Ambient PM_2.5 sampling was conducted in both the urban center of Shanghai (Fudan University, Yangpu Jiangwan Campus; 31.344° N, 121.518° E) and a suburban area (Qingpu Dianshan Lake; 31.136° N, 121.092° E). The urban site is representative of typical urban environments, characterized by high population density and heavy traffic emissions. The suburban site, located approximately 62 km northeast of the urban site, is influenced by a mix of regional pollutant transport and local biogenic emissions. Sampling campaigns were carried out during summer (22 July to 2 August 2023) and winter (9 to 20 January 2024) at the urban site, and during summer (18 July to 2 August 2023) at the suburban site. Sampling was conducted separately during daytime (07:00–18:30, local time, UTC+8) and nighttime (19:00–06:30) to capture the day-night variations in the formation of OrgSs. It should be noted that winter samples were available only at the urban site; therefore, the seasonal comparison in this study is restricted to the urban campaign, whereas the urban-suburban comparison is discussed only for the summer period. All PM_2.5 samples were collected onto pre-baked quartz fiber filters (Whatman Inc.) using high-volume samplers (TH-1000C, Tianhong, China) at a flow rate of 1.05 m³ min⁻¹. Additionally, 11.5 h field blank samples were also collected by turning off the sampling pump. A total of 48 PM_2.5 samples and 4 field blank samples were obtained from the urban site, while 32 PM_2.5 samples and 2 field blank samples were collected from the suburban site. All filter samples were stored at -20 °C before analysis. The real-time monitoring of O₃, NO, NO₂, and meteorological parameters was also conducted during each sampling campaign (Table S1 in the Supplement).

A 24 cm² aliquot was excised from each PM_2.5 filter sample and subjected to ultrasonic extraction three times, each with 3 mL of methanol for 30 min. The combined extracts were filtered through a 0.45 µm PTFE syringe filter to remove insoluble particles, then concentrated to near dryness under a gentle stream of nitrogen. The residue was reconstituted in 80 µL of methanol containing internal standards (IS, 200 ppb D₁₇-Octyl sulfate). The resulting solution was centrifuged at 10 000 rpm for 10 min, and the supernatant was collected for subsequent mass spectrometry analysis.

Sample extracts were analyzed using an ultra-high performance liquid chromatography system coupled with an Orbitrap Exploris 120 mass spectrometer (UHPLC-Orbitrap MS, Thermo Scientific, Germany). The electrospray ionization source was operated in negative ionization mode with the following settings: capillary temperature of 320 °C, auxiliary gas flow at 10 units, sheath gas flow at 35 units, sweep gas flow at 2 units, and spray voltage at 3.0 kV. The scan range was m/z 60–900, with a resolution of 120 000 at m/z 200. Chromatographic separation was performed on an Acquity UPLC HSS T3 column (1.8 µm, 100 mm × 2.1 mm, Waters) with a C18 guard column (HSS T3, 1.8 µm). The flow rate was set to 0.3 mL min⁻¹, the column temperature was maintained at 40 °C, and the injection volume was 5 µL. The mobile phase used a binary gradient, with solvent A as 0.1 % formic acid in water and solvent B as 0.1 % formic acid in acetonitrile. The gradient program was as follows: 10 % B was held for 2 min, linearly increased to 30 % B over 2–25 min, further increased to 90 % B over 25–50 min, held at 90 % B for 10 min (50–60 min), then decreased to 10 % B over 60–61 min, followed by a 9 min column re-equilibration at 10 % B prior to the next injection. External mass calibration was performed every 3 d using a commercial calibration solution (Ultramark 1621, Thermo Fisher, Germany) covering an m/z range of 74–1922. Due to the limited volume of the final extract (80 µL), duplicate injections were not performed. Instead, the remaining sample solution from the first injection was retained for potential targeted analyses rather than systematic re-injection. All chromatographic peaks exhibited typical Gaussian shapes, indicating satisfactory column performance and separation efficiency. Field blank samples were extracted and analyzed following the same procedure to account for potential background contamination.

Organic carbon (OC) and elemental carbon (EC) were determined using a thermal/optical carbon analyzer (DRI model 2001, Desert Research Institute, USA) with the well-established IMPROVE-A thermal/optical reflectance protocol. Organic matter (OM) was calculated by multiplying the OC by 1.6 (Liu et al., 2018; Xing et al., 2013; Turpin and Lim, 2001). Water-soluble inorganic ions (e.g., SO42-, NO3-, Cl⁻, NH4+, K⁺, Na⁺, Ca²⁺, and Mg²⁺) were quantified by an ion chromatograph (940 Professional IC Vario, Metrohm) following procedures described by Cai et al. (2020).

The aerosol liquid water content (ALWC) and acidity (pH) were calculated using the ISORROPIA-II thermodynamic model. ISORROPIA-II was run in forward mode, assuming the particles were in a “metastable” state. The input parameters included environmental relative humidity (RH), temperature (T), and the inorganic components of the particle phase (SO42-, NO3-, Cl⁻, NH4+, K⁺, Na⁺, Ca²⁺, and Mg²⁺) (Hennigan et al., 2015; Weber et al., 2016; Guo et al., 2015; Song et al., 2018b). Gaseous NH₃, measured by an ammonia analyzer (G2103, Picarro, California, USA), was taken into consideration in the model during the summertime and wintertime campaigns at the urban site. The simulated NH₃ concentrations agreed well with the observations (Fig. S1a). Due to the lack of gaseous NH₃ data at the suburban site, an iteration-based method was applied to reduce potential underestimation of aerosol pH (He et al., 2025; Li et al., 2024b). Only measured aerosol NH4+ was used as the total ammonia input for the first run. The predicted gaseous NH₃ from each run was then added to the original NH4+ concentration to update the total ammonia input for the subsequent run. The sixth iteration produced the closest outcome to pH values constrained by observed NH₃ and was therefore adopted for suburban-site calculations (Fig. S1b, c). At the suburban site, the variance in the predicted NH4+ mass concentrations between successive iterations fell below a threshold of 0.01 by the sixth iteration, supporting the convergence and applicability of the iterative approach. Moreover, significant correlations between the results of the first run and the sixth run were observed for pH (R2= 0.95) and ALWC (R2= 0.99, Fig. S1), further indicating the stability and reliability in estimating the pH and ALWC by ISORROPIA II.

Raw data were acquired using Xcalibur software (V2.2; Thermo Scientific). The HPLC-HRMS data were processed with MZmine-2.33 software (http://mzmine.github.io, last access: 3 April 2026) to obtain the m/z ratios, formulas, retention times, and peak areas of detected organic compounds (Wang et al., 2017; Hu et al., 2016; Cai et al., 2024). The molecular formula assignment was mainly based on detected m/z values with a mass tolerance of ±2 ppm. The compounds assigned as C_1–40H_1–100O_1–20N_0–4S_0–2 with s= 1 or 2 will be collectively referred to as organosulfur compounds, including CHOS (n= 0) and CHONS (n= 1 or 2).

For chemical formula C_cH_hO_oN_nS_s, the double bond equivalent (DBE) value, representing the degree of unsaturation, can be calculated by Eq. (1): 1DBE=2c+2+n-h2 The aromaticity equivalent (Xc) was introduced by Yassine et al. (2014) for the identification and characterization of monocyclic and polycyclic aromatic compounds. The Xc for compounds containing only carbon, hydrogen, nitrogen, oxygen, and sulfur can be calculated following Eq. (2): 2Xc=3DBE-i×o+j×s-2DBE-i×o+j×s where i and j correspond to the fractions of oxygen and sulfur atoms involved in the π-bond structures of a compound, respectively. Here, i=j= 0.5 was applied for the calculation of Xc values of compounds detected in ESI-, as carboxylic compounds with p=q= 0.5 are preferably ionized in negative mode (Tong et al., 2016; Wang et al., 2017). If DBE ≤i×o+j×s, then Xc was defined as zero. The ranges of 2.50 ≤Xc< 2.71, 2.71 ≤Xc< 2.80, 2.80 ≤Xc< 2.83, and 2.83 ≤Xc are proposed as unambiguous minimum criteria for the presence of mono-, di-, tri-, and polycyclic aromatics, respectively (Yassine et al., 2014).

Given the large number of detected OrgSs, it is impractical to evaluate the ionization efficiencies of each compound. In this study, qualitatively comparing the variation trends of different compounds detected by HR-MS is a commonly used and valid approach for characterizing the overall molecular composition of OA across different sites or seasons (Zhang et al., 2024; Wang et al., 2024b; Cai et al., 2020). Accordingly, despite the potential influence of matrix effects, analyzing and comparing the number or signal intensity percentages of different OrgS categories can, to some extent, reflect the variability in the molecular composition of OrgSs under varying atmospheric conditions. To further assess the impact of matrix effects in different samples, we spiked all samples with 200 ppb of D₁₇-Octyl sulfate as an internal standard (IS). The IS intensity exhibited minimal variability across different samples, indicating that matrix effects on ionization efficiency can be considered negligible. The abundance of each compound refers to its chromatographic peak area, and all detected abundances were blank-corrected. To facilitate more accurate comparisons, the relative abundance of each OrgS compound was defined as the ratio of its blank-corrected peak area to that of IS (set as 100 %) within the same sample. We note that the OrgS results reported here are based on IS-normalized peak areas rather than absolute mass concentrations. Because authentic standards are unavailable for most OrgS species and ESI response factors vary among different molecular structures, the present non-target dataset is mainly suitable for comparing molecular composition and relative abundance patterns, rather than for direct concentration-based comparison with other studies and model evaluations. Future work using authentic standards or improved semi-quantitative approaches is needed to better constrain the absolute concentrations of OrgS species and facilitate direct comparison with model outputs.

To evaluate the associations between environmental variables and OrgSs in suburban summer, urban summer, and urban winter, nonmetric multidimensional scaling (NMDS) analysis was conducted based on Bray–Curtis distances in R using the vegan package (Jiang et al., 2022). For all three sample sets, the OrgS compounds were dimensionally reduced to two components (NMDS1 and NMDS2), and all stress values were less than 0.15. The selected environmental parameters, including meteorological parameters (T and RH), gaseous pollutants (NO₂, CO, SO₂, O₃, and NO), chemical tracers (EC, K⁺, and Cl⁻), inorganic nitrogen and sulfur species (SO42-, NO3- , and NH4+) as well as ALWC and pH, were fitted onto the ordination plot to evaluate the key environmental drivers influencing the molecular distributions of OrgSs. The significance (p-values) of the correlations between the environmental parameters and the NMDS dimensions was assessed using 999 permutations. Only the factors significantly correlated with the NMDS dimensions were reserved and could be considered as the possible drivers associated with molecular distribution. Score and loading plots were constructed according to NMDS variables from each OrgSs compound (gray triangles and circles). The potential driving factors associated with the molecular distribution of OrgSs were indicated by arrows, with their direction and angle representing the relationship between the variables and each dimension. Spearman correlation between the sum-normalized peak areas of individual molecules and some important environmental variables and chemical tracers was performed in R, and then VK diagrams were plotted for each variable based on the Spearman correlation coefficients (Kellerman et al., 2014). Molecules found in at least four samples were adopted for correlation analysis. A false discovery rate-adjusted p-value was applied to avoid errors arising from using a large dataset.

Accurate prediction of volatility is primarily based on the structural information of the organic compounds; however, it is often difficult to obtain from field measurements. Li et al. (2016) predicted C0 as a function of elemental composition that is often determined by soft – ionization high – resolution mass spectrometry (Wang et al., 2024b; Xie et al., 2021, 2020). The parameter of log⁡10C0=fc,o,n,s to be applicable to the sulfur–containing compounds can be calculated as shown in Eq. (3): 3log⁡10C0=nC0-nCbC-nObO-2nCnOnC+nObCO-nNbN-nSbS where nC0 is the reference carbon number; nC, nO, nN, and nS denote the numbers of carbon, oxygen, nitrogen, and sulfur atoms, respectively; bC, bO, bN, and bS denote the contribution of each atom to log⁡10C0, respectively; and bCO is the carbon-oxygen nonideality (Donahue et al., 2011). According to the values of C0, OrgSs can be classified as (1) volatile organic compounds (VOC; C0>3× 10⁶ µg m⁻³), (2) intermediate volatility OC (IVOC; 300 < C0<3× 10⁶ µg m⁻³), (3) semivolatile OC (SVOC; 0.3 < C0 < 300 µg m⁻³), (4) low-volatile OC (LVOC; 3 × 10⁻⁴ < C0 < 0.3 µg m⁻³), and (5) extremely low-volatile organic compounds (ELVOC; C0< 3 × 10⁻⁴ µg m⁻³) (Donahue et al., 2011; Murphy et al., 2014). The peak area-weighted average log⁡10C0 can be calculated as 4log⁡10C0avg=∑log⁡10C0i×Ai∑Ai Here, Ai is the peak area for each individual compound i.

By using UHPLC-Orbitrap MS analysis, a total of 1964, 1914, and 2689 organosulfur formulas were detected in suburban summer, urban summer, and urban winter aerosol samples, respectively. The intensity and number of detected particulate-phase OrgSs were significantly higher in winter than in summer, which aligns with the seasonal variations in PM_2.5 and OM concentrations (Table S1). Nevertheless, the comparable OM-normalized OrgS intensities (OrgS / OM ratios) between urban summer and winter (Fig. S2) indicate that the elevated OrgS intensity in winter largely accompanied the increase in bulk OM rather than reflecting preferential enrichment of OrgSs. Accordingly, the larger number of formulas detected in winter may may partly result from higher OM loading and associated detection-threshold effects. Moreover, formula-based volatility analysis showed that the relative peak area-weighted contribution of SVOC to total OrgSs increased from 37 % in summer to 41 % in winter (Fig. S3), consistent with the enhanced partitioning of semivolatile OrgSs into the particle phase at lower temperatures (Huang et al., 2019). Further class-specific analysis revealed that the SVOC contribution to CHOS increased from 47 % in summer to 54 % in winter, whereas the ELVOC contribution to CHONS increased from 81 % in summer to 90 % in winter. This increased relative importance of low-volatility CHONS may also be associated with seasonal variations in precursor sources and atmospheric chemical processing. Overall, these results suggest that the wintertime particle-phase OrgS composition may be partly influenced by temperature-dependent gas–particle partitioning and seasonal variations in sources and chemical processing (Li et al., 2016; Xie et al., 2021). Because only particle-phase samples were analyzed, the relative contributions of chemical formation versus gas-particle partitioning could not be quantitatively distinguished.

The identified OrgSs can be classified into six categories: CHOS₁, CHOS₂, CHON₁S₁, CHON₁S₂, CHON₂S₁, and CHON₂S₂. CHOS₁ and CHON_1–2S₁ account for ∼ 88 % of the molecular number and ∼ 96 % of the total intensity of OrgSs (Table S3), suggesting their dominant roles in OrgS composition. Notably, the intensity contribution of CHOS₁ in suburban summer was lower than those in urban summer and winter, whereas CHON_1–2S₁ showed an opposite trend. This disparity underscores the distinct influences of geographical location and seasonality on OrgS composition, as discussed in detail below. Furthermore, as many as 79 %–92 % of OrgS species contained enough oxygen atoms to enable the assignment of -OSO₃H and -ONO₂ groups (e.g., (4s+3n) /o≤ 1) in their formulas, suggesting that they were potential OSs or nitrooxy-OSs (NOSs), which is consistent with previous studies (Cai et al., 2020; Wang et al., 2016). However, this formula-based criterion provides only a tentative classification and does not constitute unambiguous structural confirmation. For compounds with available MS/MS spectra, OS / NOS assignments were further conducted based on diagnostic sulfate- and nitrate-related fragment ions (e.g., SO3⚫- m/z 79.9574, HSO3- m/z 80.9652, HSO4- m/z 96.9601, and NO3- m/z 61.9884) (Hettiyadura et al., 2015; Riva et al., 2016b). Specifically, OSs were selected based on compounds with O/S ≥ 4 and HSO4- (m/z 96.9601) or HSO3- (m/z 80.9652) fragments observed in their corresponding MS/MS spectra (Figs. S4, S5), whereas CHONS species satisfying (4s+3n) /o≤ 1 and exhibiting both NO3- (m/z 61.9884) and SO3⚫- (m/z 79.9574) fragments were defined as NOSs (Figs. S6). These compounds subjected to MS/MS analysis accounted for more than 50 % of the total intensity of all potential OSs / NOSs (Table S2). It should be noted that many low-abundance CHOS and CHONS species could not trigger reliable data-dependent MS/MS acquisition and were therefore classified as potential OSs or NOSs based solely on elemental composition, potentially leading to an overestimation of the total OS / NOS intensity.

The detected OSs were classified into four groups based on their potential precursors, including BVOC-derived OSs (e.g., isoprene/monoterpene/sesquiterpene-derived OSs and other BVOC-derived OSs from the precursors of green leaf volatiles), anthropogenic VOCs (AVOC)-derived OSs from the precursors of PAHs and anthropogenically emitted alkanes, multiple-sources-derived OSs, and OSs with unidentified precursors. The potential precursors of multiple-sources-derived OSs include carbonyl compounds (e.g., glyoxal and glycolaldehyde), unsaturated fatty acids, and long-chain alkenes emitted from biogenic and/or anthropogenic sources (Passananti et al., 2016; Shang et al., 2016; Fu et al., 2008). The precursor assignment of OSs was performed by matching the measured OS molecular formulas to the published OSs whose precursors have been verified. This approach has been widely used because its feasibility relies on the high mass resolution of HR-MS, which enables reliable mass peak detection and the assignment of formulas within a narrow mass tolerance (Cai et al., 2020; Wang et al., 2019; Yang et al., 2023). Details of these OS formulas with the determined precursors are listed in Tables S6–S10.

Figure 3e shows the intensity distribution of OSs derived from various precursors among the three campaigns. A total of 270 OS molecular formulas could be assigned to potential precursors, accounting for approximately 70 % of the total CHOS intensity. Among them, BVOC-derived OSs contributed 32 % ± 13 % in suburban summer, 30 % ± 12 % in urban summer, and 28 % ± 6.7 % in urban winter, whereas AVOC-derived OSs accounted for 16 % ± 4.6 %, 18 % ± 4.7 %, and 20 % ± 2.4 %, respectively. These results are consistent with the molecular signatures discussed above, suggesting relatively stronger biogenic contributions in suburban and urban summer, but enhanced anthropogenic influences in urban winter. Monoterpene-derived OSs constituted the dominant fraction of BVOC-derived OSs. Several highly abundant formulas of monoterpene-derived OSs, including C₉H₁₆O₆S (m/z 251.0595), C₁₀H₁₈O₅S (m/z 249.0802), and C₉H₁₆O₇S (m/z 267.0544), and C₁₀H₁₈O₇S (m/z 281.0701), have been frequently reported in ambient aerosols and are commonly attributed to acid-catalyzed reactions of monoterpene oxidation products (Wang et al., 2020b; Brüggemann et al., 2020b; Surratt et al., 2008). By contrast, isoprene-derived OSs contributed a much smaller fraction to the total CHOS intensity than monoterpene-derived OSs.

For identified AVOC-derived OSs, aromatic-derived OSs contributed only minor fractions to the total CHOS intensity in suburban summer (1.3 % ± 0.4 %), urban summer (1.5 % ± 0.6 %), and urban winter samples (1.9 % ± 0.7 %), whereas long-chain alkane-derived OSs accounted for much larger fractions (15 % ± 4.5 %, 17 % ± 4.6 %, and 18 % ± 2.4 %, respectively). Polycyclic aromatic hydrocarbons have been recognized as important precursors of aromatic OSs based on laboratory and field evidence (Riva et al., 2015; Kundu et al., 2013). Aromatic OSs with benzyl and polycyclic aromatic carbon backbones, such as C₆H₆O₄S, C₇H₆O₄S, C₇H₈O₄S, C₈H₈O₄S, and C₉H₁₂O₅S derived from the photooxidation of naphthalene and 2-methylnaphthalene (Riva et al., 2015), have been widely observed in urban and semi-rural fine aerosols worldwide (Jiang et al., 2022; Yang et al., 2023; Wang et al., 2019; Brüggemann et al., 2020b), and were also detected in our samples. Nevertheless, despite their frequent occurrence, only a few aromatic OSs with relatively low intensity could currently be unambiguously classified. In contrast, aromatic CHOS species with Xc≥ 2.5 accounted for 16 % ± 3.8 % and 19 % ± 2.9 % of total CHOS species by number in suburban and urban summer, respectively, and increased to 22 % ± 3.4 % in urban winter, suggesting enhanced anthropogenic emissions in winter. The intensity fractions of aromatic CHOS in Shanghai (7 %–18 %) were comparable to those reported for Guangzhou (9 %–20 %) (Jiang et al., 2022), further underscoring a non-negligible influence of anthropogenic emissions on the molecular composition of atmospheric OrgSs in Chinese megacities. In addition, multiple-source OSs contributed similarly to total CHOS intensity in suburban summer (22 % ± 7.2 %) and urban summer (23 % ± 6.8 %), but exhibited a moderately higher contribution in urban winter (27 % ± 6.9 %). This may suggest a greater influence of anthropogenic emissions on their sources and formation processes in winter.

The precursors for 28 % ± 15 % of OS molecular formulas cannot yet be confidently recognized. Among these, subgroup B, defined as organosulfates with C > 8, DBE < 3, and 3 < O < 7 for CHOS, contributed 6.7 % ± 3.3 %, 16 % ± 8.3 %, and 8.7 % ± 2.5 % to the total CHOS intensity in suburban summer, urban summer, and urban winter, respectively. Subgroup B is characterized by a high molecular weight, long alkyl chains, and a low degree of unsaturation and oxidation. Tao et al. (2014) speculated that the precursors of subgroup B could be long-chain alkanes from traffic emissions. Long-chain alkanes can undergo rapid photooxidation under typical urban conditions (Yee et al., 2013; Lim and Ziemann, 2009), yielding hydroxylated or carbonylated intermediates that were further esterified to form alkyl OSs in acidic aerosol. Riva et al. (2016b), through smog-chamber photooxidation experiments of long-chain alkanes, revealed that alkane-derived OSs can also form via gaseous epoxide precursors with subsequent acid-catalyzed reactive uptake onto sulfate aerosols and/or heterogeneous reactions of hydroperoxides (Fig. 6d). Moreover, the formation of OSs from heterogeneous reactions of gaseous SO₂ with unsaturated fatty acids (USFA) was also important for these highly saturated OSs (Zhu et al., 2019; Shang et al., 2016). Large amounts of USFA have been observed in high-temperature cooking emissions, especially Chinese-style cooking with more frying, which is among the important emission sources in urban areas (Zhu et al., 2019; Zhao et al., 2015). Accordingly, subgroup B contributed more to total CHOS at the urban site (16 % ± 8.3 %) than at the suburban site (6.7 % ± 3.3 %) during the summer, indicating a greater influence of urban hydrocarbon emissions. Moreover, the intensity fraction of subgroup B showed positive correlations with RH (Fig. S10a), suggesting that humidity-driven heterogeneous or multiphase processing may facilitate the formation of these long-chain OSs. Although aromatic VOCs have been widely studied as drivers of particulate air pollution in densely populated regions like Shanghai (Han et al., 2023a; Gao et al., 2019), our results suggest that long-chain hydrocarbons and their corresponding OS products also play an important role in the formation of particulate organosulfurs in urban atmospheres.

In contrast to the relatively well-established precursor assignments for CHOS compounds, only a limited number of CHONS species could be linked to specific precursors (Tables S6–S10). Therefore, we further evaluated the potential precursor origins of CHONS species based on their Van Krevelen distributions, the precursor-based classification of NOSs, and molecular fingerprints from representative source samples. As shown in Fig. 4a–c, CHONS compounds, particularly NOSs, were predominantly distributed in the unsaturated aliphatic region (H / C≈ 1.5–2.0, (O - 3S - 2N) / C ≈ 0.1–0.6), suggesting that their precursors are mainly derived from the oxidation products of BVOCs, such as isoprene and monoterpenes (Surratt et al., 2008; Lin et al., 2012b; Hamilton et al., 2021), with additional contributions from traffic-related alkenes and long-chain hydrocarbon oxidation intermediates (Tao et al., 2014; Riva et al., 2016b). The distributions of suburban and urban samples in summer were broadly similar, indicating comparable dominant precursors and formation mechanisms. In contrast, urban winter samples exhibited a more constrained distribution toward lower (O - 3S - 2N) / C values, indicating a less oxygenated CHONS composition. This seasonal feature may partly reflect a greater contribution from NO₃-initiated oxidation under wintertime conditions characterized by elevated NO_x, lower temperatures, and weaker photochemical activity (Brown and Stutz, 2012; Hamilton et al., 2021). Figure 5e shows that seasonal variability was more influential than spatial differences in shaping NOS composition. Monoterpene-derived NOSs predominated in most samples (11 %–78 %), underscoring monoterpenes as important precursors of NOSs, which aligns with previous studies (Surratt et al., 2008; Lin et al., 2012b). In contrast, the relative contribution of isoprene-derived NOSs decreased markedly in urban winter, likely due to reduced isoprene emissions under low-temperature conditions (Guenther et al., 2012; Li et al., 2011). Furthermore, monoterpene-derived NOSs exhibited noticeable day-night variability, especially in summer, with generally higher contributions at night (Figs. 4e and S9), which is consistent with an important role of nighttime NO₃-initiated oxidation in NOS formation (Ng et al., 2017; Brown and Stutz, 2012).

To further constrain the possible precursor origins of CHONS, we compared their molecular characteristics with those reported for representative source samples in previous studies, including coal combustion organic aerosols (CCOAs), biomass burning organic aerosols (BBOAs), vehicle emissions, tunnel aerosols, and emissions from nonroad excavators and ships (Cui et al., 2019; Tang et al., 2020). As shown in Fig. 4a–d, substantial overlap in CHONS species was observed between our field samples and previously reported source samples, particularly in the high H / C region. However, ambient samples extended toward more oxidized regions, indicating that CHONS species are predominantly formed via secondary processes rather than direct primary emissions. The most abundant CHONS species in Shanghai were dominated by unsaturated aliphatic compounds (Fig. 4a–d), whose overall characteristics more closely resembled those of tunnel aerosol samples that had likely undergone atmospheric aging processes. By contrast, CHONS compounds in fresh vehicle emissions were abundant in aromatics, with 59 % of the identified CHONS species showing Xc≥ 2.50 (Table S12). Likewise, a large number of aromatic and highly unsaturated CHONS species were detected in both BBOAs and CCOAs, but their molecular characteristics differed substantially from the dominant CHONS components observed in Shanghai. Notably, although approximately 36 %, 40 %, and 44 % of CHONS species by number could be classified as aromatic organosulfur compounds in suburban summer, urban summer, and urban winter, respectively, most of these species exhibited relatively low abundances (Fig. 4a–d and Table S12). Thus, although combustion sources can directly emit a large variety of CHONS compounds, the low-oxidation and aromatic CHONS species that dominate the source samples contributed only weakly to the MS intensity of the ambient CHONS pool observed here. These results suggest that the CHONS in Shanghai may be less influenced by primary emissions and more likely driven by secondary formation processes, such as secondary formation via combustion-emitted precursors. Similar features were also observed in the comparison between CHOS species detected in our field samples and source samples (Fig. 3 and Table S11).

Taken together, the precursor assignment and source comparison highlight the important role of secondary formation from both biogenic and anthropogenic precursors in shaping the molecular composition of particulate OrgSs in Shanghai. Given that isoprene- and monoterpene-derived OSs and NOSs are representative BVOC-derived species and are sensitive to variations in precursor emissions, oxidant regimes, NO_x / NO₃ chemistry, and sulfate-related multiphase processing, their spatial, seasonal, and day-night variability was further examined.

During summer, isoprene/monoterpene-derived OSs at both suburban and urban sites exhibited distinct day-night variations, with significantly higher relative abundance during the day than at night (p< 0.05; Fig. 5a, b). This daytime enhancement is consistent with stronger photochemical processing of biogenic precursors under higher temperatures and more intense solar radiation, potentially involving ⚫OH-initiated oxidation (Surratt et al., 2010; Xu et al., 2015). In contrast, no significant day-night variation was observed in winter urban samples (p> 0.05), suggesting that the formation of these OS species in winter was less controlled by a single photochemical pathway and more influenced by multiple pathways under low-temperature, weak-radiation, and high-NO_x conditions. Notably, isoprene-derived OSs showed only weak seasonal differences, whereas monoterpene-derived OSs were more abundant in winter than in summer (Fig. 5a, b). This appears to suggest that enhanced nighttime and/or heterogeneous production in winter may partially compensate for the decline in photochemical efficiency. Recent studies have suggested that OSs could also be formed through heterogeneous or multiphase reactions between SO₂ and monoterpene-derived SOA or unsaturated fatty acids in the particle phase (Fig. 6b) (Yao et al., 2019; Zhu et al., 2019). The level of SO₂ during the winter campaign was 1.6–1.8 times higher than those during the summer campaigns (Table S1), suggesting the possible involvement of SO₂-related heterogeneous chemistry in the wintertime enhancement of monoterpene-derived OSs, although this pathway cannot be confirmed from SO₂ concentrations alone. We cannot exclude the possibility that aqueous-phase processing, aerosol acidity, and temperature-dependent gas-particle partitioning also contributed to the observed seasonal variation. These SO₂-involved heterogeneous pathways may warrant consideration under high-organic peroxide and high-SO₂ conditions (Yao et al., 2019) as well as under weakly acidic or low-RH conditions, when acid-catalyzed aqueous-phase OS formation could be limited by weak aerosol acidity or slow particle-phase reaction kinetics (Liu et al., 2017; Shiraiwa et al., 2011). Further field observations of these OSs and intermediates indicative of pathways are needed to assess the plausibility and relative importance of the proposed pathways under different atmospheric conditions. In addition, anthropogenic combustion sources (e.g., biomass burning) may provide additional inputs of isoprene and monoterpenes, which could help sustain OS formation under cold conditions (Akagi et al., 2011; Gilman et al., 2015; Yang et al., 2025). Overall, these results suggest that photochemistry was the dominant driver of isoprene/monoterpene-derived OS formation in summer, whereas wintertime formation was governed by a more complex, multi-pathway regime involving nocturnal NO3⚫ chemistry and heterogeneous or multiphase processes.

The day-night patterns of monoterpene/isoprene-derived NOSs differed markedly from those of OSs. The day–night variations of isoprene-derived NOSs were not obvious across the three field campaigns (Fig. 5c), suggesting that their formation may be jointly influenced by daytime ⚫OH oxidation and nighttime NO3⚫ oxidation. In contrast, monoterpene-derived NOSs exhibited obvious nighttime enhancements (p< 0.05; Fig. 5d), with significantly higher relative abundances at night than during the day in both suburban and urban summer samples (p< 0.01). This trend underscores that ⚫NO₃-initiated nighttime oxidation is likely an important pathway for their formation (Fig. 6e), consistent with previous field and laboratory studies highlighting the important role of NO3⚫ oxidation of monoterpenes in NOS formation (Hamilton et al., 2021; Surratt et al., 2008; Yang et al., 2023). Elevated nighttime NO_x levels likely further favored monoterpene-NOS formation via nighttime NO3⚫ oxidation (Table S1). Moreover, the lower daytime relative abundance of monoterpene-derived NOSs could be exacerbated by daytime photolysis and decomposition processes (He et al., 2011; Wang et al., 2023). Spatial and seasonal variations of monoterpene-derived NOSs further highlight these possible drivers (Figs. 5d and S9). In summer, monoterpene-derived NOSs were more abundant in suburban than in urban samples, likely reflecting a stronger biogenic influence and the greater importance of monoterpene precursors at the suburban site in the YRD (Ma et al., 2022; Wang et al., 2020a). Notably, their overall relative abundance in urban samples was higher in winter than that in summer, suggesting that a more favorable oxidation environment under high-NO_x conditions may promote the production efficiency of monoterpene-derived NOSs, partly compensating for the typically lower biogenic monoterpene emissions in winter (Wang et al., 2020b). Additionally, previous studies have shown that monoterpenes can also be emitted from biomass burning, besides their well-known biogenic emission (Akagi et al., 2011; Gilman et al., 2015), which may further contribute to the accumulation of monoterpene-derived NOSs during the cold season.

Atmospheric OrgSs in Shanghai originate from diverse sources, with secondary formation serving as the dominant contributor. Despite the identification of multiple reaction pathways for OS formation, their formation mechanisms in the real atmosphere remain incompletely understood. Among the known pathways, the acid-catalyzed ring-opening reaction of epoxides derived from the oxidation of VOCs is widely recognized as a kinetically feasible and key mechanism for the atmospheric formation of OSs (Fig. 6a) (Brüggemann et al., 2020b; Cai et al., 2020). For instance, previous studies have demonstrated that high concentrations of isoprene-derived OSs observed in Beijing and Guangzhou were mainly produced through the acid-catalyzed ring-opening of isoprene epoxydiol intermediates (Wang et al., 2018b; Jiang et al., 2022). Furthermore, NOSs can be formed via the nighttime NO₃-initiated oxidation of monoterpenes on acidic sulfate seed particles (Surratt et al., 2008; Ng et al., 2017). The formation and molecular distribution of OS and NOS products may be influenced by multiple environmental factors, including precursor abundance (e.g., organic precursors and anthropogenic pollutants such as NO_x and SO₂), aerosol acidity, RH, and oxidant levels. To disentangle how these complex variables impact the molecular distribution of OrgSs across different spatiotemporal settings (suburban summer, urban summer, and urban winter), NMDS analysis of OrgSs was conducted (Fig. 7). Overall, samples from different seasons and urban/suburban sites exhibit distinct separation in the two-dimensional ordination space, indicating that the molecular composition of OrgSs is highly sensitive to changes in precursor sources, oxidation conditions, and the particulate-phase reaction environment.

For suburban summer samples, Cl⁻ was positively correlated with the NMDS1 dimension, whereas tracers indicating combustion emissions (e.g., EC and CO), biomass burning (e.g., K⁺), and secondary inorganic nitrogen species (e.g., NO3-), were mainly distributed on the negative direction of NMDS1 (Fig. 7a). This suggests that the NMDS1 dimension may reflect the integrated influence of different air mass sources on the molecular composition of OrgSs. It should be noted that Cl⁻ can originate from both combustion sources (e.g., coal smoldering) and marine aerosols (Wu et al., 2024). Cl⁻ was positively correlated with Na⁺, a commonly used marine tracer (R2= 0.69, p< 0.01; Fig. S11), suggesting a potential contribution from marine air masses or sea-salt input. Nevertheless, the observed Cl⁻ / Na⁺ mass ratios were far below the seawater value of approximately 1.80 (Keene et al., 1986; May et al., 2016), indicating that particulate Cl⁻ was likely affected by chloride depletion during atmospheric aging (Yao et al., 2003; Hsu et al., 2007). Given that the sampling site was situated in a coastal region and that Cl⁻ exhibits an opposite trend to typical combustion tracers like EC, Cl⁻ is more likely to represent the influence of marine air masses or sea-salt input. Therefore, the divergent trends of Cl⁻ and EC along the NMDS1 axis may highlight the distinct impacts of anthropogenic versus marine sources on the molecular composition of OrgSs in suburban summer. Notably, the suburban summer samples exhibited a pronounced day-night separation along the NMDS2 dimension. Daytime samples primarily clustered along the positive axis of NMDS2 and were associated with T and O₃, whereas nighttime samples aggregated in the negative direction and were associated with RH and NO₂. These associations are consistent with a greater influence of high temperature and photochemical oxidation processes on the molecular composition of OrgSs during the daytime. In summer, elevated daytime temperatures facilitate the emission of VOCs, especially BVOCs, while enhanced photochemical activity and O₃ levels may promote the oxidation of VOCs and the subsequent formation of OSs. In contrast, RH and NO₂ appeared to be the environmental factors most strongly associated with nighttime OrgS composition. On one hand, RH may affect OS formation by influencing the uptake of gaseous precursors like SO₂ by VOCs (Ye et al., 2018; Li et al., 2024a; Nestorowicz et al., 2018). On the other hand, NO₂ and its associated nighttime oxidative system (e.g., NO₃ radical) may jointly influence the formation of both OSs and NOSs by modifying the oxidation pathways and product branching of organic precursors, as well as through competition or coupling with heterogeneous processes involving sulfate and acidic particles (Hamilton et al., 2021; Fan et al., 2022).

For urban summer samples, primary combustion sources related to anthropogenic activities also appeared to play an important role in shaping the molecular composition of OrgSs, whereas the influence of marine air masses was comparatively weak. In contrast to the prominent role of inorganic nitrogen in suburban summer samples, SO42- showed a stronger association with the molecular distribution of urban OrgSs (Fig. 7b). This observation aligns with previous studies in Shanghai and other polluted regions that have highlighted the importance of sulfate-related processes in OS formation (Yang et al., 2023; Wang et al., 2018b). SO42- acts not only as a critical reactant but also as a regulator of particle acidity and ALWC, which in turn affects the reactive uptake and heterogeneous conversion efficiency of precursors (Wang et al., 2021b). The distinct influences of different anthropogenic emission species on OrgSs distribution between urban and suburban summer underscore the varying impact of human activities on OS formation across different environments. Nevertheless, the environmental factors associated with day-night variations in OrgS composition were generally similar between the two sites. Daytime samples clustered with T and O₃ along the positive axis of NMDS2, while nighttime samples aggregate with RH along the negative axis. This result suggests that site type had only a limited influence on the day-night distribution of OrgSs in summer. The main processes potentially governing the day-night variability of OrgSs appeared to be broadly similar across the two sites, mainly involving enhanced photochemical oxidation during the daytime and RH-dependent multiphase processing at night.

For urban winter samples, the potential factors influencing the molecular distribution of OrgSs were more complex than those in summer (Fig. 7c). In addition to the impact of primary anthropogenic emissions, inorganic nitrogen and sulfur species were significantly associated with wintertime OrgS composition, clustering along the negative axis of NMDS1. ALWC is also distributed in the negative NMDS1 direction, suggesting that it was another environmental factor associated with the wintertime molecular distribution of OrgSs. ALWC is influenced not only by RH, but also by the loading of hygroscopic inorganic components such as NO3- and SO42-. Under the humid and polluted conditions typical of winter, the accumulation of these species can significantly enhance ALWC and thereby favor aqueous-phase processing (Wang et al., 2021b). Previous mechanistic and field studies suggest that elevated ALWC may promote the mass transfer and reactive uptake of SO₂ and oxidized organic intermediates into the particle phase, while also providing a medium for subsequent aqueous-phase transformations (Li et al., 2025; Yao et al., 2019; Wang et al., 2021b). Aerosol pH was also significantly associated with wintertime OrgS composition. Lower pH can favor the reactive uptake of epoxides and other oxidative intermediates onto acidic sulfate particles and may facilitate aqueous-phase esterification via acid-catalyzed mechanisms (Wang et al., 2018b). Taken together with previous mechanistic evidence, the observed associations of wintertime OrgS composition with ALWC and aerosol acidity are consistent with the potential importance of multiphase and acid-catalyzed processing in the secondary formation and molecular evolution of urban OrgSs in winter. Overall, source-related associations differed between sites, with suburban OrgSs linked to marine and combustion influences and urban OrgSs more closely associated with anthropogenic emissions and sulfate-related processes. The molecular composition of OrgSs in summer was more closely associated with photochemical oxidation during the daytime and RH-dependent heterogeneous processes at night, whereas wintertime OrgSs appeared to be more strongly influenced by the combined effects of enhanced primary emissions, the accumulation of secondary inorganic species, and the synergistic regulation of the particle-phase reaction environment by ALWC and pH.