Formation of organic sulfur compounds through SO₂-initiated photochemistry of PAHs and dimethylsulfoxide at the air-water interface

The presence of organic sulfur compounds (OS) at the water surface acting as organic surfactants, may influence the air-water interaction and contribute to new particle formation in the atmosphere. However, the impact of ubiquitous anthropogenic pollutant emissions, such as SO₂ and polycyclic aromatic hydrocarbons (PAHs) on the formation of OS at the air-water interface still remains unknown. Here, we observe large amounts of OS formation in the presence of SO₂, upon irradiation of aqueous solutions containing typical PAHs, such as pyrene (PYR), fluoranthene (FLA), and phenanthrene (PHE) as well as dimethylsulfoxide (DMSO). We observe rapid formation of several gaseous OSs from light-induced heterogeneous reactions of SO₂ with either DMSO or a mixture of PAHs and DMSO (PAHs/DMSO), and some of these OSs (e.g. methanesulfonic acid) are well established secondary organic aerosol (SOA) precursors. A myriad of OSs and unsaturated compounds are produced and detected in the aqueous phase. The tentative reaction pathways are supported by theoretical calculations of the Gibbs energy of reactions. Our findings provide new insights into potential sources and formation pathways of OSs occurring at the water (sea, lake, river) surface, that should be considered in future model studies for a better representation of the air-water interaction and SOA formation processes.

1 Introduction

Organic sulfur compounds (OSs) are ubiquitous in atmospheric aerosols, and organosulfates are considered as important tracers of secondary organic aerosols (SOA). Based on the occurrence of hydrophilic and hydrophobic moieties in the same molecule, OSs are surface-active compounds that cause reduction of surface tension and enhance the formation potential of cloud condensation nuclei (CCN) in aerosol particles (Bruggemann et al., 2020).

Both biogenic and anthropogenic sources, such as biomass and fossil fuel burning release OSs into the atmosphere (Bruggemann et al., 2020). The contribution of aromatic organosulfates could be up to two thirds of the sum of the identified OSs (Riva et al., 2015, 2016), and their input is highest during winter (Ma et al., 2014). Although some aromatic organosulfates have similar chemical structures as those of potential aromatic precursors of volatile organic compounds (VOCs), such as toluene and xylene (Kundu et al., 2013), monocyclic aromatics were not regarded as aromatic organosulfate precursors because they are more inclined to oxidize into ring opening products (Staudt et al., 2014; Kamens et al., 2011). Polycyclic aromatic hydrocarbons (PAHs), such as naphthalene (NAP) and 2-methylnaphthalene (2-MeNAP) have rather been postulated as the precursors of aromatic OSs (Riva et al., 2015; Staudt et al., 2014).

Among all aromatic compounds, PAHs are ubiquitous organic compounds enriched at both the sea and freshwater (lakes, river) surface (Cincinelli et al., 2001; Vácha et al., 2006; Chen et al., 2006; Lohmann et al., 2009; Seidel et al., 2017), where they can reach even 200–400 times higher concentrations compared to the water bulk (Cincinelli et al., 2001; Vácha et al., 2006; Chen et al., 2006; Lohmann et al., 2009; Seidel et al., 2017; Hardy et al., 1990). The origin of PAHs accumulated at the water surface stems from combustion processes, such as biomass burning, coal-based and petroleum-based combustion (Lammel, 2015). At the surface of freshwater and seawater, the PAHs concentrations vary from 11.84 to 393.12 ng L⁻¹ (J. Li et al., 2017), and from 5 to ∼ 1900 ng L⁻¹, respectively (Valavanidis et al., 2008; Otto et al., 2015; González-Gaya et al., 2019; Pérez-Carrera et al., 2007; Ma et al., 2013). Phenanthrene (PHE), fluoranthene (FLA) and pyrene (PYR) are the most commonly detected PAHs in the coastal surface microlayer (SML) (Guitart et al., 2007; Stortini et al., 2009), and they accounted for 92 %–96 % of the total amount of PAHs in the coastal surface waters of Nigeria (Benson et al., 2014).

Dimethylsulfoxide (DMSO) is an ubiquitous OS compound at the sea surface (Lee et al., 1999), which derives from degradation of phytoplankton (Andreae, 1980), photodegradation of dimethylsulfide (DMS) (Barnes et al., 2006; Brimblecombe and Shooter, 1986), and microbial oxidation of DMS (Zhang et al., 1991). Because of a high Henry's Law coefficient (≈ 10⁷ M atm⁻¹) and mass accommodation coefficient (0.10), gaseous DMSO can be deposited on the seawater and freshwater surface where one finds enhanced DMSO concentrations (Legrand et al., 2001; Davidovits et al., 2006; González-Gaya et al., 2016). The highest levels of DMSO are detected in the ocean (1.5–532 nmol L⁻¹) (Hatton et al., 1996; Lee and De Mora, 1996; Andreae, 1980; Asher et al., 2017), followed by rivers (< 2.5–210 nmol L⁻¹) (Andreae, 1980), lakes (up to 180 nmol L⁻¹) (Richards et al., 1994), rainwater (2–4 nmol L⁻¹) (Ridgeway et al., 1992), and aerosols (69–125 pmol m⁻³) (Harvey and Lang, 1986). One of the main degradation pathways of DMSO is the reaction with hydroxyl radicals (OH) (Barnes et al., 2006), yielding methanesulfinic and methanesulfonic acids (Librando et al., 2004). Because the concentrations of DMSO are 1–2 orders of magnitude higher than those of DMS, the photochemical oxidation of DMSO may be a relatively more important process than the photo-oxidation of DMS (Lee et al., 1999).

Sulfur dioxide (SO₂) is directly emitted into the atmosphere by anthropogenic sources, such as fossil fuel combustion, coal, oil, and industrial processes (Smith et al., 2001). In addition, it can also be formed during the oxidation of DMS (Hoffmann et al., 2016; Chen et al., 2018). As a key contributor to aerosol nucleation, the role of SO₂ at the air-water interface is also recognized as an efficient precursor of HOSO and OH radicals, following light absorption below 340 nm and its excitation to a very reactive triplet state (³SO${}_{\mathrm{2}}^{\ast }$) (Martins-Costa et al., 2018; Kroll et al., 2018). The majority of studies employed sulfate-containing seed particles to explore the formation pathway of OSs, while only a few were focused on the organosulfur formation pathway that is unique to SO₂ chemistry (Blair et al., 2017; Shang et al., 2016; Passananti et al., 2016). Some previous studies have found that the heterogeneous reaction of SO₂ with unsaturated acids can lead to the formation of OSs in the atmosphere (Shang et al., 2016; Passananti et al., 2016). However, it has also been found that monocyclic compounds such as terephthalic acid are not reactive via direct SO₂ addition (Passananti et al., 2016), and there is still a knowledge gap concerning other OSs formation pathways involving heterogeneous SO₂ oxidation. Indeed, air quality models cannot explain the quick increase of sulfate amounts in going from clean air to hazy events, when applying only the gas phase and aqueous phase chemistry of SO₂ (Wang et al., 2014; G. Li et al., 2017).

Several studies have assessed the heterogeneous chemistry of atmospherically relevant oxidants with PAHs (Donaldson et al., 2009; Monge et al., 2010; Styler et al., 2011; Zhou et al., 2019). Recently, Mekic et al. (2020a) and Jiang et al. (2021) have shown that the photosensitized degradation of DMSO by the excited triplet state of typical PAH compounds (fluorene (³FL^∗), ³PHE^∗, ³FLA^∗ and ³PYR^∗) leads to the formation of OSs, among others, in both the gas and aqueous phases.

In this study we investigated the formation of OSs from aqueous DMSO and/or PAHs/DMSO, initiated by gaseous SO₂ in the dark and in the presence of simulated sunlight irradiation (300 nm < λ < 700 nm). The gaseous OS products were assessed by membrane inlet single photon ionization-time of flight-mass spectrometry (MI-SPI-TOFMS) (Zhang et al., 2019; Mekic et al., 2020a). The formed aqueous phase products were evaluated by means of ultrahigh-resolution electrospray ionization Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) (Jiang et al., 2016). The tentative reaction pathways of the formed OSs during the heterogeneous SO₂ oxidation of PAHs/DMSO under actinic illumination are supported by theoretical calculations of the Gibbs energy of the reaction. We show that oxidation by SO₂ of PAHs/DMSO can release gaseous OSs, such as methanesulfonic acid (MSA), which are known precursors of secondary organic aerosols (SOA) in the atmosphere. The formation of an important number of OSs and unsaturated compounds, among others, was observed in the aqueous phase. We highlight the large amounts of generated linear and aromatic OSs, with potential to greatly influence the air-water exchange of organic compounds.

2 Experimental

2.1 Photoreactor

A double-wall rectangular (5 × 5 × 2 cm) photoreactor was used to assess the reaction of gaseous SO₂ with a water/organic film containing DMSO or DMSO/PAHs (Mekic et al., 2020a, b) The photoreactor was thermostated at ambient temperature (T = 293 K) in a thermostat bath (LAUDA ECO RE 630 GECO, Germany).

A SO₂ flow of 150 mL min⁻¹ mixed with an air flow of 750 mL min⁻¹ (0–1 L min⁻¹ HORIBA METRON mass flow controller; accuracy, ±1 %) allowed dilution of SO₂ from a standard gas cylinder with 5 ppm concentration to a mixing ratio of ∼ 800 ppb that was flowing through the photoreactor. The applied mixing ratio of 800 ppb is higher compared to the atmospheric background mixing ratio that ranges from 1 to 70 ppb, but it is smaller than the SO₂ mixing ratio in dilute volcanic plumes (for instance, a mixing ratio of ca. 10 ppm is observed about 10 km away from volcanic sources) (Oppenheimer et al., 1998), and it is also smaller compared to some previous laboratory studies that used 7.7–500 ppm of SO₂ (Librando et al., 2014). The applied mixing ratio of 800 ppb would probably amplify the intensity of the detected product compounds but the formation profiles would still remain the same as in the case of smaller SO₂ mixing ratios.

The concentration of PHE, FLA, and PYR (Sigma-Aldrich) used separately was 1 × 10⁻⁴ mol L⁻¹. The three compounds were dissolved individually in a mixture of DMSO and ultrapure water with a proportion of 10 : 90 $v/v$ (Mekic et al., 2020b). Such a high DMSO concentration was necessary to dissolve the poorly water soluble PHE, FLA, and PYR (Mekic et al., 2020a, b). Several previous studies also used high concentrations of the organic co-solvent to assess the co-solvent effect on PAHs photolysis (Donaldson et al., 2009; Librando et al., 2014; Grossman et al., 2016). The reactor was filled with 10 mL freshly prepared DMSO/PAHs solution and irradiated with a xenon lamp (Xe, 500 W, 300 nm < λ < 700 nm) during the SO₂ oxidation of PAHs/DMSO. The spectral irradiance of the Xe lamp was measured with a calibrated spectroradiometer (Ocean Optics, USA) equipped with a linear-array CCD detector, and compared to the sunlight radiation (Mekic et al., 2020a, b). The irradiation time of the aqueous solution was 2 h.

Blank experiments of SO₂ reaction were carried out with aqueous DMSO in all experimental conditions and with ultrapure water in the dark. All experiments were performed at least twice.

2.2 Single photon ionization-time of flight-mass spectrometry (SPI-ToF-MS)

A SPI-ToF-MS instrument (SPIMS 3000, Guangzhou Hexin Instrument Co., China) was used to detect the gas phase compounds formed during the light-induced heterogeneous reaction of SO₂ with PAHs/DMSO. The SPI-ToF-MS instrument was explained in details in our previous studies (Deng et al., 2021; Mekic et al., 2020a), so here only a brief description is given in the Supplement.

2.3 Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS)

As FT-ICR-MS was described in our previous paper, a brief description only is given in the Supplement. An in-house software was used to calculate all mathematically possible formulae for all ions with a signal-to-noise ratio above 10, using a mass tolerance of ±0.2 ppm. Data of blank samples were subtracted from those of all samples according to the same possible formulae.

The double bond equivalent (DBE) of the chemical formula C_cH_hO_oN_nS_s is calculated using the following equation (Yassine et al., 2014):

$$\begin{array}{}\text{(1)}& \mathrm{DBE}={\displaystyle \frac{\mathrm{2}c+\mathrm{2}-h+n}{\mathrm{2}}}.\end{array}$$

The aromaticity equivalent (X_c) is calculated using the following equation:

$$\begin{array}{}\text{(2)}& X_c={\displaystyle \frac{\mathrm{3}\left[\mathrm{DBE}-\left(m\mathrm{O}+n\mathrm{S}\right)\right]-\mathrm{2}}{\mathrm{DBE}-(m\mathrm{O}+n\mathrm{S})}},\end{array}$$

where X_c is introduced to improve the identification and characterization of aromatic and condensed aromatic compounds, while m and n are the respective fractions of oxygen and sulfur atoms that are involved in the π-bonds of the molecular structure (Yassine et al., 2014). According to the model structural classes, the values of m and n were all set to 0. Threshold values of X_c between 2.5 and 2.7 (2.5 ≤ X_c < 2.7) and equal or greater than 2.7 (X_c≥2.7) were set as minimum criteria for the presence of aromatics or condensed aromatic compounds in an identified molecule (Blair et al., 2017; Jiang et al., 2016). It should be highlighted that electron spray ionization (ESI) and desorption electrospray ionization (DESI) of OSs is especially favourable in the negative-ion mode, making the observed fraction of OSs further amplified (Blair et al., 2017).

2.4 Theoretical calculations

The proposed structures of all identified tentative OSs are based on the reasonable inferred elemental compositions for a single mass, following speculation through the NIST Chemistry WebBook (https://webbook.nist.gov/chemistry/mw-ser/, last access: 18 December 2020) and database of MI-SPI-ToF-MS or FT-ICR-MS. Considering that PHE, PYR, FLA and SO₂ can all absorb Xe lamp radiation, theoretical calculations were carried out to obtain some insights into the degradation pathways of DMSO initiated by the excited triplet states ³PHE^∗, ³PYR^∗, ³FLA^∗ and ³SO${}_{\mathrm{2}}^{\ast }$ in the aqueous and gaseous phases as well as the interaction between light-excited PAHs and SO₂.

All calculations shown here were assessed by the Gaussian 16W package (Frisch et al., 2016). The level of theory B3LYP/6-311G(d,p) was applied for geometry optimizations and frequency calculations for all molecules depicted in the reaction scheme (Mclean and Chandler, 1980; Binning and Curtiss, 1990). There were no imaginary frequencies for all molecules optimized. Single-point energy calculations were performed at a more expensive level, i.e. M06-2X/Def2-TZVP level (Zhao and Truhlar, 2008; Weigend and Ahlrichs, 2005; Weigend, 2006). The existence of possible geometric isomers and conformers for each species was considered and investigated and those with the lowest calculated Gibbs free energy were selected. Molecular oxygen (O₂), carbon dioxide (CO₂) and water molecules (H₂O) were placed as reactants or products (if needed) to balance atoms in the schemes. Detailed Gibbs free energies for all molecules are presented in Tables S1 and S2 in the Supplement, and the corresponding Gibbs energies of the reaction are shown in Scheme 1.

Scheme 1 Detailed reaction mechanism describing the formation of OSs gas phase products initiated by ³PAHs^∗ and ³SO${}_{\mathrm{2}}^{\ast }$. Numbers in brackets written below molecules represent compound designations to better follow the discussion in the main text.

3 Results and discussion

3.1 Gaseous OSs detected by MI-SPI-ToF-MS

To detect the gas phase compounds formed by heterogeneous SO₂ oxidation of PAHs/DMSO in the dark and under light irradiation, we applied MI-SPI-ToF-MS as a novel promising technology for the real-time monitoring of VOCs (Zhang et al., 2019; Mekic et al., 2020a).

Figure 1 Venn diagram of the detected $m/z$ signals in the gas phase for the heterogeneous reaction of SO₂ with DMSO and PAHs/DMSO under light irradiation (300 nm < λ < 700nm) (a). The total number of identified $m/z$ signals between the heterogeneous reactions of SO₂ with DMSO and PAHs/DMSO, under light irradiation (b).

Figure 1a shows the Venn diagram of the observed number of $m/z$ signals corresponding to the gas phase products of the light-induced heterogeneous SO₂ oxidation of PAHs/DMSO. The Venn diagram showing the comparison of the gaseous products formed under different conditions is presented in Fig. S1. The biggest contributor to the total number of secondarily formed products is the light-induced SO₂ oxidation of DMSO (Fig. 1b). Among all detected $m/z$ signals (Table S4), we tentatively identified a number of unsaturated multifunctional molecules and OSs released in the gas phase from the reaction of SO₂ with either DMSO or PAHs/DMSO, which are summarized in Table S5. The possibility for the existence of isomers and of different molecular formulae for the compounds with the same molecular weight should be noted (Nizkorodov et al., 2011). We emphasize the formation of gaseous OSs, and especially of those that are known to be SOA precursors. For example, the $m/z$ signals of 80, 94, 96, 112, 124, 126 were inferred as methanesulfinic acid (CH₃SO₂H, MSIA), methylsulfonylmethane ((CH₃)₂SO₂, MSM), methanesulfonic acid (CH₃SO₃H, MSA), hydroxymethanesulfonic acid (CH₄O₄S, MSAOH), ethyl methanesulfonate (CH₃SO₃C₂H₅, EMS) and 2-hydroxyethenesulfonic acid (C₂H₅O₄SH, ESAOH) (Berresheim et al., 1993; Berresheim and Eisele, 1998; Karl et al., 2007; Hopkins et al., 2008; Gaston et al., 2010; Ning et al., 2020; Dawson et al., 2012).

3.2 Formation profiles of gaseous OSs

In this study, we observed rapid formation of MSA, MSIA, MSM, EMS, MSAOH, and ESAOH (Figs. 2 and S2). Figure 2 shows typical time evolution profiles of MSA formed by the reaction of SO₂ with PAHs/DMSO under light irradiation (300 nm < λ < 700 nm). The formation profiles of MSM, EMS, MSIA, MSAOH and ESAOH are shown in Fig. S2.

Figure 2 Formation profiles of $m/z=\mathrm{96}$ (methanesulfonic acid, MSA) upon light-induced heterogeneous reactions of SO₂ with DMSO and PAHs/DMSO.

The SO₂ oxidation of FLA/DMSO leads to MSA formation, the signal of which increases during the first 1 h and then slowly decreases (Fig. 2). The intensities of the product compounds (Figs. 2 and S2) decrease after 1 h most probably due to their reaction with SO₂ and/or their photodegradation. The signal profile of MSA due to reaction of SO₂ with PHE/DMSO is shorter than the others due to a technical problem of the instrument.

Intriguingly, the light-induced heterogeneous reaction of SO₂ with DMSO, PYR/DMSO, and PHE/DMSO produces MSA that remains approximately stable over the course of the reaction. These results indicate that the suggested reaction between SO₂ and DMSO or DMSO/PAHs under sunlight irradiation can produce MSA, which can be persistent enough to affect the new particle formation (NPF) process in the atmosphere. In Sect. 3.4 we suggest a tentative reaction pathway for the production of MSA and other OSs that is supported by theoretical calculations of Gibbs free energies.

3.3 Aqueous OSs detected by FT-ICR-MS

Numerous unsaturated multifunctional molecules and OSs were identified in the liquid phase during the reaction of SO₂ with either DMSO or PAHs/DMSO by using FT-ICR-MS. The number of detected product compounds in the aqueous phase was significantly higher compared to those detected in the gas phase, due to very high sensitivity of FT-ICR-MS. Figure S3 shows all the formulae detected upon the reaction of SO₂ with PAHs/DMSO under light irradiation. The shared formulae of the compounds detected upon reactions of SO₂ with PYR/DMSO and FLA/DMSO or PHE/DMSO were more abundant than those individually formed by the reaction of SO₂ with FLA/DMSO or PHE/DMSO. The number of detected compounds formed by reaction of SO₂ with PYR/DMSO was dominant among all the products of SO₂ oxidation of PAHs/DMSO, both in the dark and under light irradiation (Figs. S3b and S4). Interestingly, the light-induced heterogeneous reaction of SO₂ with PYR/DMSO released a small number of gaseous products but the highest number of aqueous phase products among all the studied PAHs/DMSO. In general, more C_cH_hO_o (CHO) than C_cH_hO_oS_s (CHOS) products were formed in the aqueous phase with SO₂ + PAHs/DMSO under irradiation (Fig. S5), with the only exception of PHE/DMSO. Even when subtracting the formulae detected upon SO₂ oxidation of DMSO in the dark from those detected in the corresponding light-induced heterogeneous reaction, more compounds were produced under irradiation than in the dark.

The analysis of iso-abundance plots of DBE vs. carbon numbers for the detected CHO and CHOS formulae is shown in text Sect. S3 and Figs. S6–S7. The two-dimensional van Krevelen (VK) plots for all CHOs and CHOSs formed during light-induced SO₂ oxidation of PAHs/DMSO are shown in Fig. 3.

Figure 3 The van Krevelen (VK) graph and aromaticity equivalent (grey with X_c < 2.5, black with 2.5 ≤ X_c < 2.7, and silver withX_c ≥ 2.7) for the detected CHO and CHOS compounds in ESI⁻ mode, formed upon light-induced heterogeneous reaction of SO₂ with PAHs/DMSO. The X_c value is illustrated by the colour bar of each VK diagram, while the pie chart shows the number in different X_c intervals during these reactions.

The CHO and CHOS compounds formed during the reaction of SO₂ with DMSO in the dark and under irradiation are shown in the VK plot depicted in Fig. S8. The same classes of compounds, detected upon heterogeneous reactions of SO₂ with PAHs/DMSO under irradiation are illustrated in the VK plots depicted in Figs. S9 and S10, respectively. Most of the CHO and CHOS products identified with DMSO had high H $/$ C ratios (1.5–2.1) but O $/$ C ratios lower than 0.4, suggesting the formation of aliphatic compounds (Mekic et al., 2020a). However, the presence of aromatics could be additionally highlighted according to the analysis of a mathematical parameter, the aromaticity equivalent (X_c) (Yassine et al., 2014). About one half of the observed product compounds and especially CHOSs exhibited X_c ≥ 2.5, further indicating the formation of unsaturated compounds during the oxidation of DMSO by SO₂ in the presence of light (Fig. S8).

The CHO and CHOS products displayed in the VK diagrams of Fig. 3 could be all separated into two different clusters. The O $/$ C ratios of CHOs were smaller than 0.7 for SO₂ reaction with PYR/DMSO, and smaller than 0.4 for SO₂ reactions with FLA/DMSO and PHE/DMSO. Nonetheless, a cluster with relatively few observed products was located in the upper part of the VK diagrams with high H $/$ C ratios (1.2–2.0), suggesting the formation of saturated aliphatic CHO compounds.

The other cluster that includes most of the detected products is located in the lower part of the VK diagrams, exhibiting low H $/$ C ratios (0.4–0.8) and indicating a degree of unsaturation (Mekic et al., 2020a; Lin et al., 2012). Most of the CHO compounds are probably condensed aromatic compounds as suggested by their X_c values higher than 2.5, especially in the case of products observed upon reaction of SO₂ with FLA/DMSO.

The observed CHOS products emerging from the reaction of SO₂ with PHE/DMSO, also depicted in Fig. 3 could be divided into two groups based on their distribution of H $/$ C and O $/$ C ratios. Generally, the formed CHOS products that are located in the upper part of the VK diagrams (Fig. 3) have a broad range of O $/$ C ratios spanning from 0.2 to 0.8, H $/$ C ratios in a range of 1.4–2.0, and low DBE (1–4). This observation implies the formation of long-chain aliphatic-like CHOSs during the light-assisted oxidation of PHE/DMSO by SO₂. In addition, a large fraction of compounds detected during the reaction of SO₂ with PHE/DMSO exhibits X_c < 2.5, and is thus consistent with the formation of long-chain aliphatic-like CHOSs (Wang et al., 2021). The CHOS products located in the lower part of the VK diagrams have H $/$ C ratios of 0.4–1.0, which are similar to those of the observed CHO products. However, the CHOS compounds exhibit higher O $/$ C ratios that range between 0.2 and 0.8 for the reaction of SO₂ with PYR/DMSO, and 0.2–0.6 for the reaction of SO₂ with FLA/DMSO and PHE/DMSO. A possible reason is the occurrence of sulfates with R-OSO${}_{\mathrm{3}}^{-}$ groups, sulfonates with R-SO${}_{\mathrm{3}}^{-}$ groups and sulfones with R-SO₂-R^′ groups (Bruggemann et al., 2020). These CHOS products partly overlap with the oxidized aromatic hydrocarbons (Kourtchev et al., 2014). Moreover, the majority of the CHOS products exhibit a condensed aromatic structure as indicated by their X_c values higher than 2.7.

Based on the VK plots, here we used the different spacing patterns for Kendrick mass defect (KMD) analysis (Lin et al., 2012). The subunits including CH₂, C₂H₂, CH₃, C₄H₂, C₆H₂ for the same CHO homologous series, and CH₂, C₂H₂ for CHOS homologues were identified during the light-induced SO₂ oxidation of PAHs/DMSO. More subunits were found for CHOS homologues during the SO₂ oxidation of PAHs/DMSO under all experimental conditions (Fig. S11). The literature suggests that subunits of CH₂ and C₂H₂ were the most repeating mass building increments for either CHO or CHOS oxidized aromatic compounds (Lin et al., 2012). These identified subunits were responsible for the increase in the molecular mass of the detected compounds (Altieri et al., 2006), presumably leading to straight-chain alkanes and olefins.

3.4 Tentative reaction mechanisms and the production of OSs

The proposed general pathway suggested in the literature is the direct addition of SO₂ to a double bond or the separate addition of SO₂ to the cleavage of a double bond. This may also apply to PAHs/DMSO, because of the chromophoric nature of the SO₂ adduct with C=C bonds (Passananti et al., 2016). Herein, we show that this reaction pathway of SO₂ may indeed proceed on PAHs and is supported by theoretical calculations of the Gibbs energy of the reaction, based on the information obtained from the detected tentative products.

The reaction pathway suggests that photoexcitation of PAHs and SO₂ proceeds through the $\mathit{\pi }\to {\mathit{\pi }}^{\ast }$ electronic transition, followed by intersystem crossing to produce the corresponding excited triplet states (³PAHs^∗ and ³SO${}_{\mathrm{2}}^{\ast }$) (Mekic et al., 2020a), which most likely play an important role during the oxidation of PAHs/DMSO. Previous work has shown that the photochemical reaction of ³PAHs^∗ with DMSO in water would lead to the formation of singlet oxygen (¹O₂) via energy transfer reaction with ground triplet-state oxygen (³O₂) (Wilkinson et al., 1995), the formation of hydroxyl radical upon water oxidation by other triplets states (Brigante et al., 2010) and the formation of more radicals through electron transfer. Many triplet states work in the presence of oxygen as O₂ quenches most but not all of them. The excited triplet state of SO₂ (³SO${}_{\mathrm{2}}^{\ast }$) formed upon light irradiation reacts with water molecules yielding OH radicals as follows (Martins-Costa et al., 2018; Kroll et al., 2018):

$$\begin{array}{}\text{(R1)}& {}^{\mathrm{3}}{\mathrm{SO}}_{\mathrm{2}}^{\ast }+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\to \mathrm{OH}+\mathrm{HOSO}.\end{array}$$

Alternatively, SO₂ can form π complexes with C=C bonds of PAHs upon ring opening, which may undergo transformation to diradical organosulfur intermediates, which in turn can react with dissolved O₂ leading to production of reactive oxygen species (ROS) such as OH radicals. (Shang et al., 2016) The formation of diradical organosulfur intermediates and ROS have been suggested for reactions of SO₂ with alkenes and fatty acids (Shang et al., 2016; Passananti et al., 2016) but here we suggest that the same pathway may occur for the reaction of SO₂ with PAHs. While the SO₂ addition to the C=C bond would be responsible for the OSs, the CHO oxidation products could be explained by radical chain reactions triggered by ROS (Shang et al., 2016; Passananti et al., 2016). The OH radical attack on PAHs/DMSO could also yield carbon-centred radicals or alkyl radicals which further react with O₂ yielding peroxyl (RO₂) and hydroxyperoxyl radicals (HO₂) (Von Sonntag et al., 1997).

Although RO₂ could directly transform into organic sulfates by SO₂, it has been shown that SO₂ could accelerate heterogeneous OH oxidation rates by 10–20 times, originating from the radical chain reactions propagated by alkoxy radicals formed by the reaction of peroxyl radicals with SO₂ (Richards-Henderson et al., 2016). Peroxyl radicals undergo self-reactions leading to the formation of stable products (ketone or alcohol) or alkoxy radicals (Richards-Henderson et al., 2016) In the presence of ³SO${}_{\mathrm{2}}^{\ast }$, the OH production rate increases by several orders of magnitude at the air-water interface, thus resulting in the increase of radical chain length by alkoxy radicals (Martins-Costa et al., 2018).

Although it is difficult to distinguish which mechanism is prevalent in the environment, in this study, the comprehensive reaction schemes explain the detected sulfur-containing unsaturated multifunctional compounds emerging from light-induced SO₂ oxidation of PAHs and DMSO at the air-water interface. The tentative reaction pathways describing the formation of aqueous phase products including their description are given in the text Sect. S4 and Scheme S1. The suggested reaction mechanism for the formation of gas phase product compounds is shown in Scheme 1 in the following section.

3.5 Reaction mechanism of the gaseous compounds

A detailed mechanism for the ³SO${}_{\mathrm{2}}^{\ast }$ oxidation of PAHs/DMSO is presented in Scheme 1, which could be divided into two proposed general pathways including (1) self-oxidation of ³SO${}_{\mathrm{2}}^{\ast }$ and oxidation of DMSO initiated by ³SO${}_{\mathrm{2}}^{\ast }$ and ³PAHs^∗, and (2) photodegradation of sulfur-containing PAH compounds that were initially formed from PAHs by ³SO${}_{\mathrm{2}}^{\ast }$.

• Pathway A: in this pathway we emphasize the formation and transformation of MSA that plays a key role during the formation of the detected OSs (Scheme 1a).

  The self-oxidation of ³SO${}_{\mathrm{2}}^{\ast }$ could yield sulfuric acid (H₂SO₄) (1). Meanwhile, DMSO can be oxidized by oxygen and the OH radicals formed by ³SO${}_{\mathrm{2}}^{\ast }$ and ³PAHs^∗, yielding MSIA (CH₄O₂S) (2) and MSM (C₂H₆O₂S) (3). Dehydration of MSIA (2) (Urbanski et al., 1998; Kukui et al., 2003; Allen et al., 1999; Arsene et al., 2002) and sulfuric acid (1) would lead to the production of MSA (CH₄SO₃) (4). More radical species such as the methyl radical (CH₃), would be formed during the OH-addition route in the gas phase atmospheric oxidation of DMSO, resulting in the formation of SO₂ that would participate in the subsequent oxidation reactions (Falbe-Hansen et al., 2000; González-García et al., 2006; Arsene et al., 2002; Urbanski et al., 1998).

  Further transformation would occur upon oxidation of MSA (4) or self-oxidation of ³SO${}_{\mathrm{2}}^{\ast }$ to yield MSAOH (CH₄O₄S) (5) via the OH₂ radical. The OH and CH₂OH radicals could oxidize MSA (4) into ESAOH (C₂H₆O₄S) (6) and C₂H₆O₃S (7).

• Pathway B: although linear OS products would also be generated upon addition of ³SO${}_{\mathrm{2}}^{\ast }$ to double bonds, formed after ring opening of the PAH molecules, we stress that the occurrence of gaseous OSs mostly emerges from the photodegradation of aromatic OS products. Under our experimental conditions, ³SO${}_{\mathrm{2}}^{\ast }$ oxidation of PAHs would prevail over PAHs photodegradation and would lead to sulfur-containing PAHs. Here we only present the general degradation process and one proposed pathway for the given species. The details of intermediate products are not shown in this scheme.

  For example, C₉H₈O₄S (9), C₁₀H₆O₃S (10), and C₁₀H₆O₄S (11) (Scheme 1b) appear as the photodegradation products of C₁₆H₁₀O₃S or C₁₆H₁₂O₃S, C₁₄H₁₀O₃S and C₁₄H₁₀O₄S (Scheme S1). These compounds were further transformed upon oxygen attack, then followed by cleavage of the phenyl ring through a highly exoergonic process. The S(IV) in sulfite group of C₁₀H₆O₃S (10) would first undergo oxidation by the strong oxidizing agents in the system to result in more stable S(VI) in C₁₀H₈O₃S (12). Cleavage of the 5-member ring would yield C₁₀H₈O₄S (13) and C₁₀H₈O₄S (14), followed by the yield of phenyl hydrogen sulfate (C₆H₆O₄S) (15) via phenyl ring cleavage. Meanwhile, phenyl ring cleavage of condensed aromatics would also yield C₅H₆O₄S (16) and 1,3,2-dioxathiole 2,2-dioxide (C₂H₂O₄S) (17). Subsequently, the attack of oxygen or radicals initiates the exoergonic opening of a 5-member ring or a phenyl ring, resulting in linear OSs under rapid oxidative scission.

3.6 Comparison with OSs identified within atmospheric aerosols

Table S6 shows the intercomparison of the OSs detected here, mainly formed upon light-induced reactions of SO₂ with PAHs/DMSO, with OSs identified during various field campaigns. It should be stressed that the agreement between the chemical formulae of the OSs detected in this study and those from field campaigns does not necessarily imply the same molecular structure, because multiple structural isomers are plausible for each formula (Mekic et al., 2020a; Nizkorodov et al., 2011). A total of 81 tentative OSs from this study overlapped with those identified in ambient aerosol samples, of which 33 OS formulae were detected in the aqueous phase. Only 4 liquid phase OS formulae were detected following the reactions of SO₂ with PAHs/DMSO in the dark. The blue-shaded OSs are generated exclusively by the reaction of SO₂ with DMSO. The group of compounds with yellow-shaded area are produced by the reaction of SO₂ with PAHs/DMSO. Finally, the green-shaded area summarizes the OSs detected with both SO₂ + DMSO and SO₂ + PAHs/DMSO. Most of the liquid phase OS formulae (chemical formulae in bold) were formed individually upon SO₂ oxidation of PYR/DMSO and DMSO. The product compounds individually formed by the reaction of SO₂ with PHE/DMSO and the shared formulae generated by the reaction of SO₂ with DMSO and PAHs/DMSO make the biggest contribution to the gas phase OSs. These observations highlight the importance of the SO₂ oxidation reactions of DMSO and/or PAHs/DMSO at the freshwater and seawater surfaces or in the liquid films of the aerosol particles, which would represent an important source of OSs.

Tentatively identified VOC precursors are also highlighted in Table S6. The tentative precursors of more than half of the identified OSs in the atmosphere still remain unexplained and unknown. Most of the identified overlapping OSs are probably long alkyl chain OSs, thus their tentative precursors were designated as alkyl-containing OSs from anthropogenic sources (Tao et al., 2014). There are three overlapping compounds which were assumed to be aromatic OSs, thus their precursors were considered as 2-MeNAP and methylbenzyl sulfate (Staudt et al., 2014; Riva et al., 2015).

Figure 4 shows the iso-abundance plot of DBE vs. carbon numbers according to the corresponding X_c of liquid-phase OS formulae, intuitively indicating that most of the identified OS formulae with low DBE are presumably long chain aliphatic-like compounds. Moreover, there were seven detected OSs with aromatic structure that were consistent with those identified in aerosol samples from field studies. Among those aromatic OSs, only C₈H₈O₅S has a tentatively identified VOC precursor, 2-MeNAP (Riva et al., 2015). It has been shown that aromatic OSs are produced from the interaction of aromatics with sulfur-containing species. For example, the gas-phase photo-oxidation of PAHs in the presence of sulfate aerosol, and the aqueous phase reactions of several monocyclic aromatics with sulfite in the presence of Fe³⁺ mediated by sulfoxy radical anions can generate aromatic OSs (Riva et al., 2015; Huang et al., 2020). The reaction of SOA particles formed from photo-oxidation of biodiesel and diesel fuel with SO₂ was demonstrated to yield a large number of aromatic OS species (Blair et al., 2017).

Figure 4 Iso-abundance plot of DBE vs. carbon numbers according to the corresponding aromaticity equivalent (green with X_c < 2.5, red with 2.5 ≤ X_c < 2.7, and blue withX_c ≥ 2.7) for the detected organic sulfur species in the liquid phase, which had also been identified in ambient aerosol samples.

In our previous study we have identified the OSs formed in the absence of SO₂ as a precursor, during the photodegradation of DMSO initiated by the excited triplet state of fluorene in both liquid and gas phases (Mekic et al., 2020a, b). It is evident that the inclusion of SO₂ as a precursor in the reaction system enhances the formation of OSs, including those having similar structures as those detected in field studies. It can be considered that the light-induced reaction of SO₂ with DMSO or a mixture of DMSO and PAHs is a previously unknown, additional atmospheric source of OSs (Berresheim et al., 1993; Berresheim and Eisele, 1998; Karl et al., 2007; Hopkins et al., 2008; Gaston et al., 2010; Ning et al., 2020; Dawson et al., 2012).

4 Atmospheric implications

​​​​​​​The reaction of DMSO with OH leads to formation of MSIA (Barnes et al., 2006), which in turn may react again with OH radicals to form MSA (Librando et al., 2004; Barnes et al., 2006; Rosati et al., 2021). Gaseous MSA is a particularly important compound that can participate in the initial nucleation and growth step of particles, known as the NPF process (Schobesberger et al., 2013; Chen et al., 2016; Zhao et al., 2017; Perraud et al., 2015). Declining gaseous MSA concentrations were actually reported during marine NPF events, thereby suggesting that MSA may enter the aerosol particles at the earliest possible stage and significantly assist in cluster formation (Dall'osto et al., 2012; Bork et al., 2014). MSA is the simplest organosulfate compound, which can be mainly formed during the OH oxidation of DMS (Barnes et al., 2006; Rosati et al., 2021). Gas-phase MSA above the oceans and in coastal areas represents about 10 %–100 % of the gas phase sulfuric acid (SA) concentration (Berresheim et al., 2002). It has been suggested that MSA can enhance particle formation from SA by 15 %–300 %, if equal quantities of SA and MSA are present (Bork et al., 2014). There is a discrepancy between the modelled and measured concentration profiles of MSA, which indicates a missing MSA source. This source seems to be much stronger than the estimated production stemming from OH oxidation of DMS (S. H. Zhang et al., 2014). Y. Zhang et al. (2014) suggested a strong daytime source of MSA, the precursor of which may be DMSO (Y. Zhang et al., 2014). The model estimations indicated that higher DMSO concentrations would lead to enhanced chemical production of MSA to reach 4.9  × 10⁷ molecules cm⁻² s⁻¹, which is similar to the strength of the missing source of MSA.

Here, we show that during daytime the reactions of light-excited SO₂ and aqueous DMSO or DMSO/PAHs could represent an important source of gaseous MSA in the atmosphere near the water (ocean, lake and river) surface. In particular, the reaction of SO₂ with DMSO leads to enhanced formation of organic sulfur compounds compared to the photosensitized degradation of DMSO initiated by the excited triplet states of PAH compounds (Fig. S13). The results of this research point to the complexity of the chemical processes responsible for the formation of organic sulfur compounds. Considering the abundance of SO₂ in the atmosphere and the omnipresence of water-adsorbed PAHs and DMSO compounds, the future modelling studies should consider both pathways, i.e. (1) photosensitized degradation of DMSO initiated by the excited triplet states of PAHs (Zhang et al., 2019; Jiang et al., 2021), and (2) heterogeneous chemistry of SO₂ with PAHs/DMSO and DMSO, as alternative formation pathways of organic sulfur compounds in the atmosphere. In this study, many detected aromatic and linear OSs formed during the light-induced SO₂ oxidation of PAHs/DMSO are reported for the first time. An important number of detected compounds overlapped with those of ambient OSs identified during field measurements. We suggest that a plausible mechanism for OSs formation via direct addition of SO₂ to the C=C double bond is not only limited to alkenes and unsaturated fatty acids (Passananti et al., 2016; Shang et al., 2016) but is also valid for anthropogenic precursors such as PAHs. A large amount of organosulfates and especially aromatic organosulfates could be formed through this pathway and released into water and the ambient atmosphere. These OSs can form surfactant films on aerosol particles in the boundary layer, by which means they influence the surface tension and hygroscopicity of particles (Decesari et al., 2011; Tao et al., 2014). Indeed, the OSs formation pathway in the heterogeneous reaction between SO₂ and PAHs/DMSO is of great significance because OSs generated at the water surface would influence the air-water exchange and enhanced gaseous OSs would result in the formation of SOA. Moreover, the OS products formation from PAHs initiated by SO₂ is also of importance in urban areas where PAHs concentrations are usually high in ambient air (Zhu et al., 2019; Cai et al., 2020). Finally, aromatic OSs occurring in urban aerosols represent a still unrecognized source of toxic products (Riva et al., 2015).

Based on the observed emission rates of OSs in this study, we estimate emission fluxes of MSA, and MSIA, among others, considering realistic environmental conditions, SO₂ mixing ratios ranging between 2 and 50 ppb, surface UV irradiation (Brüggemann et al., 2018), surface microlayer coverage with PAHs/DMSO, to account for the potential impact of the heterogeneous SO₂ (photo)chemistry with PAHs/DMSO, on the aerosol production in marine boundary layer, the results of which will be published elsewhere.