Reactive species formed upon interaction of water with fine particulate matter 1 from remote forest and polluted urban air 2

Abstract. Interaction of water with fine particulate matter leads to the formation of reactive species (RS) that may influence the aging, properties, and health effects of atmospheric aerosols. In this study, we explore the RS yields of fine PM from remote forest (Hyytiälä, Finland) and polluted urban air (Mainz, Germany and Beijing, China) and relate these yields to different chemical constituents and reaction mechanisms. Ultrahigh-resolution mass spectrometry was used to characterize organic aerosol composition, electron paramagnetic resonance (EPR) spectroscopy with a spin-trapping technique was used to determine the concentrations •OH, O2•−, and carbon- or oxygen-centered organic radicals, and a fluorometric assay was used to quantify H2O2 concentration. The mass-specific yields of radicals were lower for sampling sites with higher concentration of ambient PM2.5 (particles with a diameter 


abundance of HOMs or aromatic compounds is defined to be the sum chromatographic area of HOMs or 237 aromatics divided by the sum chromatographic area of all assigned organic compounds, with < 30% of 238 totally detected organic compounds not assigned (Wang et al., 2018a). 239

Determination of water-soluble transition metal concentrations 240
Based on the same extraction method as the H2O2 analysis in section 2.6, the concentration of five selected 241 water-soluble transition metal species (Fe, Cu, Mn, Ni and V) in the supernatants of PM2.5 extracts was 242 quantified using an inductively coupled plasma mass spectrometer (ICP-MS, Agilent 7900). These five 243 transition metal species were chosen for analysis due to their prominent concentrations and higher oxidative 244 potential (Charrier and Anastasio, 2012). A calibration curve for the ICP-MS analysis was made by 245 measuring standard multi-element stock solutions (Custom Grade, Inorganic Ventures). An aliquot of the 246 supernatants was diluted and acidified using a mixture of nitric acid (5%) and hydrofluoric acid (1%), which 247 was finalized to be 5 mL before analysis. The measured transition metal concentrations were blank-248 corrected and shown in corresponding figures. The detection limit of the ICP-MS analysis in this study was 249 typically < 40 ng L -1 . The PM2.5 samples collected on 2 June, 7 June, 9 June, 12 June in 2017 in Hyytiälä, 250 on 22 August, 26 August, 28 August, 25 September, 25 October, 14 November in 2017 in Mainz, and all found that the total ion concentration of Fe, Cu, Mn, Ni, and V showed a rapid rise during the first 15 min 254 ( Figure S2a), but at a much slower rate afterwards ( Figure S2b). respectively. The spectrum of Hyytiälä PM2.5 is dominated by peaks attributable to C-centered radicals. In 263 contrast, the spectrum of Mainz PM2.5 comprises strong peaks attributable to  OH and C-centered radicals, 264 with  OH exhibiting stronger signals. Finally, the spectrum of Beijing winter PM2.5 is mainly composed of 265 four peaks attributable to  OH. 266 Figure 2b shows the averaged relative fractions (RF) of  OH, O2 -, C-and O-centered organic radicals 267 generated by multiple PM samples from each site. In line with visual inspection of the spectra in Figure 2a, 268 the PM2.5 from clean forest site generates relatively more C-and O-centered organic radicals but less  OH, 269 vice versa for the radical yield by PM2.5 from polluted areas. Specifically, the mean RF of C-and O-centered 270 organic radicals, ordered from highest to lowest are: Hyytiälä (66% and 11%) > Mainz (46% and 10%) > 271 Beijing (39% and 5%). Note that, the significantly higher RF of C-centered radicals than O-centered organic 272 radicals may be induced by the higher yield  and stability of BMPO-C-centered radical adduct in the liquid

Mass-specific and air sample volume-specific yields of RS from ambient PM2.5
285 Figure 3 shows the mass-specific and air sample volume-specific yields of reactive species (RS) including 286 radicals, H2O2, and the sum of radicals and H2O2 by PM2.5 from Hyytiälä, Mainz, and Beijing. The mass-287 specific yields of RS are shown in the unit of pmol µg -1 of PM2.5, reflecting the redox activities of PM2.5 288 irrespective of filter loadings. The air sample volume-specific yields of RS are shown in the unit of pmol 289 m -3 of air, indicating that the redox activities of PM2.5 scale with atmospheric concentration of PM2.5. We 290 note that, while the more polluted sampling sites led to higher mass loadings, the concentrations of PM in 291 extracts were found to have a tiny impact on the radical yields ( Figure S1c and S1d). 292 Figure 3a shows that the mass-specific radical yields are negatively correlated with PM2.5 mass 293 concentrations. The mean concentrations of PM2.5 are lower to higher in the order of 5 (Hyytiälä) < 16 294 (Mainz) < 202 µg m -3 (Beijing), whereas the radical yields are in a reverse order of 0.58 > 0.33 > 0.07 pmol 295 µg -1 . The higher mass-specific radical yield of PM2.5 from Hyytiälä may be associated with the higher 296 abundance of particulate organic matter, which includes quinones and organic hydroperoxides that undergo 297 thermal, photonic, or hydrolytical dissociation as well as redox chemistry such as Fenton-like reactions to 298 produce radicals (Badali et al., 2015;Tong et al., 2016a;Tong et al., 2019). More than 70% of PM2.5 in 299 Hyytiälä forest is composed of organic matter (Jimenez et al., 2009;Maenhaut et al., 2011), whereas the 300 abundances of organic matter in Mainz autumn and Beijing winter PM2.5 are ~40% (Jimenez et al., also shows that the mass-specific H2O2 yields of PM2.5 from Hyytiälä (~2.2 pmol µg -1 ), Mainz (~3.4 pmol 303 µg -1 ), and Beijing (~3.4 pmol µg -1 ) exhibit a weak positive correlation with PM2.5 mass concentrations, 304 agreeing with previous measurements of the H2O2 formation by fine PM from different districts of Los < 53%, whereas the relative fraction of C-centered radicals is in the reverse order of 66% > 46% > 39%. 354 The consistently higher abundance of water-soluble transition metals and RF of  OH of urban PM2.

365
To investigate the influence of biogenic-anthropogenic organic matter interaction on the formation of 366 aqueous radicals, we measured the radical yield of SOA generated from oxidation of mixed naphthalene 367 and β-pinene precursors. Figure 5a shows that the mass-specific radical yields of SOA decrease with 368 increasing relative concentrations of naphthalene (i.e., [naphthalene]/([naphthalene]+[β-pinene])). As the 369 relative concentration of naphthalene is increased from 0 to 9, 23, and 38%, the radical yields of SOA 370 decrease in the order of ~8.4 > ~3.0 > ~2.3 > ~1.9 pmol µg -1 . This is because the naphthalene SOA has a 371 lower radical yield than β-pinene SOA with the same mass concentration in water extracts (Tong et al., 372 2016a;Tong et al., 2017;Tong et al., 2018;Tong et al., 2019). Moreover, the mass-specific radical yield of 373 β-pinene SOA in Figure 5a is the mean value of SOA from ~1 ppm and ~2.5 ppm of β-pinene (see Sect. 374 2.3). Therein the SOA from ~2.5 ppm β-pinene exhibits higher radical yield (11.5 pmol µg -1 ) than the SOA 375 generated from ~1 ppm β-pinene (4.5 pmol µg -1 ), which may be associated with the increasing partition of 376 oligomers into the particle phase with higher starting concentration of β-pinene (Kourtchev et al., 2016). containing HOMs. It is notable that PM2.5 from polluted Beijing contains substantial amount of aromatics 388 (Figure 4b), but mainly generates  OH upon interaction with water, which seems to contradict our finding 389 that naphthalene SOA generates  OH only to a small extent. This may be related to the more complex 390 composition of the ambient PM compared to laboratory-generated SOA. For example, conversion of O2 -391 to  OH, H2O2, and O2 by transition metals or other redox-active PM constituents through Haber-Weiss 392 reactions or other related redox chemistry (Kehrer, 2000;Tong et al., 2016a) is expected to occur in ambient 393 samples, but would not be observed in laboratory-generated SOA that does not contain significant fractions 394 of transition metals. 395

Radical yield of surrogate mixtures comprising transition metals, CHP, HA, FA and H2O2
396 Figure 6a shows the concentration of radicals formed in aqueous mixtures comprising 0-25 µM cumene 397 hydroperoxide (CHP), 43 µM Fe 2+ , 3 µM Cu 2+ , 4 µg mL -1 humic acid (HA) and 7 µM H2O2, with mixtures 398 containing 0, 5, and 25 µM CHP to be treated as surrogates of redox-active constituents in PM from Beijing, of  OH formed at higher concentration of CHP resembles the radical yield of ambient fine PM from cleaner 404 areas (Figure 2b), which contains a large fraction of HOMs (Tong et al., 2019). Moreover, Figure S5 shows To compare the Fenton-like reactions initiated by different transition metal ions related to ambient PM2.5, 410 we measured the absolute and relative racial yields of aqueous mixtures containing CHP and different 411 transition metal species, such as Fe 2+ , Cu 2+ , Mn 2+ , or Ni 2+ . We found that Fe 2+ is most efficient in initiating 412 Fenton-like reactions (Deguillaume et al., 2005) and the BMPO-radical adduct concentrations varied along 413 the reaction time ( Figure S6). Of note, the abundance, chemical composition, and physicochemical 414 properties of the redox-active constituents in ambient PM (e.g., transition metals and organic matter) can 415 be different from the surrogate mixtures, causing partially different radical yields between surrogate 416 mixtures and ambient PM2.5 (e.g., less comparable RF of  OH than the RF of C-centered radicals), which 417 warrants follow-up studies. To simplify the discussion, we only show the radical yields as mean values 418 within ~25 minutes of extraction and measurement. 419 To assess the influence of humic acid on Fenton-like reactions, we measured the radical yields of 420 mixtures comprising 100 µM CHP, 300 µM Fe 2+ , and 0-180 µg mL -1 HA. As the concentration of HA is 421 increased from 0 to 36 µg mL -1 , the concentration of total formed radicals decreased by ~52% from 15.5 to 422 7.4 µM (Figure 6c). This may be associated with the following properties of HA. First, HA exhibits 423 pronounced iron binding capacity of 32 nmol Fe per milligram of HA, preferentially toward Fe 3+ rather than 424     The error bars represent uncertainties of signal integration of EPR spectra (for y-axis) or experimental 914 uncertainties of the solution concentration (for x-axis). CHP: cumene hydroperoxide. HA: humic acid. FA: 915 fulvic acid.