CESAM is a 4.2 m3 temperature- and pressure-controlled, stirred, stainless steel chamber (Wang et al., 2011) with solar simulator lamps (Harris et al., 2017) and held just above ambient pressure with automated flows of high-purity O2 and liquid N2 boil-off at a respective 20/80 v/v ratio. The chamber-gas-phase contents were monitored by a relative humidity (RH) sensor (Vaisala HMP234 Humicap), long-path Fourier-transform infrared (FTIR) spectroscopy (Bruker Tensor 37, 182.5±0.5 m path length; Wang et al., 2011; glyoxal integrated band intensity at 2950–2700 cm-1=6.34×10-18 cm molec.-1; Eurochamp, 2020), and high-resolution proton transfer reaction mass spectrometry (PTR-MS, KORE Tech. Series II, inlet temperature 100 ∘C, proton transfer reactor P=1.64 mbar, glow discharge P=1.94 mbar, PTR entry voltage =400 V, E/N ratio =130). Polydisperse seed particles (TSI 3076 atomizer) were diffusion-dried before addition to the dry chamber. They were then continuously sampled through a 1 m Nafion drying tube to scanning mobility particle sizing (SMPS; TSI, 20–900 nm) and cavity-attenuated phase shift single-scattering albedo (CAPS-ssa; 450 nm; Aerodyne; Onasch et al., 2015) spectrometers. A particle-into-liquid sampler (PILS; Brechtel Manufacturing) sampled N2-diluted chamber aerosol through an activated carbon denuder into a capillary waveguide UV–vis spectrometer (LWCC-100, 0.94 m path length). Water vapor was added in bursts from a stainless steel boiler (Wang et al., 2011), and chamber RH was subsequently stabilized by routing inlet N2 flow through a heated high-purity water bubbler. A droplet spectrometer (Palas Welas Digital 2000, 0.5 to 15 µm diameter, on chamber flange; Wang et al., 2011) extended the size range of detected aerosol into supermicron particles. CAPS-ssa aerosol extinction and scattering signals were zeroed against filtered chamber air every 5 min to ensure that any gas-phase species absorbing light at 450 nm does not influence measurements and averaged to SMPS scan frequency. SMPS number and concentrations and PTR-MS signals were corrected for dilution caused by flows into the chamber. SMPS size distributions were also corrected using size-dependent wall losses measured for AS particles in the chamber.

Additional experiments were conducted in a 300 L collapsible Tedlar chamber. Aerosols were generated from 0.1 % w/w aqueous solutions (TSI 3076 atomizer) and diffusion-dried (except in experiments on “wet” aerosol). Glyoxal production was monitored at the inlet by absorbance at 405 nm using a cavity ring-down (CRD) spectrometer and a cross section of 4.491×10-20 cm2 molec.-1 (Volkamer et al., 2005b). Glyoxal concentrations at the chamber outlet were measured in test experiments to determine wall loss rates (∼6.7×10-4 s-1). Glyoxal inlet concentrations, flow mixing ratios, and wall loss rates were then used to estimate glyoxal chamber concentrations. Aerosol particles were sampled via diffusion driers by a quadrupole aerosol mass spectrometer (Q-AMS; Aerodyne), CAPS-ssa (Aerodyne, 450 nm), SMPS (TSI), CRD (405 and 530 nm; Ugelow et al., 2017), and photoacoustic spectrometers (PASs; 405 and 530 nm; Ugelow et al., 2017), all of which were periodically baselined through filters to eliminate interferences by gas-phase species. RH sensors monitored humidity levels at the aerosol inlet, chamber outlet, and dried chamber outlet flows. Water vapor was added in certain experiments by passing inlet flows through Nafion humidifiers.

Reagents were used as received from Sigma-Aldrich unless otherwise mentioned. Solutions for aerosol generation were generated by dilution of glycine (>99 %) to 5 mM, AS (>99 %) to 1.2–10 mM, or sodium sulfate (≥99 %) to 7 mM in deionized water (> 18 MΩ, ELGA Maxima). MeAS was generated by mixing methylamine and sulfuric acid (Mallinckrodt) solutions at a 2:1 molar ratio; after dilution to 6.3 mM, solution pH was 4.5. Gas-phase glyoxal was generated by heating solid mixtures of glyoxal trimer dihydrate (Fluka, > 95 %) and P2O5 (99 %) to 110–150 ∘C; the glyoxal produced was flushed into the chamber with dry N2 (Volkamer et al., 2009).

Experiment 1, in which dry AS aerosol was sequentially exposed to 0.05 and then 0.50 ppm glyoxal at t= 4:47 and 5:16 h, respectively, is summarized in Fig. 1. Both glyoxal additions were detectable by PTR-MS at m/z 31 and 59. The m/z 59 signal, however, is elevated in the clean and dry chamber before glyoxal is added, indicating background interference by another chemical species or its fragment in the mass spectrometer. SMPS data, which have been corrected for wall losses and for dilution, show no observable aerosol growth after either glyoxal gas addition. This lack of observed growth at < 5 % RH is consistent with previous studies under very dry conditions (Kroll et al., 2005; De Haan et al., 2017). However, as the optical and chemical measurements described below show, this lack of growth does not indicate a lack of glyoxal reactivity.

The addition of 0.05 ppm glyoxal gas at t= 4:47 h triggered a short-lived drop in albedo by 0.034 that was observed by CAPS-ssa spectrometry at 450 nm. A second, larger glyoxal addition (0.50 ppm) at t= 5:16 h caused albedo to plummet to 0.75, a change 7 times greater than the first. Even though particle sizes did not increase after either glyoxal addition, the significant albedo declines indicate that glyoxal reactions rapidly produced light-absorbing products at dry AS particle surfaces. Over the next 30 min, as glyoxal gas-phase concentrations decreased by half (likely to due to chamber wall losses), aerosol albedo recovered proportionately, indicating that this surface brown carbon formation under dry conditions is fully reversible.

At t= 5:46 h (Fig. 1), the chamber was humidified to 50 % RH, a level which would not deliquesce the AS seeds (Biskos et al., 2006) but which may produce as many as 1–2 monolayers of adsorbed water at aerosol surfaces consisting of solid AS (Denjean et al., 2014; Romakkaniemi et al., 2001) or glyoxal reaction products (Hawkins et al., 2014). Humidification to 50 % RH caused significant changes to both the gas and aerosol phases. Glyoxal PTR-MS signals and aerosol albedo returned back to near-baseline levels within a few minutes, while the dried aerosol mass measured by SMPS spectrometry jumped downward by 15 %. The loss of gas-phase glyoxal, again without aerosol growth, suggests that water greatly accelerated glyoxal loss rates to the steel chamber walls. The simultaneous albedo recovery and SMPS mass loss indicate that humidification destroyed all brown carbon products that absorb 450 nm light, converting some fraction of them to gas-phase products. A proposed mechanism for this process is discussed below. It is significant that no browning was observed in PILS-sampled aerosol at any point during Experiment 1, presumably due to the same mechanism occurring during wet sampling.

The 15 % aerosol mass loss upon humidification to 50 % RH is surprising given that no corresponding mass gain was recorded during exposure to glyoxal under dry conditions. However, the lack of mass gain under dry conditions cannot be interpreted as a lack of glyoxal uptake or reactivity given the large observed drop in albedo. Instead, the mass loss upon humidification suggests that at least 15 % of the volume of AS seeds had been replaced under dry conditions by glyoxal reaction products that could break down into gas-phase species once water was added. Simultaneous increases in gas-phase PTR-MS signals for m/z 47 (formic acid) and 61 (acetic acid) indicate that these acids were two of the gas-phase products generated by humidification. Formic acid is a known byproduct of imidazole production by aqueous-phase glyoxal–ammonia reactions (De Haan et al., 2009a; Yu et al., 2011).

Dried seed aerosol particles atomized from AS–glycine mixtures were also exposed to 0.25 ppm glyoxal under dry conditions in Experiment 2 (Fig. S1 in the Supplement). The response of these internally mixed seeds to glyoxal exposure was comparable to that of pure AS seeds. No growth was observed by SMPS spectrometry, and aerosol albedo at 450 nm was anticorrelated with PTR-MS glyoxal signals at m/z 59, as before. The most significant difference between the experiments is that there was no 15 % loss of aerosol mass observed by SMPS spectrometry upon humidification of the chamber to 50 % RH even though acetic and formic acid were again released into the gas phase. This may be due to the lower volatility of deprotonated seed particle materials (glycine vs. ammonia). In addition, most glycine-derivatized imidazole products have permanent positive charges and are not in equilibrium with volatile neutral forms (De Haan et al., 2009a).

To better understand the reactive processes happening in the dry aerosol particles, further experiments were conducted in a 300 L Tedlar chamber probed by Q-AMS, SMPS, CAPS-ssa, CRD, and PAS spectrometry. Figure S2 shows an AMS ion correlation plot comparing average signals before and after 40 ppb glyoxal was added over a period of 25 min to the dry chamber containing AS aerosol in Experiment 3. Unsurprisingly, the slow addition of this smaller amount of glyoxal did not cause observable net particle growth or a decline in aerosol albedo at 450 nm. A marginal (0.9 µg m-3, S/N =1.6) increase in total organic aerosol was observed by Q-AMS spectrometry during glyoxal addition, associated with significant increases (+30 % or more relative to conserved aerosol species) in ion signals at m/z 15 (CH3+ or NH+ fragments), 23 (Na+), 29 (CHO+ fragment), 47 (formic acid + H+ or a CH3O2+ fragment), 69 (imidazole-H+), 81 (pyrazine-H+), and 97 and 119 (imidazole carboxaldehyde, IC, ionized by H+ or Na+, respectively). Slight decreases in aerosol water signals were observed at m/z 16. The detection of particle-phase imidazole, pyrazine, and IC suggests the presence of larger, light-absorbing molecules such as 2,2'-biimidazole and N-heterocycle derivatives that are typically associated with the products detected here (Kampf et al., 2012; Hawkins et al., 2018; Grace et al., 2019). However, detection of the larger product molecules directly from dry aerosol may require a soft, direct ionization technique such as extractive electrospray ionization mass spectrometry (EESI-MS).

Proposed chemical mechanisms for brown carbon production at AS particle surfaces are summarized in Schemes 1 and S1 in the Supplement. Except for steps where new N-heterocyclic rings are formed, all processes are reversible (Kampf et al., 2012). Thus, a reduction in gas-phase glyoxal concentrations will shift reversible reactions away from brown carbon back towards simple N-heterocycle products, which do not absorb 450 nm light. Humidification to 50 % RH accelerates this shift by removing more glyoxal from the gas phase and perhaps also by hydrolysis of double bonds and dilution effects (Rincón et al., 2010; Phillips and Smith, 2014, 2015). Humidification also triggers the observed evaporation of formate as formic acid and perhaps the evaporation of other small N-containing products.

The anticorrelation of albedo with glyoxal concentrations in Experiments 1–4 is summarized in Fig. 2. Although AS–glycine–glyoxal bulk aqueous mixtures have been shown to brown more than mixtures without glycine (Trainic et al., 2012; Powelson et al., 2014), here we see that dry AS and AS–glycine aerosol particles brown similarly for a given concentration of gas-phase glyoxal. This may indicate that glycine is not at the aerosol surface or that glycine surfaces, when present, are less able to retain adsorbed water in the dry chamber. We therefore fit the combined dataset from all four experiments. Albedo shows a clear downward curvature at high glyoxal concentrations such that the relationship is best fit by a second-order polynomial. This suggests that the formation of the compounds absorbing at 450 nm is proportional to [glyoxal]2. While glyoxal–ammonium reactions are first-order in glyoxal in dilute solution when [glyoxal] × [NH4+] < 1.2 M, they switch to second-order at higher concentrations (Noziere et al., 2009), which are likely in these dry experiments.

Reversible surface browning of AS aerosol under dry conditions was also recently observed during exposures to methylglyoxal gas (De Haan et al., 2017). The albedo values observed before and after two methylglyoxal additions are shown for comparison in Fig. 2. Although the data show a slight negative offset due to particle size effects, judging by the slope it is clear that methylglyoxal's effect on the albedo of dry AS aerosol is significantly less than glyoxal. This is the opposite of the trend in brown carbon production in bulk aqueous solutions at pH 5, where methylglyoxal is much more effective in generating light-absorbing products (Powelson et al., 2014), perhaps due to the fact that its ketone functional group is far less likely to be inactivated by hydration than the aldehyde groups on both molecules. However, in these dry aerosol experiments in which water is scarce, glyoxal's greater attraction to water (seen in its much higher Henry's law coefficient; Betterton and Hoffmann, 1988; Ip et al., 2009; Kampf et al., 2013) may allow it to interact with small amounts of adsorbed water at the AS aerosol surface far more effectively than methylglyoxal.

Gas-phase glyoxal was added to a few other types of seed particles in the small chamber. In experiments on MeAS seeds (Experiment 5, Fig. 3, top panel), the slow addition of 140 ppb of glyoxal caused a matching drop of -0.15 in aerosol albedo measured by CAPS-ssa spectrometry at 450 nm and an increase in aerosol absorbance to 28 Mm-1 measured by the PAS at 405 nm. The albedo decline at 450 nm (Fig. 2, green circles) is 4 times greater than observed on AS seeds at similar glyoxal concentrations. This result is consistent with earlier aqueous-phase studies showing greater browning in glyoxal–methylamine mixtures than in glyoxal–AS mixtures at the same pH (Powelson et al., 2014).

Albedo at 405 nm was calculated from PAS and CRD signals in Experiment 5 (Fig. S3), showing that albedo had dropped to 0.30 by 13:11 and remained at this level for 45 min. These albedo values indicate that maximum light absorbance at 405 nm was 4.7 times greater than at 450 nm and persisted for a longer period of time after glyoxal gas concentrations declined. At even longer wavelengths (530 nm), PAS aerosol absorbance reached only 0.9 Mm-1, further indicating highly wavelength-dependent light absorption. The absorption spectra of atmospheric brown carbon are typically well fit by exponential decay functions. Such featureless spectra can be characterized by an Ångstrom absorption coefficient α, which is the slope of a log(absorbance) vs. log(wavelength) plot. Comparing the amount of light absorbance at 405, 450, and 530 nm at 13:11 in Experiment 5 gives α=-12.7±0.8 (Fig. S5). A similar analysis of Experiment 4 (Figs. S6 and S7, dry AS–glyoxal), the only other experiment with measurable absorbance at all three wavelengths, gives a comparable α=-16±4. Thus, it appears that brown carbon formed by glyoxal under dry conditions on AS and MeAS aerosol absorbs light with similar wavelength dependence.

In an experiment on dry sodium sulfate seeds at ∼35 % RH in the small Tedlar chamber (Experiment 6, Fig. 3) glyoxal concentrations were increased from zero to ∼2000 ppb over 30 min. During this time, albedo at 450 nm remained at 1±0.005, and no aerosol absorbance was measured by PASs at either 530 or 405 nm. The lack of browning observed even at such high glyoxal concentrations confirms that ammonium or methylammonium ions (or ammonia or methylamine) are necessary reaction partners with glyoxal in the browning process observed in Experiments 1–5. It also confirms that our CAPS-ssa and PAS measurements are not biased by absorbance due to gas-phase glyoxal.

Finally, three experiments exploring browning on wet rather than dry AS aerosol were conducted at RH ranging from 38 % to 81 %. The highest-humidity experiment (Experiment 7) is summarized in Fig. S8. In Experiments 7–9, albedo declines of 0.013 or less were observed following addition of 1.1, 1.2, and 0.12 ppm of glyoxal gas, respectively. If graphed in Fig. 2, the resulting slopes for Experiments 7–8 would be more than 1000 times flatter than the methylglyoxal data shown for comparison. While some of the glyoxal gas added may have been quickly lost to the walls of the humid chambers as an equilibrium is established (Kroll et al., 2005), especially in Experiments 7 and 9, it is clear that wet AS aerosol particles brown much less than dry AS, AS–glycine, or MeAS aerosol upon exposure to glyoxal.

Enhanced AS aerosol browning under dry conditions is surprising given that glyoxal Maillard chemistry is normally considered to be an aqueous-phase process. One clue to the nature of the dry browning process is seen in the slight depletion of water signals observed in all dry experiments probed by Q-AMS spectrometry (nos. 3, 4, 5, Figs. S2 and S9) after browning caused by glyoxal exposure (most water is removed from aerosol particles in the AMS inlet). The extra water depletion associated with glyoxal exposure of dry aerosol, which was not observed in deliquesced aerosol experiments probed by Q-AMS spectrometry (nos. 7–8), suggests that even under dry conditions, glyoxal is able to access and deplete trace amounts of aerosol-phase surface water. Any adsorbed water would be saturated with ammonium (or methylammonium) sulfate, and the presence of dissolved AS is known to greatly increase glyoxal uptake via a “salting-in” effect (Kampf et al., 2013; Waxman et al., 2015), while methylglyoxal solubility is reduced by salting out (Waxman et al., 2015). Thus, both glyoxal and AS are expected to be concentrated in any surface-adsorbed water present. In a previous study, similar reasoning was used to explain glyoxal uptake on solid seed particles at RH levels as low as 10 % (Corrigan et al., 2008). Furthermore, the scarcity of water will favor dehydration of products, helping to form light-absorbing conjugated double bonds.