Aqueous phase behavior of glyoxal and methylglyoxal observed with carbon and oxygen K-edge X-ray absorption spectroscopy.

. Glyoxal ( CHOCHO ) and methylglyoxal ( CH 3 C(O)CHO ) are well-known components of atmospheric particles and their properties can impact atmospheric chemistry and cloud formation. To get information on their hydration states in aqueous solutions and how they are affected by addition of inorganic salts (sodium chloride ( NaCl ) and sodium sulfate ( Na 2 SO 4 )), we applied carbon and oxygen K-edge X-ray absorption spectroscopy (XAS) in transmission mode. The recorded C K-edge spectra show that glyoxal is completely hydrated in the dilute aqueous solutions, in line with previous studies. For methylglyoxal, 5 we identiﬁed, supported by quantum chemical calculations, not only C-H, C=O and C-OH bonds, but also ﬁngerprints of C-OH(CH 2 ) and C=C bonds. This implies the presence of both mono- and dihydrated forms of methylglyoxal, as well as products of aldol condensation, and enol tautomers of the monohydrates. The addition of salts was found to introduce only very minor changes to absorption energies and relative intensities of the observed absorption features, indicating that the organic-inorganic interactions at the studied concentrations are not strong enough to affect the spectra in this work. The identiﬁed structures of 10 glyoxal and monohydrated and dihydrated forms. We also calculated energies of enol forms of methylglyoxal and hydrated methylglyoxal because they have been identiﬁed as possible structures in aqueous methylglyoxal solutions by Krizner et al. (2009). They applied DFT calculations to study the hydration of 1 M of methylglyoxal at 298 K and showed that formation of enol structures and aldol condensation products, which have in their chemical structure a C=C bond, are the thermodynamically most stable species in the methylglyoxal system. condensation reactions. and dihydrated forms of the compound but also other chemical species with enol structures. The results contribute to explain the relatively high contribution of glyoxal and methylglyoxal to atmospheric SOA, despite their high saturation vapor pressures. Author contributions. NLP conceived, planned and supervised the project and secured beamtime and funding. GM, HY, MP, NLP and MH carried out the experiments on liquid samples. MN performed the gas-phase measurements of glyoxal and methylglyoxal. NK performed the quantum chemical calculations. GM analyzed the experimental data with supervision and assistance of MP. All the authors contributed to 290 discussions on the results. GM wrote the paper with assistance of JJL and contributions from all co-authors. Competing interests. authors declare they have no conﬂict of interest.

monohydrate formation reaction above 1000. Upon hydration, glyoxal and methylglyoxal can also react with themselves to form dimers and oligomers (Ervens and Volkamer, 2010;Liggo et al., 2005;Kroll et al., 2005;Wang et al., 2010;Galloway et al., 2009). Both the hydrates and dimers/oligomers are covalently bound species, not hydrogen-bonded clusters.  Figure 1. Formation of mono-and dihydrated glyoxal (first row) and methylglyoxal (second row). The values of equilibrium constants, K hyd1 and K hyd2 , are shown in Table 1.
In this work, we have used synchrotron radiation (SR) based soft X-ray absorption spectroscopy (XAS) combined with quantum chemical calculations, to study hydrate formation of glyoxal and methylglyoxal in aqueous solutions, and their aqueous 45 2 https://doi.org/10.5194/acp-2020-393 Preprint. Discussion started: 26 May 2020 c Author(s) 2020. CC BY 4.0 License.
phase interactions with various salts commonly found in atmospheric aerosol. The high brightness and element selectivity, in terms of core level excitation energy and energy resolution makes SR a convenient tool to study molecular level interactions in aqueous phase using absorption spectroscopy.

Sample preparation 50
Aqueous solutions of each of the organic compounds, glyoxal, methylglyoxal and glycerol (C 3 H 8 O 3 ), were prepared freshly before each experiment. Glycerol solution is used for comparison with the glyoxal and methylglyoxal solutions because it has similar structure to their hydrated forms. The aqueous solutions were prepared using water purified by Sartorius arium pro UV system (18.2 MΩ). Methylglyoxal was purchased from Sigma-Aldrich and all the other chemicals from Wako. All chemicals were used without further treatment. Glyoxal and methylglyoxal were delivered as 40 wt% aqueous solutions, while glycerol 55 was a viscous liquid of 99.5% purity. The organic concentrations of the prepared aqueous solutions ranged from about 0.1 to 2 M. Ternary solutions of the organic compounds, water and inorganic salt (sodium chloride (NaCl, 99.5%) or sodium sulfate (Na 2 SO 4 , 99.0%)), and an aqueous Na 2 SO 4 solution were also prepared. Salts were added to yield inorganic concentrations between 1 to 2 M, and different mixing ratios with respect to the organics. All samples were well-mixed before use, yielding homogeneous aqueous solutions. Details of the studied compounds are given in Table 1, and of the samples in Table 2.

Liquid cell experiments
XAS measurements of liquid samples were performed at the BL3U beamline (Hatsui et al., 2004) of the UVSOR-III Synchrotron Facility, Okazaki, Japan. The 750 MeV storage ring was operated in a top-up mode with 300 mA current. The end station with the liquid flow cell system connected to the beamline has been described extensively by Nagasaka et al. (2018a).
Briefly, the liquid cell consists of two Si 3 N 4 membranes (area: 2 × 2 mm 2 , thickness: 100 nm) in a chamber which is sepa-65 rated from the soft X-ray beamline under ultrahigh vacuum conditions (< 10 −5 Pa) by using one Si 3 N 4 membrane (0.2 × 0.2 mm 2 ). The chamber is filled with flowing helium (He) at atmospheric pressure which can be varied by using outlet valve of He flow line. The liquid sample thickness was optimized for better transmission signal by changing the helium flow rate, and, therefore, the pressure exerted on the walls of the cell. The liquid sample in the cell can be exchanged using a pumping system (Cole-Parmer Masterflex L/S) with adjustable flow rate. Here, the samples were pumped in the liquid cell with the flow rate of 70 5 mL min −1 . XAS spectra in transmission mode can be obtained based on the Beer-Lambert formula: I = I 0 e −µx , where I 0 is the intensity of the incident radiation, I is the transmitted intensity and x is the thickness of the sample. The absorption intensity depends on the material linear attenuation coefficient µ (Stöhr, 1992). During each measurement the incident radiation was monitored with a gold mesh placed before the liquid cell, so that the flux variations (<1%) due to the top-up mode were removed.

75
The transmitted radiation I is detected by a photodiode (IRD AXUV100) placed right after the liquid cell. The transmission signal was also measured without the sample, using a blank cell, to normalize the absorption of the membranes and possible contamination. The thickness of the liquid layer (x) was not precisely estimated. In order to avoid additional uncertainty on our results, the intensities of the XA spectra are given in arbitrary units. Here, we concentrate on the shape of the spectrum and the energies of resonances rather than absolute absorption coefficients.

80
Carbon and oxygen K-edge (1s core level) absorption spectra are recorded by scanning photon energies between 282-305 eV and 525-550 eV, respectively. The energy range can be changed from the control system of the beamline, by changing the angle of a varied-line-spacing plane grating monochromator. Photon energy resolutions were approximately 0.2 eV for C and 0.4 eV for O K-edges, respectively. Reproducibility of the recorded spectra was verified by recording a few low statistics cases several times. No significant changes were observed in the peak positions and the intensities of our spectra between the scans.

85
During the experiment, we regularly recorded carbon and oxygen K-edge absorption spectra of pure water. The spectra were used for inspection of possible contamination in the system and as a reference for the characterization of the spectra of the aqueous solutions.

Data treatment
The C K-edge spectra of the samples were subtracted by the C K-edge spectra of pure water to remove any undesired absorp-90 tions arising from e.g. the absorption of water in the solution, that of Si 3 N 4 membranes and that of carbon contamination in the beamline optics.
For the calibration of the C K-edge XAS spectra, we used XAS spectra of polymer films (proLINE) (Nagasaka et al., 2018a) scanned for energies from 284.5 eV to 288.5 eV with steps of 0.02 eV and 0.05 eV. After interpolation of the energy values with step of the data points approximately 0.05 eV, the treated data were calibrated by the absorption peak of the polymer films 95 at 285.07 eV (Nagasaka et al., 2018a).
In the C K-edge spectra of glyoxal, there were no pre-edge absorption peaks and the ionization edge was identified by fitting a sigmoid in the step of the spectra and by identifying the mean energy value from the inflection point. For the fitting of the spectra of methylglyoxal, we used the SPANCF curve fitting macro package (Kukk et al., 2001(Kukk et al., , 2005 which gives the energy and intensity of the absorption peaks. The shape of the curves was an asymmetric Voigt shape and the ionization thresholds 100 were modeled by an arctangent function. The oxygen signal of the O K-edge spectra is overwhelmed by the signal from the water. Therefore, the signal of water was not subtracted from the O K-edge spectra, as in the case of the C K-edge spectra. The data were calibrated by the first peak of the polymer films at 530.88 eV (Nagasaka et al., 2018a). For comparison, we also measured the oxygen K-edge of glycerol hydrated forms of glyoxal and methylglyoxal (three hydroxyl groups bonded to three carbons and single C-H bonds), while excluding the possibility for reversible water addition or removal.

Quantum chemical calculations
The core ionization and excitation energies were evaluated within the ∆SCF method by using the GSCF3 code (Kosugi and Kuroda, 1980). Geometries of free glyoxal, methylglyoxal, hydroxides, and enol forms were optimized by ground-state SCF 110 calculations within the second order Møller-Plesset perturbation theory using 6-311G* basis set (Frisch et al., 2004). According to our established method (Kosugi et al., 1992), core ionized states of these molecules were obtained by SCF calculations of their single core hole states with an extended basis set plus polarization basis functions (31111/21/3112/1*) for carbon and oxygen atoms and a double zeta basis set (42) for hydrogen (Huzinaga et al., 1984), and then by freezing the resultant core hole orbitals, singlet core excited states of these molecules were obtained by SCF calculations of their single core-to-valence 115 excitations. No diffuse functions to describe Rydberg states are included assuming there is no strong Rydberg-valence mixing and Rydberg states are negligible.

Results and discussion
Here we present results in terms of recorded XAS spectra for each sample of glyoxal, methylglyoxal and their mixtures with inorganic salts. The carbon K-edge spectra are presented together with calculations for unhydrated and hydrated forms of 120 glyoxal and methylglyoxal for better interpretation of the spectra. We also present oxygen K-edge spectra of aqueous solutions with glycerol and their ternary solutions containing inorganic salts.

Quantum chemical calculations
We calculated ionization potentials (IP) and C 1s-π* bound-state excitation energies for unhydrated and hydrated species of 125 glyoxal and methylglyoxal in gas phase. Quantum chemical calculations are extremely difficult to be accurate for complex systems (e.g. liquids) due to computational limitations. Gas phase calculations can be used instead to describe the electronic excitations. Typically, the core-to-valence excitation energies of molecules in gas phase show only small shifts compared to the excitation energies of these molecules in liquid phase or as a part of a cluster. However, the IP of the molecule is higher in the gas phase compared to condensed phases. The Rydberg series converging to the ionization threshold are shifted in 130 higher energies for the lower Rydberg and shifted to lower energies or quenched for the higher Rydberg by the surroundings (Flesch et al., 2004;Nagasaka et al., 2018b). The results of the quantum chemical calculations are presented in Table 3. In the case of glyoxal, we observe that the excitation energies of C=O bonds of the unhydrated and monohydrated form differ by approximately 0.5 eV. The IP of the C=O sites (295.4 for the unhydrated form and 295.15 eV for the monohydrated form) are higher than these of the C-OH sites by 0.54 and 0.29 eV, respectively.
We also calculated energies of enol forms of methylglyoxal and hydrated methylglyoxal because they have been identified as possible structures in aqueous methylglyoxal solutions by Krizner et al. (2009). They applied DFT calculations to study the hydration of 1 M of methylglyoxal at 298 K and showed that formation of enol structures and aldol condensation products, which have in their chemical structure a C=C bond, are the thermodynamically most stable species in the methylglyoxal system.

140
This is a very favorable chemical path when methylglyoxal is hydrated, and thus taken into account in our calculations as well.  In Table 3 we also present the calculated oscillator strengths (f ).

Glyoxal
150 Figure 3 presents the recorded spectra for the C 1s absorption edge of binary glyoxal-water and ternary solutions formed by addition of either NaCl or Na 2 SO 4 together with the excitation energies of our quantum chemical calculations. The results of the quantum calculations are presented with colored lines that correspond to hydrated and unhydrated molecules in gas phase, and the length of the lines is scaled in respect with the calculated oscillator strengths. Apart from the absorption probability of a molecule, the absorption intensity of the presented spectra is also related to the relative amount of the species in the solution 155 and the interactions that take place in the liquid phase, and thus we cannot directly use oscillator strengths as expected intensity ratios in the experimental spectra.
The spectra with and without addition of inorganic salts have only a single feature at around 289.6 eV. The absorption edges of the recorded spectra are presented in Table 4. The absorption energies of different solutions differ by 0.1-0.2 eV, which is just within the photon energy resolution of the measured spectra. We observe that there is a trend to increase the 160 absorption edge energy upon addition of salt or decrease in concentration. Based on the calculated energies (Table 3) glyoxal with Nuclear Magnetic Resonance (NMR) spectroscopy. In agreement to our study, they did not detect unhydrated glyoxal and concluded that the glyoxal in the solution is mainly in dihydrated form and contains oligomers. Kua et al. (2008) and Yu et al. (2011) identified potential oligomers as hydration products of aqueous glyoxal. Therefore, oligomers of glyoxal that contain C-OH bonds may be present in our solutions. 175

Methylglyoxal
In Fig. 4 we present the spectra of C 1s of methylglyoxal in aqueous mixtures. We observe three distinct absorption features at 285.8, 287.9 and approximately 290 eV. Figure 5 shows five peaks, denoted as A, B, C 1 , C 2 and C 3 , that were identified during the fitting process of the spectum of methylglyoxal in aqueous phase. We correlate these peaks with the absorption energies of functional groups presented in our solutions (Table 5). We also present the excitation energies and ionization potentials of our 180 quantum chemical calculations.

Peak A
The first peak designated as A (Fig. 5) is found at 285.8 eV. We assign this peak to the C=C-OH moiety of the enol form of the monohydrated methylglyoxal, in good agreement with the predicted C=C excitation energy of 285.91 eV. Usually, the C=C peaks are found at 285 eV, but increase in the energy of C=C transitions has been observed before in organic compounds with 185 similar structure. Phenols and quinones in solid phase were studied by Solomon et al. (2009), using Scanning Transmission X-ray Microscopy (STXM) under a He atmosphere. These compounds have the same characteristic moiety, and excitations between 286.05 and 286.35 eV were attributed to C 1s → π * C=C . C=C bonds can also result from radiation damage of the sample. However, consecutive scans of the solutions did not show any significant changes in the spectra, and thus radiation damage is considered to be small in our experiments.

Peak B
The next strong peak in Fig. 4 at 287.9 eV (peak B in Fig. 5) can be assigned to the C-OH(CH 2 ) transitions in the enol form of monohydrated methyglyoxal. The calculated energy of this transition is at 287.53 eV (lower by 0.4 eV).

Peaks C 1 -C 3
The asymmetrical shape of our spectra suggests underlying peaks. We fit three peaks, according to the observed shoulders on the two main peaks, at 286.5 eV (peak C 1 ), 287.1 eV (peak C 2 ) and 288.6 eV (peak C 3 ). The predicted C=O absorption energies are at 286.51 eV and 287.05 eV for the unhydrated and monohydrated methylglyoxal in gas phase, respectively. Latham et al. (2017) studied solid organic compounds in ultrahigh vacuum conditions (UHV), applying near-edge X-ray absorption fine structure (NEXAFS) and has also assigned the C=O transitions at 286.5 eV. Russell et al. (2002) applied soft X-ray spectromicroscopy at atmospheric pressure and Tivanski et al. (2007)  to 288.4 eV by Russell et al. (2002). In conclusion, in this energy range we expect peaks related to C=O excitations associated to the unhydrated methylglyoxal, C=O(CH 3 ) excitations of the monohydrated methylglyoxal and also C-H excitations of the hydrated species. Based on previous studies, we expect to have mainly monohydrated and dihydrated forms of the compound and a negligible fraction of unhydrated methylglyoxal (Nemet et al., 2004;Krizner et al., 2009). C=O and C-H excitations 210 from hydrated species that contain enol structures and originate from monohydrated methylglyoxal should also be considered as they can co-exist in the solutions. Due to the variety of possible species containing C=O and C-H bonds in the aqueous solutions, we suggest an energy range from 286.5 to 288.6 eV for the transitions of C=O and C-H groups.

Feature at 290 eV
The wide feature at around 290 eV in Figs. 4 and 5 is related to C-OH transitions and ionization of C 1s electrons. The 215 calculated C-OH absorption energy of the monohydrated form at 290.27 eV matches well with the energy from where the broad structure starts. The IP are expected to be shifted by some eVs compared to the calculated, depending on the surroundings where nearest neighbor is dominant.
In the case of methylglyoxal, addition of salts did not introduce observable shifts to absorption energies. Due to the presence of overlapping peaks in the spectra of the salt-free systems, we cannot estimate the absolute concentrations of unhydrated 220 and hydrated methylglyoxal and thus, we cannot observe how the addition of salts affect the concentration of the hydrates.
Changes in the relative ratios of unhydrated and hydrated forms would be expected due to changes in the molecular interactions (water-methylglyoxal, water-water) when ions co-exist in the solution. Such phenomena have been reported in previous experimental (Waxman et al., 2015) and computational (Toivola et al., 2017) studies, which have shown decrease of the cosolubility of methylglyoxal in aqueous solutions with addition of NaCl and Na 2 SO 4 (salting out effect). Here we observe a 225 small change in the relative intensities between the peaks A and B in pure methylglyoxal solution compared to solution spiked with Na 2 SO 4 , but the shape of the background also changed in the spectrum and thus these changes remain inconclusive.

Oxygen K-edge
Figures 6, 7 and 8 show the O K-edge absorption spectra of organic-water binary solutions and comparison of these when Na 2 SO 4 or NaCl is added. The spectra have been normalized to the background preceding the water pre-peak. The two 230 structures can be characterized as a pre-peak and a main peak.
As mentioned in Sect. 2.3, in the case of oxygen, the water background has not been subtracted. Due to the abundance of water in the solutions, the spectra are similar to that of water. The pre-edge peak at 535 eV is related to the transition to an orbital with the 4a 1 character in a water molecule, and the main edge is related to the transition to the continuum threshold of bulk water (Sellberg et al., 2014;Pylkkänen et al., 2010). Changes of the pre-edge peak can be related to changes in the 235 hydrogen bond network (Nilsson et al., 2010;Velasco-Velez et al., 2014;Bluhm et al., 2002). We observe that for all the organic compounds (Figs. 6,7 and 8), the ratio of the pre-peak feature to main peak is smaller than in water. We do not observe any relative changes in spectral features as a function of concentration of the organic compound. In ternary solutions with Na 2 SO 4 the ratio of the main edge region (537-540 eV) to post edge (540-544 eV) differs from all the other solutions. This has been previously described by Niskanen et al. (2015) to be related to tetrahedral geometry of the system. In solutions with Na 2 SO 4 240 we also observe that the pre-edge peaks are blue-shifted similarly to the Na 2 SO 4 aqueous solution. The energy difference between the pre-edge peaks of pure water and the ternary solutions with Na 2 SO 4 is approximately 0.1 eV for the solutions of glyoxal containing 2 M of Na 2 SO 4 and 0.05 eV for the solution of methylglyoxal containing 1 M of Na 2 SO 4 . This shift is plausible considering that Na + ions can have strong interactions with oxygen of water (Nagasaka et al., 2015(Nagasaka et al., , 2017 and that SO 2− 4 tends to form hydrogen bonds with water molecules (Niskanen et al., 2015).
245 Figure 8 presents O K-edge spectra of 2 M aqueous glycerol and its mixture with NaCl and Na 2 SO 4 . Similarly to glyoxal and methylglyoxal spectra, the shape of the pure solution and the one spiked with NaCl are very close to pure water. Again, when Na 2 SO 4 is added, the prepeak shifts to higher energies (by 0.1 eV) and the ratio of the main and post edge features changes, thus getting closer to the shape of pure Na 2 SO 4 solution.  For comparison we present also the O K-edge spectrum of pure water and of aqueous solution with 2 M of Na2SO4. Right: Zoom on the pre-peak of the spectrum.
In conclusion, O K-edge spectra were found to be sensitive neither to the organic component at studied concentrations nor to 250 addition of NaCl. However, addition of Na 2 SO 4 affected the spectra as they started to resemble more pure Na 2 SO 4 solution without any organics. Thus, no Na 2 SO 4 -organic interaction can be confirmed using this method.

Atmospheric implications
Both glyoxal and methylglyoxal are volatile organic compounds (Kielhorn et al., 2004;Liggio et al., 2005;Kalberer et al., 2004) and thus expected to be significantly present in the gas phase in the atmosphere. Nevertheless, both contribute significantly to 255 atmospheric SOA (Volkamer et al., 2006;Fu et al., 2008). Here, we confirm experimentally that glyoxal, to the sensitivity of our measurements, exists entirely in its fully hydrated form in aqueous solution, in agreement with previous studies (Yu et al., 2011;Malik and Joens, 2000;Kua et al., 2008). Our study shows that hydrated methylglyoxal furthermore takes part in enol formation and aldol condensation reactions which contributes further to shifting the gas-particle phase partitioning towards the aqueous particle phase. Both reactions can therefore lead to significant SOA formation.

260
The presence of organic solute, such as glyoxal or methylglyoxal, in the aqueous solution decreases the equilibrium water vapor pressure p over the solution compared to the equilibrium vapor pressure of a water solution (solute effect). According to Köhler theory (Köhler, 1936), increased uptake of SOA to the aqueous phase promotes hygroscopic growth, as well as  and 1 M of NaCl. We also present the recorded spectra of pure water and binary water-2 M Na2SO4 solution. Right: Zoom on the pre-peak of the spectrum. atmospheric cloud formation by lowering the humidity threshold for droplet activation (e.g., Prisle et al., 2010;Hansen et al., 2015).

Conclusions
We applied carbon and oxygen K-edge XAS using synchrotron radiation to study aqueous solutions of atmospheric highly oxygenated organics, glyoxal and methylglyoxal. We recorded absorption spectra for the carbon and oxygen K-edges and we measured as reference also oxygen K-edge spectrum of aqueous glycerol.
Glyoxal was found to be fully hydrated in aqueous solutions, as no peaks related to transitions in C=O group were identi-270 fied in our spectra. C 1s XAS of methylglyoxal had two well-defined peaks before the broad absorption continuum, and were assigned using quantum chemical calculations to C=C and C-OH(CH 2 ) moieties of the enol form of monohydrated methylglyoxal. Transitions in C=O and C-H functional groups contributed to the C 1s XAS as broad shoulders in the energy range from 286.5-288.6 eV. They are found in the monohydrated and dihydrated forms of the compound and also from products of keto-enol tautomerism followed by aldol condensation reactions.  The oxygen K-edge spectra were similar to water, due to low solute concentrations and we only observed small changes in the case of mixtures with Na 2 SO 4 : a small shift of the pre-peak to higher absorption energies and increase of intensity of the first feature after the main absorption edge at 538 eV due to strong interactions of Na + and SO 2− 4 ions with the atoms of water. In both carbon and oxygen K-edge spectra, we did not observe any effects of salting in/out interactions on the shape of the spectra. This indicates that XAS might not be sensitive enough to see additional changes to the hydrate formation equilibrium, 280 as e.g. demonstrated by Kurtén et al. (2014). As a rough estimate of the sensitivity, even at the highest 2 M concentration, we were not able to identify any unhydrated form of glyoxal, even if 2% of it has been reported before (Malik and Joens, 2000).
However, our finding that glyoxal exists entirely on covalently bond dihydrate form in aqueous solution is in line with the results of Yu et al. (2011).
Our study presents the first experimental verification that methylglyoxal in water solution contains not only monohydrated 285 and dihydrated forms of the compound but also other chemical species with enol structures. The results contribute to explain the relatively high contribution of glyoxal and methylglyoxal to atmospheric SOA, despite their high saturation vapor pressures. discussions on the results. GM wrote the paper with assistance of JJL and contributions from all co-authors.
Competing interests. The authors declare that they have no conflict of interest.