Glyoxal secondary organic aerosol chemistry: effects of dilute nitrate and ammonium and support for organic radical–radical oligomer formation

Atmospheric waters (e.g. clouds, fogs and wet aerosols) serve as a medium in which water-soluble organic compounds react. These water-soluble organics are abundant in the atmosphere due to the gas-phase fragmentation and oxidation of primary emissions. Reactions in atmospheric waters can produce low volatility products that remain in the particle phase upon droplet evaporation (e.g. oxalate, glycolate, oligomers, epoxide sulfates and imidazoles).^[1–11] As a result, gas-phase followed by aqueous-phase chemistry contributes to secondary organic aerosol (SOA).^[12,13] SOA is also formed when semi- and low-volatility products of gas-phase chemistry sorb to particulate organic matter^[14,15] (here SOA_gas is used to denote SOA formed by this latter pathway and SOA_aq to denote SOA formed with a contribution from aqueous-phase chemistry).

Although the magnitude of SOA_aq remains uncertain, the inclusion of aqueous organic chemistry in global models has led to a substantial enhancement of aerosol burden,^[16–19] improved correlations between models and measurements in the North-eastern United States^[20] and helped to explain the highly oxidised nature of atmospheric organic aerosols.^[13,21] (SOA_gas is not as oxidised as measured atmospheric aerosol.) SOA_aq and SOA_gas form from different precursors under different atmospheric conditions and have different properties and behaviour. For example, because the precursors of SOA_aq are water-soluble gases with high O/C ratios, SOA_aq is expected to have high O/C ratios, be more hygroscopic and be better able to serve as cloud condensation nuclei than other organics.^[22] For these reasons, an improved understanding of SOA_aq is needed to link emissions to ambient aerosol concentrations, composition and effects through predictive models that enable the development of effective air quality management strategies. This ultimate goal motivates the work reported herein.

Glyoxal (GLY) is ubiquitous in the atmosphere and an important SOA_aq precursor because GLY is highly water soluble, reactive to OH^• radicals and forms oligomers at high concentrations (millimolar and above). GLY is the smallest dicarbonyl and forms through the gas-phase photooxidation of alkene and aromatic compounds.^[23] The largest global source of GLY is the gas-phase oxidation of isoprene (biogenic).^[24] A major anthropogenic source of GLY is the oxidation of acetylene.^[25] Although its marine sources are uncertain, satellite measurements reveal substantial concentrations of GLY in marine air.^[26] Acetylene, with an atmospheric lifetime near 18 days,^[16] is one possible GLY source in marine air.^[27] GLY (H_GLY: 3 × 10⁵ M atm^–1)^[28] readily partitions into atmospheric waters where it is present at concentrations from one to several hundred micromolar and can react further.^[29–31] GLY can react in atmospheric waters to form lower volatility products (e.g. oxalate, oligomers, organosulfate and imidazole).^{[6,8,32–34]} Because these compounds are found predominantly in the condensed phase in the atmosphere, aqueous reaction of GLY contributes to SOA_aq.^[32,33]

Organic nitrogen and sulfur compounds have been observed to form in highly concentrated aqueous solutions relevant to wet aerosols.^[6,34–37] Evidence for the photochemical formation of organosulfates from GLY and other aldehydes is provided through studies of bulk sulfate solutions and through smog chamber experiments with ammonium sulfate seed particles and UV irradiation.^[6,35,36] Aerosol mimic solutions containing GLY and inorganic ammonium salts have yielded products with carbon–nitrogen bonds in the absence of major atmospheric oxidants (i.e. O₃, OH^• and NO₃^•).^[37] In addition, imidazole formation has been detected in (dark) concentrated aldehyde–ammonium sulfate solutions and chamber experiments.^[6,34] Inorganic and organic concentrations are orders of magnitude smaller in clouds and fogs. Although experiments have been performed to examine the effect of dilute (cloud-relevant) concentrations of acidic sulfate on GLY chemistry,^[33] to our knowledge the degree to which GLY + OH^• chemistry is altered by the presence of cloud-relevant concentrations of nitrate or ammonium has not been examined. Inorganic nitrate and ammonium are found in atmospheric cloud and fog waters (e.g. ~1 µM–3 mM for NO₃^– and NH₄⁺ in fogs),^[38–41] and organic nitrogen species have been measured in California fogs^[42] and New Jersey rainwater.^[41] Although organic nitrogen is clearly formed through gas-phase chemistry and there is evidence for its formation in aerosols, it could plausibly also form through cloud and fog chemistry. In this work, we conducted chemical modelling and aqueous GLY + OH^• experiments with cloud and fog-relevant concentrations of inorganic nitrogen (i.e. ammonium sulfate or nitric acid) to determine whether the commonly understood dilute GLY + OH^• chemistry leading to oxalate (and therefore SOA) is altered by the presence of cloud-relevant concentrations of ammonium sulfate or nitric acid.

Tan et al.^[33] studied the aqueous OH^• radical oxidation of GLY at cloud-relevant and higher concentrations (30– 3000 µM). At cloud-relevant (dilute) concentrations, oxalate was the major product and the addition of cloud and fog-relevant concentrations of H₂SO₄ had little effect on the chemistry. At higher (3000 µM) concentrations Tan et al.^[33] observed the formation of products with higher carbon numbers than the precursor, GLY. These higher carbon number products only formed in the presence of OH^• radicals. By accounting for organic radical–radical reactions in the GLY oxidation mechanism, Lim et al.^[21] were able to reproduce the concentration dynamics of the major products in the experiments of Tan et al.^[33] Using GLY as a surrogate for the behaviour of dissolved organics at the high (1–10 M) concentrations found in wet aerosols, Lim et al.^[21] predicted that oligomers would be the main products of organic OH^• radical oxidation in wet aerosols. Tartarate is a key product in the organic radical–radical mechanism proposed by Lim et al.^[21] However, tartarate was not definitively measured in the Tan et al.^[33] experiments because tartarate and malonate co-eluted and thus could not be distinguished. In model runs that simulated the experimental conditions and included organic radical–radical chemistry,^[21] tartarate forms early in the experiment, peaking after ~25 min. Malonate formation is minor and slower, because it is formed through radical–radical reactions followed by acid catalysed dehydration. The experimental results reported herein provide definitive evidence for the formation of tartarate from GLY + OH^•.

This paper examines for the first time the effect of cloud and fog relevant concentrations of inorganic nitrogen on the aqueous oxidation of GLY. The finding that oxalate is not affected by cloud-relevant concentrations of ammonium sulfate and nitric acid increases confidence that the currently understood dilute GLY + OH^• chemistry can be used to predict GLY SOA formation in clouds and fogs. This paper also provides critical laboratory evidence for oligomer formation through organic radical–radical reactions; this chemistry is evident in 1 mM GLY experiments and is the dominant OH^• radical oxidation product at the much higher (1–10 M) organic concentrations present in wet aerosols.^[21]

Bulk aqueous phase reactions of GLY (1 mM) and the hydroxyl radical (~10^–12 M, modelled OH^•) with nitric acid (1.68 mM) or ammonium sulfate (840 μM) and control experiments were conducted in a 1-L glass reaction vessel, as described in detail elsewhere.^[1] OH^• radicals were continuously produced in situ by direct photolysis of hydrogen peroxide (5 mM) using a 254-nm mercury lamp (Heraeus Noblelight, Inc. Duluth, GA) placed in a quartz immersion well in the centre of the reaction vessel. Surrounding the reaction vessel was a water jacket that maintained a temperature of 26 ± 2 °C. The reaction vessel was wrapped in aluminium foil to eliminate exposure to ambient light. The following control experiments were conducted: GLY + OH^• without nitric acid or ammonium sulfate, GLY + HNO₃ + UV, GLY + (NH₄)₂SO₄ + UV, GLY + HNO₃ + H₂O₂, GLY + (NH₄)₂SO₄ + H₂O₂, GLY + HNO₃, GLY + (NH₄)₂SO₄ and H₂O₂ + UV.

Aqueous hydroxyl radical concentrations in clouds and fogs are predicted to be 10^–12 to 10^–13 M.^[43] Hydroxyl radical concentrations in these GLY experiments were as follows: [OH^•]_average = (1 ± 2) × 10^–12 M, [OH^•]_initial = 7.8 × 10^–13 M and [OH^•]_final = 6.0 × 10^–12 M. These concentrations were estimated using the GLY photooxidation model of Lim et al.^[21] and a H₂O₂ photolysis rate constant (k₁) equal to 1.0 × 10^–4 s^–1. The value of k₁, which depends on the photon flux of the lamp, was determined by fitting modelled to measured H₂O₂ concentrations in H₂O₂ + UV control experiments (Fig. S1). Hydrogen peroxide was quantified by the triiodide method and analysed with a UV-Vis spectrometer.^[44]

Discrete samples from aqueous experiments were analysed by ion chromatography (IC, Dionex ICS 3000; IonPac AS11-HC column; 30 °C, AG11-HC guard column; conductivity detector; 35 °C), IC coupled to electrospray ionisation mass spectrometry (ESI-MS) (HP–Agilent 1100; negative ion mode) and Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) (Thermo-Finnigan LTQ-XL: positive and negative ion modes, Woods Hole Oceanographic Institution Mass Spectrometry Facility) as described in detail previously and briefly below.^[41,45] Discrete samples for IC and IC/ESI-MS (~1.2 mL) were analysed immediately (same day), whereas discrete samples for FT-ICR-MS (~5 mL; 40-min time points only) were frozen at –20 °C for later analysis as recommended by Seitzinger et al.^[46] for preservation of stored atmospheric water samples for mass spectrometric analysis.

Carboxylic acids and inorganic anions were quantified by IC with five point calibration curves (accept R² > 0.98) developed from authentic standards, including: glycolate (5.8 min), formate (6.7 min), glyoxylate (9.5 min), succinate (20.2 min), tartarate (20.8 min) and oxalate (24.2 min). Note that glycolate co-elutes with acetate, succinate with malate, and tartarate with malonate. These peaks were quantified as glycolate, succinate and tartarate respectively. Nitrate (16.7 min) and sulfate (22.8 min) are also detected within the 40-min IC program.

Negative ionisation mode mass spectrometric analysis (50–1000 amu) of each peak in the sample chromatograms was performed by IC/ESI-MS as described by Tan et al.^[45] Briefly, the IC conductivity detector effluent (0.4 mL min^–1) was directed into the ESI-MS with 40 % of 0.05 % formic acid in water and 60 % methanol mobile phase (0.22 mL min^–1). After IC separation, mass spectrometric analysis was performed with a fragmentor voltage of 40 V and a capillary voltage of 3000 V (nitrogen drying gas; 10 L min^–1; 350 °C). Mass assignments were verified with a mixed standard analysed at the beginning and end of each sequence.

A discrete sample (40 min into the reaction) from each experiment, control experiment and a 1 mM GLY standard was also analysed by FT-ICR-MS in positive and negative ionisation modes (mass resolution 100 000) to determine elemental formulae of reaction products from 50 to 500 amu. A syringe pump delivered analyte into the spray chamber at 4 µL min^–1 with a capillary temperature of 260 °C and a capillary voltage of –17.50 V. The FT-ICR-MS was calibrated using a Thermo Scientific LTQ-FT external calibration mix. For both positive and negative ion modes, at least 200 scans were collected using the parameters described in Kido Soule et al.^[47] The transients were processed using SimStich^[48] and aligned with MATLAB code provided by Mantini et al.^[49] as described in Bhatia et al.^[50] Exact masses were processed with Midas Molecular Formula Calculator (v1.1) as described by Altieri et al.^[41] to provide elemental formulae of detected species.

Water blanks and dynamic blanks were respectively sampled from the water source (17 MΩ cm Barnstead E-Pure MilliQ water) and reaction vessel, and analysed in the same manner as samples, described above. Blanks had negligible (signal <0.5 µS) IC peaks, thus no subtraction was performed. IC/ESI-MS analysis of the blanks revealed that m/z –113 had a strong signal (abundance >5000) throughout all spectra and thus was subtracted from the spectra. The remaining peaks were small (abundance <2000) and did not interfere with products (abundance of major products >10 000; e.g. tartarate, oxalate, nitrate and sulfate). Mixed standards were analysed with samples to assess the accuracy of quantified organic acids and to check for shifts in retention times. The retention times varied less than 10 %. Greater than 20 % of samples were run in duplicate. Experiments were run in triplicate. Error bars for oxalate and tartarate concentrations (Figs 1, 2) are the coefficients of variation across experiments.

Fig. 1.  Concentration of oxalate (μM) with reaction time (min) for glyoxal (GLY) oxidation experiments: GLY + OH^• (solid circle), GLY + OH^• + HNO₃, and GLY + OH^• + (NH₄)₂SO₄ (asterisk). Error bars represent the coeffecient of variation (<10 %) across three experiments.

Fig. 2.  Tartarate + malonate by ion chromatography for glyoxal (GLY) oxidation experiments: GLY + OH^• (solid circle), GLY + OH^• + HNO₃ (open circle), and GLY + OH^• + (NH₄)₂SO₄ (asterisk). Note tartarate and malonate coelute. Error bars represent the coefficient of variation across experiments, 55 % for GLY + OH^•, 49 % for GLY + OH^• + HNO₃, and 38 % for GLY + OH^• + (NH₄)₂SO₄.

The effect of OH^• production by nitrate photolysis on the batch experiments was examined by adding Reactions 1–7 to the aqueous GLY + OH^• chemistry documented in Lim et al.^[21]:

where k_F is the rate of the forward reaction and k_R is the rate of the reverse reaction. Nitric acid dissociates to H⁺ and nitrate (Reaction 1). The photolysis of nitrate produces OH^• radicals (Reactions 2 and 3). Nitric acid and nitrate react with OH^• radicals to form nitrate radicals (Reactions 4 and 5). GLY reacts with nitrate radicals by H-atom abstraction to give a GLY^• radical and the rate constant is assumed to be 1 × 10⁴ M^–1 s^–1 based on Neta and Huie^[54] (Reaction 6). Finally, nitric acid can be regenerated by the reaction of OH^• radicals with nitrite radicals (Reaction 7). This chemistry was used to simulate batch experiments (Figs S2–S5).

The addition of ammonium sulfate (840 μM) had no effect on the production of oxalate (Fig. 1), the major product of aqueous GLY photooxidation. The addition of nitric acid (1.68 mM) decreased oxalate production slightly (~10 % at peak) when compared with oxalate production in the absence of nitric acid. Mean oxalate concentrations during GLY + OH^• experiments in the presence of nitric acid are significantly less than (95 % confidence) mean oxalate concentrations in GLY + OH^• experiments at every time point. This finding can be understood by examining model simulations. The addition of HNO₃ decreased oxalate formation (Fig. S3) because HNO₃ lowered the pH (Fig. S5) affecting glyoxylic acid dissociation and subsequently oxalate formation. Simulated OH^• radical concentrations were unchanged by the addition of HNO₃ (Fig. S2) because OH^• produced by nitrate photolysis reacts with nitrite radicals to regenerate nitric acid.^[52] The production of oxalate was much slower in GLY + HNO₃ + UV control experiments (Fig. S6) compared with OH^• radical experiments, also suggesting that OH^• radical production by HNO₃ photochemistry was modest (Fig. S7). Note that simulated and measured nitrate in the GLY + OH^• + HNO₃ system remained constant over the course of the reaction (Figs. S4, S8) and no substantial change in other quantified GLY oxidation products (i.e. tartarate plus malonate) was observed either with the addition of ammonium sulfate or nitric acid (Fig. 2). This suggests that the presence of ammonium and inorganic nitrate in clouds and fogs has little effect on the formation of cloud SOA_aq from GLY, and supports the use of dilute GLY + OH^• chemistry (e.g. Lim et al.^[21]) to predict GLY cloud SOA.

When beginning this work, we hypothesised that the addition of cloud-relevant concentrations of nitric acid might lead to the formation of organonitrates by radical–radical reactions or condensation reactions shown in Fig. 3. It is plausible that aqueous nitrate radical chemistry might be a source of oxygenated organic nitrogen compounds measured in fog and rain water,^[41,42] although these compounds could also have been formed through gas-phase chemistry. If the nitrate radicals were formed by photolysis in the aqueous phase and if these reactions occur at dilute concentrations, we should see organonitrate products in experimental samples. Note that uptake from the gas phase provides a source of night-time nitrate radicals in atmospheric waters; this source is not accounted for in our experimental samples. Below we look for evidence of the formation of these and other organic nitrogen products in experimental samples.

Fig. 3.  Hypothesised reactions and organic nitrogen products.

Organonitrates are known to form by gas-phase OH^• radical reactions at high NO_x during the daytime and gas-phase NO₃^• radical reactions at night.^[55,56] In the aqueous phase, organonitrates are not expected to form photochemically with NO or NO₂ due to low NO and NO₂ water solubilities (i.e. the Henry’s Law constants for the species are H_NO = 1.9 × 10^–3 M atm^–1 and H_NO2 = 7.0 × 10^–3 M atm^–1).^[52] However, organonitrate formation by NO₃^• radical reactions could occur because NO₃^• radicals are water soluble (H_NO3 = 2 M atm^–1)^[52] and reactive with respect to organics with rate constants of ~10³ to ~10⁴ M^–1 s^–1.^[54] In the night-time atmosphere, we expect that some NO₃^• radicals will dissolve in cloud, fog and aerosol liquid water. Photolysis of HNO₃ in the aqueous phase can also form NO₃^• radicals. In the presence of NO₂, NO₃^• radicals form N₂O_5, which is highly water soluble (H_N2O5 = 1.4 × 10¹⁰ M atm^–1).^[52] HNO₃ forms by hydrolysis when N₂O₅ dissolves in the aqueous phase. Thus, GLY + OH^• + HNO₃ is an appropriate system to examine to look for the formation of organonitrates in clouds and fogs.

Experiments conducted with nitric acid versus without nitric acid provide no evidence for organonitrate formation. Modelled (Fig. S4) and measured nitrate (Fig. S8) concentrations show no decrease with time across the experiments. IC/ESI-MS results provide no evidence of unique IC peaks or unique masses during GLY photooxidation in the presence of nitrate or ammonium (Fig. 4c, d), when compared with GLY photooxidation without nitrate or ammonium (Fig. 4b). Hypothesised organonitrate products shown in Fig. 3 were not detected by FT-ICR-MS analysis.

Fig. 4.  Ion chromatography–electrospray ionisation mass spectrometry data from the mixed standard (a) and the 40-min time point samples from glyoxal (GLY) oxidation experiments: GLY + OH^• (b), GLY + OH^• + HNO₃ (c), GLY + OH^• + (NH₄)₂SO₄ (d). In each case, the ion chromatogram is shown to the left. The mass spectrum of each lettered peak is shown to the right. The mixed standard contains glycolate (m/z –75; peak A), formate (m/z –45, this MS does not detect masses <50 amu; peak B), glyoxylate (m/z –73; peak C), succinate (m/z –117; peak D), tartarate (m/z– 149; peak E) and oxalate (m/z –89; peak F). Note, the following organic acid pairs coelute: glycolate + acetate (m/z –59), succinate + malate (m/z –133) and tartarate + malonate (m/z –103). Nitrate (m/z –62) and sulfate (m/z –97) are peaks N and S respectively.

IC/ESI-MS results are shown in Fig. 4. Fig. 4a is a mixed standard analysed by IC/ESI-MS. The lettered peaks in the IC chromatogram at the left of the figure correspond to the mass spectrum labelled with the same letter at the right. Fig. 4b–d show the IC/ESI-MS analysis of 40-min samples. The IC chromatogram of the 1 mM GLY + OH^• sample (Fig. 4b) is consistent with the IC chromatogram of similar (3 mM) GLY + OH^• experiments conducted by Tan et al.,^[33] showing a large peak with the retention time of oxalate (F) and small peaks with the retention time of malonate + tartarate (E) and with the retention time of mesoxalate (Y). In the 3-mM experiments of Tan et al.,^[33] an even smaller peak with the retention time of succinate + malate (D) was detectable; this peak was not detected in the current (1 mM) experiments. The mass spectra of Fig. 4b verify that peak F is in fact oxalate (m/z –89) and peak Y is mesoxalate (m/z –117) as proposed by Tan et al.^[33] Both malonate and tartarate are detectable in the 40-min GLY + OH^• sample (peak E, m/z –103 and –149 respectively). The IC/ESI-MS of GLY + OH^• + HNO₃ (Fig. 4c) and GLY + OH^• + (NH₄)₂SO₄ (Fig. 4d) were identical to that of GLY + OH^• with the exception of the nitrate peak (Fig. 4c; peak N; m/z –62) and the sulfate peak (Fig. 4d; peak S; m/z –97). Thus, the IC/ESI-MS provides no evidence for the formation of organic nitrogen species when GLY is oxidised in the presence of cloud-relevant concentrations of HNO₃ or (NH₄)₂SO₄. It is possible that the nitrate peak quantified by the IC includes organic nitrates that formed and elute at the same retention time. However, the IC/ESI mass spectrum for peak N has a single signal at m/z –62, where we expect nitrate ion detection.

Hypothesised organonitrate products shown in Fig. 3 were not detected by FT-ICR-MS analysis. Some nitrogen-containing elemental formulae were identified in samples from experiments conducted both in the presence and in the absence of nitrogen, suggesting the presence of trace nitrogen-containing contaminants. Ultra-high resolution FT-ICR MS has sub-parts per million mass accuracy and a resolution >100 000, enabling the determination of elemental formulae for thousands of compounds in a complex organic mixture. However, ion abundance depends both on compound concentration and on the matrix of other compounds present. Thus quantification is challenging and the ion abundance above a signal-to-noise threshold is used to determine the presence or absence of that compound. In this case, positive and negative mode FT-ICR-MS analyses did not provide definitive evidence for the formation of organic nitrogen species, including those proposed in Fig. 3, in the presence of cloud-relevant concentrations of HNO₃ or (NH₄)₂SO₄ when compared with samples from GLY + OH^• alone.

This finding is, perhaps, not surprising. In order to form organic nitrogen, NO₃^• radical reactions (Fig. 3a) require a C=C double bond. The enol form of GLY has this C=C double bond. But because the GLY enol forms from monohydrated or unhydrated GLY, not dihydrated GLY, the reaction in Fig. 3a is unlikely to occur in dilute aqueous solution. Although the nitration of alcohols with HNO₃ is common, there is little evidence of GLY nitration by HNO₃. A recent study suggests that HNO₃ acts as an oxidising agent that oxidises GLY to glyoxylic acid,^[57] although we did not see evidence of this using cloud-relevant HNO₃ as a source of NO₃^• radicals and in the presence of OH^• radicals.

In the daytime atmosphere, NO₃^• radical reactions are slow relative to OH radical reactions. The rate constant for GLY with NO₃^• is 1 × 10⁴ M^–1 s^–1[54] and the rate constant with OH radicals is 1 × 10⁹ M^–1 s^–1. Plus, daytime gas phase nitrate radicals are removed by photolysis in seconds. For these reasons, nitrate radical reactions are unlikely to be important to daytime cloud chemistry. At night, NO₃^• radicals are taken up into the aqueous phase from the gas phase. They can slowly oxidise organics by hydrogen atom abstraction and peroxy radical formation (e.g. forming glyoxylic acid from GLY) or they can react at the carbon–carbon double bond to form organonitrates. Because the GLY enol forms from monohydrated or unhydrated GLY, not dihydrated GLY, the Fig. 3a reaction is more likely to occur during droplet evaporation or in wet acidic aerosols.^[58] The nitration of hydrated GLY (diols) with HNO₃ (Fig. 3b) is also more likely to occur in very concentrated solutions, as might be present in evaporating droplets and in wet aerosols.

Likewise, although reactions with NH₄, including formation of imidazole, occur in wet aerosols, these reactions are slow relative to OH^• radical reactions. Our experimental conditions (1 mM GLY + 5 mM H₂O₂ + UV + 840 µM (NH₄)₂SO₄) produce hydroxyl radical concentrations of ~10^–12 M at pH 3 and the rate constant for organic nitrogen formation (k[GLY + NH₄⁺]) given by Noziere et al.^[34] is 3.6 × 10^–7 M^–1 s^–1. This is much smaller than the rate constant for GLY + OH^• (1.1 × 10⁹ M^–1 s^–1).^[21] This suggests that the presence of NH₄⁺ at cloud-relevant concentrations will not affect GLY + OH^• chemistry. In addition, we found no evidence of the imidazoles reported by Galloway et al.^[6] by FT-ICR MS in the positive mode.

Organic nitrogen is detected in fine and course particles.^[59] Previous studies have reported the formation of organonitrates in the gas phase or in the aerosol phase, where concentrations of ammonium and nitrate are quite high. For example, imine and imidazole formation have been observed in chamber studies and bulk aqueous aerosol mimics.^[6,34,60] Ammonium and nitrate concentrations are several orders of magnitude lower in clouds and in these experiments, and we did not see evidence of such products in the experiments reported herein.

IC/ESI-MS analysis permits the identification of compounds that co-elute in the IC system, for example, tartarate and malonate. The IC peak at that retention time (21.5 min; tartarate + malonate) is quantified as tartarate throughout these experiments (Fig. 2). Peak E in the IC/ESI-MS of the mixed standard (Fig. 4a) is tartarate. The corresponding mass spectrum of peak E shows that the m/z –149 ion (tartarate) is the only major compound detected in the mixed standard. Peak E was not present in the blanks. Peak E in Fig. 4b–d is a product of GLY photooxidation and reveals the presence of both tartarate (m/z –149) and malonate (m/z –103). Thus, both are products formed during GLY photooxidation in the presence and absence of nitric acid and ammonium sulfate. According to the chemical mechanism of Lim et al.,^[21] tartarate forms first and malonate forms more slowly during GLY photooxidation. Fig. 5 supports this claim by showing that the abundance of the m/z –149 ion in peak E (tartarate) is dominant early in the reaction. The abundance of the m/z –103 ion in peak E (malonate) reaches a maximum later in the experiment (between 40 and 120 min) when peak E (tartarate + malonate) is quite small (Fig. 2). Thus, we conclude that peak E is predominantly tartarate.

Fig. 5.  Relative abundances of m/z –149 (tartarate) and m/z –103 (malonate) ions in chromatographic peak E as a function of reaction time during the OH^• radical oxidation of glyoxal (GLY). Because peak E reaches its maximum between 10 and 30 min, we conclude that peak E is predominantly tartarate.

Verification that tartarate is the major first-generation radical–radical product of GLY + OH^• provides critically important support for the oligomer formation mechanism proposed by Lim et al.^[21] Although Tan et al.^[33] identified a peak with the same IC retention time as tartarate–malonate, the IC/ESI-MS data presented here is the first confirmation that this peak is predominantly tartarate. The slower formation of malonate is also consistent with the organic radical–radical chemistry proposed by Lim et al.,^[21] because malonate forms through radical–radical reactions followed by acid catalysed dehydration. Lim et al.^[21] proposed that, at the high organic concentrations that are found in wet aerosols (1–10 M), organic radical–radical reactions dominate, producing higher carbon number products. Note that Lim et al.^[21] used GLY as a surrogate species to represent total dissolved organics in wet aerosols, in order to better understand the differences in chemistry in clouds and in wet aerosols. It should be recognised that both organics and inorganics are present at high concentrations in wet aerosols. Although organic–inorganic interactions do not appear to play an important role in clouds, they can play an important role in droplet evaporation chemistry and in wet aerosols. The chemistry leading to SOA formation in evaporating droplets and in aerosols is clearly complex and remains poorly understood.

The biogenic and anthropogenic source strengths of isoprene and aromatics and the water-solubility of GLY, a gas-phase product of these sources, make GLY a potentially important precursor to SOA_aq formation. This research suggests that the currently understood chemistry leading to the formation of SOA through in-cloud OH^• radical oxidation of GLY^[21,52] is robust in the presence of ammonium sulfate and nitric acid. In contrast, others have demonstrated that organic nitrogen products can form at the much higher concentrations found in wet aerosols.^[6,34] They might also form during cloud and fog droplet evaporation, where GLY may be present in its monohydrated form.^[58] Detection of tartarate during the GLY + OH^• experiments herein provides key verification of the organic radical–radical oligomer formation mechanism proposed by Lim et al.^[21] This chemistry is detectable in experiments with 1 mM of organic, and is predicted to be dominant at the high (1–10 M) concentrations of organics found in wet aerosols. Thus, oxalate is the major product of dilute (in cloud) GLY oxidation and oligomers^[21] and organic nitrogen^[34] are major products of GLY chemistry in wet aerosols.

Additional experimental measurements and modelling results are provided in the Supplementary material. This includes measured species (i.e. H₂O₂, oxalate, NO₃^–) and modelled species (i.e. H₂O₂, oxalate, NO₃^–, OH^•, pH) for selected experiments and control experiments. The Supplementary material is available from the journal online (see http://www.publish.csiro.au/?act=view_file&file_id=EN13074_AC.pdf).