From 16 October to 16 November 2021, aerosol mixing states were continuously and concurrently monitored using different techniques at the Gucheng site in Dingxing County, Hebei Province, China, as part of a campaign to investigate AQueous Secondary aerOsol Formation in Fogs and Aerosols and their Radiative effects in the NCP (AQ-SOFAR). The observation site, located at 39°09′ N, 115°44′ E, is an Ecological and Agricultural Meteorology Station of the Chinese Academy of Meteorological Sciences, situated between the megacities of Beijing (approximately 100 km away) and Baoding (approximately 40 km away) and surrounded by farmlands and small towns. This site provides a representative view of the background atmospheric pollution conditions in the NCP (Kuang et al., 2020; Li et al., 2021).

Different measurement techniques were used to simultaneously obtain the aerosol mixing state through CCN activity, hygroscopicity, volatility, and BC particle observations. In addition to aerosol mixing-state measurements, the AQ-SOFAR campaign includes measurements of the aerosol number size distribution, chemical composition, aerosol scattering, and absorption properties. Aerosol number size distributions in the diameter range of 13 nm–4 µm were measured by the scanning mobility particle sizer (13–550 nm) and the aerodynamic aerosol classifier (100 nm–4 µm), and they are merged by assuming an aerosol density of 1.6 g cm-3. The total BC mass concentrations were determined using an Aethalometer (Magee, AE33; Drinovec et al., 2015); more information on the correction of absorption measurements and mass concentration calculations is available in Luo et al. (2022). All aerosol measurement instruments were housed in a temperature-controlled container at 24 °C. The inlet was switched among three impactors: TSPs (total suspended particles), PM2.5 (particulate matter with an aerodynamic diameter smaller than 2.5 µm), and PM1 (particulate matter with an aerodynamic diameter of less than 1 µm). Inlet changes among impactors affect dry-state aerosol sampling owing to ambient aerosols being enlarged through aerosol hygroscopic growth or activation. However, the aerosol mixing-state and aerosol chemical composition measurements were made on submicron aerosols, and the inlet change almost did not affect those measurements under conditions of RH that was less than 90 %. The sampled aerosol was dried by two parallelly assembled Nafion dryers with a length of 1.2 m. Two Nafion dryers used because of the high RH and sample flow rate (∼ 16 L min-1) during the campaign to ensure drying efficiency. The flow rate is carefully adjusted in the inlet in order to ensure accurate aerosol particle size cutoff. In addition, during autumn and winter in the NCP, ambient air temperature (< 20 °C and sometimes < 0 °C) can be significantly lower than the room temperature (∼ 24 °C). Therefore, this dryer system can maintain the RH of the sampled aerosols to below 20 %. Meteorological data such as temperature, pressure, wind speed, wind direction, and RH were obtained from an automatic weather station operated by the station.

The chemical composition of the submicron aerosols was analyzed using a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS). The ionization efficiency (IE) was calibrated using 300 nm diameter pure NH4NO3 particles and following the standard protocols outlined in Jayne et al. (2000) in the middle of the campaign, with the relative ionization efficiency (RIE) of ammonium determined to be 5.26. The RIE of sulfate was 1.28, using pure (NH4)2SO4 particles, and the default RIEs of 1.4 for organic aerosols, 1.1 for nitrates, and 1.3 for chloride were used as the organic aerosols. The composition-dependent collection efficiency reported by Middlebrook et al. (2012) was used. Elemental ratios were derived using the “Improved-Ambient (I-A)” method as described in Canagaratna et al. (2015), including hydrogen-to-carbon (H / C), oxygen-to-carbon (O / C), and organic-mass-to-organic-carbon (OM / OC) ratios. Two primary organic aerosol (POA) and two oxygenated organic aerosol (OOA) factors were identified by high-resolution positive matrix factorization (HR-PMF; Ulbrich et al., 2009; Paatero and Tapper, 1994). This study used the summation of the two OOA factors to represent secondary organic aerosols (SOAs). The mass spectra of the organic aerosol (OA) factors and their correlations with external species are shown in Figs. S1 and S2 in the Supplement. The biomass-burning organic aerosol (BBOA) spectrum was characterized by abundant fragments of m/z 60 (mainly C2H4O2+) and 73 (mainly C3H5O2+) with two indicators of biomass burning (Mohr et al., 2009). BBOA correlated well with C2H4O2+ (R2 = 0.91) and C3H5O2+ (R2 = 0.90). Consistent with previous studies in Beijing (Xu et al., 2019), the PMF analysis revealed a mixed factor named fossil fuel organic aerosol (FFOA) which comprises traffic emissions and coal combustion and is characterized by a typical hydrocarbon ion series. FFOA had a relatively high f44 (0.083) value, which was likely due to aging during regional transportation, similar to the results observed in the winter of 2016 in Beijing (Xu et al., 2019) and coal combustion organic aerosols in Gucheng (Chen et al., 2022). Secondary organic aerosol formation from volatile organic compound precursors could occur in different formation pathways, such as aqueous-phase, heterogeneous, or gas-phase reactions. It might also be oxidized under different conditions, such as oxidation under different nitrogen oxide conditions with different oxidation capacities and oxidants. The two resolved OOA factors displayed different spectral patterns, correlations with tracers, and diurnal variations, suggesting that they resulted from different chemical processes. However, their formation mechanisms remain to be explored in future studies. In general, the OOA factor 1 (OOA1) has higher CO2+/C2H3O+ (3.9) and O / C (0.91) ratios than OOA factor 2 (OOA2) with 2.1 and 0.78, respectively. The mass fraction (MF) of each chemical composition is calculated as the bulk mass fraction of each chemical composition in non-refractory PM1 (NR-PM1).

This study did not consider losses in the inlet line and sampling systems for the following reasons: (1) investigated mixing-state parameters are represented by number fractions (NFs) of different diameters which are much less affected by losses in sampling systems compared with absolute number concentrations, and (2) good consistency was achieved between measurements of particle number size distributions (PNSDs) and mass concentrations measured by AMS. The average ratio between volume concentration derived from AMS and rBC measurements (densities of compounds are the same as Kuang et al., 2021) and the volume concentration derived from PNSD measurements was 0.79 (R = 0.97; as shown in Fig. S3), consistent with previous reports as AMS cannot detect aerosol components such as dust (Kuang et al., 2021).

The time series of the meteorological parameters, aerosol mixing-state measurements using different techniques, and mass concentrations of the aerosol chemical components are shown in Fig. 1. In detail, the measurements of aerosol mixing states include SPAR at an SS of 0.08 % by DMA–CCNC, GFPDF (PDF of GF) at 200 nm by HTDMA, SFPDF (PDF of SF) at 200 nm and 200 °C by VTDMA, and lag time PDF of 200 nm BC-containing particles by DMA–SP2. The secondary inorganic aerosol (SIA), SOA, POA, and BC mass concentrations are shown in Fig. 1b. Three periods with significantly different aerosol pollution conditions were identified during the campaign. As shown in Fig. 1b, before 23 October (moderately polluted period), the accumulation of aerosols led to SIA mass concentrations < 20 µgm-3. In contrast, the highest mass concentrations of SOA, POA, and BC all reached 10 µgm-3. The mass concentrations of different aerosol components increased significantly from 23 October to 6 November (heavily polluted period with an average non-refractory PM1 mass concentration of 49.5 ± 22.5 µgm-3) and decreased to much lower levels after 6 November (clean period with a non-refractory PM1 mass concentration of 5.1 ± 3.3 µgm-3). Two particle groups were identified concerning the CCN activity, hygroscopicity, volatility, and coating thickness, as demonstrated by the SPAR, GFPDF, SFPDF, and lag time PDF of BC-containing particles. Significant variations in the aerosol mixing states were also observed during the three periods with different pollution conditions, as demonstrated by the variations in SFPDF measured by VTDMA. For example, the SF of the non-volatile particle group decreased during the heavily polluted period. Aerosol mixing states may have changed because of various transformations in existing aerosol particles and distinct secondary formation processes under different pollution conditions (Kuang et al., 2020; Tao et al., 2021; Shi et al., 2022; Yang et al., 2022). Diurnal variations in the mass concentrations of different aerosol chemical components and mixing states can be observed in the variations in the SPAR measurements, as previously observed in this region (Liu et al., 2011; Ma et al., 2012; Kuang et al., 2015; Tao et al., 2020).

Figure 2 shows the campaign-averaged SPAR at the three SSs, PDF of the lag time of BC-containing particles, GFPDF, and SFPDF at 200 °C for different particle sizes. The sigmoidal SPAR curves were characterized by a rapid increase, followed by a gradual increase to unit 1, similar to the measured SPAR curves previously observed in this region (Deng et al., 2011; Zhang et al., 2014; Ma et al., 2016; Tao et al., 2018). At lower SSs, the particle size required for CCN activation was larger; thus, rapid increases in the SPAR curves occurred at larger particle sizes, as expected. In addition, the maximum activation ratio of the SPAR curves decreases as fewer particles are CCN-active under low SSs. For the three measured SSs, the particle sizes where SPAR equals approximately 0.5 are approximately 90, 120, and 180 nm for the three SSs of 0.08 %, 0.14 %, and 0.22 %, respectively, consistent with the average Da (see Eq. 7) values of the campaign. The NF of CCN-active particles in large-diameter ranges (which varies with SS and, for example, is greater than 200 nm for 0.08 %) can be indicated by the gradual increase in the SPAR curves and quantified by the fitting parameter, MAF (see Eq. 7). The PDFs of the lag time, GF, and SF were all characterized by a bimodal distribution, which indicates two particle groups of BC-containing particles with different coating thicknesses, hygroscopicity, and volatility. The variations in the aerosol mixing states were further analyzed based on the measured mixing-state parameters.

The size-dependent characteristics of the aerosol mixing-state parameters derived from the measurements of the four techniques and the MFs of different aerosol chemical components during the three pollution periods are shown in Fig. 3. In general, the size-dependent characteristics of MAF, NFH, NFV, and NFnoBC were similar, suggesting that they were likely dominated by the same particle group, namely BC-free particles. This particle group had the highest number fraction (> 0.7) during the heavily polluted period and the lowest number fraction (down to 0.5) during the clean period, with the fraction decreasing with increasing particle size. This suggests that primary emissions tend to have higher number fractions of BC-containing particles in larger diameter ranges; for example, the number fraction of BC-containing particles increases from ∼ 0.1 to ∼ 0.4 as the particle size increases from 200 to 500 nm during the clean period. Because the bulk aerosol MF is mostly contributed by particles > 300 nm, there may have been more hydrophilic, volatile, CCN-active, and BC-free particles with larger sizes (> 300 nm) during the heavily polluted period owing to strong SA formation in larger diameter ranges (Kuang et al., 2020), resulting in a higher NF of these particles compared to the clean period. As for RexBC, which is defined as the number concentration ratio of externally mixed BC particles in the total BC-containing particles, the small size dependence of RexBC during the moderately polluted period might have been associated with stronger primary emissions, while the decrease in RexBC with increasing particle diameter in the polluted period confirmed that SA formation is more efficient for particles with larger diameters.

As for the difference among the aerosol mixing-state parameters, NFV and NFnoBC agreed with each other with a < 0.1 difference, and both were higher than NFH by at least 0.1 in the moderately polluted period. Compared with NFnoBC, NFV was higher during the heavily polluted period when the nitrate mass fraction was the highest (∼ 30 %). The SOA mass fraction was the lowest (∼ 7 %) among all three periods, suggesting that some BC-containing particles in this period were also identified as volatile, consistent with the fact that the formation of semi-volatile nitrate in BC-containing particles increases their volatility. However, during the clean period, NFV was even lower than NFnoBC, suggesting that some BC-free particles were characterized as low-volatility, and non-negligible number fractions of BC-free particles dominated these less volatile aerosol components which were likely less volatile organic aerosols (not likely contributed by BC-containing particles with a BC smaller than the SP2 detection limit because this type of volatile BC-containing particle has an SF lower than 0.4 (= 80 nm/200 nm), which is substantially lower than the threshold SF of 0.85 for the NFV calculation). In addition, the MAF values generally agreed with NFH during the clean period. However, they were larger than NFH during the moderately and heavily polluted periods (by ∼ 0.2) when the POA/SOA mass fractions were higher (∼ 40 % vs. ∼ 35 %). POA generally has a lower hygroscopicity than SOA. The critical κ of hydrophilic mode aerosols was 0.07, suggesting that a higher number fraction of aerosols had κ below 0.07 (i.e., hydrophobic mode aerosols in this study) during the moderately polluted period. However, under supersaturated conditions, they demonstrate enhanced hygroscopicity by becoming CCN-active. NFH was consistently lower than NFV and NFnoBC (the average difference between NFH and NFnoBC was approximately 0.2). As mentioned above, NFH was also lower than MAF during moderately polluted periods, and there may be a significant number fraction of volatile BC-free particles with hygroscopicity lower than the critical κ value of 0.07; however, they were still CCN-active and therefore not fully hydrophobic.

The diurnal variations in MAF, NFH, NFV, and NFnoBC, along with the MFs of the aerosol chemical components during the three periods, are shown in Fig. 4. Except for a particle size of 50 nm, the diurnal variations in these four mixing-state parameters were generally similar for all measured sizes. The different diurnal variations at a particle size of 50 nm may be due to the different effects of emissions and aging processes on the different aerosol modes, as particles < 100 nm mainly reside in the Aitken mode, while particles > 100 nm mainly reside in the accumulation mode (Wang et al., 2022). For particles > 100 nm (Figs. 4 and S5), there was a maximum in the afternoon for MAF, NFH, NFV, and NFnoBC, indicating a peak during this time due to an increase in SA compositions, such as nitrate and SOA, and a decrease in POA and BC. Diurnal variations in the aerosol mixing-state parameters and chemical compositions were more pronounced during the moderately polluted period. During heavily polluted periods, the diurnal variation was least pronounced for NFV and most pronounced for NFH. In the clean-air period, there was another maximum at midnight for MAF and NFnoBC, which may be attributed to the diurnal variations in SA compositions, such as sulfate and SOA, and the decrease in BC and FFOAs. The average size dependence of the aerosol mixing-state parameters over different time ranges during a heavily polluted period is shown in Fig. S6. It can be seen that the differences among the four parameters were the least from 12:00 to 18:00, with the most SOA and the least POA. This is consistent with the results shown in Fig. 3, where the difference between the MAF and NFH decreased when the POA mass fractions were the smallest. RexBC tended to be lower during the daytime, and its diurnal variation was more significant for larger particle sizes. In general, the diurnal variations for RexBC were opposite to those of NFnoBC and agreed better with those of the primary aerosol MFs. This is because BC particles originate from primary emissions and are mainly mixed externally. After aging in the atmosphere, BC particles can be coated by SAs, resulting in more coated BC particles and fewer externally mixed BC particles. As SAs tend to form on larger particles, the diurnal variations in SA formation may significantly affect the RexBC of larger particle sizes.

As summarized in Table S1 in the Supplement, the comparison among MAF, NFH, NFV, and NFnoBC was conducted based on their correlations with different particle sizes. Note that the MAF at SSs of 0.08 %, 0.14 %, and 0.22 % was used for comparison at 200, 150, and 100 nm particle sizes. This is because the diameter range of rapid increases in the SPAR curves is determined by aerosol hygroscopicity in this particle size range. The midpoints of the rapidly increasing diameter ranges of the SPAR curves at SSs of 0.08 %, 0.14 %, and 0.22 % were approximately 180, 120, and 90 nm, respectively (as shown in Fig. 2). In general, there were moderate correlations (r = ∼ 0.5) between MAF, NFH, and NFV, suggesting that a similar particle group contributed to the dominance of CCN-active, hygroscopic, and volatile aerosols (Zhang et al., 2016). The agreement between MAF and NFV was slightly higher than that between MAF and NFH or between NFH and NFV. In detail, compared to the other two, the agreement between MAF and NFV has a similar correlation coefficient (r ∼ 0.65) and a smaller systematic difference (slope and intercept were much closer to 1 and 0, respectively). This is consistent with the previous finding that a substantial number fraction of volatile but less hygroscopic aerosols is CCN-active. For smaller particle sizes, the correlation became weaker (r = ∼ 0.4), whereas the degree of reduction was the lowest for the correlation between MAF and NFV (r ∼ 0.529).

Figure 5 presents the correlation between each aerosol mixing-state parameter at 200 nm and the MF of each primary organic aerosol composition during the three periods. The four mixing-state parameters (MAF, NFH, NFV, and NFnoBC) were negatively correlated with MFFFOA and MFBBOA. However, the anticorrelation with MFFFOA (-0.45 to -0.74) was much stronger than MFBBOA (-0.10 to -0.45). Biomass-burning emissions and fossil fuel emissions are the two major sources of BC in the NCP (Yang et al., 2022), and NFnoBC was negatively correlated with MFFFOA (r = -0.49) and weakly correlated (r = -0.18) with MFBBOA, suggesting that fossil fuel emissions were likely the dominant source of BC during this field campaign. The negative correlation between MAF and MFFFOA was weaker than that of NFnoBC with MFFFOA (-0.62 vs. -0.49). In particular, at the same MFFFOA, the MAF was lower than NFnoBC, demonstrating that some BC-free particles were CCN-inactive and were likely mainly composed of organic aerosols from fossil fuel combustion emissions. The negative correlation between NFV and MFFFOA was slightly weaker than between NFnoBC and MFFFOA (-0.56 vs. -0.49). At the same MFFFOA, NFnoBC was close to NFV, and considering that BC-containing particles were dominated by thinly coated BC most of the time (Fig. 5), this demonstrates that the non-volatile population identified by VTDMA was mainly contributed by BC-containing particles. NFH had the lowest negative correlation with MFFFOA (r = -0.74), demonstrating significant contributions from fossil fuel emissions to nearly hydrophobic aerosol populations. At the same MFFFOA, for example, when conditions of MFFFOA > 0.1 were met, NFH (< 0.7) demonstrated a noticeable decrease compared to NFnoBC (> 0.7), and NFH showed a negative correlation with both MFBBOA and MFFFOA, suggesting that a substantial portion of nearly hydrophobic particles originated from FFOA- or BBOA-dominant rather than BC-containing particles. Additionally, markedly different correlations were observed between MAF and MFFFOA (r = -0.62) and between MAF and MFBBOA (r = -0.2), implying that nearly hydrophobic but CCN-active aerosols likely originated from biomass burning. The correlations between the ratio of thinly coated BC in the total BC-containing particles (RexBC) and the MFs of BBOA and FFOA are shown in Fig. 6. Weak correlations (r < 0.3) between RexBC and MFBBOA and MFFFOA were observed. However, RexBC tended to increase with MFFFOA, suggesting that BC-containing particles emitted from fossil fuel combustion tended to be more externally mixed with other aerosol components than those emitted from biomass burning, which is consistent with the results of previous studies (Schwarz et al., 2008; Laborde et al., 2013; Liu et al., 2017; Y. Zhang et al., 2020). These results demonstrate remarkably different mixing states and the physicochemical properties of fossil fuel combustion and biomass-burning aerosols.

The impact of primary emissions on the differences among the four aerosol mixing-state parameters at a particle size of 200 nm was analyzed and is shown in Fig. 7. The difference between NFnoBC and NFH (NFnoBC - NFH) was significantly positively correlated with MFFFOA and MFBBOA (r > 0.5), suggesting that a substantial proportion of POA resided in BC-free particles and was volatile but contributed substantially to nearly hydrophobic aerosols, as did the differences between NFV and NFH (NFV - NFH). The MFs of BBOA and FFOAs were poorly correlated with the differences between the MAF and NFV (MAF - NFV), MAF and NFnoBC (MAF - NFnoBC), and NFV and NFnoBC (NFV - NFnoBC) (Fig. S7). The difference between MAF and NFH was positively correlated with MFBBOA, further suggesting that BBOA contributed to nearly hydrophobic aerosols under subsaturated conditions; however, their hygroscopicity was enhanced, and they became CCN-active under supersaturated conditions. The enhanced hygroscopicity of BBOA under supersaturated conditions may be attributed to (1) the surface tension being lowered by surface-active organic solutes (Hodas et al., 2016; Ruehl et al., 2016), (2) liquid–liquid phase separation (Ovadnevaite et al., 2017; Liu et al., 2018), (3) the dissolution of sparingly soluble compounds at higher saturated conditions (Wex et al., 2009; Dusek et al., 2011), and (4) highly viscous organic aerosol which takes up water by surface water adsorption under sub-saturated conditions and by absorption of water under supersaturated conditions (Pajunoja et al., 2015). The correlations between the mixing-state parameters and primary aerosol composition during the campaign and different pollution periods are summarized in Fig. S7.

In general, both field and laboratory studies have shown that primary organic aerosols from the combustion of biomass and fossil fuels are less hygroscopic. In the laboratory experiment, it is found that organic aerosols produced by fossil fuels have very low hygroscopicity, significantly less than 0.1 (Vu et al., 2015, 2017; Fofie et al., 2018; Zhang et al., 2018b; Mukherjee et al., 2021). Observations have also found that the organic aerosols associated with fossil fuel combustion have low hygroscopicity which may be due to the poorly water soluble substances in FFOAs (Qiu et al., 2019; Li et al., 2021). The aerosol composition produced by biomass burning is complex, with a large number of organic aerosols (BBOAs) and inorganic components being produced at the initial stage, making important contributions to CCN (Spracklen et al., 2011; Bougiatioti et al., 2016; Pöhlker et al., 2018). These primary organic aerosols (i.e., BBOA) are generally semi-volatile (May et al., 2013) and less hygroscopic (Engelhart et al., 2012; Hennigan et al., 2012); this has a negative contribution to the overall hygroscopicity (Bougiatioti et al., 2016; Kuang et al., 2020, 2021; Cai et al., 2022), resulting in weaker overall aerosol hygroscopicity in the initial stage of the biomass burning (Engelhart et al., 2012; Hennigan et al., 2012; Pöhlker et al., 2018). However, laboratory experiments found that BBOAs may contain organic substances with different hygroscopicity under different saturation ratios (Malek et al., 2022) and lead to the increased hygroscopicity and enhanced CCN activity of BBOAs under supersaturation conditions (Hersey et al., 2013). Our results generally agree with previous studies and provide evidence of the enhanced CCN activity of BBOA under supersaturation conditions in field campaigns. Furthermore, the different impacts of aerosols emitted from biomass burning and fossil fuel combustion on CCN are directly observed in this campaign using the newly developed advanced aerosol–cloud sampling system which show that biomass-burning aerosols are efficient CCN even under low supersaturations (< 0.05 %); however, aerosols from fossil fuel combustion can only activate at higher supersaturations (> 0.14 %). These results suggest simultaneous measurements of aerosol GF distributions, SPAR curves, and BC mixing states, and their comparisons could provide novel insights into different synergistic hygroscopic, volatile, and activation properties of aerosols from different sources in the atmosphere.

The correlations between the aerosol mixing-state parameters at 200 nm and the MF of each SA component are presented in Fig. 8 for three periods, and the entire campaign is presented. The analysis is conducted at only 200 nm, where all four aerosol mixing-state parameters were measured to compare the four aerosol mixing-state parameters and their relationships with aerosol chemical components simultaneously. Generally, MAF, NFH, NFV, and NFnoBC exhibited strong positive correlations with MFNH4 (r > 0.5). This is likely because ammonium was mainly formed through neutralizing sulfuric and nitric acids with ammonia; therefore, variations in ammonium better represent the overall secondary inorganic aerosol formation. As shown in Fig. 3, the secondary inorganic aerosol components dominated SA (the mass ratio between SIA and SA is approximately 70 %), indicating that SA formation was primarily composed of secondary inorganic aerosol formation, which explains the weaker correlation with SOA (r = ∼ 0.3), as shown in Fig. 8.

During the clean-air period when the MFs of SOA and sulfate were both above 15 %, all four parameters had a strong positive correlation with MFSO4 and MFSOA (r > 0.5), suggesting that when a clean background air mass with higher mass fractions of sulfate and SOA prevailed, the local primary emissions that contributed substantially to BC-containing and less hygroscopic POA aerosols became less significant. The positive correlations between the MAF and SA components have been extensively discussed by Tao et al. (2021), who found that SA formation enhances the hygroscopicity of nearly hydrophobic aerosols, thereby increasing CCN activity. This also explains the strong correlation between the NFH or MAF and ammonium formation. The strong positive correlations between NFV and SA formation (r = ∼ 0.6) are consistent with the fact that nitrate dominates SA formation during this campaign and is semi-volatile. For the first time, strong positive correlations between NFnoBC and SA formation were observed (r = 0.6). NFnoBC depends primarily on the relative variation between BC-containing and BC-free particles. The increase in NFnoBC at 200 nm as a function of the SA MF suggests that SAs migrated the higher mass fraction of BC-free particles with particle sizes smaller than 200 nm to particle sizes of 200 nm, suggesting that SAs tended to form more quickly on BC-free particles than on BC-containing particles with BC higher than SP2 detection limit. Recent studies reported that catalyst or photochemical reactions on BC particles can contribute the formation of secondary aerosols (F. Zhang et al., 2020, 2021). Our results may indicate that SA formation on BC particles might not be a significant pathway that contributes substantially to haze formation, and the underlying mechanisms need to be further resolved.

The effects of SA formation on the differences between the four aerosol mixing-state parameters were studied and are illustrated in Fig. 9. The two OOA factors (OOA1 and OOA2) were formed through different chemical pathways. The difference between NFnoBC and NFH (NFnoBC - NFH) showed a strong negative correlation with MFNH4 and MFNO3 (mainly -0.6), as did the differences between NFV and NFH (NFV - NFH). Ammonium nitrate is a pure-scattering, semi-volatile component with strong hygroscopicity, and the increase in its mass fraction can enhance both aerosol volatility and hygroscopicity, therefore resulting in a smaller difference among NFnoBC, NFH, and NFV.

Furthermore, the difference between NFV and NFH showed a positive correlation with MFOOA2 and a negative correlation with MFOOA1, indicating the different volatility and hygroscopicity of the two SOA factors. The differences between NFV and NFH concerning the MF of OOA1 and OOA2 are shown in Fig. 9e and f, respectively. As previously noted, NFV was generally higher than NFH, and the difference between the two decreased with increasing MFOOA1, which was generally smaller than 0.3. This suggests that the formation of OOA1 enhances the hygroscopicity of volatile particles, which aligns with the highest oxidation state of OOA1 (higher O / C but lower H / C compared to OOA2) and has a significant and overall positive impact on aerosol hygroscopicity (Cerully et al., 2015; Thalman et al., 2017; Zhang et al., 2023). A positive correlation was observed between NFV and MFOOA2 (r = ∼ 0.25).

In contrast, the correlation between NFH and MFOOA2 was weak (R was close to 0), implying that OOA2 might be semi-volatile but only weakly hygroscopic, which could contribute to NFV being higher than NFH as OOA2 increases. The difference between NFnoBC and NFV (NFnoBC - NFV) was negatively correlated with MFNO3, which is consistent with the semi-volatile nature of nitrate. The negative correlation between NFnoBC - NFV and MFOOA2 indicates that the difference is smaller when there is more OOA2, implying that OOA2 is also a semi-volatile compound and is likely formed mainly on BC-free particles (particles with BC mass lower than detection limit are not excluded). The correlations between the differences between NFV - MAF and NFnoBC - MAF and the MF of each SA composition were very weak. The impacts of SA formation on BC mixing states are shown in Fig. S8. In general, the NF of thinly coated BC has a negative correlation with SIA and a weak association with SOA, suggesting that SIA formation mainly enhances the thickness of the BC coating. The correlations between the mixing-state parameters and SA composition during the campaign and different pollution periods are summarized in Fig. S9. Our results on OOA agree with previous studies in which OOA is reported to be volatile (Kim et al., 2020; Cai et al., 2022) but can have a positive or negative impact on hygroscopicity, depending on its oxidation level (Kim et al., 2020; Kuang et al., 2021; Cai et al., 2022).

In addition to changes in the MFs of SA compositions, the accumulation of SA pollution may provide insights into the impact of SA formation on aerosol mixing states. As shown in Fig. 10a, during the heavily polluted periods, there were two distinct pollution accumulation processes from 23 to 27 October and from 28 to 31 October, respectively. During the pollution accumulation process, the mass concentration of SAs increased approximately 3-fold, indicating the rapid formation of secondary compositions and a significant increase in the NR-PM1 mass concentration. Figure 10b and c illustrate that this increase in SAs significantly enlarged the value of aerosol mixing-state parameters, including MAF, NFV, NFH, and NFnoBC, which increased from approximately 0.5 to 0.8 with evident diurnal variations. This highlights the impact of SA formation on the aerosol mixing states and the importance of studying the pollution accumulation processes of SAs. The enhancements in the different aerosol mixing-state parameters during the pollution accumulation process were not uniform. MAF and NFH initially exhibited lower values than NFV and NFnoBC; however, their later enhancement was stronger than that of NFnoBC. Figure 10d and e show the difference between NFnoBC and NFV at 200 and 300 nm as a function of SA mass concentrations during these two pollution periods, which clearly shows how, during SA formation, NFV became higher than NFnoBC while NFV remained close to the NF of thickly coated BC-containing particles (NFCBC) plus NFnoBC (NFCBC + NFCBC). These results suggest that SA formation increases the volatility of BC-free and BC-containing particles, leading to an increased NFV compared with NFnoBC. Almost all BC-free particles and some BC-containing particles become volatile during the accumulation of pollution.