et al., 2013; Calvert et al., 2002). These aromatics could react with oxidants (e.g., O₃, OH, and NO₃ radicals) and undergo multi-step oxidative processes to form multifunctional products, which have sufficiently low volatility to contribute to SOA via gas–particle partitioning (Hallquist et al., 2009; Atkinson and Arey, 2003).

Research has shown that SA makes a significant contribution (30 %–77 %) to PM_2.5 (particles with a diameter less than 2.5 µm) during severe haze events in China (Huang et al., 2014; Guo et al., 2014; Jimenez et al., 2009). However, there is still a significant gap between the predicted SA derived from current atmospheric quality models and that observed in field observations (Zhao et al., 2018a; Yang et al., 2018; Zheng et al., 2015). Therefore, considering the characteristics of complex pollution in China, it is crucial to study the synergistic effects of SO₂ and NH₃ on the formation of SA, which have been considered an important potential source of SA formation (Zhao et al., 2018b; Chu et al., 2016; Liu et al., 2016; Santiago et al., 2012; Na et al., 2007).

A few studies have focused on the influence of SO₂ or NH₃ on SA formation. Jang and Kamens (2001) first reported the acid catalytical effect of acidic H₂SO₄ on the oxidation of atmospheric carbonyls. The promotion effect of SO₂ was further found in SA formation from typical biogenic (e.g., isoprene and α-pinene) (Lin et al., 2013; Jaoui et al., 2008; Kleindienst et al., 2006; Edney et al., 2005) and anthropogenic (e.g., toluene, o-xylene, 1,3,5-trimethylbenzene, and gasoline vehicle exhaust) precursors (Chu et al., 2016; Liu et al., 2016; Santiago et al., 2012) through acid-catalyzed heterogeneous reactions (Jang et al., 2002, 2003a, b; Czoschke et al., 2003), which promote the reactive uptake process of organic species or enhance the formation of high-molecular-weight compounds (Liggio and Li, 2008, 2006; Liggio et al., 2007). With regard to the role of NH₃ in SA formation, knowledge is still limited. In previous studies, inconsistent impacts of NH₃ on SA formation have been reported under different precursor systems. For example, NH₃ could elevate SA formation in the α-pinene–ozone oxidation system through acid–base reactions (Na et al., 2007), while the effects of NH₃ neutralization were masked by other multiple factors and did not show significant influence on isoprene-derived SOA formation (Lin et al., 2013). The addition of NH₃ even significantly reduced SA formation in the styrene–ozone system, which was caused by nucleophilic attack from the NH₃ molecule, leading to rapid decomposition of the major aerosol products (Na et al., 2006). For the photo-oxidation of aromatic VOCs (e.g., toluene, o-, m-, and p-xylene), the presence of NH₃ could facilitate new particle formation (NPF) and particle growth, subsequently leading to increased SA formation (Li et al., 2018; T. Liu et al., 2015b).

At the present time, the effects of SO₂ and NH₃ on SA formation have rarely been studied under highly complex pollution conditions (Chu et al., 2016). Vehicular evaporative emissions have been reported to be non-negligible contributors (39.20 %) to ambient VOCs from anthropogenic sources compared with vehicular tailpipe emissions (H. Liu et al., 2017). In addition to short-chain alkanes, a certain proportion of aromatics and alkanes (C₆ to C₁₂) were also contained in the evaporative emissions (Liu et al., 2008; Zhang et al., 2013). Previous studies have reported that aromatics and long-chain (C₆ to C₁₉) alkanes, which are intermediate-volatility organic compounds (IVOCs) (Donahue et al., 2006), could contribute to SOA formation (Pye and Pouliot, 2012; Tkacik et al., 2012; Lim and Ziemann, 2005). Therefore, it is necessary to study the influence of SO₂ and NH₃ on SA formation from evaporative emissions.

In this study, unburned gasoline vapor was used as a substitute for evaporative emissions, and the roles of SO₂ and NH₃ in SA formation from the photo-oxidation of unburned gasoline vapor were investigated in a 30 m³ indoor smog chamber, in order to understand the formation potential of SA from the oxidation of gasoline vapor in the cocktail of pollutants in Beijing. The respective influences of SO₂ and NH₃ on both the microphysics and chemistry of SA formation were examined. Meanwhile, the chemical compositions of the formed SOA in the presence of SO₂ and NH₃ were further explored by applying positive matrix factorization (PMF) analysis. The formation potentials of SA and sulfur- and nitrogen-containing organics from vehicular evaporative emissions in the presence of SO₂ and NH₃ were evaluated and discussed.

2 Materials and methods

2.1 Gasoline fuel

The utilized gasoline fuel with grade 92 was collected (refer to the standard method for manual sampling of petroleum liquids, GB/T 4756-2015) from a gas station located in Beijing. The gasoline complies with the China V gasoline fuel standard. It contains 65.1 % (v∕v) alkanes (C₆ to C₁₂), 22.8 % (v∕v) aromatics (mainly including benzene, toluene, xylene, and trimethylbenzene), and 12.1 % (v∕v) alkenes. The composition of the gasoline is similar to the gasoline collected in northern China reported by Tang et al. (2015) and could represent the gasoline used in most areas of China for studying SA formation potential. Details on the gasoline composition are given in Table S1 in the Supplement.

2.2 Smog chamber facility

A series of photochemical experiments with unburned gasoline vapor in the absence or presence of SO₂ or NH₃ were performed in a 30 m³ indoor smog chamber at the Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences (RCEES-CAS). The detailed schematic structure of the indoor smog chamber is given in Fig. S1 in the Supplement and described elsewhere (Chen et al., 2019a, b). Briefly, the cuboid chamber reactor ($L\times W\times H=\mathrm{3.0}\times \mathrm{2.5}\times \mathrm{4.0}$ m, $S/V=\mathrm{1.97}$ m⁻¹) was irradiated by 120 UV lamps (Philips) with peak intensity at 365 nm, providing an NO₂ photolysis rate of 0.55 min⁻¹. The interior was coated with 125 µm thick FEP100 film (DuPont^™, US) and the chamber was located in a temperature-controlled room in which the temperature (T) and relative humidity (RH) could be controlled mechanically. A three-wing stainless-steel fan coated with Teflon was installed inside the reactor to guarantee that the gas- and particle-phase species mix sufficiently before photochemical reaction.

The chamber was also equipped with a series of gas- and particle-phase monitoring instruments. For gaseous NO_x, O₃, and SO₂, a chemiluminescence analyzer (model 42i-TL, Thermo Fisher Scientific, USA), a UV photometric analyzer (model 49i, Thermo Fisher Scientific, USA), and a pulsed fluorescence analyzer (model 43i, Thermo Fisher Scientific, USA) were used to monitor the concentrations in real time, respectively. The VOC species in gasoline were measured with a gas chromatograph (7890B GC, Agilent, USA) equipped with a DB-624 column (60 m × 0.25 mm × 1.40 µm; Agilent, USA) and a mass spectrometry detector (5977A MS, Agilent, USA) (GC-MS). In addition, proton-transfer reaction time-of-flight mass spectrometry (PTR-ToF) (Ionicon Analytik GmbH, Austria) was also used for the measurement of gas-phase hydrocarbons and their intermediate products (Yuan et al., 2017). The size distribution and number concentration of the formed particulate matter (PM) were measured using a scanning mobility particle sizer (SMPS; TSI, USA), which was composed of a differential mobility analyzer (DMA; 3080 Classifier, TSI, USA) coupled with a condensation particle counter (CPC; 3776, TSI, USA). The mass concentration was estimated based on the volume concentration and the density of PM calculated from the equation $\mathit{\rho }=d_{\mathrm{va}}/d_{\mathrm{m}}$, where d_va is the mean vacuum aerodynamic diameter measured by an Aerodyne high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) and d_m is the mean electrical mobility diameter measured by SMPS (DeCarlo et al., 2004). The calculated density of PM ranged from 1.5 to 1.6 g cm⁻³ in the different reaction systems, which was in the range of density of SOA derived from aromatic hydrocarbons (1.24–1.48 g cm⁻³) (Sato et al., 2010) and ammonium nitrate (NH₄NO₃, 1.72 g cm⁻³) (Bahreini et al., 2005) and was comparable with previous studies (Li et al., 2018). The mass concentration and chemical composition of PM were simultaneously monitored using a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS; Aerodyne Research Inc. USA). For all experiments, the HR-ToF-AMS operated in a cycle including two modes, 3 min V mode and 2 min W mode. Specifically, the V mode (higher signal) can obtain the mass concentrations of the aerosols and the W mode (higher resolution) can obtain high-resolution mass spectral data. The inlet flow rate, ionization efficiency (IE), and particle sizing were calibrated according to the standard protocols (Drewnick et al., 2005; Jimenez et al., 2003; Jayne et al., 2000) using the size-selected pure ammonium nitrate (AN) particles. All HR-ToF-AMS data were analyzed with the ToF-AMS analysis toolkit SQUIRREL 1.57I/PIKA 1.16I version in Igor Pro Version 6.37. HR-ToF-AMS results were also corrected using the mass concentration derived from SMPS according to the same method as Gordon et al. (2014); the details of this correction are shown in the Supplement. As for the RH control system, it is achieved by vaporizing Milli-Q ultrapure water contained in a 5.0 L high-pressure-resistant container, and the water vapor is flushed with purified dry zero air into the chamber. T and RH were monitored in real time using a hydro-thermometer (Vaisala HMP110) during the entirety of each experiment.

2.3 Wall loss corrections

The measured particle concentration was corrected in accordance with the relationship between the deposition rate (k_dep) and particle diameter (D_p, nm) (i.e., $k_{\mathrm{dep}}=\mathrm{4.15}\times {\mathrm{10}}^{-\mathrm{7}}\times D_{\mathrm{p}}^{\mathrm{1.89}}+\mathrm{1.39}\times D_{\mathrm{p}}^{-\mathrm{0.88}}$), which was described by Takekawa et al. (2003). The wall loss rates of NO₂, NO, O₃, SO₂, and VOC species were determined to be $(\mathrm{1.67}\pm \mathrm{0.25})\times {\mathrm{10}}^{-\mathrm{4}}$, $(\mathrm{1.32}\pm \mathrm{0.32})\times {\mathrm{10}}^{-\mathrm{4}}$, $(\mathrm{3.32}\pm \mathrm{0.21})\times {\mathrm{10}}^{-\mathrm{4}}$, $(\mathrm{4.52}\pm \mathrm{0.11})\times {\mathrm{10}}^{-\mathrm{4}}$, and $(\mathrm{2.20}\pm \mathrm{0.39})\times {\mathrm{10}}^{-\mathrm{4}}$ min⁻¹, respectively. Therefore, the wall loss of gas-phase species was evaluated to be less than 5 % of their maximum concentration in this study.

Wall losses of semi-volatile organic compounds (SVOCs) and low-volatility organic compounds (LVOCs) would lead to a substantial underestimation of SA formation (Krechmer et al., 2016; Ye et al., 2016; Zhang et al., 2015, 2014), which is caused by the competition between vapor types condensing onto particles versus onto chamber walls. This competition could be evaluated by the corresponding timescales associated with reaching gas-to-particle partitioning equilibrium (${\stackrel{\mathrm{‾}}{\mathit{\tau }}}_{\text{g-p}}$) and vapor wall loss (τ_g-w) (Zhang et al., 2014), and this underestimation of SA formation could be approximately quantified by the ratio of these two timescales (i.e., ${\stackrel{\mathrm{‾}}{\mathit{\tau }}}_{\text{g-p}}/{\mathit{\tau }}_{\text{g-w}}$). According to the methods described by Zhang et al. (2014), ${\stackrel{\mathrm{‾}}{\mathit{\tau }}}_{\text{g-p}}$ and τ_g-w could be estimated assuming an upper bound and a lower bound of the molecular mass of organic vapor (MW) (100–300 g mol⁻¹) (as discussed in the Supplement). In order to accurately quantify the SA formation, the underestimation caused by the loss of SVOCs and LVOCs (including gaseous H₂SO₄) to the chamber walls was taken into account in this study. In this study, the SA yields were underestimated by a factor of 1.97–2.82 when considering the ratio of these two timescales (i.e., ${\stackrel{\mathrm{‾}}{\mathit{\tau }}}_{\text{g-p}}/{\mathit{\tau }}_{\text{g-w}}$), which showed a decreasing trend with the increase in SO₂ and NH₃ initial concentrations, suggesting that an increasing proportion of vapor is partitioned onto the suspended particle surface rather than the chamber wall.

2.4 Experimental conditions

Prior to each experiment, the chamber reactor was flushed by purified and dry zero air for about 24–36 h at a flow rate of 100 L min⁻¹ until almost no gas-phase species (i.e., NO_x, O₃, and SO₂) could be detected (<1 ppb) and the particle number concentration was <10 cm⁻³. Before the experiments, the chamber was humidified to ∼50 % RH by passing purified zero air through ultrapure water (18.2 MΩ; Millipore Milli-Q). After that, a known volume of liquid gasoline (100 µL) was injected into the chamber through a heated Teflon line system (∼100 ^∘C) carried by purified dry zero air to ensure that all were evaporated into the chamber. Subsequently, NO, SO₂, and/or NH₃ were successively injected into the chamber from standard gas cylinders using mass flow controllers. The initial VOCs∕NO_x ratio (ppb C ppb⁻¹) was kept constant (Table 1). In order to reduce the adsorption of NH₃ in the pipeline, the NH₃ flow in a bypass line was balanced for about 30 min before it was injected into the chamber. The concentrations of NO and SO₂ were continuously monitored until they were stable, ensuring that the gaseous species mixed well in the chamber. For the concentration of NH₃, the value was estimated according to the amount of NH₃ introduced and the volume of the reactor chamber. The experiment was then conducted for about 8 h after turning off the fan and turning on the UV lights. All the experiments were performed at a temperature of 26±1 ^∘C and wet conditions (RH $=\mathrm{50}\pm \mathrm{3}$ %). The detailed experimental conditions are listed in Table 1. The letters in the abbreviations represent the reactants introduced into the chamber reactor for each experiment. For example, SGN is an experiment with the presence of sulfur dioxide (S), gasoline vapor (G), and nitrogen oxides (N). Four experiments (Exps. SGN1, SGN2, SGN3, and SGN4) were carried out at different SO₂ initial concentrations. AGN is an experiment with the presence of ammonia (A), gasoline vapor (G), and nitrogen oxides (N). Two experiments (Exps. AGN1 and AGN2) were carried out at different NH₃ initial concentrations.

Table 1 Summary of experimental conditions in this study.

^a Letters in the abbreviations represent the reactants introduced into the chamber reactor; i.e., “G” represents gasoline, “N” represents nitrogen oxides, “S” represents sulfur dioxide, and “A” represents ammonia. ^b The concentration of NH₃ is estimated by the amount of NH₃ added and the volume of the smog chamber. ^c The surface area of aerosol particles measured by SMPS after 480 min of each experiment. ^d SA yield was calculated after taking vapor and particle wall loss into account.

3 Results and discussion

3.1 Effect of SO₂ and NH₃ on the gas-phase species

Time-resolved concentrations of inorganic and organic gas-phase species during the photo-oxidation of gasoline ∕ NO_x in the absence or presence of SO₂ and NH₃ are shown in Figs. S2 and S3 in the Supplement, respectively. After turning on the UV lights, NO was rapidly converted to NO₂. At the same time, O₃ was gradually generated, with a maximum concentration of up to 350 ppb (Fig. S2). As shown in Fig. S2, there was no obvious difference in the variation of NO_x and O₃ in the presence of SO₂ or NH₃. Additionally, the decay of typical VOC precursors (e.g., benzene, toluene, methylcyclopentane, methylcyclohexane) measured by PTR-ToF and GC-MS are given in Fig. S3, which traced very closely with each other (Fig. S4 in the Supplement). There were also no observable differences in these precursor VOCs among these experiments. According to the decay curves of aromatic hydrocarbons, the OH radical concentrations were estimated to be (7.54–8.40) ×10⁶ molec. cm⁻³, which were also similar among these experiments. This was consistent with the previous study conducted by Chu et al. (2016), who found that the presence of SO₂ and NH₃ did not significantly impact the OH concentration during the photo-oxidation of toluene in the presence of NO_x.

However, the gas-phase intermediates formed during the photo-oxidation of gasoline ∕ NO_x under different conditions, such as small-molecule oxygenated VOCs (OVOCs), could also be measured by PTR-ToF. The time series of OVOC concentration can vary with the concentration of SO₂ and NH₃. For example, we observed that acetic acid concentration decreased with the increased concentration of SO₂ (Fig. S5 in the Supplement), suggesting that the uptake of acetic acid may be enhanced. This phenomenon was consistent with that reported by Liggio and Li (2006), who observed that the uptake of organic compounds under acidic conditions was enhanced significantly. Moreover, the presence of high concentrations of SO₂ would generate gaseous H₂SO₄, which would contribute to the formation of the particle phase, as discussed in the next section. Similarly, the concentration of acetic acid also showed an obviously decreased trend in the presence of NH₃ (Fig. S5 in the Supplement), which could be caused by the acid–base reaction or the uptake of acetic acid in the presence of NH₃ (Y. Liu et al., 2015).

3.2 Role of SO₂ in secondary aerosol formation

To investigate the effects of SO₂ on SA formation from the photo-oxidation of gasoline ∕ NO_x, smog chamber experiments with different SO₂ initial concentrations were carried out (Table 1). As shown in Fig. 1, compared to the experiments without the addition of SO₂, the SA concentration was enhanced to different degrees (1.6–2.6 times) in the presence of different SO₂ concentrations (35–151 ppb; i.e., 100–431 µg m⁻³). As for each chemical species (i.e., organics, nitrate, sulfate, and ammonium), they all showed a trend of linear increase with the increase in SO₂ concentration (Fig. 2), especially for the sulfate ($k=\mathrm{8.4}\times {\mathrm{10}}^{-\mathrm{2}}$) and organic aerosol ($k=\mathrm{2.9}\times {\mathrm{10}}^{-\mathrm{2}}$). Previous studies have also revealed its promoting role in SA formation from different precursors (Zhao et al., 2018a; S. Liu et al., 2017; Liu et al., 2016; Díaz-de-Mera et al., 2017; Chu et al., 2016).

Figure 1 Time series of secondary aerosol concentrations during the photo-oxidation experiments with different SO₂ concentrations (Exps. GN, SGN1, SGN2, SGN3, and SGN4).

Figure 2 Linear relationship between the concentration of chemical species (i.e., organic (green), nitrate (blue), sulfate (red), and ammonium (orange)) and SO₂ under different SO₂ initial concentration conditions (Exps. GN, SGN1, SGN2, SGN3, and SGN4). Each line represents a linear fitting, and the k values are the corresponding slopes for each chemical species.

Figure 3 Time series of the size distributions for the generated secondary aerosol during the photo-oxidation experiments with different SO₂ initial concentrations (Exps. GN, SGN1, SGN2, SGN3, and SGN4). $D_{\mathrm{p},max}$ and N_max represent the maximal diameter and number concentration of generated secondary aerosol, respectively, during each photo-oxidation experiment.

Additionally, the particle number concentrations and size growth were greatly enhanced by the presence of SO₂. As evident from Fig. 3, the corresponding maximal particle number concentrations (5.82×10⁴–1.91×10⁵ cm⁻³) were significantly enhanced by a factor of 2.9–3.3 in the presence of SO₂. This universal phenomenon has been reported by many studies (Díaz-de-Mera et al., 2017; S. Liu et al., 2017; Liu et al., 2016; Chu et al., 2016). For example, the maximal particle number concentrations were enhanced by the presence of SO₂ (∼130 ppb) to 1 order of magnitude in the photo-oxidation of high concentrations of toluene ∕ NO_x (Chu et al., 2016). For complex precursor systems (gasoline vehicle exhaust), Liu et al. (2016) have also found that under high SO₂ concentration (∼150 ppb) conditions, the maximum particle number concentrations increased by 5.4–48 times compared to those without SO₂ during the photo-oxidation of gasoline vehicle exhaust. This higher magnification of SO₂ might be related to the different VOC composition between evaporative emissions and gasoline vehicle exhaust, especially the aromatic and IVOCs (H. Liu et al., 2017). Our recent study demonstrated that SOA formation could be significantly enhanced by the increase in aromatic content (Chen et al., 2019b). Those unspeciated organic emissions (e.g., IVOCs) from gasoline vehicle exhaust would also make a significant contribution to SOA formation (Jathar et al., 2014; Gordon et al., 2014). Moreover, a small amount of POA was present in the initial reaction systems in Liu et al. (2016). This enhanced SOA formation and the preexisting POA would provide larger surface areas for the condensation and heterogeneous uptake of low-volatility vapor (e.g., gaseous H₂SO₄), thus promoting a higher magnification in particle number concentrations in the presence of SO₂. Higher initial mixing ratios of precursors (2.2–4.3 ppm) were also present in the reaction systems conducted by Liu et al. (2016), which would be further beneficial to SOA formation. In addition, size distributions of generated SA in smaller size ranges (4–160 nm) were also determined using another SMPS equipped with a nanometer differential mobility analyzer (nano-DMA), indicating that the new particle formation (NPF) phenomenon was enhanced significantly when the SO₂ concentration increased (Fig. S6). The presence of high concentrations of SO₂ would generate H₂SO₄, which would contribute to nucleation and increase the total particle number concentrations (Zhao et al., 2018a; Sipilä et al., 2010). As the SO₂ concentration increased from 35 to 151 ppb, the maximal particle diameters (144–172 nm) became larger, which will have a direct impact on the scattering and absorption of light (Seinfeld and Pandis, 2016). An enhancement effect of SO₂ on the surface area of particles was also observed. As shown in Table 1, the surface area of aerosol particles at the end of each experiment increased from 1.12×10³ to 2.46×10³ µm² cm⁻³ when the SO₂ concentration increased from 0 to 151 ppb. The larger surface area would be beneficial to the condensation and heterogeneous uptake of low-volatility vapor (Chapleski et al., 2016), consequently leading to higher SA yield in the presence of SO₂ (Table 1) (Santiago et al., 2012). Additionally, it is worth noting that there was a discrepancy between the magnification of particle number concentrations, surface areas, and SO₂ concentrations. On the one hand, there might be some particles, especially nanoclusters, lost to the chamber wall and not detected; on the other hand, the initial size of nanoclusters contributed from gaseous H₂SO₄ was small (sub-3 nm) (Chu et al., 2019; Sipilä et al., 2010) and could not be detected by our general SMPS. That is to say that the particle number concentrations and surface areas measured by our SMPS might be the particles after growing by collision. This could be supported by the enhancement in the particle diameters (144–172 nm) and sulfate concentrations (13–38 µg m⁻³) in the presence of SO₂. After considering the underestimation of particle formation (factor of 1.97–2.82; Sect. 2.3), the sulfate concentrations will be enhanced by a factor of 5.8 when comparing experiments SGN 1 and SGN 4.

In order to further investigate the role of SO₂ in the chemistry of SOA formation, the particle acidities were estimated using the E-AIM model (Model II: H⁺ – ${\mathrm{NH}}_{\mathrm{4}}^{+}$ – ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ – ${\mathrm{NO}}_{\mathrm{3}}^{-}$ – H₂O) (Clegg and Brimblecombe, 2005; Wexler and Clegg, 2002; Clegg et al., 1998). The concentrations of chemical components (i.e., ${\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, and ${\mathrm{NO}}_{\mathrm{3}}^{-}$) at the time when the SOA formation rate reached its peak were used as the inputs of the model. As shown in Fig. 4, the H⁺ concentration was increased from 8.5 to 32.5 nmol m⁻³ with the increase in SO₂ concentration under moderate humidity conditions (RH = 50 %), and the higher SOA concentration and SOA yield could be well explained by the enhancement of the particle acidities (R²=0.960 and R²=0.986, respectively). The higher SOA concentration and SOA yield were related to the acid-catalyzed reactions of multifunctional aldehydes (e.g., glyoxal and methylglyoxal), which were the products of aromatic hydrocarbons in the gasoline vapor through gas-phase photo-oxidation. Hemiacetals, acetals, and alcohols could be generated through the acid-catalyzed heterogeneous reactions of glyoxal (Czoschke et al., 2003; Jang et al., 2002). These low-vapor-pressure products generated from heterogeneous reactions preferentially contribute to the SOA formation (Kroll and Seinfeld, 2008; Cao and Jang, 2007; Casale et al., 2007; Jang et al., 2002).

Figure 4 Relationship between SOA concentration (left y axis), corrected SOA yield (right y axis), and H⁺ concentration, which was used to characterize the particle acidities. The H⁺ concentration presented in this plot was the value when the SOA formation rate reached the peak during each experiment (Exps. SGN1, SGN2, SGN3, and SGN4).

In addition, the sulfur-containing organics formed in the presence of SO₂ might be another reason for the increase in SOA yield (Kundu et al., 2013; Liggio et al., 2005). Jaoui et al. (2008) have reported that the acidic aerosol generated in the presence of SO₂ could lead to sulfur-incorporating reactions in the particle phase during the photo-oxidation of α-pinene ∕ toluene ∕ NO_x mixtures. Sulfur-containing organics could be generated via reactions of organic species (e.g., polycyclic aromatic hydrocarbons (PAHs), C₁₀–C₁₂ alkanes, alcohols, epoxides) with sulfate, bisulfate, or sulfuric acid, especially under high relative humidity and acidity conditions (Riva et al., 2015, 2016; Huang et al., 2015; Hatch et al., 2011; Surratt et al., 2007; Liggio et al., 2005). Huang et al. (2015) have revealed that sulfur-containing organics with R–O–${\mathrm{SO}}_{\mathrm{3}}^{-}$ functional groups will yield S-bearing organic fragments (C_xH_yO_zS) during ionization, which subsequently could be detected by HR-ToF-AMS and used as marker ions to quantify them. In our gasoline ∕ NO_x experiments in the presence of SO₂, the ions CSO⁺, ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{2}}^{+}$, and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{3}}^{+}$ could be separated (Fig. S7), although uncertainty might be induced in the peak fitting of the highly abundant ions ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{4}}{\mathrm{O}}_{\mathrm{2}}^{+}$, ${\mathrm{C}}_{\mathrm{6}}{\mathrm{H}}_{\mathrm{7}}^{+}$, and ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}^{+}$. These characteristic ions (i.e., CSO⁺, ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{2}}^{+}$, and ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{3}}^{+}$) have also been observed from sulfur-containing organics in previous field measurements (Huang et al., 2015; Farmer et al., 2010). According to the estimation method for sulfur-containing organics mentioned in Huang et al. (2015), we found that the signal of these ions and the concentrations of sulfur-containing organics increased with the SO₂ initial concentration (Fig. 5). The conservative lower-bound-estimated concentrations of sulfur-containing organics (13–26 ng m⁻³) were comparable to those (∼20 ng m⁻³) observed in the mid-Atlantic United States, which were derived from biogenic and anthropogenic hydrocarbons (Meade et al., 2016). Additionally, it should be noted that the sulfur-containing organics concentration in this study might be underestimated by the HR-ToF-AMS when considering that one cannot resolve all the sulfur-containing fragments that may exist, and some of the sulfur-containing organics might fragment into masses that do not contain sulfur and are thus quantified as organic. Furthermore, the relative ionization efficiency (RIE) for the sulfur-containing organics fragments was assumed to be equivalent to the remainder of the organics (1.3), since the RIE value for sulfur-containing organics is unknown. This may introduce an additional uncertainty to the quantitation of sulfur-containing organics. Therefore, photo-oxidation of gasoline vapor in the presence of SO₂ might be a noteworthy source of sulfur-containing organics, although the concentration was low compared to that of generated ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ (∼0.1 % of ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$).

Figure 5 Signal of fitted peaks, i.e., CSO⁺, ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{2}}^{+}$, ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{SO}}_{\mathrm{3}}^{+}$ (right y axis), and sulfur-containing organics concentration (left y axis), as a function of SO₂ initial concentration.

3.3 Role of NH₃ in secondary aerosol formation

Similarly, the role of NH₃ in SA formation was examined. It is worth noting that ammonium aerosols were formed without the addition of gaseous NH₃ (Fig. S8 in the Supplement), which signified that some NH₃ was present in the background air in the chamber or introduced during the humidification process of the chamber (Y. Liu et al., 2015). Unfortunately, appropriate instruments are unavailable to measure the exact concentration of background NH₃ in the chamber. According to the concentration of generated ammonium aerosols, the concentration of background NH₃ was estimated to be ∼15 ppb using the E-AIM model (Clegg and Brimblecombe, 2005; Wexler and Clegg, 2002; Clegg et al., 1998). Therefore, for the experiments with the presence of NH₃, the concentration of injected NH₃ (150–200 ppb) was much higher than this value to identify the effect of NH₃ on SA formation. The SA concentration was enhanced by a factor of 2.0–2.5 in the presence of NH₃, as shown in Fig. S9a. The formation of SOA, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ was enhanced to varying degrees. The increase in ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ could be attributed to the formation of inorganic NH₄NO₃ in the presence of NH₃. The ${\mathrm{NO}}^{+}/{\mathrm{NO}}_{\mathrm{2}}^{+}$ ratio, which could be derived from HR-ToF-AMS, has often been used as a proxy for the identification of inorganic nitrate and organic nitrogen compounds (Farmer et al., 2010; Sato et al., 2010; Rollins et al., 2009). Generally, the ${\mathrm{NO}}^{+}/{\mathrm{NO}}_{\mathrm{2}}^{+}$ ratio of inorganic nitrate (1.08–2.81) is lower than that of organic nitrogen compounds (3.82–5.84) (Liu et al., 2016). In this study, the ${\mathrm{NO}}^{+}/{\mathrm{NO}}_{\mathrm{2}}^{+}$ ratio became substantially lower (∼2.00) in the presence of NH₃ compared with that in the absence of NH₃ (∼5.46). Therefore, NH₄NO₃ was the dominant nitrate species in the presence of NH₃. As for the reason for SOA enhancement, the NH₃ present could react with some organic acids and subsequently contribute to SOA formation (Na et al., 2007, 2006), which could be supported by the increase in N∕C (from 0.016 to 0.033) with an increasing NH₃ concentration at similar concentrations of NO_x. In addition, we have found that the presence of NH₃ readily increased the particle diameter and number concentration of SA generated in the photo-oxidation of gasoline (Fig. S9b and c), which revealed that NH₃ played an important role in new particle formation (NPF). These are consistent with simulation results finding that NH₃ promotes atmospheric NPF and also the conversion of SO₂ and NO₂ (Jiang and Xia, 2017). The increased surface area of particles was also observed (Table 1; 2.07×10³ and 2.48×10³ µm² cm⁻³) as the NH₃ concentration increased from 0 to 150 and 200 ppb. Similarly, the larger surface area would favor the partitioning of low-volatility vapor to the particle phase, leading to the higher SA yield (Table 1).

Previous studies have reported that the reaction of carbonyl compounds (e.g., glyoxal) could be catalyzed by ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ions through a Bronsted acid pathway or an iminium pathway, which could generate N-containing products and oligomers (Nozière et al., 2009), and then contribute a substantial fraction to SOA (Y. Liu et al., 2015; Farmer et al., 2010; Cheng et al., 2006). Researchers have identified the characteristic fragments of nitrogen-containing organics as C_xH_yN_n and C_xH_yO_zN_n using HR-ToF-AMS (Lee et al., 2013; Farmer et al., 2010; Galloway et al., 2009). In this study, the typical normalized mass spectra of N-containing fragments in SOA after 480 min of photo-oxidation reaction at different concentrations of NH₃ are given in Fig. 6. The prominent peaks in the C_xH_yN_n family were at m∕z 27 (CHN⁺), 30 (CH₄N⁺), 40 (C₂H₂N⁺), 41 (${\mathrm{CHN}}_{\mathrm{2}}^{+}$, C₂H₃N⁺), 42 (C₂H₄N⁺), 43 (C₂H₅N⁺), 54 (${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{2}}{\mathrm{N}}_{\mathrm{2}}^{+}$, C₃H₄N⁺), 55 (C₃H₅N⁺), and 68 (${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{4}}{\mathrm{N}}_{\mathrm{2}}^{+}$, C₄H₆N⁺); the C_xH_yO_zN_n fragments were dominated by 45 (CH₃ON⁺), 46 (CH₄ON⁺), 59 (C₂H₅ON⁺), 63 (CH₅O₂N⁺), 73 (${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}{\mathrm{ON}}_{\mathrm{2}}^{+}$, C₃H₇ON⁺), 86 (C₃H₄O₂N⁺, ${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}{\mathrm{ON}}_{\mathrm{2}}^{+}$), 91 (C₃H₉O₂N⁺), 97 (${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{5}}{\mathrm{ON}}_{\mathrm{2}}^{+}$), and 104 (C₃H₆O₃N⁺, C₄H₁₀O₂N⁺). The N-containing fragments observed in the experiment without added NH₃ could be attributed to the reactions between organic peroxy (RO₂) radicals and NO_x (Arey et al., 2001) or the uptake of background NH₃ by SOA. Additionally, it was obvious that the signal intensities of most N-containing fragments became significantly stronger as the NH₃ concentration increased (150–200 ppb). Therefore, a considerable amount of nitrogen-containing organics (the ratio of nitrogen-containing organics to SOA was about 6.7 %–7.7 %) was formed during the photo-oxidation of gasoline vapor in the presence of NH₃. This was consistent with the previous study conducted by Y. Liu et al. (2015), who observed the formation of organic nitrogen compounds in the SOA generated from the OH oxidation of m-xylene. The promoting role of NH₃ in the formation of N-containing species was also observed in the reaction system of ozonolysis and the photo-oxidation of α-pinene (Babar et al., 2017).

Figure 6 Typical normalized mass spectra of N-containing fragments in SOA formed from the photo-oxidation of gasoline vapor at different concentrations of NH₃ (Exps. GN, AGN1, and AGN2).

In addition, elemental analysis was also carried out to elucidate the SOA chemical composition and SOA formation mechanisms (Chhabra et al., 2011; Heald et al., 2010) at different concentrations of NH₃. The time evolution of H∕C and O∕C in SOA formed from the photo-oxidation of gasoline vapor at different concentrations of NH₃ is shown in Fig. 7. As evident from Fig. 7, all data points are located in the triangular area for a slope between −1 and 0, which suggests that SOA formation from the photo-oxidation of gasoline vapor is a combination of carboxylic acid and alcohol–peroxide (Heald et al., 2010). Moreover, in the presence of NH₃, as shown in Fig. 8, N∕C increased as the reaction proceeded in the initial oxidation stage (0–120 min), accompanied by a rapid increase in O∕C (0.12–0.67), a decrease in H∕C (2.12–1.61), and rapid formation of SOA. During this stage, the photo-oxidation of VOC precursors leads to a rapid increase in O∕C and a rapid decrease in H∕C. The termination chemistry of NO_x with free radicals and the NH₃ uptake result in a rapid increase in N∕C. As the reaction proceeded further (120–300 min), an increase in H∕C, which should be caused by NH₃ uptake, resulted in an almost constant oxidation state of SOA in continuous photo-oxidation, accompanied by an increase in the SOA concentration. Nozière et al. (2009) have reported that N-containing products are generated from carbonyl compound (e.g., glyoxal) self-reactions catalyzed by ${\mathrm{NH}}_{\mathrm{4}}^{+}$ ions, which will have a dramatic impact on the volatility of oxidation products and the yield of SOA (Ortiz-Montalvo et al., 2014). In the last stage of the reaction (360–480 min), NH₃ uptake might reach saturation; therefore, H∕C and N∕C are almost constant. Comparing experiments with different concentrations of NH₃, the average H∕C shows an obvious increase (1.53–1.70), while the average O∕C (0.70–0.78) shows a slight increase with the increase in NH₃ concentration (0–200 ppb), as seen in Fig. S10. The slope in the van Krevelen diagram shows a trend from slope $=-\mathrm{1}$ to slope =0 (Fig. S10), indicating that the formed carboxylic acid could further react with NH₃ via an acid–base reaction to generate an ammonium salt of a carboxylate anion in the presence of NH₃ (Na et al., 2007). Xu et al. (2018) recently found that imidazole products containing multiple oxygen atoms could be generated through heterogeneous reactions between NH₃ and carbonyl compounds (e.g., glyoxal), which might also contribute to the increase in the O∕C of the SOA.

Figure 7 Time evolution of H∕C and O∕C in SOA formed from the photo-oxidation of gasoline vapor at different concentrations of NH₃ (Exp. GN, AGN1, and AGN2). The numbers (i.e., −1, −0.5, 0, 0.5, and 1) labeling the dashed lines show the average carbon oxidation state ($\mathrm{OSc}=\mathrm{2}\times \mathrm{O}/\mathrm{C}-\mathrm{H}/\mathrm{C}$) (Kroll et al., 2011). The black lines represent the addition of functional groups to an aliphatic carbon (Heald et al., 2010).

Figure 8 Time evolution of (a) O∕C, H∕C, N∕C, and (b) SOA concentration in the photo-oxidation of gasoline vapor in the presence of 150 ppb NH₃ (Exp. AGN1).

3.4 Different roles of SO₂ and NH₃ in SOA chemical properties

The chemical properties of the SOA generated under different concentrations of SO₂ or NH₃ were further compared by applying positive matrix factorization (PMF) analysis to the HR-ToF-AMS data (Chu et al., 2016; Liu et al., 2014). The details of the PMF analysis are given in the Supplement. For the experiments under different SO₂ concentration conditions (i.e., Exps. GN, SGN1, SGN2, SGN3, and SGN4), two factors (Factor 1-S and Factor 2-S; Fig. S11a) were identified from the PMF analysis, and the different mass spectra (m∕z 12–170) between the two factors and the time series of the mass concentrations are shown in Fig. 9. The intensity of C_xH_y and S-bearing organic fragments (C_xH_yO_zS) in Factor 1-S was obviously stronger than that in Factor 2-S. Meanwhile, fragments in the high m∕z range (>110 Da) were more abundant in Factor 1-S (Fig. 9a, marked by the red box). By contrast, the fragments containing oxygen in Factor 2-S were more abundant than in Factor 1-S, such as the typical fragment ${\mathrm{CO}}_{\mathrm{2}}^{+}$ (m∕z 44). Therefore, Factor 1-S was tentatively assigned to the less-oxygenated organic aerosol and oligomers, while Factor 2-S was more-oxygenated organic aerosol (Ulbrich et al., 2009). Similarly, for the experiments at different NH₃ concentrations (i.e., Exps. GN, AGN1, and AGN2), two factors (Factor 1-N and Factor 2-N; Fig. S11b) were also identified in the same way. According to Fig. 10, Factor 1-N was tentatively assigned to the less-oxygenated organic aerosol and oligomers, while Factor 2-N was more-oxygenated organic aerosol and nitrogen-containing organics.

Figure 9 (a) Different mass spectra (Factor 1-S–Factor 2-S) between the two factors, (b) time series of the mass concentration, and (c) relationship between the concentration of SO₂ and the maximum concentration of the two factors identified by applying PMF analysis to the AMS data derived from the experiments at different concentrations of SO₂ (Exps. GN, SGN1, SGN2, SGN3, and SGN4).

Figure 10 (a) Different mass spectra (Factor 1-N–Factor 2-N) between the two factors, (b) time series of the mass concentration, and (c) relationship between the concentration of NH₃ and the maximum concentration of the two factors identified by applying PMF analysis to the AMS data derived from the experiments at different concentrations of NH₃ (Exps. GN, AGN1, and AGN2).

As shown in Figs. 9b and 10b, these two factors both had different time series during the entire reaction. With respect to Exps. GN, SGN1, SGN2, SGN3, and SGN4, Factor 1-S was formed later (∼30 min) than Factor 2-S and then continuously increased during the entire reaction. Comparing experiments with different SO₂ concentrations, the maximum concentration of Factor 1-S, which was related to the less-oxygenated organic aerosol and oligomers, was enhanced with an increased SO₂ concentration (R²=0.881; Fig. 9c). This suggested that the presence of SO₂ was prone to decrease the oxidation state of organic aerosol via acid-catalyzed reactions and enhance the formation of oligomers (Liu et al., 2016), which was consistent with the evolution of O∕C vs. H∕C shown in Fig. S12. Moreover, the gradually increasing concentration of Factor 1-S was related to the formation of sulfur-containing organics in the presence of SO₂ (Blair et al., 2017). By contrast, Factor 2-S was first gradually increased with the progress of the reaction and then decreased after reaching a peak (i.e., inflection point). The time to reach the inflection point was affected by the SO₂ concentration (Fig. 9b). As the initial concentration of SO₂ increased from 0 to 151 ppb, the time corresponding to the inflection point decreased, which indicated that the adverse influence of acid catalysis on Factor 2-S was gradually enhanced. In addition, the maximum concentration of Factor 2-S was negatively related to the SO₂ concentration (R²=0.987; Fig. 9c); this suggested that the presence of SO₂ and acid catalysis was adverse to the formation of more-oxygenated organic aerosol, leading to the decrease in the oxidation state of organic aerosol (Fig. S12).

By contrast, for Exps. GN, AGN1, and AGN2, Factor 1-N was first increased with the progress of the reaction and then gradually decreased after reaching a peak (Fig. 10b), while Factor 2-N was formed later (∼30 min) than Factor 1-N and then continuously increased during the entire reaction. This phenomenon was consistent with the expected behavior that less-oxidized organic aerosol would be further oxidized to form more-oxidized organic aerosol. When comparing experiments with different NH₃ concentrations, it was observed that the concentration of Factor 2-N increased with an increasing NH₃ concentration. Meanwhile, Factor 2-N, which was related to the more-oxidized organic aerosol and nitrogen-containing organics, was a dominant factor in the presence of NH₃, and its maximum concentration was enhanced with the increase in NH₃ concentration (R²=0.988; Fig. 10c). Thus, the formation of more-oxygenated organic aerosol and nitrogen-containing organics will be enhanced with an increase in NH₃ concentration. In contrast, a negative correlation was observed between the maximum concentration of Factor 1-N and NH₃ concentration (R²=0.876; Fig. 10c); this revealed that less-oxygenated organic aerosol was gradually transformed to more-oxidized species and nitrogen-containing organics in the presence of NH₃.

4 Conclusions

In this study, SA formation from the photo-oxidation of gasoline ∕ NO_x in the presence of SO₂ or NH₃ was investigated. Our experimental results demonstrated that SA was enhanced by a factor of 1.6–2.6 or 2.0–2.5, respectively, with the increase in SO₂ or NH₃ concentration (0–151 and 0–200 ppb, respectively). Meanwhile, both secondary organic aerosol (SOA) and secondary inorganic aerosol (SIA) were increased by varying degrees. In the presence of SO₂, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ was the most sensitive linear increase with the increase in SO₂ concentration, and SOA was also greatly enhanced due to the acid catalytic effect and the formation of sulfur-containing organics. In the presence of NH₃, NH₄NO₃ was most enhanced, following by SOA. The formation of nitrogen-containing organics was also promoted by the presence of NH₃. Meanwhile, conspicuous new particle formation (NPF) and particle size growth were enhanced in the presence of SO₂ or NH₃.

In this study, a linear relationship between the SA yield and SO₂ or NH₃ concentration was also obtained (Fig. S13). Considering the typical concentrations of SO₂ and NH₃ of 40 and 23 ppb in haze pollution in the North China Plain (Cheng et al., 2016) and the lower aromatics content (∼10 %) in vehicular evaporative emissions (Zhang et al., 2013), the SA yield is roughly estimated to be about 0.20. Recently, an updated emission inventory of vehicular evaporative emissions was reported to be 1.65 Tg yr⁻¹ (H. Liu et al., 2017). Then, the SA formed from the photo-oxidation of VOCs emitted by vehicular evaporation in the presence of SO₂ and NH₃ is roughly estimated to be 0.33 Tg yr⁻¹, which is about 1.5 times as much as the primary PM_2.5 emissions from transportation (0.21 Tg yr⁻¹) in China (Jing et al., 2015; Zhang et al., 2007), accounting for about 21 % of SOA production (1.6 Tg yr⁻¹) from anthropogenic precursors estimated by global chemical transport models (Farina et al., 2010). In addition, the photo-oxidation of long-chain alkanes (>C₆, IVOCs) contained in evaporative emissions also contributes to SOA formation (Pye and Pouliot, 2012; Tkacik et al., 2012; Presto et al., 2009; Lim and Ziemann, 2005; Zhao et al., 2016). This estimate suggests that vehicular evaporative emissions will be a significant source of SA in the presence of SO₂ and NH₃, although the estimate might have a high uncertainty due to the fact that SA yield might vary considerably under different atmospheric conditions. Meanwhile, in the presence of NO_x, SO₂, and NH₃, vehicular evaporative emissions may be a potential source of sulfur- and nitrogen-containing organics according to the results obtained from our study. Sulfur- and nitrogen-containing organics will have an adverse influence on the climate by light absorption and/or by affecting aerosol hygroscopicity (Staudt et al., 2014; Nguyen et al., 2012), and they also make a significant contribution to SOA and nitrogen or sulfur budgets (Lee et al., 2016; Shang et al., 2016).

Therefore, under the compound pollution conditions of SO₂ and NH₃, the synergistic emission reduction of vehicular evaporative emissions, SO₂ (e.g., coal-fired flue gas), and NH₃ (e.g., emitted from agricultural nonpoint sources and traffic) should be taken into consideration by policy makers for future management, which will contribute to reducing the burden of PM_2.5 and then cut the environmental, economic, and health costs caused by PM pollution. Our work will provide a scientific basis for taking corresponding control measures to relieve haze events in China. Additionally, there might be some differences between the VOC composition of gasoline vapor directly injected to the smog chamber and vehicular evaporative emissions. Thus, further work should be focused on SA formation directly from vehicular evaporative emissions under coexisting SO₂ and NH₃ conditions to shed light on the formation mechanism of SA under more atmospherically relevant conditions.