The online field observation was carried out in the top floor of a building (the ninth floor, ∼ 27 m above the ground) at the Qingyuan campus of Beijing Institute of Petrochemical Technology in Daxing District (39.73∘ N, 116.33∘ E) from 25 July to 21 August 2019. The site is located between the 5th Ring Road and the 6th Ring Road in the south of Beijing and is a typical suburban site. Additionally, there is no additional source of pollution except for two adjacent streets (i.e., Xinghua Street and Qingyuan Street, with a distance ∼ 600 m). The aerosols and gases were sampled using a 1/4 in. stainless steel tube and 1/4 in. PFA tube, respectively, and the sampling inlet extended out of the window about 2.0 m (Fig. S1 in the Supplement). For the aerosol sampling line, the total flow rate was adjusted to 6.0 Lmin-1 using an additional pump to reduce the residence time of aerosols (∼ 0.7 s). Meanwhile, an impactor with a size cut of 2.5 µm was equipped in the front of this sampling inlet to remove coarse aerosols. A series of gas analyzers shared the gas sampling line, and the total flow rate was about 5.5 Lmin-1 with a corresponding residence time of gas of ∼ 0.8 s. During the field observation period, the average temperature and RH were 28.3 ± 3.3 ∘C (22.1–37.7 ∘C) and 62.0 % ± 17.5 % (21.7 %–93.6 %), respectively. The wind speed was in a range of 0.40 to 5.37 ms-1 with an average value of 1.62 ± 0.93 ms-1.

The mass concentration and chemical composition of NR-PM1 were simultaneously measured by a HR-ToF-AMS (Aerodyne Research Inc., USA). The detailed principles of HR-ToF-AMS can be found elsewhere (DeCarlo et al., 2006). Briefly, the aerosols are dried using a diffusion dryer containing silica gel before they enter the HR-ToF-AMS to minimize the impact of aerosol liquid water on collection efficiency (CE). Following this, the aerosols are sampled through a critical orifice and then concentrated into a narrow beam via an aerodynamic lens. The size of aerosol is determined using the flight time of particles to the thermal vaporization and ionization chamber. Then the aerosols are successively vaporized by a heated surface (∼ 600 ∘C), ionized by electron ionization (EI, 70 eV), and detected by a mass spectrometer detector. During this field observation, the HR-ToF-AMS was operated under alternation of two modes, i.e., 2 min V mode and 2 min W mode. The mass concentration of NR-PM1 was derived from V mode considering its higher sensitivity, and the elemental ratios were obtained from W mode due to its higher resolution. Meanwhile, routine quality assurance and quality control procedures, mainly including the calibration of inlet flow, ionization efficiency (IE), and aerosol sizing, were carried out regularly every week according to the standard protocols, using pure dry mono-dispersed 300 nm NH4NO3 aerosols (Chen et al., 2019b; Drewnick et al., 2005) to guarantee the credibility of the HR-ToF-AMS results. The size distribution and number concentration of aerosols with a mobility diameter from 13.6 to 736.5 nm were also measured by a custom-built scanning mobility particle sizer (SMPS, Model 3082 equipped with 3776 CPC, TSI, USA). PM1 mass concentrations also calculated based on the volume concentration measured by SMPS and the estimated PM1 density (method details can be found in the Supplement). Its time series traced well with that from HR-ToF-AMS (R2 = 0.91; slope = 0.95 ± 0.01; see Fig. S2 in the Supplement).

The gas-phase species including NOx, SO2, O3, and CO were measured in real time by a series of Thermo analyzers (Model 42i-TL, 43i, 49i, 48i, respectively). The volatile organic compound (VOC) (e.g., benzene and toluene) concentrations were measured online using a vacuum ultraviolet single-photon ionization time-of-flight mass spectrometer (SPIMS-3000, Guangzhou Hexin Analytical Instrument Co., Ltd., China). These instruments were calibrated periodically with the corresponding standard gas to ensure the accuracy of the observation data. In addition, the meteorological parameters including temperature (T), RH, and wind speed and direction were recorded by an automatic weather station (Vaisala M451).

The time series of mass concentrations of NR-PM1 species (i.e., OA, SO4, NO3, NH4, and Cl) and their relative contributions are summarized in Fig. 1. During this field observation, the hourly mass concentrations of NR-PM1 were in the range of 2.2–64.3 µgm-3, with an average of 19.3 ± 11.3 µgm-3, which is close to those observed in a suburb of Beijing in summer 2016 (14.2 ± 9.4 µgm-3, Li et al., 2019a) and summer 2018 (24.1 ± 18.0 µgm-3, Chen et al., 2020). However, this averaged PM1 concentration was much lower than those observed in urban Beijing in summertime, such as 80 ± 40.6 µgm-3 in 2006 (Sun et al., 2010), 63.1 µgm-3 in 2008 (Huang et al., 2010), 50 ± 30 µgm-3 in 2011 (Sun et al., 2012), and 37.5 ± 31.0 µgm-3 in 2012 (Hu et al., 2017). Meanwhile, among all species in NR-PM1, OA contributed the most (48.3 %), indicating the dominant role of OA in summertime PM1 pollution (Chen et al., 2020; Hu et al., 2016; Sun et al., 2015; Zhang et al., 2014). Moreover, the largest daily contribution of OA (> 80 %) was observed during the clean period, with PM1 concentrations less than 20 µgm-3 (e.g., from 16 to 18 August 2019, Fig. 1b). On the other hand, SIA accounted for 51.7 % of NR-PM1, in which sulfate was the largest contributor (23.2 %), followed by nitrate (14.3 %), ammonium (13.4 %), and chloride (0.8 %). Similar relative contributions of these species were also observed in a suburb of Beijing in summer 2016, where OA contributed a mass fraction of 42 %–71 %, followed by sulfate (15 %–27 %), nitrate (6 %–22 %), ammonium (8 %–13 %), and chloride (0.3 %–0.6 %) (Li et al., 2019a). The percentage of different chemical species in PM1 reported in previous observations in summertime of Beijing in urban and suburban was compared in Fig. S7 in the Supplement. Typically, the fraction of OA (48.3 %) in suburban Beijing is higher than those in urban Beijing (33 %–38 %), while the contribution of SIA (51.7 %) in suburban Beijing is smaller than those in urban Beijing (58 %–64 %) (Hu et al., 2017; Huang et al., 2010; Sun et al., 2010, 2012). The reason may be the difference in the emissions of gaseous precursors (e.g., SO2 and NOx) and their conversion ratio to SIA (Hu et al., 2017; Li et al., 2020). Meanwhile, the inorganic sulfate, nitrate, and chloride could be well neutralized by ammonium during this field observation since there is an extremely good linear relationship between NO3 + 2 ⋅ SO4 + Cl and NH4 (R2 = 0.998, slope = 0.992 ± 0.002, Fig. S8 in the Supplement). This implied the significant contribution of gaseous NH3 to the formation of SIA and PM2.5. Therefore, NH3 emission control in China should be strengthened in the future to mitigate PM pollution (Liu et al., 2019b).

Figure 1 also presents the time series of total oxidant (Ox = O3 + NO2) and photochemical age (ta). In this study, ta and Ox can be used to characterize the photochemical aging process undergone and the total oxidant present in the ambient atmosphere, respectively. As shown in Fig. 1c, ta showed a similar evolutionary trend to that of Ox, with a high correlation coefficient (R2 = 0.75) during this summertime observation, implying there may exist a correlation between Ox and ta. It is well known that the photochemical aging process of VOCs can form organic peroxy radical (RO2) and then accelerate the cycle of NOx and the formation of O3 in the atmosphere. Therefore, a longer aging time (ta) implies that more VOCs are consumed and more RO2 is formed. Consequently, the transform cycles of NOx and the accumulation of O3 would be accelerated under these conditions (T. Wang et al., 2017).

In order to evaluate the effect of the photochemical aging process, the relationships between Ox, PM1, OA concentration, and binned ta were analyzed and shown in Fig. 2. These species concentrations all increase positively with the increasing of ta. Meanwhile, with the increase in ta, concentrations of PM1 and O3 increased simultaneously (Fig. 2 and S9 in the Supplement), suggesting that the photochemical aging process may have an important contribution to the formation of PM1 and O3 during summertime in Beijing. The possible reason is that high temperature in summer can volatilize more VOCs from biogenic sources (Fu et al., 2010) and anthropogenic sources (Pusede et al., 2014) (such as paint, as in McDonald et al., 2018, and asphalt, as in Khare et al., 2020), which could promote the formation of O3 and PM1 (Li et al., 2019c; Schnell and Prather, 2017). This was also verified by the positive correlation between ta and T (Fig. S6). Further analysis showed that the concentration of O3 significantly decreased with increasing NOx in this observation period (Fig. S9b), implying the VOC-sensitive regime of O3 formation. Thus, increased VOCs will produce more RO2 and O3 locally (Li et al., 2018b), as well as more OA, since SOA (predominant species in OA) mainly originates from the photochemical process of VOCs (Fan et al., 2020; Fu et al., 2014; Hallquist et al., 2009).

The time series of total OOA (= LO-OOA + IO-OOA + MO-OOA) and Ox was also compared (Fig. 3) and they presented similar temporal changes (R2 = 0.85). As shown in the inset of Fig. 3, the regression slope for OOA vs. Ox was determined to be 0.130 ± 0.002 µgm-3ppb-1, which was slightly lower than those observed in Pasadena, CA (0.146 ± 0.001 µgm-3ppb-1) (Hayes et al., 2013); Riverside, CA (0.142 ± 0.004 µgm-3ppb-1) (Docherty et al., 2011); and Mexico City (0.156 ± 0.001 µgm-3ppb-1) (Aiken et al., 2009). This might be related to the observed high concentrations of light alkenes during this period. These species will cause high O3 concentrations but will not contribute greatly to the formation of SOA. The scatter data are colored by the time of day, and the slope observed in the morning is steeper than that in the afternoon, which was also observed in other field measurements and was mainly due to the increased evaporation of IO-OOA (Hayes et al., 2013; Herndon et al., 2008; Wood et al., 2010). These results implied that the formation and evolution of PM pollution and OA were closely related to the photochemical aging process and total oxidant present during the summertime, which need to be deeply explored to understand the growth of PM1.

To further investigate the role of ta in aerosol formation, all NR-PM1 species and OA factors were normalized to HOA to exclude the accumulation and/or dilution effects in the atmosphere, and the corresponding ratios as a function of ta are given in Fig. 4. Although the HOA emissions vary throughout the day and might bring some uncertainty, Sun et al. (2013a) indicated that this uncertainty might be reduced and become insignificant when these databases are averaged (as a function of ta in this study). The normalized ratios of all NR-PM1 species positively increased as a function of ta with different increase rates. At low ta levels (< 40 h), relatively small ratios of all species could be observed, and all ratios showed a slight increase trend. However, at ta=70–80 h, the normalized concentration of total NR-PM1, organic aerosol, and sulfate increased by a factor of > 6 compared with that at low ta levels (< 40 h). This indicated that these species were susceptible to photochemical aging processes in the summer with strong solar radiation. Exponential fitting (y = A⋅exp⁡(x/B)+C) was applied to quantitatively describe the relationship between these species and ta, as given in Fig. 4. The fitting parameter B could be used to characterize the sensitivity of each species to the increase in ta, while smaller B means more sensitivity to ta. According to the fitting results, the formation of NR-PM1 was most sensitive to ta, followed by sulfate and organic aerosol in sequence. Their average increase rates in absolute mass concentration (Fig. S10 in the Supplement) showed the same trend as the normalized ratios in Fig. 4. The average increase rate of sulfate was the largest (0.4 µgm-3h-1) among NR-PM1 species, while that of NR-PM1 was 0.8 µgm-3h-1. These results also suggested that the photochemical aging process plays an important role in sulfate formation in summertime (Li et al., 2020). Previous studies have revealed that an aqueous-phase process plays an important role in sulfate formation (Sun et al., 2015, 2013a). However, it appears that photochemical oxidation was the primary pathway of sulfate formation in the present summertime observation, as seen in the ta- and RH-dependent distributions of sulfate (Fig. S11 in the Supplement). Both the sulfate mass concentration and its proportion in PM1 were highly dependent on ta rather than RH. As for nitrate, the corresponding ratio first increased and then slightly decreased as ta increased, which was similar to the evolution trend of absolute mass concentration (Fig. S10). It is worth noting that the concentration of NOx decreased with increasing ta (Fig. S12 in the Supplement), suggesting the photochemical oxidation of NOx to HNO3 or nitrate. This was in agreement with the observation of high concentrations of NOz (NOz = NOy - NOx) at elevated ta levels (Fig. S12). Meanwhile, high temperature accompanied high levels of ta (Fig. S12), which could cause the evaporation of nitrate or adsorbed HNO3 into the gas phase (Xu et al., 2019a). As demonstrated by the ta- and RH-dependent distributions of nitrate (Fig. S11), the aqueous-phase process made a significant contribution to the formation of nitrate (Duan et al., 2020).

Regarding the OA factors, they exhibited different dependence on ta, as presented in Fig. 4. Among these four OA factors, MO-OOA was most affected by the increase in ta, with a factor of ∼ 13 compared with that at low ta levels (< 40 h). The enhancement factor of LO-OOA with increased ta was ∼ 10, which might be due to its lower oxidation state (OSc = -0.34) than that of MO-OOA (OSc = -0.23). These results are consistent with the exponential fitting parameter B of MO-OOA and LO-OOA in Fig. 4. The average increase rate of MO-OOA (0.3 µgm-3h-1) was also significantly larger than that of LO-OOA, with a factor of ∼ 7. As for IO-OOA, ta exhibited a slight enhancement effect, suggesting that photochemical aging may be not its main formation pathway. In contrast, the normalized concentration of IO-OOA and its contribution to PM1 significantly depended on RH (Fig. S13 in the Supplement), which was consistent with the results of Herrmann et al. (2015).

Additionally, the evolution of OA/ΔCO as a function of ta was also presented in Fig. S14 in the Supplement, in which ΔCO was obtained by the measured CO concentrations subtracting its background concentration, and the latter was determined to be 0.1 ppm according to the method described by DeCarlo et al. (2010). As shown in Fig. S14, the concentration ratio of organic aerosol and OOA (especially for MO-OOA) to ΔCO also increased significantly with ta, which could be attributed to the SOA formation from the photochemical process and is also similar to the evolution trend of the ratios of OA components to HOA with ta (Fig. 4). These results further indicated the positive role of the photochemical aging process in aerosol formation.

Figure 5 shows the averaged contributions of NR-PM1 species and OA factors at low and high ta levels. The average mass concentration of NR-PM1 at ta > 40 h was 25.3 µgm-3, which was 1.4 times that at ta < 40 h. All NR-PM1 species were enhanced in mass concentration by a factor of more than 1.1 at elevated ta, of which sulfate had the largest enhancement factor of 1.7. OA was the dominant species in NR-PM1 at both ta levels, although its percentage decreased from 48 % to 44 % with the increase in ta. This decreased percentage could be attributed to the significantly decreased contribution of HOA (from 37 % to 20 %) with the enhancement of ta, suggesting that the photochemical aging process was not conducive to the accumulation of HOA. Comparing the contribution of other species to NR-PM1 at low and high ta levels, it was found that the contribution of sulfate was enhanced from 22 % to 28 % as ta increased, which confirmed that the formation of sulfate was closely related to the photochemical process. The contribution of nitrate presented a decreasing trend (from 16 % to 12 %) as ta increased, which could be due to the enhanced production of OOA and sulfate as ta increased, as well as the dilution effects and evaporation of ammonium nitrate (DeCarlo et al., 2008; Nault et al., 2018).

The OA factors also presented different change trends and different compositions at high ta levels compared with those at low ta levels. The OOA accounted for more than half of OA, and showed much higher contribution at high ta levels than at low ta levels (63 % vs. 80 %). Among them, MO-OOA showed the largest enhancement, by a factor of 1.7 from 3.2 to 5.3 µgm-3 in absolute mass concentration and a factor of 1.6 from 29 % to 48 % in relative contribution, indicating the importance of MO-OOA in PM1 pollution at elevated ta during the summertime. The contributions of LO-OOA and IO-OOA were almost equal (∼ 22 % and ∼ 10 %, respectively) between low and high ta levels, while the absolute mass concentrations were enhanced by a factor of > 1.1 at high ta levels. On the contrary, the absolute mass concentration of HOA decreased from 3.5 to 2.1 µgm-3 during high ta periods, indicating a negative effect of ta on the accumulation of HOA. The effects of dilution and evaporation might also be another reason for the lower HOA contribution and concentration, which tended to be observed at higher temperature conditions and/or in the afternoon with the higher planetary boundary layer (PBL) (Fig. S9).

The role of ta in the evolution of OA during the summertime was further examined. Figure 6 shows the variations of H/C, O/C, N/C, S/C, OM/OC, and OSc as a function of ta. All of these elemental ratios were determined from W mode due to its higher resolution. And all except H/C increased with the increase in ta. The decrease in H/C was related to the decreased contribution of HOA (Fig. 5), since HOA had the highest H/C among these four OA factors (Fig. S4). In contrast to H/C, O/C and OM/OC increased from 0.52 to 0.77 and 1.88 to 2.20 with the increase in ta, respectively. Moreover, a high correlation coefficient between OM/OC and O/C was observed (R2 = 0.997, Fig. S15 in the Supplement). These results suggest that OA was highly oxidized due to the progress of atmospheric photochemical aging. Meanwhile, the analogous variation between OSc and O/C confirmed the high oxidation state of OA, since both are metrics of the oxidation degree of OA (Kroll et al., 2011). As for the variation of N/C and S/C, they presented a slight increase trend with the increase in ta. This indicated that photochemical processes could contribute to the formation of N-containing and S-containing organics, which is consistent with the fact that both of them are the important products of gas-phase oxidation of VOCs and have also been detected in SOA (Farmer et al., 2010; Chen et al., 2019c, d).

The binned H/C and O/C as a function of ta are given in Fig. 7. The fitted slope of -0.74 is similar to those reported in both field observations and laboratory simulations on the photochemical aging of anthropogenic primary OA or biomass burning OA (Chen et al., 2015). This slope can be obtained by a combination of the simultaneous addition of both alcohol (slope = 0) and carbonyl (slope = -2) functional groups, or the addition of carboxylic acid with (slope = -0.5) and without (slope = -1) fragmentation (Peng et al., 2016; Chen et al., 2015; Ng et al., 2011; Heald et al., 2010). Additionally, as ta increased, the OA evolved from the top-left region to the bottom-right region of the Van Krevelen diagram, accompanied by a decrease in H/C and increase in O/C. This further indicated that the evolution process of OA was from less oxidized to more aged as the atmospheric photochemical aging progressed, leading to higher PM1 concentrations.