Three types of commonly used biomass (RS, WS, and CS) were cut into small uniform pieces (∼ 10 cm in length) and dried. About 100 g of the dried biomass materials (∼ 10 % moisture content) was then introduced into a traditional iron stove commonly used in rural areas (Fig. S1 in the Supplement). The stove was covered with a hood, and the biomass was ignited using a propane lighter. The BB smoke generated was collected onto 90 mm quartz filters at 0.9 m³ min⁻¹ for 10 min by a custom-made aerosol sampler under mixed combustion conditions (include flaming and smoldering, modified combustion efficiency (MCE) 0.85 ≤ΔCO2/(ΔCO2+ΔCO)≤0.95) (Ting et al., 2018). The sampler was placed 1 m above the ground and connected to a PM_2.5 sampling head through a sampling pump. For incense burning (IS), laboratory-generated smoldering smoke was collected on 47 mm quartz filters at a flow rate of ∼ 6.0 L min⁻¹ for 80 min using a stainless-steel combustion chamber. Note that the different combustion modes of IS and BB are intentionally used to represent real-world combustion conditions. Our previous study demonstrated that IS was representative of BB based on GC × GC (two-dimensional gas chromatography) chromatograms and pixel-based partial least-squares discriminant analysis (Tang et al., 2023). Hereafter, we will use BB to represent both the real BB materials and the surrogate materials (IS) unless otherwise specified. After sampling, the collected BB samples (fresh BB) were wrapped by pre-baked aluminum foil (550 °C for 6 h) and stored at -20 °C until further analysis.

To achieve atmospheric OH• aging, the collected fresh BB filter samples were placed in a pre-flushed chamber (zero air, more than 24 h) and illuminated with UV lamps for 40 min. We used lamps of 185 and 254 nm, the combination of which has been widely used in oxidation flow reactor design and experiments for mimicking atmospheric OH• aging conditions (Peng and Jimenez, 2020; Rowe et al., 2020; Tkacik et al., 2014; Hu et al., 2022). The estimated OH exposure was ∼ 2.0 × 10¹² molecules cm⁻³ s, equivalent to an atmospheric aging period of 15 d (assuming an average atmospheric OH concentration of 1.5×106 molecules cm⁻³) (Mao et al., 2009). Detailed characterization of the OH exposure can be found in our previous study (Tang et al., 2023). We will discuss the sulfate formation of these fresh and OH• aged samples later.

Aqueous stock solutions of BB samples were prepared by dissolving the collected filters in ultrapure water and subjecting them to ultrasonication in a cooled-water bath three times, each for 20 min. The resulting water extracts of the BB were then filtered through 0.22 µm PTFE filters and stored in brown vials at 4 °C in a refrigerator. The anions, i.e., the chloride, sulfate, and nitrate of the BB extracts, were analyzed by Dionex ion chromatography (ICS-1100, USA). An aliquot (∼ 0.5 mL) of the BB or IS extracts was used for water-soluble organics detection by ultra-high-performance liquid chromatography (UHPLC; Thermo Fisher Scientific, Dionex UltiMate 3000) coupled with high-resolution Orbitrap Fusion Lumos Tribrid mass spectrometry (Orbitrap HRMS, Thermo Fisher Scientific, USA). The particulate organic matter was characterized by directly heating the filter samples using a thermal desorption module (TDS 3, GERSTEL) coupled to a comprehensive two-dimensional gas chromatograph–mass spectrometer (GCMS-TQ™8050 NX, Shimadzu, Japan). UV–Vis spectrometry (UV-3600, Shimadzu, Japan) was employed to examine the absorbance of BB extracts. Total organic carbon (TOC) was measured by a total carbon analyzer (TOC-L CPH, Shimadzu, Japan). Metal concentrations were measured by inductively coupled plasma–mass spectrometry (ICP-MS, Agilent 7800). Detailed analysis can be found in Sect. S1 in the Supplement. Aqueous stock solution of sodium chloride (≥ 99.8 %, Unichem) was prepared by dissolving the corresponding salt in ultrapure water to obtain a concentration of 1 M. The study utilized high-purity-grade synthetic air and nitrogen supplied by Linde HKO Ltd., while sulfur dioxide was obtained from Scientific Gas Engineering Co., Ltd.

In SO₂ uptake experiments, the stock solution of BB (fresh and aged) extracts was premixed with sodium chloride solution (1 M) at a volume ratio of 1:1, and the solutions had a pH of 4–6. A droplet generator (model 201, Uni-Photon Systems, Inc.) was then utilized to deposit droplets onto a hydrophobic substrate (model 5793, YSI Inc.) for SO₂ uptake experiments. Reactive SO₂ uptake experiments were performed via a flow cell–in situ Raman system at controlled room temperature (23–25 °C). The top and bottom quartz windows of the flow cell were used for Raman analysis and UV irradiation, respectively. The light experiment was performed using a xenon lamp (model 6258, ozone-free, 300 W, Newport), with a photon flux of 9.8×1015 photons cm⁻² s⁻¹ at 280–420 nm received by particles in the flow cell (Zhang and Chan, 2023). Identical experiments were conducted in the dark with the lights off and the experimental area kept in complete darkness. The relative humidity (RH) inside the flow cell was adjusted to 80 % by mixing dry and wet synthetic air or nitrogen. The particles were then equilibrated at 80 % RH for over 60 min and remained liquid throughout the experiment period. SO₂ was introduced into the system to reach a concentration of 8.0 ppm. The prescribed size used in our in situ Raman research was 60±5 µm. Despite using particles for droplet experiments that were larger than ambient fine particles, we employed the SO₂ uptake coefficient (γSO2) as a kinetic parameter to account for the particle size effects. Comprehensive calculation of γSO2 can be found in our previous studies (Gen et al., 2019b, a; Tang et al., 2023; R. Zhang et al., 2020).

Aqueous-phase photochemical reactions were performed using a quartz photo reactor (Go et al., 2023, 2022). Specifically, a 500 mL solution containing 100 ppm bisulfite and 1 ppm BB TOC extracts was continuously mixed using a magnetic stirrer throughout the experiments. Note that the 1 ppm BB TOC and 100 ppm bisulfite align well with the atmosphere-relevant ranges in aqueous aerosols, fog, and clouds, where PS concentration can reach hundreds of micromolar and total sulfur concentration can exceed several millimolar (Anastasio et al., 1997; Guo et al., 2012; Shen et al., 2012; Rao and Collett, 1995). To achieve air-saturated conditions, synthetic air was continuously introduced into the solutions at a flow rate 0.5 L min⁻¹ throughout the experiments. The above-described mixed solutions were then exposed to radiation via the same xenon lamp (light intensity of 1318 mW cm⁻²) as in the droplet experiments. Samples were collected at 1 h intervals for a total of 8 h for sulfate and bisulfite analysis using ion chromatography. Unlike droplet experiments, NaCl was not added to the bulk solution in most studies, unless specified otherwise.

As no sulfate was detected under dark conditions for any of the experiments, we have focused on the light experiments. Figure 1 depicts the sulfate production by (panel a) fresh BB–NaCl droplets and (panel b) aged BB–NaCl droplets as a function of time in the presence of light, air, and SO₂ at 80 % RH. Our previous study (Tang et al., 2023) found significantly higher sulfate formation of IS–NaCl droplets compared to NaCl droplets. Here, we only focus on the comparison of sulfate formation between different kinds of BB–NaCl droplets and IS–NaCl droplets. Regardless of whether the extracts were fresh or aged, the sulfate production by BB–NaCl droplets was higher than IS–NaCl droplets. Specifically, sulfate formed by fresh (F) BB–NaCl droplets followed the trends of CS_F–NaCl (16.8±2.6 mM (ppmC)⁻¹) > RS_F–NaCl (9.8±0.1 mM (ppmC)⁻¹) > WS_F–NaCl (4.2±0.2 mM (ppmC)⁻¹) > IS_F–NaCl (0.8 mM (ppmC)⁻¹) after illumination for 1080 min. In aged (A) samples, while BB_A–NaCl is more efficient than IS_A–NaCl in sulfate formation, the order of sulfate formation was different from that of the fresh samples: RS_A–NaCl (35.2±0.6 mM (ppmC)⁻¹) > CS_A–NaCl (13.0±0.1 mM (ppmC)⁻¹) > WS_A–NaCl (6.0±1.6 mM (ppmC)⁻¹) > IS_A–NaCl (0.6 mM (ppmC)⁻¹). The sulfate enhancement factors of RS_F–NaCl, WS_F–NaCl, and CS_F–NaCl compared to IS_F–NaCl after 18 h of SO₂ uptake (SulfateBBF-NaCl/ISF-NaCl) were 11.7, 5.0, and 20.0, respectively. The enhancement of sulfate can also be observed in aged BB samples, with values of 54.3, 9.2, and 20.1 for RS_A–NaCl, WS_A–NaCl, and CS_A–NaCl, respectively. The lower sulfate formation of IS–NaCl droplets compared to BB–NaCl droplets can be explained by the significantly higher TOC concentration of IS due to the incomplete and smoldering combustion (Table S1 in the Supplement). The TOC concentration of the IS extracts (> 550 mg L⁻¹) was nearly an order of magnitude higher than that of the BB extracts (34.0–69.9 mg L⁻¹), and the sum of anions per water-soluble organic carbon (WSOC) exhibited a more than 10-fold increase in BB extracts compared to IS extracts. Previous studies have confirmed that the smoldering condition of BB results in significantly more organic compounds and fewer ions than the flaming condition (Y. Wang et al., 2020; Fushimi et al., 2017; Kalogridis et al., 2018; Kim et al., 2018). Additionally, a significantly higher polycyclic aromatic hydrocarbon (PAH) proportion (12.2 %–16.6 % by intensity) in BB compared to IS (∼ 5.0 %) were observed by GC × GC–mass spectrometry (MS). G. Huang et al. (2022) reported higher PAHs in BB particles (CS, WS, and RS > 262.5 mg kg⁻¹ and > 3.7 % of organic matter) than in IS particles (3.3 mg kg⁻¹, 0.9 % of organic matter) (Song et al., 2023). Fushimi et al. (2017) and Kim et al. (2021) demonstrated that more PAHs can be emitted under flaming than smoldering conditions. PAHs like pyrene, fluoranthene, and phenanthrene have been recognized as PSs (Jiang et al., 2021; Yang et al., 2021) and are mainly from combustion processes, e.g., pyrosynthesis from aliphatic and aromatic precursors in biomass burning processes, and the constituents vary with temperatures and oxygen contents (Pozzoli et al., 2004). The higher percentage of PAHs in BB together with the collection procedure (mixed combustion and higher temperature for real BB and smoldering and lower temperature for IS) suggested that the BB materials would generate more PAHs at high temperatures and may contribute to sulfate formation.

Table 1 and Fig. S2 present the reactive (γSO2) and normalized reactive SO₂ uptake coefficients (nγSO2) of different BB–NaCl droplets. The γSO2 values obtained in our study are 0.9×10-6–6.6×10-6, which are consistent but fall on the low side of the reported heterogeneous SO₂ oxidation processes, including nitrate photolysis (10⁻⁶–10⁻⁵) (Gen et al., 2019b), transition metal ion (TMI)-catalyzed oxidation (10⁻⁶–10⁻⁴) (Zhang et al., 2024), NO2/O3 oxidation (10⁻⁶–10⁻⁴) (S. Zhang et al., 2021; Zhang and Chan, 2023), and peroxide oxidation (10⁻⁶–10⁻¹) (S. Wang et al., 2021; Ye et al., 2018; Yao et al., 2019). Additionally, the reported γSO2 in our study aligns well with the results obtained from ambient samples in Beijing (Y. Zhang et al., 2020). The large discrepancy of the reported γSO2 values can be attributed to the differences in aerosol components, particle size, RH, SO₂, and oxidant concentrations. Higher nγSO2 values were found for fresh and aged real BB–NaCl droplets than IS–NaCl droplets, following the trends of CS_F–NaCl (8.8×10-8 (ppmC)⁻¹) > RS_F–NaCl (6.2×10-8 (ppmC)⁻¹) > WS_F–NaCl (2.0×10-8 (ppmC)⁻¹) > IS_F–NaCl (0.61×10-8 (ppmC)⁻¹) and RS_A–NaCl (2.2×10-7 (ppmC)⁻¹) > CS_A–NaCl (6.2×10-8 (ppmC)⁻¹) > WS_A–NaCl (3.5×10-8 (ppmC)⁻¹) > IS_A–NaCl (0.46×10-8 (ppmC)⁻¹), respectively.

In our previous study, we observed a significant increase in sulfate formation in IS–NaCl droplets compared to NaCl droplets, which we attributed to PSs present in the IS (Tang et al., 2023). Considering the fact that BB–NaCl droplets produced sulfate more efficiently than IS–NaCl droplets and NaCl droplets, we explore the underlying mechanisms driving this phenomenon. Possible reasons include nitrate (from BB extracts or newly formed) photolysis; [Cl⁻–H₃O⁺–O₂] photoexcitation (Cl⁻ from BB extracts); H₂O₂ oxidation; black carbon (BC)-catalyzed oxidation; reactive-nitrogen-species oxidation; and organics-driven pathways, e.g., HCHO, photosensitizing components, organic peroxide, and TMI-organic complexes (Ye et al., 2023).

Since there was no nitrate peak in our Raman spectra in all experiments, the potential impact from nitrate photolysis was excluded. Besides, the very low Cl⁻ concentration (0.0002–0.001 M) in the original BB extracts (compared to 1 M NaCl, Table S1) has minimized the influence of chloride photoexcitation of [Cl⁻–H₃O⁺–O₂] (Cl⁻ from BB extracts) on the sulfate formation. Reactive nitrogen species, e.g., NO_x, HONO, and NH₃, were neither introduced nor detected in our system, indicating that the oxidation pathway involving reactive nitrogen species was insignificant. Additionally, the water extraction process excluded the possibility of BC-catalyzed oxidation. The absence of sulfate formation in dark conditions ruled out the involvement of direct H₂O₂ oxidation and organic peroxide oxidation pathways. The concentrations of TMI did not exhibit a consistent relationship with the sulfate formation observed in both BB_F–NaCl and BB_A–NaCl droplets (Fig. S3), suggesting that the TMI-catalyzed oxidation pathway may not be responsible for the observed phenomenon. Therefore, the most probable reason for the enhancement of sulfate formation by BB–NaCl droplets over NaCl droplets would be the photosensitizing components. Given the complexity and the lack of a method to quantify PSs in BB aerosols, using the total TOC concentration as an upper limit for estimating PS concentration is considered a compromise that allows systematic comparison. The sulfate formation reported here can be considered the lower limit of photosensitizing capacity. Our goal is to compare the photosensitizing ability in different chemical systems but not to quantify their absolute values.

State-of-the-art mass spectrometry analysis including UHPLC–Orbitrap MS and GC × GC–MS showed the existence of possible PSs such as PAHs (e.g., fluoranthene, pyrene, cyclopenta[cd]pyrene, 4-methylphenanthrene, benzo[a]pyrene, perylene; Table S2) and aromatic carbonyls (SyrAld, VL, 3,4-dimethoxybenzaldehyde, acetophenone, acetosyringone; Table S2). Photosensitizing components can directly or indirectly (by forming secondary oxidants in the presence of oxygen) oxidize S(IV) to S(VI). X. Wang et al. (2020) proposed a direct oxidation process of S(IV) to sulfate by excited triplet states of photosensitizers (³PS*). To explore the contribution of the direct ³PS* oxidation to sulfate formation, we performed the same sets of experiments in N₂-saturated condition, shown in Fig. S4. Under N₂-saturated conditions, secondary oxidants such as HO2• and the OH• oxidation pathway can be ruled out due to the lack of oxygen. Chlorine radicals from droplets can also react with dissolved SO₂ to generate sulfite radicals, but O₂ is required to form sulfate. Despite initial molecular oxygen in the droplets possibly also participating in sulfate formation under N₂-saturated conditions, its contributions are likely minimal. Consequently, the sulfate formed under N₂-saturated conditions can be considered the upper limit of direct ³PS* oxidation. The BB–NaCl droplets showed a direct ³PS* oxidation contribution of only 3.6 % to 22.7 %, highlighting the predominant role of secondary oxidants (Tang et al., 2023). For BB_F–NaCl droplets, the contribution of direct ³PS* followed the trend of WS_F–NaCl (22.7 %) > RS_F–NaCl (15.7 %) > CS_F–NaCl (7.0 %), while for BB_A–NaCl droplets, WS_A–NaCl (10.2 %) > CS_A–NaCl (6.7 %) > RS_A–NaCl (3.6 %) was observed. In summary, regardless of whether fresh or aged, the secondary oxidants triggered by indirect ³PS* oxidation were the main reason for sulfate formation, highlighting the importance of O₂ in ³PS*-mediated oxidation processes.

To investigate the aging effects on sulfate formation across various BB materials, we subjected the collected BB filters to OH radical aging by irradiating them with UV lights at wavelengths of 185 and 254 nm. This combination effectively generated OH radicals (Tang et al., 2023). Figure S5 exhibits the differences in sulfate formation rates in droplets of different fresh and aged BB materials. RS and WS show sulfate formation enhancement, while CS and IS show reduction after aging. Figure 2a shows that the 18 h sulfate enhancement factor (Sulfate_A / Sulfate_F) followed the trend of RS–NaCl (3.6) > WS–NaCl (1.4) > CS–NaCl (0.8) ≈ IS–NaCl (0.8), which is not consistent with the trends of sulfate formation for BB_F–NaCl or BB_A–NaCl, indicating the varying effects of the aging of BB materials. A similar trend was found for nγSO2, showing the highest and lowest sulfate enhancement for RS–NaCl (3.5) and IS–NaCl (0.7), respectively.

Bulk aqueous reactions using fresh/aged BB extracts without NaCl were performed to investigate the aging effects on the sulfate formation in the cloud phase (Fig. S6). As the experiment proceeded, sulfate concentrations accumulated while bisulfite concentrations decreased. Concurrently, the pH of the aqueous solution decreased from approximately 5.0 to 3.0, reflecting enhanced acidity. Bulk reactions have lower sulfate formation rates than droplets reactions, which may be attributed to the accelerated reactions induced by PSs at the air–water interface (W. Wang et al., 2024; Martins-Costa et al., 2022), as well as the differences in concentrations of S(IV) and NaCl between bulk and droplet surfaces. However, given that interfacial reactions are closely linked to particle size (Wei et al., 2020; Chen et al., 2022b), additional research is needed to better understand their influence. Our experiments involve large droplets of the size of 60 µm. The interfacial effects of such large droplets may not be evident. Future work to examine the interfacial effects of submicron- and nanometer-sized particles is needed.

In bulk experiments, all BB extracts have higher kSO42- values after aging. The increased sulfate formation of BB extracts after aging may be due to changes in their chemical compositions. Compared to RS_F (28.3 % for CHON- and 67.3 % for CHN+ in total intensity), RS_A has higher CHON- (36.1 %) and CHN+ (88.3 %) percentages (Figs. S7–S8). Zhao et al. (2022) observed a slight increase in the CHON percentage for RS from 53.4 % to 56.2 % after aging. A similar trend was observed for CS extracts, where the CHON- and CHN+ percentage increases from 26.7 % and 65.2 % to 31.5 % and 68.8 %, respectively, after aging. To semi-qualitatively distinguish BrC chromophores from the rest of dissolved organic carbon, we used the double-bond equivalent (DBE) value in the range of 0.5C ≤ DBE ≤ 0.9C as a criterion (Lin et al., 2018). For example, RS_A−BrC denotes the BrC chromophores in RS_A according to the above. Larger intensity fractions of CHON- species were found in RS_A−BrC (41.9 %) and CS_A−BrC (35.5 %) than RS_F−BrC (32.3 %) and CS_F−BrC (34.7 %). One of the key categories of CHON- is nitrated aromatics, which have been widely identified in lab-generated BB smoke (R.-J. Huang et al., 2022; X. Wang et al., 2017; Zhang et al., 2022; Xie et al., 2019) and field campaigns (Salvador et al., 2021; Mohr et al., 2013; Chen et al., 2022a). A series of CHON- species, e.g., C₆H₅NO₃, C₆H₅NO₄, C₇H₇NO₃, and C₈H₉NO₃, which were tentatively identified as nitrophenol, nitrocatechol, methyl-nitrophenol, and dimethyl-nitrophenol, have been detected in our BB extracts. Nitrophenol photolysis has been found to be a potential source of OH radicals (Sangwan and Zhu, 2018; Guo and Li, 2023; Cheng et al., 2009; Sangwan and Zhu, 2016), and it may partially contribute to the increase in sulfate formation by RS_A and CS_A.

Approximately 80 % of the CHN+ species identified exhibited a diatomic nitrogen composition in their molecular formula. The precise determination of the molecular structures of these compounds solely based on elemental composition is challenging due to the presence of stable isomers. However, the N bases, which contain two nitrogen atoms, can be attributed to various N-heterocyclic alkaloids (Fig. S9). For example, homologs of C₅H₆N₂(CH2)n were likely pyrazine, pyrimidine, or aminopyridine, which are composed of six-membered heterocyclic rings with N atoms and alkyl side chains (Lin et al., 2012; Laskin et al., 2009). C₅H₈N₂(CH2)n were likely alkyl-substituted imidazole compounds, featuring a five-membered heterocyclic ring with two nitrogen atoms as the core structure and alkyl side chains (Lin et al., 2012; Laskin et al., 2009). For C₇H₆N₂(CH2)n homologs, the core skeleton was C₇H₆N₂, with an AI_mod of 0.8, indicating its distinctive characteristics of compounds containing fused five-membered and six-membered rings, such as benzimidazole or indazole (Y. Wang et al., 2017). Redox-inactive heterocyclic nitrogen-containing bases, e.g., pyridine, imidazole, and their derivatives, have been shown to enhance the redox activity of the humic-like substance (HULIS) fraction by hydrogen-atom transfer, with the degree of enhancement directly correlated to their concentrations (Dou et al., 2015; Kipp et al., 2004). Thus, the increased CHN+ percentage may also contribute to the enhanced sulfate formation of RS_A and CS_A by acting as an H-bond acceptor to facilitate the ³PS*-mediated oxidation by generating more oxidants.

However, the CHON- and CHN+ percentages in WS_A were lower than in WS_F, indicating that the sulfate enhancement in WS_A was not due to the CHON and CHN species. Instead, CHO- accounted for a higher proportion in WS_A (68.5 %) and WS_A−BrC (68.9 %) than in WS_F (65.0 %) and WS_F−BrC (64.8 %). This aligns with a prior aerosol mass spectrometer (AMS) study showing increased CHO proportions in aged wheat burning emissions (Fang et al., 2017). We suppose that CHO- compounds, particularly photosensitizing compounds with carbonyl groups, would explain the difference in sulfate formation in WS extracts (Gómez Alvarez et al., 2012; Go et al., 2023; Felber et al., 2020; Fu et al., 2015). Therefore, we filtered the chemical formula of CHO- species from UHPLC–Orbitrap HRMS by applying the maximum carbonyl ratio (MCR) (Y. Zhang et al., 2021; K. Wang et al., 2024; Calderon-Arrieta et al., 2024; D. Liu et al., 2023), H/C, and O/C as well as the modified aromaticity index (AI_mod) to focus on potential PSs (Zherebker et al., 2022; Koch and Dittmar, 2006). In short, molecular formulae were classified into six groups, namely, condensed aromatics (AImod≥0.67), polyphenolic compounds (0.50 < AI_mod < 0.67), highly unsaturated and phenolic compounds (AImod≤0.5, H/C < 1.5), aliphatics (H/C ≥ 1.5, O/C ≤ 0.9, N=0), peptide-like compounds (H/C ≥ 1.5, O/C ≤ 0.9, N>0), and sugar-like compounds (H/C ≥ 1.5, O/C > 0.9); details can be found in Sect. S1. As aliphatics, peptide-like compounds, and sugar-like compounds are unlikely to be PSs, we exclude them as potential PSs. By applying further data filtration involving CHO-, condensed aromatics, polyphenolic compounds, and highly unsaturated and phenolic compounds based on the aforementioned criteria, as well as MCR ≥ 0.9 (which includes oxidized unsaturated and highly unsaturated compounds such as PS-like imidazole-2-carboxaldehyde (2-IC) and PAHs) (Y. Zhang et al., 2021), 52.6 % and 49.7 % of the compounds (by intensity) can be considered potential PSs in WS_A and WS_F, respectively. The main compositional difference lies in polyphenolic compounds, comprising 26.3 % and 21.8 % of WS_A and WS_F, respectively. Therefore, the higher sulfate formation in WS_A may be related to the higher contributions of the polyphenolic compounds, e.g., C₈H₈O₃.

To summarize, we propose that in aqueous reactions, the enhanced sulfate formation in CS_A and RS_A was likely due to the increased proportions (by intensity) of CHON and CHN species, potentially nitrophenols and N-heterocyclic compounds. Conversely, the increased sulfate formation in WS_A appears to be linked to a higher percentage of CHO species. However, an association between detailed chemical characteristics and sulfate formation is not able to be discerned in this study due to the complexity of the interactions between different chemical categories and difficulties in the interpretation of the coefficients. Future studies are needed to elucidate the relationships between sulfate formation and the chemical characteristics.

Unlike the droplet experiments where RS–NaCl has the highest sulfate enhancement factor after aging, aqueous reaction results (without NaCl) show a sulfate enhancement trend of WS > CS > RS > IS, suggesting that chloride may have an effect in the droplet experiments, especially in the RS–NaCl system. To evaluate the effects of chloride on sulfate formation, we conducted bulk reaction experiments using RS extracts as an example with 100–200 ppm NaCl additions, where the NaCl-to-TOC ratio ranged from 100:1 to 200:1 to match the range of 100:1 to 1000:1 in droplet experiments. Interestingly, incorporating NaCl yielded contrasting results for RS_F and RS_A (Fig. 3). While the addition of NaCl enhanced sulfate formation in RS_A, it showed the opposite trend in RS_F. The nature of the cations and ionic strength may affect the sulfate formation rate; however, previous studies have indicated that their effects are negligible (Zhang and Chan, 2024; Parker and Mitch, 2016). The opposite effect of the NaCl addition on RS_F and RS_A, to some extent, explains the significantly higher sulfate and SO₂ uptake coefficient enhancement factor for RS–NaCl in Fig. 2. Compared to the RS-based system, the NaCl control experiment showed minimum (but non-zero) sulfate formation (Table 1 and Fig. 3). On one hand, it supported the findings that chloride participated in the sulfate formation under light conditions but no sulfate formation under dark conditions (Cao et al., 2024; Tang et al., 2023; Zhang and Chan, 2024). On the other hand, the opposite trend of Cl⁻ effects on RS_F and RS_A reflects its complex interactions with BB extracts under light and air conditions. While direct reaction between S(IV) species and ³PS* may occur (X. Wang et al., 2020), other pathways, i.e., interactions among halide ions, PSs, and oxygen, should also be considered. Detailed mechanisms will be discussed later.

Statistical analysis using the Spearman correlation coefficients, as guided by the Shapiro–Wilk test (Table S4), revealed that the CHO, CHON, and CHN species exhibited significant correlations (|R| > 0.7) with the sulfate formation rate (p<0.01, Fig. S10). While PSs can be the main CHO species contributing to sulfate formation, N-containing organic compounds (NOCs), i.e., CHN and CHON species, may affect the chloride contribution to the sulfate formation rate. Therefore, we selected SyrAld and VL as model CHO species (PSs), pyrazine (Pyz) as a model CHN, and 4-nitrocatechol (4-NC) as a model CHON to elucidate how the N-containing species can alter the effects of chloride on sulfate formation rate by studying the CHO + Cl⁻, CHO + CHN + Cl⁻, and CHO + CHON + Cl⁻ systems. For SyrAld and VL, as the [Cl-]0/[PS]0 increases, kSO42- initially decreases and then increases. The initial decrease in kSO42- may be attributed to the quenching of ³PS* by electron transfer from Cl⁻ or the loss of OH radials by forming ClOH^•− through the reaction of OH• + Cl-↔ ClOH^•− (Anastasio and Newberg, 2007). Excessive chloride (e.g., 100 and 200 ppm) may generate Cl and OH radicals through photoexcitation in the presence of air and water and compensate for the loss of ³PS* or OH radicals. Previous studies have shown a contradictory influence of halides on the photosensitized oxidation of organic compounds or bisulfite. Parker and Mitch (2016) and Zhang et al. (2023) attributed the significantly higher photodegradation of dienes, thioethers, and acetaminophen to the formation of reactive halogen species generated by the reactions of PSs and halides. Zhang and Chan (2024) reported that [Cl-/PS]0 in the range of 1:2 to 4:1 did not lead to significant differences in sulfate formation. The differences between the current results and the aforementioned study might be attributed to the higher [Cl-/PS]0 ratio used in this study (up to 200:1), which may have been sufficient to initiate the relevant reactions, as well as the difference in photosensitizing capacities of the PSs studied (triplet quantum yield of 0.86 ± 0.05 for 2-IC and 0.21 ± 0.01 for VL) (Felber et al., 2021, 2020). Safiarian et al. (2023) reported that increasing chloride concentrations facilitated anthracene photosensitization by producing high-level reactive oxygen species (ROS). N. Wang et al. (2023) found that the effects of chloride on sulfate formation depended on the specific PS: enhancing sulfate production for benzophenone (BP) and 3,4-dimethoxybenzaldehyde (DMB) but decreasing it for 1,4-naphthoquinone.

Incorporating CHN species yielded a 2- to 3-fold increase in kSO42- due to the enhanced H transfer by CHN acting as an H-bond acceptor (Dou et al., 2015). With the addition of NaCl, the enhanced H-transfer effect by CHN was inhibited, possibly due to the consumption of ³PS* by Cl⁻. On the other hand, the addition of model CHON species into PSs decreased kSO42- due to the consumption of ³PS* by CHON species, in agreement with Y. Wang et al. (2023), who reported an increased effective quantum yield of 4-NC under co-photolysis with VL. Further addition of NaCl increased the kSO42-, possibly due to the consumption of 4-NC by reactive chlorine species (RCS; T. Wang et al., 2024), which, to some extent, reduced the loss of ³PS*. Generally, the addition of chloride increased kSO42- values of PS–CHON but decreased kSO42- values of PS–CHN. However, the ambient air is characterized by the presence of tens of thousands of chemical compounds. As a result, the interplay among this diverse array of species may occur in ways that exceed current understanding, necessitating additional research to investigate the interactions between different organic compounds more thoroughly.

A conceptual diagram of PS- and chloride-mediated ROS and RCS production in the oxidation of S(IV) to S(VI) is shown in Fig. 4. Initially, the PSs (e.g., SyrAld and VL) absorb solar radiation and produce the singlet state ¹PS*, which then undergoes a spin conversion through intersystem crossing, leading to the formation of the triplet state ³PS*. The ³PS* can react with molecular oxygen through energy transfer and generate singlet state ¹O₂*, while the ³PS* returns to ground state. The ¹O₂* can then transform to O2•- via electron transfer. The ³PS* can also react with an H donor (RH, e.g., organic acids, syringol, guaiacol; Table S3), leading to the formation of the alkyl or phenoxy radical (R•) and the ketyl radical (PSH•). R• can react with O₂ and form RO₂ radicals, while PSH• can transfer an H atom to O₂ and form HO2•, returning to its ground state PS. Additionally, ³PS* can react with an electron donor, e.g., Cl⁻, and form chlorine radicals and PS^•−. The PS^•− formed then reacts with O₂ and forms O2•-, which undergoes a series of reactions and forms HO2•, H₂O₂, and OH•. RO2• can further react with RO2• or HO2• and form organic peroxides (ROOR/ROOH). The abovementioned reactions are the main processes in the ROS pathway. Recently, Zhang and Chan (2024) proposed that the reactive chlorine species (RCS) would contribute to sulfate formation. Cao et al. (2024) proposed a mechanism of OH and Cl radicals formation by [Cl⁻–H₃O⁺–O₂] under light irradiation through an electron transfer process. Our results also demonstrate that the addition of Cl⁻ will affect the oxidation process of S(VI) (Figs. 3, S11–S13). Combining the above, the RCS pathway is shown in orange arrows in Fig. 4. The Cl• can be formed in two pathways, photoexcitation of the [Cl⁻–H₃O⁺–O₂] complex that generates Cl radicals in deliquescent BB–NaCl droplets or aqueous BB–NaCl solution (Cao et al., 2024) and ³PS*-mediated Cl• formation via electron transfer by Cl⁻ (Corral Arroyo et al., 2019). The Cl• radicals formed can then react with each other through radical–radical reactions and produce molecular Cl₂. The Cl^•− can also react with Cl⁻ or Cl2•-, forming Cl2•- or Cl₂. Cl• and Cl2•- can also react with OH and form HOCl. The ³PS* itself can also oxidize the S(IV) (e.g., dissolved SO₂ or bisulfite) to S(VI). However, significantly lower sulfate formation was found in the presence of N₂ than air (Fig. S4), highlighting the importance of secondary oxidants compared to direct ³PS* oxidation. As a consequence, these reactive species, e.g., OH^• / HO2• / RO2• / O2•- / ROOH / ROOR and Cl^• / Cl2•-, may all participate in the oxidation of S(IV) to S(VI). In addition, nitrogen-containing heterocyclic compounds such as pyrazine can act as H-bond acceptors and facilitate the H transfer, which then generates more ROS (Dou et al., 2015). Note that although ROS and RCS pathways both contribute to the oxidation from S(IV) to S(VI), they may act in competitive relationships due to the co-consumption of ³PS*. Therefore, different Cl effects may occur regarding various combinations of reactants (Fig. 3, promoting effect on RS_A and inhibiting effects on RS_F).