αpOS-249 (in the form of its sodium salt) was synthesized by Yu and her co-workers (Wang et al., 2017) through the Upjohn dihydroxylation and sulfation of α-pinene. The purity (∼99 %) has been tested using liquid chromatography–electrospray ionization–high-resolution mass spectrometry and nuclear magnetic resonance (1H and 13C) spectrometry. Heterogeneous OH oxidation of αpOS-249 aerosols was carried out using a 13 L aluminum OFR at 50±2.0 % RH and 298.0±0.5 K. Experimental details together with a schematic diagram (Scheme S1 in the Supplement) are given in the Supplement. Briefly, αpOS-249 was first dissolved in deionized water (0.1 wt %) followed by a 30 min sonication. Aqueous aerosols were generated by passing the solution through an atomizer (TSI Model 3076) using 3 L min-1 of nitrogen (N2). The aerosol stream was then directly mixed with ozone (O3), dry and humidified nitrogen (N2), and oxygen (O2) to control the RH. A total flow of ∼5 L min-1 was fed into the reactor, corresponding to a residence time of ∼156 s (Xu et al., 2020a).

Gas-phase OH radicals were generated by photolyzing O3 with UV light at 254 nm in the presence of water vapor inside the OFR. The concentration of gas-phase OH radicals was varied by changing the O3 concentrations. The OH exposure, a product of gas-phase OH radical concentration and the residence time, was in the range of 0–17.4×1011 molecule cm-3 s. It was determined by measuring the decay of sulfur dioxide (SO2) in independent calibrating experiments (Teledyne SO2 analyzer, Model T100) based on the reaction rate between gas-phase OH radicals and SO2 (=9.0×10-13 molecule-1 cm3 s-1) at 298 K (Kang et al., 2007). It acknowledges that the presence of aerosols did not significantly affect the generation of gas-phase OH radicals and the determination of OH exposure (less than ∼10 %). The aerosol stream leaving the reactor passed through an annular Carulite catalyst denuder (manganese dioxide and copper oxide catalyst; Carus Corp.) and an activated charcoal denuder to remove residual O3 and other gas-phase species. Aerosols were collected onto the Teflon filters (47 mm, 2.0 µm pore size, Pall Corporation) through filtration at a sampling flow rate of 3 L min-1 using an air sampling pump (Gilian 500, Sensidyne) for 30 min, with a total gas sampling volume of ∼90 L. Duplicate filters were collected from each of oxidation experiments for subsequent chemical analysis. After collection, filters were immediately stored at -20 ∘C in the dark and analyzed within 3 months. Part of the remaining stream was introduced into a scanning mobility particle sizer (SMPS, TSI, CPC Model 3775, Classifier Model 3081) to measure the size distribution of the aerosols. The aerosol mass was determined from measured volume concentration assumed for spherical aerosols with a unit density. Before oxidation, the mean surface weighted diameter for aerosol distribution was about 181.3±0.5 nm with a geometric standard deviation of 1.3, and the aerosol mass loading was measured to be ∼2000 µg m-3.

Scheme S2 shows the overview of the chemical analysis. First, the filters were extracted twice with 5 mL methanol in an ultrasonic bath for 1 h. Extracts were then filtered through a 0.2 µm polytetrafluoroethylene (PTFE) syringe filter and combined. A total of 300 µL of the extract was blown to dryness under a gentle stream of N2 at room temperature and then reconstituted in 1 mL methanol–water (1:1 v/v) containing 200 ppb D17-octyl sulfate as an internal standard.

The amount of inorganic sulfate (SO42-) formed upon oxidation was quantified using the IC method. Operating conditions have been given by Huang et al. (2018). Briefly, the filters were extracted using 5 mL of double de-ionized water (18.2 MΩ cm-1) from the ultrapure water system (Nanopure Diamond UV/UF) and were sonicated for 60 min and then mechanically shaken for 60 min. After filtering, these extracts were subsequently analyzed by an ion chromatograph (Dionex ICS-1100). The separation of anions was accomplished using an AS11-HC analytical column (IonPac, 4×250 mm) and an AG11-HC guard column (IonPac, 4×50 mm) with 15 mmol L-1 NaOH eluent. It is known that bisulfate ion (HSO4-) is being converted into SO42- upon mixing with the alkaline eluent. The concentration of SO42- quantified by the IC method thus represents a total amount of HSO4- and SO42-. Our previous study has shown that the peak from the Na2SO4 standard has the same retention time in the IC chromatogram as that of the sodium bisulfate (NaHSO4) standard (Xu et al., 2020a). Furthermore, the responses of the NaHSO4 and Na2SO4 standards are about the same. These would justify the use of the Na2SO4 standard for the quantification of HSO4- and SO42-. In this work, the amount of HSO4- and/or SO42- produced upon oxidation at a given OH exposure was proportional to its peak area in the chromatogram and was determined using the Na2SO4 standard calibration curve. The extraction efficiency was determined to be 90.3±1.5 % by measuring the recovery of the Na2SO4 standard spiked onto blank filters. The uncertainty for the measurement of SO42- is discussed in the Supplement.

We also note that no distinct peak was detected in the ion chromatogram for αpOS-249 standard while a small peak corresponding to SO42- was observed. The absence of an αpOS-249 peak could be attributed to IC detection for ionic and ionizable species likely being limited to small compounds (up to C5) (Surratt et al., 2008; Domingos et al., 2012). The presence of a SO42- peak may be due to the hydrolysis of αpOS-249. However, this peak comprised 2.1±0.8 % of the total mass, suggesting that the hydrolysis of αpOS-249 was not significant. This observation is consistent with the literature that αpOS-249 does not readily undergo hydrolysis (Hu et al., 2011).

The high recovery of αpOS-249 suggests the sample preparation and extraction methods are effective. Wang et al. (2017) have also reported that there was no degradation for αpOS-249 after 2 years of storage at low temperature (-20 ∘C). A recent study by Hughes et al. (2019) examined the stability of a range of OSs (e.g., methyl sulfate, hydroxyacetone sulfate, two α-pinene-derived OSs: m/z=279 (C10H15O7S-), and m/z=281 (C10H17O7S-)) on filters frozen at -20 ∘C over 1 year. The filters were extracted via a similar procedure to that applied in this study, and the extracts were analyzed by HPLC-ESI-HRMS. They found that the investigated OSs with different functional groups (e.g., alkyl, carboxylate, and hydroxyl groups) showed no degradation during the storage. Taken together, αpOS-249 and its oxidation products (i.e., OSs) which have similar carbon skeletons while possessing different functional groups (alcohol and/or ketone) are likely stable during the storage and pre-treatment processes for chemical analysis.

We note that organic compounds such as carbonyls and carboxylic acids could undergo reactions with methanol during extraction, storage, and possibly during the electrospray process (Bateman et al., 2008). For instance, Bateman et al. (2008) suggested carboxylic acids could react with methanol to form esters and with carbonyls to hemiacetals and acetals. We checked the presence and relative abundance of these potential products (to that of our precursor, αpOS-249) in our aerosol mass spectra. At the maximum OH exposure, only a few products that could be potentially formed from the reactions of αpOS-249 with methanol were detected, and they had negligible intensities. This would suggest that the influence of methanol is not significant on the identification of the major reaction products.

As shown in Fig. 3, αpOS-249 decays at a significant rate upon oxidation. Approximately 30 % of αpOS-249 remains unreacted at the highest OH exposure. Oxidation kinetics can be quantified by measuring the decay of αpOS-249 at different OH exposures. The normalized decay can be fit with an exponential function (Smith et al., 2009): 1ln⁡II0=-kOH⋅t, where I is the concentration of αpOS-249 quantified by HPLC/ESI-QTRAP-MS at a given OH exposure, I0 is the concentration before oxidation, k is the effective second-order heterogeneous OH rate constant, and [OH]⋅t is the OH exposure. The rate constant determined using Eq. (1) is found to be (6.72±0.55)×10-13 cm3 molecule-1 s-1. Assuming a 24 h averaged [OH] of 1.5×106 molecules cm-3 (Mao et al., 2009), the atmospheric lifetime of αpOS-249 against heterogeneous OH oxidation, τ=1/k[OH] is calculated to be 11.5±0.9 d. Considering atmospheric aerosols with a similar size (∼200 nm) having a typical lifetime of 10–14 d against wet or dry deposition (Kanakidou et al., 2005), heterogeneous OH oxidation could be a competitive sink for αpOS-249. Lam et al. (2019) and Chen et al. (2020) recently reported that isoprene-derived OSs (i.e., 2-methyltetrol sulfate and 3-methyltetrol sulfate) can undergo heterogeneous OH oxidation efficiently. Altogether, these results may suggest that transformation kinetics and pathways would need to be considered in chemical transport models in order to better predict the abundance and composition of atmospheric OSs.

In this study, the heterogenous OH reactivity of aqueous αpOS-249 aerosols at a single RH (50 %) was investigated. In the atmosphere, the complex interplay between aerosol phase state (e.g., solid or aqueous), morphology, and the types and concentrations of salts could significantly alter the heterogenous OH reaction kinetics and mechanisms under different environmental conditions (e.g., RH and temperature).

The aerosol physical state can play a key role in determining the heterogeneous kinetics and chemistry of pure organic aerosols (Koop et al., 2011; Shiraiwa et al., 2011, 2013; Chan et al., 2014). For instance, Chan et al. (2014) reported that aqueous succinic acid aerosols reacted about 40 times faster than in solid aerosols towards heterogeneous OH oxidation. These could be explained by the more rapid diffusion of succinic acid to the surface of aqueous droplets for oxidation than solid aerosols. Moreover, for aqueous droplets, aerosol water content can vary considerably, depending on atmospheric conditions. The change in aerosol-phase water and solute concentrations would influence the reactivity by varying aerosol viscosity (Slade and Knopf, 2014; Chim et al., 2017; Marshall et al., 2016, 2018). For instance, oxidation kinetics in highly concentrated aqueous organic aerosols are found to be much slower than those in diluted ones. This is because aerosol viscosity generally increases with the solute concentration, thereby slowing down the diffusion of organic molecules within the aerosol and lowering the overall reactivity.

Atmospheric aerosols are comprised of organic compounds, inorganic salts, and many other species. To date, large uncertainty remains in how inorganic salts alter the heterogeneous kinetics and chemistry (McNeill et al., 2007, 2008; Dennis-Smither et al., 2012). A few laboratory studies have revealed that the presence of dissolved inorganic ions (e.g., ammonium sulfate) can reduce the heterogeneous OH reactivity of organic compounds but does not significantly alter the reaction mechanisms (Mungall et al., 2017; Kwong et al., 2018; Lam et al., 2019). More recently, Xu et al. (2020b) reported that the change in the heterogeneous reactivity strongly depends on the concentration of organic compounds and inorganic salts. They found that the rate of the reactions decreases when the organic-to-inorganic mass ratio (OIR) decreases. This could be explained by the colliding probability between OH radical and organic species at the aerosol surface becoming lower in the presence of salt, resulting in a smaller overall reaction rate. We also acknowledge that different inorganic ions could have different propensity for air–aerosol interface depending on their polarizability and interactions with other components (Jungwirth and Tobias, 2002; Gopalakrishnan et al., 2005). Jungwirth and Tobias (2002) investigated the preference of sodium cation and chlorine anion to the interface or bulk in sea salt aerosols based on polarizable MD simulation. They found that chlorine anion has a stronger propensity for the interface than sodium cation and is proportional to its polarizability. Therefore, the types and concentration of inorganic ions (e.g., ionic strength) within aerosol could potentially alter the overall heterogeneous kinetics and reaction pathways. To date, it remains an open question whether the salts alter heterogeneous reactivity chemically, physically, or both.

We would like to note that additional uncertainties in heterogeneous reactivity of organic compounds in the presence of inorganic salts could also arise when these organic–inorganic droplets undergo phase separation, depending on environmental conditions and aerosol composition (e.g., different types of inorganic salts, the average oxygen-to-carbon (O:C) elemental ratio of organic compounds, and OIR) (You et al., 2014; Qiu and Molinero, 2015; Freedman, 2017, 2020). These phase-separated droplets typically exhibit two distinct liquid phases: an inorganic-rich inner phase and an organic-rich outer phase. Different morphologies have also been observed (e.g., core–shell morphology, partially engulfed morphology, and transitions between different types of morphology). Lam et al. (2021) recently reported that phase-separated organic–inorganic droplets have a slightly higher reactivity towards gas-phase OH radicals, compared to single-phase ones. As phase separation occurred, an uneven distribution of organic species within the droplets increased the collision probability between organic molecules and OH radicals at or near the droplet surface. Overall, further studies emphasizing the effects of aerosol composition, phase transition and separation, and morphologies on heterogeneous reactivity of OSs are needed to better understand their transformation rates and chemistry.

Scheme 1 shows the proposed formation pathways for the detected products summarized in Table 2. Prior to oxidation, αpOS-249 tends to dissociate readily. Oxidation is initiated via hydrogen abstraction by an OH radical, forming an alkyl radical (R⚫), which reacts with an oxygen (O2) molecule quickly to form a peroxy radical (RO2⚫), which can react via different pathways. For instance, the reactions of two RO2⚫ can generate two carbonyl products via the Bennett and Summers mechanism (Reaction R1) (Bennett and Summers, 1974) or an alcohol product and a carbonyl product via Russell reactions (Reaction R2) (Russell, 1957), or two alkoxy radicals (RO⚫). RO⚫ can react with an O2 molecule (Reaction R3), undergo intermolecular hydrogen abstraction (Reaction R4), and/or decompose involving the cleavage of a C-C bond or a C-O bond (George and Abbatt, 2010; Carrasquillo et al., 2015; Kroll et al., 2015).

R1RO2⚫+RO2⚫→2R=OR2RO2⚫+RO2⚫→R=O+ROHRO2⚫+RO2⚫→2RO⚫R3RO⚫+O2→R=O+HO2R4RO⚫+R′H→ROH+R′⚫R5RO⚫→decomposition

Based on detected products (Table 2 and Fig. 2), OH oxidation tends to increase the functionalities and oxygen content of αpOS-249. Moreover, these products (C10OSs) can be classified as functionalization products and are formed via the addition of one or two oxygenated functional groups to αpOS-249. For instance, ions corresponding to first-generation products produced from OH oxidation of αpOS-249 are detected at m/z=263 (C10H15O6S-) and 265 (C10H17O6S-). When the oxidation proceeds, these products can react with OH radicals to form second-generation products, which give m/z=277 (C10H13O7S-), 279 (C10H15O7S-), and 281 (C10H17O7S-) products.

Figure 4 shows the evolution of the intensity of the reaction products as a function of OH exposure. Different kinetic profiles have been observed for first- and second-generation products. It can be seen that first-generation products (m/z=263 (C10H15O6S-) and 265 (C10H17O6S-)) reach a maximum intensity at an OH exposure of about 14.3×1011 molecules cm-3 s. Their intensities then decrease slightly at the highest OH exposure. The concentration of second-generation products (m/z=277 (C10H13O7S-), 279 (C10H15O7S-), and 281 (C10H17O7S-)) follows a different kinetic profile. At low OH exposures, the intensities of the products are observed to increase at a slightly slower rate relative to first-generation products. Their concentrations always increase with increasing OH exposure and reach their maximum values at the highest OH exposure. This trend is consistent with the multigenerational reactions that the formation of second-generation products is from the reactions of OH with first-generation products. In general, the products detected, shown in Table 2, are consistent with reactions (described in detail below) that form hydroxyl and carbonyl functional groups.

During oxidation, the carbon site where the hydrogen atom is abstracted by OH radical governs the following formation of products. More recently, molecular dynamics (MD) simulations showed that heterogeneous reaction is likely not initiated by the direct collision between a gas-phase OH radical and an organic molecule near the aerosol surface (Xu et al., 2020b; Lam et al., 2021). Instead, the reaction may occur after a number of collisions between the absorbed OH radical and the organic molecule. As a first approximation, the reactivity of αpOS-249 towards OH radical is estimated by a structure–activity relationship (SAR) model proposed for aqueous-phase OH reactions with organic compounds (Monod and Doussin, 2008). We acknowledge that the SAR model does not include the parameterization of the sulfate group (-OSO3-) for OSs. As the sulfate group (-OSO3-) exhibits a resonance electron-withdrawing effect (Bahl and Bahl, 2010) as the carboxylate anion (-COO-) and bears a negative charge, here the effect of the sulfate group on the OH reactivity is evaluated using the descriptor of the carboxylate anion (-COO-) in the SAR model. Overall, the goal of this simple analysis is to qualitatively assess which carbon site is more favorable for hydrogen abstraction in reactions of OH radicals with αpOS-249 in order to gain more insights into the reaction mechanisms.

Table S2 shows the reactivity of a hydrogen atom at different reaction sites upon OH oxidation of αpOS-249. The model predicts that the hydrogen abstraction likely occurred at two secondary carbons (4-C and 7-C) and a tertiary carbon (3-C), in which the abstraction rates were significantly larger than other sites. For instance, the largest hydrogen abstraction rate is predicted to occur at a secondary carbon site (4-C). This could be explained by the electron donor effect of the carboxylate anion (-COO-) group in β position when the OH attacks the reactive site as an electrophilic reaction (Monod and Doussin, 2008). A comparable rate is predicted for a tertiary carbon (3-C) as the adjacent α-CH2 and β-CH3 groups exhibit significant electron donor properties. A large abstraction rate for 7-C could also be due to strong electron-donating effect of surrounding α-CH groups. The model suggests a large difference in the OH reactivity for the three tertiary carbon sites (1-C, 3-C, and 5-C). For instance, the slowest abstraction rate is predicted for 5-C. This could be attributed to a combined effect of the resonance electron-withdrawing effect of the α-carboxylate anion (-COO-) group and the field electron-withdrawing effect of the β-OH group. The slow abstraction rate predicted for 1-C could be due to the field electron withdrawing effect of the β-OH group. In general, the OH reactivity for the primary carbons (8-C, 9-C, and 11-C) and hydroxyl group (10-O) were predicted to be slower than that of the secondary and tertiary carbons except 5-C. We would like to acknowledge that uncertainties could arise from the fact that the sulfate group (-OSO3-) is currently not being considered in the SAR model and differences between the concentrated aqueous aerosols and the SAR model that is formulated for dilute aqueous solutions. Other factors such as steric hindrance and long-distance electronic effects of the sulfate group (-OSO3-) on the reactivity have not been considered (Yamazaki et al., 2019) and warrant further study.

Depending on the initial reaction site, a number of first-generation products with various structural isomers can be formed as indicated by multiple peaks observed in the EICs of detected ions (Fig. 2). Different isomers of second-generation products can also be formed from the reactions of OH with first-generation products. The possible formation pathways of detected products are further discussed below. Given a number of reaction pathways could possibly lead to the formation of the products with different isomers, only a general reaction scheme is depicted in Scheme 1 for simplicity and clarity.

On the basis of reaction products, we attempt to examine whether the heterogeneous OH oxidation of αpOS-249 freshly formed from the photooxidation of α-pinene could explain the formation of OSs detected in laboratory-generated α-pinene SOA and ambient aerosols. As shown in Table 2, five out of six product ions (m/z=247, 263, 265, 279, and 281) have been detected in ambient aerosols (Iinuma et al., 2005; Surratt et al., 2008; Ma et al., 2014; Hettiyadura et al., 2019; Wang et al., 2021), while four out of six product ions (m/z=247, 265, 279, and 281) have been observed for SOA formed from the photooxidation of α-pinene in the presence of acidic sulfate aerosols in laboratory studies (Liggio et al., 2006; Surratt et al., 2008; Ma et al., 2014; Zhang et al., 2015; Hettiyadura et al., 2019). Reaction mechanisms have been proposed for the formation of some of these products such as m/z=265 (C10H17O6S-) and 279 (C10H15O7S-) in the literature. For instance, in chamber studies, Liggio et al. (2006) suggested that C10H17O6S- (m/z=265) can be formed through the reactive uptake of pinonaldehyde, a major semi-volatile product generated from α-pinene oxidation, onto acidic sulfate aerosols via sulfate esterification reactions. Similar to C10H17O6S- (m/z=265), C10H15O7S- (m/z=279) can be produced through the reactive uptake of hydroxypinoinc acid on acidic sulfate aerosols via sulfate esterification reactions (Surratt et al., 2008). For field studies, ions corresponding to m/z=247 (C10H15O5S-) and 263 (C10H15O6S-) have been observed in ambient aerosols, but their sources and formation mechanisms have remained unclear. Here, we show that these OSs previously observed in ambient aerosols and α-pinene-derived SOA in laboratory studies could be first- and/or second-generation products formed upon heterogeneous OH oxidation of αpOS-249.

Recent studies have reported that heterogeneous OH oxidation of small OSs (C1, C2, and C5) can produce inorganic sulfates (Kwong et al., 2018; Lam et al., 2019; Chen et al., 2020; Xu et al., 2020a). It has been proposed that the decomposition of an alkoxy radical formed at the α-position of the sulfate group can generate a sulfate radical anion (SO4⚫-), which subsequently reacts further to generate inorganic sulfur species (HSO4- and SO42-) (Schemes 1 and S4). We here examine the significance of this conversion from organosulfur to inorganic sulfur upon heterogeneous OH oxidation of αpOS-249 (a C10OS). Figure S3 shows the IC chromatograms of αpOS-249 before and after heterogenous OH oxidation. Before oxidation (Fig. S3a), a small quantity of SO42- (contributing only 1.9±1.3 % of total sulfur mass) was detected in the ion chromatogram due to the hydrolysis of αpOS-249, which has been corrected for the determination of sulfate yield. After oxidation (Fig. S3b), an increase in the SO42-signal was observed, suggesting that some sulfur is being converted from its organic form (i.e., αpOS-249) into its inorganic form (HSO4- and/or SO42-) upon OH oxidation. The amount of inorganic sulfur formed from oxidation was then quantified to calculate the yield, defined as the total number of moles of HSO4- and SO42- formed per mole of αpOS-249 reacted at a given OH exposure (Xu et al., 2020a): yield=ΔSO42-ΔαpOS-249×100%, where the SO42- concentration was quantified by IC, and the αpOS-249 concentration was determined by HPLC/ESI-QTRAP-MS before and after oxidation. As shown in Fig. 5, the yields are determined to be small as indicated by the small SO42- peak detected after oxidation and range from 3.5±11.3 % at OH exposure of 4.8×1011 molecules cm-3 s to 13.8±4.7 % at 14.3×1011 molecules cm-3 s. The relatively large uncertainty at low OH exposures is mainly attributed to the small change in αpOS-249 and SO42- concentrations at low OH exposures (the determination of uncertainties of the yield is given in the Supplement). The small yields could be explained by the fact that the hydrogen abstraction at 5-C is likely not favorable compared to other carbon sites upon OH oxidation of αpOS-249 at the first place as predicted by the SAR model (Table S2). The generation of sulfate radical anions (SO4⚫-) via the breakage of the C-O bond in the 5-C alkoxy radical and the further reaction of SO4⚫- to generate inorganic sulfates might not be favorable (Schemes 1 and S4). Furthermore, we cannot rule out the possibility that the formation of SO42- could be from the hydrolysis of reaction products, as some secondary and tertiary OS could readily undergo hydrolysis and may form upon oxidation.

It is also worthwhile to note that upon oxidation, the decomposition of alkoxy radicals (RO⚫) formed at 5-C would lead to SO4⚫- and a fragmentation product. For instance, a C10 aldehyde product (C10H16O2, m/z=167) can be formed (Scheme 1). The effective saturation vapor pressure, C∗, of this product is estimated to be 2.27×105 µg m-3 based on its saturation vapor pressure (3.29 Pa) predicted from EVAPORATION (Compernolle et al., 2011). The volatility of this product is found to be about 5 orders of magnitude larger than that of αpOS-249 (C∗=1.67 µg m-3), which is estimated using a saturation vapor pressure (7.96×10-6 Pa) predicted by COSMOtherm (Hyttinen et al., 2020). Given its high volatility, this fragmentation product likely partitions back to the gas phase and has not been detected in our chemical analysis.