As part of the APHH winter campaign, we conducted measurements at the Tower Division of Institute of Atmospheric Physics (IAP), Chinese Academy of Science (39∘58′ N, 116∘22′ E) in Beijing (Fig. S1 in the Supplement), from 15 November to 13 December of 2016. The site was surrounded by residential infrastructures and a freeway in the east (360 m).

The SP-AMS was deployed on the rooftop of the Herong building (∼8 m above the ground), with a PM2.5 cyclone (model URG-2000-30EN) and a diffusion dryer in front of the inlet. The single-particle soot photometer (SP2, Droplet Measurement Technology, Inc., Boulder, CO, USA) was operated simultaneously inside another container nearby (∼20 m away) on the ground. The SP2 incandescence signal was calibrated for BC mass by using Aquadag^® black carbon standard (Aqueous Deflocculated Acheson Graphite, Acheson Inc., USA) (Laborde et al., 2012). For the SP-AMS, since the filament that ejects electrons can still heat the tungsten vaporizer up to ∼200 ∘C (Willis et al., 2014) even if it is turned off, the tungsten vaporizer was thus physically removed to make sure only BC and its associates were vaporized by the laser and to eliminate the influence of uncoated species on BC cores.

The tuning and calibration procedures of SP-AMS followed the procedures described previously (Lee et al., 2015; Willis et al., 2016; Massoli et al., 2015; J. Wang et al., 2017). During the campaign, the SP-AMS was run with a 10 min cycle: one W mode with high chemical resolution (2.5 min) and two mass sensitive V modes, including one with particle time of flight (PToF) mode (2.5 min) and another one (5 min) with a large mass-to-charge (m/z) range (up to 2000) (J. Wang et al., 2016). The filtered air measurement was performed for a day to determine the detection limits (DLs) of various aerosol species and to adjust the fragmentation table. The ionization efficiency (IE) and relative ionization efficiency (RIE) of sulfate and nitrate were calibrated by using pure ammonium nitrate and ammonium sulfate according to Jayne et al. (2000), respectively. RIE of BC was calibrated by using REGAL black particles (RB, REGAL 400R pigment black, Cabot Corp.) (Onasch et al., 2012), and the average ratio of C1+ to C3+ was determined to be 0.53 to minimize the influence of C1+ from non-refractory organics. However, it should be aware that the laser-only SP-AMS cannot vaporize ammonium nitrate or sulfate if they do not coat BC; thus the IE and RIE calibrations were done before removal of the tungsten vaporizer and the values were assumed to be unchanged after the tungsten heater's removal (Willis et al., 2016). Note that the RIE of BC was calibrated before the campaign and was repeated in the middle and end of the campaign. RIEs of nitrate, ammonium, sulfate and BC were determined to be 1.1, 3.82, 0.82 and 0.17, respectively. The default value of 1.4 was used as the RIE of organics (Canagaratna et al., 2007). Polystyrene latex (PSL) spheres (100–700 nm) (Duke Scientific Corp., Palo Alto, CA) were used to calibrate the size before the campaign (Canagaratna et al., 2007) .

Standard AMS data analysis software (Squirrel and Pika) based on Igor Pro 6.37 (Wavemetrixs, Lake Oswego, OR, USA) were used to obtain the concentrations, mass spectra and size distributions of BC and its coating species. All data were calculated based on high-resolution fitting results. Due to different vaporization schemes between the SP-AMS and HR-AMS, the mass spectra from these two instruments even for the same population of particles are not entirely the same. Laser-only SP-AMS can result in less fragmentation overall; therefore the mass profile may contain more large m/z fragments and less small m/z fragments compared with that from HR-AMS (Massoli et al., 2015). Therefore, here the elemental ratios of organics, i.e., oxygen-to-carbon, hydrogen-to-carbon and nitrogen-to-carbon ratios (O/C, H/C and N/C) were determined by the Aiken approach first (Aiken et al., 2008), and then O/C and H/C were corrected by using factors of 0.83 and 1.16, respectively (Canagaratna et al., 2015).

Source apportionment for organics coated on BC was conducted by using positive matrix factorization (PMF) (Paatero and Tapper, 1994) evaluation tool written in Igor (Ulbrich et al., 2009). In this study, high-resolution mass spectra (HR-MS) of organic (including BC) and inorganic species were combined together to perform the PMF analyses (Sun et al., 2012; J. Wang et al., 2017, 2018). It should be noticed that only fragment ions from polycyclic aromatic hydrocarbons (PAHs) were included for m/z range of ∼150 to ∼250 in the PMF analysis because of the limited mass resolution of SP-AMS. All PMF solutions were evaluated following the standard instruction (Zhang et al., 2011). Finally, four types of organic aerosol (OA) associated with BC were determined eventually, including a fossil fuel combustion OA (FFOA), a biomass-burning OA (BBOA) and two oxygenated OA (OOA1 and OOA2) (a diagnostic plot was provided in Fig. S2).

Supporting data such as meteorological parameters including relative humidity (RH), wind speed (WS), wind direction (WD) and temperature (T), as well as concentrations of gaseous species such as O3, SO2, NO, NO2, NOx, NOy, NOz, and CO were measured in parallel. All data here are reported in local time (Beijing Time, UTC+8).

Figures 1 and 2 show the temporal variations in meteorologic parameters, mass loadings of gaseous pollutants (CO, NOx, SO2 and O3), BC and its associated coating components (sulfate, nitrate, ammonium, chloride, total OA and four PMF-resolved OA factors). The campaign-averaged composition of BC-containing particles and mass contributions of the four OA factors to total OA were also displayed in Fig. 2. Overall, wind directions and speeds had close associations with overall mass loadings of BC-containing particles. The polluted periods (characterized by concentrations of BC-containing particles above 10 µgm-3) were accompanied by relatively low wind speeds (< 4 m s-1) and in a relatively large part from southern air masses, since Beijing is at the foot of the mountains, which facilitates the accumulation of pollutants from the southern North China Plain (NCP). The clean periods (characterized by the concentrations of BC-containing particles below 10 µgm-3) were mainly under the control of northwesterly strong winds (>4 ms-1) (Fig. S3). During the campaign, the mass loadings of BC cores and BC-containing particles ranged from 0.11 to 26.54 µgm-3 and 0.71 to 174.40 µgm-3, with averages of 4.9 µgm-3 and 29.4 µgm-3, respectively. We also compared BC concentrations determined by the SP-AMS with those from SP2, and they correlated quite well with each other (r2 of 0.93; Fig. S4), indicating that the quantification of BC by the SP-AMS is reliable.

The coating species occupied on average about 83.4 % of the mass of BC-containing particles, indicating that BC was generally thickly coated throughout the whole campaign, with an average mass ratio of coatings to BC (RBC) of ∼5.0. Organic aerosol (OA) was the most abundant coating component, taking up 59.4 % of the total mass, followed by nitrate (8.8 %), sulfate (6.5 %), ammonium (4.7 %) and chloride (4.0 %). OA correlated quite well with BC (r2 of 0.97), suggesting that many OA species were co-emitted and mixed with BC, and indeed, primary OA (POA = FFOA + BBOA) was found to dominate the OA mass (66.3 % = 43.9 % + 22.4 %). Chloride (Cl-) had a great correlation with BC (r2 of 0.94), suggesting it was mainly associated with primary emissions, for example, gasoline, diesel and coal combustion during wintertime in urban Beijing. Sulfate and nitrate are typically secondarily formed; therefore their correlations with BC were relatively weak (r2 of 0.64 for SO42- vs. BC and 0.60 for NO3- vs. BC). Their properties are discussed in detail in the following sections.

Figure 3a shows the campaign-averaged mass-based size distributions of major BC-coating species, including organics (BC-org), sulfate (BC-sulfate), nitrate (BC-nitrate), chloride (BC-Chl) and BC core itself. It should be noticed that the size distribution of BC was scaled from that of m/z 24 (C2+), as other major carbon cluster ions might be significantly affected by other ions; for example, C1+ at m/z 12 can be influenced by fragments from non-BC organics, C3+ at m/z 36 by HCl+, C4+ at m/z 48 by SO+ and C5+ at m/z 60 by C2H4O2+, etc. Similarly, the size distribution of BC-Chl was scaled from Cl+ signal at m/z 35. As shown in Fig. 3a, on average, size distributions of BC-sulfate, BC-nitrate and BC-org displayed similar patterns with a major peak at ∼550 nm (vacuum aerodynamic diameter, Dva), suggesting that they were relatively well mixed internally. However, the BC presented a remarkably different pattern with a much broader distribution and smaller peak sizes than its coating species, and in particular, relatively small particles tended to have thin coatings.

Figure 3b–f further present image plots of size distributions of the major aerosol components as a function of RBC (a surrogate of coating thickness). In contrast to the average data shown in Fig. 3a, the coating species can be roughly classified into two modes separated by RBC of ∼4.5. Most sulfate and nitrate concentrated at RBC>4.5 (Fig. 3b and c): sulfate peaked in a narrow RBC range of 5.5–6.5, while significant nitrate mass could distribute across a wider RBC range (even to RBC of ∼8.0). Only organics and chloride had a significant portion of mass distributed on relatively thinly coated BC-containing particles at RBC<4.5 (Fig. 3e and f). Specifically, they both showed a submode locating in the regime with RBC of ∼3.5–4.5 and Dva of ∼200–700 nm. These submodes suggest that organics or chloride are partially from primary sources as freshly emitted BC are more likely thinly coated. This is consistent with organics including species from fossil fuel and biomass-burning combustion, revealed by the PMF analysis. Similarly, coal burning might contribute to chloride during wintertime in Beijing (Sun et al., 2016). As for sulfate and nitrate, since they are predominantly secondary species, they would coat BC cores due to chemical aging and are therefore mostly distributed at higher RBC.

The high-resolution mass spectra of different factors of the organic coating, resolved from PMF analyses, their relative contributions and diurnal cycles of temporal variations relative to BC, are shown in Fig. 4. Figure 4a illustrates the mass profile of the fossil fuel combustion OA with BC carbon clusters (FFOA + BC). This factor had a low O/C ratio of 0.16. In this work, this factor might include emissions from both traffic and coal combustion, as it contained a series of significant PAH ion fragments in the mass spectrum (PAH fragments are negligible in other factors), indicative of coal burning (Sun et al., 2014, 2016), and presented a good correlation with C4H9+ (r2 of 0.72) – an AMS tracer ion of vehicle emissions (Zhang et al., 2005). Temporal variations in FFOA also correlated well with C9H7+ (m/z 115, r2 of 0.92) and Cl- (r2 of 0.60), which have been proposed as possible coal combustion tracer species (Yan et al., 2018; Sun et al., 2014). The FFOA / BC (Fig. 4f) appeared to be higher at nighttime than during the daytime. Note that the diurnal pattern of BC itself (Fig. 5c) was similar to that of FFOA / BC. The diurnal variations in BC might be influenced by both fossil fuel combustion activities and relatively low PBLH at nighttime. The fossil fuel combustion included coal burning and vehicle emissions (gasoline cars and the heavy-duty diesel vehicles that are only allowed to enter the city late at night). The mass ratios of different factors to BC have a smaller influence from PBLH; therefore high levels of FFOA / BC strongly indicate that co-emitted organic species with BC from fossil fuel combustion were enhanced at nighttime.

Figure 4b shows the mass spectrum of BBOA and related BC clusters. One feature of this factor is that it had relatively high fractional contributions of C2H4O2+ (1.47 % of total) and C3H5O2+ (0.95 %), which are often regarded as AMS marker ions from levoglucosan emitted from biomass-burning (Cubison et al., 2011; Mohr et al., 2009). Note that the FFOA also contained appreciable C2H4O2+ and C3H5O2+ signals, partially because coal burning (such as lignite) can emit some levoglucosan as well (Yan et al., 2018). Nevertheless, the mass fractions of C2H4O2+ and C3H5O2+ in FFOA were smaller than those in BBOA, and they correlated much better with BBOA than those with FFOA (for example, r2 of 0.90 for BBOA vs. C2H4O2+, and 0.72 for FFOA vs. C2H4O2+). The BBOA correlated very well with another biomass-burning tracer – K+(r2 of 0.90). In addition, BBOA had negligible PAH ion fragments, while the FFOA contained remarkably high PAH signals. Such characteristics are generally in agreement with previous AMS findings at the same location during wintertime in Beijing (Sun et al., 2016). For these reasons, the second factor was identified as BBOA. The diurnal pattern of BBOA / BC reached minimum during afternoon and was high overall at nighttime, similarly to FFOA / BC, indicating the nighttime enhancement of BB-related organics emissions in wintertime Beijing.

Besides the two POA factors, we also identified two secondary OA factors (OOA1 and OOA2), the O/C ratios of which were 0.45 and 0.28. OOA1 was the most oxidized OA factor that had a higher CO2+/C2H3O+ ratio than that of OOA2. The correlation between OOA1 and sulfate was better than with nitrate (r2 of 0.99 vs. 0.86). As a comparison, the less oxygenated OOA2 correlated better with nitrate than with sulfate (r2 of 0.59 vs. 0.34). These characteristics are consistent with previous AMS–PMF results (Zhang et al., 2011). In contrast to the diurnal cycles of FFOA / BC and BBOA / BC, the OOA2 / BC ratio rose significantly from early morning and peaked in the afternoon (∼ 15:00). The diurnal pattern of OOA1 / BC presented a similar peak at ∼ 15:00. This result demonstrates clear evidence of the important role of afternoon photochemical reactions to the formation of secondary organic species. However, the precursors leading to the formations of OOA1 and OOA2 remain to be elucidated. Interestingly, for OOA1 / BC, in addition to the peak during the afternoon, it increased during early evening and remained at high levels until early morning. This result indicates that nighttime aqueous-phase processing (high levels of RH at nighttime shown in Fig. 5a) can also contribute to OOA1 production. As such behavior was not observed for OOA2 / BC, it agrees with previous field and laboratory findings that aqueous-phase reactions tend to produce more highly oxygenated species (Ervens et al., 2011; Ge et al., 2012; Herrmann et al., 2015; Xu et al., 2017).

Overall, the mass fractions of BC cores that were associated with fossil fuel combustion, biomass burning, less and more oxygenated secondary processes were 32.7 %, 31.8 %, 18.7 % and 16.9 % (Fig. 4e). The organic coating of BC was predominantly primary species.

Figure 5 presents the diurnal cycles of meteorological parameters (T, RH, WS and WD), BC concentrations and RBC, mass ratios of major species to BC, gaseous species (CO, SO2 and NOx), O/C and OSc (oxidation state, defined as 2×O/C-H/C) (Kroll et al., 2011). Note that BC did not present a peak at 08:00, yet RBC, org/BC, SO42-/BC, NO3-/BC and Cl-/BC were all low at ∼08:00. This was likely attributed to an increase in the mass fractions of fresh and barely coated BC-containing particles (rather than the increase in absolute concentrations of fresh BC-containing particles) emitted during morning rush hours from traffic emissions, etc. This was consistent with the decreases in O/C and OSc and increases in CO and NO2 at 08:00 on the day. On the contrary, the RBC drop at ∼16:00 was unlikely due to the influence of the afternoon rush hour, as there were no increases in CO, NO2, and both O/C and OSc were at high levels. In fact, the 16:00 RBC drop was mainly caused by the large decrease in org / BC (as SO42-/BC, NO3-/BC and Cl-/BC did not decrease at 16:00, Fig. 5d), which were mainly the portions of fossil fuel and biomass-burning OA (Fig. 4f).

The diurnal variation in NO3-/BC peaked at ∼ 15:00–16:00, consistent with the variation in T and similar to those in the previous reports during wintertime in Beijing (Ge et al., 2017a; Sun et al., 2016), reflecting the dominant contribution of photochemical formation of nitrate. SO42-/BC showed a relatively small afternoon increase, indicating partial sulfate was produced from photochemical activities; it also presented a nighttime enhancement, similar to OOA1 / BC, suggesting the sulfate formation in aqueous-phase, consistent with the nighttime increase in RH and decrease in temperature (Fig. 5a). Due to increases in FFOA / BC, BBOA / BC and OOA1 / BC (the portion likely from aqueous-phase production), org / BC remained at high levels at nighttime. All these increases were added together, leading to high RBC at nighttime. In addition, Cl-/BC varied generally similar to those of FFOA / BC and BBOA / BC, again indicating its strong association with primary emissions.