Atmospheric evolution of environmentally persistent free radicals in the rural North China Plain: effects on water solubility and PM_2.5 oxidative potential

Environmentally persistent free radicals (EPFRs) represent a novel class of hazardous substances, posing risks to human health and the environment. In this study, we investigated the EPFRs in ambient fine, coarse, and total suspended particles (PM_2.5, PM₁₀, and TSPs) in the rural North China Plain, where local primary emissions of EPFRs were limited. We observed that the majority of EPFRs occurred in PM_2.5. Moreover, distinct seasonal patterns and higher g factors of EPFRs were found compared to those in urban environments, suggesting unique characteristics of EPFRs in rural areas. The source apportionment analyses revealed atmospheric oxidation as the largest contributor (33.6 %) to EPFRs. A large water-soluble fraction (35.2 %) of EPFRs was determined, potentially resulting from the formation of more oxidized EPFRs through atmospheric oxidation processes during long-range or regional transport. Additionally, significant positive correlations were observed between EPFRs and the oxidative potential of water-soluble PM_2.5 measured by dithiothreitol-depletion and hydroxyl-generation assays, likely attributable to the water-soluble fractions of EPFRs. Overall, our findings reveal the prevalence of water-soluble EPFRs in rural areas and underscore the fact that atmospheric oxidation processes can modify their properties, such as increasing their water solubility. This evolution may alter their roles in contributing to the oxidative potential of PM_2.5 and potentially also influence their impact on climate-related cloud chemistry.

1 Introduction

Environmentally persistent free radicals (EPFRs) are a new class of risk substances that have garnered significant attention in recent years (Yi et al., 2023; Vejerano et al., 2018). Unlike short-lived radicals, EPFRs are characterized by their stability, with lifetimes ranging from days to months and even being indefinite (Gehling and Dellinger, 2013; Runberg et al., 2020). They have been identified in various environmental matrices, including soil, sediment, leaves, industrial fly ash, household dust, and atmospheric particulate matter (PM) (Jia et al., 2017; Vejerano and Ahn, 2023; Zhao et al., 2019b; Filippi et al., 2022; Arangio et al., 2016). Of particular interest is their presence in inhalable atmospheric PM, which serves as a crucial carrier for EPFRs, amplifying their detrimental health impacts. Toxicological studies have demonstrated that EPFRs can induce lung damage by triggering oxidative stress in lung cells, primarily through the reactive oxygen species (ROS) generated from the catalytic cycling of EPFRs (Wang et al., 2011; Balakrishna et al., 2009; Yi et al., 2023). Moreover, the persistent nature of EPFRs and their activation potential by certain environmental factors (e.g., water molecules and ultraviolet light) enable them to participate in diverse atmospheric reactions, initiating or propagating subsequent radical reactions (Truong et al., 2010; Sarmiento and Majestic, 2023; Comandini et al., 2012). Thus, understanding the sources and properties of EPFRs is essential for assessing their atmospheric and health impacts.

The occurrence of EPFRs was initially detected in aerosols originating from cigarette smoke (Pryor et al., 1983) and later also in combustion-derived particles as well as ambient PM (Dalal et al., 1991). Various combustion and thermal processes, including industrial processes, coal combustion, biomass burning, engine exhaust, and waste incineration, have been identified as important sources of EPFRs (Liu et al., 2021; Yang et al., 2017; C. Wang et al., 2020). Mechanistic studies suggest that EPFRs can be generated and stabilized on the surfaces of transition-metal-doped particles in the post-flame and cool-zone regions of combustion systems (Cormier et al., 2006). In addition, secondary chemical processes of organic molecules may also contribute to the presence of EPFRs in atmospheric PM (Chen et al., 2019; C. Wang et al., 2020). Previous work has shown that EPFRs can be formed through the oxidation of organic molecules by ozone and photochemical reactions of PM (Borrowman et al., 2016; Sarmiento and Majestic, 2023; Qin et al., 2021; Tong et al., 2018). Various factors, including solar irradiation, humidity, and the types of precursors, may influence the formation of secondary EPFRs (Sarmiento and Majestic, 2023; Chen et al., 2019; Liu et al., 2023).

Multiple studies conducted in recent decades have investigated the composition of EPFRs in atmospheric PM (Yang et al., 2017; Arangio et al., 2016; Xu et al., 2020). The identified types of EPFRs include carbon-centered and oxygen-centered free radicals, as well as carbon-centered free radicals with a nearby heteroatom. These radicals often manifest as cyclopentadienyl, semiquinone, and phenoxyl radicals (Ai et al., 2023). Notably, Chen et al. (2018b) revealed that the dominant fraction of EPFRs existed within nonsolvent-extractable (unable to be extracted by water, methanol, dichloromethane, and n-hexane) organic matter of urban PM_2.5, underscoring the need for further exploration into the organic molecules associated with ambient EPFRs.

EPFRs exhibit redox activity, capable of reducing oxygen and facilitating the formation of ROS, such as hydroxyl radicals (•OH) and superoxide radicals (${\mathrm{O}}_{\mathrm{2}}^{-}\mathrm{•}$) (Hwang et al., 2021; Guo et al., 2020). Consequently, EPFRs may serve as crucial hazardous components contributing to the toxicity of atmospheric PM. Li et al. (2023) found that EPFRs can contribute to the oxidative toxicity of both water-soluble and water-insoluble fractions of atmospheric PM. The types of EPFRs and their extractability may influence their roles in ROS formation (Zhao et al., 2019b). Moreover, ambient EPFRs have been detected in both fine and coarse particles, with observed seasonal and spatial variations in their size distribution (Jia et al., 2023; Wang et al., 2022). The presence of EPFRs in different particle sizes may pose various health risks to humans due to differences in deposition efficiency within the respiratory tract.

The investigations of airborne EPFRs in urban areas, heavily influenced by traffic, industrial, and residential emissions, have been the primary focus of previous studies (Yang et al., 2017; C. Wang et al., 2020). However, EPFRs, characterized by their long lifetimes, can undergo transport over considerable distances and reach rural areas with minimal local emissions. Indeed, the long-range transport of EPFRs has been demonstrated (Chen et al., 2018a). During the transport, the characteristics of EPFRs may undergo evolution through atmospheric chemical processes, potentially altering their roles in ROS formation. Despite this, investigations into the characteristics of airborne EPFRs in areas with limited local emissions remain sparse. Insights into airborne EPFRs in such areas will allow a better understanding of the atmospheric transformation and fate of EPFRs, as well as their atmospheric and health effects.

Therefore, to enhance our understanding of EPFR evolution during atmospheric transport and its effects on ROS formation by the corresponding PM, we collected yearlong PM samples in a typical rural area located in the North China Plain (NCP), where local combustion emissions contributing to EPFRs are minimal. In addition, northern China, including the NCP, is one of the most polluted regions in China, characterized by significant combustion sources, making the selected location ideal for our research objectives. Specifically, our study aims to (i) investigate the characteristics of EPFRs, including their concentration, size distribution, and seasonal variations in the studied region; (ii) determine the sources of EPFRs; and (iii) explore the roles of EPFRs' speciation in contributing to the oxidative potential (OP) of PM.

2 Methods

2.1 PM sample collection

Ambient PM samples were collected at a rural site (36°51^′48^′′ N, 115°00^′58^′′ E) in the county of Quzhou in the NCP from April 2022 to March 2023. The sampling site represents a typical rural environment, predominantly surrounded by croplands and devoid of significant local industrial sources (Fig. S1 in the Supplement). Sequential 24 h fine, coarse, and total suspended particles (PM_2.5, PM₁₀, and TSPs) were collected using a high-volume sampler (TH-1000CII, Tianhong, China) at a flow rate of 1.05 m³ min⁻¹. The PM samples were collected onto prebaked (900 °C) quartz filters (Munktell, type MK360) and stored at −20 °C until analysis. In total, 95 PM samples were collected during the whole sampling period. A summary of the number of PM_2.5, PM₁₀, and TSP samples collected in each season is provided in Table S1 in the Supplement. In addition, data of other air quality parameters (SO₂, NO₂, O₃, and CO) were obtained from the monitoring site nearest to the sampling site from the local environmental monitoring center.

2.2 EPFR analysis

Three punches (1.2 cm² per punch) of each filter were inserted into a quartz tube (5 mm i.d., SP Wilmad-LabGlass) for EPFR measurements using an electron paramagnetic resonance (EPR) spectrometer (A300-9.5/12, Bruker). The detection parameters for EPFRs were set as follows: a modulation frequency of 100 kHz, a microwave frequency of 9.8485 GHz, a microwave power of 1.76 mW, a modulation amplitude of 1.00 G, a sweep width of 150 G, a time constant of 81.92 ms, and a receiver gain of 1 × 10³ G. All EPR measurements were conducted at room temperature. To minimize noise in EPR signals, a baseline correction was performed, followed by fitting the signals using a Gaussian function via the least-squares method. EPFR concentrations for the filters were determined by comparing the peak area with a calibration curve (Fig. S2) generated using a common radical stand, 4-hydroxy-2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPOL) (Arangio et al., 2016).

In addition, we conducted water extraction and acidification experiments on the samples to determine the reduction of EPFR contents on the filter samples after treatment. The procedures of water extraction and acidification were adopted from previous studies (Chen et al., 2018b; Yang et al., 2017). For water extraction, all PM_2.5 samples were processed alongside two randomly selected PM₁₀ and TSP samples from each season. Briefly, three punches (1.2 cm² per punch) of each filter were immersed in 3 mL of deionized water for 14 h under dark conditions. Regarding acidification, two PM_2.5, PM₁₀, and TSP samples in each season were randomly selected, and two punches (1.2 cm² per punch) of each filter were immersed in 2 mL of 6 M HCl solution for 1 h under dark conditions. Subsequently, the filter punches were dried by a vacuum freeze dryer before conducting EPFRs measurements using the aforementioned detection parameters. All the EPFR measurements were conducted within 1 year of sampling.

2.3 Carbonaceous fractions and element analyses

A 1.0 cm² punch of each sample filter was analyzed for organic carbon (OC) and elemental carbon (EC) following the Interagency Monitoring of Protected Visual Environments (IMPROVE) thermal–optical reflectance (TOR) protocol using the DRI model 2001 carbon analyzer. Different carbonaceous fractions, including OC1–OC4 and EC1–EC3, were isolated and quantified based on their thermal stability.

A total of 12 elements (Li, Mg, Al, Si, K, Ca, Cr, Mn, Fe, Cu, Zn, and Pb) were analyzed using inductively coupled plasma mass spectrometry (ICP-MS; Thermo Scientific iCAP RQ). Prior to analysis, a 3.6 cm² punch of each filter was digested using 1 mL of aqua regia (HNO₃+3HCl, v:v) at 99 °C and a rotational frequency of 350 rpm for 24 h. After digestion, the extracts were filtered through a 0.22 µm PTFE syringe filter and then diluted to 5 mL deionized water with 2 % HNO₃.

2.4 Oxidative potential measurements

The OP of PM samples was measured by two techniques, the dithiothreitol-depletion assay and •OH-production assay (Verma et al., 2012; Son et al., 2015). In the dithiothreitol (DTT) assay, the decay of 100 µM DTT by PM in phosphate buffer was monitored over a 40 min incubation at 37 °C. The remaining DTT after the incubation was quantified by its reaction with dithiodinitrobenzoic acid, yielding ultraviolet-detectable 2-nitro-5-thiobenzoic acid (Verma et al., 2012; Fang et al., 2015). In the •OH-production assay, the terephthalate (10 mM) in phosphate buffer was used to measure •OH formation by PM throughout 2 h incubation at 37 °C. At pH 7.4, terephthalate reacted with •OH to form stable and highly fluorescent hydroxyterephthalic acid (2-OHTA), and the production rate of •OH was calculated based on the produced 2-OHTA, as the formation of 2-OHTA is proportional to the generation of •OH (Yu et al., 2022; Li et al., 2019). Note that samples were incubated at a PM concentration of 100 µg mL⁻¹ for both assays. Meanwhile, both total OP (Total-OP) and water-soluble OP (WS-OP) were determined in this work. For total OP determination, unfiltered PM extracts with filter punches left in the extracts were directly incubated with the probes, while for water-soluble OP, the extract was filtered through a 0.22 µm PTFE syringe filter before incubating with probes. The OP contribution from water-insoluble PM components (water-insoluble OP, WIS-OP) was considered the difference between Total-OP and WS-OP. A detailed description of the DTT and •OH assays can be found in the Supplement (Sect. S1).

2.5 Statistical analysis

2.5.1 Positive matrix factorization

The EPA positive matrix factorization (PMF 5.0) model was employed to apportion the sources of EPFRs and PM in this study. The PMF model is an advanced multivariate factor analysis tool widely utilized for source apportionment of environmental pollutants (Heo et al., 2013; Wang et al., 2019). The input data include the concentrations and uncertainties of PM, EPFRs, organic carbon fractions (OC1, OC2, OC3), elemental carbon fractions (EC1, EC2, EC3), SO₂, NO₂, CO, O₃, and the 12 elements. The uncertainties of each variable were calculated using the following equation: uncertainty = K × concentration, where K denotes analytical uncertainty (Wang et al., 2019). For PM, EPFRs, OC, and EC, K was set as 10 % (Jang et al., 2020). For metal elements and meteorological parameters, K was set as 15 % (Ikemori et al., 2021). Missing values and associated uncertainties were estimated by substituting the median concentrations of the components and 4 times the median value of the components to mitigate their impact on the results (Wang et al., 2019).

The PMF model was run with four to seven factors and with random seeds. The six-factor result was considered the optimal one based on a minimal Q value (indicating the variation between the observation and the model is the least), $Q_{\text{Robust}}/Q_{\text{Theo}}$ (< 2), scaled residuals (within ± 3), and the interpretable profiles from the literature (Reff et al., 2007; Ramadan et al., 2000). Bootstrap and displacement analyses were conducted to estimate the uncertainty of the PMF model with six factors (Brown et al., 2015), and the results are shown in Fig. S3. The bootstrap factor mapping exceeded 93 % for all factors without any displacement run exchanges.

2.5.2 Correlation analysis

Correlation analysis was performed using Pearson's correlation coefficients and two-tailed significance tests by IBM SPSS Statistics 27.

2.6 Backward trajectories

To characterize the origins and transport pathways of the air masses to the sampling site, 48 h backward trajectories of the air masses were simulated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model. The input meteorological data were acquired from the Global Data Assimilation System (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1, last access: 7 April 2023).

3 Results and discussion

3.1 Characteristics of EPFRs in different sizes of PM

Figure 1a illustrates the box plots of volume-normalized (EPFRv) concentrations of EPFRs in different sizes of PM during the entire sampling. EPFRv levels in PM_2.5, averaging (5.6 ± 1.1) × 10¹² spins m⁻³, accounted for over 95.2 % of those in PM₁₀ ((5.8 ± 1.0) × 10¹² spins m⁻³) and TSPs ((5.9 ± 1.1) × 10¹² spins m⁻³). However, the average mass concentration of PM_2.5 only represented 61.8 % of that in PM₁₀ and 47.5 % of that in TSPs (Fig. S4). Similar results were found for EPFRv and PM concentrations in each season (Fig. S5). These results suggest that the majority of airborne EPFRs are present in PM_2.5, with only a small portion occurring in the 2.5–10 µm size range.

Figure 1 The concentrations of EPFRv (a, c) and EPFRm (b, d) in different sizes of PM samples and different seasons. The boxes represent the 25th percentile (lower edge), median (solid line), mean (solid dot), and 75th percentile (upper edge). The whiskers represent the minimum and maximum.

This is further evident by the results of mass-normalized (EPFRm) concentration of EPFRs, as depicted in Fig. 1b. The average EPFRm in PM_2.5 (6.5 ± 3.5) × 10¹⁶ spins g⁻¹) was approximately 1.7 times that in PM₁₀ ((3.9 ± 2.1) × 10¹⁶ spins g⁻¹) and 2.2 times that in TSPs ((3.0 ± 1.7) × 10¹⁶ spins g⁻¹). Consistent with our finding, previous studies have also observed a predominant fraction of EPFRs in smaller sizes (< 2.5 µm) of PM (Chen et al., 2020; Dugas et al., 2016). The strong association of EPFRs with fine particles may stem from multiple factors. Firstly, airborne EPFRs primarily originate from the combustion of organic materials and multiphase chemistry of gaseous organics in the atmosphere, processes that predominantly form fine particles (Tian et al., 2009; Arangio et al., 2016). Additionally, the larger specific surface area and porous structure of fine particles facilitate the attachment of EPFRs, leading to their higher retention within this fraction (Yang et al., 2017).

Table S2 presents a summary of the concentrations of EPFRv and EPFRm in this study compared to the existing literature. Both EPFRv and EPFRm in this study were lower than in most urban and suburban environments (Wang et al., 2019; Jia et al., 2023; Yang et al., 2017). This implies that there were only limited local sources directly emitting EPFRs at this rural site. This is further supported by the seasonal variation of EPFRs. Notably, higher EPFRv levels were observed in summer than in winter (Fig. 1c), with the highest EPFRm occurring during summer (Fig. 1d) in the studied area. This contrasts with findings from previous studies, which indicated that winter typically exhibits the highest EPFRv, especially for cities in northern China with prominent local emission sources of EPFRs, such as large-scale coal consumption (Wang et al., 2019; Jia et al., 2023; Ai et al., 2023). The distinct seasonal characteristic of EPFRs observed in our study could be attributed to the limited residential and industrial activities in the rural NCP. On the other hand, the highest EPFRm levels in summer may be linked to enhanced atmospheric oxidation, promoting EPFR formation due to elevated concentrations of photooxidants and stronger solar irradiation (Borrowman et al., 2016; Sarmiento and Majestic, 2023; Shiraiwa et al., 2011).

In addition to examining the concentrations of EPFRs, we also investigated the g factors of EPFRs across different sizes of PM, as the g factor is an important indicator of EPFR characteristics. We observed that the g factors for all PM samples ranged between 2.0033 and 2.0037, with a mean value of 2.0035 ± 0.0001 (Fig. S6). These g factors are higher than the values obtained from PM samples collected from highway and urban locations (Hwang et al., 2021; Fang et al., 2023). A g factor below 2.003 signifies the presence of carbon-centered radicals (e.g., cyclopentadienyl), while a g factor exceeding 2.004 indicates oxygen-centered radicals (e.g., phenoxy, semiquinone radicals). A g factor falling between 2.003–2.004 suggests a carbon-centered radical with adjacent oxygen atoms (Ai et al., 2023; Yang et al., 2017). The observed g factors in this work indicate the presence of carbon-centered radicals with adjacent oxygen atoms and oxygen-centered radicals. Furthermore, we noted a decreasing trend in the g factor with particle size (i.e., PM_2.5 > PM₁₀ > TSPs), indicating closer proximity of the unpaired electron to the oxygen center in smaller particles. This trend could be attributed to the increased exposure of porous structures in smaller particles, rendering them more susceptible to oxidation (Yang et al., 2017).

3.2 Sources of EPFRs

The PMF analysis was utilized to identify and quantify the sources contributing to EPFRs. As shown in Fig. 2a, six primary source factors were determined. Factor 1 was attributed to atmospheric oxidation, characterized by its elevated proportion of O₃. The presence of O₃ signifies atmospheric oxidative capability (Ainur et al., 2023; Y. Q. Wang et al., 2020). Factor 2 exhibited high proportions of Al, Fe, Cr, SO₂, CO, and EC, suggesting industrial emissions. Al and Fe could be associated with iron and steel production and the metal industry (Khobragade and Ahirwar, 2022). The availability of Cr is linked to fossil fuel combustion or oil combustion (Alleman et al., 2010; Begum et al., 2011). SO₂ and CO are typical tracers of coal combustion (Johnson et al., 2006; Kundu et al., 2010; Wang et al., 2019), and they (and also EC) could be emitted from coal-based industries. Factor 3, characterized by high proportions of SO₂, NO₂, and CO, was identified as coal combustion (Johnson et al., 2006; Kundu et al., 2010; Wang et al., 2019). Previous research indicates that NO₂ emissions, although often associated with vehicle emissions, are also prevalent in coal combustion (Ainur et al., 2023; Lei et al., 2016; C. A. Wang et al., 2018). In factor 4, high proportions of OC3, OC4, EC1, EC2, EC3, Mn, Cu, and Pb were observed, suggesting motor vehicle emissions. EC1, OC2, OC3, and OC4 primarily originate from gasoline vehicle emissions, while EC2 and EC3 are closely associated with diesel vehicle emissions (Kim and Hopke, 2004; Ai et al., 2023). Additionally, Mn (from unleaded gasoline additives), Pb (from gasoline additives), and Cu (from brake linings) contribute to this factor (Sharma et al., 2014). Factor 5 displayed a high loading of OC1 and a considerable loading of OC2, indicative of biomass burning. OC1 and OC2 have been linked to biomass combustion in previous studies (Stanimirova et al., 2023; Cao et al., 2005; Dong et al., 2022). Factor 6 exhibited high proportions of Mg and Ca, indicative of soil dust as these elements are major constituents of the Earth's crust (An et al., 2015; Y. Liu et al., 2022).

Figure 2 (a) Factor profile obtained by positive matrix factorization analysis. (b) Seasonal and annual contributions of the six factors to EPFRs.

In summary, the six primary sources contributing to EPFRs were identified as atmospheric oxidation, industrial emissions, coal combustion, motor vehicle emissions, biomass burning, and soil dust. As shown in Fig. 2b, atmospheric oxidation emerged as the largest contributor (33.6 %) to EPFRs during the entire sampling period. The formation of EPFRs through atmospheric oxidation could occur via photooxidant oxidation (reactions of polycyclic aromatic hydrocarbons (PAHs) with solar radiation, O₃, and •OH) and metal-catalyzed redox reaction (PAHs interacting with metals under visible and UV light irradiation). Solar radiation and O₃ can directly generate EPFRs by disrupting the stable covalent bonds of organic substances, while metals act as electron acceptors, facilitating electron transfer with PAHs to form EPFRs, as suggested by previous studies (Borrowman et al., 2016; Shiraiwa et al., 2011; Chen et al., 2019; C. Wang et al., 2020).

The significant contribution of atmospheric oxidation observed in our study contrasts with findings from urban areas, where primary sources typically dominate airborne EPFRs (Ainur et al., 2023, 2022). The results of seasonal source apportionments (Fig. 2b) further illustrate much higher contributions of atmospheric oxidation in spring (42.2 %) and summer (50.3 %) compared to autumn (21.3 %) and winter (18.6 %). The elevated contribution in summer aligned with the expectation of stronger photochemistry. The relatively high contribution in spring could be attributed to the dominance of long-range transport of air mass (60 %, Fig. S7a), which was markedly higher than the proportions (< 23 %) observed in the other seasons. Previous studies have suggested that long-range transport of air mass favors atmospheric oxidation occurrence more than the air mass from regional/short-range transport (Ramya et al., 2023; Zhong et al., 2022).

Industrial emissions (30.8 %) were the second-largest contributor to annual EPFRs, followed by coal combustion (14.5 %), motor vehicle emissions (10.3 %), biomass burning (6.6 %), and soil dust (4.2 %). Industrial emissions and coal combustion have been widely suggested as significant sources of EPFRs (C. Wang et al., 2020; Yang et al., 2017; P. Wang et al., 2018). Although industrial activities were limited at this rural site, emissions might originate from surrounding industrial cities such as Handan in the south, as indicated by backward trajectory analysis (Fig. S7). Over 17 % of air mass was identified as regional transport from the south, particularly notable in autumn (54 %), corresponding to a higher contribution (44.7 %) of industrial emissions to EPFRs compared to other seasons. Winter exhibited notably higher contributions from coal combustion and biomass burning, as coal and biomass are primary fuels for residential heating in the rural NCP during this season.

While motor vehicle emissions accounted for a small contribution to EPFRs, they constituted the largest contributor (43.8 %) to PM (Fig. S8), suggesting differing components contribute to EPFRs and PM. For instance, volatile organic compounds (e.g., toluene) and nitrogen oxides emitted from vehicle exhausts may significantly contribute to PM formation but are generally not considered major contributors to EPFRs (Nagpure et al., 2016; Gao et al., 2022; Wang et al., 2019). Regarding soil dust, recent studies indicate that PAHs readily adsorb on mineral surfaces, likely forming “cation–π” interactions with active sites. This interaction promotes electron transfer from aromatic compounds to surface cations on clay surfaces, facilitating the formation of intermediate radicals/final products (Zhao et al., 2019a; Ni et al., 2023).

3.3 Associations of EPFRs with OP of PM

3.3.1 OP of PM

Oxidative potential, defined as the catalytic generation of ROS by PM components, serves as a plausible metric for assessing PM toxicity (Abrams et al., 2017; Weichenthal et al., 2016; Daellenbach et al., 2020; Bhattu et al., 2024). EPFRs have been identified as significant contributors to the OP of PM due to their ability to catalytically generate ROS (Gehling et al., 2014; Hwang et al., 2021). In this work, we also measured the OP of the PM samples using two commonly used assays: dithiothreitol-depletion (OP^DTT) and hydroxyl-generation (OP^•OH) assays. OP^DTT is found to be a good indicator of the production of ${\mathrm{O}}_{\mathrm{2}}^{-}\mathrm{•}$ and H₂O₂ but does not capture •OH generation (Fang et al., 2019). Thus, the combined application of two assays provides complementary insights into the role of EPFRs in ROS generation. Further, WS-OP, WIS-OP, and Total-OP were all determined to explore their potential correlations with EPFRs.

Tables S3 and S4 summarize the OP results in this work. A detailed discussion of OP values between the present work and literature can also be seen in the Supplement (Sect. S2). In short, consistent with the findings of EPFRm, the mass-normalized OP (OPm) values in this work were also lower than in most urban and suburban environments. This disparity suggests fewer redox-active PM components in the studied rural area, likely due to the absence of significant local emission sources of pollutants.

Figure 3a illustrates the box plots of mass-normalized WS-OP, WIS-OP, and Total-OP determined by the two assays. Generally, Total-OP levels were higher than WS-OP and WIS-OP, demonstrating measurable contributions of both water-soluble and water-insoluble species to the overall OP of PM. Additionally, WS-OP accounted for a larger fraction of Total-OP in both assays (WS-OP^DTT: 67.8 ± 20.5 %; WS-OP^•OH: 56.1 ± 22.6 %). Furthermore, a reverse relationship between particle size and the contribution of WS-OP to Total-OP was observed, with the largest contribution in PM_2.5, followed by PM₁₀ and TSPs (Fig. 3b). A significantly higher contribution of WS-OP^DTT to Total-OP^DTT has also been observed for ambient PM_2.5 at multiple locations worldwide (Gao et al., 2017; Yang et al., 2024; Li et al., 2024). Yet, PM samples near highways and road dust have shown a higher fraction of WIS-OP^DTT than WS-OP^DTT (Zhang et al., 2024; Li et al., 2023). This could be due to a lower solubility of PM at the sites near primary emissions. These divergent results suggest that atmospheric oxidation processes may alter the role of PM components in contributing to OP.

Figure 3 (a) The concentrations of total, water-soluble (WS), and water-insoluble (WIS) fractions of OP^DTT and OP^•OH. The boxes represent the 25th percentile (lower edge), median (solid line), mean (solid dot), and 75th percentile (upper edge). The whiskers represent the minimum and maximum. (b) Proportions of WS-OP and WIS-OP to Total-OP in different sizes of PM samples; the bar indicates the standard error.

3.3.2 Associations among EPFRs, OP, and PM components

To identify individual chemical species influencing the intrinsic redox activity of ambient PM, correlation analyses between mass-normalized OP (${\mathrm{OP}}^{\mathrm{DTT}/\mathrm{•}\mathrm{OH}}$) and the mass fraction of determined chemical species were performed. The results are shown in Fig. 4, and the detailed information is also listed in Tables S5–S7 (the volume-normalized correlation results are also included in Fig. S9 in the Supplement in case the readers are interested).

Figure 4 Correlation coefficients (Pearson's r) of mass-normalized OP (Total-OP, WS-OP, and WIS-OP) with mass fractions of selected chemical species.

Notably, the chemical species associated with Total-OP and WS-OP exhibited similarities, particularly for PM_2.5 (Fig. 4). This may be partly because Total-OP was primarily comprised of WS-OP (Fig. 3b). Also, good correlations were observed between Total-OP and WS-OP, except for the OP^•OH of TSP samples (Fig. S10). In addition, we observed that OP^DTT and OP^•OH were sensitive to different chemical species (Fig. 4). Specifically, OC, EC, EPFRs, Fe, and Cr showed good correlations with Total-OP^DTT and WS-OP^DTT, while Cu exhibited a good correlation with Total-OP^•OH and WS-OP^•OH (r > 0.5, p < 0.05). As mentioned before, OP^DTT is a good indicator of the production of ${\mathrm{O}}_{\mathrm{2}}^{-}\mathrm{•}$ and H₂O₂ (Fang et al., 2019). Thus, these results indicate that different chemical species contribute to the production of distinct types of ROS (Campbell et al., 2021; Jin et al., 2019; Calas et al., 2018). In terms of WIS-OP, only OC showed a moderate correlation (r = 0.43, p < 0.05) with WIS-OP^DTT. No species displayed a positively good correlation with WIS-OP^•OH, implying that WIS-OP might arise from complex synergistic–antagonistic interactions among PM components (Charrier and Anastasio, 2015; Yu et al., 2018). However, further research is warranted to elucidate the mechanisms and PM components responsible for WIS-OP.

Interestingly, EPFRs in PM_2.5 exhibited significant correlations with WS-OP but not with WIS-OP (Fig. 4). This observation contradicts previous studies reporting that the majority of EPFRs are not water-extractable (Chen et al., 2018b; Guo et al., 2023; P. Wang et al., 2018). Given the source apportionment results showing atmospheric oxidation as the largest contributor to EPFRs, we hypothesized that atmospheric oxidation processes may have increased the water solubility of rural EPFRs, leading to the observed significant correlations.

3.3.3 Solubility of EPFRs and its linkages with OP

To demonstrate the possibly high solubility of EPFRs, additional experiments were conducted with PM filters extracted by water prior to EPFR analysis. On average, a 35.2 % reduction in EPR signals was observed after water extraction (Fig. 5), indicating that the same percentage of EPFRs was water-soluble (a detailed breakdown of the water-soluble fraction of EPFRs in each filter is listed in Table S8). This result is substantially higher than the reported water-soluble fraction (0.2 %) of EPFRs in urban PM_2.5 of Xi'an (Fig. 5b), where the majority of EPFRs likely consisted of graphene oxide analogues (Chen et al., 2018b). In addition, the result is also higher than that (11 %) of EPFRs emitted from biomass burning (Guo et al., 2023), suggesting a high water-soluble fraction of EPFRs in the studied rural area.

Figure 5 (a) Gaussian fitting EPR spectra of a selected PM_2.5 filter sample (sampling date: 12 March 2023) before (black) and after water extraction (orange), as well as after acidification (purple). A typical EPR spectrum of a blank filter (grey) is also present for reference. (b) The percentages of different fractionated EPFRs in the rural NCP (this study), in urban PM of Xi'an (Chen et al., 2018b), and in biomass burning particles (Guo et al., 2023).

It is widely acknowledged that EPFRs from combustion processes are formed through the electron transfer mediated by transition metals on organic combustion byproducts such as PAHs (S. T. Liu et al., 2022; Vejerano et al., 2018). The EPR signal of these metal–EPFR complexes can be substantially reduced (> 90 %, Fig. 5c) after acidification, as demonstrated by previous work (Guo et al., 2023). We also conducted the same acidification procedure and observed, on average, a 70 % reduction in EPR signals (Fig. 5c; detailed information is provided in Table S9). Although the reduction in our work is not as substantial as that of biomass burning particles (Guo et al., 2023), the result indicates that there was a large fraction of EPFRs in the form of metal–EPFR complexes, likely originating from combustion sources (e.g., coal and biomass combustion). In addition, EPFRs showed significantly good correlations (r > 0.5, p < 0.01; Table S10) with certain transition metal species (i.e., Fe, Zn, Cr), suggesting their formation was likely associated with transition metals. The higher water-soluble fraction of EPFRs in our work compared to that of biomass burning particles could be relevant to the increased polarity of EPFRs by atmospheric oxidation processes (i.e., the aging of metal–EPFR complexes). It is also worth noting that the reduction of EPR signals is at a much greater extent than that (24 %, Fig. 5c) of EPFRs (mostly consisting of graphene oxide analogues) in urban PM_2.5 of Xi'an (Chen et al., 2018b).

In addition to the potentially oxidized/aged metal–EPFR complexes, another significant source of water-soluble EPFRs could be EPFRs directly generated by secondary chemical processes. Chen et al. (2019) found that water-extracted humic-like substances were the primary precursors of secondary EPFRs formed by visible-light illumination in ambient PM. A laboratory study by Tong et al. (2018) demonstrated that secondary organic aerosol formed by photooxidation of naphthalene contained EPFRs at levels comparable to those found in ambient PM. Given that atmospheric oxidation was the main source of EPFRs and PM in our studied region, these secondary chemical processes likely had a substantial impact on ambient PM, contributing to the formation of secondary EPFRs. However, further research is needed to fully elucidate and quantify the contribution of secondary formation pathways to water-soluble EPFRs.

Figure 6 Correlation of WS-EPFRs and Total-EPFRs with WS-OP in PM_2.5: (a) WS-EPFRs with WS-OP^DTT, (b) WS-EPFRs with WS-OP^•OH, (c) Total-EPFRs with WS-OP^DTT, and (d) Total-EPFRs with WS-OP^•OH. The Pearson correlation coefficients (r) and associated p values are illustrated in the figure. The lines and shaded areas are linear regressions with their 95 % confidence intervals.

Figure 7 Implication of the atmospheric evolution of EPFRs. EPFRs, primarily emitted from combustion sources, may undergo atmospheric oxidation during long-range or regional transport. This transformation may enhance the water solubility of EPFRs, potentially heightening their bioavailability and consequently elevating risks to human health.

Importantly, compared to total EPFRs, water-soluble EPFRs (WS-EPFRs) showed stronger correlations with WS-OP (Fig. 6) of PM_2.5. In addition, no significant correlation was observed between water-insoluble EPFRs (WIS-EPFRs) and WS-OP (Fig. S11a and b). These results demonstrate our hypothesis that the significant correlations between EPFRs and WS-OP were driven by the water-soluble fractions of EPFRs, while atmospheric oxidation processes had increased the water solubility of EPFRs. On the other hand, the lack of significant correlation between WIS-EPFRs and WIS-OP (Fig. S11c and d) could be attributed to the complex interplay of organics and metals affecting WIS-OP (Gao et al., 2020). Nonetheless, the significant correlation between WS-EPFRs and WS-OP suggests that atmospheric oxidation processes may have contributed to the increased water solubility of EPFRs, thereby affecting their roles in PM_2.5 oxidative potential.

4 Conclusions and implications

In this study, EPFRs in fine, coarse, and total suspended particles in a typical rural region of the NCP have been investigated. The majority of EPFRs occurred in fine particles. EPFRs exhibited seasonal patterns distinct from those observed in urban environments. In addition, higher g factors of EPFRs were identified compared to those reported for highway and urban PM, suggesting a greater extent of oxidation in rural EPFRs. These findings underscore a unique characteristic of EPFRs in rural areas, where local primary emissions of EPFRs are limited.

Additionally, atmospheric oxidation was resolved as the largest contributor to EPFRs. The atmospheric oxidation processes occurring during long-range or regional transport were suggested as the primary driver behind the more oxidized EPFRs observed in our work. We also demonstrated that the rural EPFRs contained a higher water-soluble fraction compared to those found in biomass burning particles and urban PM. Our results thus emphasize the importance of considering the water-soluble fraction of EPFRs, alongside their occurrence in organic solvent-extractable or non-extractable organics, particularly in regions without significant primary emissions of EPFRs. Future studies are warranted to investigate airborne EPFRs in other rural regions to yield a more complete understanding of the sources and properties of EPFRs, as only a typical rural site in the NCP was examined in our work.

The WS-EPFRs could be an important contributor to the oxidative potential of PM_2.5, as suggested by their significant positive correlations. While prior research has predominantly focused on EPFRs in urban environments or originating from primary combustion emission sources, our findings revealed the evolution of EPFRs through atmospheric oxidation during transport (Fig. 7). The atmospheric evolution of EPFRs may modify their properties, such as water solubility, thereby altering their roles in contributing to the oxidative potential of PM (S. Y. Wang et al., 2018). It is also important to note that water-soluble EPFRs are more bioavailable than their insoluble forms. They are more likely to reach the tissues or organs beyond the lung deposition site, amplifying the risk to human health (Liu and Ng, 2023). Furthermore, beyond influencing the oxidative toxicity of PM, the role of evolved EPFRs in climate-related cloud chemistry may also change and should be explored in future research.

Generally, soluble EPFRs are assumed to have a relatively short lifetime. For instance, it has been found that secondary EPFRs formed by photoexcitation are highly susceptible to reduction by oxygen, leading to short lifetimes ranging from approximately 30 min to 1 d (Chen et al., 2019). However, measurements conducted on several PM_2.5 filters stored at −20 °C for roughly 1 year showed that EPR signals only decreased by approximately 11 % (Fig. S12), suggesting the persistency of EPFRs, including their water-soluble fraction. Previous research has indicated that semiquinone-type radicals adsorbed into a polymeric carbonaceous core (Valavanidis et al., 2005) or by electron transfer with transition metals (Truong et al., 2010) are stable and can have a lifetime longer than months. However, their water solubility remains unclear. Further efforts to identify the types of water-soluble EPFRs and investigate their lifetimes are warranted.

Overall, this study revealed the characteristics of EPFRs in rural regions, highlighting the importance of considering atmospheric oxidation processes in understanding their behavior and impacts. Future research efforts should aim to further elucidate the mechanisms governing the formation, evolution, and fate of EPFRs, as well as their interactions with other atmospheric components. Such insights are essential for developing effective strategies to mitigate the adverse effects of EPFRs on air quality and public health, as well as for advancing our understanding of their broader implications for atmospheric and environmental science.