Simulations were performed in the gas and particle phase chemistry box model ADCHAM . ADCHAM considers state-of-the-art aerosol dynamics, treating the formation, growth, ageing and deposition of aerosol particles in the atmosphere. New particles are formed in the model by use of the Atmosphere Cluster Dynamics Code (ACDC) . ACDC takes into account both neutral and ion-induced pathways when calculating nucleation rates for various clustering systems, simulating the formation of clusters consisting of maximum five acid molecules and five base molecules. Clusters that manage to grow into the upper size range in ACDC are introduced as particles in ADCHAM in the smallest size bin at 1.08 nm. The setup used in this study models NPF from SA and NH3, taking into account the loss of clusters to collisions, evaporation, coagulation, and scavenging. The SA-NH3 system was calculated using the domain based local pair natural orbital (DLPNO) method at the DLPNO-CCSD(T)/aug-cc-pVTZ//ωB97X-D/6-31++G(d,p) quantum chemical level of theory . The DLPNO method was used in the default model setup as opposed to the resolution-of-the-identity second-order coupled cluster (RI-CC2) method as RI-CC2 has been shown to over-predict cluster binding energies . Nucleation rates obtained with the RI-CC2 method should therefore be interpreted as the upper limit for SA-NH3 driven NPF, whereas nucleation rates obtained with the DLPNO method comprises the lower (and more realistic) limit.

Newly formed or pre-existing aerosol particles grow in the model, either through vapour condensation or through the partitioning and processing of water soluble species in the deliquesced aerosol particles and cloud droplets. The condensation and evaporation of species is solved with the analytical predictor of condensation (APC) scheme , taking into account the dissociation of condensing compounds in the aerosol particle aqueous-phase. Other water soluble compounds, including oxidation agents, are allowed to partition to and (for certain compounds) dissociate in the aqueous-phase in accordance with their Henry's law solubility and pKa values, respectively. In this study, ADCHAM considers the partitioning of 52 gaseous species to the aerosol particle aqueous-phase. Some of these species contribute to the growth of the aerosol particles by undergoing aqueous-phase oxidation to form low-volatile compounds that remain in the particle-phase, even upon a decrease in the water-content of deliquesced aerosol particles or through the evaporation of cloud droplets. Henry's law solubilities and pKa values for the key DMS-derived water-soluble compounds HPMTF, DMSO, MSIA and MSA are based on COSMOtherm calculations obtained from .

Once formed, aerosol particles in ADCHAM leave the atmosphere through either wet or dry-deposition. Wet-deposition, including below-cloud scavenging of particles, is calculated for different particle sizes in accordance with the parameterization by . The parameterization considers the intensity of precipitation in mm h⁻¹. Dry-deposition of aerosol particles in an ocean environment relevant for this study is calculated with the parameterization by .

To investigate the influence of DMS on NPF in the pristine MBL, a model scenario reproducing conditions over the remote open ocean was constructed. The simulation was made to imitate the conditions during the flight-campaign described in , where frequent NPF within the MBL was observed following the passage of cold fronts. In the BaseCase scenario, the movement of an air-parcel within the MBL is simulated for 144 h, starting and ending at midnight. During the first 48 h, the air-parcel passes through cloud cover four times – twice during midnight and twice during mid-day. To form clouds, the air-parcel is lifted, causing the air to become supersaturated with respect to water vapour through adiabatic cooling. The clouds are allowed to persist for 2 h, until the air-parcel descends and they evaporate. During the fourth cloud event, heavy precipitation at 10 mm h⁻¹ is introduced in the model. The cloud is allowed to precipitate for the first hour of its lifetime. During the remaining 4 d of the simulation, no cloud or precipitation events take place. In this time (and in between cloud periods during the first 2 d), the relative humidity is kept at 90 % and the aerosol particles are assumed to be deliquesced.

The prescribed scenario is designed to spur NPF after the precipitation event on day 3 of the simulation. Before this event, the model is initiated with a size distribution consisting of NaCl in the accumulation and coarse mode in addition to (Na)2SO4 in the Aitken mode. The initial aerosol concentration is kept high to ensure that new particles are less likely to form during the first 2 d of the simulation. The precipitation event on day 2 effectively lowers the aerosol particle concentration, allowing DMS-derived oxidation products to form and grow new particles. The initial size distribution and sea spray emissions are calculated using the parameterization by . This parameterization takes into account the surface layer wind-speed, which is kept fixed at 6 m s⁻¹ in the BaseCase simulation but varied in additional sensitivity runs.

DMS emissions in ADCHAM are treated by considering an oceanic DMS dataset and a DMS transfer velocity parameterization. The oceanic DMS dataset by was used to construct four scenarios with five different surface ocean concentrations of DMS, including low (5 nmol L⁻¹), moderate (10 nmol L⁻¹), high (15 nmol L⁻¹) and very high (20 nmol L⁻¹) concentrations. High to very high concentrations of DMS are representative for phytoplankton blooms in the North Atlantic Ocean during May–July, and in the Southern Ocean from December–February . It should be noted that even higher concentrations can be found in the ocean surrounding Antarctica during sea-ice breakout and melting. A super-linear transfer velocity specifically made for DMS was used to obtain the flux of DMS to the model . We acknowledge that alternative linear and quadratic transfer velocities are also commonly used to calculate the DMS sea-air flux . All parameterizations consider the sea surface temperature (SST) and wind-speed when calculating the transfer velocity. Emissions of NH3 were treated similar to DMS by considering an oceanic dataset and a transfer velocity. The sea surface concentration of total ammonia NHx was set to 0.3 mmol m⁻³. This values represents a middle ground between NHx concentrations observed in the North Atlantic Ocean (May–July: ∼ 0.2 mmol m⁻³) and the Southern ocean around Antarctica (December–February: ∼ 0.4 mmol m⁻³) when the phytoplankton blooms are at their peak in each respective region . Additional sensitivity runs are performed using both higher and lower concentration of sea surface NHx. The transfer velocity of NH3 was calculated in accordance with the method by . This approach employs a bidirectional exchange of NH3, meaning that the air-sea flux of NH3 in the model is regulated by the gas-phase concentration of NH3. In the case that the concentration of NH3 in the gas-phase exceeds the so-called “compensation-point”, the sea-air flux of NH3 can also become negative. It should be noted that the sea surface concentration of both NHx and DMS can reach considerably higher concentrations near the coast, especially in regions with strong upwelling or nutrient run-off . Emissions of biogenic volatile organic compounds (BVOCs) over the oceans (including isoprene and α-pinene) is treated in accordance with the rates reported in .

The DMS oxidation reaction scheme used in this study has its basis in the mechanism presented in . Here, this mechanism is updated to account for the recent advances in the representation of DMS chemistry. In addition to these updates, we also use quantum chemical calculations to present new reaction pathways and temperature dependencies, as a means to address reactions in the DMS oxidation mechanism that still remain uncertain. The full gas-phase and aqueous-phase mechanism is found in Tables S1 and S2 in the Supplement, respectively. Alternative reaction rate coefficients from the literature for key reactions in the mechanism can be found in Table S3. A simplified scheme for the gas-phase reactions that impact the direct formation of SA and thus DMS-drived NPF in the MBL is shown in Fig. .

The main part of the DMS-derived gas-phase chemistry is taken from the Master Chemical Mechanism version 3.3.1 (MCMv3.3.1) . This includes the gas-phase oxidation of DMS by OH and NO3 in addition to the fundamental processing of oxidation intermediates and products in the so-called addition and abstraction pathway. Here, the addition pathway refers to the addition of OH to DMS while the abstraction pathway refers to the H-abstraction from DMS by OH. The oxidation of DMS by halogen radicals Cl and BrO (through H-abstraction and addition, respectively) is treated in accordance with the study by . The processing of halogen radicals, including the activation of chlorine and bromine species in the sea spray particles, is obtained from . This chemistry is essential for the representation of halogen radicals in the MBL and thus the oxidation of DMS. In addition to the halogen driven oxidation of DMS, also provides the reactions that govern the aqueous-phase processing of DMS and its water soluble oxidation products in both deliquesced aerosol particles and cloud droplets. Amongst others, these reactions include the uptake of DMS, DMSO and MSIA to the aqeoues-phase and their reaction with OH and O3 to form MSA. The uptake of MSIA has been found to be particularly impactful for the production of MSA particle mass (PM), as MSIA is able to partition to deliquesced aerosol particles in between cloud periods . The aqueous-phase processing of SO2 is treated in accordance with .

The formation of HPMTF through autoxidation, proposed in theory by , shown first in experiments by and observed in the atmosphere by , has undergone numerous revisions in the last couple years. The compound is formed from the CH3SCH2OO peroxy radical (produced initially in the DMS abstraction pathway), which undergoes two hydrogen shifts (H-shifts) to form HPMTF. The first of the two H-shift is considerably slower than the second and thus constitutes the rate-limiting step in the formation of HPMTF. used multiconformer transition state theory to calculate temperature dependant rate coefficients for both H-shifts. Using this approach, the first and rate-limiting H-shift was determined to be 0.047 s⁻¹ at 295 K. This estimate remains considerably slower than the some of the reaction coefficients determined in experiments, e.g. by (0.23 ± 0.12 s⁻¹ at 295 K) and (0.13 ± 0.03 s⁻¹ at 295 K). In this study, we use the temperature dependant rate coefficient by and scale it to match the rate coefficient obtained by . The rate by is used as opposed to the rate by , as the uncertainty associated with the rate is lower. To stay consistent with the use of reaction coefficients from , the subsequent reaction of HPMTF with OH is also implemented from this study. Said rate is considerably higher than the original one presented in the study by (2.1 × 10⁻¹¹ and 1.4 × 10⁻¹² cm³ s⁻¹, respectively), ensuring that a higher fraction of HPMTF is converted to SO2. HPMTF has also been found to oxidise in the aqueous phase to form SO42- . However, as the mechanism for these reactions remains unknown, this pathway has been omitted from the mechanism used in this study. Still, we acknowledge that this pathway is likely to reduce the production of SO₂ from HPMTF in the gas phase.

Another aspect of the DMS oxidation mechanism that has undergone revision in recent years is the fate of the products formed in the addition-pathway. In MCMv3.3.1, DMS undergoes OH addition to form DMSO, MSIA and finaly SO2. Recent studies, combining smog chamber experiments on DMS oxidation with box modelling, are nevertheless unable to account for the production of MSA in the gas-phase and particle-phase without considering the H-abstraction from MSIA by OH to form the intermediate radical CH3SO2 . CH3SO2 subsequently undergoes thermal decomposition to form SO2 (similar to the fate of MSIA in MCMv3.3.1), but may also react with O3 or O2 and NO to form CH3SO3 and thus SA and MSA. Here it is essential to distinguish between the formation of SA from CH3SO3 and from SO2, the latter pathway being considerably slower than the first (2.8 × 10⁻² and 8.8 × 10⁻⁷ s⁻¹ at 283 K and an OH concentration of 10⁶ cm⁻³). With a direct impact on the production of SA and MSA, the H-abstraction from MSIA by OH therefore has the potential to impact the formation and growth of aerosol particles in the MBL. The reaction coefficient for said reaction was set to match the experimentally determined rate by . As this rate was determined at 298 K, quantum chemical calculations are used in this study to provide temperature dependency for the reaction (Sect. ). These calculations demonstrate that the H-abstraction is more likely to occur at lower temperatures. While the H-abstraction from MSIA by OH has been shown to improve the model representation of MSA in previous publications, others continue to report a missing source of MSA in their model simulations . Consequently, this study also includes the addition of OH to MSIA, ultimately forming MSA through O2 hydrogen abstraction. This pathway has been proposed theoretically in previous publications , but the rate of the reaction has so far only been based on estimates or taken from similar reactions. Here, we implement quantum chemical calculations to determine the temperature dependant rate coefficient for the production of MSA through OH addition to MSIA. The reaction pathway is incorporated together with the work by , which suggest that H2SO3 also forms from the addition of OH to MSIA. In this mechanism, H2SO3 is assumed to decompose directly to SO2. An alternative source of MSA through the reaction of CH3SO3 with DMS is also considered .

The isomerization of CH3SOO to CH3SO2 was included in the mechanism in accordance with the study by , which ensures that SA and MSA forms in the abstraction pathway, even during low NOx conditions . In MCMv3.3.1, CH3SO2 is only able to form via the abstraction pathway through the reaction of CH3SO with NO2 or via O2 addition followed by oxygen abstraction by NO. As NOx concentrations are known to be low in the pristine MBL, the isomerization of CH3SOO therefore plays a role in the production of CH3SO2, SA and thus NPF. The fate of CH3SO2 nevertheless remains uncertain. CH3SO2 either undergoes thermal decomposition to form SO2 or reacts with O3, NO2 or O2 followed by NO, HO2, or RO2 to form CH3SO3. While the reaction coefficients for the reactions forming CH3SO3 are consistent in the literature, the coefficient for the thermal decomposition of CH3SO2 are not. MCMv3.3.1 reports a temperature dependant reaction coefficient of 0.29 s⁻¹ at 295 K, which is considerably lower than the estimates used in the box modelling studies by (6 s⁻¹ at 295 K) and (20 s⁻¹ at 295 K). As the two latter reaction coefficients are in fact estimates obtained by fitting box model results to measurements, the rate presented in MCMv3.3.1 is used in this study. Due to the impact of the reaction on the distribution of SO2, SA and MSA (and thereby NPF), the different rates are tested in ADCHAM through sensitivity runs. Similar to CH3SO2, the thermal decomposition of CH3SO3 also remains uncertain. Again, MCMv3.3.1 reports a relatively low rate for the reaction (0.11 s⁻¹ at 295 K) while another study suggests that the reaction coefficient might be considerably higher (9.71 s⁻¹ at 295 K) . The rate from MCMv3.3.1 is implemented as the default rate in this study while sensitivity runs are performed with the rate by .

In addition to the default simulation (named BaseCase), different sensitivity runs were performed to test the impact of meteorological and hydrological conditions along with alternative rate coefficients and reaction pathways in the DMS oxidation mechanism on the role of DMS in NPF in the MBL. Table  gives an overview of the different simulations. The TEMP₂₇₃, TEMP₂₇₈, TEMP₂₈₈ and TEMP₂₉₃ sensitivity runs are performed at air temperatures both lower and higher than in the BaseCase simulation, respectively. This is done to demonstrate the impact of temperature on the DMS chemistry in addition to its effect on particle formation and particle growth. WS₂, WS₄, WS₈ and WS₁₀ showcase the impact of wind-speed on the emissions of compounds to the model, especially emissions of DMS, NH3 and sea-spray. These simulations represent conditions of low, moderate and high wind-speed. PRCP_2.5, PRCP₅ and PRCP₁₅ show the impact of varying the amount of precipitation on day 2 of the simulation. Here, precipitation at 2.5–5.0 mm h⁻¹ is classified as moderate rain, while precipitation from 10–15 mm h⁻¹ is classified as heavy rain. The sea surface concentration of DMS is varied from the concentration used in the BaseCase simulation (10 nmol L⁻¹) to 5 nmol L⁻¹ in [DMS]₅, 15 nmol L⁻¹ in [DMS]₁₅ and 20 nmol L⁻¹ in [DMS]₂₀. This represent conditions ranging from weak to moderate and strong phytoplankton blooms. [NHx]_0.1, [NHx]_0.2 and [NHx]_0.4 demonstrate the effect of the sea surface concentration of total dissolved ammonia, which in turn impacts the emissions of NH3 to the marine atmosphere. NHx concentrations of 0.2 mmol m⁻³ are representative of conditions in the North Atlantic Ocean between May and July, where phytoplankton blooms are at their peak in the region . The same is true in the Southern Ocean during December to February, where NHx concentration of 0.4 mmol m⁻³ are common. Both higher and lower concentrations of NHx can be found in different regions of the Worlds oceans, or in different periods of the year. The range of NHx concentrations presented here are nevertheless representative of pristine ocean regions during periods with active phytoplankton blooms. In the BaseCase simulation, no clouds are introduced in the model after the precipitation event at the end of day 2. In the CLOUD sensitivity run, mid-day and mid-night clouds continue to be present after the precipitation event to test the impact of clouds on NPF in the simulation.

In addition to the sensitivity runs focusing on emissions and different meteorological and hydrological conditions, certain aspects of the DMS oxidation chemistry and its impact on DMS-derived NPF are also examined. In ΔCH3SO2, the rate for the thermal decomposition of CH3SO2 from is used as opposed to the rate from MCMv3.3.1, to test how an increase in SO2 at the expense of SA and MSA impacts NPF. In ΔCH3SO3, the rate from for the thermal decomposition of CH3SO3 is used as opposed to the one from MCMv3.3.1 to demonstrate the impact from an increase in SA from CH3SO3. In noMSIAabs and noMSIAadd, the reactions for the H-abstraction from MSIA by OH and OH addition to MSIA are removed from the simulation. This is done to showcase the role of the MSIA compound in the formation and growth of aerosol particles from DMS. In addition to the results depicted in Figs. – from the main text, further figures can be found in the Supplement, Sects. S4–S12. The mean source flux of H₂SO₄, MSA, SO₂, MSIA, DMSO, and HPMTF depicted in these figures are summarized in Table S4.

In this study, two key reactions on the oxidation of MSIA by OH were modified based on quantum chemical calculations. This was done due to their impact on SA and MSA production in the gas-phase and thus DMS-derived NPF in the MBL. The two reactions concern the H-abstraction from MSIA by OH and the addition of OH to MSIA. Quantum chemical calculations were used on the following principles: (1) If the reaction coefficient had previously been based on estimates, then the quantum chemical calculations were used to obtain both the reaction rate and the temperature dependency for the reaction. (2) If the reaction coefficient has been determined experimentally but at a fixed temperature, then the quantum chemical calculations were used to obtain the temperature dependency for the reaction. In the context of this work, quantum chemical calculations were therefore used to obtain the temperature dependency for the H-abstraction from MSIA by OH, since the rate of reaction had already been determined at a fixed temperate (298 K) by . As no experimental rates exists for the addition of OH to MSIA, both the reaction rate and temperature dependency were calculated using quantum chemical calculations.

For the quantum chemical calculations, all geometry optimizations and harmonic vibrational frequency calculations of the reactants, transition states (TS), intermediates (IM), and products were performed at the M06-2X/6-311+G(3df,2p) level of theory . Intrinsic Reaction Coordinate (IRC) calculations at the same level were carried out to confirm that each TS connects the intended reactants and products. Single-point energy (SPE) calculations were performed at the ROHF-ROCCSD(T)-F12a/cc-pVDZ-F12 (hereafter denoted as F_12a) level , including zero-point energy corrections obtained at the M06-2X/6-311+G(3df,2p) level. This combination of methods has been widely applied to study the transformations of peroxy and alkoxy radicals, and its accuracy is well established as a “gold-standard” approach . All quantum chemical calculations were performed using the Gaussian 16 and Molpro 2024.3 software packages .

To account for the effects of multiple conformers of the reactants and transition states on the reaction mechanism, both single and multi-conformer approaches were employed to balance computational cost and accuracy. Based on the single-conformer results, multi-conformer effects were explicitly considered for key reaction channels, including H-abstraction and OH-addition reactions of MSIA with OH, as well as the subsequent reactions of the addition intermediates (IM_add). Following previous work , a conformational sampling scheme using Molclus and Gaussian 16 was adopted to explore the low-energy conformers of the relevant reactants and transition states. For kinetic analysis, the electronic energies were taken from the F_12a calculations, while the partition functions were derived from the M06-2X/6-311+G(3df,2p) results. Rate constants were obtained using multiconformer transition state theory (MC-TST). One-dimensional unsymmetrical Eckart barriers were applied to account for quantum tunnelling in reactions involving H-shift or H-abstraction . All rate constants were calculated with the KiSThelP program . The calculations are discussed in more detail in the Supplement, Sect. S4.

At the conditions given in the BaseCase model setup, a distinct NPF event takes place on the third day of the simulation following the precipitation event (Fig. a). The formation and growth of particles is driven by emissions of DMS, which enters the model at 8.3 × 10⁹ molec. cm⁻² s⁻¹ (6 m s⁻¹ wind-speed) and gradually builds up to reach a concentration of 0.92 ppb at the end of the simulation (Fig. b). Atmospheric DMS concentrations in the MBL have been shown to range from tens of ppt in regions with low surface-ocean DMS concentrations to a couple of ppb over waters with high surface-ocean DMS concentrations . The scenario prescribed in the BaseCase simulation is representative of a marine environment with moderate to high surface-ocean DMS concentrations, and the simulated gas-phase concentrations of DMS are therefore in line with observations from the MBL. The flux parametrization for the sea-air exchange of NH3 results in a mean NH3 gas-phase concentration of 17 ppt throughout the simulation with a maximum concentration of 50 ppt. These concentrations fall within the range observed in the Arctic MBL , and constitute a realistic representation of NH3 in the MBL without impact from anthropogenic emissions. As the simulation is initiated with high concentrations of sea spray aerosol particles (18.6 µg m⁻³) and partial cloud cover, NPF is suppressed on day 1 and day 2. The suppression of NPF is driven by the high SA condensation sink (1.4 × 10⁻³ s⁻¹ before the precipitation event) and the uptake of SA and SA precursors to the cloud droplets. These precursors include MSIA and SO2 which partition to the cloud droplets during cloud cover (Fig. c) to form MSA and SO42-, respectively. This means that MSIA and SO2 are almost completely removed from the gas-phase during cloud cover. Consequently, the mean SA gas-phase concentration is lower on day 2 of the simulation (1.2 × 10⁶ cm⁻³) compared to day 3 (1.6 × 10⁷ cm⁻³). The uptake and processing of water soluble species in the activated cloud droplets causes a small but distinct Hoppel minima to form in the size distribution during days 1–2 (Fig. a), driven predominantly by the uptake and aqueous-phase processing of DMSO, MSIA and SO2. The minima persists after the precipitation event and highlights the impact that aqueous-phase chemistry in cloud droplets has on the aerosol particle population over the ocean. MSA is also removed during cloud cover, but since it possesses semi-volatile characteristics (Henry's law solubility of 6.7 × 10⁸ mol kg⁻¹ atm⁻¹ at 283 K) a fraction of it will evaporate back to the gas-phase after each cloud period (Fig. d). During the cloud free conditions representative of days 3–6, MSA follows a similar diurnal trend as SA with peak production and concentration during midday and lower concentrations at night. HPMTF builds up in the gas-phase in the same way as DMS, reaching a concentration of 166 ppt at the end of the simulation (Fig. b).

On day 3 of the simulation, the precipitation event removes a considerable fraction of the aerosol PM, thereby decreasing the condensation sink to 1.1 × 10⁻⁴ s⁻¹. This spurs an increase in the SA gas-phase concentration which causes SA and NH3 to cluster and form new particles. The continuous condensation of SA and NH3 ensures that the newly formed clusters grow, first into a distinct nucleation mode and thereafter into the Aitken mode which reaches a mode particle diameter of 73 nm (25th–75th prct.: 60–88 nm) at the end of the simulation. Additional NPF on days 4–6 is suppressed due to the increase in the condensation sink from the additional sea-spray PM and the particles formed in the NPF event on day 3. The gas-phase concentration of MSIA, MSA and SA decreases correspondingly as they are lost to the increasing concentration of said particles (Fig. c and d). SO2, on the other hand, accumulates in the gas-phase during days 3–6 as the lack of clouds prevents it from partitioning efficiently to the particle-phase. This also means that the SA source flux from SO2 increases towards the end of the simulation whereas the SA source flux from CH3SO3 decreases in correspondence with the decrease in MSIA (Fig. S6). Consequently, SO2 comprises 21 % of the SA source flux during day 6 compared to 3 % during day 3. As a result, SO2 becomes less important for the initial NPF as the conversion of SO2 to SA is considerably slower than the conversion of CH3SO3 to SA. When SO2 finally reaches a concentration where it contributes significantly to the production of SA, the sea-spray and DMS-derived PM concentration has become too high for new particles to form efficiently. Nucleation rates for the SA-NH3 system obtained with the RI-CC2 method are significantly higher than those obtained with the DLPNO method (maximum nucleation rate of 2.0 cm⁻³ s⁻¹ on day 3 of the simulation using RI-CC2 as opposed to 0.13 cm⁻³ s⁻¹ when using DLPNO)(Fig. S12). This rate leads to a 2.2 times increase in the total PN concentration at the end of the simulation. The increase in the PN concentration stunts the growth of the particles as more SA is needed to grow said particles into the nucleation and Aitken mode. As a result, the growth mode only reaches a mode particle diameter of 56 nm (25th–75th prct.: 46–68 nm) at the end of the simulation. Consequently, using the RI-CC2 method as opposed to the DLPNO method to obtain nucleation rate for the SA-NH3 system will increase the total number of particles but decrease the number of particles that may reach the CCN size range.

SO42- and NH4+ PM dominates the chemical composition of the aerosol particles formed during the NPF event and in the subsequent growth mode (Fig. e). This is caused by the ability of SA to condense irreversibly to the particle-phase, with NH3 aiding the growth by neutralizing the otherwise acidic aerosol particles. According to ADCHAM, DMS-derived SA and natural emissions of NH3 from the surface ocean are capable of sustaining the growth of newly formed aerosol particles into the nucleation and Aitken mode. MSA does not contribute significantly to the growth of the smaller particles, as its semi-volatile characteristics mainly allows it to condense on the larger sea-spray particles. MSA PM nevertheless aids the growth of the particles in the Aitken and upper end of the nucleation mode, with particularly high PM concentrations in the accumulation and coarse mode particles (Fig e). This happens as 85 % of MSA is formed in the aqueous-phase of the deliquesced aerosol particles and cloud droplets from the oxidation of dissolved and dissociated MSIA by O3 (Fig. S7). The remaining MSA production happens in the gas-phase through the addition of OH to MSIA (8 %) and from the reaction of CH3SO3 with HO2 (6 %) and DMS (2 %). While MSA formed in the gas-phase is able to condense on particles in the nucleation and Aitken mode size range, MSIA mainly partitions to the coarse mode deliquesced sea-spray particles where it is converted to MSA. This tendency is evident from the decrease in the MSA PM during the precipitation event. Here, the MSA PM decreases by 95 % as opposed to 86 % for the SO42- PM as the coarse mode particles containing most of the MSA PM are removed more efficiently through wet-deposition than particles in the lower size ranges (Fig. g). After the precipitation event, the MSA PM increases disproportionally to the SO42- PM as the lack of clouds prevents SO2 from being converted to SO42-. As MSA continues to be produced in the deliquesced sea spray particles, the MSA / SO42- PM ratio increases throughout the simulation from an average ratio of 0.3 on day 3 to 2.3 on day 6 (Fig. f).

Although this study has demonstrated that DMS is able to drive the formation and growth of aerosol particles within the pristine MBL under various meteorological and hydrological conditions, some uncertainty still remains due to the present representation of the DMS oxidation mechanism. Most of this uncertainty lies in the fate of MSIA and its successors CH3SO2 and CH3SO3, due to their role in the production of SA and MSA. CH3SO2 and CH3SO3 are also formed in the DMS abstraction pathway, but since HPMTF and SO2 make up the major products from said pathway the production of CH3SO2 and CH3SO3 remains small. At 283 K, MSIA oxidation by OH makes up 96 % of the CH3SO2 source flux while the isomerization of CH3SOO makes up 4 %. At the low NOx conditions representative of the pristine MBL, CH3SO2 formation through the reaction of CH3SO with NO2 or CH3SOO2 with NO becomes insignificant. The role of MSIA in the production of SA through CH3SO2 and CH3SO3 is evident from the “noMSIAabs” sensitivity run. Here, the source flux of SA decreases by 87 % throughout the simulation, thus decreasing the strength of the NPF event following the precipitation event on day 3 (Figs. a and S84). This is caused by SA being formed predominantly through SO2 which is slower than the production of SA through the thermal decomposition of CH3SO3. Consequently, SA is not formed fast enough to initiate a strong NPF and growth event before the concentration of sea spray becomes to high. Some new particles are able to form on days 4–6 of the simulation when the SO2 concentration has had time to accumulate, but the nucleation and growth rate remains limited. The rate of the H-abstraction from MSIA by OH was originally determined experimentally at 298 K by . In this work, we extend the rate determined by with a theoretical temperature dependency based on quantum chemical calculations. Said temperature dependency predicts a slight increase in the rate of H-abstraction at lower temperatures, e.g. 9.0 × 10⁻¹¹ cm³ s⁻¹ at 298 K and 1.2 × 10⁻¹⁰ cm³ s⁻¹ at 273 K. It should be noted that in the simulation the increase in the H-abstraction rate is counterbalanced by the decrease in the OH concentration at lower temperatures.

In addition to adding temperature dependency to the rate for H-abstraction from MSIA by OH, this study also presents the temperature dependant reaction rate for the addition of OH to MSIA and subsequent production of MSA through H-abstraction by O2. This pathway has been discussed in previous publications , but so far no experimental nor theoretical reaction rates have been determined. Here, quantum chemical calculations are used to provide the reaction rate. The noMSIAadd sensitivity run demonstrates the impact of the reaction, which has little influence on the formation and initial growth of the aerosol particles as it does not significantly influence the production of SA. It does however impact the gas-phase production of MSA which decreases by 47 % throughout the simulation (Fig. b). As a result, including this reaction pathway might help to improve model representations of MSA in the gas-phase. The HO-MSIA intermediate that forms after the OH-addition to MSIA has also been shown to form H2SO3 (and thus eventually SO2) in other studies by the loss of a methyl group . In accordance with the calculations performed in this study, however, the H-abstraction by O2 forming MSA is considerably faster than the decomposition to H2SO3 (2.4 × 10¹³ and 6.4 × 10¹⁰ s⁻¹ at 283 K and 1 atm, respectively).

Another aspect of the DMS oxidation mechanism that adds uncertainty to the role of DMS in NPF in the MBL is the rate for the thermal decomposition of CH3SO2 and CH3SO3 into SO2 and SO3, respectively. Unfortunately, it is challenging to calculate the transition state for the decomposition of both these species and therefore a theoretical rate has not been determined in this study. Instead we demonstrate the impact that these rates have on DMS-derived NPF in the MBL through sensitivity runs performed with the experimentally determined or estimated rates that are currently available in the literature. In this study we use the rates prescribed by MCMv3.3.1 in the BaseCase setup as they are in line with the experimental rate for the CH3SO2 decomposition obtained by (e.g. < 1 s⁻¹ at 298 K). Higher rates for the decomposition of CH3SO2 have been estimated in the studies by (8.3 s⁻¹ at 298 K) and (20 s⁻¹). As for the decomposition of CH3SO3, calculated an upper limit of 12.7 s⁻¹ at 298 K while an estimate based on experimental results by falls in line with the rate presented in MCMv3.3.1 (0.1 s⁻¹ at 298 K).

In the ΔCH3SO2 sensitivity run, the rate for the thermal decomposition of CH3SO2 was increased in accordance with the rate estimated by . This increase favours the production of SO2, the source flux of which increases by 24 % throughout the simulation at the expense of CH3SO3 and thus SA, the source flux of which decreases by 72 % (Fig. c and a). The decrease in SA lowers the nucleation rate on day 3 from 0.13 to 8.7 × 10⁻² cm⁻³ s⁻¹ and causes the particles to reach a mode diameter of 52 nm (25th–75th prct.: 41–65 nm) at the end of the simulation compared to 73 nm (25th–75th prct.: 60–88 nm) in BaseCase. Due to the decrease in the CH3SO3 concentration, SO2 becomes more important for the production of SA making up 52 % of the SA source flux as opposed to 11 % in the BaseCase simulation. The conversion of SO2 to SO3 and thus SA nevertheless remains slow, meaning that new particles are not able to form efficiently before the concentration of sea spray particles and thus the SA condensation sink becomes to high (Fig. S84). Using the rate by also impacts the source flux of MSA in the gas-phase which decreases by 44 % as a result of the decrease in the CH3SO3 concentration. As a consequence, MSA production through OH addition to MSIA becomes the dominating source of MSA in the gas-phase comprising 86 % of the MSA gas-phase source flux compared to 50 % in the BaseCase simulation. In summary, a higher rate for the thermal decomposition of CH3SO2 would make DMS-driven NPF in the MBL less favourable as the direct production of SA is suppressed.

A higher rate for the thermal decomposition of CH3SO3, however, is seen to have the opposite effect. Increasing said rate favours the production of SO3 and SA as opposed to MSA. This is evident from the ΔCH3SO3 sensitivity run where the rate for the thermal decomposition of ΔCH3SO3 was increased in accordance with the study by . As a consequence, the total SA source flux showed a 25 % increase compared to the BaseCase run at the expense of the MSA source flux in the gas-phase which decreases by 49 %. Said decrease originates from the reaction of CH3SO3 with HO2 and DMS, respectively, which make up ∼ 1 % of the MSA source in the ΔCH3SO3 sensitivity run. Consequently, MSA production through OH addition to MSIA as proposed in this study becomes the almost exclusive source of MSA in the gas phase. The impact of the increase in the SA source flux from the increase in the thermal decomposition of CH3SO3 is evident from the SA-NH3 derived nucleation rate. On day 3 of the simulation, the nucleation rate increases by 19 %, ultimately increasing the total PN concentration at the end of the simulation from 1.1 × 10³ cm⁻³ in BaseCase to 1.2 × 10³ cm⁻³ in ΔCH3SO3. The increase in the PN concentration counterbalances the increase in the growth for the individual aerosol particles as more SA is needed to grow the newly formed particles into the nucleation and Aitken mode size range. As a result, the mode diameter of the growth mode at the end of simulation is lower at 68 nm (25th–75th prct.: 57–82 nm) in ΔCH3SO3 compared to 73 nm (25th–75th prct.: 60–88 nm) in BaseCase, despite the increase in the SA source flux.

The rate calculated by nevertheless remains uncertain. In the review article by , it is mentioned that the rate for the thermal decomposition of CH3SO3 should be similar or much lower than the one by CH3SO2. This is supported by a recent experimental study by , where they estimate said rate to be similar to the one presented in the MCMv3.3.1. It remains evident however that the rate for the thermal decomposition of both CH3SO2 and CH3SO3 has an impact on the formation and growth of DMS-derived aerosol particles in the MBL. At the moment, any claim that these rates should be higher than those presented in the MCMv3.3.1 are based on estimates and theoretical calculations. It remains essential therefore that the thermal decomposition of both CH3SO2 and CH3SO3 are examined further in laboratory studies so as to decrease the uncertainty linked to their role in the formation and growth of aerosol particles from DMS.