Revisiting the reaction of dicarbonyls in aerosol proxy solutions containing ammonia: the case of butenedial

. Reactions in aqueous solutions containing dicarbonyls (especially the  -dicarbonyls methylglyoxal, glyoxal, and 10 biacetyl) and reduced nitrogen (NH X ) have been studied extensively. It has been proposed that accretion reactions from dicarbonyls and NH X could be a source of particulate matter and brown carbon in the atmosphere and therefore have direct implications for human health and climate. Other dicarbonyls, such as the 1,4-unsaturated dialdehyde butenedial, are also produced from the atmospheric oxidation of volatile organic compounds, especially aromatics and furans, but their aqueous phase reactions with NH X have not been characterized. In this work, we determine a pH-dependent mechanism of butenedial 15 reactions in aqueous solutions with NH X that is compared to  -dicarbonyls, in particular the dialdehyde glyoxal. Similar to glyoxal, butenedial is strongly hydrated in aqueous solutions. Butenedial reaction with NH X also produces nitrogen-containing rings and leads to accretion reactions that form brown carbon. Despite glyoxal and butenedial both being dialdehydes, butenedial is observed to have three significant differences in its chemical behavior: (1) as previously shown, butenedial does not substantially form acetal oligomers, (2) the butenedial/OH - reaction leads to light-absorbing compounds, and (3) the butenedial/NH X reaction is fast and first order in the dialdehyde. Building off of a complementary study on butenedial gas-particle partitioning, we suggest that the behavior of other reactive dialdehydes and dicarbonyls may not always be adequately predicted by  -dicarbonyls, even though their dominant functionalities are closely related. The carbon skeleton (e.g., its hydrophobicity, length, and bond structure) also governs the fate and climate-relevant properties of dicarbonyls in the atmosphere. If other dicarbonyls behave like butenedial, their reaction with NH X could constitute a regional source of brown 25 carbon to the atmosphere.

been shown to be greatly accelerated in evaporating cloud droplets (Lee et al., 2013) and produce strongly absorbing 65 compounds even at low concentrations (Shapiro et al., 2009). Methylglyoxal/NHX reactions in bulk solutions are faster and are shown to be linearly dependent on methylglyoxal and ammonium, but they too may be limited in the atmosphere (Powelson et al., 2014;Sareen et al., 2010;Sedehi et al., 2013). Biacetyl/NHX reactions are also shown to be linearly dependent on the dicarbonyl and ammonium and can form light-absorbing species, although biacetyl is much less hydratable and is not as relevant to aqueous aerosol (Grace et al., 2020). Further supported by field observations, modeling, and other laboratory work, 70 the consensus is that dicarbonyl/NHX reactions probably are too slow to contribute substantial particle mass in the atmosphere (Ervens and Volkamer, 2010;Laskin et al., 2015).
Beyond reaction with NHX, past work with -dicarbonyls suggests that the dicarbonyl family tends to behave similarly in aqueous solutions. At basic pH, methylglyoxal and glyoxal both react with OHto form colorless oligomeric products. Both are hydratable when dissolved in an aqueous solution, and therefore have suppressed vapor pressures over 75 aqueous media. However, there are key differences in the behavior of these compounds, which have been attributed to the dicarbonyl moiety. Glyoxal favors a dihydrate state in dilute solutions that can oligomerize with mono-or unhydrated glyoxal to form cyclic hemiacetals. Methylglyoxal hydration occurs predominantly at the aldehyde, resulting in a ketodiol that is much less reactive than glyoxal's monohydrate, which possesses an aldehydic group. Oligomerization of methylglyoxal and biacetyl proceeds predominantly through aldol condensation rather than through the formation of acetals. In the presence of anions like 80 sulfate, methylglyoxal vapor pressure is increased (salting out) whereas glyoxal's decreases (salting in) (Kampf et al., 2013;Wang et al., 2014;Waxman et al., 2015). The accumulation of such dissimilarities can lead to different overall reactivities and solubilities between glyoxal and methylglyoxal. For example, glyoxal has an effective Henry's Law constant that can be ~4 orders of magnitude larger than methylglyoxal when partitioning over salt-containing solutions (McNeill, 2015;Waxman et al., 2015), which can impact the loading and composition of atmospheric aerosol. The Henry's law constant of biacetyl is even 85 lower than that of methylglyoxal. In general, it is thought that behavioral differences can be explained by the ketone versus aldehyde content of dicarbonyls and, thus, butenedial would be expected to behave similarly to glyoxal.
The influence of carbon backbone structure on dicarbonyl reactivity has received far less attention, including in reactions with NHX. Kampf et al. (2016) observed that the reactions of two larger dicarbonyls, 2,5-hexadione (an unsaturated 1,4-diketone) and glutaraldehyde (an unsaturated 1,5-dialdehyde), with NH X are fast, and their products may be more light-90 absorbing than those of -dicarbonyls/NH X reactions. Building off of this work, we investigate butenedial reactions in aqueous solutions containing NHX and compare them to those of -dicarbonyls. Butenedial is an unsaturated 1,4-dialdehyde observed as a major product of aromatic and furan degradation in laboratory studies Aschmann et al., 2014;Bierbach et al., 1994;Coggon et al., 2019;Stockwell et al., 2015;Strollo and Ziemann, 2013;Volkamer et al., 2001) and in ambient air . As butenedial is an electron-poor dialdehyde, its chemical reactivity is expected to be more similar 95 to that of glyoxal than methylglyoxal or biacetyl. However, we have shown in a past study that butenedial acetal oligomers are insignificant in water and that it salts out in the presence of anions, behavior that is unlike that of glyoxal and more similar to methylglyoxal (Birdsall et al., 2019). Butenedial is studied first in aqueous solutions without NHX, in which it reacts reversibly with H2O and irreversibly with OH -. Then it is studied in AS aqueous solutions, in which it can additionally react irreversibly with NHX. Reaction 100 products and reaction rates are observed with NMR to characterize the chemical mechanism. The reactivity of butenedial is compared to glyoxal as well as methylglyoxal and biacetyl. Implications for unsaturated 1,4-and other dicarbonyls are considered. The atmospheric relevance of dicarbonyl/NHX reactions is re-examined, including as a source of brown carbon.

Methods
A chemical mechanism was developed for butenedial in an aqueous NHX solution. Butenedial reaction with OHor NHX was 105 studied in bulk solutions with conditions given in Table 1. As is justified later, reaction with OHwas found to be negligible at the pH and NHX conditions of the butenedial/NHX reaction studies, and vice versa, as no overlapping products were observed in solutions. No other reactions were observed. Measurements of the composition of the bulk solutions were taken with NMR or MS and identified reaction products were used to formulate a chemical scheme. A custom kinetic mechanism was fit to the temporal evolutions of reactants and reaction products (via NMR only), from which rate constants were empirically 110 determined. Predictions with the kinetic mechanism were compared against additional experimental measurements at different pH and NHX.

Materials and instruments
All chemicals were obtained from Sigma Aldrich unless otherwise specified. The synthesis of butenedial is described elsewhere in greater detail (Avenati and Vogel, 1982;Birdsall et al., 2019). Mixtures containing 2.4 M 2,5-dihydro 2,5-115 dimethoxyfuran (TCI America, 98%) and 3.4 M glacial acetic acid (HAc, VWR, 99.7%) were prepared. After about 10 days of reaction at room temperature, we purified mixtures with rotary evaporation to 75% butenedial by weight (w/w). The remaining 25% was predominantly residual water and HAc. All mixtures were well mixed at the start of reactions. Reacting mixtures were kept in capped glass vials or capped NMR tubes without further stirring.
All 1D 1H-NMR spectra were collected with Varian 400 MHz spectrometers, 64 scans per recorded spectrum. 120 Deuterium oxide (D 2 O, 99.9 atom % D) was the solvent for NMR experiments. The methods described by Yu et al. (2011) were followed to estimate molarities quantitatively in 1H-NMR spectra with 1% w/w dimethyl sulfone (DMS, 99%). The pH was estimated from the spectra by tracking the proton shift of an acid near 50% dissociation, according to the methods described by Yu et al. (2011). The acid that was tracked depended on the solution pH: if solution pH 3-6, residual acetic acid was tracked; if solution pH 6-8.5, 1% w/w methylphosphonic acid (MPA, 98%) was added and tracked; and if solution pH 9-125 11, 1% w/w 2,4,6-trimethylphenol (TMP, 99%) was added and tracked. analyte solution with a droplet-on-demand injector. The particles were briefly (< 1 s) in contact with dry nitrogen gas. Reaction 130 and evaporation of solutes were negligible while the particles traveled from injector to a glass slide heated at 220 C. The glass slide heated the particle, and a corona discharge ionized the resulting vapors to be drawn into the inlet of the TOF-MS. Mass spectral signals were recorded as counts per integer mass-to-charge (m/z) channel. 10-20% w/w hexaethylene glycol (PEG-6, 99%) was used as an internal standard for MS measurements, as done previously (Birdsall et al., 2018). Previous work demonstrated there is no reaction between butenedial and PEG-6 (Birdsall et al., 2019). Distilled H2O was the solvent for all 135 MS experiments. The TOF-MS was operated in positive mode while the LC-MS was operated in negative mode.

Butenedial in aqueous solutions without NHX
0.55 M butenedial in D2O solution and a six-fold dilution of this solution, resulting in 0.09 M butenedial in D2O, were studied with NMR under acidic conditions. In both solutions, butenedial strongly favored the dihydrate form with only minimal 140 formation of acetal oligomers (see Supplement Section 2, Figure S2 and Table S2), as shown previously (Birdsall et al., 2019).
Additionally, four solutions containing 0.2-0.3 M butenedial were prepared in D2O buffered with 1 M Na2CO3-NaHCO3 to pH 8.8-10.4. All solutions immediately turned dark brown. Butenedial was measured quantitatively with NMR throughout its reaction with OH -. Disproportionation products were observed ( Figure S4), including the growth of broad peaks embedded in the baseline that were indicative of accretion reactions. 145

Butenedial in aqueous solutions with NHX
0.9 M butenedial/0.45 M AS (VWR, > 99%) mixtures were prepared in water and D2O with the internal standards PEG-6 or DMS and 0.5 M sodium carbonate (Na2CO3) -sodium bicarbonate (NaHCO3) buffer. The solution immediately turned orange brown. After 20 min of reaction, mass spectra of the mixtures indicated nitrogen-containing products with signals at m/z 84, 149, 150, and 168, assumed to be adducts with H + ( Figure S7). The most reasonable chemical formulas of these products were 150 C4H5NO (83 Da), C8H8N2O (148 Da), C8H7NO2 (149 Da), and C8H9NO3 (167 Da). C8H9NO3 was the parent molecule for the C 8 H 7 NO 2 fragment.
The 1H-NMR spectra ( Figure S8) showed distinct groups of quantitative related signals that had similar temporal behavior. Each group of peaks whose quantitative signal strength behaved as integers and had the same temporal behavior was presumed to arise from a single compound. A molecular structure was proposed for each cluster of peaks and the molecular 155 formulas mentioned above. The inferred products were as follows: 2-pyrrolinone (pyrrolinone, PR, C4H5NO), and a butenedialpyrrolinone "dimer" (BD-PR, C8H9NO3). We propose that 2-butenal-1,3-diazepine (diazepine, DZ, C8H8N2O) is a minor product that is observable with MS but was not detected with the less sensitive NMR. The growth of broad peaks embedded in the baseline suggested substantial formation of N-containing accretion products that were likely strongly π-conjugated and should therefore absorb light and could explain the dark color of the solution.

Kinetic mechanism
The kinetic mechanism was formulated as a system of ordinary differential equations, one ordinary differential equation per identified chemical compound. The butenedial reaction with OHwas quantified and used as foundation for the mechanism with NHX. Rate laws were formulated based on known reactions of related species and then adjusted to optimize agreement between the mechanism and observations (see Section 3). Python's scipy package was used to parameterize each fit rate 165 constant's mean value and standard error. The lmfit library with the Levenberg-Marquardt algorithm was used to perform the least squares minimization. A Monte Carlo simulation was performed to derive the reported 95% confidence intervals on model runs.

Butenedial/OHreaction
Dicarbonyl/OHreactions are known to be effectively irreversible and are characterized by a rate law that is first order in the 170 dicarbonyl (R1). According to Fratzke & Reilly (1986), the dicarbonyl/OHreaction rate constant (k1) is a function of OHas defined by the following relationship: where aI and aII are related to the role of the hydrated anion or dianion, respectively (Fratzke and Reilly, 1986). The coefficients aI, aII, and aIII were fit with four unique [OH -]/k1 pairings, each corresponding to a different experimental run. k1 was derived 175 from the first order loss of butenedial in each experiment ( Figure S12). Subsequently, aI-aIII were determined using the scipy optimize.curve_fit library implemented with the Trust Region Reflective minimization method.

Butenedial/NHX reaction
A system of three ordinary differential equations (DE1-DE3) was used to model butenedial, pyrrolinone, and butenedialpyrrolinone dimer concentrations, with pH and initial concentrations of reactants and products as inputs. Rate constants for 180 five rate laws (R2-R6) were fit to experimental data, with starting conditions of 0.9 M butenedial and 0.9 M NHX, and pH ranging 4.2-5.7. k 1 was implemented according to the fitting described in Section 2.3.1. pH was estimated with an empirical formulation that agreed closely with measurements ( Figure S13). Model performance was assessed against measurements taken in bulk liquid experiments with different initial conditions and pH (Figures S15-S16). One limitation was that reaction rates of unmeasured species had to be approximated with a proxy, i.e., in the cases of NHX and accretion product 185 concentrations. As is discussed in Section 3, the approximations have minimal impact on the prediction of butenedial loss and on the estimation of most parameters.

Results
A chemical scheme for butenedial in an aqueous solution with NHX is proposed in Section 3.1. Butenedial reaction with OHand reaction with NHX are described in Section 3.2 and Section 3.3, respectively. The fate of butenedial in aqueous solutions 190 with and without NHX is summarized in Section 3.4.

Chemical scheme
Butenedial in aqueous NHX solutions of pH 3.6-10.4 is proposed to obey the chemical scheme shown in Figure 1, which demonstrates that it can undergo three reactions. First, butenedial can be reversibly hydrated and is observed to prefer the dihydrate form in aqueous solutions without evidence for significant acetal oligomer formation. Second, butenedial reacts with 195 OHto form irreversible reaction products such as a hydroxy acid, which ultimately lead to oligomeric, light-absorbing compounds. Third, butenedial reacts with NH3 to produce an imine intermediate which forms irreversible reaction products (pyrrolinone, a diazepine, and a butenedial-pyrrolinone "dimer") and also ultimately lead to oligomeric, light-absorbing compounds. These accretion products are observed to be reactive with butenedial, pyrrolinone, and the butenedial-pyrrolinone dimer. 200 As discussed in a previous study, like glyoxal, the hydration equilibrium of butenedial is strongly shifted toward the dihydrate (>95% of the total butenedial on a molar basis). Birdsall et al. (2019) also showed that the ratio of unhydrated or monohydrated to dihydrated butenedial appeared to be unaffected by the availability of water. It is expected that the hydration behavior of butenedial will affect its reactivity, and possibly differentiate it from glyoxal, which has been observed to form a highly reactive yet soluble monohydrate form. Additionally, in contrast to glyoxal, which readily forms glyoxal acetal 205 oligomers, no butenedial acetal oligomers were observed by Birdsall et al. (2019). Our experiments support that only minimal acetal oligomer formation is possible and is much less pronounced than for glyoxal. This behavior is not typical for dialdehydes but has been observed for adipaldehyde (Hardy et al., 1972). One explanation is that butenedial, like adipaldehyde, has a hydrophobic center that influences the ability of its hydrates to oligomerize like glyoxal.
The 1H-NMR spectra of butenedial in basic aqueous solutions and solutions with NHX both exhibit the buildup of 210 signal in the baseline in the vicinity of possible monomer peaks (shifted to lower ppm, Figure S2, S5). This buildup increases in intensity and spreads out with respect to chemical shift over time. Thus, accretion reactions take place that are the ultimate sink for butenedial and its reaction products, producing low-volatility compounds that can explain the brown color. Reaction products of dicarbonyl/OHreactions are not thought to be reactive to the dicarbonyl or to products of dicarbonyl/NHX reactions, such as in the case of glyoxal (Yu et al., 2011). Butenedial loss is observed to be first order in butenedial at all time 215 scales and [OH -], indicating that butenedial/OHreaction does not result in additional butenedial removal from solution, e.g., via products reacting with butenedial. On the other hand, butenedial is reactive with butenedial/NHX reaction products, as has been observed in analogous reactions of glyoxal, methylglyoxal, and biacetyl, and they and higher accretion products increase butenedial loss.

Butenedial/OHreaction 220
Estimates of the a-coefficients of k1 are shown in Table 2. The dependence of the pseudo first-order rate constant k1 on OHis shown in Figure 2. The rate constant is < 1×10 -4 s -1 at pH < 9, indicating that butenedial/OHreaction is insignificant except at basic pH. No evidence of reaction has been observed in standard butenedial solutions (pH ~ 4) that we have kept on the shelf for months. At solution pH 8.5-8.8, butenedial loss from butenedial/OHreaction is negligible compared to butenedial/NHX reaction ( Figure S14). We therefore conclude that butenedial/OHreaction is insignificant at neutral and acidic conditions 225 relevant to the atmosphere, especially when NHX is present, although the parametrization is included in the butenedial/NHX kinetic mechanism.

Butenedial/NHX reaction
Five chemical reactions (R2-R6) explain the evolution of butenedial (BD) and its major reaction products with NHX, pyrrolinone (PR) and butenedial-pyrrolinone "dimer" (BD-PR). Table 3 shows the proposed reactions, their rate laws and fitted 230 rate constants. The reactions and their rate laws are discussed here and the fitted rate constants in Section 3.3.2.
The initial, irreversible reaction between butenedial and NH3 (R2) is linearly dependent on both species and produces pyrrolinone. While the reaction is acid catalyzed, the rate constant of dialdehyde/NHX reactions is pH dependent (Yu et al., 2011), resulting in a pH-independent rate law. At constant NH3, the reaction is pseudo first order in butenedial. In analogy to the related glyoxal and methylglyoxal reactions, we propose an imine intermediate for this reaction. Reaction of pyrrolinone 235 with butenedial is pH-dependent and produces a butenedial-pyrrolinone dimer (R3). We anticipate that the proposed imine can also undergo an acid-catalyzed reaction to produce a diazepine (DZ).
Reactions between each of butenedial, pyrrolinone, butenedial-pyrrolinone dimer and an accretion product term (R4-R6) are included to represent the removal of these species through accretion reactions. Accretion reactions have been observed in studies with glyoxal (Kampf et al., 2012;Yu et al., 2011). These reactions typically involve oligomer-like molecules made 240 up of precursor compounds (in this case, butenedial), its reaction products (pyrrolinone and butenedial-pyrrolinone dimer), and products of subsequent reactions. The resulting accretion products are diverse, as is known for similar chemical systems, and were not quantified directly with 1H-NMR. Therefore, to include these reactions in the kinetic mechanism, the accretion product (AP) concentration is approximated with butenedial-pyrrolinone dimer as a proxy: [AP] = [BD-PR]. Setting the AP concentration equal to BD-PR concentration involves several assumptions, namely that the number of AP reactive sites scales 245 with BD-PR concentration, the molecular weight distribution of AP members is independent of pH, and any reversibility in accretion reactions can be accounted for with this approximation. However, strong agreement is still observed between butenedial observations and model predictions under different pH and initial reactant conditions, which suggests that this approximation does not significantly affect mechanistic portrayal of butenedial reactivity.

[BD] = − 1 [BD] − 2 [BD][NH 3 ] − 3 [BD][PR][OH -] − 2 4 [BD][AP]
(DE1) The butenedial/NHX kinetic mechanism contains three differential equations (DE1-DE3), one per explicitly measured 255 species: butenedial (BD), pyrrolinone (PR), and butenedial-pyrrolinone dimer (BD-PR). The finalized mechanism contains the best fit reaction rate constants to experimental data; the resulting model output and experimental data are shown in Figure 3. artificially consumes NHX when butenedial dimerizes with pyrrolinone or accretes, and NHX loss could be overestimated. To assess the sensitivity of the parametrization to NHX, the model fitting was also performed assuming NHX is not consumed during reaction (i.e., zero NHX is lost per burtenedial). The parameter fitting with this scenario provides a maximum deviation 265 as it is effectively the opposite extreme. The differences between these produced parameters and those of the original fitting were small, typically falling <5% of the parameter estimates with one exception ( Figure S14). The employed simplification of NHX is therefore assumed to have minor effects on the kinetic mechanism.
Model predictions using this mechanism compare well with measurements from two additional experiments with different pH and initial conditions (Figure S15-S16). One was performed with 0.4 M BD0, 0.4 M NHX,0, and pH 3.6 and the 270 other with 0.9 M BD0, 0.2 M NHX,0, and pH 8.5-8.8. This indicates that the kinetic mechanism is robust across a relevant range of pH and initial conditions.

Comparison of butenedial loss processes in aqueous aerosols
The lifetime of condensed-phase butenedial from aqueous reaction with OHor NHX, with AS as NHX source, is compared to that of wet deposition, which for tropospheric aqueous particles is about one week (Seinfeld and Pandis, 2016). The dominant 275 first order loss process of butenedial is shown as a function of NHX and pH in Figure 4. Butenedial/OHis the main sink if < 1 mM NHX and above ~pH 7, although this pH is not particularly atmospherically relevant. Reaction with NHX can be fast at typical NHX, even under somewhat acidic conditions, and is therefore competitive with wet deposition. Butenedial loss is increased through accretion reactions in the NHX pathway but this effect is not included in this analysis. Thus, the figure represents a lower limit for butenedial loss via the NHX pathway.

Comparison of dicarbonyl reactivity in aqueous NHX solutions
The reactivity of dicarbonyls in aqueous NHX solutions was previously understood primarily on the basis of the dominant functionality through studies of -dicarbonyls. Following the work of Kampf et al. (2016), this study provides an additional perspective that considers the role of the carbon skeleton on dicarbonyl behavior. Biacetyl is the least hydratable of the -285 dicarbonyls, and because it is a diketone, is the least similar to butenedial. Methylglyoxal is less hydrated than glyoxal and butenedial because it is a ketoaldehyde that has a hydrophobic methyl moiety. Butenedial is an electron-poor dialdehyde like glyoxal and is therefore strongly hydrated, but it also has a hydrophobic alkene group. In addition, it has a four-carbon chain, making the formation of stable five-membered organic molecules thermodynamically and especially kinetically favorable in comparison to glyoxal, methylglyoxal, and biacetyl, which provides evidence for the importance of the carbon backbone to 290 dicarbonyl chemistry.
The dicarbonyl moiety leads to several similarities in the behavior of butenedial, glyoxal, methylglyoxal, and, to a lesser degree, biacetyl in aqueous NHX solutions: all hydrate reversibly, react with OHunder basic conditions, and react with NHX to produce heterocycles and subsequently undergo accretion reactions. Brown carbon is produced in solution even if minimal reaction has occurred and accelerates with increasing pH. However, we demonstrate three important differences 295 between butenedial and glyoxal in particular that showcase the variability possible between dicarbonyls and even electronpoor dialdehydes.
First, unlike butenedial, glyoxal shows a strong tendency to form acetal oligomers in pure aqueous solutions. Biacetyl and methylgloyxal can also form acetal oligomers in aqueous solutions, although aldol condensation products are more common (Grace et al., 2020;Sareen et al., 2010). Somewhat surprisingly, butenedial acetal oligomers are much less 300 pronounced, despite butenedial having two reactive aldehyde groups and being predominantly dihydrated in aqueous solution.
This behavior was reported by Birdsall et al. (2019) who did not find any acetal products from butenedial itself or, perhaps even more surprisingly, in the presence of high concentrations of polyethylene glycol. One explanation could be that butenedial, like methylglyoxal and biacetyl, has a substantial hydrophobic component that influences the ability of its hydrates to oligomerize like glyoxal, which was not expected based on the similarity with glyoxal in hydration behavior. 305 It is also observed that the double bond within the carbon backbone affects the properties of the products. Polymers of lactic acid and glycolic acid (the products of methylglyoxal and glyoxal reaction with OH -) are colorless, presumably because they lack π-conjugated double bonds. Accretion products from butenedial/OHreaction on the other hand efficiently absorb light even at relatively low quantities (browning was observed immediately after introducing butenedial to basic conditions). The light absorptivity of the products can be attributed to the alkene bond they inherit from butenedial and 310 potentially the presence of carbonyls. Therefore, it is suggested that other unsaturated dicarbonyls could lead to the production of light-absorbing compounds, although unsaturated compounds and basic pH are not expected to be relevant in atmospheric aerosols. The third important difference is the rate and rate law for the dominant products of the dicarbonyl/NHX reactions. We do not suggest that NH4 + is a catalyst (as a source of Bronsted acid) for butenedial/NH3 reaction, which has been proposed by 315 previous studies for glyoxal (Nozière et al., 2009). While imidazole production is second order in glyoxal and NH3 and explicitly pH dependent, rendering it inefficient at low ambient concentrations (Yu et al., 2011), pyrrolinone production is linearly dependent on butenedial and NH3, which is similar to methylglyoxal (Sareen et al., 2010) and biacetyl (Grace et al., 2020). This may not be surprising because in contrast to the two-carbon glyoxal, reaction between one butenedial and one NH3 already results in a stable heterocycle. We suggest that dicarbonyls with a favorable separation of reactive aldehyde groups 320 can form heterocycles with bimolecular rate laws, which means they can occur even if ambient concentrations are low. They could include four-carbon dicarbonyls (e.g., succinaldehyde, 4-oxopentanal, other unsaturated or saturated 1,4-dicarbonyls) and phthalaldehyde, which can form five-membered rings, and five-carbon dicarbonyls (e.g., glutaraldehyde) that are capable of forming pyridines and other six-membered heterocycles, as shown by Kampf et al. (2016). Electron poor dialdehydes with longer carbon backbones (six-carbon or more) may also be able to produce stable products from self reactions. Notably, 325 butenedial reaction with NH3 is much faster than for glyoxal, methylglyoxal, and biacetyl. Yu et al. (2011) showed that only 12% of the initial glyoxal in a 1 M glyoxal/1 M AS solution had been consumed over 5.5 months, whereas at comparable pH we observed 11% removal of butenedial in a 0.4 M butenedial/0.2 M AS solution after 8 hours. Similarly, the lifetime of methylglyoxal with respect to reaction with NH4 + in a 14 M AS solution is 29.8 hours (Sareen et al., 2010), however we calculate a corresponding lifetime of 4.2 hours at pH 4, and 2.5 minutes at pH 6 for butenedial. The fast rate of butenedial/NHX 330 reaction supports other work that has shown more rapid brown carbon formation from larger dicarbonyls than for -dicarbonyls (Kampf et al., 2016).
In sum, this work complements a previous study (Birdsall et al., 2019) to show that it is difficult to extrapolate the physicochemical properties of reactive dicarbonyls from their -dicarbonyl prototypes. Not only the reactive functional groups affect reactivity but the structure of the carbon backbone as well. In the case of glyoxal, it is likely that its vicinal two aldehydes 335 cause its chemical and physical properties to be unique and dissimilar to the rest of the dialdehydes/dicarbonyls. Further studies should be conducted with other more complex dicarbonyls to elucidate patterns in chemical behavior that are related to the carbon skeleton.

Atmospheric implications
This work shows that the aqueous reaction of butenedial with and without NHX can form low-volatility chromophores that will 340 be retained in the condensed phase and absorb radiation. The results, especially the rapid reaction of butenedial with NHX, show that similarly reactive dicarbonyls could impact chemical composition and optical properties of particles, and thus directly influence the human health and climate impacts of particles.
Butenedial was recently shown to have a gas-phase photochemical lifetime of 10-15 minutes due to photolysis (Newland et al., 2019)  contribute significantly. It is, however, likely that condensed phase reaction of butenedial with NHX could regionally be important, specifically at close to neutral pH and high NHX, such as in agricultural areas in India where NH3 emissions are high (Kuttippurath et al., 2020) and rainwater is observed to be alkaline (Kulshrestha et al., 2001). At pH 6 and 4 M NHX, the lifetime of butenedial against reaction with NHX is only 18 minutes, even excluding enhancements from reaction with 350 pyrrolinone and other accretion reactions or, in the atmosphere, other reactive organics. The results show that accretion reactions, and therefore the accumulation of chromophores, increase strongly with pH. Dimers and accretion products correspond to the vast majority of products (and quickly pull more butenedial out of solution) at slightly acidic or neutral pH.
An important future research step is to refine the relationship between pH and oligomerization/accretion reaction rates.
We do not propose that butenedial alone contributes significant brown carbon. However, the reactivity of butenedial 355 can be extrapolated to dicarbonyls for which condensed phase chemistry has not been studied. Such dicarbonyls are rarely measured but could be abundant. For example, 4-oxopentanal (a saturated 1,4-ketoaldehyde) was recorded at particulate concentrations averaging 62.7 ng m -3 over a Japanese forest (Matsunaga et al., 2004); for a typical aerosol liquid water content of 1-10 mg m -3 , this corresponds to particulate 4-oxopentanal concentrations of approximately 6-60 mM, which was similar to that of glyoxal and methylglyoxal. If 4-oxopentanal is representative of a range of dicarbonyls that react with NHX like 360 butenedial, then the reaction could be fast enough that the sum of dicarbonyls may constitute a regional source of brown carbon in regions with high NHX and alkalinity. The vast majority of studies of the condensed-phase atmospheric chemistry of dicarbonyls have focused on glyoxal, methylglyoxal, and biacetyl, in part because they are abundant and commercially available. The fact that butenedial has much faster reaction rates of forming brown carbon that are first order indicates that additional studies of larger dicarbonyls with hydrophobic moieties are needed, especially to further evaluate the role of 365 dicarbonyls in the formation of brown carbon.
Code and data availability. The Python package pyrosolchem used as the kinetic model of droplet evaporation is available at https://github.com/jackattack1415/pyrosolchem (last access: 23 February 2021). Data used to generate paper figures are available upon request.   535 Table 2: Butenedial/OHreaction, kinetic expression of the rate law, and corresponding estimates of the coefficients in the rate constant and their standard errors. aI and aIII are well constrained. Although aII has a large standard error, it appears to not severely impact agreement between model and measurement (Figure 3). See Equation 1 for the expression of k1 in terms of its coefficients, aI, aII, and aIII, and OH -.

555
NHX was not measured explicitly but was assumed to be consumed at a 1:1 ratio with butenedial. pH was estimated empirically outside of the model fit. Only the best model fit of pH was taken as input into the kinetic mechanism, although the 95% confidence interval is reported for the empirical fitting of pH.

Figure 4:
The dominant butenedial loss pathway, reaction with OH -(blue) or NHX (red) and wet deposition (green), is shown as a function of pH and NHX. AS is the NHX source. Loss via wet deposition is considered to have a one-week lifetime, typical of atmospheric particles.

560
The range of pH (3-6) and NHX concentration (<28 M) relevant to the atmosphere is overlaid on the plot, as well as the NHX concentrations at which phase separation in the mixtures was observed and the AS solubility limit.