Characterizing the volatility and mixing state of ambient fine particles in summer and winter of urban Beijing

Understanding the volatility and mixing state of atmospheric aerosols is important for elucidating the formation of fine particles and to help determining their effect on environment and 10 climate. In this study, the volatility of the fine particles is characterized by the size-dependent volatility shrink factor (VSF) for summer and winter in the urban area of Beijing using measurements of a volatility tandem differential mobility analyzer (VTDMA). We show the volatility of aerosols is always with one high-volatile and one lessor non-volatile mode both in the summer and winter. On average, the particles are more volatile in the summer (with mean VSF of 0.3) than in the winter (with mean VSF 15 of 0.6). The outstanding high-volatile mode around noontime illustrates the role of nucleation in producing more volatile particles in the summer. We further retrieve the mixing state of the ambient fine particles from the size-resolved VSF and find that the non-black carbon (BC) particles that formed from nucleation processes accounted for 52–69 % of the total number concentration in the summer. While, particles containing a refractory core that is thought to be BC-containing particles dominate and 20 contribute 67–77 % toward the total number concentration in the winter. The diurnal cycles of the retrieved aerosol mixing state for the summer further supports the conclusion that nucleation process is the main contributors to non-BC particles. In addition, the extent of aging of BC particles was characterized as the ratio of the BC diameter before and after heating at 300 oC (Dp/Dc), showing that the average ratio of ~2.2 in the winter is higher than the average of ~1.5 in the summer, which indicates 25 that BC aging is more efficient in wintertime, with resulting differences in light absorption enhancement between cold and warm seasons. https://doi.org/10.5194/acp-2021-454 Preprint. Discussion started: 5 July 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction
The volatility of atmospheric aerosols affects their effect on climate, visibility and human health (Dzubay et al., 1982;Pöschl, 2005;Baklanov et al., 2016) by modulating mass concentrations and size 30 distributions of aerosol particles via gas-particle partitioning. Aerosol measurement could be largely biased under different temperatures because of volatility (Meyer et al., 2000;Grieshop et al., 2006;Chen et al., 2010). In addition, volatility influences the partitioning of aerosols in gas and particle phases, and thus affects dry and wet deposition rates (Bidleman, 1988), chemical reaction mechanisms, and atmospheric lifetime (Huffman et al., 2009;Glasius and Goldstein, 2016). Therefore, it is important 35 to study the volatility of aerosols in different regions and environments, including in polluted urban areas.
Laboratory and field measurements have shown that aerosol volatility is correlated with chemical composition of the particles, which is impacted by emission sources and atmospheric processes (Wehner et al., 2004;Yeung et al., 2014). Therefore, the volatility of fine particles varies greatly with 40 time and location. A common measure of volatility is the shrink factor (VSF, the ratio of the particle diameter after and before being heated), which is measured by volatility tandem differential mobility analyzer (VTDMA). The VSF can range from as low as 0 (completely volatile compounds) to 1 (indicating completely non-volatile substances, e.g. black carbon), reflecting heterogeneity in particle composition in diverse environments Chen et al., 2020). In addition, the dependence 45 of particle volatility on particle size is complex. For example, Wang et al. (2017) found that ambient aerosol volatility typically decreases as particle size increases in urban Beijing, whereas Levy et al. (2014) showed the opposite dependence on size near the California-Mexico border. Numerous studies have linked aerosol volatility to the presence and abundance of refractory carbonaceous compounds by inferring aerosol mixing state and the degree of aging from measured volatility (Wehner et al., 2009;50 Cheung et al., 2016;Zhang et al., 2016;Zhang et al., 2017;Chen et al., 2020). Mixing state has been found to vary significantly between clean and heavily polluted days (Wehner et al., 2009), with a corresponding decrease in the fraction of externally mixed black carbon (BC) particles from 37 % during clean to 18 % during heavily polluted periods. Cai et al. (2017) showed that nearly all particles partially volatilized at about 300 ℃ in Okinawa, while 15-21 % did not in Pearl River Delta. Saha et al. (2018) found the non-volatile fraction in roadside aerosols was mostly externally mixed. The results from these studies show that aerosol mixing state and degree of aging may differ greatly under diverse ambient conditions. Considering current uncertainties in assessing the radiative forcing of BC particles, which are largely due to uncertainties in the model treatment of BC mixing state, emissions, and removal processes (Cappa et al., 2012;Nordmann et al., 2014), an improved understanding of aerosol 60 volatility and mixing state is hence needed.
Most previous studies in north China have focused on aerosol chemistry, sources, and transport (Wang et al., 2010;Gao et al., 2011;Sun et al., 2016b), but few have linked the volatility and mixing state of the aerosol to its sources, formation and growth. In north China, severe haze usually occurs in the winter season, with extremely high PM2.5, while in the summer it occurs much less frequently and 65 with much lower PM2.5, with the contrast resulting from influences of multiple factors such as regional and local emissions, particle formation, meteorology, and photochemistry. A comprehensive study on investigating the aerosols volatility and mixing state in different seasons may help to elucidate the fine particles formation mechanisms.
In this study, a VTDMA system (Sakurai et al., 2003;Wehner et al., 2009;Cheung et al., 2016;70 Wang et al., 2017) was extensively employed in field observations to measure size-resolved volatility at different heating temperatures during wintertime and summertime in Beijing. At the maximum employed heating temperature of 300 ℃, volatile components tend to evaporate while leaving refractory materials such as black carbon (Cheng et al., 2009). Therefore, the VTDMA is also used to determine the mixing state of refractory carbonaceous particles (Wehner et al., 2009;Zhang et al., 2016;75 Chen et al., 2020). Here, we used the size-dependent VSF as a parameter to quantify the volatility behaviour of fine particles in cold and warm seasons of urban Beijing; the mixing state of ambient fine particles, which is retrieved from size-resolved VSF, in the two seasons is investigated; the seasonal contrast in the degree of aging of the aerosol, characterized by the coating thickness on BC, is also discussed to elucidate the effects of particle growth mechanisms on volatility and mixing state.

Instruments and measurements
The volatility and mixing state of fine particles were measured with a VTDMA (Fig. S1). Wang et 95 al. (2017) provided a brief description of the custom-made VTDMA system. Here, we give more details.
The instrument mainly consists of the following seven parts: neutralizer, the first differential mobility analyzer (DMA1), temperature control module, the second DMA (DMA2), water-based condensation particle counter (WCPC, TSI model 3787), auxiliary components (electric steering valve, vacuum pump, proportional valve, etc.), and software for control and data acquisition. During the measurement, 100 ambient aerosols were first sampled by a PM2.5 inlet and subsequently passed through a Nafion dryer that reduced the sample flow relative humidity to below 30 %. The dried aerosols were then directed through a neutralizer to neutralize the charge carried by the particles and entered the DMA1 to produce quasi-monodisperse aerosols by setting the voltage. The dry diameters (Dp) selected in this study were 40,80,110,150,200 and 300 nm respectively. The selected quasi-monodisperse particles went either to 105 the WCPC to obtain particle counts or through the heating tube for volatility measurements, sequentially at 80, 150, 200 and 300 ℃. Here, we focus on the data derived at 300 ℃. After heating, the sample flow entered the DMA2 to scan the multi-dispersed particles, and finally entered the WCPC to get the particle size distributions after heating, thereby obtaining the volatility shrink factor measured distribution function (VSF-MDF). Through inversion, the volatility shrink factor probability distribution 110 function (VSF-PDF) could be further obtained. The VSF-PDF was retrieved based on the TDMAinv algorithm developed by Gysel et al. (2009). The residence time in the heated region was 2.4 s (Cheung et al., 2016). Compared with that of 0.3 to 1.5 s for other VTDMA systems (e.g., Brooks et al., 2002;Philippin et al., 2004;Villani et al., 2007;Jiang et al., 2018), the residence time in this VTDMA is sufficient for the volatile materials to be effectively vaporized. The relative humidity was calibrated 115 periodically with ammonium sulfate during the measurement period. Because this study investigates only the fine mode particles below 300 nm, refractory components that are present mostly in coarse mode particles, e.g. dust and sea salt, are expected to be negligible. At around 300-350 °C, the refractory component in sub micrometer aerosols in continental and urban areas has been considered to be mainly BC and a small contribution by charred organic material, which is often negligible (Rose et particle phase upon heating to 300 °C, while the rest of the aerosol components tend to evaporate, resulting in a reduction in particle size (Cheng et al., 2009).
In addition to the volatility system, some other auxiliary instruments were used for simultaneous observation, including an aerosol chemical speciation monitor (ACSM) for measuring non-refractory 125 submicron aerosols in PM2.5, an aethalometer (AE33, Magee Scientific) for measuring the mass concentration of BC, and a scanning mobility particle sizer (SMPS) for measuring the particle number size distribution (PNSD) of aerosols. Before the field measurement, all instruments used were calibrated to ensure the data obtained during the study period were accurate and reliable. The meteorological variables from the meteorological observation station were also used, including the ambient temperature 130 (T), relative humidity (RH), wind direction (WD), and wind speed (WS).

Data analysis method
In this study, we use the abbreviations Ex-BC, In-BC, and Non-BC to denote externally mixed, internally mixed, and non-BC-containing particles, respectively. The number fraction ( ) of the completely volatile particles was obtained by considering the number concentrations of the residual 135 particles after heating ( ) and the number concentrations of DMA1-selected particles before heating Where , is the transportation efficiency of the sampled particles, which represents particle losses between DMA1 and DMA2 due to diffusion and thermophoretic forces , 140 and varies as a function of particle size and heating temperature. In this study, , at each particle size is determined from the previous experiment results at 300 ℃ (Cheung et al., 2016 -high-volatile (HV): VSF < 0.45, considered to be Non-BC.
The VSF-PDF ( ( , ) ) was normalized as ∫ ( , ) = 1 . Then, the number fraction ( ) for each volatile group with the boundary of [ , ] is defined as: Where = Ex-BC, In-BC or Non-BC. It is worth noted that, when = Non-BC, those particles that completely evaporate were assumed to be included in the HV mode (considered as Non-BC), so − is calculated as: The number concentrations ( ) of In-BC, Ex-BC, and Non-BC from the VSF distributions combined with the total PNSD simultaneously measured by the VTDMA, are calculated as follows: Here, is the number concentrations of ambient fine aerosol particles. Therefore, the actual number fractions of Ex-BC, In-BC and Non-BC particles before heating could be obtained. The 160 retrieval result, which has been compared with the measurements by single particle aerosol mass spectrometer (SPAMS) in previous study , is reliable for deriving the mixing state of BC. In the summer, the VSF distributions of 40-nm particles were almost always bimodal, with a nonvolatile mode (the VSF was approximately equal to 1) and a high volatile mode (with VSF of about 0.2-175 0.5). As stated above, we attribute the non-volatile group to refractory BC particles. In the winter, the VSF-PDF was bimodal only occasionally, and mostly on polluted days, which could be caused by changes in meteorology and enhanced primary BC emissions during polluted days in the winter. For the 150-nm particles, the distributions are generally unimodal, with VSF of about 0.3-0.6 both in the summer and winter, likely resulting from mixing and aging of the primary particles during growth to 180 larger sizes. The VSF in the summer fluctuated a bit less than that observed in the winter, which will be further discussed in the following section.

Comparison of the VSF-PDF between summer and winter
Comparison of the average VSF-PDF distribution for all measured dry particle sizes during the winter and summer period is illustrated in Fig. 3.

Comparison of diurnal variation of particles volatility between summer and winter
To obtain further insights into the effect of the formation and growth of particles on their volatility, 210 we compare the diurnal variations of the observed mean VSF and VSF-PDF between the summer and winter (Fig. 4). During the summer, low VSF during the daytime (08:00-18:00 LT) and high VSF during the nighttime were observed (Fig. 4a). Accordingly, the VSF-PDF shows the HV mode dominated around noontime and early afternoon and the LV mode dominated during nighttime (Figs. 4c and 4e). The diurnal variation is more evident for small particles (e.g. 40 nm) than for larger particles 215 (e.g. 150 nm). The diurnal variations illustrate that particles are more volatile during the daytime than at night, with VSF decreasing dramatically after ∼10:00 LT when new particle formation (NPF) events usually occurred (Fig. 4g). It has been shown that ∼97 % of newly formed particles are volatile because they are dominated by non-refractory sulfate and organics (Wehner et al., 2009). In addition, during daytime atmospheric aging processes facilitated the mixing of primary particles (e.g. BC) with 220 secondary species, leading to the transformation of externally mixed particles to internally mixed particles. In the evening and the early morning, the number fraction of LV-mode particles increased because of increased emissions of refractory particles (like BC) from traffic and other primary sources, coupled with slower particle aging and weaker vertical mixing that concentrates the externally mixed BC close to the surface (Zhang et al., 2016). 225 Compared with that in the summer, there was little diurnal variation in VSF during the winter period (Fig. 4b), and an MV mode was present in the VSF-PDFs for both 40-and 150-nm particles during the daytime (Figs. 4d and 4f). This is likely because of weakened photochemistry during the daytime in the cold season, when fewer NPF events were observed (Fig. 4h). In addition, the number fraction of the LV mode for both 40-and 150-nm particles is much lower during the winter. This may 230 reflect the fact that BC particles can be coated and aged quickly through heterogeneous reactions of VOCs and other precursor gases (like SO2 and NOx) , which are usually more concentrated during polluted days in the winter (Sun et al., 2016a). However, the aging process is expected to be slowed in the summer when precursor concentrations are lower. Such an explanation is reasonable and can be supported by the observed thicker coating layer in the winter, as characterized by 235 Dp/Dc (shown in Fig. 8

Figure 4.
Diurnal variation of (a-b) mean VSF for all measured dry particle sizes, (c-f) mean VSF-PDF for 40and 150-nm particles, and (g-h) mean particle number size distribution during the summer (left) and winter (right) periods. The shade regions in (a-b) denote the standard deviations.

Number concentrations and fractions of Non-BC, In-BC, and Ex-BC
To study the aerosol mixing state, we retrieved the number concentrations of Non-BC, In-BC, and Ex-BC from the VSF data (Fig. 5).   In summary, during the winter most ambient aerosol particles were BC-containing, suggesting that BC particles are a dominant component in urban Beijing. In the summer, however, BC-containing particles contributed much less (only 31-48 %) toward the total number concentration in the measured 275 size range, while Non-BC particles originating from nucleation are the dominant particle type.

Coating thickness characterized by Dp/Dc ratio
The ratio of the BC diameter before and after heating at 300 º C (Dp/Dc) was used as a quantitative 310 index to characterize the coating thickness (degree of aging) of BC-containing particles. In the winter, most BC particles have Dp/Dc ratios of 1.6-2.6 (Fig. S6). The large Dp/Dc ratios suggest that the BC particles are thickly coated and likely have a compacted structure following atmospheric aging that results from additional emissions from the residential heating sector and favorable condensation because of the low temperature. While during the summer, the coatings on BC were thinner, with an 315 average Dp/Dc of 1.5 (Fig. 8). In addition, in the winter the Dp/Dc ratio was highly dependent on particle size, with decreasing coating layer thickness with increasing particle size. In the summer, it was independent of particle size.
Our results are similar to those reported previously in similar urban environments (Wehner et al., 2004;Zhang et al., 2018;Liu et al., 2019). For example, Liu et al. (2019) found BC coating thickness 320 was more variable in the winter than in the summer, and that the average coating thickness on BC particles was higher in the winter. Zhang et al. (2018) showed that the size-dependence of the Dp/Dc ratio was associated with air pollution and indicated that the aging of smaller BC cores was more sensitive to air pollution levels.
Previous modelling studies have reported that coating materials on BC particles can significantly 325 enhance the light absorption of BC via the lensing effect (Jacobson, 2001;Moffet and Prather, 2009;Lack and Cappa, 2010;Zhang et al., 2020

Conclusion
In this study, the volatility of the fine particles is characterized as VSF and the results from wintertime and summertime are shown and compared. Results show that the measured VSF-PDF is almost always bimodal, with one high-volatile and one less-or non-volatile mode, both in the summer and winter. The mean VSF-PDF has a pronounced HV mode in the summer, generally with a VSF