To generate the input files for the COSMOtherm calculations, we used the COSMOconf program version 4.3 . COSMOconf contains conformer generation algorithms, different levels of theory of quantum chemical calculations for both the condensed and the gas phase, and various methods for reducing the number of conformers in a way that does not compromise the accuracy of the COSMOtherm calculations.

Including multiple conformers in the COSMOtherm calculations is important when the conformers have different polarities, as is the case for molecules that are able to form intramolecular hydrogen bonds . For finding an initial set of conformers, COSMOconf uses various conformer-generating algorithms. However, none of these methods allow for the systematic conformer sampling of the molecules. The nonsystematic conformer generation in COSMOconf has been shown to lead to significantly different results in COSMOtherm depending on the initial geometry with molecules containing hydroxy and hydroperoxy functional groups . Based on the recommendation by , we therefore used the systematic conformer sampling with the Merck molecular force field (MMFF94) in the Spartan '14 program . In addition to the most common carbon and oxygen atom types, the MMFF94 is parameterized for the atom types of a sulfate group sulfur and oxygens . This ensures that all unique conformers are found using the systematic sampling.

The conformers from Spartan '14 were used as input to COSMOconf, and the TURBOMOLE program package version 7.11 was used for the quantum chemical calculations. Our calculation template in COSMOconf follows the BP-TZVPD-FINE-COSMO.xml template found in the program, omitting the conformational sampling step at the beginning and setting the cutoffs (based on energy and the number of conformers) of conformers high enough that no conformers were discarded. The gas-phase conformers were obtained by BP/def-TZVP gas-phase geometry optimizations and BP/def2-TZVPD single-point energy calculations of the condensed-phase geometries from COSMOconf using the calculate function in TURBOMOLE. The BP/def2-TZVPD-FINE//BP/def-TZVP level .cosmo and .energy files from COSMOconf and TURBOMOLE were used in COSMOtherm calculations. In addition, .cosmo, .energy and .vap files for H2O and the inorganic ions were taken from the COSMObase17 database .

In our COSMOtherm calculations, we have used the most recent BP_TZVPD_FINE_19 parameterization. All calculations were done at 298.15 K. To the best of our knowledge, experimental information on the pure component phase state of most atmospherically relevant organics is not available. We therefore assume that all of the organosulfates (OS) and the isoprene epoxydiols (IEPOX) are liquid at 298.15 K. Without melting point and heat of fusion data, we are not able to accurately estimate the solubilities of solid-phase organosulfates. If the OS and IEPOX are solid at 298.15 K, the solubility results shown here are the mole fractions of the virtual liquid of the solute in the two liquid phases of a solid–liquid–liquid equilibrium. Sodium salts of the organosulfates (R−OSO3Na, NaOS) are similar to sodium dodecyl sulfate (SDS) with regard to molar mass and functionality. SDS is solid at 298.15 K, and we therefore assume that NaOS is solid at this temperature. The organic compounds are treated as solutes and the aqueous solutions (pure water or binary aqueous ammonium salt mixtures) as the solvent.

To select the maximum number of conformers needed for COSMOtherm calculations, convergence on the number of conformers was tested by calculating the activities of isoprene-OS-1. In these test calculations, the change in activity of isoprene-OS-1 and H2O (in different mole fractions of isoprene-OS-1 in water) was at most 0.005 between 40 and 45 conformers of isoprene-OS-1. Based on this, the maximum number of conformers was set to 40 for larger monoterpene-derived organosulfates and 50 for the smaller isoprene-derived molecules.

In COSMOtherm calculations, conformers are weighted according to the Boltzmann distribution based on the sum of their solvated energy and chemical potential in the solution. However, normally only conformers with the lowest solvated energies are selected for COSMOtherm calculations. If the total number of unique conformers is high, not all conformers can be included in the COSMOtherm calculation. When only a fraction of all conformers is used in a COSMOtherm calculation, only those containing intramolecular hydrogen bonds are used, as they have the lowest solvated energies. However, the interaction between a compound and water is more favorable for conformers containing no intramolecular hydrogen bonds. Therefore, in aqueous solutions, the chemical potential of conformers containing no intramolecular hydrogen bonds is much lower than that of conformers that contain multiple hydrogen bonds. If a compound contains more unique conformers than can be included in COSMOtherm calculations, more attention should be paid to selecting the conformers to represent the conformer distribution in the studied solutions.

found that COSMOtherm (release 18; ) overestimates the effect of intramolecular hydrogen bonds and recommended that only conformers containing no intramolecular hydrogen bonds should be used in saturation vapor pressure calculations. We tested the difference in saturation vapor pressures calculated using releases 18 and 19 (parameterizations BP_TZVPD_FINE_18 and BP_TZVPD_FINE_19, respectively) and found that differences between the two parameterizations are larger using all conformers than when only conformers containing no intramolecular hydrogen bonds are used. Variation between estimates using different conformer sets is also smaller in release 19 than in release 18. We therefore omitted all conformers containing intramolecular hydrogen bonds from the calculations of OS and IEPOX. Generally, the omission of conformers containing intramolecular hydrogen bonds leads to lower saturation vapor pressures .

The number of intramolecular hydrogen bonds in the condensed phase was determined using release 18 of COSMOtherm. For isoprene-OS-3 and isoprene-OS-4, we only found two and zero conformers containing no hydrogen bonds, respectively. For these two species, we used all conformers containing no full and any number of partial intramolecular hydrogen bonds or one full and no partial intramolecular hydrogen bonds. Many of the deprotonated organosulfates (sodium salt anions) have only conformers that contain intramolecular hydrogen bonds. For this reason, we chose to use all of their lowest-energy conformers in the COSMOtherm calculations involving the NaOS. In the calculation of pKa we used all conformers, since the calculation uses both the neutral and the ionic species.

Solubilities of organics in pure water and of water in the organic-rich phase were calculated using COSMOtherm as the respective mole fractions at the liquid–liquid equilibrium of OS–water mixtures. Results are shown in Fig. .

The LLE was not found for isoprene-derived organosulfates, IEPOX isomers or methyl bisulfate, indicating that these compounds are fully miscible with pure water at 298.15 K. We therefore also calculated the pure water solubilities relative to the organosulfate solubility in a 0.09 mole fraction salt solution by solving the LLE of ternary systems in which the solvent contains 0.09 mole fraction of either ammonium sulfate (AS, (NH4)2SO4) or ammonium bisulfate (ABS, NH4HSO4). Solubility calculations for ternary systems are described in more detail in Sect. . This is done to get a quantitative estimate of the relative solubilities of the compounds that are fully soluble in pure water. The 0.09 mole fraction is below the solubility limit of both (NH4)2SO4 (xSOL,AS=0.094) and NH4HSO4 (xSOL,ABS=0.33) in water at 298.15 K . The specific inorganic salt mole fraction was chosen to be as high as possible while within the aqueous solubility limit of the salt to ensure that the organic compounds are typically not fully miscible with the salt solution. Results are shown in Fig.  together with corresponding binary organic solubilities. Compared to the binary LLE solubility, the aqueous solubility calculated as a relative solubility for monoterpene-derived organosulfates is on average 3.1 times higher (1.8–5.5) using (NH4)2SO4 solutions as a reference and 2.2 times higher (1.7–2.9) using NH4HSO4 solutions. The LLE was not found in the ternary systems containing IEPOX and 0.09 mole fraction of NH4HSO4.

Based on LLE calculations, the monoterpene-derived organosulfates are less soluble in the ammonium sulfate and bisulfate solutions than in pure water, which means that the ammonium salts have a salting-out effect on the OS. From the solubilities calculated as relative solubility compared to (NH4)2SO4 and NH4HSO4 solutions, we can see that the relative solubility calculation in COSMOtherm overestimates the salting-out effect of both ammonium salts compared to the more accurate LLE calculation. In addition, the salting-out effect of (NH4)2SO4 is overestimated more than that of NH4HSO4. The relative solubility calculation uses the zeroth-order solubility approximation, which means that the estimate is less accurate when the absolute solubility of the solute is high. The largest difference between the aqueous LLE and the solubility calculated relative to the ternary LLE is seen for the OS with the higher absolute solubilities.

We see in Fig.  that α-pinene-OS-1 has the lowest solubility of all the organosulfates. There are only minor structural differences between α-pinene-OS-1 and α-pinene-OS-2, but this still leads to a factor of 3.6 difference in the calculated solubility. All the β-pinene and limonene organosulfates, with the same functional groups as α-pinene-OS-1 and α-pinene-OS-2, have solubilities between those of α-pinene-OS-1 and α-pinene-OS-2. These results show that even minor differences in the molecular structure, such as the placement of functional groups, can have a large impact on the solubility of organosulfates.

The most soluble monoterpene-derived organosulfates are α-pinene-OS-5 and α-pinene-OS-6, which each have both a carboxylic acid group and a carbonyl group. α-Pinene-OS-4 has a flexible carbon backbone and three carbonyl functionalities; however, it still has a relatively low solubility compared to the other α-pinene-OS. The effect of the different types of oxygen-containing functional groups on the solubilities is caused by their ability to form intermolecular hydrogen bonds with the solvent water. This explains the lower solubility of α-pinene-OS-4, which has mainly hydrogen-bond-accepting carbonyl groups, compared to α-pinene-OS-3, α-pinene-OS-5 and α-pinene-OS-6, which contain hydroxy groups that can act as both H-bond acceptors and donors.

We calculated acid constants (pKa) for all organosulfates to capture the effect of the dissociation of the neutral molecules in water. Estimated pKa values of the organosulfates are between -4.57 and -2.37, indicating that all of the organosulfates are strong acids that will likely be strongly dissociated in water. For comparison, we estimated the first pKa of sulfuric acid with COSMOtherm to be -3.51. The organosulfates are therefore estimated to be of equivalent strength or even stronger acids than H2SO4, and thus for all practical purposes they fully dissociate in near-neutral solutions and even solutions at the most atmospherically relevant pH. The pKa values for all organosulfates and sulfuric acid are shown in Table S1 of the Supplement.

Dissociation-corrected solubilities were calculated from Eq. () using the LLE solubilities in pure water and pKa estimated with COSMOtherm. Molar liquid volumes of the pure organic compounds used to calculate the densities of organic-water solutions for Eq. () are shown in Table S7. For all organic compounds, dissociation-corrected solubilities correspond to mole fractions higher than 1. This unphysical result is likely caused by the inability of Eq. () to accurately capture the solution behavior of very strongly acidic compounds. This equation is only used to calculate dissociation-corrected solubilities and has no effect on other property calculations. The dissociation of strong acids is expected to be high in solutions with higher pH than the pKa of the solute , as is the case here.

Since the organosulfates strongly dissociate in water, we also calculated the aqueous solubilities of their sodium salts (NaOS). For these sodium organosulfate salts, we used the heat capacity of fusion estimate (ΔCp,fus=ΔHfus/Tmelt) with a melting point of 478.15 K and a heat of fusion of 50 kJmol-1 . Calculated solubilities of the NaOS salts are shown in Fig.  and Table S1. For systems in which a solid–liquid equilibrium was found, the solubility of the organosulfate sodium salt is around 0.065 mole fraction.

Solubilities of both OS and NaOS were calculated by solving the LLE or the SLE, respectively, in aqueous solvents containing 0.09 mole fraction of either ammonium sulfate or ammonium bisulfate. Solubility values for the OS and NaOS in these solvents, and the aqueous ammonium sulfate and bisulfate salt solutions in the OS phase, are given in Table S2.

Organic solubilities in aqueous inorganic solutions ranging from pure water to 0.09 mole fraction of inorganic salt were calculated using relative screening. These relative solubilities were then scaled using the absolute solubility values of the 0.09 mole fraction binary solvents to obtain the final relative solubilities of the OS and NaOS with respect to each binary system at the different inorganic salt mole fractions. The procedure is described in detail in Sect. S2. Relative solubilities are shown in Fig.  (OS in (NH4)2SO4), Fig.  (NaOS in (NH4)2SO4), Fig. S7 (OS in NH4HSO4) and Fig. S8 (NaOS in NH4HSO4).

At low (<10-3) (NH4)2SO4 mole fractions, the molecular organosulfates salt in, meaning that the presence of the inorganic salt enhances the total amount of the organosulfate soluble in the aqueous phase. At higher inorganic salt mole fractions the organosulfates salt out. All IEPOX isomers and NaOS salts salt out in the presence of co-solvated (NH4)2SO4 across the whole concentration range. At 0.09 mole fraction of (NH4)2SO4, the organic compounds can be grouped into three categories based on their relative solubilities: methyl bisulfate with the highest relative solubility, isoprene-derived organosulfates and IEPOX in the middle, and all monoterpene-derived organosulfates with the lowest relative solubilities with respect to the pure aqueous solubility.

All of the organic compounds salt out in ternary aqueous solutions with NH4HSO4 (see Figs. S7 and S8), but the salting-out effect of NH4HSO4 on the organic compounds is weaker than that of (NH4)2SO4. This is due to the stronger salting interactions of the doubly charged sulfate ion compared to the singly charged bisulfate ion.

As was mentioned above, the salting-out of organosulfates from 0.09 mole fraction (NH4)2SO4 solution is overestimated by a factor of 3.1 using the relative solubility calculation compared to the LLE calculation. found that COSMOtherm overestimates the salting-out effect of (NH4)2SO4 on average by a factor of 3 compared to experiments. They described the salting behavior using Setschenow constants calculated from COSMOtherm-estimated (release 14) partition coefficients, which are comparable to relative solubilities. We used COSMOtherm19-estimated relative solubilities to calculate corresponding Setschenow constants for the compounds used by that are in COSMObase17 and the same 5 % (NH4)2SO4 solution (w/v, corresponding to xAS=0.007) with solvent densities by . We found that COSMOtherm19 overestimates the experimental Setschenow constant of these compounds in 5 % (NH4)2SO4 solution on average by a factor of 1.5 (see Fig. S11), which is an improvement to the factor of 3 of COSMOtherm14. The overestimation might be decreased by calculating LLE solubilities as opposed to relative solubilities that use the zeroth-order solubility approximation. However, finding the LLE of multiple systems is computationally infeasible and not certain to improve the results.

Liquid–liquid-phase separation (LLPS) has been detected in several aerosol experiments . For example, observed LLPS for ammonium sulfate aerosol that contained organic compounds with O:C below 0.8, whereas no LLPS was seen with O:C above 0.8, depending on the functional groups. In these experiments, organic compounds contained hydroxy, carbonyl and carboxylic acid groups . In binary aerosol systems containing water and organic compounds (without inorganic salt), observed LLPS for O:C below 0.44 or 0.58 in systems with one or two different organic compounds, respectively. The compounds in this study contain ester, ether and hydroxy functional groups . With O:C ratios of the monoterpene- and isoprene-derived organosulfates in the ranges 0.5–0.7 and 1.2–1.4, respectively, these results are consistent with the present work. On the other hand, in experiments with an OH-oxidized α-pinene and water system only a single organic-rich phase was observed, whereas LLPS was seen between water and ozone-oxidized α-pinene products or OH-oxidized isoprene products . There are small differences in the partial charges of the oxygen atoms associated with a sulfate group compared to a carboxylic acid group (see Sect. S3 for a comparison of the σ potentials) that may influence the O:C ratio of organosulfates required for LLPS. From our results we can also see that other structural factors further affect the thermodynamic properties, in addition to the O:C ratio or the types of functional groups.

Activities were calculated for organosulfates, IEPOX isomers and water in binary aqueous mixtures with different organic-to-water molar ratios (see Table S3). Figure  shows, as examples, binary mixing diagrams similar to that presented by for water and (a) α-pinene-OS-5, (b) β-pinene-OS-1, (c) limonene-OS-1, (d) isoprene-OS-2, and (e) δ1-IEPOX. Diagonal dashed lines illustrate the ideal mole-fraction-based activities (ai=xi) with respect to a pure compound (i=OS, water) reference state. Since the solubility of the organics in water is much smaller than the solubility of water in the organics, the mixing diagrams for monoterpene-derived organosulfates (Fig. a–c) are divided into two sections (note the different scales of the two phase regions): the aqueous phase (left sides) and the organic phase (right sides). In between is a composition range corresponding to the miscibility gap. From Fig. a–c we see how the calculated water and organosulfate activities fulfill the liquid-phase equilibrium condition of Eq. () at the solubility limit.

Activities for the monoterpene-derived organosulfates display three different types of behavior. The most common is exemplified in Fig. a, where in the organic-rich phase, the organosulfate activity is lower than the mole fraction of the organics (aOS≤xOS). A low activity indicates that the organosulfate is more stable in the organic-rich phase than in the ideal pure organosulfate. The water activity is below the ideal activity (aw<xw) at low mole fractions of water and above the ideal activity (aw>xw) at higher water mole fractions in the organic-rich phase. The organic activity at the solubility limit is low (aOS<0.28 when xOS=xSOL) compared to the other monoterpene-derived organosulfates. Similar behavior is seen in α-pinene-OS-3, α-pinene-OS-4, α-pinene-OS-6, β-pinene-2 and limonene-OS-4. A comparison between the activities of α-pinene-OS-5 and H2SO4 calculated using COSMOtherm and literature values of H2SO4 activities are shown in Fig. S9.

The opposite is seen in α-pinene-OS-1, β-pinene-OS-1 (Fig. b) and limonene-OS-3, for which the activity of the organosulfate in the organic-rich phase is very close to or above the ideal activity. In addition, the activity at the solubility limit (both the solubility of the water and the organic) for these compounds is above 0.36. The third behavior type seen in Fig. c is between the first two cases, in which the water activity follows the ideal activity in small mole fractions of water. Here the organic activity at the solubility limit is around 0.3. The other compounds in this group are α-pinene-OS-2 and limonene-OS-2.

Since the isoprene-derived organosulfates, IEPOX isomers and methyl bisulfate are fully miscible with pure water, liquid–liquid-phase separation was not observed for these systems. The mixing diagrams for all isoprene-derived organosulfates and methyl bisulfate are similar to the one shown in Fig. d. Calculated activities for all IEPOX isomers are close to the ideal activities at all mixing states (Fig. e).

Figure a–e show mixing diagrams for the same organic compounds as Fig. a–e but now with a solvent that is a 0.09 mole fraction binary aqueous solution of NH4HSO4 instead of pure water. Here, COSMOtherm also predicts LLPS for systems containing the isoprene-derived organosulfates (Fig. d). Again, activities for the organosulfates are higher than their mole fractions in the water-rich phase. Here we can also see that the predicted activity of water in the binary solvent is 0.78. The corresponding activity coefficients γi=ai/xi for the organic compounds and water in each system in Fig.  are tabulated in Table S4.

The calculated activity of each organic compound in the aqueous phase is higher in the ternary OS+aqueous ammonium bisulfate systems compared to the binary OS+water systems. This means that the inorganic salt decreases the stability of the organosulfate in the aqueous phase. At the same time, the stability of the organosulfate in the organic-rich phase also decreases in the presence of the inorganic salt.

Similar mixing diagrams for 0.09 mole fraction aqueous ammonium sulfate solvent are shown in Fig. S10 and tabulated values in Table S5. In ammonium sulfate solutions, COSMOtherm predicts a water activity of 1.14 in the aqueous solvent-rich phase, indicating that according to COSMOtherm, the 0.09 mole fraction aqueous solution of ammonium sulfate is unstable. This discrepancy with the experimental solubility of xSOL,AS=0.094 is possibly caused by an inadequate representation of the solvation of ionic liquids in COSMO-RS theory .

We calculated saturation vapor pressures for the neutral organic compounds at 298.15 K (Table ). Comparing the studied organosulfate compounds based on their functional groups, those containing carboxylic acid groups, i.e., α-pinene-OS-5 and α-pinene-OS-6, have the lowest saturation vapor pressures. α-Pinene-OS-4, also having an O:C ratio of 0.7, has an order of magnitude higher saturation vapor pressure, indicating that two carbonyl groups are less effective at lowering the vapor pressure than a single carboxylic acid group. In addition, α-pinene-OS-3 (one carbonyl and one hydroxy group) has a lower saturation vapor pressure than α-pinene-OS-4 with one more oxygen atom.

The saturation vapor pressure of sulfuric acid (extrapolated from experimental data using ab initio data) is 2.10×10-3 Pa at 298.15 K , while COSMOtherm estimates the vapor pressure of the pure sulfuric acid to be 7.21×10-2 Pa (about 34 times higher). Due to the previously demonstrated systematic overestimation of absolute saturation vapor pressures by COSMOtherm , we show both absolute vapor pressures and the vapor pressures relative to the estimated sulfuric acid saturation vapor pressure in Table . The saturation vapor pressures of monoterpene- and isoprene-derived organosulfates are all 4 to 8 orders of magnitude lower than that of sulfuric acid. On the other hand, the saturation vapor pressures of IEPOX isomers and methyl bisulfate are higher than for sulfuric acid.

Compared to previously calculated saturation vapor pressures for α-pinene autoxidation products using COSMOtherm, the organosulfates studied here are significantly less volatile . It should be noted, however, that in the study of , the number of intramolecular hydrogen bonds was not limited in the COSMOtherm calculations, which likely led to higher saturation vapor pressure estimates . Furthermore, as we have seen here, the acidic organosulfates readily dissociate in the particle phase, forming ionic species, which will effectively suppress their partitioning to the gas phase.

δ1-IEPOX has a higher saturation vapor pressure than the other IEPOX isomers. This can be understood from a structural point of view, as the lowest-energy conformer (highest weight in the COSMOtherm calculations) of δ1-IEPOX seems to have two intramolecular hydrogen bonds. COSMOtherm does not count either of these as full or partial intramolecular hydrogen bonds in the condensed phase. However, the gas-phase free energy (G(g)) of the conformer is lower than for the other IEPOX conformers, leading to about 5 kJmol-1 of difference in the energy between the condensed and gas phase of δ1-IEPOX and δ4-IEPOX. This in turn leads to a relatively higher saturation vapor pressure (Eq. ) compared to the other IEPOX isomers.

The activity coefficients at infinite dilution in water, the free energies of solvation and Henry's law solubilities in pure water calculated using the different methods (explained in Sect. ) at 298.15 K are given in Table S6. Among the studied organics, Henry's law solubility is the highest for monoterpene- and isoprene-derived organosulfates containing the highest number of oxygen atoms and the lowest for methyl bisulfate and the IEPOX isomers. The infinite dilution Henry's law solubilities (Hsol∞) were calculated by COSMOtherm using Eq. (). We also calculated LLE-based Henry's law solubilities (HsolLLE) using Eq. () with the pure water solubilities of the organic compounds. A comparison between the infinite dilution and the LLE-based Henry solubilities is shown in Fig. . The LLE-based Henry's law solubility for monoterpene-derived OS+water is on average 4.4 times lower than the corresponding infinite dilution Henry's law solubility. Henry's law solubility is the equilibrium ratio between the abundance in the gas phase and in the aqueous phase for a dilute solution. For the fully miscible compounds, including the dissociation correction, the solution is no longer dilute. We therefore did not calculate the LLE-based Henry's law solubility of the isoprene-derived compounds and methyl bisulfate, which are fully miscible with pure water at 298.15 K.

Additionally, we calculated the infinite dilution Henry's law solubilities of all compounds in two organic solvents, hexanoic and cis-pinonic acids (see Fig.  and Table S6). The densities of these organic acids (ρhexanoic=0.9400 gcm-3 and ρcis-pinonic=1.0739 gcm-3) were estimated using COSMOtherm. The Henry's law solubilities of the monoterpene-derived organosulfates are the lowest in water and the highest in cis-pinonic acid. The isoprene-derived compounds (OS and IEPOX) are all less soluble in hexanoic acid than in water. The more oxygenated isoprene-OS-3 and isoprene-OS-4 are also less soluble in cis-pinonic acid than in water, opposite to the less oxygenated isoprene-OS-1 and isoprene-OS-2, which are the most soluble in cis-pinonic acid. The epoxydiols are least soluble in hexanoic acid and the most soluble in water. This means that the phase separation behavior of OS from different precursors will be different in multiphase atmospheric aerosol, leading to different OS aerosol-phase states depending on the predominant precursor.

Figure  shows the infinite dilution Henry's law solubilities for the organic compounds in the aqueous mixtures with different mole fractions of ammonium sulfate (panel a) and ammonium bisulfate (panel b). The decrease in Henry's law solubility is steeper with the increase in ammonium sulfate than in ammonium bisulfate. This is due to the stronger salting-out effect on the organics of ammonium sulfate than of ammonium bisulfate, also seen in the relative solubility calculations. In the case of both inorganic salts, all of the hydroxy sulfates, i.e., α-pinene-OS-1 and α-pinene-OS-2, and all β-pinene and limonene isomers have similar Henry's law solubilities and trends as a function of salt mole fractions. In ammonium sulfate solutions, the Henry's law solubility of isoprene-derived organosulfates decreases more slowly with the increase in ammonium salt concentration than the solubility of monoterpene-derived organosulfates.

COSMOtherm-estimated Henry's law solubility has previously been reported for isoprene-derived 2-methyltetrol , which is similar to isoprene-OS-3 and isoprene-OS-4 with the difference that the sulfate group is replaced by a hydroxy group. We calculated the Henry’s law solubility of 2-methyltetrol in water using COSMOtherm19 and found that the compounds containing a sulfate group (isoprene-OS-3 and isoprene-OS-4) have Henry's law solubilities that are 4 orders of magnitude higher than the compound containing only hydroxy groups (2-methyltetrol). The higher Henry's law solubility of the organosulfate compared to 2-methyltetrol is caused by 5 orders of magnitude lower saturation vapor pressure and an order of magnitude higher activity coefficient at the infinite dilution of the solute. Similar differences are seen between the IEPOX isomers, isoprene-OS-1 and isoprene-OS-2, although the functional groups in isoprene-OS-1 and isoprene-OS-2 (hydroxy and carbonyl) are different than those in IEPOX (hydroxy and epoxy). This means that the presence of sulfate in SOA and the formation of organosulfate compounds enhance SOA formation, since organosulfates are less likely to evaporate than non-sulfate organics.