Characteristics and Sources of Environmentally Persistent Free Radicals in PM 2.5 in Dalian: Important Role of Polycyclic Aromatic Hydrocarbons

Environmentally persistent free radicals (EPFRs) are an emerging class of environmental hazardous contaminants that extensively, stably exist in airborne particulate matter and pose harmful effects on human health. However, there was little research about the sources of EPFRs in actual atmospheric conditions. This study reported the occurrence, characteristics and sources of EPFRs and polycyclic aromatic hydrocarbons (PAHs) in PM 2.5 collected in Dalian, China. The concentrations of PM 2.5 -bound EPFRs ranged from 1.13×10 13 to 8.97×10 15 spins/m 3 (mean value: 1.14×10 15 spins/m 3 ). Carbon-centered radicals and carbon-centered radicals with adjacent oxygen atoms were detected. The concentration of ∑ PAHs ranged from 1.09 to 76.24 ng/m 3 and PAHs with high molecular weight (HMW) were predominant species in PM 2.5 . The correlation analysis and PMF result showed that coal and biomass combustion are the top contributors to EPFR, followed by vehicle emission. The secondary sources to EFPRs was negligible. The nding of present study provides an important evidence for further study on the formation mechanism of EPFRs in actual atmospheric to control the air pollution. organic carbon-centered free radicals and carbon-centered radicals with adjacent oxygen atoms. The source appointment result indicated that the primary sources, i.e. fuel and biomass combustion, vehicle emission, are the dominate sources of PM 2.5 -bound EPFRs in Dalian. As for actual atmospheric environment, the portion of secondary EFPRs associated O 3 is negligible. Over all, this study provides a better insight the sources of PM 2.5 -bound EPFRs in Dalian. The study warrants further attention on the formation mechanism of PM 2.5 -bound EPFRs in atmospheric environment.


Introduction
Environmentally persistent free radicals (EPFRs), present within ne atmospheric particulate matter, have caused concern because they are demonstrated to be detrimental to human health (Chen et al., 2018a;Fahmy et al., 2010;Reed et al., 2015). EPFRs, in contrast to common short lifetime radicals (superoxide radicals and hydroxyl radicals for instance), have longer lifetimes up to months or years Gehling and Dellinger, 2013).Toxicological studies have con rmed that EPFRs can generate reactive oxygen species (ROS) in different media, which lead to oxidative stress on cells (Gehling et al., 2014;Khachatryan et al., 2014;Wang et al., 2020). Thus, investigation on the occurrence, characteristics and sources of particulate EPFRs will be of great importance for assessing their potential health risks.
Studies have shown that ambient particulate matters with different sizes could contain large amounts of EPFRs with different characteristics (Arangio et al., 2016;Chen et al., 2020;Shaltout et al., 2015;Yang et al., 2017b). For example, Yang et al. (2017b) detected the EPFRs in different size of ne particulate matter during haze days in winter in Beijing with concentration ranging from 1.49 to 1.95×10 20 spins/g, and found the particulate matter with aerodynamic diameters < 1 µm had a highest EPFRs level. Chen et al. (2020) also found the level and type of EPFRs in ne particulate matter (< 2.1 µm) are different with those in coarse particles (2.1-10 µm). Other studies also found the EPFR levels in airborne particulate matter in different cities, such as Taif, Saudi Arabia, Denver, United States and Xuanwei, China (Runberg et al., 2020;Shaltout et al., 2015;Wang et al., 2018). Till now, a few researches have been published on evaluating the concentrations and sources of EPFRs, but information are merely limited to several cities with limited samples. Furthermore, atmospheric particulate matter may contain different kinds of EPFRs according to the g factor (g > 2.004, oxygen-centered radicals; 2.003 < g < 2.004, carbon-centered radicals with adjacent oxygen; g < 2.003, carbon-centered radicals) in different regions and in different seasons (Chen et al., 2018b;Ruan et al., 2019). Thus, further investigations on the characteristics of EPFRs in different regions are still necessary.
Currently, su cient investigation on the formation mechanism and the sources of particulate matterbound EPFRs is still lack. Many studies have reported that particulate matter bound EPFRs were related to formation and adsorption of substituted aromatic molecules and transitional metal oxide (such as copper oxides, ferric oxides) on the particle formed in thermal or combustion process (Gehling and Dellinger, 2013;Reed et al., 2014;Vejerano et al., 2012;Yang et al., 2017a). In addition, EPFRs could also be formed during the photochemical decomposition process of organic pollutants, indicating that burning process was not the sole source of EPFRs (Borrowman et al., 2016;Chen et al., 2019b;Jia et al., 2019b;Zhu et al., 2019). Furthermore, the sources of PM 2.5 -bound EPFRs are different in different cities with different meteorological conditions, which need further exploration (Chen et al., 2018b;Shaltout et al., 2015;Wang et al., 2019a). PAHs, an important precursor of EPFRs, is a typical class of organic species released into atmospheric environment from different sources resulting ubiquitous presence in atmospheric environment (Jia et al., 2016;Jia et al., 2017;Jia et al., 2018). PAHs are regarded as "seasonal contaminants" because of the variations of concentrations and composition between different seasons (Ravindra et al., 2008). The variations are caused not only by the different meteorological conditions (e.g., humidity, temperature) but also by the presence of different sources (Ravindra et al., 2008;Tian et al., 2009). Although previous studies have con rmed that PAHs can form EPFRs during photochemical reaction or oxidation by O 3 in laboratory condition, the relationship between EPFRs and PAHs in actual atmospheric is still unknown (Borrowman et al., 2016;D'Arienzo et al., 2017;Jia et al., 2016;Jia et al., 2017;Jia et al., 2019b;Jia et al., 2018;Yang et al., 2017a). Moreover, PAHs from different sources have different congener compositions which is also a key factor in the relationships. In addition, criteria pollutants (e.g., ozone, sulfur dioxide, carbon monoxide, nitric oxide and PM) in atmospheric environment are typical pollutants that may also play an important role during the formation process of EPFRs (Borrowman et al., 2016;Chen et al., 2018b;Wang et al., 2019a). However, information on the source apportionment of EPFRs by the precursors-PAHs and criteria pollutants in actual atmospheric condition is still missing.
The motivation for the present work was an interest in the sources of PM 2.5 -bound EPFRs in Dalian. In present study, the concentrations and characteristics of PM 2.5 -bound EPFRs and PAHs samples collected in non-heating and heating periods were investigated. In addition, the statistic correlation between the concentrations of PAH, six kinds of criteria pollutants and EPFRs and the positive matrix factor (PMF) model was also performed. Based on the results of this work, the dominant source of EPFRs in PM 2.5 in Dalian was identi ed.

Chemicals and reagents
Page 4/21 The details of the chemicals and reagents and their supplier was shown in the Supplementary Material.

Sampling Information
PM 2.5 samples were collected for 48 h in quartz lter at a ow rate 100 L/min by a low-volume air sampler (Zhongrui, ZR-3920C, China). The lter was baked at 400 ℃ and stored in valve bag at -20 ℃ before the sampling. A total of 46 PM 2.5 samples were collected during heating and non-heating period from November 2018 to May 2019. The site was located west of the Dalian University of Technology campus in Dalian, a northeast city in China. All lters were wrapped with aluminized paper, and stored at -20 ℃ until pretreatment. The concentrations of six conventional atmospheric pollutants, namely CO, NO 2 , PM 2.5 , PM 10 , SO 2 , and O 3 were recorded from the daily report according to local environmental monitoring stations during sampling period.

EPFRs analysis
The PM 2.5 -bound EPFRs were measured with a Bruker EXM A-200 spectrometer (Bruker, Bermen, Germany). The experiment was performed based on the method of previous report (Yang et al., 2017b).
The lters with collected PM 2.5 were cut into pieces and steeped in 20 ml dichloromethane (DCM) for 12 h in dark environment. And then use ultrasonic extraction to extract for 30 min. After transfer the supernatant to another sample bottle, the lter pieces were rinsed in triplicate. The supernatant liquid and rinse solutions were fully mixed and then evaporated to approximately 100 μL with a nitrogen blower. The typical operation parameters of EPR were set as follows: scan width, 100 Gauss; center eld, 3400 G; microwave frequency, 9.84 GHz, detection time 178 s, number of detections, 5; modulation amplitude, 0.20 mT; microwave power, 8.0 mW. Radical quanti cation for PM 2.5 -bound EPFRs was calculated by comparing the DI/N, as calculated from double integration of the rst derivative signal divided by normalized constant, to a 2,2-diphenyl-1-picrylhudrazyl (DPPH) standard, and the concentrations were normalized to the volume of air. The calibration of DPPH was shown in Fig. S2. The nal result was given in spins/m 3 which means the radical concentration per m 3 of air. The value of g factor was estimated using Bruker's WINEPR software. All measurements were performed at room temperature.
The spin number is calculated using equation (1): Here S DPPH is the total number of DPPH spin. DI/N EPFRs and DI/N DPPH were obtained by double integration of the rst derivative signal divided by normalized constant.

PAHs analysis
Brie y, half of the lter samples were cut into pieces, spiked with ve labelled PAHs (Chrysene-D12, Perylene-D12, Acenaphthene-D10, Naphthalene-D8, Phenanthrene-D10), and placed in dark environment for 12 h subsequently, nally extracted twice with n-hexane: Dichloromethane (DCM) (1:1, v/v) using ultrasonic. The extract solutions were condensed to ~0.5 mL with nitrogen blower and then cleaned up by a multilayer column lled with 1 cm anhydrous sodium sulfate and 4 cm neutral silica gel from top to bottom. PAH congeners were eluted with 20 mL mixture (DCM: n-hexane=1:1, V/V). The eluate was concentrated to about 0.3 mL with nitrogen blower. Before GC/MS analysis, a known amount hexamethylbenzene as the internal standard was added in the samples for quanti cation .
PAHs were determined by GC-MS (Shimadzu, GCMS-QP2020, Japan) with electron ionization (EI) in SIM model. The 15 kinds of PAH congeners were separated in a DB5-MS capillary column (0.25 mm i.d. × 30 m, 0.25 μm lm thickness). 1 μL sample was injected in the splitless mode. The helium carrier gas was at a ow rate of 1 mL/min. The oven temperature started at 80 ℃ for 3 min, increased to 300 ℃ at 6 ℃/min, and then held for 3 min. The temperature of transfer line was 300 ℃ and that of ion source was 250 ℃.

Quality assurance and quality control (QA/QC)
The limit of quanti cation (LOQ) and the limit of detection (LOD)were de ned as ten and three times of the ratio of single to noise. The LOD and LOQ were in range of 0.32-16 pg/m 3 and 1.08-53.6 pg/m 3 (Table   S1). Target compounds in samples with concentration below the LOQ were assigned a value of nondetected (n. d.). The recoveries of surrogate standards of PYR-d12, CHR-d12, PHE-d10, ACE-d10, NAP-d8, and were 97.2%, 108.9%, 84.6%, 67.6%, and 53.5%, respectively. The result was corrected by the surrogate recoveries.

Statistical analysis
Person' correlation analysis was performed between the concentrations of EPFRs, conventional pollutants (CO, SO 2 , PM 2.5 , PM 10 , NO 2 and O 3 ) and PAH congeners for investigating the potential sources of PM2.5-bound EPFRs in Dalian. Statistically signi cant correlation was set at a p < 0.05. The concentration of some PAH congeners below the LOQs, were set as half the LOQ for statistical analysis. All statistical analyses were conducted with IBM SPSS Statistics 24.
Furthermore, the EPA PMF 5.0 was used for EPFRs source appointment. Time series of PAH congeners and conventional pollutants and EPFRs were used as input le of this model. Generally, 3 to 6 factors were compared to optimize the result, and the 5 factors and 20 of runs were chosen. PAHs concentrations below the method detection limit (MDL) were half of the detection limit and their overall uncertainties (Unc) were set with the formula where k are analytical uncertainty and C is the measured chemical concentrations. K of six conventional pollutants is 5%, and k of PAHs and EPFRs is 10%. Missing values and uncertainty were input by the median concentration of the component and four times the median values to decrease their in uence on the results Qian et al., 2020;.
3 Results And Discussion

Temporal variation of EPFRs
To probe the potential difference of PM 2.5 -bound EPFRs between the non-heating period and heating period, samples were collected every month and monitored by EPR at room temperature, the g factor, concentration and ΔH p−p of EPFRs associated with PM 2.5 were displayed in Table S3. The EPFRs associated with PM 2.5 with a g factor of 2.00312 ± 0.00034 (range: 2.00222-2.00393). The single, unstructured peak (shown in Fig. S1) and the g factor indicated both carbon-centered radicals and carbon-centered radicals with adjacent oxygen atoms were exist in PM 2.5 in Dalian. The ΔH p−p of PM 2.5bound EPFRs in present study was from 4.3 to 9.1 G (average value: 6.7 G). The temporal variation of concentration of EPFRs was shown in Fig. 1 and  (Guo et al., 2020). Therefore, the concentrations of EPFRs in Dalian are one or two orders of magnitude lower than those in Beijing, and one or two orders of magnitude higher than those in Nanjing.   (Niu et al., 2019). Figure 3 showed the abundance pro les of individual PAHs during different periods. The 4-ring PHAs (FLA, PYR, BaA, CHR) had the highest abundance pro le (49%), followed by 5-ring PAHs (BkF, BaP, BbF, DahA) (30%), while 3-ring (FLU, ANT, ACY, ACE, PHE,) and 6-ring (IcdP, BghiP) PAHs showed a relatively low pro les in heating period. In non-heating period, 5-ring and 4-ring PAHs also had high pro les. Compared with the heating period, there was a signi cant increase in the pro les of 3-ring (from 9-18%),

Temporal variation of PAHs
while the proportions of 4-ring and 5-ring PAHs decreased to 45% and 29%, respectively in non-heating period. The various pro les of PAH congeners in different periods were mainly caused by the different emission sources. The PAHs were divided into two groups: high-molecular-weight (HMW, four or more aromatic rings) and low-molecular-weight (LMW, with two or three aromatic rings) congeners (Delgado-Saborit et al., 2014;Janssen et al., 2014). HMW PAHs were dominant PAH congeners associated with PM 2.5 during the whole sampling period in Dalian. For PAH congeners, BbF was the dominant congener (16.02%), followed by FLA (14.08%), CHR (12.37%) and BaA (11.9%) for all samples during the entire sampling period, and ACE had the lowest proportion (0.2%). In addition, it was found the most abundant components were BbF (16%), FLA (14%), CHR (13%), PYR (12%) in heating period, and BbF (20%), FLA (14%), PHE (13%), and PYR (12%) were the most abundant compounds in non-heating period. in non-heating period as a result of the photochemical pollutant that occurred frequently in summer, which is consistent with previous studies (Guo et al., 2019). SO 2 is mainly discharged during coal combustion, while NO 2 was formed during the combustion of petroleum, as a result, the criteria pollutants could play as indicators of air pollution.

Statistical correlation between criteria air pollutants, EPFRs and PAHs
The above results showed that PM 2.5 -bound EPFRs and PAHs had different levels and characteristics in non-heating and heating periods, showing the temporal trends of different kinds of pollutants. To better understand the possible sources of EPFRs during different period in actual atmospheric environment in Dalian, this paper performed the statistical correlations between the concentrations of PM 2.5 -bound EPFRs, PAHs and criteria pollutants, i.e. PM 2.5 , SO 2 , O 3 , NO 2 , PM 10 , and CO by Person's correlation.
The statistical correlation results between the concentrations of criteria pollutants and EPFRs in PM 2.5 were shown in Fig. 5 and Table S4. The concentrations of EPFRs had a signi cant positive correlation with those of SO 2 (r = 0.602, p < 0.01, n = 46), NO 2 (r = 0.453, p < 0.01, n = 46) and CO (r = 0.405, p < 0.01, n = 46) for all samples, re ecting that coal and petroleum combustion were the predominant sources of PM 2.5 -bound EPFRs in Dalian. The correlation between EFPRs and SO 2 is higher than those with NO 2 indicating that coal combustion had a more signi cant in uent on the EPFRs. The metal oxide and the organic molecules formed from the fuel combustion process, would generate abundant EPFRs in post ame zone and cool zone. And then the EPFRs were discharged into environment with particulate matters (Vejerano et al., 2018). Previous studies also showed that under irradiation of light, PAHs can turn into EPFRs in the presence of O 3 (Borrowman et al., 2016). Nonetheless, there was a negative association between O 3 and EPFR during the whole sampling period (r = -0.558, p < 0.01, n = 46). As shown in Fig. 1 and Fig. 4, the concentrations of O 3 were higher in non-heating period than those in heating period, whereas the concentrations of EPFRs were higher in heating period than those in non-heating period, causing the negative association between the concentrations of O 3 and EPFRs during the whole sampling period. Other study found there was few EPR signals of EPFRs in PM 2.5 with a high concentration of O 3 , either (Xu et al., 2020). With a high level of O 3 , a high oxidative pollutant, the EPFRs may be oxidized quickly, causing the negative correlation between the EPFRs and O 3 , which indicates the O 3 can both accelerate the formation and decay the EPFRs (Xu et al., 2020). Unpredictably, the concentrations of EPFRs has no signi cant correlations with those of PM 2.5 , indicating that ne particulate matters were only the carrier of EPFRs and have no signi cant effect on the formation of EPFRs, which is consistent with previous results Schendorf et al., 2019).
Many reports showed that the PAHs were important precursors of EPFRs Zhu et al., 2019). Thus, this statistical correlation was also performed between the concentrations of 15 kinds of PAH congeners and EPFRs in PM 2.5 to better illustrate the potential sources of PM 2.5 -bound EPFRs in Dalian. The result was presented in Table S1. The concentrations of PM 2.5 -bound EPFRs had a signi cant positive correlation (r = 0.560, p < 0.01, n = 46) with the concentrations of ∑ 15 PAHs, which implied that PAHs and EPFRs might have similar sources. As for PAH congeners, a signi cant correlation was found between EPFR concentrations and those of 12 kinds of PAH congeners (0.422 < r < 0.606, p < 0.01, n = 46) except for ACE, FLU and ANT in the whole sampling period. These PAH congeners are the maker for coal and biomass combustion and vehicle emission , indicating the combustion process of fuel and biomass source may contribute to EPFRs in PM 2.5 .

Source apportionment
To investigate the potential sources and formation process of EPFRs in PM 2.5 , the PMF model was used to statistically analyze EPFRs, PAH congeners and six conventional pollutants in samples. The factors obtained by PMF model would re ect the different sources and formation processes of EPFRs. Five main contributing factors to EPFRs in PM 2.5 were shown in Fig. 6 and Fig. 7. The typical characteristic of factor 1, dominated by PM 10 and PM 2.5 , was identi ed as the dust source. There was no contribution to EPFRs in PM 2.5 , which is consistent with previous study (Qian et al., 2020). Factor 2 indicates coal combustion source, which has high proportions of BghiP, InD and DiB, associated with coal combustion emission (Qin et al., 2013;Wang et al., 2019b). Coal combustion source contributed about 58.7% to EPFRs. The central heating in winter in Dalian consume lots of coal, leading the high level of EPFRs in PM 2.5 during heating period. Factor 3 indicates secondary source, which is dominated by O 3 (Lengyel et al., 2004;Wang et al., 2019b). This factor contributed about 0.6% to EPFRs, indicating secondary sources barely contributed to the EPFRs in PM 2.5 . Wang et al. (2019a) also reported that secondary reaction only contributed 3.42% to the PM 2.5 -bound EPFRs in Xi'an, a northwest city of China, which was in line with our result. Factor 4 indicates vehicle emission due to the presence of high levels of NO 2 , FLA, PYR, BaA, CHR, BbF and BkF , which contributed 11.7% to PM 2.5 -bound EPFRs. Factor 5 indicates biomass combustion due to the presence of high level of ACY, ACE and CO , which contributed 28.7% to EPFRs in PM 2.5 . Overall, the sources of EPFRs in PM 2.5 in Dalian were mainly coal and biomass combustion, followed by the vehicle emission, which mean that the primary source made a dominate contribution to PM 2.5 -bound EPFRs in Dalian. Even though PAHs can be oxidized by O 3 and form EPFRs under laboratory conditions, the EPFRs derived from PAHs had a shorter lifetime in aerobic environment than in anoxic environment (Borrowman et al., 2016;Jia et al., 2019a). Our study also showed that the extraction solution of PM 2.5 could generate an extensive signal under the irradiation of light and then decayed quickly after removing the light source, as shown in Fig. S3. which indicated the photo-formation of EPFRs in PM 2.5 has a short lifetime. Thus, the secondary sources to PM 2.5 -bound EPFRs, which are involved in the photoreaction of O 3, are negligible. Further studies are necessary to illustrate the formation mechanism of EPFRs in the actual atmosphere.

Conclusion And Environmental Implications
The concentration levels, species characteristics of EPFRs and PAHs were investigated in PM 2.5 in Dalian during heating and non-heating period. The concentration of PAHs and EPFRs in PM 2.5 samples in Dalian ranged from 1.88 to 76.24 ng/m 3 and 1.22×10 13 to 8.97×10 15 spins/m 3 , respectively. The g factors of PM 2.5 bounded EPFRs in Dalian ranged from 2.00222-2.00393, which are the major representative of organic carbon-centered free radicals and carbon-centered radicals with adjacent oxygen atoms. The source appointment result indicated that the primary sources, i.e. fuel and biomass combustion, vehicle emission, are the dominate sources of PM 2.5 -bound EPFRs in Dalian. As for actual atmospheric environment, the portion of secondary EFPRs associated O 3 is negligible. Over all, this study provides a better insight the sources of PM 2.5 -bound EPFRs in Dalian. The study warrants further attention on the formation mechanism of PM 2.5 -bound EPFRs in atmospheric environment.

Declarations
Ethics approval and consent to participate  Figure 1 Temporal variation of concentrations of environmentally persistent free radicals (EPFR) bounded to PM2.5 measured in Dalian, China, during heating and non-heating period.