Atmospheric Chemistry and Physics Discussions

Abstract. It is now accepted that one of the important pathways of secondary organic aerosol (SOA) formation occurs through aqueous phase chemistry in the atmosphere. However, the chemical mechanisms leading to macromolecules are still not well understood. It was recently shown that oligomer production by OH radical oxidation in the aerosol aqueous phase from α-dicarbonyl precursors, such as methylglyoxal and glyoxal, is irreversible and fast. Methyl vinyl ketone (MVK) was chosen in the present study as it is an α,β-unsaturated carbonyl that can undergo radical oligomerization in the aerosol aqueous phase. We present here experiments on the aqueous phase OH-oxidation of MVK, performed under various conditions. Using NMR and UV absorption spectroscopy, high and ultra-high resolution mass spectrometry, we show that the fast formation of oligomers up to 1800 Da is due to radical oligomerization of MVK, and 13 series of oligomers (out of a total of 26 series) are identified. The influence of atmospherically relevant parameters such as temperature, initial concentrations of MVK and dissolved oxygen are presented and discussed. In agreement with the experimental observations, we propose a chemical mechanism of OH-oxidation of MVK in the aqueous phase that proceeds via radical oligomerization of MVK on the olefin part of the molecule. This mechanism highlights in our experiments the paradoxical role of dissolved O 2 : while it inhibits oligomerization reactions, it contributes to produce oligomerization initiator radicals, which rapidly consume O 2 , thus leading to the dominance of oligomerization reactions after several minutes of reaction. These processes, together with the large range of initial concentrations investigated show the fundamental role that radical oligomerization processes likely play in polluted fogs and atmospheric aerosol.


Introduction
Although secondary organic aerosol (SOA) represents a substantial part of organic aerosol, which affects air quality, climate and human health, the understanding of its formation pathways and its properties is still limited due to the complexity of the physicochemical processes involved.It is now accepted that one of the important pathways of SOA formation occurs through aqueous phase chemistry (Hallquist et P. Renard et al.: Radical mechanisms of methyl vinyl ketone oligomerization al., 2009;Carlton et al., 2009;Ervens et al., 2011).In particular, a number of studies have observed the formation of large molecular weight compounds in atmospheric aerosols (see for example Claeys et al., 2004Claeys et al., , 2010;;Baduel et al., 2010;Liao and Seinfeld, 2005;Nozière et al., 2006) and in cloud/fog droplets (Herckes et al., 2002(Herckes et al., , 2007)).
Recent studies have shown that (aerosol) aqueous phase chemistry of glyoxal (Volkamer et al., 2007(Volkamer et al., , 2009;;Ervens and Volkamer, 2010;Lim et al., 2010), methylglyoxal (Tan et al., 2012), pyruvic acid (Guzmàn et al., 2006;Tan et al., 2012) glycolaldehyde (Ortiz-Montalvo et al., 2012), methacrolein and methyl vinyl ketone (El Haddad et al., 2009;Liu et al., 2012) can produce significant amounts of SOA.In particular, Volkamer et al. (2007Volkamer et al. ( , 2009) ) and Ervens and Volkamer (2010) have shown that SOA production can occur via cloud and aerosol aqueous phase processes of glyoxal in deliquesced particles named wet aerosol, where ambient relative humidity (RH) ranges from 50 to 80 %.These findings give an extremely large set of conditions where organic aqueous phase processes can occur, i.e., from rain drop, cloud and fog droplet to wet aerosol, for which atmospheric lifetimes (< 1 min -days), liquid water content (LWC ∼ 10 8 -1 µg m −3 ), surface area of a particle, drop or droplet (A = πD 2 ∼ 10 −2 -10 −10 cm 2 ), particle number concentration (∼ 10 −4 -10 4 cm −3 ) and individual organic and inorganic chemical concentrations (∼ 10 −2 -10 6 µM) vary over orders of magnitude (Ervens and Volkamer, 2010).In their review, Lim et al. (2010) report that aerosol phase reactions of glyoxal with OH radicals performed under high initial concentrations tend to be faster and form more oligomers than non-radical reactions.They conclude that in clouds/fog conditions (i.e., diluted concentrations of 10 −2 -1 mM), radical reactions yield organic acids, whereas in wet aerosols (i.e., concentrated conditions of 10 mM-10 M) they yield large multi-functional species, or oligomers, formed via radical-radical reactions.An oligomer is a molecule that consists of a few monomer units (from 2 to up to 30).Lim et al. (2010) and Tan et al. (2012) propose that radical-radical reactions to form oligomers are alkyl-alkyl radical additions, which always compete with O 2 addition reactions.This explains why oligomer formation is observed only at high initial precursor concentrations, inducing high alkyl radical concentrations (after initial OH-oxidation of the precursor) which are required for radical-radical reactions to take place in competition with the reaction of O 2 .However, O 2 concentrations were supposed to stay constant at saturation (i.e., Henry's law equilibrium) in these studies, as they were only measured and the beginning and at the end of the reaction.
In the present study, in order to determine the atmospheric relevance of radical reactions, the radical mechanisms and the influence of O 2 concentrations on this chemistry is explored in details, using a different precursor, i.e., methyl vinyl ketone.This compound is an α,β-unsaturated carbonyl that is water soluble, it bears a highly reactive function (i.e., carbon-carbon double bond) that is likely to play a major role on radical chemistry and oligomer formation, as it was preliminarily shown by Liu et al. (2012).The reactivity of olefin compounds has been scarcely studied in the aqueous phase up to now, although a number of field measurements have observed them in atmospheric waters: unsaturated diacids were detected in rain and fog samples (7-14 % of the total mass of diacids: Kawamura and Ikushima, 1993;Sempéré and Kawamura, 1994;Kawamura et al., 1996) and in marine aerosols (2-7 % of the total mass of diacids; Fu et al., 2013).In clouds, it was observed that 1-18 % of the total mass of carbonyls were unsaturated carbonyls (among which methyl vinyl ketone) (van Pinxteren et al., 2005), and in biogenic aerosols, unsaturated polyols (C5-alkene-triols) represented 2-5 % of the total mass of identified polyols (Claeys et al., 2010).Finally, using NMR spectroscopy, Decesari et al. (2005) detected that 10-35 % (respectively 7-37 %) of the organic chemical functions were unsaturated in fog samples (respectively aerosols) in the Po Valley.In view of these numbers, one can reasonably suppose that 2-20 % of the organic matter concentration is unsaturated in atmospheric waters.Therefore, assuming total water soluble organic compounds (WSOC) concentrations of 0.01-1 µM in rain drops, 1-100 µM in cloud droplets, 1-100 mM in fog droplets and 1-10 M in wet aerosol, one obtains a range of unsaturated organic compounds of 0.0002-0.2µM in rain drops, 0.02-20 µM in cloud droplets, 0.02-20 mM in fog droplets and 0.02-2 M in wet aerosol.
The aim of the present study was to determine the radical mechanisms involved in the oligomerization of MVK, and to identify the oligomers formed via this chemistry.MVK was used as a model compound for unsaturated organic compounds present in atmospheric waters, its initial concentrations were varied from 0.2 to 20 mM, thus representing the total concentrations of unsaturated organic compounds in fog droplet and wet aerosol.In order to determine the atmospheric relevance of this radical chemistry, the influence of temperature and dissolved oxygen concentrations were studied.

Experimental
A photoreactor was used to simulate the aqueous phase photooxidation of MVK.HO • radicals were generated from H 2 O 2 photolysis.In order to determine the reaction mechanism, a complete set of analytical strategies was used to identify the oligomers produced.Aqueous aliquots sampled at different photoreaction times were analyzed by mass spectrometry, UV absorbance spectroscopy and NMR.In order to test our radical mechanism for MVK oligomerization, its aqueous phase photooxidation was studied under various conditions of temperature, MVK initial concentrations, and above all, dissolved oxygen concentrations.2012), the photolysis rate constant of H 2 O 2 increased by an order of magnitude due to the more powerful lamp used (Table 1).This resulted in higher HO • concentrations using lower H 2 O 2 initial concentrations, and thus led to faster MVK degradation and kept artifacts due to H 2 O 2 reactivity to a minimum.

Atmos
A 50 mL gas space was left over the liquid level.The loss of aqueous MVK to the gas phase was insignificant, based on its Henry's law constant (41 M atm −1 at 25 • C; Iraci et al., 1999) and on control experiments.

Experimental conditions
All experiments started with irradiation of UHQ water (18.2M cm, Millipore), then H 2 O 2 (30 %, non-stabilized, Acros) was introduced, photolysed for 10 min until photostationary conditions were reached.MVK (99 %, Sigma Aldrich) was finally introduced at time noted 0 min.The experimental conditions (UV-visible irradiation, temperature, concentrations) were chosen in order to be representative of fog droplet and/or wet aerosol conditions.Additionally, a detailed study of the influence of oxygen initial concentrations was performed, in order to determine its role on radical chemistry, and its atmospheric relevance.

Control experiments
To check that the observed products resulted from the aqueous phase OH-oxidation of MVK, two control experiments were conducted: (1) MVK (20 mM) + H 2 O 2 (400 mM) under dark conditions, and (2) direct photolysis of MVK (20 mM).MVK was not significantly consumed either in presence of H 2 O 2 in the dark (1) or under direct irradiation (2).Additionally, we performed (3) direct photolysis of H 2 O 2 (400 mM) in absence of MVK, and the observed decrease of H 2 O 2 concentrations allowed us to determine its photolysis rate constant J (Table S1).We used UPLC-UV at 265 nm for H 2 O 2 detection.The duration of these three control experiments was 300 min, consistent with the reaction time of the actual MVK photooxidation experiments.Tan et al. (2010) have shown the important impact of initial concentrations on oligomer formation for α-dicarbonyls.The experiments were thus carried out with various MVK initial concentrations, i.e., 0.2 mM, 2 mM and 20 mM (corresponding to 9.6 to 960 mg C L −1 ), which can be regarded as representative of the total concentrations of unsaturated organic compounds in fog droplet and wet aerosol (see Introduction).

Initial concentrations of reactants
The ratio [H 2 O 2 ] 0 /[MVK] 0 = 20 was chosen to favor HO • reaction toward MVK over its reaction with H 2 O 2 (Reaction R2, in Table S1) by more than 90 %.Under these conditions, HO • concentrations were estimated to approximately 5 × 10 −14 M, which fall in the range of the estimated values for cloud and fog droplets (Herrmann et al., 2010;Ervens and Volkamer, 2010).

Temperature, pH and dissolved oxygen
Temperature, pH and dissolved oxygen concentrations were continuously monitored in the solution using a Consort C3020 multi-parameter analyzer.Three relevant temperatures were tested, 5 • C, 9 • C and 25 • C. The pH of the unbuffered solution decreased from 6 to 3 within 90 min of photoreaction (Experiments A, B and C: [MVK] 0 = 20 mM, at 25 • C, 9 • C and 5 • C; Table 1).
At 25 • C, in experiments where dissolved oxygen concentrations were not constrained (Experiments A, B, C, E and F in Table 1), these concentrations were at Henry's law equilibrium with atmospheric O 2 in UHQ water.After H 2 O 2 introduction the dissolved O 2 concentrations were supersaturated by a factor of 155 % (corresponding to 400 µM of O 2 in experiment A), due to H 2 O 2 photooxidation (Reactions R3 and R4 in Table S1), and then they decreased down to nearly 0 during MVK photooxidation.Thus, for these experiments, O 2 was supersaturated (by a factor of 155 %) when MVK was introduced (called "supersaturated initial O 2 concentrations" experiments in the following).Due to this high variability, and also because oxygen is known as a radical polymerization inhibitor (Odian, 2004), the influence of initial dissolved oxygen concentrations were investigated.
To complete the investigation of the role of oxygen, we also performed experiments with low initial O 2 concentrations (experiments D) by bubbling a flow of argon in the solution before injecting MVK in order to evacuate as much O 2 as possible before the reaction started.Then the argon flow was stopped in order to prevent MVK from evaporation, and the samples were collected using a syringe through a septum to avoid atmospheric O 2 entering the reactor.

Mass spectrometric analyses of the solution
An ultra-high performance liquid chromatographic column coupled to a time of flight mass spectrometer equipped with a soft ionization electrospray source was used (Synapt-G2 HDMS, Waters).It is noted UPLC-ESI-MS hereafter.This instrument enabled us to demonstrate the oligomerization dependence on oxygen, and to follow the oligomerization kinetics.One of the experiments A (at 25 • C, Table 1) was performed using an ultra-high resolution mass spectrometer coupled to an electrospray ionization source, noted ESI-UHRMS (LTQ-Orbitrap-XL, Thermo).

UPLC-ESI-MS analyses
Aqueous samples (taken from the photoreactor) were analyzed for organic species using Quadrupole-Time-of-Flight Mass Spectrometry and Ion Mobility Spectrometry (Synapt G2 HDMS, Waters, MA, USA) combined with electrospray ion (ESI) source and coupled with an Acquity UPLC system (Waters).The mass spectrometer was used in its resolution mode, up to 18 000 FWHM (Full width at half maximum) at 400 Da and allowed the determination of elemental composition of some organic species, using the I-FIT software.The I-FIT isotope predictive filtering is a strategy to reduce the number of proposed elemental compositions using algorithms to estimate the number of carbon, oxygen or sulfur atoms in an unknown molecule based on the mass of the molecular ion and the relative intensity of the 1st and 2nd isotopes (Hobby, 2005).
The chromatographic separations were carried out on a HSS T3 UPLC column (Waters Acquity High Strength Silica T3, 2.1 × 100 mm, 1.8 µM) at 40 • C. The mobile phases consisted in (A) 0.1 % formic acid in water (Fluka, 98 %) and (B) Acetonitrile (CAN; Optima LC-MS, Fischer).The gra-dient elution was performed at a flow rate of 600 µL min −1 using 5 to 95 % of B within 7 min and held at 95 % of B for 1.5 min.The sample injection volume was 10 µL.
The ESI source of this instrument contains two individual orthogonal sprays.One spray is for the column eluent, and the other one is for the internal standard called lockmass.During each chromatographic run, leucine enkephalin (2 ng µL −1 , C 28 H 37 N 5 O 7 , MW 555.27,Waters Q-ToF product) was used for lock-mass correction to obtain accurate masses for each organic component eluting from the column.The lock-mass syringe pump was operated at 7 µL min −1 .A solution of sodium formate (CO 2 HNa, Waters Q-ToF product) was infused daily in the ESI source to calibrate the instrument.Optimum ESI conditions were found using a 2.5 kV capillary voltage, 40 V sample cone voltage, 450 • C desolvation temperature, 120 • C source temperature, 20 L h −1 cone gas flow rate and 800 L h −1 desolvation gas flow rate.The ESI source has been optimized directly with the samples.These parameters allow increasing the signal to noise ratio, keeping soft ionization technique.
Data were collected from 50 to 1800 Da in the positive and negative ionization modes.All products were detected as their protonated molecules ([M + H] + ) or sodium adducts ([M + Na] + ) in the positive mode, and their deprotonated molecules ([M − H] − ) in the negative mode.
For experiments A (Table 1), complementary analyses were performed using MS/MS fragmentation to confirm the structure of the products.MS/MS experiments (below 200 Da) were carried out with a trap collision energy ramp from 5 to 20 eV.Additionally, traveling wave ion mobility allowed us to separate ions which had the same elemental formula.The main parameters were the IMS Wave Velocity 650 m s −1 and the IMS Wave Height 40 V.

ESI-UHRMS analyses
One of the experiments A (at 25 • C, Table 1) was performed using an ultra-high resolution mass spectrometer coupled to an ESI source (LTQ-Orbitrap-XL, Thermo).Aqueous aliquots sampled from the photoreactor were diluted by a factor of 2 with ACN, and directly infused into the ESI at a 3 µL min −1 flow rate.Using a mixture (1 : 1) of ACN and aqueous samples helps the desolvation process in the ESI source, thereby ensuring more stable operations.ACN was chosen against methanol as this latter was shown to occasionally induce esterification during the ionization (Bateman et al., 2008).Each sample was measured in the negative and positive ionization modes, with the following optimized settings: source voltage -3.5 kV; capillary temperature -275 • C; tube lens voltage -50 V in the positive mode; and source voltage -3.7 kV; tube lens voltage -90 V and same capillary temperature in the negative mode.
Transient acquisition time was set to 1 s, which corresponds to a nominal resolution of 100 000 at 400 Da, and to observe individual peaks resolution (FWHM) typically better than 200 000 at 200 Da.Each spectrum was obtained by averaging 20 min of acquisition time, so as to increase the S/N ratio of the lower intensity peaks.Acquisition was performed over two overlapping mass ranges: 50-300 Da and 150-1500 Da.This method overcomes transfer limitations of very wide mass ranges in the ion optics which guide the molecular ions through the LTQ to the Orbitrap mass analyzer.
External mass calibration was made daily by infusing a calibration mixture (L-methionine-arginyl-phenylalanylalanine, caffeine, and ultramark: MSCAL 5-10EA Supelco) having peaks in the range 195 to 1921 Da.The obtained mass accuracy was as low as 2-5 ppm, and better for peaks with a relative intensity above 0.5 % (Makarov et al., 2006).
For our moderately complex mixture (∼ 6000 peaks), containing only O, C, H, and Na, unambiguous elemental formula assignments were established below 350 Da within 1 mDa using a custom computer software (Attributor v1.05) developed by Orthous-Daunay (2011) and described in Danger et al. (2013).It is based on a suite of scripts that allow to evaluate the most probable attribution of a high resolution peak, based on its mass, the presence and relative intensity of isotopic peaks, as well as chemical rules (nitrogen rule, even electron ions, etc.) Using a Kendrick analysis (Hughey et al., 2001) based on the MVK pattern (i.e., C 4 H 6 O) found by Liu et al. (2012) and on the unambiguous attribution of most peaks below 350 Da, series of regularly spaced peaks were identified in the spectrum, and used to extend the formula attribution toward higher masses.This extension into Kendrick series was also checked against isotopic patterns.Typical standard deviation of the Kendrick mass defect in any such identified series was lower than 0.5 mDa, which ensures the proper attribution of all the members of a series to within less than 1 mDa in the full mass range explored (i.e., 50-1500 Da).

NMR spectroscopy
One of the experiments A was analyzed using NMR spectroscopy (Table 1).The instrument used was a Bruker Avance 500 MHz equipped with a 5 mm inverse-triple tuned (TXI) 1 H/ 13 C/ 15 N with z gradient coil probe.For 1-D 1 H-Spectra, 128 scans were collected with an impulsion time of 7.5 µs, a relaxation delay of 5 s, an acquisition time of 4.67 s, a spectral window of 5000 Hz and 64 K data points zero filled to 128 K before Fourier transformation with 0.3 Hz line broadening.For 2-D homonuclear (COSY and TOCSY) and heteronuclear ( 1 H/ 13 C HSQC and HMBC) experiments were performed with quadrature phase detection in both dimensions, using state-TPPI or QF detection mode in the indirect one.For each 256 (homonuclear experiments) or 512 (heteronuclear experiments) increments in the indirect dimension, 2 K data points were collected and 16 or 32 transients were accumulated in the direct dimension. 13C decoupling (GARP) was performed during acquisition time for heteronuclear experiments.A π/2 shifted square sine-bell function was applied in the indirect dimension before Fourier transformation.Spectra were treated with Topspin version 2.0.All NMR spectra were recorded at 25 • C.
Aliquots of 540 µL of the aqueous solution sampled from the photoreactor (at 5, 10, 25, 70 and 90 min of reaction) was supplemented with 60 µL of a buffer containing 100 mM phosphate and 5 mM of sodium tetra deuterated trimethylsilyl propionate (TSPd 4 , Eurisotop).TSPd 4 constituted a reference for chemical shifts (0 ppm) and quantification.Final volumes of 600 µL of prepared samples were put in 5 mm-diameter NMR tubes.
Additionally, an aliquot of 5 mL sampled at the end of the kinetic (90 min of reaction) was freeze-dried, re-suspended in 600 µL of CDCl 3 and transferred in a 5 mm NMR tube for further 1-D and 2-D NMR analysis.D 2 O and CDCl 3 were used for locking and shimming.

Evidence for the formation of oligomers by radical oligomerization of MVK
The aqueous phase OH-oxidation of MVK leads to the formation of series of oligomers as indicated by the analyses of our solutions, and is in good agreement with the study by Liu et al. (2012).Furthermore, the whole set of analyses showed that the produced oligomers were formed by radical oligomerization of MVK, as discussed below.The ionic oligomerization of MVK (through carbocations or carbanions) was unlikely under our experimental conditions because the protic and nucleophilic characters of the solvent (water) inhibit oligomerization by reacting instantly with carbocation and carbanion initiators (Odian, 2004).

Series of oligomers evidenced by mass spectrometry and NMR spectroscopy
Figure 1 shows a mass spectrum (obtained using UPLC-ESI-MS) for the retention time range 0-7 min, recorded in the positive mode, at 50 min of MVK photooxidation at 5 • C (experiment C in Table 1).Spectra typically contain hundreds of peaks in both modes, and it was out of the scope of this study to explore their full complexity.Yet, oligomer systems are clearly visible, with very regular spacing of 70.0419 Da, which correspond to the exact mass of the precursor, MVK, as confirmed by the ESI-UHRMS analysis.These systems extend up to 1800 Da, thus containing up to 25 monomers.Using a lower intensity xenon lamp, and a lower resolution mass spectrometer (Table 1), Liu et al. (2012) also found oligomer series at the same nominal masses, with slightly different intensities.The main differences observed here were the kinetics.The maximum intensity was reached around 50 min here at 25 • C, instead of 20 h previously (Liu et al., 2012) at the same temperature, certainly due to the higher HO • concentrations obtained here with a more powerful irradiation lamp (Table 1).Experiment A was repeated 12 times and analyzed using the same UPLC-ESI-MS conditions.The exact masses measured for the oligomers were very repeatable, as was the kinetics of their appearance and of MVK consumption.
Experiment A was also monitored once using 1 H NMR spectroscopy (Table 1).Figure 2 shows 1 H NMR peaks resonating at 2.36 ppm (s), 6.13 ppm (dd), 6.37 ppm (m) and 6.39 ppm (m) that were assigned to the 1 H of MVK.These 1 H peaks decreased with time while new 1 H NMR signals resonating at 1.15-1.88ppm and 2.13-2.34ppm increased with time, indicating the formation of reaction products, to the detriment of MVK.These signals are very wide and are consistent with the presence of overlapping signals due to a variety of oligomers (Alarifi and Aouak, 2009;Ziaee et al., 2009).Overall, both mass spectrometry and NMR measurements indicate that a wide variety of oligomers was formed from the reactivity of MVK and that this formation of high mass oligomers occurs through an extremely fast mechanism, as in radical propagation systems.

Indications on the chemical structure of the monomers according to NMR and UV-spectroscopic measurements
In order to identify the structure of the oligomers observed by 1 H NMR spectroscopy in Fig. 2, the sample taken at the end of the reaction (experiment A at 90 min) was freezedried and re-suspended in CDCl 3 leading to a concentrated solution of the oligomers allowing the measurement of 1-D and 2-D NMR spectra (Fig. 3).The analysis of the 1 H NMR resonances in the 1-D 1 H NMR spectrum (Fig. 3a) and of the 1 J 1 H-13 C correlations present on the 2-D 1 H-13 C HSQC NMR spectrum (Fig. 3b) allowed proposing assignments of CH, CH 2 and CH 3 functional groups of the putative oligomeric structure shown in Fig. 3. Long range J 1H−13C correlations observed on the 2-D 1 H-13 C HMQC NMR spectrum (Fig. 3c) clearly demonstrated the presence of C=O functional groups belonging to the proposed oligomeric structure.NMR experiments also showed that MVK was largely converted into oligomer forms and that most of the signals of C=C bonds have disappeared at 70 and 90 min of reaction, thus indicating that the oligomers are mostly aliphatic.
Using UV absorption spectroscopy from 190 to 400 nm during experiment A, the evolution of the two intense absorption bands of MVK are shown in Fig. 4.During the photooxidation process, the n → π * transition (Fig. 4a) was clearly shifted to shorter wavelengths (blue shift; i.e., from 296 nm at time 0 to 269 nm at 90 min of reaction), with increasing absorption intensity.The π → π * transition did not show any shift (in the range of the investigated wavelengths) (Fig. 4b), but a clear decrease of its intensity (at 211 nm) was observed.The observed blue shift shows that, in the reaction products formed, the corresponding transition needs higher energy (than in MVK) to get excited (i.e., at shorter wavelengths).This can be due to the loss of the conjugation, probably due to the loss of the vinyl function, in good agreement with the NMR analyses.Furthermore, the absorption intensity at 211 nm was directly proportional to the MVK concentrations during the reaction, as it was verified with 1 H NMR (at 2.36 ppm) and with mass spectrometry.

Time profiles of MVK, and reaction products
The observed kinetics of the MVK degradation (Fig. 5) were characteristic of those of the monomers in radical polymerization (Pearce et al., 1982).During the initiation step of the reaction (∼ 13 min at 25 • C), MVK was slowly degraded, and low molecular weight reaction products were formed, such as acetate as observed by 1-D 1 H NMR (at 1.90 ppm, Fig. 2).More generally, organic acids were produced, as denoted by  the fast decrease of the pH, from 6 to 4, during this step.After 13 min, oligomerization started, and the kinetics of MVK degradation significantly increased (Fig. 5).The maximum of oligomers intensity was reached at 50 min of reaction, for the major series.After 90 min of reaction, more than 95 % of MVK was consumed and the oligomers started to decrease.

Comparison of the oligomers formed by OH-oxidation of MVK with a synthetic oligomer of MVK
In order to confirm the radical MVK oligomerization during OH-oxidation, oligomers of MVK were synthesized using a water-soluble radical initiator (V50: 2,2'-Azobis(2-methylpropionamidine) dihydrochloride, Wako) which forms two symmetric radicals by thermal homolysis (Reaction 6 in Supplement Table S2), under oxygen-free conditions at T ≥ 50 of MVK as in our photooxidation experiments (Table S2).
UV absorption spectra obtained during the OH-oxidation of MVK in the aqueous phase (Fig. 4) and after the synthesis of MVK oligomers in the aqueous phase (Fig. S1) both showed the same blue shift on the n → π * transition band (from 296 nm), and a significant decrease of the intensity at 211 nm.
The mass spectra obtained for the synthesized oligomer (Fig. S2) were very similar to those obtained during the OHoxidation of MVK in the aqueous phase (Fig. 1). Figure S2 shows very regular spacing of 70.0419 Da, and a Poisson distribution, typical of polymers (Pearce et al., 1982).Additionally, the method used to identify the initiator radicals (see Sect. 3.2 below), applied to the high resolution mass spectrum obtained from the synthesized oligomers, allowed us to determine the chemical formula of the initiator radical formed from the V50 molecule (Reactions 6 and 7 in Supplement Table S2).
Overall, this comparison confirms the radical character of the MVK oligomerization during its OH-oxidation in the aqueous phase under similar conditions.

Proposed mechanism of radical oligomerization of MVK
In view of the observations performed in Sect.3.1, we proposed a simple chemical mechanism of radical oligomerization of MVK that is likely to occur during its OH-oxidation in the aqueous phase (Fig. 6).In the initiation step, the reactivity of MVK towards the HO • radical yields (directly or indirectly: see below Sect.3.4) initiator radicals iR • (Odian, 2004).Each radical adds on another MVK molecule by opening its vinyl double bond, leading to the formation of another radical, iR-MVK • which can then add to another MVK molecule in the same way, and so on, leading to the formation of a large radical, iR-(MVK) n -MVK • .The propagation keeps on increasing the chain length, until the radical termination.Termination occurs by bimolecular reaction between two radicals, by coupling or by disproportionation (Fig. 6), thus yielding, for each initiator radical iR • , two oligomer series including one saturated and one unsaturated terminal group (Gibian and Corley, 1973).This chemical mechanism is in very good agreement with our experimental observations.The regular mass spacing of 70.0419 Da observed in mass spectra (Fig. 1) corresponds exactly to the mass of the monomer MVK in the mechanism.Radical oligomerization is in good agreement with a very fast propagation with prompt growth of the chain www.atmos-chem-phys.net/13/6473/2013/Atmos.Chem.Phys., 13, 6473-6491, 2013 Fig. 6.Scheme of the proposed mechanism of radical oligomerization of MVK in the aqueous phase, initiated by OH-oxidation.iR • is an initiator Radical; MVK is the monomer or the repeating unit, n is the degree of polymerization.For clarity, only the external radical additions are shown.to high molecular weight oligomers after an initiation step (Fig. 5).The MVK oligomerization is in good agreement with the significant loss of the vinyl group observed in Figs. 2  and 3, which is correlated with a significant loss of double bond conjugation as observed in Fig. 4. Additionally, this is in agreement with the fact that oligomerization of carboncarbon double bonds is by far more likely than carbonyl groups because of the polarized nature of the latter (Odian, 2004).
Finally, the termination step of the mechanism is in good agreement with the observation of the same degree of polymerization (DP) (i.e., n = number of monomers) for series of oligomers in mass spectra distant by 2.0156 Da (Fig. 1).
The scheme of the proposed mechanism (Fig. 6) only shows radical additions on the β-carbon of MVK for clarity.Although both α-β-additions are possible at each step, β-additions are likely favored because resonance with the carbonyl group gives it a stronger positive character.These additions lead to the formation of a number of isomers that grows exponentially with the DP.This is clearly observed in Fig. S3 showing the extracted UPLC-ESI-MS chromatogram of one of the most intense series of oligomers R-OH-(MVK) n (S174 colored in red in Fig. 1) formed during OH-oxidation of MVK after 50 min of reaction.The retention time of the oligomers regularly increases with the DP, certainly due to the growth of the corresponding molecules.Additionally, the width of the chromatographic peaks increases with the DP (from n = 1 to n = 5) unambiguously indicating that there are different products at the same mass, probably because of different tacticities of oligomers.After n = 5, the peaks become sharper and their area decreases, indicating that one way of oligomerization may be privileged due to steric hindrance for example.In this series, which is one of the most intense ones, oligomers with a DP up to 23 (n = 23) were detected.

Effects of initial conditions on the oligomer formation
The parameters varied, oxygen concentrations, temperature and MVK initial concentrations were chosen because they are known to significantly influence radical polymerization (Odian, 2004).A second reason for testing the influence of the initial concentrations of MVK was that organic concentrations can vary by orders of magnitude in the atmosphere from cloud droplet to wet aerosol (see Introduction).
Dissolved O 2 is known as a major inhibitor of radical polymerization, and also atmospheric dissolved O 2 concentrations in the condensed phase are not known, but are supposed to be saturated.Temperature is known to influence not only the kinetics of the reactions but also the oxygen Henry's law equilibrium (Table 1).The results of these experiments allowed us to confirm the mechanism of radical oligomerization of MVK, and also to identify a number of radical initiators (Sect.3.4).

Effect of initial concentrations of MVK
Between experiments A, E and F (Table 1), initial concentrations of MVK varied from 20 to 0.2 mM with identical temperature (25 • C) and similar initial dissolved O 2 concentration (supersaturated).These three experiments show that different initial concentrations of MVK induced important effects on the oligomer formation, as observed by UPLC-ESI-MS (Fig. 7).The same major series of oligomers were observed at the three initial concentrations, but the chain length increased with the initial MVK concentration.For example, for the series depicted in red in Fig. 1 and in Fig. 7, the observed DP was 23 at 20 mM initial MVK concentration, 9 at 2 mM, and 4 at 0.2 mM.This correlation between the DP and the initial concentration of MVK is in very good agreement with the kinetic chain length of a radical chain oligomerization (Odian, 2004).
Finally, Fig. 7 clearly shows that for the high initial MVK concentrations (2 and 20 mM), the main reaction products were the oligomers (up to 800 and 1800 Da, respectively), while at the low initial MVK concentration studied (0.2 mM), low molecular weight reaction products were mainly formed (up to 230 Da).It is probable that at the latter initial concentration, the observed formation of initiators recombined because the MVK concentration was too low to propagate the radical oligomerization to higher chain lengths.

Effects of initial concentrations of dissolved oxygen and temperature
Experiments A and D (Table 1) were performed under the same initial concentrations of MVK (20 mM) and temperature (25 • C), with very different initial dissolved O 2 concentrations: supersaturated and low initial O 2 concentrations, respectively, i.e., higher and lower than Henry's law equilibrium (by a factor of 1.7 and 4.5, respectively; Table 1).
The results showed that the initial dissolved O 2 concentrations had a drastic effect on oligomerization, and on the kinetics of MVK consumption, in very good agreement with a mechanism of radical oligomerization of MVK.Dissolved O 2 is known to inhibit free radical photo-oligomerization by reacting with radical species to form chain terminating oxygenated compounds such as hydroperoxides (Decker and Jenkins, 1985).In our experiments, a strong inhibition of the oligomer formation by 155 % supersaturated dissolved O 2 was observed (Fig. 8a): at low initial O 2 concentrations (experiment D), the whole oligomer series developed within 5 min of reaction, while only first elements of the series appeared at the same reaction time under "supersaturated initial O 2 concentrations" (experiment A).This phenomenon, which was observed for all the other series of oligomers, shows that the reactivity of the initiator radicals (iR • ) underwent a competition kinetic between their reaction towards O 2 and towards MVK for further oligomerization propagation.
When O 2 was present, iR • consumed O 2 very rapidly (i.e., faster than O 2 renewal from the gas phase and from H 2 O 2 reactivity), and when O 2 concentration was sufficiently low, the competition furthered the oligomerization pathway which was extremely fast.Figure 9 shows a very contrasting be-havior of MVK time profiles between "low initial O 2 concentrations" and "supersaturated initial O 2 concentrations" conditions.Under "supersaturated initial O 2 concentrations", two kinetic steps were observed: a slow one during which the oligomers were poorly formed, followed by a fast one where oligomers were actively formed (Figs. 5 and 9b).In contrast, under "low initial O 2 concentrations", the kinetic of MVK consumption was fast from the beginning (only one step was observed), and oligomerization started as soon as MVK was introduced (Figs.9a and 8).S1).Then, as soon as MVK was introduced, the time profiles of O 2 concentration followed those of MVK, thus indicating that O 2 reacted with the radicals formed from MVK, even when oligomerization was the dominant pathway.Interestingly, under "supersaturated initial O 2 concentrations", the oligomerization acceleration started when O 2 concentrations were depleted from their initial values, but still they were higher than their Henry's law equilibrium (by a factor of 120 % and 140 % at 25 and 5 • C, respectively), thus showing the atmospheric relevance of these oligomerization processes (see atmospheric implications).Experiments A, B and C differ for the temperature (i.e., 25 • C, 9 • C and 5 • C) with identical initial concentrations of MVK (20 mM) and similar dissolved O 2 concentration (supersaturated).The kinetic decay of MVK was clearly a For each series, the name is given by "S" followed by a number corresponding to the nominal mass of the first oligomer, i.e., iR-(MVK)n = 1.b The chemical structures are given in Fig. 10.R • corresponds to the radical: 30 a : For each series, the name is given by "S" followed by a number corresponding to the nominal mass of the first oligomer, i.
, with: TSI = total Series' intensity influenced by the temperature during the initial step (Fig. 9b  and c).Additionally, the lower temperature induced higher initial dissolved O 2 concentrations, and both these phenomena explain the time delay of the acceleration of oligomerization at 5 • C compared to 25 • C (at around 25 and 13 min, respectively).In contrast, the fast decay of MVK was not influenced by the temperature, thus showing that the radical mechanism of oligomerization of MVK at 20 mM is the dominant pathway for MVK consumption at this step, and it has a very small activation energy.
Finally, when most of the MVK was consumed, all oligomer series reached their maximum intensities, with slight differences from one condition to the other as illustrated for one series in Fig. 8b.These slight differences are particularly interesting to investigate the comprehension of the initial steps of the radical mechanism.

Identification of oligomers and their initiator radicals
In Fig. 10, starting from the known mechanism of OHoxidation of MVK in the aqueous phase (Liu, 2011;Zhang et al., 2010;Huang et al., 2011), the possible oligomerization developments were added according to the experimental observations discussed above.The chemical structures of the identified oligomers are shown in Fig. 10 as well as their series names.The latter are also reported in Table 2 together with the corresponding m/z and chemical formulas obtained for the first oligomer (n = 1) of each series (i.e., iR-(MVK) n=1 ) by UPLC-ESI-MS and ESI-UHRMS analyses.
The identification was performed on the smallest oligomer of each series because the resolution and sensitivity of both instruments are optimum for 100 < m/z < 350 Da.The identifications of the most intense peaks were further confirmed by MS/MS analyses.This method of identification, applied to the synthetic oligomers of MVK formed from the V-50 initiator, allowed us to identify the most intense series of oligomers as the one shown in Reaction (R7) (in Supplement Table S2), thus showing the robustness of the method.The identification of the longer chain oligomers (n ≥ 2) was further completed by the Kendrick analysis based on the MVK (C 4 H 6 O) pattern, using the method mentioned in Sect.2.3.The oligomers were likely formed according to the mechanism shown in Fig. 6, each one being initiated by an initiator radical iR • formed from the OH-oxidation of MVK (Fig. 10).The OH-oxidation of MVK proceeds via OH-addition (1a) and H-abstraction (1b).For clarity, only OH-addition on the β-carbon of MVK is fully developed in Fig. 10, as it is more likely than α-addition, due to resonance with the carbonyl group, and steric hindrance.Furthermore, both β-and α-additions lead to isomeric forms of oligomers with the same masses which could not be discerned by mass spectrometry.In the proposed mechanism, each step leads to the formation of a radical iR • , which undergoes either O 2 addition or oligomerization with MVK, in good agreement with our experimental observations.A detailed description of the mechanism is shown for pathways 1a and 1b (Fig. 10).

Pathway 1a
OH-addition leads to the formation of radical R • , which undergoes either oligomerization with MVK (2a) or O 2 addition (2b).Pathway 2a leads to the formation of radicals R-(MVK) • n which terminate either directly to form R-(MVK) n (i.e., series S156 and S158) via the disproportionation mechanism indicated in Fig. 6 or after O 2 addition (3a) followed by H-abstraction (on HO • 2 , MVK or an oligomer) to form hydroperoxides R-(MVK) n -OOH (series S190) (Decker and Jenkins, 1985).An isomer of the latter oligomer series (S190) may also be produced from the oligomerization of ROO • radicals, which proceeds via the disproportionation mechanism (Fig. 6).However, this pathway (not shown) is of minor importance because the corresponding unsaturated series (S188) showed a very weak signal by both mass spectrometers (below 1/100 of the most intense series shown in Table 2).ROO • radicals can also undergo self reaction (4a) leading to a tetroxide which decomposes to form lowmolecular-weight compounds (LMWC) oxygenated molecular compounds (as those observed by Zhang et al., 2010) (5c), and RO • radicals (von Sonntag and Schuchmann, 1997) (5a).ROO • radicals can also react with MVK (or HO • 2 ) (4b) to form hydroperoxides ROOH.The latter can be photolyzed (5b) to yield HO • and RO • radicals.These alkoxy radicals undergo either rearrangement (6a) or decomposition (6b).The rearrangement of alkoxy radicals in aqueous medium is a well-known process (Schuchmann and von Sonntag, 1984), it produces (via pathway 6a) more stable radicals (R-OH) • which undergo either O 2 reactivity (leading to some  of the oxygenated molecular compounds observed by Zhang et al., 2010) or oligomerization with MVK via the disproportionation mechanism (Fig. 6) and form R-OH-(MVK) n (series S172, S174).The decomposition (pathway 6b, detailed in Zhang et al., 2010) produces smaller oxygenated compounds, such as methylglyoxal and acetic acid.The latter products react further with HO • radicals to form radicals (as described by Tan et al., 2012) which undergo either O 2 reactivity or oligomerization with MVK via the disproportionation mechanism (Fig. 6) and form series S140 and S142, and series S128 and S130, respectively.
In this mechanism, the formed low-molecular-weight compounds (LMWC, such as formaldehyde, methylglyoxal, glyoxal, oxalic, formic, pyruvic, malonic and acetic acids, observed by Zhang et al. (2010) during the OH-oxidation of MVK under similar conditions) can also further react with HO • and produce other iR • radicals, which can oligomerize.However, the corresponding oligomers were observed here only for acetic acid and methylglyoxal, which were among the major low-molecular-weight products observed by Zhang et al. (2010).A confirmation of these observations was brought by two extra experiments (A and A ) where 2 mM of methylglyoxal and acetic acid (Sigma Aldrich), respectively, were injected in the vessel at the same time as MVK injection.The results showed a significant increase (by 42 % and 77 %, respectively) of the corresponding oligomer series (S140 and S142 and series S128 and S130, respectively), compared to experiment A. Furthermore, experiment C (performed at 5 • C) allowed us to clearly observe the kinetics of oligomer formation (Fig. 11).The series R-(MVK) n (S156) were directly formed, immediately followed by the series R-(MVK) n -OOH (S190) (which require O 2 addition), and the series (Methylglyoxal)-(MVK) n (S142) and (Acetic acid)-(MVK) n (S130) were formed later, in very good agreement with the proposed mechanism (in Fig. 10).
The role of molecular O 2 was evidenced by the intensities of the identified series (Table 2).Comparing experiments A and D (i.e., from supersaturated to low initial O 2 conditions, respectively), after 50 min of reaction, a systematic increase of the intensities of all series not requiring O 2 addition were observed, as well as a systematic decrease of those requiring O 2 addition.This observation is also in very good agreement with the proposed mechanism (Fig. 10) described above, and it shows the prevailing role of oxygen on the processes.

Atmospheric implications
The proposed mechanism allowed for explaining the particular role of dissolved O 2 under our experimental conditions.Each iR • radical underwent competition kinetics between O 2 addition (Reaction R1) and oligomerization (Reaction R2): . Assuming that the ratio k R2 k R1 does not vary from the laboratory conditions to the atmospheric ones, one can simply predict the oligomerization to occur from the [MVK]/[O 2 ] ratio.In our experiments, the detailed study of the time profiles of O 2 and MVK together with the kinetics of oligomer formation allowed us to determine that radical oligomerization dominates over O 2 addition for [MVK]/[O 2 ] ratios (in M M −1 ) equal or higher than 32 (at 5 • C) and 54 (at 25 • C).In atmospheric waters, assuming that dissolved O 2 concentrations are saturated (i.e., at Henry's law equilibrium) everywhere from 0 to 5 km in altitude, and from −20 to +25 • C, gives a range of 190-391 µM for [O 2 ].Furthermore, taking the concentrations of unsaturated organic compounds ([UNS]) in atmospheric waters as stated in the introduction, one obtains [UNS]/[O 2 ] ratios as indicated in Fig. 12 (Ervens et al., 2012)   Table 3.Comparison of MVK atmospheric lifetimes between its gas phase reactivity only and its multiphase reactivity, taking into account its air/water partitioning at Henry's law equilibrium, its gas and aqueous phase reactivity: oligomerization is considered only in fogs and aerosol media, with k oligo values derived from our experimental results.are equal or higher than 32 or 54.It is thus concluded that radical oligomerization will always occur in wet aerosols, and in sometimes in fogs: in the most polluted fogs, where [UNS] > 6 mM.This result, added to the fact that the lifetime of wet aerosols in the atmosphere are several days, shows the extreme relevance of radical oligomerization of unsaturated organic compounds in the atmosphere.Another point of view for atmospheric implications is the fate of MVK.In general, aqueous phase OH-oxidation is known to drastically reduce WSOCs atmospheric lifetimes, compared to their gas phase reactivity (Monod et al., 2005).As it was shown in the present study, once in the aqueous phase, MVK can undergo OH-oxidation.In fogs and wet aerosols, it can additionally undergo oligomerization with a first order kinetic rate constant of k oligo = 7.6 (±0.3) × 10 −4 s −1 , (which is not temperature dependent between 5 and 25 • C) as derived in the present work from the MVK decay during oligomerization, under all conditions (Figs. 5  and 9).Although MVK is weakly water soluble, its (aerosol) aqueous phase reactivity may impact its overall atmospheric lifetime.In Table 3, MVK atmospheric lifetimes are compared between its gas phase reactivity only (taking into account both OH-oxidation and ozonolysis) and its multiphase reactivity.The latter takes into account MVK air/water partitioning at Henry's law equilibrium, and its aqueous phase reactivity: oligomerization is considered only in fogs and aerosol media.Table 3 shows that aqueous phase reactivity impacts the overall atmospheric lifetime of MVK by 2 to 13 %.Compared to these numbers, the rate of heterogeneous ozonolysis of MVK on SiO 2 or α-Al 2 O 3 particles under various relative humidity (γ = 10 −10 to 10 −9 , Shen et al., 2013) calculated for a number of 100 nm particles of 5000 particles cm −3 , would deplete its atmospheric lifetime by less than 0.00006 %.Thus, aqueous phase photooxidation seems more efficient, but this needs to be confirmed by more studies of both bulk and heterogeneous reactivity of olefin compounds.
The results obtained in Fig. 12 and Table 3 show the atmospheric relevance of (aerosol) aqueous phase reactivity of unsaturated water soluble organic compounds (even for low soluble ones like MVK), and their ability to activate radical oligomerization chemistry, which is extremely fast and is able to form macromolecules as high as 1800 Da in polluted fogs and wet aerosols.For an unsaturated compound 10 times more soluble than MVK, it is anticipated that its overall atmospheric lifetime would be depleted by 13 to 79 %, thus showing the need for further studies of oligomer formation from other relevant unsaturated compounds.In particular, the influence of other unsaturated compounds such as methacrolein on the chemistry detailed here is currently under study (G. Salque et al., personal communication, 2013).Furthermore, relevant mixtures of unsaturated compounds should be studied to investigate cross-polymerization reactions that are likely in the atmosphere.Finally, further studies are also needed to investigate the oligomer yields, their oxidizing states, and their aging (F. Siekmann et al., personal communication, 2013).

Fig. 1 .
Fig. 1.Mass spectrum (obtained using UPLC-ESI-MS from 0 to 1800 Da) for the retention time range 0 to 7 min, in the positive mode, at t = 50 min of experiment C. The most intense series (S174) is highlighted in red.Each group of peaks represents a degree of polymerization (DP).Left zoom: main series detected for DP = 4. Right zoom: MVK regular pattern of mass differences of 70.0419 Da.Another regular pattern of mass difference of 2.0157 Da corresponding to 2 H is also observed (in green in the left zoom).

Fig. 3 .Fig. 3 .
Fig. 3. Proposed structure identification of oligomers.1D 1 H NMR spectrum (A) and 2D 1 H-13 C NMR spectra (B and C) of a sample taken at of reaction during experiment A, freeze-dried and re-suspended in CDCl 3 .

Fig. 4 .
Fig. 4. UV absorption spectra of the solutions sampled during experiment A (with catalase added prior to analysis), (A): from 240 to 360 nm; and (B): diluted by 100 from 191 to 241 nm.

Fig. 6 .
Fig. 6.Scheme of the proposed mechanism of radical oligomerization of MVK in the aqueous phase, initiated by OH-oxidation.iR • is an initiator radical; MVK is the monomer or the repeating unit, n the degree of polymerization.For clarity, only the external radical additions are shown.

38Fig. 7 .
Fig.7.Mass spectra (obtained using UPLC-ESI-MS, for the retention time range 0-5 min, in the positive mode) during experiments A, E and F (i.e. at 25°C, for different initial MVK concentrations and at different reaction times (corresponding to the maximum oligomers intensity): 5 min for 0.2 mM; 15 min for 2 mM and 50 min for 20 mM.The most intense series (S174) is highlighted in red, as in Figure1).

Fig. 7 .
Fig. 7. Mass spectra (obtained using UPLC-ESI-MS, for the retention time range 0-5 min, in the positive mode) during experiments A, E and F (i.e., at 25 • C, for different initial MVK concentrations and at different reaction times (corresponding to the maximum oligomers intensity): 5 min for 0.2 mM; 15 min for 2 mM and 50 min for 20 mM.The most intense series (S174) is highlighted in red, as in Fig. 1).

Figure 9
also shows the behavior of dissolved O 2 under three different conditions (experiments D, A and C).For all of these conditions, before the injection of H 2 O 2 , dissolved O 2 was at its Henry's law equilibrium with ambient O 2 in the gas phase.When H 2 O 2 was added, dissolved O 2 concentrations increased due to the self reaction of HO • 2 and/or O •− 2 radicals in the aqueous phase (see Reactions R3 and R4 in Table e. iR-(M VK)n=1.b : The chemical structures are given in Figure 10.R• corresponds to the radical: c : Ions (and corresponding m/z) identified by UPLC-ESI-M S and/or ESI-UHRM S in experiment A; d : Intensity ratio of the series after 50 min of OH-oxidation of M VK (20mM ): , with: TSI = total Series' intensity

Fig. 10 .
Fig.10.Proposed scheme of the chemical mechanism of OH-oxidation of MVK in the aqueous phase with implementation of radical oligomerization pathways (for clarity, only external OH-addition on MVK is developed).RadH = MVK, HO2 • or molecular reaction products (in red).Radicals (in blue).LMWC: low-molecular-weight compounds.

Fig. 10 .
Fig. 10.Proposed scheme of the chemical mechanism of OH-oxidation of MVK in the aqueous phase with implementation of radical oligomerization pathways (for clarity, only external OH-addition on MVK is developed).RadH=MVK, HO • 2 or molecular reaction products (in red).Radicals (in blue).LMWC: low-molecular-weight compounds.

Fig. 11 .
Fig. 11.MVK kinetics and time profile of main oligomers series (obtained using UPLC-ESI-MS, for the retention time range 0-7 min, in both modes) at 5 • C. The series intensity is the sum of the peaks intensity.Each series intensity is expressed as a percentage of its maximum.
R2) Supersaturated (by a factor of 155 %) initial O 2 concentrations inhibited radical oligomerization by fast addition on iR • resulting in the formation of LMWC (such as acetic acid and methylglyoxal), which were further OH-oxidized and formed other iR • radicals.The fast O 2 addition reactions resulted in a fast decrease of O 2 concentrations in the vessel, faster than O 2 renewal from the gas phase and from the reactivity of H 2 O 2 , and even faster than MVK consumption.At initial MVK concentrations higher than 0.2 mM, the decrease of O 2 concentrations resulted in the dominance of Reaction (R2) after several minutes, and oligomerization started, even when O 2 concentrations were still higher than Henry's law equilibrium with atmospheric O 2 .The paradoxical role of O 2 resides in the fact that while it intensely inhibits oligomerization, it produces more iR • radicals, which contribute to O 2 consumption, and thus lead to oligomerization.These processes, together with the large ranges of initial concentrations investigated (60-656 µM of dissolved O 2 and 0.2-20 mM of MVK concentrations) show the fundamental role that O 2 likely plays in atmospheric waters.In order to scale the relative importance of Reactions (R1) and (R2) from the laboratory to the atmospheric conditions, one has to compare the rates of Reactions (R1) and (R2):

Fig. 12 :
Fig. 12: Estimated ranges of the ratios of unsaturated dissolved organic carbon concentration to oxygen concentration (in M M -1 ) in atmospheric waters.The straight lines delimit the values for which radical oligomerization dominates over O 2 addition, as determined by the present work (see text).

Fig. 12 .
Fig. 12.Estimated ranges of the ratios of unsaturated dissolved organic carbon concentration to oxygen concentration (in M M −1 ) in atmospheric waters.The straight lines delimit the values for which radical oligomerization dominates over O 2 addition, as determined by the present work (see text).

Renard et al.: Radical mechanisms of methyl vinyl ketone oligomerization 6475 2.1 Photoreactor The
Liu et al. (2009)up is based on the one described by YaoLiu et al. (2009)andLiu et al. (2012)with some modifications.It is a 450 cm 3 Pyrex thermostated photoreactor.The arc light source (LSH 601, Lot Oriel) is equipped with a 1000 W xenon arc lamp (LSB 551, Lot Oriel).A glass filter (ASTM 892 AM 1.5 standard) was used to remove the UV irradiation below 300 nm, resulting in an irradiance spectrum comparable to that of the sun at sea level, for a 48.3 • zenith angle, but approximately twice more intense.Compared to the set-up used by YaoLiu et al. (2009)andLiu et al. ( . Chem.Phys., 13, 6473-6491, 2013 www.atmos-chem-phys.net/13/6473/2013/P.

Table 1 .
Liu et al. (2012)itions of OH-oxidation of MVK, and comparison with our previous work, the study byLiu et al. (2012).In all studies, all experiments were unbuffered (free pH).The HO • concentrations were calculated assuming the steady state approximation at time 0: [ d

Table 2 .
Oligomer series identified by UPLC-ESI-MS and ESI-UHRMS in positive and negative modes.