Winter sampling was conducted in CD, located in the SCB, from 14 December 2024–9 January 2025. The observation site was situated on the roof of the Civil Engineering Building at Southwest Jiaotong University's Jiuli Campus (30.7° N, 105.05° E) (Fig. S1 in the Supplement), at a height of 25 m above ground level. It was located in the central area of CD, near the North Second Ring Road, surrounded by residential areas, shopping malls, gas stations, and restaurants, During the sampling period, meteorological conditions in CD were stagnant with low wind speeds and high RH (Table 1). The SCB features a typical deep-basin topography (with a basin depth exceeding 2000 m), surrounded by mountains on all sides, which severely hampers the dispersion of air pollutants (Rao et al., 2026). In winter, the Qinling Mountains block the intrusion of cold air, allowing warm and moist air masses to accumulate within the basin, leading to high humidity, frequent fog, and recurrent temperature inversions (with a winter inversion frequency of 95.4 % and an inversion layer depth of 289.4 m) (Feng et al., 2020). During pollution episodes, the boundary layer is significantly compressed, with its height reduced by approximately 50 % (Lu et al., 2024), resulting in a stable atmospheric stratification that is extremely unfavorable for pollutant diffusion. The synergistic effects of increased emissions, stagnant meteorological conditions, and enhanced aqueous-phase reactions under high relative humidity (>80%) can readily cause extreme PM2.5 episodes (Shu et al., 2025), and the humid environment further promotes the hygroscopic growth and secondary transformation of particulate matter (Wang et al., 2021). In CD, the major sources of PM2.5 are vehicle emissions (25.6 %) and secondary sources (40.5 %) (Zhang et al., 2024). In winter, secondary inorganic ions account for 40 %–60 % of PM2.5 mass, of which approximately half is nitrate (Rao et al., 2026), providing abundant precursors for the secondary formation of NOCs (Li et al., 2023). The combined topographic-meteorological-emission characteristics impart unique regional representativeness to the study of gas-aerosol partitioning of NOCs under the humid and stagnant conditions typical of CD. Therefore, the observation site can effectively capture the influence of the basin-specific atmospheric chemical processes on the partitioning of WSON.

Continuous measurements of inorganic ions, WSON, methylamine (MA) and dimethylamine (DMA) in the gas and aerosol phases were conducted at the campus site. Online samples of the aforementioned compounds were collected via the In-situ Gas and Aerosol Compositions monitor (IGAC, model 63GA, Fortelice International Co., Ltd., Taiwan, China) coupled with an ICS-5000+ ion chromatograph (Thermo Scientific,USA) and a TOC/TN analyzer (model TOC/TN-L CPH, Shimadzu, Inc., Japan) for simultaneous measurements of online samples. WSON was measured with a time interval of 3 h, while other species were measured hourly. In this study, the equation for calculating WSON concentration is as follows: 1WSONg,p=WSTNg,p-WSINg,p where, g and p represent the gas phase and particle phase, respectively; WSTN represents total water-soluble nitrogen measured by TOC/TN analyzer; WSIN represents the total water-soluble inorganic nitrogen measured by IGAC–IC, when calculating this value, the average is taken over a 3 h WSON sampling intervals. The unit is µgNm-3. Due to the differing forms in which inorganic nitrogen exists in the gas phase and particulate phase, the calculation methods will vary accordingly, as follows: 2WSINg=[NH3]17+[HNO3]63+[HNO2]47×143WSINp=[NH4+]18+[NO3-]62+[NO2-]46×14. The gas-particle phase partition coefficient (Fp) for WSON is calculated as follows, similarly, similar calculations apply to amines: 4Fp=CpCg+Cp.

Five anions (SO42-, NO3-, NO2-, Cl-, F-), five cations (NH4+, Na+, K+, Mg2+, Ca2+), and two amines (MA and DMA) were measured in this study. PM2.5 and meteorological parameters (including T and RH) were measured by the E–BAM continuous environmental PM2.5 monitor (Met One Instruments, United States) and an automatic weather station (MILOS520, Vaisala, Inc., Finland), respectively. In this study, pollution periods (PP) were defined as days with a daily PM2.5>75µgm-3, clean periods (CP) as PM2.5<35µgm-3, and transition periods (TP) as PM2.5 between 35 and 75 µgm-3. The organic carbon (OC) and elemental carbon (EC) of the PM2.5 filter samples were analyzed by DRI Model 2015 Carbon Analyzer (Atmoslytic, Inc., Calabasas, USA) with IMPROVE_A protocol (Chow et al., 2007; Li et al., 2025).

To minimize background interference and systematic errors in the instrument, a HEPA filter was installed at the IGAC sampling port. Measurement results indicated that the background levels in the tube system were comparable to the blank concentration of deionized water. The background concentrations of WSOC and WSTN in the tubes were 0.049±0.030mgL-1 and 0.007±0.002mgL-1, respectively, significantly lower than the sample concentrations. Additionally, the collection efficiency of IGAC system for both gas and particle phase species is >73% (Lv et al., 2022a, b), and the data during the observation period were corrected accordingly. For detailed information on instrument applications and method accuracy tests, please refer to the Support Information Text S1.

The ALWC and pH of PM2.5 were estimated using the thermodynamic ISORROPIA-II model based on hourly data of aerosol chemical composition (including Na+, NH4+, K+, Ca2+, Mg2+, SO42-, NO3-, and Cl-), gaseous NH3 concentration, and meteorological parameters (T; RH) (Fountoukis and Nenes, 2007). The forward metastable mode was employed for the estimation due to its lower sensitivity to measurement errors compared to the reverse mode (Hennigan et al., 2015). Data corresponding to RH greater than 95 % were excluded on the basis that such conditions are associated with increased uncertainty in the model estimation of both ALWC and pH (Guo et al., 2015). The contribution of organic compounds to ALWC (ALWCorg) was determined using the mass concentration of organic matter (OM) which was calculated on the basis of OC (OM=OC×1.6) (Lv et al., 2022a, b), as described in detail in the Supporting Information Text S2.

To further elucidate the effects of pH and ALWC on the gas-particle phase distribution of organic amines, this study assumes that the gas and particle phases of amines are in thermodynamic equilibrium. ϵ(CH3NH3+) and ϵ((CH3)2NH2+), the theoretical values of the gas-particle phase distribution coefficients, are calculated using a thermodynamic model and analyzed via S-curve analysis with aerosol pH. The calculation equations are as follows: 5ϵ(CH3NH3+)=γH+γCH3NH3+10-pH+KaγCH3NH2HCH3NH2KaWiRT×0.987×10-141+γH+γCH3NH3+10-pH+KaγCH3NH2HCH3NH2KaWiRT×0.987×10-14

6ϵ((CH3)2NH2+)=γH+γ(CH3)2NH2+10-pH+Kaγ(CH3)2NHH(CH3)2NHKaWiRT×0.987×10-141+γH+γ(CH3)2NH2+10-pH+Kaγ(CH3)2NHH(CH3)2NHKaWiRT×0.987×10-14 where, Hi is the Henry's law constant for organic amine i at 298.15 K, Ka is acid dissociation constant at 298.15 K, γi represents the activity coefficient, which can be calculated using the E-AIM model; Wi is the ALWC of the particles; R is the gas constant, and T is the temperature. Detailed derivations of Eqs. (5) and (6) are provided in the Support Information Text S3.

The higher nitrogen oxidation ratio (NOR) and sulfur oxidation ratio (SOR) represent more efficient conversion of gaseous species into secondary aerosols. The calculation equations are as follows: 7NOR=[NO3-]NO3-+[NO2]8SOR=[SO42-][SO42-]+[SO2] where, [i] refers to the molar concentration of species i.

The temporal variations of meteorological parameters, major atmospheric pollutants during the whole campaign are shown in Fig. 1 and Table 1, respectively. During the campaign, the average PM2.5 concentration in CD was 76±38µgm-3 (Table 1), with 14 pollution days accounting for 51.9 % of the total sampling days. The average PM2.5 concentration during PP was 100±33µgm-3, with the highest pollution level reaching 199 µgm-3. Concurrently, NO3- concentrations (Table S1 in the Supplement) and contributions (Table S1) began to exceed those of SO42- and exhibited an overall increasing trend, indicating that NO3- has become the dominant inorganic salt in PM2.5. In contrast, SO42- remained relatively stable and was significantly lower in 2024 compared to previous years (Table S1). NH4+ has remained relatively stable since its decline began in 2016, as has Cl-. In this study, SO42-, NO3-, and NH4+ accounted for 7.2±3.1%, 22.8±6.6%, and 11.5±4.6% of the PM2.5 mass, respectively (Fig. 1b). Additionally, the observed concentration of OM in CD was 21.1±9.6 µgm-3 (Table 1), contributing 27.3±7.4% to the total PM2.5 mass. As shown in Table S1, previously reported values in the SCB also exceeded 25 %. These results clearly indicate that NO3- and OM are the major components of PM2.5 in the SCB, accounting for over 50 % of the total PM2.5 mass.

The temporal variation of WSONp (Fig. 1c) exhibited the same trend as PM2.5 (Fig. 1b), with a significant positive correlation between WSONp and PM2.5 (R=0.78, p<0.05, Fig. S2). WSONp contributed to 0.5 %–12.4 % of the PM2.5 mass, with the average value was 3 % (Fig. S3a–d). The concentration of WSONp (2.3±1.4µgNm-3) was more than twice that of WSONg (1.2±0.9µgNm-3) (Fig. 1c and Table 1), indicating that WSON in CD mainly exists in the particle phase, as observed at Mt. Fuji in Japan (Matsumoto et al., 2020). WSON contributed 9.3 % and 14.7 % to the TN in the gas phase and particulate phase, respectively. In CD, the WSON/WSOC ratios for the gas and particle phases were 0.05 and 0.15 (Fig. S3e–l), respectively. Notably, the WSON/WSOCp ratio increases as pollution worsens (Fig. S3i–l), indicating that more NOCs are enriched during the accumulation of OM. Our study shows that the wintertime WSONp concentration of PM2.5 in CD (2.3±1.4µgNm-3) was slightly higher than those in central Chinese cities such as Nanjing (1.7±0.6µgNm-3) (Guan et al., 2022) and Shanghai (SH, 1.9±1.8µgNm-3) (Lv et al., 2022a), as well as in the northern city of Wuhai (1.7±1.1µgNm-3) (Xie et al., 2025), but substantially higher than those in southern coastal cities such as Guangzhou (0.48–0.53 µgNm-3) (Yu et al., 2017) and Hong Kong (0.6±0.6µgNm-3) (Leung et al., 2024). This indicates that the wintertime NOCs of PM2.5 in CD are notably prominent, likely due to its unique basin topography, which allows local emissions (e.g., biomass burning (Yu et al., 2017; Xie et al., 2025), cooking fumes (Özel et al., 2010; Ditto et al., 2022), vehicle exhaust (Yu et al., 2017; Xie et al., 2025)) and secondary formation (Liu et al., 2023; Yu et al., 2017; Xie et al., 2025) to continuously accumulate near the surface, resulting in concentrations that exceed those in other cities.

As shown in Fig. 1d and Table 1, the gas-phase concentrations of methylamine (MAg) and dimethylamine (DMAg) (1.23±0.37µgm-3 and 0.41±0.03µgm-3, respectively) were 3–5 times higher than their particle-phase concentrations (0.45±0.24µgm-3 and 0.09±0.05µgm-3, respectively). This trend was consistent with observations reported from the Nanling Mountains in Guangzhou, China (MAg: 60.6±35.0ngm-3, MAp: 8.8±7.8ngm-3; DMAg: 71.3±54.5ngm-3, DMAp: 2.4±3.2ngm-3) (Liu et al., 2018), suggesting that amines predominantly exist in the gas phase in the atmosphere. Furthermore, the observed amine concentrations were much higher than those reported in other regions (Liu et al., 2018; Yang et al., 2022; Zhu et al., 2022; Yang et al., 2023; Brean et al., 2025). Studies show that vehicle emissions elevate roadside C1–C2 amine concentrations by more than a factor of four relative to background levels (Brean et al., 2025), with a contribution exceeding 70 % (Yang et al., 2022; Zhu et al., 2022). As shown in Fig. S4, amines in CD exhibited significant correlations with the vehicular tracers NO2 and CO (r=0.4 and 0.5, respectively, p<0.05), indicating that vehicle exhaust contributes to amine emissions to some extent. Except primary emissions, secondary formation also was a major source of amines (Yang et al., 2023). In CD, particulate amines exhibited strong correlation with SNA (sulfate, nitrate, ammonium) (r>0.5, p<0.01; Fig. S5), indicating that secondary formation driven by acid–base neutralization is also an important source.

Figure 2 illustrates the morphological characteristics of aerosols under two different temperature conditions. The FWSON values were lower during the high-T period (0.61±0.11) than during the low-T period (0.71±0.13) (Fig. 2a), suggesting different partitioning processes under the two T regimes. During the high-T process, the OM fraction (fOM=0.22±0.08) (Fig. 2b) exceeded the ALWC fraction (fALWC=0.15±0.08) (Fig. 2c), indicating that under high-T and low-RH conditions, the aerosol predominantly existed in an organic phase. In contrast, during the low-T and high-RH process, the fALWC was substantially higher than the fOM, with fALWC being approximately 4.5 times that of fOM, suggesting that under low-T and high-RH conditions, the aerosol was mainly present in an aqueous phase. Furthermore, Fig. 2d shows that the SNA-to-OM ratio during the low-T process (1.16±0.45) was considerably lower than that during the high-T process (1.62±0.42). Although the high-T process exhibited high concentrations of hygroscopic inorganic components, the RH was only 53±13% (Fig. 2e), significantly lower than that during the low-T process (92±4%) (Fig. 2e), which limited the increase in ALWC. Consequently, fALWC during the high-T process was much lower than that during the low-T process (0.59±0.11) (Fig. 2c). Therefore, the elevated OM concentrations during the high-T process may have facilitated the formation of an organic aerosol phase, whereas the combination of high RH and abundant inorganic components was likely the primary driver for the high ALWC content observed during the low-T process. In summary, water-soluble organic compounds were preferentially partitioned into the organic phase during the high-T process, whereas they were mainly distributed into the aqueous phase during the low-T process.

Table 2 presents the differences in FWSON across three Chinese cities, namely CD (representing the SCB), SH (representing the YRD), and Xianghe (XH, representing the NCP). As shown in the Table 2, FWSON was highest in CD (0.70±0.14), followed by SH (0.49±0.26), and lowest in XH (0.43±0.23). The variation in FWSON among the three cities did not follow the same trends as T and RH, and no significant differences in pH were observed, suggesting that T, RH, and pH exerted only weak influences on FWSON. ALWC showed a decreasing trend across the three cities, which was consistent with the regional variation trend of FWSON. In addition, it was also observed that within each city, FWSON increased with increasing ALWC, implying that ALWC may play a substantial role in modulating FWSON variations. As mentioned earlier, CD is located at the bottom of the SCB, and its unique deep-basin topography leads to unfavorable meteorological conditions in winter, such as stagnant conditions (Table 1) and the frequent occurrence of low boundary layers, resulting in the continuous accumulation of pollutants. The strong temperature inversions in CD promote increases in the low-level cloud liquid water path and near-surface RH, exerting a pronounced regulatory effect on the accumulation of ALWC (Guo et al., 2026). The coupling of these topographic and meteorological conditions makes the wintertime ALWC level in CD significantly higher than that in other regions, making the regulatory effect of ALWC on WSON gas-aerosol partitioning more pronounced than that in coastal or northern Chinese cities.

To further investigate the influence of ALWC on the partitioning of WSON in China, we conducted a statistical analysis of the relationship between the spatial distribution of the fALWC and FWSON across the country (Fig. 3a). A significant linear correlation was observed between fALWC and FWSON (R=0.78, p<0.05), indicating that higher ALWC levels promote the partitioning of WSONg into the particle phase. As illustrated in Fig. 3b, ALWC increased with increasing SNA concentrations during the observation period, particularly under conditions of high RH. Calculations using the ISORROPIA-II thermodynamic model indicate that NH4NO3 contributed approximately 61.3 % to ALWC, followed by (NH4)2SO4 (14.4 %) and OM (7.7 %) during the whole campaign in CD (Fig. 3c). Figure 3d demonstrates that ALWC increased with the ratio of NH4NO3 to PM2.5 at a given RH, suggesting that ALWC in CD during winter is primarily driven by NH4NO3. Therefore, controlling emissions of NH4NO3 precursors (NOx and NH3) is crucial for both ALWC and WSON.

Amines contribute significantly to WSONg,p in CD. Specifically, amines-N (MA-N and DMA-N) account for 48.7 % and 10.5 % of WSON in the gas and particle phases, respectively (Fig. 4). During the winter haze periods in CD, the total concentration of amines increased substantially, and the gas-particle partitioning coefficient (F) nearly doubled (Table 1), indicating enhanced gas-to-particle conversion. It is also evident from Table 1 that the gas-aerosol partitioning coefficients of the basic nitrogen-containing species (NH4+, MA, DMA) followed a decreasing order of FAM(0.36)>FMA(0.25)>FDMA(0.18), indicating that as the hydrogen atoms on the amino group (-NH2) are progressively substituted by hydrophobic methyl groups (-CH3), these basic species increasingly tend to remain in the gas phase rather than partitioning into the aerosol phase. Previous studies have demonstrated that with increasing degree of methylation of amines, their relative hydration free energy increases, thereby weakening their affinity for water and enhancing their partitioning from the aerosol phase to the gas phase (Rizzo and Jorgensen, 1999). In addition, as shown in Table S2, the reaction rate of NH3 with hydroxyl radical (•OH) in the atmosphere is relatively slow (1.6×10-13cm3molecule-1s-1), corresponding to an atmospheric lifetime of up to 72.3 d. In contrast, MA ((2.20±0.22)×10-11cm3molecule-1s-1) and DMA ((6.54±0.66)×10-11cm3molecule-1s-1) react extremely rapidly with •OH, with lifetimes of only a few hours (12.6 and 4.3 h for MA and DMA, respectively) (Table S2). This implies that the fast chemical loss of their gas-phase concentrations directly reduces their accumulation potential in the aerosol phase, leading to the observed decline in the measured F values with increasing methylation degree.

To investigate the partitioning behavior of amines, we calculated their partitioning coefficients under different diel conditions, as shown in Fig. 5. As depicted in Fig. 5a and c, OM exhibited no significant correlation with the Fp of amines during either daytime or nighttime (R<0.5, p>0.05), indicating that OM exerted only a weak influence on the gas-aerosol partitioning of MA and DMA, and that their partitioning into the aerosol organic phase was relatively low. In contrast, both FMA and FDMA showed a clear increasing trend with increasing ALWC (R>0.5, p<0.05), suggesting that MA and DMA preferentially partitioned into the aerosol aqueous phase.

Furthermore, we found significant linear relationships between Fp and acidic anions (nitrate and sulfate) (R>0.5, p<0.01), and the coefficients increased with NOR and SOR (Fig. 6). This indicates that inorganic salts significantly promote the partitioning of MA and DMA, and that reactive uptake may also serve as one of the driving forces facilitating the partitioning of amines into aerosol. According to the Reactions (9–12) (Ge et al., 2011), gaseous amines synergistically neutralize H2SO4/HNO3 together with NH3 to generate particulate amine salts (Almeida et al., 2013; Glasoe et al., 2015; Liu et al., 2018). Figure S5 presents a comparison between amines and inorganic salts. Our results show that amines exhibited good linear correlations with both nitrate and sulfate (R>0.5, p<0.01), suggesting that MA and DMA are likely present in particles predominantly as amine nitrate and/or amine sulfate salts (Huang et al., 2022). Furthermore, ALWC was found to promote their formation (Fig. S5). The humid and stagnant meteorological conditions in CD (Table 1) provide an ideal environment for the above transformations, promoting efficient enrichment of amines in the aerosol phase and making them a significant component of water-soluble ions in PM2.5. 9RnNH3-n(g)⇌RnNH3-n(aq)10RnNH3-n(aq)+H2O(l)⇌RnNH4-n+(aq)+OH-(aq)11RnNH4-n+(aq)+NO3-/SO42-(aq)⇌RnNH4-nNO3(s)/(RnNH4-n)2SO4(s)12RnNH3-n(g)+HNO3(g)/H2SO4(g)⇌RnNH4-nNO3(s)/(RnNH4-n)2SO4(s)

To further identify the key factors governing the partitioning of amines, we analyzed the correlations between Fp and meteorological conditions (i.e., RH and T) as well as pH. As shown in Fig. S6a–f, no significant linear or nonlinear correlations were observed between FMA and T, RH, or pH during either daytime or nighttime (R<0.5). Similarly, FDMA did not show notable correlations with T, RH, or pH during the daytime (R<0.5, p>0.05). However, during the nighttime, FDMA exhibited a significant nonlinear relationship with RH, with an exponential increase when RH exceeded 80 %, while the influences of T and pH remained relatively weak (R<0.5, p>0.05). Figure 7a and b present the theoretical values of the gas-particle partitioning coefficients for MA and DMA (ϵ(CH3NH3+) and ϵ((CH3)2NH2+), respectively) calculated using a thermodynamic model and plotted against aerosol pH in an S-curve analysis. The observed partitioning coefficients for amines in CD fell within the range predicted by theoretical calculations; however, they did not exhibit the anticipated decrease with increasing pH. Theoretically, the partitioning of amines is co-regulated by pH, ALWC, and T. Under acidic conditions with low pH, amines are protonated to form ionic species, with ϵ approaching 100 %, favoring their partitioning into the aerosol phase. As pH increases, deprotonation occurs, forming volatile free bases, with ϵ approaching 0, leading to their transfer to the gas phase. However, given the acidity of atmospheric aqueous phases (pH 0–5), amines exist almost entirely in their protonated form (R-NH3+), within the observed pH range and the uncertainty of the available measurements, no significant independent pH dependence was detected (Tilgner et al., 2021). Generally, lower temperatures enhance the effective solubility of trace gases. However, (Tilgner et al., 2021) explicitly noted that aqueous-phase partitioning coefficients are higher in clouds (typically around 285 K) compared to near-surface conditions during hot summers, and higher in winter compared to summer. This indicates that the influence of T on partitioning coefficients is manifested in vertical altitude gradients and seasonal variations. Within a single season, its impact is less pronounced compared to that of ALWC (Tilgner et al., 2021). In this study, the partitioning coefficients of amines showed no significant correlation with T but exhibited a significant positive correlation with ALWC (Fig. 5). Consequently, variations in pH and T exert minimal influence on amine partitioning (Fig. S6), with ALWC emerging as the dominant limiting factor (Fig. 5).

According to the Pankow (1994), aerosol mass loading may also be a factor affecting the gas-particle partitioning of organic compounds (Pankow, 1994; Lutz et al., 2019; Li et al., 2025). This relationship is typically described using Raoult's law, with the calculation equation as follows (Pankow, 1994): 13Ki=[i]particle[i]gasTSP where Ki is the partitioning constant of compound i, [i]particle and [i]gas represent the concentrations of compound i in the particle and gas phases, respectively, and TSP is the concentration of total suspended particulate material (PM2.5 mass concentrations are used in this study). Figure S7 shows the relationship between the [i]particle/[i]gas ratios of MA and DMA and PM2.5 mass concentrations during the clean, transition, and pollution periods, respectively. For MA, the correlations between its [i]particle/[i]gas ratio and PM2.5 were weak across all the three periods, with R<0.5 and p<0.01. The [i]particle/[i]gas ratio of DMA during the clean period showed a correlation with PM2.5 (R=0.56, p<0.01); however, according to the application criteria for the Ki value proposed by (Lutz et al., 2019), an R2 greater than 0.7 is required for accurate Ki determination and meaningful interpretation, indicating that the results did not fall within the applicability range in this study. During the transition and pollution periods, the [i]particle/[i]gas ratio of DMA also showed no significant correlation with PM2.5 (R<0.4, p<0.01), suggesting that aerosol mass loading did not exert a significant influence on amines uptake during the sampling period.