Effects of SO2 on optical properties of secondary organic aerosol generated from photooxidation of toluene under different relative humidity

Abstract. Secondary organic aerosol (SOA) have great impacts on air quality, climate change and human health. The composition and physicochemical properties of SOA differ a lot for they originated under different atmospheric conditions and from various precursors and oxidations. In this work, photooxidation experiments of toluene were performed under four conditions (dry, dry with SO2, wet, and wet with SO2) to investigate the effect of SO2 under different relative humidity on the composition and optical properties of SOA at the wavelength of 375 nm and 532 nm. According to our results, the increase of humidity enhances not only light absorption, but also scattering property of SOA. Highly conjugated oligomers formed through multiphase reaction might be the reasons of this phenomenon. Adding SO2 slightly lower the real part of complex refractive index (n) of SOA: ndry, SO2 



Introduction
Secondary organic aerosol (SOA) accounts for a major fraction of the atmospheric fine particulate matters (PM2. 5), and has significant impacts on air quality, climate change and human health (Seinfeld and Pandis, 2006;Jimenez et al., 2009;Hallquist et al., 2009;Huang et al., 2014). Globally, SOA could influence the radiative balance directly by scattering and absorbing solar and terrestrial radiation and 5 indirectly by impacting cloud formation and their lifetimes (IPCC, 2013;Andreae and Gelencser, 2006;Moise et al., 2015). On the urban and regional scales, SOA contributes to the degradation of visibility and adversely influences human health (Watson, 2002;Pope et al., 2002). Light-absorbing aerosols (including black carbon (BC), mineral dust, and brown carbon (BrC)) are recognized as playing important roles in climate radiative forcing because of the strong dependence of their optical properties on the 10 aerosol composition, the complexity of their production and the poor constraints on their contribution to radiative forcing (Peng et al., 2016;Wang et al., 2013). Quantifying the optical properties of SOA is one of the key problems in accessing anthropogenic pollution data for visibility, air quality, and climate change, as well as one of the most urgent issues in atmospheric sciences (Laskin et al., 2015;Moise et al., 2015;Andreae and Gelencser, 2006). 15 Aromatic compounds, accounting for 20-40% v/v of gasoline fuel, are one of the representative 3 urban SOA" (Surratt et al., 2010). M. Jaoui et al. (2008) showed that the addition of SO2 would make the colour of the SOA generated via the photooxidation of α-pinene and toluene darker and browner, and oligomers and nitrogenous organic compounds were detected in the SOA extracts (Jaoui et al., 2008). Nakayama et al. (2018) investigated the effect of SO2 on the optical properties of isoprene under various NOx concentrations and oxidation pathways (Nakayama et al., 2015;Nakayama et al., 2018). However, 5 more research is needed on the role of SO2 in the subsequent optical properties of SOA formed from various VOCs.
Field studies have shown that haze events were often accompanied by high relative humidity (RH) (Sun et al., 2016a;Sun et al., 2016b), and organic aerosols were mostly liquid (Shiraiwa et al., 2017;Liu et al., 2017). Multiphase reactions also play an important role in SOA formation, as shown in previous 10 studies. For example, Jia et al. (2018) have shown that the yield of toluene SOA almost doubled at relative humidity of 85 % compared to dry conditions (Jia and Xu, 2018). Li et al. (2017) found that multiphase reactions would enhance light scattering and radiative forcing of the m-xylene SOA (Li et al., 2017c).
Understanding the impact of complicated air conditions on SOA generation, especially the effect of polluting gases (e.g., SO2) and phase state, is urgently needed for forecasting, assessing, and controlling 15 air pollution.
In this work, we investigated the effect of SO2 on the optical properties and chemical composition of the SOA derived from toluene under different humidity. The results will greatly help the evaluation of the toluene-derived SOA on atmospheric visibility and climate change under complicated pollution conditions. More importantly, the data in the current study will be highly useful for the simulation of 20 models and field observations performed under various pollution conditions.

Smog chamber experiments and online measurements
Experiments were performed in a 5 m 3 dual-reactor smog chamber, the details of which were described elsewhere Li et al., 2017c;Li et al., 2017b). Briefly, the reactors were placed in a 25 thermally isolated enclosure in which the temperature and RH could be well controlled by blowers and air conditioners. The variation range of temperature was 0.5 °C, and the limit for variability of RH was below 5%. Experiments were performed using multiple UV (ultraviolet) light sources, which resulted in a similar spectrum as that of solar radiation . During the experiments, the concentrations of NO, NO2 and SO2 were measured by corresponding gas analysers (Teledyne API  30 T200UP, and Thermo 48i). The concentrations of toluene were measured by a proton transfer reaction quadrupole mass spectrometry in H3O + mode (PTR-QMS, Ionicon). The size distributions, number concentrations and mass concentrations of the SOA were determined by a scanning mobility particle sizer (SMPS), which consisted of an electrostatic classifier (EC, TSI 3080), a differential mobility analyser (DMA, TSI 3081) and a condensation particle counter (CPC, TSI 3776). A density of 1.4 g· cm -35 4 The extinction coefficients at 532 nm were detected with custom-built cavity ring-down spectrometers (CRDS), the details of which were given previously (Wang et al., 2012;Li et al., 2014;Li et al., 2017b).
The scattering, absorption, and extinction coefficients at 375 nm were measured by a photoacoustic extinctiometer (PAX-375, Droplet Measurement Technologies).

Experimental conditions and off-line measurements 5
As shown in Table 1, four sets of experiments were conducted: D, dry condition (RH < 5%, SO2 <1 ppb); DS, dry condition with SO2 (RH < 5%, SO2 = 30-50 ppb); W, wet condition (RH > 80%, SO2 < 1 ppb); and WS, wet condition with SO2 (RH > 80%, SO2 = 30-50 ppb). To ensure that the results were reproducible, each experiment was performed twice. Before each experiment, the chamber was flushed at least three times using zero air (AADCO 737-15) to achieve clean conditions in which number 10 concentration of particles lower than 50 cm -3 and concentrations of NOx and SO2 lower than 1 ppb. After preparation, a known volume of organic precursor was added to the chamber by a glass U-tube through zero air. HONO was then added into the chamber by bubbling a small flow of zero air through the immediately mixed solution of 1 mL of 1 wt.% NaNO2 and 2 mL of 10 wt.% H2SO4. The by-products NO and NO2 were also introduced into the chamber. The concentrations of toluene in the chamber were 15 approximately 180 ppb, while the total NOx levels were approximately 500 ppb. SO2 was introduced into the chamber from a 29.6 ppm gas cylinder. For experiments under wet condition, zero air was passed through a fuel cell humidifier (FC-200-780-7MP, Perma Pure), and the RH was above 80% before the experiments started. It is important to highlight that dew drops did not appear in the chamber during the experiments due to accurate control of the chamber temperature (± 0.5 °C). The contents in the chamber 20 were mixed for approximately 10 minutes, and then the UV lamps were turned on to initiate the photoreaction. When the SOA mass concentration stopped rising, the lights were switched off to stop the reaction. All experiments were conducted at 25.0 ± 0.5 °C.
After each experiment, the SOA was collected by PTFE filters (0.2 μm, 47 mm, MILLIPORE FELP) and then extracted in 5 mL methanol in an ultrasonic bath for 30 minutes for mass spectrometry analysis and 25 in 10 mL for UV-Vis absorption analysis. The methanol solutions were analysed by electrospray ionization time-of-flight mass spectrometry (ESI-TOF-MS) in positive and negative mode. A typical mass-resolving power of > 30000 was achieved at m/z 200, with an absolute mass error below 3 ppm.
Each sample extract was measured in triplicate with an injection volume of 10 μL. The ions detected in the filter blank were subtracted, and molecular formulas were assigned to all ions in the samples having 30 a signal-to-noise ratios greater than 10. The settings for ESI positive mode were as follows: capillary voltage, 3 kV; nebulizer gas pressure, 0.4 bar; dry gas flow rate, 3 litres per minute; dry gas temperature, 180 °C; and mass detection range, m/z 100-1000. For negative mode, the capillary voltage was set to 2.5 kV, the nebulizer gas pressure was 0.3 bar, the dry gas flow rate was 4 litres per minute, the dry gas temperature was 150 °C, and the mass detection ranged from m/z 200 to 1000. 35 The following chemicals were used without further purification: toluene (≥99.5%) [

Calculation of the complex refractive indexes
The complex refractive index (RI, RI = n + ki) is the only intrinsic optical property of particles, including the real part of complex refractive index (RI(n), standing for scattering) and the imaginary part of complex refractive index (RI(k), standing for absorption) (Bohren and Huffman, 1983;Bond and Bergstrom, 2006).The RI calculation method used in this study has been applied and approved in 5 previous studies (Li et al., 2014;Li et al., 2017b;Li et al., 2017c;Li et al., 2018b;Peng et al., 2018;Li et al., 2018a). Briefly, the extinction coefficients (αext) at 532 nm measured by CRDS and the scattering, absorption, and extinction coefficients at 375 nm measured by PAX-375 were calculated on the basis of Mie theory. For monodispersed spherical particles, αext can be represented by: where σext is the extinction cross section, N is the number concentration of particles, D is the diameter of 10 particles, and Qext is the extinction efficiency, which is the ratio of the extinction cross section to the geometric area of the particles.
For polydispersed particles with a log-normal size distribution and a geometric standard deviation (σg) smaller than 1.5, αext can be expressed as follows: where Stot is the total surface concentration. A hypothesis was made for simplifying data processing: the 15 Qext value of the polydispersed particles in the whole size distribution range was the same as the Qext value of particles with the surface mean diameter (Dsm). The Stot and Dsm were given by SMPS directly, of which the uncertainties were ± 1% and ± 5%, respectively.
The value of Qext of particles with a given Dsm were subsequently calculated through the Mie program.
The best-fit complex index was determined by minimizing the following reduced merit function (χr): 20

Calculation of RI of products
The RI(n) of the products were calculated by the quantitative structure-property relationship, developed by Redmond and Thomson (Redmond and Thompson, 2011). In brief, the RI(n) values were estimated by polarizability (α), degree of unsaturation (μ), mass density (ρm) and molecular weight (M) via the application of equation (4). 25 ( ) = 0.031717( ) + 0.0006087( ) − 3.0227 ( ) + 1.38708 The mass density of each molecule was estimated by the E-AIM model (Extended Aerosol Inorganics Model, http://www.aim.env.uea.ac.uk/aim/density/density.php), which had high accuracy for 166 organic compounds (Girolami, 1994). It should be noted that the model should be used at 550 nm. It would cause only small, acceptable errors for its use at 532 nm. This method was employed in our previous studies as well (Li et al., 2017b;Li et al., 2018b;Li et al., 2017c;Peng et al., 2018). Similar to 6 previous studies, negligible light absorption was found for toluene SOA at wavelength > 500 nm, as shown in Figure S3 as well, so only the real part of RI is considered at 532 nm (Li et al., 2017b;Li et al., 2018b;Li et al., 2014). For brevity, at 532 nm, "RI" represents the real part of the complex refractive index, RI(n), in the following text.

General results of the experiments
As shown in Figure 1, particles were generated only several minutes after the lamps turned on. Although the concentrations of NO differed under different conditions, the maximum concentrations of NOx were approximately the same, ~500 ppb. The consumption of the organic precursor was similar in all experiments (~100 ppb), indicating similar OH exposures and oxidation levels. As shown in Figure S1, 10 the maximum total number concentrations of the toluene SOA were almost doubled under conditions with SO2, which implied that the addition of SO2 could promote new particle formation. At the end of the experiments, the Stot of particles were approximately 2.9×10 9 , 4.3×10 9 , 3.8×10 9 , 6. As shown in Figure S3, the absorption of toluene-derived SOA under the D condition at 375 nm was approximately 0.002 a.u., from which we could calculate that RI(k) under the D condition as approximately 0.001 (Sun et al., 2007). PAX-375 showed no absorption of the toluene-derived SOA 5 under dry condition, in other words, the absorption of SOA under dry condition was below the detection limit, and the calculated RI(k) was approximately 0.0009. Under this circumstance, RI(k) was set to zero (±0.001), which was lower than other studies. Table S1  Tables 2, S2 and S3), and some of them contained organonitrogen groups, which could be seen more clearly in negative mode. However, previous studies mainly focused on a lower mass range (< 200 Da) (Hinks et al., 2018;Forstner et al., 1997;Staudt et al., 2014;Wang et al., 2018;Birdsall et al., 2010), e.g., 35 Birdsall et. al. (2010 detected benzaldehyde, cresol, phenol, and butenedial in the photooxidation process of toluene under both low-NOx and high-NOx conditions (Birdsall et al., 2010). In this work, we analysed the products around m/z 200 400 and discussed the effect of SO2 under different humidity. The representative identified molecular weight, molecular formula, and calculated RI(n) values of toluenederived SOA are shown in Figure 5 and Table 2. More details are given in Table S2 and Table S3.

Effects of high humidity
High relative humidity causes a substantial increase in the RI values for toluene-derived SOA (Figure 2 and Figure 3). The complex refractive indexes increased from 1.45 to 1.566 at 375 nm and 1.412 to 1.504 at 532 nm in absence of SO2. Previous studies reported similar findings, e.g., Li et al. (2017)   To clearly see the change of chemical compositions under the D and DS conditions, the subtraction plots of mass spectra are shown in Figure S6. As we can see in Figure S6 (a), adding SO2 caused the relative intensities of products with lower molecular weight (< 200 Da) to increase. These products are mainly alcohols and esters with smaller molecular weight and RI(n) values of which are lower than 1.4. The reason for this is most likely that the addition of SO2 caused large amounts of new particle formation and 35 high particle number concentrations in the system (Chu et al., 2015;Chu et al., 2016;Deng et al., 2017;Liu et al., 2018a), which could provide a larger mass concentration and adsorb high-volatile small molecules into the particle phase (Li et al., 2018b). These small molecule products usually possess high volatility and low oxidation state, which would also reduce the oxidation state of aerosols generated by toluene. Previous studies have found similar phenomena, for example, Zhao et al. (2018) found that high SO2 concentrations decreased the ratio O/C in α-pinene or limonene systems (Zhao et al., 2018), while Liu et al. (2016) discovered that the oxidation state of carbon was -0.51±0.06 for SOA formed from light-5 duty gasoline vehicle exhaust with SO2 and -0.19±0.08 without SO2 (Liu et al., 2016b), which all implied that adding SO2 reduced the oxidation state of the resulting SOA. It should be noted that our off-line analytical method caused the loss of a large part of alcohols and esters with small molecular weights, which might overestimate the values of their RI.

Effects of SO2 under dry condition
Light absorption properties of SOA are related to its composition, the contribution of each product to 10 light absorption and so on (Laskin et al., 2015;Moise et al., 2015). Small α-dicardonyls compounds such as glyoxal and methylglyoxal are important intermediate products of toluene that undergo polymerization to produce low-volatility oligomers (Ji et al., 2017;Fu et al., 2009;Fu et al., 2008). These products might undergo particle phase reactions, e.g., acid-catalysed aldol condensation reactions under SO2 conditions, such as hydroxyls, to an acceptor group, such as a ketone or aldehyde (Phillips and Smith, 2014). For 20 organosulfate, another kind of BrC, we did not detect them under the DS condition, which is in accordance with previous studies (Staudt et al., 2014).

Effects of SO2 under wet condition
Under wet condition with SO2, the average RI(n) values of toluene-derived SOA were 1.51 at 375 nm and 1.468 at 532 nm, which were higher than those of the DS condition and lower than those of the W 25 condition (Figures 2 and 3). As for absorption, RI(k) under the WS condition was lower than the W condition, and similar to that under the DS condition. Figures 4 and S5 show the results of mass spectra difference of toluene-derived SOA under the DS or W condition minus the WS condition. Low-oxidation state organic matters and oligomers were both found in the mass spectra. Under the WS condition, relative intensities of products above 400 Da (oligomers) 30 were higher than those under the DS condition (negative values), suggesting more types of oligomers were produced. More types of low organic matters were observed compared to those under wet condition without SO2. Under these combined effects, the values of RI(n) of toluene-derived SOA were lower than those under the W condition and higher than those under the DS condition.
The RI(k) under the WS condition is almost equal to that under the DS condition, while lower than the 35 values under the W condition. Products with lower oxidation state and less conjugated oligomers caused by addition of SO2 might be the reason for this phenomenon, as proved in previous studies (Nakayama et al., 2015;Liu et al., 2016a). The concentrations of donor group under the WS condition were lower than those under the DS condition, resulting in lower concentrations of CT complexes, which might another reason of reduce of k. Organosulfate compounds were not found under the WS condition. The combined effect of SO2 and wet condition on optical properties of toluene-derived SOA are first described to the best of our knowledge, and has a significant influence on light absorption, extinction, visibility and direct radiative forcing of regional air, especially in complex polluted area. Our results 5 provide some explanations for the observed variation, and further research is needed to quantify this synergistic effect.

Atmospheric and Climate Implication
The values of mass cross section (MAC) were calculated for the DS, W and WS conditions at 375 nm, and the method was described in Supporting  under wet conditions and 0.01 g/m 2 at dry conditions at 380 nm, higher than our results, which is likely 15 due to the high ratios of NO to NO2 in their study.
The impacts of the atmospheric and climate were assessed by comparing the ratio of light extinction efficiency and simple forcing efficiency (SFE) under the four different conditions, and the results are shown in Figures 6 and S7. Aerosol sizes between 100 nm and 250 nm were circled because these are atmospherically relevant aerosols sizes Tao et al., 2017). As shown in Figure 6, 20 adding SO2 caused a light extinction efficiency reduction of approximately 16%-35% with an average of 25% at 532 nm, and humid conditions enhanced the light extinction efficiency by approximately 36%-64%, with an average of 50%. For the comprehensive impact made by SO2 and high humidity, the light extinction efficiency increased about 16%-47% with an average of 30%. These results confirm that SOA generated under synergistic pollution conditions, conditions that contained high emissions of primary 25 particles and gaseous precursors from multiple sources, efficient secondary matter formation, as well as adverse meteorological and climate conditions and regional transport, might have a greater impact on the visibility reduction, atmospheric photochemical reactions and secondary species formation (Dickerson et al., 1997;An et al., 2019).
We estimate the clear-sky direct radiative forcing per unit optical depth with the help of the SFE concept, 30 as mentioned in the Supporting Information. Figure S7 shows the SFE of toluene-derived SOA under the D, DS, W and WS conditions at 375 nm and 532 nm, which could also reflect the change of direct radiative forcing (DRF). The forcing efficiency crossed over from negative (warming)  and 30% compared to that under under the D condition, respectively, which is a similar trend to that of the extinction efficiency. The SFE values in our system coincided with those of organic aerosols from wood combustion and the burning of boreal peatlands (Chen and Bond, 2010;Chakrabarty et al., 2016).
Our model does not include hygroscopicity and other factors, which would increase particle size and 5 negative forcing. In this situation, our forcing is much lower than those in global climate models (Schulz et al., 2006). The combined effects of SO2 and humidity should be considered in the modified climate model.

Conclusion
The effect of SO2 under different humidity on the optical properties of SOA photooxidized by toluene 10 was investigated in this study. The results show that for the experimental system with RH greater than 80%, as expected, the increase in humidity greatly enhanced the real part of RI, from 1.412 to 1.504 at 532nm and from 1.45 to 1.566 at 375 nm, the imaginary part of RI was enhanced as well, which is probably because of the oligomers formation from multiphase reactions. Adding SO2 can reduce the RI(n) values of toluene-derived SOA at 375 nm and 532 nm, whether under low or high humidity. The RI 15 values of toluene-derived SOA produced under dry conditions with SO2 are 1.37+0.014i at 375 nm and 1.348 at 532 nm, while the RI values under dry and SO2-free conditions are 1.412 at 532 nm and 1.45 at 375 nm. The reason for this phenomenon might be that adding SO2 caused large amounts of new particle formation and high particle mass concentrations in the system, which could have absorbed high-volatile molecules into the particle phase. High-volatile molecules produce lower oxidation state and lower RI(n) 20 values, resulting in the decrease of RI(n) for toluene-derived SOA. The increase in RI(k) is probably related to acid-catalysed reactions on acidic particles. For the experimental system under high humidity condition with sulfur dioxide, the RI(n) was higher than SOA derived from dry condition without SO2.
The RI(k) under this condition are lower than those under wet condition without SO2 because fewer oligomers formed. The extinction properties under the WS condition are approximately 30% higher than 25 under the D condition. The results here highlighted that the combined effect of SO2 and high humidity could greatly enhance the refractive index, light scattering, and direct radiative forcing of toluene-derived SOA and, potentially, others. These results will improve our understanding of SOA optical properties, especially under complex atmospheric conditions.