Aqueous phase oligomerization of methyl vinyl 1 ketone through photooxidation 2 Part 1: Aging processes of oligomers 3

11 It has recently been established that unsaturated water soluble organic compounds (UWSOC) 12 might efficiently form oligomers in polluted fogs and wet aerosol particles, even for weakly 13 soluble ones like methyl vinyl ketone (MVK). The atmospheric relevance of these processes 14 is explored by means of multiphase process model studies in a companion paper. In the 15 present study, we investigate the aging of these aqueous phase MVK-oligomers formed via 16 • OH-oxidation, as well as their ability to form secondary organic aerosol (SOA) upon water 17 evaporation. The comparison between aqueous phase composition and aerosol composition 18 after nebulization of the corresponding solutions shows similar trends for oligomer formation 19 and aging. The measurements reveal that oligomer aging leads to the formation of organic 20 diacids. Quantification of the SOA mass formed after nebulization is performed, and the 21 obtained SOA mass yields seem to depend on the spectral irradiance of the light used to 22 initiate the photochemistry. Investigating a large range of initial MVK concentrations (0.2 – 23 20 mM), the results show that its • OH-oxidation undergoes competition between 24 functionalization and oligomerization that is dependent on the precursor concentration. At 25 high initial MVK concentrations (≥ 2mM), oligomerization prevails over functionalization,


Introduction
1 Organic aerosol plays an important role in many atmospheric processes and has an important 2 impact on climate and human health. Globally, about 20 % of the organic aerosol mass is 3 emitted directly (Kanakidou et al., 2005;Spracklen et al., 2011), which conversely indicates 4 the relevance of aerosol formed by transformation of organic gas phase species, i.e. secondary 5 organic aerosol (SOA). The most commonly studied mechanism of SOA formation is the 6 oxidation of volatile organic compounds (VOC), which can lead to the formation of less 7 volatile species that subsequently partition into the condensed phase (Donahue et al., 2011;8 Kanakidou et al., 2005;Kroll and Seinfeld, 2008;Hallquist et al., 2009). Nevertheless, the 9 oxidation of VOCs also results in more water soluble products that readily partition into the 10 aqueous phase (Blando and Turpin, 2000;Ervens et al., 2011;Epstein et al., 2013). Due to 11 further reactivity in the liquid phase, higher molecular weight and less volatile compounds 12 can be formed, which can remain at least in part in the condensed phase upon water 13 evaporation, thus leading to additional secondary organic aerosol formation through aqueous 14 phase reactions (aqSOA) (El Haddad et al., 2009;Carlton et al., 2009;Ervens et al., 2011;15 Ortiz-Montalvo et al., 2012). In particular, Lee et al. (2012) observed a significant 16 enhancement of organic mass during the initial stage of oxidation of cloud water organics, 17 that they explained by functionalizing dissolved volatile organics via hydroxyl radical ( • OH) 18 oxidation. Aqueous phase processes can be very different from those in the gas phase, thus 19 leading to aqSOA with likely very different physical and chemical properties (Ervens et al., 20 2011;Ortiz-Montalvo et al., 2012). These differences can explain that the oxidation state of 21 SOA formed during dry smog chamber experiments is significantly lower than that of ambient 22 SOA (Kroll and Seinfeld, 2008;Aiken et al., 2008;De Carlo et al., 2008;Ng et al., 2010;Lee 23 et al., 2012). 24 Volkamer et al. (2007) suggested that chemical processes in the aqueous phase of hygroscopic 25 particles, such as wet aerosol, can efficiently contribute to aqSOA mass. Besides, wet aerosol 26 provides higher precursor concentrations than in cloud and fog water droplets and reside in 27 the atmosphere over hours or days (Ervens et al., 2011), suggesting a significant role for 28 aqSOA formation in wet aerosol, in particular, in regions with high relative humidity (Carlton 29 and Turpin, 2013) and hygroscopic aerosol. Isoprene has the largest global atmospheric 30 emissions, estimated at 600 Tg yr −1 of all non-methane VOCs (Guenther et al., 2006). Its key 31 oxidation products, i.e. methacrolein (MACR) and hydroperoxides (Kroll et al., 2006) are 32 known to contribute directly to the formation of SOA in the atmosphere. Methyl vinyl ketone 33 evaporation. It is thus likely that the atmospheric impact of MVK reactivity, and especially its 23 ability to form SOA, is very different under dry and humid conditions. 24 The aim of the present study is to investigate the aging of the oligomers formed through 25 aqueous phase photooxidation of MVK. We determine the SOA chemical composition during 26 the formation and aging of the aqueous phase oligomers and we revisit the corresponding 27 SOA mass yields. A large range of initial precursor concentrations (from 0.2 to 20 mM) is 28 investigated in order to study the competition between functionalization and oligomerization. 29 30 2. Experimental 1 A photoreactor was used to simulate the aqueous phase photooxidation of MVK. • OH 2 radicals were generated from H 2 O 2 photolysis (Table 1). The liquid phase was analyzed using 3 a variety of analyzers for qualitative and quantitative characterization of the solution (detailed 4 in section 2.2). 5 For aerosol generation, aliquots of the solution were sampled from the photoreactor at specific 6 reaction times, then nebulized and dried prior to aerosol characterization using a scanning 7 mobility particle sizer (SMPS) and a high resolution time-of-flight aerosol mass spectrometer 8 (AMS) (Figure 1). Each experiment, i.e. aqueous phase photooxidation and aerosol 9 generation, was repeated at least once. 10

Photoreactor 11
The photoreactor set-up was based on the one described by Renard et al. (2013). It was a 450 12 cm 3 Pyrex thermostated photoreactor, equipped with a 1000 Watt Xenon arc lamp  Oriel, LSH 601) and a glass filter (ASTM 490 AM 0). The resulting spectral irradiance into 14 the reactor is compared to that of the sun at sea level for a 48.3° zenith angle in Figure S1 (Table 1), i.e., 0.2, 0.5, 2, 5 and 20 mM, corresponding to 9.6 to 22 960 mgC L −1 . Considering MVK as a proxy for UWSOC, this concentration range is 23 comprised in the range of the estimated total UWSOC concentrations from fog droplets to wet 24 aerosol (Renard et al., 2013). 25 The 50 cm 3 gas phase head space of the photoreactor was opened to ambient air for a few 26 seconds during each sampling. We verified in control experiments that this procedure induced 27 insignificant losses of MVK from the solution. Since the photoreactor was closed most of the 28 time, this procedure also induced a decrease in the dissolved O 2 concentrations, at a rate that 29 was dependent on the initial reactant concentrations, as shown in the companion paper by 30 Ervens et al. 31 The initial H 2 O 2 concentrations were chosen in order to obtain a ratio , in order 1 to favor • OH reaction with MVK rather than with H 2 O 2 by more than 90 %. Under these 2 conditions, • OH concentrations were estimated in the range (2 -6) ×10 −14 M (see 3 supplementary information 3), which falls in the range of the estimated values for • OH 4 concentrations in cloud and fog droplets (Herrmann et al., 2010;Ervens andVolkamer, 2010 5 andArakaki et al., 2013). 6

Aqueous phase characterization 7
Aliquots of the solution sampled from the photoreactor were analyzed for qualitative structure 8 elucidation of the oligomers using ultra-performance liquid chromatography mass 9 spectrometry (UPLC-ESI-MS); and for quantitative studies of the concentrations of i) MVK 10 and H 2 O 2 by liquid chromatography coupled to UV detection (UHPLC-UV), ii) carboxylic 11 acids by ion chromatography-mass spectrometry (IC-ESI-MS), and iii) oligomers using 12 preparative liquid chromatography associated to total organic carbon (TOC) analyses. 13

UPLC-ESI-MS analyses 14
Aliquots of the solution sampled from the photoreactor were analyzed for organic species 15 using an ultra-high performance liquid chromatographic system coupled to a time of flight 16 mass spectrometer equipped with an electrospray source and an ion mobility cell (Synapt-G2 17 HDMS, Waters). The mass spectrometer was tuned to V-mode with a resolving power of 18 18000 at m/z 400. The mass accuracy (< 5 ppm) allowed for the determination of elemental 19 composition of organic species (Renard et al., 2013 and2014), using the I-FIT software. The 20 I-FIT isotope predictive filtering is a strategy to reduce the number of proposed elemental 21 compositions using algorithms to estimate the number of carbon, oxygen atoms in an 22 unknown molecule based on the mass of the molecular ion and the relative intensity of the 23 first two most abundant isotopes (Hobby, 2005). 24 All parameters used are detailed in Renard et al. (2013)

IC-ESI-MS analyses 24
Quantification of organic acids in the solutions was performed with an ion chromatography 25 system (Dionex ICS3000) driven by Chromeleon ® software (6.80 version), composed of a 26 gradient pump (Dionex SP-5), an autosampler (Dionex AS40), a conductivity detector 27 (Dionex, CD25) and coupled to a quadrupole mass spectrometer (Thermo Scientific Surveyor 28 MSQ) operated in the negative electrospray ionization (ESI) mode, using nitrogen gas flow of 29 6 L h -1, 40 psi, temperature 500°C; capillary voltage 3,5 kV; sample cone voltage 75 V. An 30 electrolytic suppressor (Dionex, 4 mm ASRS 300) operated in external water mode (7 mL 31 min -1 ) was placed before the conductivity cell. An additional peristaltic pump was used during 32 measurements to wash the entrance cone of the mass spectrometer with water at a flow rate of 1 0.4 mL min -1 . The chromatographic separations were carried out on a column (Dionex,  2 IonPac AS11-HC, 4 x 250 mm) coupled to a guard column (Dionex, AG11-HC, 4 x 50 mm). 3 A 25 µL sample was injected automatically using a 25 µL loop injection valve. The analysis 4 was performed at 35°C, with a flow rate set at 0.8 mL min -1 . Eluent A (Ultra High Quality 5 water) and eluent B (100 mM NaOH) were flushed with purified helium gas for 30 min and 6 kept under nitrogen atmosphere during the procedure. Separation was carried out using the 7 following gradient (min, B % ): 0, 1 %; 12, 5 %; 30, 19 %; 40, 40 %, 50, 1 %. The analytes 8 were monitored using the selected ion-monitoring (SIM) mode, and signal areas (counts min -9 1 ) of each peak were used for quantification. 10

TOC analyses 11
TOC measurements were associated to preparative liquid chromatography to separate the 12 oligomers from the small and/or volatile reactants and reaction products in the liquid samples, 13 in order to measure the oligomer mass yields in experiment A (see section 3.2.3). A total 14 organic carbon / total nitrogen (TOC/TN) analyzer (Analytik Jena, N/C2100S) with the non-15 purgeable organic carbon (NPOC) method was used to quantify the produced oligomers in our 16 liquid samples. 17 The NPOC method consists in pre-purging samples with oxygen and pre-acidifying (at pH=2 18 with HCl) to remove the inorganic carbon and purgeable organic carbon. TOC is measured by 19 injecting the sample into a heated combustion tube (800°C) with an oxidation catalyst. The 20 CO2 produced is measured by a non-dispersive infrared (NDIR) gas analyzer. TN is measured 21 in parallel using chemiluminescence detection (CLD). 22

Particle generation and characterization 23
For aerosol generation, 35 mL of the solution was sampled at specific reaction times (Table  24 1), and nebulized using an atomizer (TSI, 3079) with a flow rate of 3.5 L min -1 (Figure 1). 25 The generated droplet flow was led through a silica gel diffusion dryer and diluted with 26 filtered ambient air (at 5 L min -1 , using a HEPA capsule filter). A small fraction of the sample 27 (≈ .4 L min -1 ) was passed through a Nafion dryer (Permapure, MD-110), before entering a 28 small 100 mL glass mixing chamber and the on-line analytical devices. The obtained relative 29 humidity was constant during all experiments at ca. 15 % measured at the entrance of the 30 AMS ( Figure 1). The nebulization time for each sample was 30 min and, to ensure constant 31 and reproducible aerosol generation, only the last 15 min of nebulization were employed for 32 data analysis. To avoid memory effects, before each nebulization experiment, the system was 1 flushed by nebulizing UHQ water for 30 min. 2 The number size distribution was measured using a scanning mobility particle sizer (SMPS), 3 (Grimm, SMPS+C) consisting of a differential mobility analyzer (L-DMA) with a 4 condensation particle counter (Grimm, CPC, 5.403). The analyzed particle size ranged from 5 11 to 1083 nm (scanned within 6 min and 43 s). 6 A high resolution time-of-flight aerosol mass spectrometer was used to measure the bulk 7 chemical composition of the non-refractory submicron particulate matter (De Carlo et al., 8 2006;Canagaratna et al., 2007). The instrument was used under standard conditions 9 (vaporizer at 600°C and electron ionization at 70 eV), in the high sensitivity V-mode with a 10 resolving power of 2000 at m/z 200. Each measurement point was averaged for 2 min and 40 s 11 (MS-and PToF-cycle, 40 s each, 2 cycles per run). Mass spectra of filtered air, using a HEPA 12 capsule filter, were taken prior each series of nebulizing experiments in order to adjust the 13 m/z 44 entry of the fragmentation table due to gas phase CO 2 . 14 The standard fragmentation table with the corrected air fragment column for our carrier gas 15 and the default values of relative ionization efficiency were used in the AMS data analysis 16 (Squirrel 1.51H and the software PIKA 1.10H). 17

Evidence for oligomer formation and aging 19
During MVK-• OH oxidation, the aqueous phase composition was monitored and compared to 20 the composition of the corresponding nebulized solutions. 21

Aqueous phase analyses 22
For each experiment, the solution was directly monitored using UPLC-ESI-MS and UHPLC-23 UV for reaction times up to 150 min (Table 1). This time was higher than the complete 24 consumption of MVK in order to study the formation of oligomers and their aging processes, 25 as illustrated in Figure 2. oligomers, corresponding to several initiator radicals identified by Renard et al. (2013) under 2 similar conditions. As an example, a tentative molecular structure of the most intense series is 3 given in Figure 2c and is highlighted in red in the mass spectrum. At that time, 90 % of MVK 4 was consumed. Finally, the intensities of all the oligomer series decreased simultaneously for 5 all masses with no change in the oligomer pattern up to 90 minutes. From this reaction time, 6 the mass spectra show a collapse of the regular pattern in both negative (Figure 2d, 150 min) 7 and positive modes, possibly corresponding to a drastic aging process in which oligomers 8 formed smaller molecules. This hypothesis is confirmed by a more global approach, using the 9 SMPS and the AMS analysis of the SOA formed after nebulization of the solutions. 10

Aerosol composition of SOA generated after nebulization of the solutions 11
Under similar conditions, we verified as done in a previous study (Liu et al., 2012) that 12 nebulization of the reacted solutions and subsequent aerosol particle drying processes induced 13 negligible chemical transformations of the oligomers compared to the aqueous phase 14 composition. It was thus meaningful to compare the compositions of aqueous phase and SOA 15 after nebulization. 16 The AMS mass spectra ( nebulizing an aqueous solution containing the reactants before reaction, with m/z fragments 23 lower than 100. Then, the total mass increases to reach a maximum at 50 min (Figure 3c), an 24 order of magnitude higher than at 5 min. The mass spectrum is dominated by the m/z 43 25 fragment ( Figure 3c). This observation is likely due to fragmentation by electronic impact of 26 oligomers containing repetitive carbonyl functions such as those identified in the aqueous 27 phase as shown by the example of a tentative molecular structure in Figure 2c. Finally, the 28 intensity of both the total organic mass and that of m/z 43 fragment decrease, the one of m/z 29 44 increases, and they both dominate the AMS mass spectrum with the same intensity at the 30 end of our investigation (150 min, Figure 3d). 31 Furthermore, comparing the AMS mass spectra between 50 and 150 min at higher masses 32 (m/z 100 -200) (Figure 4), it is clear that at 50 min of reaction, the mass spectrum contains 33 more fragments in this range, than at 150 min. It is thus likely that the oligomers are being 1 significantly photooxidized through a fragmentation mechanism that forms smaller acidic 2 compounds, as observed by Aljawhary et al. (2013) for different precursors, and it confirms 3 the oligomer aging process suggested in Figure 2. After 50 min, oligomer fragmentation 4 prevails over oligomer formation. 5 For the quantitative study (see section 3.2), we used the data provided by the SMPS analysis. 6 Note that the overall collection efficiencies (CE) of the AMS in our experiments varied from 7 0.07 to 0.21, related to the SMPS signal. These low CE values, compared to chamber studies 8 or ambient aerosols, can be due to particle bounce at the vaporizer surface before 9 volatilization and to the shape and size-dependent transmission of the aerodynamic lens. As a 10 result, the studied compounds did not volatilize sufficiently fast at standard AMS vaporizer 11 temperatures to be fully detected (Liu et al., 2007;Docherty et al., 2013;Miyakawa et al., 12 2013). In addition to these effects, it is possible that our low CE values were also due to the 13 particle size range (50 nm -150 nm mass distribution), as the lowest part of this size range 14 corresponds to the region where the AMS transmission curve varies greatly (Liu et al., 2007). 15 This effect is confirmed by the fact that our lowest values for CE (0.07) were obtained for the 16 lowest MVK initial concentrations (0.2 -2 mM) where the smallest particles were formed (50 17 nm mass distribution). 18

SOA mass 20
For experiment B, Figure 5a shows a continuously increasing number size distribution with 21 reaction time from 5 to 150 min, with an increasing mode size during the two first kinetic 22 steps (up to 50 min), and a decreasing mode size during the third one, which corresponds to 23 oligomer aging. In order to determine the particle mass concentrations, we used the method 24 analysis of the SOA formed in our system. These ratios extend to the same ranges as those 31 used by Kuwata et al. (2012), and the resulting particle densities are reported in Table 2 and  1   Table 3. In particular, Table 2 shows a substantial change in the H/C (decrease) and O/C 2 (increase) after 50 min of reaction, t max , for which the maximum SOA mass is reached, 3 denoting the oligomer aging and inducing an increase of the aerosol density. 4 Using these particle densities, the total mass concentrations were determined, and the time 5 evolution of the resulting distribution particle mass concentrations is shown in Figure 5b for 6 experiment B. The blank signal was determined prior to each individual experiment by 7 nebulizing pure water samples and was subtracted in the results for the mass calculation. At 8 the initial reaction time (0 min), the particle size distribution was determined by nebulizing an 9 aqueous mixture of the reactants (using experiment B concentrations), it showed a mass 10 concentration (11.0 ± 1.4 µg m -3 ) not statistically different from the one obtained by 11 nebulizing pure water, assuming a density of 1.1 g cm -3 . This confirms that the reactants are 12 too volatile to form substantial amounts of organic aerosol by nebulization of the solution 13 prior to reaction. 14 Confirming the UPLC-ESI-MS aqueous phase analyses (see section 3.1.1) and the AMS 15 results (see section 3.1.2), a similar kinetic behavior is also observed on the SMPS total mass 16 concentrations ( Figure 5b and Figure 6). A slow increase is observed during the first step (0 -17 10 min). Then oligomerization takes place corresponding to a fast increase of the SMPS mass, 18 until 50 min. Finally, after this maximum of oligomerization, a significant decrease of the 19 SMPS mass is observed. This decrease may be related to the decrease in the particle size 20 ( Figure 5a), which can be due to the decrease of the oligomer size, by fragmentation of the 21 oligomers. It is thus likely that the oligomer aging forms more volatile compounds that the 22 SMPS does not measure. The high correlation between the total aerosol mass concentration 23 and the consumed MVK observed in Figure 6 from 0 to 50 min, allows for the determination 24 of the SOA mass yield, as discussed in section 3.2.3. 25

Influence of initial MVK concentrations 26
The influence of the initial aqueous phase concentration of MVK on the SOA formation was 27 investigated over a wide range, i.e. from 0.2 to 20 mM (Table 1). Not surprisingly, Figure 7  28 shows that the total aerosol mass concentration increases with increasing initial MVK 29 concentration. This observation is in very good agreement with the influence of MVK initial 30 concentration on the oligomerization process observed in the aqueous phase by Renard et al. 31 (2013). For experiments D and E, corresponding to the lowest initial MVK concentrations, the 32 SMPS and AMS signals were low, and they could be influenced by water impurities, whereas 33 no such influence was observed for experiments A, B and C. This is why the signal obtained 1 from the blank experiments was subtracted only for experiments D and E in Figure 7. 2 Moreover, Figure 7 clearly shows a different kinetic behavior of the SOA mass concentration 3 from the lowest initial concentration experiments (D and E), compared to the three highest 4 ones (experiments A, B and C). For experiments A, B and C, the SOA mass concentration 5 increases rapidly, reaches a maximum, and then decreases, while for experiments D and E, the 6 signal slowly increases and does not reach a maximum. This particular evolution may be due 7 to different chemical mechanisms occurring at different initial concentrations. We 8 hypothesized the predominance of oligomerization at 2 mM initial concentration and above, 9 this is further discussed in section 4. 10 The continuous increase of the particle number (shown in Figure 5a for experiment B) with 11 reaction time was observed for all initial concentrations (experiments A to E), whereas the 12 decrease of the size mode (in the number size distributions, after t max ) was observed for the 13 three highest initial concentrations only (experiments A, B and C) and not for experiments D 14 and E, i.e. only during oligomer aging. 15

SOA mass yields 16
The SOA mass yields, Y t , were calculated at each reaction time step t from eq. 2. 17 (eq. 2) 18 Where [MVK] t is the consumed [MVK] in mg L -1 at reaction time t; and [SOA] t is the 19 formed SOA mass at reaction time t, in mg per L of evaporated water. This term takes into 20 account the SOA mass (M SMPS ) measured by the SMPS at time t (in µg m -3 ), the atomizer flow 21 (F atomizer in L m -3 ), the dilution (f dil ), and the transmission efficiency in our nebulizing system 22 (T eff in %) (see Table S1). 23 (eq. 3) 24 The yields obtained at t max for experiments A, B and C are shown in Table 3. Although the 25 total SOA mass (at t max ) increases linearly with the initial concentration for these three 26 experiments, the yields are statistically identical as well as their H/C and O/C ratios. Due to 27 the very large uncertainties of our yield determinations (see below), it is not possible to use 28 these data (Table 2 and 3) to provide any interpretation on the possible effect of initial 29 concentrations on the yields. In contrast, the O/C and H/C ratios clearly show statistically 30 stable values when the total particle mass increases from 100 to 900 µg/m 3 (Table 3). It is thus 31 likely that the total mass loading does not influence the relative oxygenation of the SOA 1 produced (at t max ) under our experimental conditions. 2 Although the particle mass loadings (M SMPS ) were accurately measured, our yield 3 determinations were affected by large uncertainties due to the estimation of the transmission 4 efficiency in our nebulizing system (see supplementary information 1, Table S1). In order to 5 confirm these yields' values, another method was tested for experiment A at 9 min of 6 reaction (i.e. close to t max ). Preparative chromatography was performed using UPLC, where 7 small molecules were separated from the oligomers using a divert valve, at retention times 8 lower than 2 min. The solution containing oligomers was accumulated, concentrated and 9 analyzed using a TOC analyzer. From the carbon mass, we deduced the total mass using the 10 H/C and O/C ratios given by the AMS. The yield was then directly calculated from the total 11 mass of sample (in mg L -1 ) divided by the mass of consumed MVK at the same reaction time. 12 A yield of 59 ± 5 % (in mass) was obtained with this method at 90 min of reaction, thus 13 statistically similar from the one obtained by the nebulizing method (70 ± 50 %) at t max . 14 These yields are significantly higher than those obtained by Liu et al. (2012) who obtained 15 yields up to 9.9 % under similar experimental conditions as ours. It is important to note that 16 these values were obtained assuming all the particle densities were 1 g cm -3 in Liu et al. 17 (2012), and also the transmission efficiency of the nebulizing system was calibrated with 18 NaCl solutions. However, it is likely that succinic acid or ammonium nitrate are more 19 adequate for the calibration, and we show in the supplementary information 1 (Table S1) that 20 the transmission efficiency of NaCl solutions are significantly different from the two other 21 solutions. The nebulizing system was slightly different, with a teflon bag in Liu et al. (2012) 22 that could enable i) larger amounts of wall losses for organic particles as compared to the 23 system presented here; but ii) longer particle residence times, leaving more time for gas-24 particle equilibrium than in our system. However, our control experiment using preparative 25 chromatography confirms the high yield value obtained here, independent on the nebulizing 26 system and its calibration. The different yields obtained here as compared to the study by Liu 27 et al. (2012) may be due to the different irradiation Xe lamp used: 300 W, with a pyrex filter, 28 in Liu et al. (2012), and 1000 W, with a ASTM 490 AM 0 filter, in the present study. The 29 influence of the lamp spectra on SOA mass yields of other systems, i.e. gas phase 30 photooxidation of biogenic and anthropogenic precursors, have been previously observed in 31 atmospheric simulation chambers (Bregonzio-Rozier et al., 2014). We verified, using a 32 spectroradiometer (SR-501, LOT-Oriel), that the spectral irradiance of the 300W and the 33 1000W Xe lamps at  ≥ 400 nm represent respectively half and twice the solar irradiance 1 intensity at sea level, for a 48.3° zenith angle (Supplementary information 2, Figure S1). Due 2 to the high variability of the irradiance in the atmosphere at  ≥ 4 nm, as shown by the 3 Tropospheric Ultraviolet and Visible Radiation Model (http://cprm.acd.ucar.edu 4 /Models/TUV/Interactive_TUV/), both lamps can be seen as representative of the natural 5 irradiance in this wavelength range. However at 300 nm, the spectral irradiance of the 1000W 6 Xe lamp is 7 and 9 times higher than that of the direct solar irradiance (for a 48.3° zenith 7 angle) and the 300W Xe lamp respectively (Supplementary information 2, Figure S1). This 8 part of the spectrum is essential for photochemistry, and may induce different photochemical 9 processes: we verified that we observed the same series of oligomers as in Liu et al. (2012), 10 but with different relative intensities. The different spectral irradiance of the lights used at 300 11 nm may be the reason for the different yields obtained, but it needs to be confirmed by a 12 thorough study of the influence of the spectral irradiance in the UV, on the oligomer mass 13 yields. 14 It is interesting to note that the yields and densities obtained in the present study are in the 15 same range as those of a similar study with a different precursor, i.e. glycolaldehyde and a 16 different irradiation system even more intense in the UV, i.e. a 254-nm mercury lamp (Ortiz-17 Montalvo et al., 2012). They reported aqSOA yields for oxidation products of glycolaldehyde 18 (1 mM) which decrease gradually with reaction time from about 120% to 50%; while the 19 calculated densities increase from 1.3 to 1.6 g cm -3 . 20  (acrylonitrile) similarly leads to stabilization of the monomer and decreases enthalpies of 27 polymerization (Odian, 2004). It is thus likely that a large number of atmospherically relevant 28 molecules can follow the same process either in the bulk or at the wet aerosol interface 29 (Kameel et al., 2013 and2014). In this context, our results suggest that this class of 30 compounds can impact the aerosol composition, and contribute to aqSOA formation upon 31 water evaporation. The corresponding aqSOA mass yields seem to depend on the spectral 32 irradiance of the light used to initiate the photochemistry, but further studies are needed to 33 confirm this point. Finally, the aging of the oligomers formed could be an explanation, at least 1 in part, for the presence of diacids, such as oxalic, malonic and succinic acids, observed in the 2 ambient aerosol (Legrand et al., 2007;Kawamura et al., 2010). In Part 2 of this study, the 3 atmospheric relevance of these processes is explored by means of multiphase box model 4 studies. Oligomers formed through in-cloud methylglyoxal reactions: Chemical composition, 17

Supplementary Information 1: Calibration experiments of our set up
Calibration experiments of our set up were performed using aqueous solutions of three different compounds (NaCl, NH 4 NO 3 , and succinic acid) at various concentrations covering those of the consumed MVK during its reaction (Table S1). In these calibration experiments, for each compound, the obtained numbers of particles increase with increasing solution concentrations, and the corresponding total particle mass (using the corresponding densities) increases linearly with the solution concentration. This result was used to evaluate the transmission efficiency of our set up. Assuming a similar behavior for the nebulized MVK-oligomers up to t max (i.e. increasing oligomer concentrations with reaction time, assuming no major change in the oligomer composition), the calibration experiments were used to determine the SOA mass yields according to equations 2 and 3. The differences obtained between the three calibration experiments may be due to the different physical properties of the particles (e.g. volatility, surface tension and hygroscopicity) linked to their chemical composition. Table S1 shows that the solutions of NaCl generated a significantly higher transmission efficiency than the solutions of NH 4 NO 3 and succinic acid. It is likely that the properties of the SOA generated from the nebulized solutions of oxidized MVK were closer to that of NH 4 NO 3 and succinic acid than NaCl. Finally, the transmission efficiency obtained for NH 4 NO 3 solutions was used to calculate the SOA mass yields according to equations 2 and 3. Supplementary Information 2: Average spectral irradiance of the Xe lamps Figure S1: Figure S1: Average spectral irradiance of the Xe lamps 1000 W with a ASTM 490 AM 0 filter (in red, used in the present study), 300 W with a pyrex filter (in blue, Liu et al., 2012), and as compared to the direct solar irradiance at sea level, for a 48.3° zenith angle (in orange). Lamp Xe 300 W (Liu et al., 2012)