In this study, six biomass materials and five types of coal were collected and used to generate smoke samples. The biomass materials consisted of three types of crop straw (wheat straw (WS), rice straw (RS), and corn straw (CS)) and three types of wood branches (pine wood (PW), Chinese fir (CF), and white poplar (WP)). These materials are usually used as fuels for heating and cooking in rural areas and are also occasionally burned in the field (Fan et al., 2018; V. Kumar et al., 2018). The combustion of these crop straws and woody fuels is reported to make a significant contribution to atmospheric aerosols in China (Shen et al., 2013). Five types of coal were used for the collection of CC smoke samples. They consisted of four types of bituminous coal (B-1, B-2, B-3, and B-4) and one anthracite coal (AN), representing the major types of coal used for residential CC in China. The details of these samples are provided in the Supplement.

Samples of the smoke emitted from BB and CC were collected in a combustion and sampling system. The system consisted of a combustion hood, clean background air dilution and injection ports, smoke pipe, mixing fan, mixing chamber, PM2.5 sampler (JCH-120F, Juchuang Environmental Protection Group Co., Ltd., Shandong, China), and an exhaust port. The details of the sampling procedure are described in our previous study (Fan et al., 2018; Li et al., 2018) and the Supplement.

Blank quartz filters were collected before each group of combustion experiments prior to the fuels being ignited. Blank filters were used to correct the mass of smoke, the optical signals, and DTT consumption by BrC. To prevent contamination of the following sample, the collection system was cleaned before each new combustion experiment.

In this study, the WSOC, HULIS-C, and MSOC fractions were obtained using the solvent extraction method, as described in our previous studies (Fan et al., 2016; Li et al., 2018). Initially, the filter samples were cut into small pieces and ultrasonically extracted three times with 20 mL ultrapure water for 30 min. The extract was filtered through a 0.22 µm polytetrafluoroethylene (PTFE) syringe filter (Jinteng, Tianjin, China), which collected the WSOC fraction. The HULIS-C fraction in WSOC was further isolated using the SPE method (Oasis HLB, 200 mg, Waters, Milford, MA, USA). The detailed procedure is provided in Sect. S3 of the Supplement.

The MSOC fraction was obtained using a method developed by Cheng et al. (2016). Briefly, a portion of the filter was immersed in methanol (Macklin, > 99.9 %, Shanghai, China) for 2 h and then filtered through a 0.22 µm PTFE syringe filter. Static digestion without ultrasonic treatment can avoid the loss of PM and facilitate the determination of the dissolved organic matter (DOM) content. Finally, the residual filters were dried in a vacuum dryer. The organic carbon (OC) content of MSOC was obtained by subtracting the OC concentration of the extracted filters from untreated filters.

The UV–visible (vis) absorption spectra of the BrC solutions were analyzed using a UV–vis spectrophotometer (UV-2600, Shimadzu, Kyoto, Japan). The BrC solution was placed in a 0.01 m quartz cuvette, and the UV–vis spectra were recorded from 200 to 700 nm at 1 nm intervals. Milli-Q water was used as a blank reference for the WSOC and HULIS-C solutions, while pure methanol was used as the blank for the MSOC fraction. The corresponding background was used to determine the interference from the instrument and operational blank sample.

To describe the optical properties of BrC fractions, the AAE and MAE365 were calculated in this study. The AAE is a measure of the spectral dependence of chromophores in BrC, while the MAE365 can indicate the light-absorbing capacity of BrC (Fan et al., 2016; Cheng et al., 2016). The detailed calculations are described in the Supplement.

The EEM fluorescence spectra of BrC fractions were recorded by an F-4600 fluorescence spectrometer (Hitachi, Tokyo, Japan) using a 0.01 m width quartz cuvette with a 400 V xenon lamp at room temperature and a 2400 nm/min scanning speed. The scanning ranges for excitation (Ex) and emission (Em) were 200–400 and 290–520 nm, respectively. The slit width and intervals for Ex and Em were both set to 5 nm. According to the different solvents used for sample extraction (water and methanol), all EEM spectra were divided into two groups for analysis (66 samples for water-soluble WSOC and HULIS-C and 33 samples for MSOC). The PARAFAC modeling procedure was conducted in EFC v1.2, which is an application software based on MATLAB that has the functions of conversion, correction, cognition, comparison, and calculation for processing the fluorescence spectra (He and Hur, 2015; Murphy et al., 2011, 2013). The PARAFAC analysis method that was included in the software was consistent with the calculation made by the drEEM toolkit when using MATLAB (Murphy et al., 2010, 2013). The PARAFAC was computed using two to seven component models, with non-negativity constraints and a residual analysis, and split-half analysis was used to validate the number of fluorescence components. According to the results of the split-half and core consistency analysis, four component models were chosen for both the WSOC and HULIS-C fractions and the MSOC. The EEM was normalized to the area under the ultrapure water Raman peak (Ex = 350 nm, Em = 365–430 nm) collected before the measurement of samples to produce corrected fluorescence intensities in Raman units (Lawaetz and Stedmon, 2009). The relative contribution of individual chromophores was estimated by calculating the maximum fluorescence intensities (Fmax: maximum fluorescence intensity of identified fluorescence components, relative content % = Fmax / ΣFmax) (Matos et al., 2015; Chen et al., 2016).

Approximately 5 mg of the BrC fractions (i.e., HULIS-C, WSOC, and MSOC) derived from BB and CC were used for 1H-NMR measurements. The water-soluble BrC fractions (WSOC and HULIS-C) were redissolved in 500 µL deuterium oxide, and MSOC was redissolved in 500 µL deuterated methanol and then transferred to a 5 mm NMR tube. 1H-NMR spectra were obtained at a frequency of 400 MHz using a spectrometer (Avance III 400, Bruker Daltonik GmbH, Bremen, Germany). Data were acquired from 100 scans, with a recycling time of 2 s for a condensed water sample. The length of the proton 90∘ pulse was 8.87 µs. A 1.0 Hz line-broadening weighting function and baseline correction were applied. The identification of the functional groups in the NMR spectra was based on their chemical shift (δH) relative to that of tetramethylsilane (0 ppm), which was applied as an internal standard (Zou et al., 2020).

The oxidative potential of BrC emitted from the BB and CC processes (i.e., WSOC, HULIS-C, and MSOC) was measured by a DTT assay. This protocol mainly followed the methods introduced by Fan et al. (2018) and Gao et al. (2020), with some minor modifications. Briefly, 3 mL of extracted sample solution (MSOC was a mixture of 100 µL sample and 2.9 mL of 18.2 MΩ Milli-Q water, and the corresponding blank was the same solution as that of the water blank) and 3 mL of 1 mM DTT were mixed in a 20 mL brown vial and then placed in a 37 ∘C water bath to maintain the samples at a constant temperature. At specific time intervals (0, 5, 10, 15, and 20 min), 1 mL of the well-mixed sample was transferred to another 4 mL brown vial, and 1 mL trichloroacetic acid (TCA 1 % w/v) was added to stop the reaction. Then, 0.5 mL 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB, 1 mM) was added to react with the remaining DTT to produce 2-nitro-5-thiobenzoic acid (TNB). After 5 min, 1 mL of tris(hydroxymethyl)methyl aminomethane buffer (0.4 mM Tris buffer, pH 8.9 in 4 mM) containing diethylene triamine pentaacetic acid (DTPA) was added, and the yellow color of TNB was visible in the mixed samples. The absorbance was measured at 412 nm with a UV–vis spectrometer (UV2600, Shimadzu). The DTT, TCA, and DTNB were all dissolved in 0.1 M phosphate buffer (pH 7.4) containing 1 mM DTPA, and the corresponding filter blank was analyzed to correct the DTT activity of the sample fractions. The DTT consumption rate after subtracting the field blank was determined using the absorbance and normalized by the particulate mass (DTTm, pmol/min/µg) (Verma et al., 2012; Fan et al., 2018). In this study, 1,4-phenanthraquinone was used to conduct a positive control, of which the DTT consumption rate was 0.46 ± 0.03 µM DTT/min (n=10). The rate was similar to those reported in the previous studies (Fan et al., 2018; Lin and Yu, 2019).

Finally, principal component analysis (PCA) was performed to investigate the key factors that may affect the DTT activities from a series of characteristics of the BrC fraction. The details are described in Sect. S6 of the Supplement.

Table 1 summarizes the abundance of BrC fractions, including WSOC, HULIS-C, and MSOC, in BB and CC smoke PM2.5 samples. As shown in Table 1, the average contribution of WSOC to smoke PM2.5 was 2.9 %–12 % and 2.3 %–22 % for BB and CC, respectively. These results were comparable to the results obtained for smoke samples from the combustion of cherry leaves (16 %), gingko tree leaves (6.0 %) (Park et al., 2013), corn straw (5.9 %), pine branches (6.4 %) (Fan et al., 2016), and residential coal (4 %–11 %) (Li et al., 2018) and in the ambient PM2.5 from rural and urban sites (4 %–13 %) (Matos et al., 2015; Qin et al., 2018; Wu et al., 2020). This suggests that both BB and CC can release substantial amounts of water-soluble BrC into atmospheric aerosols. As the hydrophobic fraction of WSOC, the carbon content of HULIS (HULIS-C) accounted for 1.0 %–7.8 % and 0.5 %–10 % of BB and CC smoke PM2.5, respectively. These values are comparable to the results obtained for BB smoke (5.9 %–15.2 %) (Fan et al., 2018; Huo et al., 2018), CC smoke (1.9 %–4.8 %) (Li et al., 2018), and atmospheric aerosols in Beijing (4.8 %–9.4 %) (X. Li et al., 2019), with an average value of 7.2 % ± 3.3 %, therefore confirming the important contributions made by BB and CC to atmospheric HULIS. As a comparison, the contribution of MSOC to smoke PM2.5 was 6.4 %–47 % and 9.4 %–73 % for BB and CC, respectively, with both values being much higher than the contributions of the water-soluble fractions (WSOC and HULIS-C) in the same smoke samples. Similar results have been reported in previous studies (Li et al., 2018; Cheng et al., 2016), which suggest that there are more organic compounds that could be extracted by methanol than by water, and it could therefore be a better indicator of total BrC. This result also indicated that BB and CC both released large amounts of water-insoluble BrC compounds, including hydrophobic polycyclic aromatic hydrocarbons (PAHs) and nitrogen/sulfur-containing heteroatomic PAHs (Geng et al., 2014; Dong et al., 2021; Huang et al., 2020).

Some differences were observed among the different types of smoke samples. As shown in Fig. 1, the average contributions of the WSOC and HULIS-C fractions to the total carbon (TC) were 22 % ± 7.3 % and 11 % ± 3.8 %, respectively, for BB smoke, which were higher than the corresponding values of 19 % ± 9.4 % and 8.2 % ± 4.0 % for CC smoke. The contribution of MSOC to OC was 69 % ± 19 % for BB, which was significantly lower than the value of 97 % ± 1.8 % for CC. These results suggested that BB generally released the more water-soluble OC fraction, whereas more water-insoluble OC fraction was contained in the smoke particles emitted from CC. These differences can be explained by the fact that biomass fuels generally contain large amounts of biopolymers, such as carbohydrates (cellulose, hemicellulose, etc.); the burning of biomass fuels produces more highly polar compounds, such as phenols, polyols, and polysaccharides; and CC emits more relatively hydrophobic and less polar components, such as coal tar and polycyclic aromatic species (Wu et al., 2014, 2021; Huang et al., 2020).

AAE and MAE365 are important optical indicators of the light absorption properties of atmospheric BrC and were investigated for BB- and CC-derived BrC in this study. As shown in Fig. 2a and c, the AAE values of the WSOC and HULIS-C fractions were 6.1–9.9 (mean 7.8 ± 1.6) and 7.2–9.6 (mean 8.5 ± 0.8), respectively, for BB smoke and 8.5–16 (mean 13 ± 2.9) and 10–16 (mean 14 ± 2.3), respectively, for CC smoke. These results were comparable to those measured for combustion-emitted aerosols with reported AAE values for HULIS of 7.4–8.3 (Park and Yu, 2016) and 6.2–8.1 (Fan et al., 2016, 2018) for BB smoke and 5.2–14 for CC smoke (M. Li et al., 2019). Moreover, the AAE values of BB WSOC and HULIS were also comparable to those reported for WSOC in urban aerosols in Beijing (mean 7.28 ± 0.24) (Cheng et al., 2016), HULIS in Amazon BB aerosols (∼ 7.10) (Hoffer et al., 2006), urban aerosols in Beijing (5.3–5.8) (Yan et al., 2015), and aerosols in the Tibetan Plateau (7.14–9.35) (Wu et al., 2020) but higher than that (1.2–5.4, mean of 3.2) of water-soluble BrC in Los Angeles (Zhang et al., 2013). However, the AAE values of the water-soluble BrC fraction from CC were almost higher than those in ambient aerosols, as described above. The AAE values for MSOC were 5.62–6.95 for BB smoke and 8.46–10.0 for CC smoke. It was obvious that the AAE value of BB MSOC was comparable to that of urban aerosols (average 7.10 ± 0.45) in Beijing (Cheng et al., 2016) and the reported value (5.0–6.5) for urban aerosols in India (Mukherjee et al., 2020), but the AAE values of CC MSOC were likely higher than those for urban aerosols. It is obvious that CC-derived BrC fractions (WSOC, HULIS-C, and MSOC) generally have relatively higher AAE values than ambient BrC, thereby suggesting that the contribution of CC may improve the AAE values of BrC in the atmosphere and should not be ignored.

As shown in Fig. 2a and c, the average AAE values of the WSOC, HULIS-C, and MSOC fractions in BB smoke were all lower than those for the same BrC fraction in CC smoke, indicating that BB-derived BrC had a weaker wavelength dependence than CC-derived BrC. This finding agreed with the results reported in a previous study (Fan et al., 2016). The AAE values of the BrC fraction also varied according to the type of BrC fraction. HULIS-C had the highest AAE values, which were slightly higher than those for WSOC but much higher than those for MSOC (Fig. 2a and c), indicating that water-soluble BrC fractions had a greater wavelength dependency than the corresponding MSOC. This was similar to the results of previous studies that found higher AAE values for WSOC than MSOC in ambient aerosols (Cheng et al., 2016; Kim et al., 2016) and can be explained by the fact that the strongly light-absorbing organic molecules are generally comprised of aromatic structures with a high degree of conjugation and low solubility in water.

MAE365 is an important parameter that characterizes the light absorption ability of atmospheric BrC. As shown in Fig. 2b and d, the MAE365 values of WSOC and HULIS-C were 0.9–1.5 (mean 1.2 ± 0.3) and 1.1–1.6 (mean 1.3 ± 0.2) m2/gC, respectively, for BB smoke and 0.2–0.8 (mean 0.3 ± 0.2) and 0.3–1.1 (mean 0.4 ± 0.3) m2/gC, respectively, for CC smoke. As the hydrophobic fraction of WSOC, the MAE365 values of HULIS-C in BB and CC smoke were slightly higher than that of the corresponding WSOC, suggesting that HULIS-C had a stronger light-absorbing ability. Moreover, the MAE365 values of WSOC and HULIS-C in BB smoke were comparable with the results of previous studies of the WSOC and HULIS-C fractions in combustion-released smoke particles and ambient aerosols. For example, the reported MAE365 values of WSOC and HULIS-C were 0.8–1.6 and 1.0–1.5 m2/gC, respectively, in BB smoke PM2.5 (Park and Yu, 2016; Huo et al., 2018); 0.3–1.0 and 0.5–1.4 m2/gC, respectively, in CC smoke particles (Li et al., 2018); and 0.1–1.5 m2/gC in ambient aerosols (Cheng et al., 2016; Yan et al., 2015; Zou et al., 2020). In contrast, the MAE365 values for MSOC were 1.9–2.7 m2/gC for BB smoke and 1.0–2.7 m2/gC for CC smoke, which were 1.3–8.5 times higher than the corresponding values for HULIS-C and WSOC and suggest that MSOC had the strongest light absorption capacity. The MAE365 values of BB and CC MSOC were comparable to the MAE365 values of urban aerosols in Beijing winter (average 1.45 ± 0.26 m2/gC) (Yan et al., 2015) and the water-insoluble BrC (0.85–2.45 m2/gC) in summer and winter ambient aerosols in Xi'an, northwest China (Li et al., 2020b). However, the values were higher than the MAE365 value of aerosol MSOC in the Central Tibetan Plateau (0.27–0.86 m2/gC) (Wu et al., 2020), which may be due to the relatively low combustion source contribution in this region.

As shown in Fig. 2b and d, some differences were observed among the BrC fractions. WSOC, HULIS-C, and MSOC in BB smoke all had relatively higher MAE365 values than the same BrC fractions from CC, which suggested that BrC components emitted from BB had a relatively higher light absorption ability than those from CC and may therefore have a higher radiative force (Alexander et al., 2008). This finding is important for accurately assessing the climate effects of BrC from different combustion sources.

1H NMR is an important analytical tool for the investigation of the functional groups of WSOC and HULIS in rural/urban aerosols (Fan et al., 2016; Zou et al., 2020) and rainwater (Santos et al., 2009, 2012). The typical 1H-NMR spectra of the WSOC, HULIS-C, and MSOC fractions in smoke emitted from BB crop straw (e.g., WS) and CC (e.g., B-1) are shown in Fig. 5, and the 1H-NMR spectra of other BB and CC BrC fractions are shown in Fig. S3. These BrC fractions had 1H-NMR spectra similar to those derived from atmospheric HULIS and/or WSOC in rainwater (Santos et al., 2009, 2012), BB aerosols (Fan et al., 2016), and ambient aerosols in urban and rural regions (Zou et al., 2020).

As shown in Fig. 5, the 1H-NMR spectra were mainly composed of several distinct sharp peaks superimposed on an unresolved broad band. According to previous studies and reference NMR spectra (Zou et al., 2020; Chalbot et al., 2014a, 2016), these sharp peaks can be ascribed to low-molecular-weight organic compounds, such as levoglucosan (δ3.52, δ3.67, δ4.08, and δ5.45 ppm), glucose (δ3.88–δ3.91 and δ3.81–δ3.85 ppm), and fructose (δ3.79–δ3.84 ppm) associated with BB emissions; phthalic acid (δ7.45–δ7.47 and δ7.58 ppm) and terephthalic acid (δ8.01 ppm) associated with anthropogenic activity; and the CH3 in trimethylamine (δ2.71 and δ2.89 ppm), dimethylamine (δ2.72 ppm), and monomethylamine (δ2.55 ppm) co-emitted with ammonia. The relatively few and/or weak sharp peaks in the 1H-NMR spectra of HULIS-C compared with those of WSOC may be the result of low-molecular-weight organic compounds that have been removed from HULIS-C through SPE isolation. In addition, all BB-derived WSOCs had a high intensity of sharp peaks associated with carbohydrates, such as levoglucosan, glucose, and fructose resonances, which may be released from the thermal reactions of biopolymers, such as cellulose. As a comparison, several peaks (δ0.90 and δ1.35 ppm) were observed in MSOC and were mainly located in the aliphatic region. These peaks were weaker in WSOC and HULIS-C, suggesting that more less polar aliphatic compounds were present in the MSOC fraction.

Despite some sharp peaks being identified, most of the signals in the 1H-NMR spectra of the BrC fractions presented a continuous unresolved distribution, suggesting that BrC consists of a complex mixture of organic substances (Fan et al., 2016; Chalbot et al., 2014a, 2016). As shown in Fig. 5, the functional groups of smoke BrC could be divided into four representative categories: (1) R-H with aliphatic protons in alkyl chains (0.6–1.9 ppm), including methyl (R-CH3) protons, methylene (R-CH2) protons, and methyne (R-CH) protons; (2) H-C-C= with aliphatic protons bound to carbon atoms adjacent to unsaturated groups (1.9–3.2 ppm), including carbonyl (H-C-C=O) and imino (H-C-C=N) groups or aromatic rings; (3) H-C-O with protons bound to oxygenated aliphatic carbons atoms in alcohols, polyols, ethers, and esters (3.4–4.4 ppm), generally indicating that carbohydrates and ethers were present in organic matter; and (4) Ar-H with protons bound to aromatic carbon atoms (6.5–8.5 ppm) (Fan et al., 2016; Zou et al., 2020). The distribution of the four types of protons was obtained by integrating the area of the observed 1H-NMR bands for each sample and is shown in Table 2. These functional groups were also observed in the 1H-NMR spectra of HULIS in ambient aerosols. In general, HULIS in ambient aerosols (Chalbot et al., 2014b, 2016) and rainwater (Santos et al., 2012) were characterized by the predominance of H-C (41 %–60 %), moderate contents of H-C-C= (25 %–34 %) and H-C-O (4.0 %–49 %), and a lesser contribution of Ar-H (2.0 %–6.0 %). However, it was obvious that the relative content of Ar-H groups (18 %–37 %) in HULIS-C from both combustion processes (BB and CC) was higher than the levels in ambient HULIS (Table 2), which suggests that BB- and CC-derived HULIS-C contained more aromatic structures than ambient HULIS. This was consistent with reports that more aromatic structures are observed in HULIS in colder season aerosol particles in northern China, which may be related to the amount of residential coal and straw combustion (Li et al., 2018; Sun et al., 2017).

As shown in Table 2, the relative contents of the four functional groups (i.e., R-H, H-C-C=, H-C-O, and Ar-H) varied with the type of BrC. For example, BB WSOC was always characterized by a relatively high level of oxygenated H-C-O groups and a relatively low level of aliphatic R-H groups compared with the corresponding MSOC extracted with methanol. As shown in Fig. 5, several strong signals in aliphatic R-H were also identified in MSOC, but they were weaker in the WSOC fraction. This was considered reasonable because the less polar aliphatic compounds were difficult to dissolve in water but could be extracted by methanol. As the hydrophobic fraction of WSOC, HULIS-C contained a relatively higher content of the Ar-H group and a relatively lower content of the oxygenated H-C-O group than the original WSOC for all BB and CC smoke samples. This was due to most of the low-molecular-weight oxygenated compounds not being retained by the hydrophilic–lipophilic balance cartridges and the enrichment of aromatic species (Fan et al., 2016; Zou et al., 2020).

Some distinct differences in the distribution of functional groups were also observed among the BrC fractions from BB and CC. As shown in Fig. 5, several oxygenated compounds (e.g., levoglucosan) were identified, with higher-intensity signals in the BB WSOC fraction, but they were weaker in the WSOC fraction from CC. The relative content of the H-C-O group was in the range of 34 %–54 % for the six BB WSOCs, which was higher than the values (9.0 %–34 %) for the five CC WSOCs. These oxygenated aliphatic compounds were mainly assigned to carbohydrates and polyols that may be caused by the degradation of biomass polymers such as cellulose (Fan et al., 2012, 2016; Lin et al., 2016). In contrast, the BrC fractions from CC indicated a relatively higher level of unsaturated functional groups (Table 2). For example, there was a relatively higher content of Ar-H (30 %–37 %) and H-C-C= (34 %–40 %) in the smoke HULIS-C from CC than from BB, indicating that CC HULIS-C contained more unsaturated structures, such as aromatic structures and unsaturated aliphatics (Wu et al., 2014; Dong et al., 2021; Huang et al., 2020).

The oxidative potential of the BB- and CC-derived BrC fractions (i.e., WSOC, HULIS-C, and MSOC) was investigated through a DTT assay, and the results are shown in Table S2 and Fig. 6. The DTTm value of WSOC ranged from 0.5 pmol/min/µg (B-3) to 7.4 pmol/min/µg (CS) with a mean of 3.8 pmol/min/µg. These DTTm values are comparable with those for the water-soluble fractions of BB, CC, and diesel soot (1.4 ± 0.6, 2.1 ± 2.3 and 1.1 ± 0.4 pmol/min/µg) (R. Li et al., 2019; Zhu et al., 2019) but were much lower than the ranges of 14–25 pmol/min/µg in Los Angeles wildfire aerosol samples, 22–68 pmol/min/µg in Atlanta PM2.5 samples, and 0.13 ± 0.10 nmol/min/µg in Beijing PM2.5 samples (Verma et al., 2012; Bates et al., 2019, Yu et al., 2019). These results suggested that the water-soluble fraction from BB and CC in this study had a weaker ROS generation capacity than ambient aerosols, which was likely due to the differences in the chemical composition of water-soluble fractions in BB and CC smoke particles and ambient aerosols (Lin and Yu, 2011, 2019; Dou et al., 2015; Wong et al., 2019). In general, ambient aerosols contain various sources, and the contribution of other sources, such as vehicle emissions or anthropogenic emissions, and transition metals (e.g., Fe, Cu) could increase the ability of atmospheric water-soluble fractions to produce ROS (Ma et al., 2018; R. Li et al., 2019). In addition, because of the evaporative loss of non- or less-DTT active semivolatile organic compounds, the DTT activities of BB-derived water-soluble fractions were enhanced during the aging process (Wong et al., 2019).

The DTTm values of BB- and CC-derived HULIS-C ranged from 0.5 pmol/min/µg (B-3) to 5.5 pmol/min/µg (RS) with a mean of 2.3 pmol/min/µg. These values were lower than the range (15–45 pmol/min/µg) previously reported for ambient HULIS measured with the same DTT assay (Lin and Yu, 2011; Ma et al., 2018; Verma et al., 2012). As an important component of WSOC, the DTT activity of HULIS-C accounted 63.1 % ± 15.5 % (41.4 %–90.6 %) for that of WSOC in the BB and CC samples. These values of DTTm,HULIS/DTTm,WSOC were always higher than the organic carbon contribution of HULIS-C to WSOC for the same sample (Table 1), therefore indicating that hydrophobic HULIS-C was an important redox-active fraction in the BB- and CC-derived WSOC compounds. This result was comparable with the higher oxidative contribution (64 %) of HULIS-C following water extracts from ambient aerosols in Atlanta (Verma et al., 2012). As reviewed by Win et al. (2018), this phenomenon can be explained by the specific organic species and functional groups with DTT activity in HULIS-C. As described in previous studies and in this study, the hydrophobic organic fractions isolated by the SPE column are mainly comprised of aromatic compounds (Sannigrahi et al., 2006; Fan et al., 2016; Huo et al., 2018). These compounds most likely include some of the redox-active species such as nitro-PAHs and quinones (Verma et al., 2012), which can catalyze the oxidation of cellular antioxidants and generate ROS (Verma et al., 2012; Lin and Yu, 2011). In addition, as the charge transfer intermediate, the reversible redox sites in HULIS lead to continuous ROS production (Ma et al., 2018; Lin and Yu, 2011).

The DTTm values of MSOC were in the range of 3.1 pmol/min/µg (B-4) to 84 pmol/min/µg (RS). These values were comparable to those reported in previous studies involving atmospheric aerosol methanol extracts (∼ 55 pmol/min/µg) (Verma et al., 2012). As shown in Fig. 6, the DTTm values of MSOC were much higher than those of WSOC and HULIS-C from the same smoke samples, which suggested that the water-insoluble components possessed significant oxidative properties that are relevant in toxicological studies (Verma et al., 2012). These results were consistent with the results of previous studies showing that water-insoluble compounds made the largest contribution to the oxidative potential (Verma et al., 2012, 2015).

The DTTm values of the BrC fractions varied with the type of fuel. As shown in Table S2, the DTTm values of BB WSOC were 4.5–7.4 pmol/min/µg, which was significantly higher than the range of 0.5–2.1 pmol/min/µg for CC WSOC. Similar results were also observed for the HULIS and MSOC fractions (Fig. 6). These results indicated that the BrC fractions from BB had higher oxidative potential values than those from CC and therefore more readily catalyzed the generation of ROS. Furthermore, no regular variations were observed for the oxidative potential of water-soluble BrC (e.g., WSOC and HULIS-C) in BB or CC smoke samples, but the MSOC in crop straw smoke had a much higher DTTmass value than the MSOC in smoke samples from wood burning and CC. These differences were associated with the differences in the amounts of redox-active compounds in each BrC fraction. There is a need for more studies to investigate the relationship between the molecular structures in BB smoke BrC and their DTT activities.

The BrC compounds produced by the BB and CC processes generally have different oxidative potentials. The oxidative potential values of water-soluble BrC (WSOC and HULIS-C) were much lower than those in MSOC, and the BB BrC fractions had higher oxidative potential values than CC BrC fractions. These results suggested that BrC from different sources exhibited distinct redox properties (Lin and Yu, 2011). To elucidate the association of chemical characteristics with the oxidative potential of BB and CC, principal component analysis (PCA) and Pearson correlation coefficients were conducted. Because the optical and chemical properties were all obtained based on organic matter rather than PM, the oxidative potential value normalized by the organic carbon mass (DTTOC) of each fraction was used here to present DTT activities, as well as the capacity to produce ROS. In addition, considering the statistical significance and quantity, the WSOC, HULIS-C, and MSOC data were analyzed together.

The results are shown in Fig. 7 and Table 3. It is obvious that DTTOC showed a positive loading for both principal component 1 (PC1) and principal component 2 (PC2), and DTTOC was grouped with fluorophores C4 and MAE365. These results are also given by the Pearson correlation coefficient analysis in which the DTTOC values showed significant positive correlations with the parameters MAE365 (R=0.697, p<0.01) and C4 proportion (R=0.560, p<0.01). These results suggested that fluorophore C4 and high-light-absorbing components may significantly contribute to the DTT activities of BrC compounds.

Moreover, a significant positive relationship was also observed for C4 and MAE365 (R=0.531, p<0.01), which indicated that C4 may be the main substance leading to the light absorption of BrC. As reported previously, MAE365 is related to the aromatic structure of the conjugated system (Andrade-Eiroa et al., 2013; Fan et al., 2018), and fluorophore C4 was considered to be a highly oxygenated species containing more carbonyl and carboxyl groups (Chen et al., 2016; Li et al., 2020a). Therefore, the C4 component may mainly comprise chemical species with a conjugated system and highly oxygenated species, such as quinones or aromatic acids, which were believed to be the key components for the enhancement of the ability of BrC to produce ROS (Lin and Yu, 2011; Jiang et al., 2016; Verma et al., 2012). These results also explained that the water-soluble BrC fractions in BB and CC smoke showed a relatively lower DTT consumption rate than those in ambient aerosols, in which distinctly higher contents of fluorophore C4 were observed in the water-soluble fraction (Matos et al., 2015; Chen et al., 2016).

We note that a positive correlation was observed between DTTOC and R-H and a negative correlation was observed between DTTOC and Ar-H; however, it is scientifically unreasonable. The main reason is that 1H-NMR spectroscopy only measures the concentrations of nonexchangeable hydrogen functional groups in BrC compounds. Some organic compounds not carrying nonexchangeable hydrogen atoms, such as carbonyl or carboxylic groups in BrC, cannot be detected by 1H NMR (Chalbot et al., 2014b; Paglione et al., 2014). However, some of these oxygenated functional groups likely have the ability to catalyze the generation of ROS (Lin and Yu, 2011; Verma et al., 2015). In addition, the H / C ratios of different hydrogen functional groups (i.e., R-H, H-C-C=, H-C-O, and Ar-H) are very different; thus, the relative abundances of hydrogen functional groups are difficult to compare with the carbon functional groups in BrC compounds (Decesari et al., 2007). Therefore, it is necessary that other NMR techniques such as solution-state 13C NMR and two-dimensional heteronuclear (1H-13C) NMR be used to explore the chemical functional groups associated with the oxidative potential of BrC in future studies.