Secondary aerosol formation from dimethyl sulfide-improved mechanistic understanding based on smog chamber experiments and modelling

Dimethyl sulfide (DMS) is the dominant biogenic sulphur compound in the ambient ::::: marine : atmosphere. Low volatile acids from DMS oxidation promote the formation and growth of sulphur aerosols, and ultimately alter cloud properties and Earth’s climate. We studied the OH-initiated oxidation of DMS in the Aarhus University research on aerosols (AURA) smog chamber and the marine boundary layer (MBL) with the aerosol dynamics, gasand particle-phase chemistry kinetic multilayer model ADCHAM. Our work involved the development of a revised and comprehensive multiphase DMS oxidation mechanism, 5 both capable of :::::: capable :: of :::: both : reproducing smog chamber and atmospheric relevant conditions. The secondary aerosol mass yield in the AURA chamber was found to have a strong dependence on the reaction of methyl sulfinic acid (MSIA) and OH :::::: causing :: a :::: 58.9 ::::::: percent ::::::: increase :: in ::: the ::::: total ::: PM : at low relative humidity (RH), while the autoxidation of the intermediate radical CH3SCH2OO forming hydroperoxymethyl thioformate (HPMTF) proved important at high RH:::::::::: temperature::: and:::: RH ::::::::: decreasing ::: the :::: total ::: PM ::: by :::: 66.2 :::::: percent. The observations and modelling strongly support that a liquid water film existed on 10 the Teflon surface of the chamber bag, which enhanced the wall loss of water soluble intermediates and oxidants ::::::: dimethyl ::::::: sulfoxide :::::::: (DMSO), MSIA, HPMTF, SO2, ::::::::::::: methanesulfonic:::: acid:MSA, SA and H2O2. The effect caused a:::: 75.0:::: and :::: 92.8 :::::: percent decrease in the secondary aerosol mass yield obtained at both dry (0-12 % RH) and humid (50-80 % RH) conditions : , :::::::::: respectively. Model runs reproducing the ambient marine atmosphere indicate that OH comprise a strong sink of DMS in the MBL :::::::::: (accounting ::: for :::: 32.0 :::::: percent ::: of ::: the :::: total :::: sink :::: flux :: of :::::: DMS), although less important than ::: the :::::::: combined ::::: effect ::: of 15 halogen species Cl and BrO ::::::::: (accounting ::: for :::: 25.7 :::: and :::: 40.4 ::::::: percent, :::::::::: respectively). Cloudy conditions promote the production of SO2− 4 particular mass (PM) from SO2 accumulated in the gas-phase, while cloud-free periods facilitate MSA formation in the deliquesced particles. The exclusion of aqueous-phase chemistry lowers the DMS sink as no halogens are activated in the sea spray particles, and underestimates the secondary aerosol mass yield by neglecting SO2− 4 and MSA PM production in the particle phase. Overall, this study demonstrated that the current DMS oxidation mechanisms reported in literature are 20 inadequate in reproducing the results obtained in the AURA chamber, whereas the revised chemistry captured the formation,


Chamber wall effects -gas to wall partitioning
Comparing secondary aerosol yields from chamber experiments performed at high and low RH revealed a significant decrease 110 in the overall PM during humid conditions. The effect may be caused by the formation of a liquid film on the chamber walls ( Fig. 1c), lowering the gas-phase concentration of oxidants and water soluble products from the DMS oxidation mechanism.
Adsorption of water onto Teflon bag surfaces is known to occur in smog chamber experiments (Sumner et al., 2004), and may lead to condensation during high RH conditions (Svensson et al., 1987). Wall contamination including HONO and HNO 3 from NO x exposure increases the uptake by fixating the adsorbed water molecules (Sumner et al., 2004). Separate experiments 115 examining 1-butanol oxidation by OH was :::: were used to quantify the thickness of the Teflon bag water film. Analogous to the method by Song et al. (2019) the OH concentration was estimated based on the decay of 1-butanol. Humid chamber conditions decreased OH production and slowed down the butanol oxidation (Fig. 1a). The effect likely arose from the enhanced water uptake to the Teflon walls, which lowered the gas-phase concentration of H 2 O 2 and thus the production of OH (Fig. 1b).
The water content needed to suppress OH formation to match the experimental observations at ∼60% RH and 293K, in the 120 centre of the chamber, corresponded to approximately 30 g m −3 or a liquid film layer of ∼10 µm on the chamber surface (see supplementary Fig. S1). The predicted concentration requires water to condensate on the Teflon bag. While Sumner et al. (2004) rejects the idea that condensation is possible, chamber experiments by Svensson et al. (1987) showed that water condensates on both polluted and clean Teflon surfaces at RH conditions beyond 70 %. The authors also observed the formation of a few (red) :::: and ::::: BUT2 :::: (blue) : and B the OH concentration at varying RH with and without the influence of a liquid water film ::::::: (woLWC : : :::::: without :::: liquid ::::: water :::::: content). The water content changed in accordance with varying temperature and RH, panel C. Outside cooling of the chamber may affect the water adsorption on the Teflon surface.
mono-layers of water during low RH (<5%) conditions. During dry experiments in the AURA chamber a small increase in RH 125 was observed caused by water contamination from instrument sampling (Table 1).
We performed several sensitivity tests with different water film thickness for the modelled DMS experiments (see Fig. S2-4).
From this, we concluded that with an effective water film concentration of 10 µg m −3 , ∼ 1% of a water monolayer, the model generally capture the observed PM mass formation and SA to MSA PM mass ratio during the dry experiments. For the humid DMS experiments the water concentration on the walls need to be ∼ 10 g m −3 and ∼ 500 g m −3 for the 293 K and 273 K 130 experiments, respectively. The found optimal values of the wall liquid water content are on the same order of magnitude as the values estimated based on the butanol experiments performed at similar RH and temperature conditions (Table 2).
On average the temperature proved smaller at the chamber bag surface as opposed to the chamber bag centre. The effect was caused by the temperature regulation setup cooling the chamber bag from the outside (Fig. 1c) In all experiments, the AMS measurements reveal that the formed PM contains a substantial mass fraction of ammonium. We expect that the ammonium mass mainly originates from NH 3 (g) gradually evaporating from the chamber walls and to a smaller extent from NH 3 (g) leaking into the chamber. To be able to capture the observed secondary ammonium mass formation and the new particle formation (Sect. 2.4) in all experiments, we assumed an initial pool of ∼100 µg m −3 ::::: (mass ::: per ::: air ::::::: volume) dissolved ammonium sulfate on the chamber walls and that the NH 3 (g) concentration :::::: mixing :::: ratio : in the air outside the 145 chamber was 2.0 ppb v . The pool of ammonium and sulfate on the chamber walls can be motivated by previous experiments in the AURA chamber with ammonium sulfate seed particles, SA(g) and NH 3 (g) depositing on the walls. While the sulfate (S(VI)) on the walls can be considered to be non-volatile, the semi-volatile ammonium (N(III)) will partition between the liquid film on the walls and the gas-phase in different extent which depend on the wall liquid water content, the acidity and the temperature. Hence, the N(III) concentration on the chamber walls most likely differs from one experiment to another. This 150 motivates why we decided to use the initial N(III) wall concentration as a unknown model fitting parameter, when comparing the modelled and observed new particle and secondary ammonium mass formation. This resulted in initial N(III) : S(VI) on the walls ranging between 1.2 and 1.8 for the different experiments (  S5 shows the modelled NH 3 (g) concentration and the NH 3 concentration on the chamber walls for all modelled experiments.

155
The modelled NH 3 (g) concentration gradually decreased during most of the experiments because of the NH 3 uptake to the formed aerosol particles and the SA and MSA deposition on the chamber walls, which resulted in a gradually more acidic liquid water film. However, in DMS4, DMS6 and DMS7 the modelled NH 3 concentration increases slightly, especially during the end of the experiments. This is because the leakage of NH 3 (g) into the chamber become ::::::: becomes : larger than the sink of NH 3 (g) to the particle phase.

160
The first order wall loss rates of gases were modelled using the theory proposed by McMurry and Grosjean (1985) (Eq. 1).
A V is the chamber surface area A to volume V ratio, α w the vapour wall mass accommodation coefficient, ν the mean thermal speed of the molecules, k e the coefficient of eddy diffusion and D the molecular diffusivity. k e and α w are the two key unknown parameters in Eq. 1. In the AURA smog chamber k w has been estimated to ∼10 −3 s −1 for low volatility highly 165 oxygenated organic molecules (HOMs) formed from α−pinene ozonolysis (Quéléver et al., 2019). For a typical α−pinene HOM monomer or SA and α w ≥ 10 −4 this corresponds to k e ≈ 0.02 s −1 . This estimated value of k e can, e.g., be compared with the value reported by Zhang et al. (2014) of 0.015 s −1 , for a Teflon chambers with a volume of 28 m 3 . Fig. S6 in the supplementary material illustrates how k w varies as a function of k e and α w for a compound with similar molecular properties as MSA and SA. For α w ≥ 10 −4 the vapour wall losses are entirely governed by the chamber mixing 170 and the molecular diffusion through a thin surface layer next to the chamber walls, while for smaller values of α w the sur-face reactivity also limits the wall uptake (McMurry and Grosjean, 1985). By default the chamber wall loss of the important condensable vapours (SA, MSA, HNO 3 , NH 3 ) and the highly water soluble H 2 O 2 were assumed to be limited exclusively by the chamber mixing and molecular diffusion across the surface layer next to the chamber walls (i.e. using α w = 1.0). For the important intermediate DMS oxidation products, i.e. DMSO, DMSO 2 , MSIA and HPMTF, the wall mass accommodation 175 coefficients were by default set to 10 −5 . Wall mass accommodation coefficients ∼10 −5 have previously been suggested when modelling wall losses of volatile and semi-volatile organic molecules in Teflon smog chambers (Matsunaga and Ziemann, 2010;Zhang et al., 2014). For DMS, O 3 , SO 2 and NO 2 : , we used a default α w of 10 −7 . This α w value is approximately one order of magnitude greater than the value reported by McMurry and Grosjean (1985) for O 3 , SO 2 and NO 2 , during dry chamber conditions. The relatively low α w values used for DMS, compared to the DMS oxidation products, was motivated by the 180 reported low mass accommodation coefficients for DMS uptake to particles and cloud droplets (Hoffmann et al., 2016;Zhu et al., 2006;Kreidenweis et al., 2003) and the observed and modelled DMS trends during the humid experiments. When using the default α w values k w become ∼ 10 −3 s −1 for MSA and SA, ∼ 7 · 10 −4 s −1 for the intermediate DMS oxidation products, and ∼ 3 · 10 −5 s −1 for DMS, O 3 , SO 2 and NO 2 , for a fully inflated chamber (V = 5 m 3 ). We performed several model sensitivity simulations with α w values ±1 order of magnitude from the default values for DMS, 185 DMSO, DMSO 2 , MSIA, HPMTF, O 3 and SO 2 , for all experiments listed in Table 1-2 (see Fig. S7-S21 in the supplementary material). From this we conclude that the modelled secondary aerosol formation is especially sensitive to the rate at which O 3 partition :::::::: partitions to the liquid water film on the chamber walls. The uptake of O 3 in the liquid water film governs the uptake and oxidation of MSIA(g) on the walls, both during the dry and humid experiments and the oxidation of DMS during the humid experiments ( Fig. S7-11). Also lower wall loss rates (i.e. lower α w ) for the intermediate oxidation products, especially MSIA,190 has relatively large impact on the modelled secondary aerosol mass formation. Decreasing α w with one order of magnitude for the intermediate oxidation products result ::::: results in increasing particle mass formation with a factor of 1.5-3 in experiment DMS1-6, but has very minor impact on the modelled particle mass (PM) formation during the humid and cold experiment conditions (Grosjean, 1985), and it is possible that the liquid water content on the chamber walls influence α w . However, for the base case model simulations we chose to use the same α w values for both the dry and humid experiments.

215
H * = LW C · R * · T · H cp ::::::::::::::::::::: (4) H cp denote :::::: denotes : the Henry's law solubility in unit kg atm −1 mol −1 , R * is the universal gas constant (82.06 cm 3 atm mol −1 ; K −1 ), T is the temperature in K and LW C is the aqueous phase liquid water content, which in the case of gas-wall partitioning is equal to the effective wall water concentration given in unit kg cm −3 . 220 Table S2 gives the values of H cp and α w for all species that were considered to undergo phase transfer between the gas and aqueous phase in the multiphase chemistry module.

Multiphase chemistry
2.3 ::::::::: Multiphase ::::::::: chemistry The multiphase chemistry mechanism was developed based on the DMS gas-phase oxidation scheme in the Master Chemical Mechanism version 3.3.1 (MCMv3.3.1) (Jenkin et al., 1997(Jenkin et al., , 2015Saunders et al., 2003), the CAPRAM DMS Module 285 1.0 (DM1.0) (Hoffmann et al., 2016), and a subset of reactions from the multiphase halogen chemistry mechanism CAPRAM Halogen Module 2.0 (HM2.0) (Braeuer et al., 2013). Additional reactions and rate constants were implemented from individual studies (Turnipseed et al., 1995;Kukui et al., 2003;Wu et al., 2014;Cao et al., 2013;Berndt et al., 2019Berndt et al., , 2020. A complete list of all reactions can be found in the supplementary material Table S1. In total the mechanism include 922 species and 2946 reactions of which 100 reactions are phase transfer reactions (forward and backward), 2542 are gas-phase reactions and 304 290 are aqueous phase reactions. However, the majority of the reactions are only relevant for the atmospheric model simulations (Sect. 3), including 1900 gas-phase reactions which is part of the MCMv3.3.1 isoprene chemistry scheme (Jenkin et al., 2015) (not listed in Table S1), 411 halogen gas-phase reactions, 216 halogen aqueous phase reactions and 58 halogen phase-transfer reactions. The concentrations of H + and OH − which are involved in many of the aqueous phase reactions were updated outside the multiphase chemistry mechanism in the ADCHAM thermodynamics module (Roldin et al., 2014) and were not 295 considered to be influenced by the irreversible aqueous phase chemistry. Hence, [H + ] and [OH − ] were multiplied directly into the reaction rate expressions in the multiphase chemistry mechanism and are not explicitly included as reactants and products in the reactions listed in Table S1. The multiphase chemistry mechanism was generated with the Kinetic PreProcessor (KPP) (Damian et al., 2002) and solved with the ordinary differential equation solver DLSODE (Hindmarsh, 1983). Figure  2.3.1 New particle formation 2.4 ::: New ::::::: particle ::::::::: formation The

Particle wall losses
The particle wall losses of particles with 0 to 3 elemental charges were calculated using the particle wall loss parametrizations described in Roldin et al. (2014). For this, the model takes into account the initial fraction of neutral and charged new particles, which was calculated with ACDC, and the evolution of the aerosol particle charge distribution over time. The key unknown parameters which govern the particle wall losses of neutral and charged particles are the friction velocity (u * ) and the electric 320 field strength (E f ield ). We used a fixed value of u * of 0.013 for all experiments. This value was chosen in order for the particle wall losses to be consistent with the gas wall loss rates calculated with Eq. 1. I.e., for a hypothetical non-charged particle or gas molecule of 0.6 in diameter both parametrizations give first order wall loss rates of ∼10 −3 . E f ield (Table 2) were set to different values in the range of 0.7-5 , depending on the observed and modelled particle number and volume concentration loss rates. The air ion concentration in the chamber was calculated from the steady state solution of the ion balance equation, taking 325 into account the ion formation rate, ion-ion recombination, condensation sink and wall losses of air ions (Kirkby et al., 2016) . The steady state air ion concentration of positively and negatively charged ions (n + − ), was used to derive the particle charge distribution by solving a system of differential equations: ] i denote the number concentration of particles with 0, 1, 2 or 3 elemental charges in each size bin (i). k qi are the aerosol particle ion attachment coefficients (unit ), which depend on the size and sign of the particle charge (q) and air ions (Fuchs, 1963;Hoppel and Frick, 1986). For example, k −2i represent the attractive air 340 ion attachment coefficients for an air ion approaching a particle, in size bin i, with 2 elemental charges when the sign of the particle and ion charge are different, while k 1i represent the repellent air ion attachment coefficient for an air ion approaching a particle with 1 elemental charge, when the sign of the charge of the ion and particle are the same. Equation 5-8 assumes an even distribution of positive and negative charged air ions (i.e. . Particles with more than 3 elemental charges were not considered in the model. Thus, the 1/2 fraction of all particles with 3 elemental charges which in reality 345 would have gained 4 elemental charges upon collision with air ions were assumed to keep their 3 elemental charges. Fig. S25 in the supplementary material illustrates how the modelled particle charge distribution and particle wall losses evolve during experiment DMS2. To evaluate the updated multiphase DMS chemistry for atmospheric realistic conditions ADCHAM was set up to reproduce the pristine marine environments of the open ocean. For this purpose an emission estimate of relevant gas-phase and particle-phase species was implemented based on the model scenarios in the work by Braeuer et al. (2013) and Hoffmann et al. (2016). While halogens are of insignificant importance in the AURA smog chamber, they comprise an important oxidant in the ambient atmosphere. Bromine and chlorine released from sea salt aerosols interact strongly with sulphur compounds including 355 DMS (Braeuer et al., 2013), and alter the mechanism presented in Figure 2. Consequently, the CAPRAM Halogen Module was implemented to address both the gas-phase and aqueous-phase oxidation of DMS by halogen compounds. A base run scenario (named AtmMain) reproduced the movement of an air parcel along a marine environment trajectory for 120 hours (Fig. 5). Starting at midnight the simulation included eight in-cloud periods, four during the day and four at night. ADCHAM clouds formed and evaporated during 75 minute adiabatic cooling and warming periods, respectively, in which the RH changed 360 slowly over time. The in-cloud residence time was chosen in accordance with the study by Pruppacher and Jaenicke (1995).
The maximum cloud liquid water content was assumed to be 0.3 g m −3 . The last cloud period included a rain event, with below cloud particle scavenging described by the parameterisation by Laakso et al. (2003) and gas scavenging described by a parameterisation from Simpson et al. (2003). Cloud conditions were introduced to illustrate the effect of multiphase DMS chemistry, and varying UV light intensity to reproduce the oxidation capacity of the marine atmosphere during both day and 365 night-time conditions. The wet deposition of particles and gases by rain was introduced to spark a NPF event. Between incloud periods the RH was kept at 90%. Consequently, the aerosols formed were treated as deliquesced particles receptive to aqueous-phase chemistry. This setup was essential to capture the gas-phase concentrations of halogens bromide and chloride mainly formed by halogen activation inside the particles. Sea spray emissions were based on a temperature and wind speed dependant parameterisation by Salter et al. (2015). Wind speed was kept fixed at 8 m s −1 in accordance with the global annual 370 average marine wind speed (Kent et al., 2013). Besides the base setup four sensitivity runs were performed to validate the effect of varying atmospheric conditions. The first (named 'PolAtm') represented a polluted marine environment with high NO x and O 3 gas-phase concentrations. The second (named 'woCloudAtm') did not include any in-cloud periods. The third run (named 'woAqAtm') disregarded all aqueous-phase chemistry. The final run (named 'lowWindAtm') kept the wind-speed at 2 m s −1 .
For the atmospheric model simulations the particle number size distribution was represented by 200 fixed size bins in the 375 diameter range 1.07 nm to 10 µm.
In the abstraction pathway, the large increase in MSA and SA production by the MSIA + OH pathway was counterbalanced to match observations by implementing a newly discovered DMS autoxidation pathway (Veres et al., 2020). The pathway SO 2 comprised the main product in the abstraction pathway, formed mainly through the thermal decomposition of CH 3 SO 2 .

Aqueous-phase DMS oxidation
During high RH conditions in the chamber experiments or cloud cover in the MBL, water affects the gas-phase oxidation of DMS (Hoffmann et al., 2016). In the atmosphere the aqueous-phase chemistry proceeds in cloud droplets and deliquesced particles (Seinfeld and Pandis, 2016). In the chamber the water adsorbs to the Teflon bag and forms a thin liquid film (see section (66.2 ::::::: percent ::::::: decrease :: in :::: the :::: total ::: PM :::::: during :::::::::: experiment ::::::: DMS6). In atmospheric relevant conditions HPMTF may oxidise ::: may ::: be ::::::: oxidized : in cloud droplets analogous to compounds with similar functional groups (Doussin and Monod, 2013). While this reaction pathway has not been validated in the literature, we propose a mechanism that incorporates the aqueous-phase 455 OH-initiated oxidation of HPMTF to HOOCH 2 SCO. The subsequent transfer of HOOCH 2 SCO to the gas-phase strongly increases the HPMTF derived production of SO 2 and thus SO 2− 4 .

Chamber contaminants
Smog chamber experiments have the advantage of elucidating atmospheric phenomena in controlled temperature, RH, UV light and VOC concentration conditions. Even so, contamination from walls and instrument sampling affects the outcome 460 (Sumner et al., 2004). pollutants has a high impact on DMS chemistry (Barnes et al., 2006). Both by direct oxidation of intermediate species in the DMS oxidation mechanisms and indirect by the formation of ozone. Chambers exposed regularly build up on the walls from the heterogeneous reaction of and water on the Teflon bag surface (Svensson et al., 1987). The wall pool comprise an additional source of during chamber experiments. When exposed to UV-lights photodissociates to form NO and ground state atomic oxygen that combines rapidly with molecular oxygen to yield ozone. In the gas-phase, ozone facilitates during which is taken up by the water film on the Teflon bag. In the aqueous phase, ozone drives the uptake of DMS by rapid oxidation (see section 3.2). concentrations were estimated based on the observed ozone formation in the chamber experiments.
Matsunaga and Ziemann (2010) described the direct uptake of gas-phase molecules and particulate matter to Teflon surfaces.

470
During long term usage, the concentration of certain sticky species may be build up on the chamber bag in spite of thorough cleaning. Previous experiments performed in the AURA chamber have used as seed particles. Therefore, we motivate the presence of a coating from seed particles deposited on the chamber walls. Depending on the pH of the aqueous solution a fraction of the dissolved will deprotonate and form (aq) which evaporates and forms a continuous source of (g). Different smog chamber studies have examined NPF from ternary nucleation involving SA, water and (Benson et al., 2011;Kürten et al., 2016) 475 , and all demonstrate that enhance SA nucleation. The influx of to the gas-phase was essential to reproduce the particle number (PN) concentration and PM observed by PSM/WCPC and HR-ToF-AMS measurements, respectively. The increasing LWC on the chamber walls during humid and cold conditions resulted in lower concentrations (Fig. S5), which enables the model to capture the observed lower particle number concentrations during the humid and cold experiment (DMS7) (Fig. S32).

Model simulations
The ADCHAM model was constrained to conditions specific for the experiments performed in the AURA smog chamber. Fig.   S26-32 compares the modelled and measured gas and particle concentrations for all AURA DMS experiments listed in Table   485 1-2. Different sensitivity runs were performed for three representative experiments (DMS2, DMS6 and DMS7) to highlight the effects of our revised DMS multiphase chemistry mechanism, compared to previous studies. An additional run examined the implemented mechanism in atmospheric relevant conditions. Table 3 provides an outline of each simulation while results are given in Table 4. = ::::: 0.96) concentrations in the AURA smog chamber at 293K and high RH conditions (Fig. 3b,c). In this context, it should be noted that the HR-ToF-AMS PM concentration was corrected using the SMPS particle volume concentration (PV) (Fig. 3e) and HR-ToF-AMS aerosol density (?) ::::::::::::::::: (Rosati et al., 2021b) analogous to the method by Bahreini et al. (2009) to account for the uncertainties = ::::: 0.36) : (Fig. 3a,d). We motivate the PN concentration underestimate in the first two hours of the experiment by the presence of organic contamination (Fig. 3b). A similar effect is seen in the modelled PV (Fig. 3e), which likewise does not consider the influence of organics in the aerosol particle formation. A water concentration of 10 g m −3 corresponding to a ∼3 µm water film was implemented based on model sensitivity runs (Fig. S3). The found optimal LWC 500 are within the same order of magnitude as the value predicted by the butanol experiment (see section 2.2.1). Consequently, H 2 O 2 partitioned strongly to the aqueous-phase reducing gas-phase concentration of HO 2 with a factor of ∼ 4 compared to the dry experiments ( Fig. S22-23). With the reduction in HO 2 , MSA production from the CH 3 SO 3 + HO 2 reaction decreased correspondingly -:: a :::::: change :::: from :::::::: 4.53·10 6 cm −3 s −1 : to ::::::::: 9.770·10 5 cm −3 s −1 :: in ::: the ::::: MSA :::::: source ::: flux. The reduced conversion from CH 3 SO 3 to MSA favoured the decomposition of CH 3 SO 3 to SO 3 and lowered the MSA/SO 2− 4 ratio ::: from ::::: 4.34 :: to :::: 1.72 compared to experiments performed in dry conditions (Table 4). The overall mass yield was strongly influenced by the uptake of DMSO and MSIA to the water film. Since the oxidation of MSIA by OH was implemented as an alternative source of CH 3 SO 2 (an important precursor of SA and MSA) in the gas-phase mechanism (Kukui et al., 2003), SA and MSA production and concentrations in the gas-phase were reduced in accordance with the DMSO and MSIA dissolution in the wall aqueous film (Fig. S22-23). Thus, the presence of a water film on the chamber bag surface strongly altered the DMS oxidation product 510 ratio and total PM yield ::::: (92.8 :::::: percent :::: total ::: PM :::::::: decrease :: in ::::::::: experiment ::::::: DMS6).
PTR-MS measurements indicated a strong DMS oxidation in the first few hours of the experiment (Fig. 3c). While OH 520 initiated gas-phase oxidation alone could not explain the observed trend, the aqueous-phase oxidation by O 3 improved the fit (Fig. S8). DMS has a small Henry's law solubility and the majority resides in the gas-phase (Campolongo et al., 1999).
However, the O 3 present in the water film ensured a steady conversion of DMS to DMSO and hence a flux of DMS between the gas-phase and aqueous-phase. The uptake of O 3 to the water film was apparent from the decrease in O 3 concentration observed in the experiment (Fig. 3c). In dry conditions, O 3 was found to increase gradually from NO x contamination (see 525 section 2.2.3). These results strongly advocate the presence of a chamber wall water film. Despite the aqueous-phase uptake O 3 remained abundant in the gas-phase, thus favouring the CH 3 SO + O 3 reaction and promoting SO 2 production. The reduced importance of SO 2 in gas-phase SA formation worked to lower the overall SA and MSA production, since SO 2 was taken up by the water film. Consequently, the gas-phase O 3 abundance decreased SA production and hence NPF.
MSIA oxidation by OH was essential to capture the observed onset in NPF from PSM and SMPS measurements (Fig. 3a).
The autoxidation of the CH 3 SCH 2 OO radical leading to the formation of HPMTF exerted a strong influence on the PM yield 540 in the humid chamber experiments, :::::::: lowering ::: the :::: total ::: PM ::: by :::: 66.2 :::::: percent :: in :::::::::: experiment :::::: DMS6. The effect was evident when implementing the rate limiting H-shift reaction constant as suggested by Yin et al. (1990) and Veres et al. (2020), respectively (Table 4), as opposed to rate proposed by Berndt et al. (2019). PM yields decreased by 48 percent in accordance with the study by Yin et al. (1990), and increased by 90 percent in accordance with the study by Veres et al. (2020). The difference coincides with the production and accumulation of HPMTF in the gas-phase. In consequence of the high sensitivity of the reaction rate to 545 the model outcome, more detailed studies of the HPMTF autoxidation pathway are needed to overcome the uncertainty related to the current mechanism.
Overall, modelling the OH-initiated DMS oxidation at high RH strongly indicated the presence of a water film on the Teflon bag. This crucial finding comprise an important consideration when performing smog chamber experiments, and could help to explain the complex interaction between gaseous species and chamber walls. The effect, however, remains uncertain (Svensson 550 et al., 1987;Sumner et al., 2004) and requires further investigation to be validated.

Dry chamber
ADCHAM captures the difference in secondary aerosol PM concentration between dry and humid experiments performed in the AURA smog chamber (Fig. 4c). The water film concentration was kept low :: at :::: 10 −5 : gm −3 in accordance with the study by Svensson et al. (1987), and had little ::: less effect on the DMS oxidation :::::::: compared :: to ::: the ::::: humid ::::::::::: experiments :::: (75.0 :::: and :::: 92.8 555 :::::: percent ::::::: decrease ::: in ::: the :::: total ::: PM ::::: yield ::::::: obtained ::: in ::: dry ::: and :::::: humid ::::::::: conditions, ::::::::::: respectively). Consequently, almost all DMSO and a large fraction of MSIA remained in the gas-phase ( Fig. S17-18, Fig. S22) and increased the total MSA and SA PM. The limited uptake of MSIA to the water film made the MSIA + OH reaction essential to reproduce the MSA PM production. By contrast, said reaction had little impact on the overall PM yield in high RH conditions since MSIA dissolve and react almost exclusively in the aqueous-phase. While different rates has been reported for the oxidation of MSIA by OH in literature, 560 1.0·10 −12 cm 3 molecule −1 (Lucas and Prinn, 2002) and 1.6·10 −11 cm 3 molecule −1 (Yin et al., 1990), the increased rate of 1.0·10 −10 cm 3 molecule −1 proposed by Kukui et al. (2003) offered a good agreement with the observed PM concentration ( ::: R 2 : = ::::: 0.99) : (see supplementary Fig. S33). This result entails that the main pathway leading to gas-phase MSA production in low RH chamber or cloud-free atmospheric conditions proceeds not by the abstraction pathway as assumed in previous studies but via the addition pathway. A similar conclusions was drawn in a study by Glasow and Crutzen (2004). The effect became more 565 profound when incorporating the production of HMPTF by autoxidation (Veres et al., 2020). In this case, the rate proposed by Berndt et al. (2019) ensured a strong decrease in SA and MSA production formed via the abstraction pathway :: (38 ::::::: percent ::: PM ::::::: decrease :::::: during :::::::::: experiment :::::: DMS2). Consequently, the OH-initiated addition to DMS and subsequent oxidation pathway proved the main source of secondary aerosol PM in the AURA chamber at low RH conditions. The effect of the MSIA + OH reaction rate on the PM yield is evident from the results in Table 4. The reaction rate proposed by Yin et al. (1990) and Lucas 570 and Prinn (2002) decreased the PM yield by 28 and 36 percent, respectively, as opposed to the rate suggested by Kukui et al. The H 2 O 2 (and hence HO 2 ) partitioned readily to the aqueous-phase in the humid experiments but not in the dry. The difference is evident from the MSA/SO 2− 4 PM ratio :: of :::: 4.24 :::: and :::: 1.72 observed in dry and humid experiments, ::::::::::: respectively the substantially lower H 2 O 2 and hence HO 2 gas-phase concentration limited the CH 3 SO 3 + HO 2 reaction and favoured the decomposition of CH 3 SO 3 to SO 3 and hence SA. Consequently, SO 2− 4 PM production matched that of MSA in humid conditions. The initial O 3 concentration appeared low during dry conditions, but increased consistently as each experiment 580 progressed (Fig. 4a). This response incites a constant influx of NO x which may arise from a HONO wall pool (see section 2.2.3) and NO 2 contamination from air leaking in from outside the Teflon bag. The decrease in the gas-phase O 3 concentration observed in humid conditions was associated with the uptake of O 3 to the Teflon bag water film. Our model results indicate that such uptake was of no or insignificant importance during dry conditions (Fig. S7).
The uptake of gaseous species to the aqueous-phase increased in accordance with the water film concentration and Henry's law theory (Fig. S24). Most important was the decrease in the gas-phase NH 3 that worked to suppress the onset and total 605 concentration of NPF (Fig. S5).

Atmospheric implication
DMS decay dominated :::::::: dominates : during the day when the UV light intensity and thus the gas-phase concentration of oxidant species OH, Cl and BrO was : is : high (Fig. 5b). to the submicron aerosol number concentrations in the marine atmosphere. This statement stands in stark contrast to the sea spray parameterisation by Salter et al. (2015) utilised in this work. Therefore, it is plausible that the emission of the sea salt aerosols and thus the halogen radical concentration may be overestimated in the model.
Halogen activation inside the sea salt aerosols comprised the main source of both Br and Cl in the gas-phase. Analogous to the work by Braeuer et al. (2013) chloride ions were activated by the iodine species HOI in the deliquesced sea-salt particles, forming ICl capable of transferring to the gas-phase. BrCl formed by bromide activation also partitioned to the gas-phase.
Both species photodissociated proportional to the UV light intensity causing Br and Cl concentrations to peak at midday in non-cloud conditions. Br and Cl gas-phase radicals reacted strongly with ozone to form BrO and ClO. While ClO had little effect on the DMS decay, BrO comprised the main oxidant and dominated the sink flux of DMS (Fig. 5b). During in-cloud periods the halogen activation terminated and both bromide and chloride stayed in the aqueous-phase. Consequently, neither

645
finding is consistent with the study by Hoffmann et al. (2016). Although essential for the formation of new particles, SA produced in the gas-phase proved insignificant to the overall formation of SO 2− 4 PM. During in-cloud periods, SO 2 partitioned strongly to aqueous-phase. The subsequent oxidation by O 3 and H 2 O 2 comprised the main source of SO 2− 4 PM throughout the simulation. MSA in the gas-phase formed almost exclusively via the CH 3 SO 3 intermediate. However, the strong uptake of MSIA to the particles made the system less sensitive to the MSIA + OH reaction. As a results, the gas-phase MSA production 650 proved insignificant to the total MSA PM concentration. Instead, MSA was almost exclusively formed via the aqueous-phase oxidation by O 3 in both the deliquesced particles and cloud droplets in between and during in-cloud periods, respectively (see supplementary Fig. S38). The uptake of MSIA to the aerosol particles, and thus the importance of the MSIA + OH reaction, does, however, depend strongly on the Henry's law solubility of MSIA, which is highly uncertain. The Henry's law solubility utilised in this work was based on COSMOtherm calculations and exceeded that suggested in previous studies (Barnes et al., 655 2006;Hoffmann et al., 2016) by more than one order of magnitude. In this context, it is important to note that the Henry's law solubility of MSIA utilised by Barnes et al. (2006) and Hoffmann et al. (2016) were assumptions based the Henry's law solubility of DMSO and MSA. The extent of the aqueous-phase MSA production increased in accordance with the increased uptake of MSIA to the particle-phase. This finding contradicts the work by Glasow and Crutzen (2004), in which the gasphase production of MSA comprised more than half of the total MSA yield. The difference was unexpected, since Glasow and 660 Crutzen (2004) implemented the MSIA + OH reaction and rate based on the experimental work by Kukui et al. (2003) also utilised in this paper. The effect was explained by the aqueous-phase oxidation of MSIA by O 3 , not considered in the study by Glasow and Crutzen (2004). Without this reaction, MSA in the aqueous-phase formed primarily by OH oxidation and thus solely during in-cloud periods when the aqueous OH concentration was high.

670
Due to its high stability, HPMTF accumulated in the gas-phase during clear conditions, an insignificant fraction being oxidised by OH. Analogous to the field measurements by Veres et al. (2020) HPMTF was taken up by aqueous cloud particles during in-cloud periods. Veres et al. (2020) claimed that HPMTF may likely contribute directly to particle formation and growth, but failed to provide adequate prove for this finding. COSMOtherm calculations quantifying the Henry's law solubility of HPMTF render its contribution to particle growth very insignificant, and neither ADCHAM nor the AURA chamber 675 experiments suggests that HPMTF resides in the aerosol-phase. Furthermore, HPMTF is unlikely to contribute to NPF. This is caused by the fact that the functional groups in the HPMTF molecule is a carbonyl group and a hydroperoxide group. Based on COSMOtherm calculations , it has been well-established that highly oxygenated organic molecules (HOMs) consisting of multiple carbonyl-, hydroperoxy-and peroxy acid functional groups may not have low vapour pressures, despite their high O:C ratios, especially if the HOM contains only few H-bond donors (Hyttinen et al., 2021). Furthermore, 680 quantum chemical calculations have shown that in order for a given organic highly oxidized organic molecule to be involved sulfuric acid based new particle formation it must contain several strong binding moieties such as carboxylic acid groups (Elm et al., 2017). The importance of HPMTF in new particle formation is further diminished by the fact that the monomer can be stabilised by an intramolecular hydrogen bond between the -OOH and O=C-groups. This implies that the intramolecular hydrogen bond needs to be broken before HPMTF can cluster with another molecule. Using quantum chemical calculations 685 it has been shown that such intramolecular hydrogen bonds hinder strong cluster formation between sulfuric acid and highly oxidized C 6 H 8 O 7 peroxy acid products formed from cyclohexene ozonolysis (Elm et al., 2015(Elm et al., , 2016. HPMTF may, however, indirectly contribute to particle growth via the aqueous-phase oxidation by OH forming SO 2 and hence SO 2− 4 . At present, no data on the aqueous-phase oxidation of HPMTF is available in the literature. Instead, a rate constant of the OH-initiated oxidation of HPMTF was estimated based on experimental data of compounds with similar functional groups (Doussin and Monod,690 2013). Unlike the slow gas-phase oxidation, the implemented rate ensured a strong sink flux of HPMTF in the aqueous-phase ::::::: (-1.3·10 2 ::::::::: molecules ::::: cm −3 ::: s −1 :::: and ::::::: -8.0·10 3 :::::::: molecules ::::: cm −3 :::: s −1 ::::: mean ::::::: HPMTF :::: sink ::: flux :: in ::: the ::::::::: gas-phase ::: and ::::::::::::: aqueous-phase, :::::::::: respectively). The formed intermediate HOOCH 2 SCO partitioned readily to the gas-phase and oxidised to form SO 2 . Consequently, the aqueous-phase oxidation of HMPTF may increase the conversion of HPMTF to SO 2 and thus the SO 2− 4 PM production. However, further investigations are required to overcome the uncertainties linked to said reaction. The weak oxida-695 tion of HPMTF by OH and consequent accumulation in the gas-phase could potentially reduce the local effect of DMS derived aerosol formation, as HPMTF may be transported large distances before partitioning to and oxidise in aqueous cloud particles.
The PN concentration in the NPF event that followed decreased accordingly, as the condensation sink of species SA and NH 3 capable of forming new particles remained high. By comparison the PN concentration in the NPF event following the rain 720 event in AtmMain was one degree of magnitude higher. Unlike AtmMain, NPF events in woCloudAtm occurred consistently throughout the simulation (see supplementary Fig. S35). Each event initiated in the morning in accordance with the increase in OH, BrO and Cl concentrations and thus DMS oxidation and SA production. The lack of clouds allowed the particles to grow to approximately 30 nm in particle mode diameter. In AtmMain, morning NPF events were terminated by the uptake of SA and NH 3 to the cloud particles. Overall, cloud free conditions promoted the gas-phase SA production and thus NPF, but impeded 725 SO 2 derived SO 2− 4 PM formation in the aqueous particle phase as species SO 2 and HPMTF accumulated in the gas-phase. The effects of aqueous-phase chemistry in both the deliquesced particles and cloud droplets were validated in the woAqAtm sensitivity run. Bromide and chloride ions did not activate as no iodine chemistry took place in the aerosol particles. Consequently, neither HOBr nor ICl entered the gas-phase and thus no BrO or Cl radicals were formed. The lack of reactive halogen species lowered the sink flux of DMS :: by :::: 35.2 ::::::: percent, causing it to accumulate in the gas-phase throughout the simulation (see 730 supplementary Fig. S36). The total secondary aerosol yield decreased :::: from :::: 1.43 :: µg ::::: cm −3 ::: to ::: 0.29 ::: µg ::::: cm −3 : in accordance with the decrease in DMS decay (Table 4). MSA PM production proved strongly reduced in woAqAtm ::::: (0.07 :: µg ::::: cm −3 :: at ::: the :::: end :: of :: the :::::::::: simulation) relative to AtmMain :::: (0.85 ::: µg :::::: cm −3 ). While woAqAtm still considered the phase transfer of soluble species, no reactions took place in the particle-phase. As a results, MSIA did not oxidise by O 3 to form MSA in the deliquesced particles.

755
We have presented ADCHAM simulation of the OH-derived oxidation of DMS and subsequent particle formation and growth in the AURA smog chamber and under relevant atmospheric conditions. New particles in the chamber experiments formed primarily by nucleation of SA (produced in the gas-phase via the CH 3 SO 3 intermediate), and NH 3 , and grew by condensation of MSA. The total production of secondary aerosol mass and the MSA to SA ratio was strongly influenced by the formation of a liquid water film on the Teflon bag chamber wall, the effect of which increased in accordance with RH. Water soluble 760 reaction products and intermediates DMS, DMSO, DMSO 2 , MSIA, HPMTF, MSA and SA were taken up by the water film.
Consequently, the secondary aerosol PM production decreased significantly during experiments performed at humid conditions (50-80% RH) compared to experiments performed at dry conditions (0-12% RH). Recently discovered autoxidation product HPMTF comprised a large fraction of the gas-phase products produced in both smog chamber and atmospheric relevant simulations, but proved insignificant to the direct formation and growth of aerosol particles. HPMTF may, however, contribute 765 indirectly to the particle growth by oxidising in the aqueous-phase to form SO 2 and thus SA and SO 2− 4 . At high RH the rate of CH 3 SCH 3 OO autoxidation leading to the formation of HMPTF had a strong impact on the secondary mass yield in the chamber experiments. At low RH, the oxidation of MSIA by OH proved essential to the total particle mass. Overall, a significant revision of the DMS oxidation mechanisms presented in literature was needed to reproduce the measurements obtained in the AURA smog chamber.

770
OH-derived oxidation of DMS in an atmospheric relevant context comprised a significant DMS sink, but proved less important than the equivalent oxidation by halogen species BrO and Cl. The relative importance of OH oxidation increased in accordance with a decrease in wind speed, which lowered sea spray emissions and thus the gas-phase concentration of reactive halogen species. The large surface area of the sea spray aerosols induced a strong condensation sink, thus impeding NPF. By introducing precipitation in the model the decrease in the condensation sink was sufficient to allow new particles to form by 775 nucleation of SA and NH 3 .
Smog chamber studies are able to elucidate atmospheric phenomena in a controlled environment, but rarely represent actual atmospheric conditions. Therefore, future studies will focus on implementing the revised DMS chemistry in the chemistry transport model ADCHEM, and test the setup on field measurements from the marine Arctic environment.
Code and data availability. All source codes, including the complete ADCHAM model version and plotting programs used to conduct 780 the analysis presented in this paper can be obtained by contacting the corresponding author R.W.J. All results presented in the paper and supplementary material and the complete DMS multiphase chemistry mechanism (Supplementary Tables S1) written in a format compatible with the Kinetic PreProcessor (KPP) Damian et al. (2002)