Misha I. Schurman This email address is being protected from spambots. You need JavaScript enabled to view it.1,2, Alexandra Boris1, Yury Desyaterik1, Jeffrey L. Collett, Jr.1 1 Department of Atmospheric Science, Colorado State University, Fort Collins, CO 80521, USA
2 Colorado Mountain College, Breckenridge, CO 80424, USA
Received:
January 15, 2017
Revised:
June 5, 2017
Accepted:
June 5, 2017
Download Citation:
||https://doi.org/10.4209/aaqr.2017.01.0029
Schurman, M.I., Boris, A., Desyaterik, Y. and Collett, Jr., J.L. (2018). Aqueous Secondary Organic Aerosol Formation in Ambient Cloud Water Photo-Oxidations. Aerosol Air Qual. Res. 18: 15-25. https://doi.org/10.4209/aaqr.2017.01.0029
Cite this article:
The current understanding of aqueous secondary organic aerosol (aqSOA) formation is based largely on laboratory investigations of very simple surrogate cloud water solutions that aid mechanistic understanding of aqueous oxidation but may not accurately reflect the influence of the complex ambient matrix present in authentic cloud waters on organic chemistry. In this study, unaltered ambient cloud water and ‘biogenically influenced’ ambient cloud water (with added pinonic acid) were photo-oxidized, atomized, and dried to simulate the formation of aqSOA in clouds, then analyzed using an Aerodyne Aerosol Mass Spectrometer. Two major chemical regimes were identified: in the first, particle organic mass is gained, then lost; sustained increases in highly oxidized fragments indicate overall organic acid formation, while increases in nominally volatile fragments suggest that evaporation may contribute to the observed mass decrease. In the second regime, the oxidation level of cloud water organic matter decreases as mass decreases, suggesting that oxidized functional groups are fragmented and lost to evaporation. Overall, the rate of aqSOA production in unaltered cloud water decreases as oxygenation increases, until organic mass loss beginning at consistent values of f44 > 0.23 ± 0.05 and O:C > 0.61 ± 0.05. We hypothesize that there may be a parameterizable ‘maximum oxidation level’ for cloud water above which functional group fragmentation is dominant. These experiments are among the first to quantify organic mass production in ambient cloud water and employ the most atmospherically relevant oxidant concentrations to date.HIGHLIGHTS
ABSTRACT
Keywords:
Secondary organic aerosol; Cloud chemistry; Photooxidation; Aqueous.
Particles in the atmosphere have wide-ranging health, environmental, and climate effects, many of which are attributed disproportionately to fine-mode (< 2.5 µm diameter) secondary organic aerosols (SOA). While gas-phase chemistry and subsequent partitioning was traditionally thought to be the main source of SOA, this mechanism fails to predict the amount, content (Chang and Thompson, 2010), distribution (Tsigaridis and Kanakidou, 2003; Heald et al., 2005), and size of ambient particles (Blando and Turpin, 2000), where ‘ambient particles’ are defined as particles found in ambient air; recent work suggests that up to ~50% of total SOA mass is formed through aqueous reactions (Ervens et al., 2011). However, most of our knowledge of SOA formation is based on laboratory investigations of very simple solutions that aid mechanistic understanding of aqueous oxidation but may not accurately reflect the influence of the complex ambient matrix (defined as the mixture of compounds comprising a cloud water sample) on organic chemistry (Boris et al., 2014). This paper seeks to investigate the relevance of previous aqueous SOA (aqSOA) studies by photo-oxidizing ambient cloud water (defined as cloud water collected in-situ from natural clouds) with and without addition of common aqSOA precursors; while the complexity of the cloud water matrix prevents precise mechanistic determination, general reaction regimes may be inferred in terms of changes in organic mass, composition, and level of oxidation. Such an approach can be very useful, especially given our incomplete understanding of the organic composition of ambient aerosols and cloud water. Important aqSOA precursors include anthropogenic and biogenic volatile organic compounds (A- and B-VOCs, respectively) oxidized to form products such as glyoxal, methylglyoxal, pinonaldehyde, and pinic/pinonic acids (Atkinson and Arey, 2003; Fu et al., 2008). Pinonic acid is used herein to represent oxidation products of α-pinene which may be absorbed from cloud interstitial spaces (Jang, 1999; Kanakidou et al., 2005), as a proxy for total interstitial BVOC absorption. We hypothesize that this addition of organic material, and its relatively low oxidation level relative to the original cloudwater, may augment and/or sustain aqSOA formation as found in the previous works (see discussion in Results Section). The hydroxyl radical (OH) is a dominant aqueous oxidant (Herrmann et al., 2010). OH can be absorbed from the gas phase, but is also formed significantly (up to ~33%; Ervens et al., 2003b; Arakaki and Faust, 1998) by aqueous photolysis of hydrogen peroxide (H2O2; Graedel and Goldberg, 1983; Zellner et al., 1990), iron(III) hydroxide (Fe(OH)2+; Weschler et al., 1986; Arakaki and Faust, 1998), nitrous acid (HONO; Fischer and Warneck, 1996; Arakaki and Faust, 1998), and nitric acid (HNO3; Graedel and Weschler, 1981); H2O2 and HOx (HO+ hydroperoxyl radical, HO2) may also undergo regenerative chemistry involving organic radicals (Anastasio et al., 1994; Valverde-Canossa et al., 2005). Direct organic photolytic oligomerization and lysing reactions are also possible and may form OH in turn through HOx cycling, though organic radical-initiated reactions are generally not competitive with OH oxidation (Guzman et al., 2006; Tan et al., 2010; Tan et al., 2012). Fenton reactions may produce OH in the absence of ultraviolet light (UV; Arakaki and Faust, 1998; Deguillaume et al., 2005), which motivated our no-UV control experiments. Reaction with OH and photolysis are important aqueous pinonic acid sinks (Lignell et al., 2013). The current understanding of aqSOA mechanisms, products, and yields is based largely on laboratory investigations of simple solutions with unrealistically high oxidant and reactant concentrations. Solute concentration strongly influences aqSOA yield and product type, with higher concentrations forming higher molecular weight compounds such as oligomers through organic self-reaction, while lower fog- and cloud-relevant concentrations often form organic acids (Tan et al., 2009; Lim et al., 2010). Solute composition is also important, as organic radical formation, hygroscopicity, acidity, and component distribution in the droplet (e.g., organic shells) may all affect aqSOA yields (Volkamer et al., 2009; Ervens and Volkamer, 2010; Galloway et al., 2011). Simple solutions of common SOA precursors, while more mechanistically constrained, may not accurately represent the complex chemistry in real atmospheric waters. This work explores how photooxidation influences ambient cloud water organic mass, average oxidation state, and reaction regimes (such as functionalization versus fragmentation), and investigates relationships between oxidative progression and solution characteristics such as total organic carbon (TOC) and oxidation level (oxygen-to-carbon elemental ratios (O:C), f44, etc.). The experiments presented here compare bulk photo-oxidations of real cloud water and cloud water with added pinonic acid (simulating ‘biogenic precursor’ addition), all featuring among the most atmospherically relevant reactant and oxidant concentrations to date. The Aerodyne High Resolution Time-of-Flight Aerosol Mass Spectrometer (AMS) measures the size and composition of submicron non-refractory particles with high time- and mass-resolution and has been described elsewhere (Decarlo et al., 2006); vaporization and electron impact ionization therein cause molecular fragmentation, complicating identification of specific reaction products as discussed later. AMS data quality is constrained by rigorous calibration; a LI-COR LI-820 monitor before the inlet constrained CO2(g) contributions to the CO2+ fragment. The ToF-AMS DAQ v 1.7.2B software was used for data acquisition; data analysis utilized the SQUIRREL (v1.52K) and PIKA (v1.11K, Sueper et al., 2011) tools in Igor Pro 6.22A (WaveMetrics Inc., Lake Oswego, OR). Measurements near the AMS detection limit were not altered or eliminated. Particulars can be found in the supplement. Briefly, bulk solutions are oxidized in a beaker-like vessel using UV light (Green Killing Machine, Model AAUV09W-UVC, 254 nm peak wavelength) and hydrogen peroxide, atomized (TSI 3076), dried (mimicking particle formation from cloud droplet evaporation, Loeffler et al., 2006; De Haan et al., 2009; El Haddad et al., 2009; Lee et al., 2011a), and analyzed via AMS with 1-minute time resolution after Lee et al. (2011b), with the addition of (a) a continuously circulating water sheath around the reaction vessel for temperature control at ~22 C (AquaBox fish-tank pump, Model AA2005), (b) a water trap after the atomizer (intermittent visible water at the atomizer outlet, see supplement), and (c) a 210Po neutralizer between the driers and the AMS (Fig. 1; Boris et al., 2014). Photo-oxidations lasted between ~30–140 minutes; control experiments ran for ~20 minutes (see supplement). Photo-reactor and stir bar were cleaned with deionized water and air-dried between each experiment. Five ambient cloud samples from Mt. Tai, China were used (named T1-4 and T7); T7 was oxidized with (T7PA) and without (T7) added pinonic acid (30 µM). See supplement section S3 and Shen (2011) for sampling, storage, and composition details. Generally, pinonic acid (PA) concentrations in cloud water are unknown; due to similarity between yields of pinonic acid (~0.1–1%), pinic acid (~1–5%), and pinonaldehyde (~0–1%) from α- and β-pinene aqueous oxidation (Yu et al., 1999), we conjecture that ambient pinonic acid concentrations might be similar to those of pinic acid and pinonaldehyde (0.11 and 0.8 µM, respectively; van Pinxteren et al., 2005). Species concentrations are normalized by total sulfate to correct noise from fluctuating atomizer output (supplement Sections S2 and S4 for detailed methodological assessment). Hydroxyl radicals were produced via photolysis of hydrogen peroxide (300 µM). The resulting [OH]aq was quantified in the cloud samples (and deionized water for comparison purposes) via absorption spectroscopy of 202 nm light by salicylic acid, which was below detection limit in the cloud samples, preceded by reverse phase liquid chromatography separation to prevent measurement interference from other absorbing species (N = 3; accounting for salicylic acid direct photolysis and using the second-order rate constant of 1.6 × 1010 M–1s–1; see Boris et al., 2014); these measurements yielded 1.9 × 10−14 M OH within the cloud water and 5.4 ± 0.5 × 10−14 M OH within pure water (Boris et al., 2014). Before hydrogen peroxide addition, these ambient cloud samples contained [HOOH]aq = 1–100 µM (Shen, 2011); assuming that the rate constant for OH production by hydrogen peroxide photolysis used herein holds in the complex ambient matrix, and that OH-organic reaction rate increases are linear within the given [OH] concentration range, this initial hydrogen peroxide content produces concentrations of [OH]aq ≈ 10–15–10–14 M which, adding the supplementary hydrogen peroxide, yields [OH]aq ≈ 10–14 M. Ambient cloud [OH] is difficult to measure, but has been estimated at ~10–14 to 10–13 M (Jacob, 1986; Arakaki and Faust, 1998; Ervens et al., 2003b; Deguillaume et al., 2005; Ervens and Volkamer, 2010), 1–2 orders of magnitude lower than used in similar laboratory experiments (Lim et al., 2010; Lee et al., 2011a). These solutions are among the most atmospherically relevant used in the laboratory to date; methodological assessment of evaporative cycling effects and OH exposure is discussed further in supplement Section S4. Control experiments evaluated contributions of dark (noUV/noH2O2, ‘baseline’), photolytic (+UV/noH2O2), and hydrogen peroxide-initiated (noUV/+H2O2) reactions to aqSOA. For ambient cloud samples from Mt. Tai, China (T1-T7), neither organic mass (Total Organics/Total Sulfate, normalized to t = 0) nor oxidation indicator f44 (mass at m/z 44/Total Organics, normalized to t = 0) change significantly under any of the control conditions, indicating probable dominance of OH-initiated reactions in the full photo-oxidations (Fig. S1). Organic mass (Org/SO4), O:C, and f44 (m/z 44/Org) for ambient cloud water photo-oxidation experiments are shown in Fig. 2; sample volume limited the duration of T1. Fig. 3 shows timelines of CO2+/SO4 (Takegawa et al., 2007), C2H3O+/SO4 (at m/z 43, indicating carbonyl content; McLafferty and Turecek, 1993), and C2H3O+/C4H7+ (ratio of most to least oxidized fragment at m/z 43), as well as CH2O2+/SO4, which is, nominally, formic acid (a major oxidation product of glyoxal compounds; see supplement S4 for caveats to formic acid quantification via AMS). Although determination of specific oxidative mechanisms is precluded by both fragmentation in the AMS and the complexity of the ambient cloud water composition, possible functionalization pathways are examined using delta analysis and movement in van Krevelen space. Experiments may be grouped according to their oxidative behavior into two groups. In Functionalization-Dominant Regime 1 (T1 and T3) samples first gain, then lose organic mass, and gain, then sustain CO2+ fragments and f44 representing carboxylic acids (and other oxygenated fragments, as seen in difference spectra Fig. S4) indicating retention of oxygenated products. In contrast, in Fragmentation-Dominant Regime 2 (T4 and T7), organic mass is unchanging for the first 30–60 minutes and decreasing significantly thereafter; there is no increase in organic mass as seen in Regime 1. Additionally, there is early formation, then pronounced loss of oxygenated fragments; together, these suggest mass loss due to functional-group fragmentation. Evidence supporting these regime groupings is detailed in the following sections. For the T1 and T3 experiments with ambient Mt. Tai cloud water, Org/SO4 increases for the first 15–30 minutes (to ~110% and ~140% of the t = 0 value, respectively), then decreases for the remainder of the experiment. Oxidation proxy f44 and carboxylic acid indicator CO2+/SO4 increase for the first 15–30 minutes (T1 and T3); the T3 experiment ran longer and these proxies were observed to stay constant thereafter through 90 minutes. Reduction in C2H3O+/SO4 beginning at ~10 min may result from oxidation of carbonyls to carboxylic acids; contemporaneous rapid oxidation of CxHy species is indicated by an increase in C2H3O+/C4H7+ (from ~10–25 minutes in T1 and ~6–16 minutes in T3), indicating that C4H7+ must be decreasing faster than C2H3O+. After ~10–15 minutes, both C2H3O+/SO4 and C2H3O+/C4H7+ decrease, indicating a net loss of hydrocarbons and carbonyls. Most chemical change discernable via AMS occurs in the first half-hour or so of oxidation. A picture emerges of rapid oxidation of hydrocarbons to carbonyls and carbonyls to carboxylic acids, consistent with other similar studies (Lee et al., 2011a). This is supported by difference spectra for the major organic fragment ‘families’ (CH, CHO, CHOn>1; Fig. S4), in which CH fragments decrease and CHO and CHOn>1 generally increase, especially at m/z 44 (CO2+); slight decreases in some CHO fragments commensurate with increases in CHOn>1 fragments at the same m/z reflect oxidation of carbonyls to carboxylic acids. These patterns can also be seen in delta mass spectral analysis (see supplement S6 and discussion below). The ‘formic acid’ fragment increases in both T1 and T3, suggesting that volatile formic acid and/or molecules fragmenting to CH2O2+ are early-generation products. Formic acid production occurs in aqueous oxidations of glyoxal and is responsible for the eventual loss of total organic mass in those experiments due to its volatility and resultant loss during drying (Lee et al., 2011b); the same mechanism could also be at play in T1 and T3, accompanied by decomposition of oxidized molecules to hydrocarbons and other non-formic-acid volatile fragments that are lost during evaporation as evidenced above. Fragmentation may be further explored through the slope of movement in van Krevelen space (O:C vs. H:C, see Fig. 4) as an isolated particle/droplet population reacts (e.g., Heald et al., 2010). T1 and T3, though with different compositions and therefore different starting positions in the diagram, move similarly toward enhanced O:C with slopes near –0.3, which is between m = –0.5, indicative of either carboxylic addition with C-C bond breakage or addition of both acid and alcohol/peroxide functional groups without fragmentation (total 1 H loss per 2 O addition, Fig. 4), and m = 0, indicative of alcohol or peroxide addition. Delta analysis can reveal which of these mechanisms are important (see supplement S6), as functional group composition dictates fragment m/z values such that delta (∆) = m/z – 14n + 1, where n is the number of methylene (CH2) bridges on the functional group (Table S3, McLafferty and Turecek, 1993). For T1 and T3, reductions in Δ-2, Δ0, and Δ6, which may be partially attributed to alcohols (Table S3), suggests at least that alcohol addition does not exceed consumption of unsaturated hydrocarbons and carbonyls (at Δ0, Δ-2), and implies that carboxylic addition with C-C fragmentation is a more important reaction pathway for this dataset; this fragmentation may also contribute to the organic mass loss observed if the non-functionalized fragments are volatile. These slopes are similar to gas-phase chamber oxidations of both generated and ambient OA (Chhabra et al., 2011; Lambe et al., 2011; Ng et al., 2011). In contrast to experiments T1 and T3, there is relatively little change in composition during oxidation of T2. Org/SO4 is unchanging for the first ~20 minutes, with subsequent decrease (Fig. 2). f44 increases slightly, but given that CO2+/SO4 and C2H3O+/SO4 are constant (statistically, with a slight CO2+/SO4 decrease toward the end), this is more a function of total organic mass loss than proportional gains at m/z 44; C2H3O+/C4H7+ increases in the first ~20 minutes, indicating decrease in hydrocarbon C4H7+, which is echoed in general (though not universal) CxHy reduction in the difference mass spectrum (Fig. S4). However, clear progression toward higher O:C in the van Krevelen diagram (Fig. 4) implies that oxygenated compounds are gained; family difference spectra reveal increases in fragments > m/z 50 from the oxidized CHOn>1 family, though fragment abundance changes are small compared to the other experiments. The progression to highly oxidized (CHO n>1) fragments > m/z 50 and decreased CO2+ suggests higher product stability against fragmentation within the AMS. By all measures, T2 started at a higher level of oxidation than the other experiments, which may reduce its ability to participate in further oxidation reactions, if, as discussed later, there exists a ‘maximum’ level of average solution oxidation above which subsequent organic mass formation is hindered. In Mt. Tai cloud water experiments T4 and T7, organic mass is unchanging for the first 30–60 minutes and decreasing significantly thereafter; there is no increase in organic mass. Fig. 2 shows that the normalized amount of Org/SO4 does not increase as in Regime 1, but only decreases after the beginning of photooxidation at t = 0. There is some ‘noise’ in this data, which is discussed thoroughly in Supplement Section S2: Description of variable atomizer output; nonetheless, the normalized Org/SO4 value begins to fall below 1 immediately after t = 0 and ‘noise’ does not exceed 1.09 in either T4 and T7, nor does it show the clear increasing, then decreasing trends seen in T1 and T3. This is in contrast to maximum normalized Org/SO4 of 1.43 and 1.12 in T3 and T1. O:C and f44 generally increase for the first 15–30 minutes and then decrease (Fig. 2). T4 and T7 show early formation, then pronounced loss (beginning at ~35–40 min) of CO2+/SO4; the first half-hour looks similar to Regime 1 (T1 and T3), with fragment ratios indicating progressive carbonyl, then carboxylic acid increases (as above), but is followed by strong decreases in C2H3O+ (to ~50% of t = 0 value in T4) and CO2+ (max ~200% at t = 20 min in T4; back to starting value by experiment’s end). This is in clear contrast to Regime 1, in which oxygenated products were sustained throughout the experiments. The formic acid fragment is near zero in T4 and with some mass in T7, but shows no significant trend in either time series. The fragment-family difference spectra show universal hydrocarbon (CH) loss and decreased C2H3O+ (m/z 43), with small increase or decrease in experiment-end CO2+ in T4 and T7, respectively, consistent with the fragment timelines. In the van Krevelen diagram, T4 and T7 initially progress toward higher O:C and lower H:C as would be expected of the carbonyl and carboxylic acid formation inferred from the fragment ratios above, but then move back toward lower O:C and higher H:C as photo-oxidation progresses; along with the loss of organic mass and oxygenated fragments, this suggests that oxidized functional groups are chemically fragmented in solution and lost during atomization/drying, leaving behind the hydrocarbon ‘backbones’ of the molecules. In contrast to T7 and the other cloud samples, aqueous photo-oxidation of T7PA (T7 with added pinonic acid) has fairly constant Org/SO4, f44, and CO2+/SO4 (Figs. 2 and 3). C2H3O+/C4H7+ and moderate C2H3O+/SO4 increases imply hydrocarbon loss and carbonyl gains that are corroborated in the difference spectra (Figs. 3 and S4). In the van Krevelen diagram, T7PA resides in a region similar to T7 (Fig. 4); T7PA’s lower starting O:C implies that unaltered T7 is on average even more functionalized than pinonic acid, which contains both a carboxylic acid and a carbonyl group. While T7PA develops slightly higher O:C and lower H:C with time (Fig. 4), these changes lack the amplitude and organization seen in T7 and the other samples in van Krevelen space. The family difference spectra reveal addition of CHO and CHOn>1 fragments (including acids and carbonyls); delta analysis also indicates carbonyl formation in increasing Δ2 (alkyls, saturated carbonyls). T7PA differs from the unaltered sample (T7) in a number of ways. Addition of pinonic acid reduces the initial level of solution oxidation and pH, enhances formation of carbonyls and larger, more stable molecules (increased CHO and CHOn>1 > m/z 50 fragments), and does not contain ‘formic acid’ signal; organic mass is maintained instead of decreased. This mass retention may be influenced by increasing total organic carbon (regardless of organic composition), as added PA constitutes a 57% increases in TOC for T7PA; increasing precursor concentration often results in higher mass production as hypothesized for this experiment and discussed in the introduction. Chamber pinonic acid-OH oxidations have resulted in highly functionalized, low-volatility molecules (Müller et al., 2012) similar to unpublished pinonic acid aqueous photo-oxidations in our laboratory (identified via ESI-ToF-MS). Our observation of dominant carbonyl formation compares favorably to single-precursor pinonic acid photo-oxidations in our photoreactor (Schurman, 2014) and other aqueous cis-pinonic acid (Aljawhary et al., 2016) and α-pinene oxidations (Bleier and Elrod, 2013); however, α-pinene products appear to vary with OH exposures, with dominant carboxylic acid formation and little change in f43 under high-OH conditions (George and Abbatt, 2010), and carbonyl formation in the form of acetone, formaldehyde, formic acid, etc. under lower [OH] (Nozière et al., 1999); Lignell et al. (2013) show that direct aqueous PA photolysis is a minor mechanism. Since our precursor, pinonic acid, is a fairly early-generation α-pinene product and our equivalent OH exposure is approximately 1-to-1 with ambient exposure, our samples experience only a few hours of ageing compared to the days-to-weeks exposures in many chamber experiments that observe oxidation to carboxylic acids (which may be less relevant to cloud lifetimes, as well). These results suggest that biogenic precursor addition may suppress or delay the oxidative mass loss of organics observed in aqueous photo-oxidations of unaltered cloud samples, and that total organic mass formation or loss may be related to the average level of solution oxidation. To investigate this, we evaluated the relationship between O:C, f44, and the incremental change in Org/SO4, finding that mass decreases begin only when O:C > 0.61 ± 0.05, f44 > 0.23 ± 0.05 and, generally, f43 < ~0.08 ± 0.02 (T1 organic mass decreases occur with f43 steady at ~0.084, but this holds true for all other ambient-cloud experiments). Because multiple fragments of different degrees of oxidation contribute to both m/z 43 and 44, the f44 and f43 ‘break-over points’ (at which organic mass begins to decrease) may not be as useful (or universal) as that of O:C. Two patterns emerge in these ambient cloud water photo-oxidations: in the first (Functionalization-Dominant Regime 1; T1 and T3), organic mass is gained, then lost, and sustained increases in highly oxidized fragments indicate overall organic acid formation; decrease in hydrocarbon fragments and production of the potentially-volatile CH2O2+ fragment (‘formic acid,’ which partitions mostly to the gas phase; see discussion below and Liu et al., 2012), suggests that volatilization of these species may contribute to mass decrease. In Fragmentation-Dominant Regime 2 experiments (T4 and T7), oxidized fragments decrease as mass decreases, indicating that oxidized functional groups are fragmented and lost to evaporation. pH varies significantly between functionalization-dominant Regime 1 (pH ~5.5–6.4) and fragmentation-dominant Regime 2 (pH ~4.6–4.7; Fig. 5). These pH variations may drive formic acid production (or at least particle partitioning, as explained below), with substantial ‘formic acid’ fragments occurring only in samples with pH > ~5.5 (T1 and T3); T3 had the highest pH, and also the greatest increase in nominally formic acid fragments, while T1 had a lower pH and lower formic acid concentrations. In Regime 2, for which solution pH values were ≤ 4.74, little or no formic acid was detected. This is logical, since the Henry’s Law partitioning of formic acid is driven toward the particle phase by higher pH (Liu et al., 2012). Furthermore, particulate formic acid is enhanced significantly beyond the Henry’s Law equilibrium in recent studies of urban-influenced particles (including one in Beijing, China; Wang et al., 2007; Liu et al., 2012); possible explanations include varying pH between particle bulk and surface (such that bulk measurements do not capture the pH relevant to the reaction) and variability in LWC (due to varying degrees of drying over the course of our experiments; Liu et al., 2012). Also, cations such as Ca2+ can undergo acid-base reactions that lead to formic acid retention in the particle phase (Fornaro and Gutz, 2003; Löflund et al., 2002; Malm et al., 2009); this chemistry has been indirectly evidenced in one case, coincidently involving long-range transport of formic acid and dust from China (Kawamura et al., 2012). Initial solution O:C (ranging from 0.35–0.53) does not vary significantly between Regime 1 and Regime 2, but Regime 1 reactions generally end at higher O:C (0.69–0.95, versus ~0.64 for Type 2) due to retention of oxygenated functional groups. On average, Regime 1 samples had higher starting concentrations of TOC (Table 1), hydrogen peroxide, and other solutes, which may explain the tendency toward higher functionalization and therefore mass retention as described in the introduction; however, these concentration differences were not significant using the Mann-Whitney-Wilcoxon rank-sum test, and samples with equivalent TOC concentrations were found between the types (Fig. 5). Meteorological back-trajectories indicate the same source region (the heavily populated, industrial South China Plain) for air masses involved in all cloud samples (Shen, 2011). Because it is difficult to both collect and to fully and continuously speciate cloud water, many studies use simple surrogate cloud water solutions to better constrain yields and reaction pathways, impeding direct comparison of our findings to the literature. However, using methods quite similar to those herein, though with higher OH concentrations, Lee et al. (2012) noted that ambient cloud samples formed highly oxidized SOA similar to that found in ambient aerosols, with organic mass production maxima of ~110–130% of the initial value followed by decreasing mass, similar to this work (Lee et al., 2012). A model simulation of rural SOA(aq+g) with monoterpene emissions suggested ~21–25% SOA mass increase in the first ~1 hour when clouds (pH = 4.5) were added, with a slight reduction in mass after ~2 hours much like our experiments (Chen et al., 2007a), while urban aerosol with anthropogenic emissions (NOx, aromatics, alkanes) increased about 10% with no subsequent decrease; as in this work, model addition of less oxidized precursors appeared to augment and/or sustain aqSOA increases, supporting the hypothesis that pinonic acid addition would do the same. This hypothesis is also supported by a methacrolein cloud chamber photo-oxidation experiment involving a continuous interstitial source of less-oxidized precursors gases, which found a non-linear 47-fold increases in SOA mass sustained over a 22-hour experiment (Monod et al., 2000, 2007; El Haddad et al., 2009; Liu et al., 2009). The methodology used herein does differ from ambient cloud processing in some ways. First, bulk-phase photooxidations disregard the continuous evaporation and re-condensation of cloud droplets, which influences evaporative losses and the cycling of interstitial VOCs, semi-volatile compounds, oxidants, and oxygen (O2) through the gas and droplet phases; this may in turn change aqSOA production (De Haan et al., 2009). With typical cloud droplet lifetimes of ~10 minutes (Roelofs and Kamphuis, 2009), a real cloud could undergo 12 droplet cycles over the course of the experiments herein, implying a more continuous source of fresh interstitial VOCs than was recreated in our bulk-phase experiments. The bulk phase also limits gas exchange by eliminating droplet surface area at which exchange takes place: this lack of droplet surface area and attendant gas exchange may suppress both precursor absorption and volatile product evaporation (although drying before AMS analysis may offset the latter effect). The effect of evaporative cycling on aqSOA is a focus of current research. In all, the bulk-aqueous/AMS analytical method is expected to lose a large and variable fraction of volatile oxidation products, the amount of which is highly dependent on the oxidation precursors and the volatility of their dominant products and is assumed to mimic the behavior of evaporating cloud droplets in the real atmosphere. Loss mechanisms may include a) evaporative loss during diffusion drying (Lee et al., 2011a), and b) evaporative loss during drying as the particle beam enters the high vacuum in the AMS (Zelenyuk et al., 2006). Mechanism a) might be more atmospherically relevant than b), but the current dataset cannot determine the contributions of each of these mechanisms to organic evaporative loss. Also, while the continuously-mixed bulk solution methodology might prevent the phase separation observed in atmospheric aerosol water (e.g., organic shells, which may change water evaporation and/or solute interaction and therefore reactivity), this effect is not thought to be important in cloud-relevant concentrations, where relative dilution leads to homogeneous mixing (Marcolli and Peter, 2005; Zuend et al., 2008; Ziemann, 2010; Shiraiwa et al., 2013). Under the assumption that the drying method used herein resemble those in the real atmosphere, the AMS does capture evaporated-cloud-relevant particulate organics. This work explores how photooxidation influences ambient cloud water organic mass, oxidation state, and reaction regimes (such as functionalization versus fragmentation), and investigates relationships between oxidative progression and solution characteristics such as oxidation level (O:C) and initial total organic carbon (TOC). Unaltered ambient cloud water and ‘biogenically influenced’ ambient cloud water (via addition of pinonic acid) were photo-oxidized using UVC light and H2O2 (to form OH), atomized, and dried to simulate the formation of aqSOA from clouds; these experiments are among the first to quantify organic mass production in ambient cloud water and employ the most atmospherically relevant oxidant concentrations to date. All ambient samples (except the heavily oxidized T2) gained carbonyl, then organic acid fragments; two of the five samples increased in sulfate-normalized organic mass (to a maximum of 110–140% of the t = 0 value). All samples eventually lost organic mass (reaching 60–80% of their initial values), which is attributed to formation of volatile products in some cases (‘Functionalization-Dominant Regime 1’ evidenced by increasing ‘formic acid’ fragment and sustained increases in CO2+) and molecular decomposition leading to loss of oxidized functional groups in others (‘Fragmentation-Dominant Regime 2’; CO2+, C2H3O+, O:C, and total organic mass decrease simultaneously); Regime 1 had significantly higher pH than Regime 2, and generally (though not significantly) higher initial total organic carbon, consistent with literature results showing formation of more functionalized products in higher-concentration solutions. The rate of aqSOA production in unaltered cloud water decreases as oxygenation increases, until organic mass loss beginning at a consistent value of O:C > 0.61 ± 0.05 and f44 > 0.23 ± 0.03 (f43 < 0.08 ± 0.02). This implies that there may be a parameterizable ‘maximum oxidation level’ for cloud water above which organic molecule fragmentation is dominant; however, we were unable to further mathematically describe the relationship between aqSOA production/loss rate and oxidation level. The consistency of this ‘maximum oxidation level’ within the Mt. Tai cloud water should prompt further investigation of cloud samples with varying composition and concentration. If widely applicable, such a parameterization would contribute greatly to the understanding and modeling of the secondary organic aerosols that contribute to so many important health, environmental, and climate effects. Lastly, addition of TOC in the form of pinonic acid to an ambient sample (meant to simulate interstitial biogenic precursor absorption) appeared to prevent organic mass loss and foster carbonyl formation, rendering the un-altered experiments herein lower-bound estimations of aqSOA formation in photooxidized clouds. Collection of Mt. Tai cloud samples was funded by NSF Grant ATM-0711102. Thanks to Taehyoung Lee and Xinhua Shen for collecting the Mt. Tai cloud samples used in this study, and Tao Wang, Wenxing Wang, and Xinfeng Wang for their help and partnership in planning the Mt. Tai cloud sampling campaign.INTRODUCTION
METHODS
Fig. 1. Schematic of photoreactor experimental setup. Circles with black triangles indicate pumps.
RESULTS
Control Experiments
Ambient Cloud Water Photo-Oxidation
Fig. 2. Timelines of f44 (red), normalized (to t = 0) Org/SO4 (green), and O:C (blue) for T1, T2, T3, T4-1, and T7 aqueous photo-oxidations. Uncertainties (not plotted): one standard deviation equals 28% of the value, from propagation of 20% AMS quantification error through the indicated calculations. Experiment time approximates ambient oxidation time as discussed in Section 2.3. Fig. 3. Timelines of (left axes) fragment ratios normalized to the t = 0 value and (right axes) the mass concentration of the ‘formic acid’ fragment for T1-4, T7, and T7PA. Uncertainties (not plotted): one standard deviation equals 28% of the value, from propagation of 20% AMS quantification error through the indicated calculations.
Functionalization-Dominant Regime 1: Oxygenated Products MaintainedFig. 4. van Krevelen diagram showing (left) T1, 3, and 4 full photo-oxidations and (right) T2, 7, and 7PA, colored by time during experiment. Dashed red lines: Estimated oxidation state, OSc = 2 × O:C – H:C (Kroll et al., 2011). Red and blue lines indicate the region usually inhabited by ambient data in the f43 vs. f44 plot, as transposed to van Krevelen space using the empirical parameterization of f43 vs. H:C and f44 vs. O:C for ambient and chamber OA (Ng et al., 2010); grey lines represent 10% error. Linear fits: T1: m = –0.28, r2 = 0.89; T2: m = 0.25, r2 = 0.32; T3: m = –0.28, r2 = 0.64; T4: m = –0.51 (see text), r2 = 0.46; T7: m = –0.32 (see text), r2 = 0.14; T7PA: m = –0.45, r2 = 0.26.
Fragmentation-Dominant Regime 2: Oxygenated Fragments Lost
Ambient Cloud Water Photo-Oxidations with Added Pinonic Acid Precursor
DISCUSSION
Fig. 5. Initial concentrations of the given cloud water constituents for given sample regimes. Units are, from left to right: Cl-Ca, HCHO, and S(IV) (µN); TC-TOC (ppm); Fe and Mn (µg L–1). Error bars are standard deviations where data from multiple samples of the same regime were averaged. T2 consists of only one sample, so error bars are omitted.
CONCLUSIONS
ACKNOWLEDGMENTS
REFERENCES
Pinene oxidation products. J. Phys. Chem. A 120: 1395–1407. [Publisher Site]
from Whiteface Mountain, New York: The role of the Fe(r) (r = II, III) photochemical cycle. J. Geophys. Res. 103: 3487–3504. [Publisher Site]
Aerosol Air Qual. Res. 18 :15 -25 . https://doi.org/10.4209/aaqr.2017.01.0029