A total of 30 oxygenated organic species, including dicarboxylic acids, polycarboxylic acids, alcohols, and multifunctional compounds, were included in this study (Table 1). Small amounts of a standard organic compound were dissolved in ultrapure water, which was subsequently used to generate pure organic particles using a constant output atomizer (Model 2076, TSI). For generating rBC–organic mixed particles, REGAL black pigment (REGAL 400R, Cabot Corp) and a standard organic compound were mixed in the bulk solution for atomization. REGAL black was used in this study because it has been suggested as an rBC standard for calibrating the LV scheme of the SP-AMS (Onasch et al., 2012). Atomized particles were subsequently dried by a diffusion dryer using silica gel to minimize the interference of particle-phase water to H2O+ signals and other related fragments (i.e., O+ and HO+). Pure argon gas was used for atomization and dilution to minimize the interference of gaseous N2 and CO2 to the quantification of CO+ and CO2+ signals in organic mass spectra (Canagaratna et al., 2015a; Corbin et al., 2014; Willis et al., 2014).

The details of the SP-AMS (Aerodyne Research, Inc.) have been reported in detail by Onasch et al. (2012). In brief, the SP-AMS is equipped with a thermal vaporizer (i.e., a heated tungsten surface) and a laser vaporizer (i.e., a continuous-wave intra-cavity 1064 nm Nd:YAG laser). While the thermal vaporizer operated at 600 ∘C can vaporize non-refractory particulate matter (NR-PM, including organic, sulfate, nitrate, ammonium, and chloride), the laser vaporizer is designed for vaporizing rBC-containing particles at which rBC cores can be gradually heated up to ∼ 4000 K. During this heating process, organic coatings can be vaporized at a lower range of temperature (likely <600 ∘C) depending on the volatility of each organic compound. The vaporized analytes are ionized using 70 eV electron impact (EI) ionization, and the ions are subsequently detected by a high-resolution time-of-flight mass spectrometer operated in V mode (Canagaratna et al., 2007; DeCarlo et al., 2006).

The SP-AMS instruments were operated in two different vaporization schemes for characterizing pure organic particles and rBC–organic mixed particles, respectively. When the laser vaporizer of the SP-AMS was off, the instrument was operated as a standard HR-ToF-AMS to facilitate flash-vaporization of pure organic particles (TV scheme). For the second part of the experiments, the thermal vaporizer was removed from the SP-AMS and the laser vaporizer was turned on for measuring standard organic compounds coated on REGAL black exclusively (LV scheme; Fig. 1c). Note that some pure organic particles might be generated through atomization of the rBC-organic mixture, but they cannot be detected using the LV scheme. Observations from three SP-AMSs were reported in this study, and they are labeled as SP-AMS 1, 2, and 3. SP-AMS 1 and 2 were used to generate new data for 18 and 20 organic species, respectively. Data of SP-AMS 3 (10 organic species) were extracted from Canagaratna et al. (2015b). Table 1 summarizes the tested species for each SP-AMS. The three SP-AMSs were operated by different researchers from the National University of Singapore (SP-AMS 1), University of Toronto (SP-AMS 2), and Aerodyne Research (SP-AMS 3).

The raw data of SP-AMS measurements were processed by the AMS data analysis software (Squirrel for unit mass resolution (UMR) data and PIKA 1.21b for high-resolution peak fitting; Sueper, 2015), and statistical analyses were processed by R (version 3.6). Given that pure argon gas was used for particle generation and dilution in all the experiments, CO+ signals could be quantified in the high-resolution aerosol mass spectra. Note that only small N2+ signals were detected in some of the experiments using SP-AMS 1 due to the residual N2 desorbed from the desiccant. For the results of rBC-organic mixed particles, the interference of refractory COx+ signals (rCOx, i.e., CO+ and CO2+) formed during rBC vaporization to organic mass spectra was corrected in the fragmentation table using C1+ : C3+, CO+ : C3+, and CO2+ : C3+ ratios obtained from pure REGAL black particles. REGAL black calibrations were performed for each instrument, and thus the applied corrections for the fragmentation table were instrument specific. Elemental analysis was performed using the I-A method (Canagaratna et al., 2015b) to calculate H:C and O:C ratios of each organic compound. The average carbon oxidation state of each organic compound was also calculated based on the method reported in Kroll et al. (2011) (i.e., OSC=2×O:C-H:C).

Figure 2 shows the normalized mass spectra of azelaic acid (a dicarboxylic acid) measured by the SP-AMS 1 for illustrating different fragmentation patterns generated using the TV and LV schemes. The comparison for azelaic acid shows that the LV approach can produce larger organic fragments (e.g., m/z 60, 69, 73, 83, and 84) compared to the TV approach. Similarly, arabitol (a sugar alcohol) and levoglucosan (a chemical marker for biomass burning OAs) show less fragmentation when they were vaporized with rBC particles by the laser vaporizer (Figs. S1 and S2). The normalized cumulative histograms of m/z for the organic compounds measured by the SP-AMS 1 are presented in Fig. 2c (i.e., the average of 18 species), which clearly shows that the curve shifts toward larger m/z when the compounds were vaporized by the LV scheme. The green dashed line represents the differences between respective fragments from the TV (blue line) and LV (red line) schemes. It shows a decreasing trend within the red shaded region, illustrating that the LV scheme could generate more organic fragments starting from m/z 55 on average. Overall, our observations support the general hypothesis and previous observations (Canagaratna et al., 2015b) that a thermal vaporizer operated at 600 ∘C tends to generate organic mass spectra with smaller molecular fragments compared to the LV scheme, in which organic vaporization and fragmentation can occur at a lower range of temperature, for the same organic compound. It is worth mentioning that although organo-nitrate compounds, which have significant contribution to SOA mass in some locations (Fry et al., 2018; Lee et al., 2019a; Xu et al., 2015, 2017), were not tested in this work, changes in fragmentation for ammonium nitrate (i.e., AMS calibration standard) due to the vaporization scheme were observed. Our observations show that the NO+/NO2+ ratios of ammonium nitrate measured using the TV and LV schemes are equal to 1.9 and 0.7, respectively, suggesting that the quantification of organo-nitrate compounds based on the NO+/NO2+ ratios approach can also be affected using the vaporization scheme (Farmer et al., 2010; Xu et al., 2015).

The organic fragments of C2H3O+ and CO2+ are the two dominant peaks observed in ambient OOA components identified by the PMF analysis of standard HR-ToF-AMS measurements (i.e., TV scheme) (Ng et al., 2011). The relative importance of the two fragments varies between OOA components identified at the same locations, and the ratio of fC2H3O+/fCO2+ (fi+= a mass fraction of m/z i+ to total organic) usually decreases with the degree of oxidative aging (Ng et al., 2011, 2010). Based on the observations from SP-AMS 1 and 2, Fig. 3a shows that most of the organic species detected using the LV scheme gave higher fC2H3O+/fCO2+ ratios compared to the TV scheme. Note that dicarboxylic acids and multifunctional organic compounds show stronger enhancement compared to alcohols. The average of fC2H3O+/fCO2+ ratios measured by the TV and LV schemes are 0.48 (±0.52) and 3.07 (±3.59), respectively, indicating less thermal-induced decarboxylation with the gradual vaporization.

The observed enhancement of fC2H3O+/fCO2+ ratios in this work suggests that the f44 vs. f43 (or fCO2+ vs. fC2H3O+) observational framework developed based on the HR-ToF-AMS datasets worldwide by Ng et al. (2010) may not be directly applicable for evaluating the degree of aging and characteristics of OOA coatings measured using the LV scheme of the SP-AMS. Figure 3a compares the ambient OOA components determined using the TV and LV schemes (i.e., concurrent HR-ToF-AMS and SP-AMS LV scheme measurements) at the same location if their time series of mass concentrations are strongly correlated (i.e., R>0.75). Table S1 summarizes the characteristics of the OOA components observed in Beijing summer (Xie et al., 2019a; Xu et al., 2019); Beijing winter (Wang et al., 2019; Xie et al., 2019b); Tibet (Wang et al., 2017; Xu et al., 2018); and Fontana, CA (Chen et al., 2018; Lee et al., 2017), that are used for our comparison. It can be found that most of the OOA factors determined using the LV scheme gave significantly higher fC2H3O+/fCO2+ ratios compared to those determined using the TV scheme by factors of 6–18, which is similar to those observed for dicarboxylic acids and multifunctional organic compounds (Fig. 3a). Therefore, the differences in fC2H3O+/fCO2+ ratios between the ambient OOA components determined using the two vaporization schemes can be partially explained by the laboratory observation reported in this section. Nevertheless, we cannot rule out the possibility that the chemical compositions of the OOA coatings on BC particles were different than those externally mixed with BC particles despite their strong temporal correlations (see more discussion on this topic in Sect. 3.6 based on the elemental analysis).

C2H4O2+ is a tracer fragment often associated with biomass burning OAs (BBOA), and fC2H4O2+ is commonly used to identify the presence of BBOA and to evaluate their degree of aging (Bozzetti et al., 2017; Cubison et al., 2011; Milic et al., 2017). Figure 3b shows that most of the species from SP-AMS 1 and 2 (i.e., 97 % of the tested species) showed the enhancement of fC2H4O2+ when they were detected using the LV scheme regardless of their functional moieties. The average enhancement factor of fC2H4O2+ for alcohol, dicarboxylic acids, and multifunctional groups are 2.62 ± 0.92, 2.90 + 0.78, and 2.69 ± 1.11, respectively. Levoglucosan, a known cellulose-derived compound produced during biomass burning (Simoneit et al., 1999), gives an enhancement of fC2H4O2+ by a factor of 2.33. Different to fC2H3O+/fCO2+ ratios that show scattered data between vaporization schemes, a strong linear correlation was obtained for all the tested species (R=0.95) with the slope equal to 2.45. This strong linear correlation suggests that the fC2H4O2+ increased to a similar extent for most of the tested oxygenated organic species.

Together with the changes in fCO2+ caused by the vaporization schemes, our observation suggests that the f44 vs. f60 (or fCO2+ vs. fC2H4O2+) observational framework developed by Cubison et al. (2011) has to be used cautiously for evaluating potential influences of biomass burning emissions on the chemical composition of OOA coatings. For example, Rivellini et al. (2020) observed that the laser vaporization approach led to the enhancement of fC2H4O2+ for total OAs, less oxidized OOA (LO-OOA), and more oxidized OOA (MO-OOA) in an urban environment by comparing their laser-off and laser-on measurements using a dual-vaporizers scheme (i.e., the SP-AMS switched between the TV scheme and TV + LV scheme during operation). Wang et al. (2019) and Xie et al. (2019b) identified BBOA factors in Beijing winter from their concurrent SP-AMS and HR-ToF-AMS measurements, respectively, and a higher fC2H4O2+ value was observed for the BBOA factor determined using the TV scheme. However, it is important to note that the two BBOA factors were weakly correlated (R=0.42), and thus they likely represented BBOA materials from different origins.

Elemental analysis of pure OAs and organic coatings on rBC particles was performed based on the I-A method. Tables S2–S4 summarize the H:C, O:C, and OSC values of all the organic compounds measured by the three SP-AMSs. Figure 4 compares the measured H:C, O:C, and OSC values of all the organic compounds generated using the LV scheme to their true values. While the measured H:C ratios are scattered around the 1:1 line (Fig. 4a), the measured O:C and OSC values are generally lower than their corresponding true values (Fig. 4b and c). Note that some organic species, such as glycolic acid and glutamic acid, give relatively large discrepancies between the measured and the true O:C values that can be due to their low contribution of CHO+ and CO2+ signals to the total organic mass (e.g., 1 %–2 %). The average relative errors of H:C and O:C ratios for individual SP-AMSs varied from -3.4 % to 11.5 % (mean = 6.6 %) and from -37.1 % to -22.0 % (mean = -26.3 %), respectively (Table 2). Note that there are no statistical differences (ANOVA, p<0.05) between the relative errors of elemental ratios determined by the three instruments, suggesting that the elemental analysis is not strongly instrument dependent. For the thermal vaporization approach, the average relative errors of H:C (mean = -5.2 %) and O:C (mean = -21.5 %) ratios determined in this work are similar to the measurement uncertainties of HR-ToF-AMSs previously reported by Canagaratna et al. (2015a).

Figure 5 compares the elemental ratios of standard organic compounds determined using the LV and TV schemes. The data points from the three independent SP-AMSs are well-aligned with each other even though a large fraction of the tested organic species (∼ 50 %) were not repeated between the different SP-AMS instruments. The linear fits of all the measured data demonstrate that O:C and H:C ratios determined by the LV scheme differ from their corresponding values determined using the TV scheme by factors of 0.89 and 1.10, respectively. Canagaratna et al. (2015b) conducted similar comparisons for 10 organic species (i.e., data from SP-AMS 3 in this work) based on the results obtained from the A-A method, reporting that the O:C and H:C values determined using the LV scheme differ from their corresponding values determined using the TV scheme by factors of 0.83 and 1.16, respectively. The uncertainty of O:C and H:C ratios is further reduced to approximately ±10 % between the TV and LV schemes on average in this study. Figure S4 shows that the I-A method with the scaling factor applied can improve the accuracy of elemental ratios of oxygenated species in general except for the H:C ratios of alcohol group. While the fragmentation of oxygenated organic species due to the TV and LV scheme can be significantly different, this work illustrates that their elemental compositions can be comparable to the I-A method applied for laboratory-generated particles with a single oxygenated organic species.

The I-A method has been widely used for the elemental analysis of ambient OAs measured using the TV scheme of HR-ToF-AMSs, and this approach has been described in detail by Canagaratna et al. (2015a). In brief, chemical standards, including dicarboxylic acids, multifunctional acids, and alcohols, were tested using the TV scheme. The elemental ratios were first determined by the A-A method developed by Aiken et al. (2008), and they were subsequently corrected by a multi-linear regression (MLR) model based on the fraction contributions of CHO+ and CO2+ fragments (i.e., fCHO+ and fCO2+) to address the composition dependence in the OA fragmentation, which is referred to as the I-A method. Note that fCHO+ and fCO2+ can be used as surrogates for alcohol and acid groups, respectively, and they are major peaks observed in ambient OOAs (Canagaratna et al., 2015a; Duplissy et al., 2011; Takegawa et al., 2007). As illustrated in the previous sections, OA fragmentation can be significantly different between the LV and TV schemes, and hence an updated multilinear regression parameter is conducted to check whether the model accuracy can be improved for analyzing data generated using the LV scheme. Following the approach for developing the I-A method (Canagaratna et al., 2015a), updated multiple linear regression parameters for determining H:C and O:C ratios of organic coatings were obtained based on the data from two SP-AMSs (1 and 2) as shown in Eqs. (1) and (2), respectively. 1H:CI-ASP=H:CA-A×[0.90+1.02×fCHO++2.78×fCO2+],2O:CI-ASP=O:CA-A×[1.74-2.50×fCHO++1.93×fCO2+], where H:CI-ASP and O:CI-ASP are the elemental ratios obtained from the new fitting parameters (denoted as I-ASP method hereafter) and H:CA-A and O:CA-A are the elemental ratios obtained by the A-A method.

Given that most of the tested organic species were not the same for each SP-AMS, a direct inter-instrument comparison was not possible. Hence the average relative errors obtained from the two instruments were used to evaluate the performance of the I-ASP method across all the tested species. The relative errors of I-ASP method for H:C, O:C, and OSC were 6.3 %, 5.8 %, and -9.8 %, respectively, for the LV scheme data. While the I-ASP method leads to a substantial improvement of the relative errors of O:C compared to the I-A method (i.e., from -26.3 % to 5.8 %), the relative average errors of H:C ratio obtained from I-ASP and I-A method are comparable. As shown in Fig. 4d–f, the H:C, O:C, and OSC values calculated using the I-ASP method are better aligned with the 1:1 line compared to those determined using the I-A method (Fig. 4a–c) with smaller root mean squared error (RMSE) reported in Table S5. However, it is worth noting that the I-ASP method gives the highest positive bias for the O:C ratio for alcohol species compared to the results from the I-A method (with and without applying the scaling factor) as illustrated in Fig. S4. For the H:C ratios, the I-ASP method can reduce relative error for alcohol species but generate a larger range of errors for dicarboxylic acids and multifunctional species.

There is an increasing number of field studies operating a standard HR-ToF-AMS and an SP-AMS concurrently (i.e., total OAs measured by the TV scheme vs. organic coatings measured using the LV scheme) to investigate the mixing state of BC particles and the effects of primary emissions and atmospheric processing on the formation of organic coatings on BC particles (Lee et al., 2017; Massoli et al., 2015; Wang et al., 2020). In particular, whether SOA materials condensed on BC particles have similar chemical characteristics to those externally mixed with BC particles remains poorly understood. Previous observations in urban environments have reported that the mass spectral features of ambient OOA components identified by the PMF analysis were significantly different between the two co-located measurements even though their temporal variabilities strongly correlated to each other (Chen et al., 2018; Lee et al., 2017; Liu et al., 2019; Massoli et al., 2015; Wang et al., 2020; Xu et al., 2019; Zhao et al., 2019). Given our observations that vaporization scheme plays a critical role in the fragmentation process of oxygenated organic species, the LV elemental analysis scaling factors (0.89 for H:C and 1.10 for O:C) and the I-ASP method obtained in this work can facilitate more direct and robust comparison between the two types of measurements based on elemental analysis. Nevertheless, it is important to note that rBC and organics were likely more homogeneously mixed within our laboratory-generated particles, which can be very different from the morphology of ambient organically coated BC particles. Therefore, such inter-instrument comparisons assume that BC morphology is not a key factor to affect organic fragmentation observed using the LV scheme.

Figure 6 shows the elemental ratios of OOA components observed from previous field studies conducted in California Research at the Nexus of Air Quality and Climate Change (CalNex) 2010 campaign (Massoli et al., 2015), Fontana, California, in 2015 (Chen et al., 2018; Lee et al., 2017), Tibet in 2015 (Wang et al., 2017; Xu et al., 2018), Beijing winter in 2016 (Wang et al., 2019; Xie et al., 2019b), and Beijing summer in 2017 (Wang et al., 2020; Xie et al., 2019a; Xu et al., 2019). Correlations of hourly-averaged mass concentrations between different PMF factors identified by the TV and LV schemes were investigated, and only strongly correlated OOA factors (R>0.85) were included in Fig. 6. If an OOA component correlated well to multiple PMF OA factors identified by another vaporization scheme, the comparison was only performed for a pair of OA factors that gave the strongest correlation (see Table S1). Note that a transported BBOA factor identified in Tibet, which was associated with a significant amount of OOA materials (R=0.96), was also included in this comparison. Figure 6a and b show the adjusted H:C and O:C ratios measured using the LV scheme (y axis) using the inter-conversion factor and the I-ASP method, respectively. The error bars of each data point represent the average absolute errors of the I-A and I-ASP method obtained in this study. Note that both I-A and I-ASP methods likely over-estimate O:C ratios of organics coated on ambient BC particles because of the contributions of refractory COx+ fragments (e.g., CO+ and CO2+) usually remain unresolved.

In Fig. 6a, although a few H:C and O:C ratios were well-aligned onto the 1:1 line, it can be found that the majority of the O:C ratios measured by the LV scheme with the LV elemental analysis scaling factors applied (i.e., LV-OOA from CalNex, OOA-2 from Fontana, BBOA from Tibet, OOA-1, OOA-2 from Beijing winter and OOA-1 from Beijing summer) were still lower than those measured using the TV scheme after considering the uncertainties of I-A method. In addition, some H:C ratios of OOA coatings were higher than the OOAs measured using the TV scheme. When the I-ASP method was used, the O:C ratios generally increased compared to those determined using the I-A method with the inter-conversion factors applied (Figs. 6b and S3). In particular, LO-OOA from CalNex gave the largest enhancement of O:C ratio, followed by OOA-2 from Fontana and SV-OOA from CalNex. The H:C ratios remained roughly the same between the two methods. Combining the results presented in Fig. 6a and b, it can be concluded that the OOA materials associated with rBC particles were likely less oxygenated (i.e., lower O:C) compared to the total OOAs measured using the TV scheme at some of the sampling locations. This observation may be due to the fact that OOAs formed from heterogeneous oxidation of POA such as HOA that are co-emitted and internally mixed with BC are likely to be less and/or non-oxygenated in nature. For example, our comparison shows the largest differences of O:C ratios between the two vaporization schemes for the observations from Beijing (OOA-2 in winter and OOA-1 in summer), where the air quality was expected to be significantly influenced by local combustion sources. The O:C ratio of an aged/transported BBOA factor detected using the LV scheme of the SP-AMS in Tibet is also noticeably lower than that of MO-OOA detected by the HR-ToF-AMS even though they are strongly correlated (R=0.96). Overall, our observations indicate that even though the time series of OOA factors determined by the TV and LV scheme are strongly correlated (e.g., R>0.9), suggesting that they were likely co-emitted or formed through similar aging processes during transport, they might contain multiple types of OA materials and their relative distribution between rBC and non-BC particles might be significantly different.