Interactive comment on “ Inorganic salts interact with organic di-acids in submicron particles to form material with low hygroscopicity and volatility ” by G .

Ault et al. 2013 directly observe the surface enhancement of both Ca and Mg, suggesting this behavior is not specific to Mg and may apply to divalent salts in general. Furukawa et al. (2011) also see higher fractions of oxalate bound zinc to total zinc with decreasing particle size, suggesting that higher surface area-to-volume ratios, and hence the potential of formation of ZnOx at the surface and corresponding surface enhancement. These references are present in the text, and citations are not used in the abstract.


Introduction
The ubiquity of di-carboxylic acids (DCA) is well documented both in urban and rural terrestrial aerosol and marine aerosol (Warneck, 2003;Wang et al., 2006;Wang and Kawamura, 2006;Tan et al., 2010;Sempéré and Kawamura, 1996;Pavuluri et al., 2010;Ma et al., 2013;Kundu et al., 2010;Kawamura et al., 1996Kawamura et al., , 2007;;Gierlus et al., 2012;Altieri et al., 2008).A recent study of marine aerosol from around the globe found that DCA contribute on average ∼ 15 % to total marine organic aerosol (OA) mass (Fu et al., 2013;Myriokefalitakis et al., 2011).DCA are commonly associated with aqueous processing of water soluble organics (Ervens et al., 2011;Tan et al., 2010) and hence readily affect aerosol-water interactions.Oxalic acid (OxA) is typically the dominant DCA found in aerosol (Neusüss et al., 2000).It was shown to contribute more than 50 % on average of the total mass of marine DCA, and mineral dust is often enriched in OxA (Sullivan et al., 2009;Fu et al., 2013).High oxidation states and the ability to strongly interact with inorganic compounds (especially crustal/marine salts) make the effects of DCA on aerosol composition and physical properties unique and complex.However, the impact of such interactions on particles' properties are not well understood.In this study we focus on DCA interactions with inorganic salts and their impacts on hygroscopicity and volatility.
Studies of both pure components of salt/DCA particles (e.g., CaCO 3 , H 2 C 2 O 4 , CaC 2 O 4 ) as well as external mixtures of particles of these pure components indicate that dust or salt particles that include DCA may show reduced hygroscopicities due to formation of low-solubility compounds (Sullivan et al., 2009;Ma et al., 2010Ma et al., , 2013;;Ma and He, 2012).While pure particles of DCA and salts and their external mixtures shed insight on the probable behavior of particles containing these compounds, internal mixtures of these components (single particles containing DCA, their organic salts, and inorganic salts) will also be important in the atmosphere.Several processes may lead to such internally mixed particles, with initial water content largely determining the route to DCA formation and thus probable particle morphology.Gas-phase water-soluble organic compounds (WSOC), such as glyoxal and glutaric acid, will partition to aerosol water.They may then be oxidized in the particle aqueous phase to produce OxA (Warneck, 2003), which can then react with mineral ions in the particle to produce insoluble material (e.g., calcium oxalate), reducing particle hygroscopicity.OxA has also been found to be significant in particles dominated by non-hygroscopic material (e.g., mineral dust) (Sullivan and Prather, 2007).For these particles, solubilizing reactions, such as uptake of nitric acid, may occur at the particle surface, and subsequent water exposure creates conditions ripe for reaction between OxA and di-valent cations (Laskin et al., 2012).Surface reactions between oxalic acid and metal chlorides can produce highly volatile hydrochloric acid, and evaporation of HCl will drive the in-particle equilibrium towards formation of the insoluble M-Ox complex (Laskin et al., 2005).Loss of volatile products occurs at the particle surface, leading to enrichment of the remaining lower volatility material at the particle surface.When present at the particle surface, only small amounts of low-solubility material are required to strongly affect particle hygroscopicity (Schwier et al., 2011).A recent study has directly shown surface enrichment of di-valent material in naturally occurring sea-salt aerosol (Ault et al., 2013).Finally, separate particle populations containing DCA and di-valent metals in an external mixture may combine via coagulation to form internal mixtures (Ma and He, 2012).Thus there are multiple formation pathways for particles consisting of soluble di-/poly-valent salts and DCA, and the hygroscopicity of these particles will depend on both the amount of DCA − −metal complexation and the final particle morphology (i.e., potential surface enrichment of insoluble complexes).
Accurately describing gas-particle partitioning and predicting atmospheric aerosol loading requires knowledge of how low-volatility salts contribute to particle growth and alter the overall volatility of aerosol particles (Lee et al., 2012).Low-volatility atmospheric species have special importance since they participate in aerosol particle growth, in particular the growth of very small, freshly nucleated particles (Kulmala et al., 2004;Pierce et al., 2011;Riipinen et al., 2012).Aerosol particles smaller than 50 nm in diameter are expected to contain compounds with equilibrium vapor pressures as low as 10 −8 -10 −7 Pa (Pierce et al., 2011).Such low vapor pressures can be achieved by organic salt formation via acid-base reactions (Riipinen et al., 2012;Barsanti et al., 2009).For example, OxA has a high vapor pressure (∼ 10 −2 Pa), but the majority of atmospheric OxA resides in the particle phase due to oxalate formation (Ervens et al., 2011 and references therein).The formation of organic salts is also a reactive sink of particle-phase organics, enhancing gas phase uptake (Pierce et al., 2011).Recently, the formation of low-volatility organic salts has been observed under various atmospheric conditions in reactions of organic acids with mineral salts (Laskin et al., 2012), ammonium (Martinelango et al., 2007;Dinar et al., 2008;Ortiz-Montalvo et al., 2012) and amines (Sorooshian et al., 2008;Smith et al., 2010).
Recent field studies, performed both in urban and remote environments, have pointed out the existence of very low-volatility (non-volatile even at temperatures as high as 280 • C) organics in submicron aerosol particles (Wehner et al., 2002;Ehn et al., 2007;Backman et al., 2010;Häkkinen et al., 2012).Häkkinen et al. (2012) observed a positive correlation between low-volatile aerosol material, atmospheric particulate organics, and organic nitrates at a boreal forest site in central Finland, owing to the presence of lowvolatility organonitrates or organic salts.Since the contribution of sea salt (or other mineral salts) in submicron particles at continental forest sites far from the sea is minor (Saarikoski et al., 2005), ammonium or aminium salts are likely main contributors to the growth of aerosol particles, even at sub-20 nm sizes (Barsanti et al., 2009;Smith et al., 2010).However, at marine sites or other mineral-rich areas the contribution of organic salts from mineral-salt-organicacid reactions may have a significant effect on aerosol chemical properties and further cloud processing (Furukawa and Takahashi, 2011;Laskin et al., 2012;Rinaldi et al., 2011;Sorooshian et al., 2013).These results emphasize the fact that atmospheric species, both organic and inorganic, can undergo drastic changes when they react with each other and these changes can alter the chemical and physical properties of particles.
This study addresses the major effects of DCA-cation interactions on two key aerosol properties, hygroscopicity and volatility.Towards understanding these effects, internally mixed particles were created by exposing several metalchloride salts to gas-phase OxA.The hygroscopicity and volatility of the resulting particles was then studied.From the results we assess the applicability of the standard volume additivity rules for hygroscopicity (κ-Köhler theory) and aerosol partitioning of organics that may bind to metals in terms of surface processes and unaccounted inorganicorganic interactions.

Aerosol generation
For hygroscopicity measurements, salt-oxalate aerosols were generated by deposition of gas-phase oxalic acid onto dry, inorganic aerosol.Salt particles were generated from aqueous solutions of CaCl2 q 6H2O (Sigma Aldrich 98 %), MgCl2 q 6H2O (Fisher Scientific 99 %), ZnCl2 (Sigma Aldrich 98 %), NaCl (Fisher Scientific 99 %), and Na2C2O4 (Fisher Scientific 99.5 %).In addition to solutions of each individual salt, three mixtures were used: 8 : 1 and 1 : 1, by mass, NaCl / MgCl 2 , and Instant Ocean.Instant Ocean is a commercially available mixture of salts used as a proxy for sea salt, with the NaCl and MgCl 2 as the main components.The total salt concentration for all solutions was 0.2 M.
Salt solutions were atomized (TSI Model 3076), then diluted with dry nitrogen to low (< 10 % RH) humidity, to induce efflorescence.Aerosols were then introduced to a glass tube coated with a layer of solid oxalic acid, which was heated to a constant temperature.Heating was provided by resistive heating tape with a set applied voltage.The aerosol mixture was then diluted using humidified nitrogen, such that the desired RH was attained.After OxA deposition, the particles were re-humidified to ∼ 10 % RH above deliquescence for the di-valent salts, which under these conditions should always remain as hydrates (e.g., CaCl 2 • 2H 2 O).Rehumidification will hydrate particle surfaces and may create aqueous droplets; causing mixing to facilitate surface chemistry prior to measurement.Just prior to entering the DMA, the particles were passed through a diffusion drier charged with silica bead desiccant.The air stream exiting the drier was always less than 10 % RH (measured via RH probe; Vaisala), leaving minimal residual particle water.Separate test experiments with a second drier to reach less than 2 % RH did not show significantly different results.
Aerosol size distributions were measured via SMPS (TSI Model 3775).Upshifting of the dry aerosol mean diameter was used to estimate the amount of adsorbed oxalic acid onto aerosols after being exposed to the heated tube.cloud condensation nuclei (CCN) activity for all particles was measured using a Constant flow rate streamwise thermal gradient CCN counter (CFSTGC, Droplet Measurement Technologies) (Lance et al., 2006;Roberts and Nenes, 2005).
For volatility studies, Na 2 Ox (Na 2 C 2 O 4 ) and oxalic acid (C 2 H 2 O 4 ) particles were generated slightly differently than the CaOx particles.Na 2 Ox aerosol was produced by atomizing a solution of sodium oxalate (Fisher Scientific, > 99.5 % purity) and deionized water and OxA aerosol by atomizing a solution of oxalic acid (oxalic acid di-hydrate, Fisher Scientific, > 99.5 % purity) and deionized water.These particles were passed through a diffusion drier charged with silica bead desiccant prior to heating.The mean number concentrations and geometric mean diameters of the generated particles (oxalic acid, Na 2 Ox and CaOx) measured with a SMPS (during oxalic acid and Na 2 Ox experiments SMPS by TSI, during CaOx experiment SMPS by Grimm Technologies) before introducing the aerosol to the temperature programmed desorption aerosol-chemical ionization mass spectrometer (TPD) aerosol-CIMS are presented in Table 1.

Temperature programmed desorption aerosol-CIMS
TPD aerosol-CIMS (temperature programmed desorption aerosol-chemical ionization mass spectrometer) consists of a quadrupole mass spectrometer coupled with a heated volatilization flow tube inlet (McNeill et al., 2007).Compared to more traditional instruments which measure aerosol volatility by detecting change in aerosol particle diameter upon heating (Philippin et al., 2004;Wehner et al., 2005;Ehn et al., 2007), TPD aerosol-CIMS detects changes in volatilized aerosol organics in the gas phase via CIMS.In the heating region, the aerosol stream flows through a PFA tube (OD 1.27 cm, length 23 cm) wrapped with a heating tape (VWR).Diluted aerosol flow of 1.8 L min −1 from the total flow of 5-6 L min −1 enters the heating tube giving the aerosol a residence time of less than a second in the tube.
The temperature in the heated region is controlled by a temperature controller (Staco Energy Products Co.) and monitored using a thermocouple.The aerosol (gas-particle mixture) is then introduced to an ionizer where the gas-phase analyte molecules (i.e., oxalic acid) react with I − reagent ion to form clusters (see details in Sareen et al., 2010).Ion clusters are guided from atmospheric pressure to ultrahigh vacuum (∼ 10 −8 Torr) using differential pumping and the masses (mass-to-charge ratio, m/z) of these clusters are determined using a quadrupole mass spectrometer.Aerosol evaporation is observed as an increase in signal when the When investigating the volatility of oxalate salts (Na 2 Ox, CaOx), the heating temperature was increased step wise using mean inlet temperatures from 25 • C to 180 • C (altogether at 8-9 temperatures).Signal response for a given molecular cluster to temperature change was rapid.It took around 30 to 90 min to stabilize per temperature setting.The studied oxalate salts were observed as clusters of oxalic acid and I − in the TPD aerosol-CIMS.Detecting organic salts in the CIMS as their corresponding acids has been observed before (McNeill et al., 2007).It is known that gas-phase oxalic acid can thermally decompose at sufficiently high temperatures (∼ 100 • C) forming formic acid and CO 2 (Higgins et al., 1997).This was taken into account when performing oxalate salt experiments; instead of tracing only oxalic acid (217 ± 0.5 amu, I − • C 2 H 2 O 4 ) also formic acid (173 ± 0.5 amu, I − • CH 2 O 2 ) was traced and the total oxalate signal was taken as the sum of these two signals (e.g., Fig. 4).The observed oxalate signal was 1-10 % of the mass spectrometer signal of the reagent ion, which was ∼ 100 kcps.
The calibration of the TPD aerosol-CIMS was done with pure oxalic acid aerosol observed as a cluster of I − (see the "Enthalpy of vaporization" and "Volatility of oxalate salts" sections for more about the calibration and other data analysis).All of the TPD aerosol-CIMS experiments were performed at relative humidities below 10 % in order to keep the contribution of water negligible (Saleh et al., 2010;Cappa et al., 2007).This was done by drying the particles using silica gel dryer after particle generation.Due to minor contamination inside the ionization regime and/or in the tubing, a constant background of oxalate was always present.Before every experiment the background was determined and removed from the observed oxalate signal in the data analysis.

Hygroscopicity and CCN activity
Hygroscopicities of aerosol mixtures were compared using κ, a single parameter representation of CCN activity (Petters and Kreidenweis, 2007), where critical supersaturation, s c , is the supersaturation required for a given particle of dry diameter, D d , to activate.
A is a constant-temperature parameter expressing particle curvature effects on vapor pressure where σ s/a is the solution surface tension at the point of activation, M w is the molecular mass of water, R is the universal gas constant, T is the solution temperature, and ρ w is the density of water.The surface tension of water, 0.072 J m −2 , was used for all κ calculations.For mixtures of soluble component species, κ can be approximated as a summation of its volume-fraction-weighted terms (Petters et al., 2007): where κ i and ε i are the individual component hygroscopicities and volume fractions, respectively.Using the volume of oxalic deposition onto the inorganic aerosol, and literature and measured values of pure substance hygroscopicities, or κ values, (Tables 1 and 2) were calculated taking into account the conversion of oxalic acid and inorganic salts to their respective oxalate salts.The calculated values

Enthalpy of vaporization
The effective enthalpies of vaporization of the aerosol constituents ( H vap , kJ mol −1 ) can be determined using TPD aerosol-CIMS, by applying the Clausius-Clapeyron equation as follows: where R is gas constant, T is mean temperature of the gas in the flow tube and the mean observed signal, signal(T ), is proportional to the effective vapor pressure of the studied aerosol species.H vap depends on temperature, however, it can be approximated as a constant in relatively narrow temperature ranges.Hence, Eq. ( 4) can be considered a linear equation and H vap can be determined from the slope of the curve fitted to the experimental data using least-square method that takes into account known covariance.The covariance of each data point (mean signal at temperature T ) was the standard deviation of the aerosol-CIMS signal squared.The curvature (Kelvin) effect on vapor pressure was found to be insignificant in the calculation of H vap (Seinfeld and Pandis, 2006).According to calculations ( 2σ M ρ l r p vs. H vap, CIMS , where σ , M and ρ l are surface tension, 72 mN m −1 , molar mass, 126.07 g mol −1 , and liquid density, 1.653 g cm −3 , respectively) the influence of Kelvin effect on the H vap is minor (below 1 % with r p of 20 nm).Equation (4) also assumes compounds do not react during the heating process.

Results and discussion
CCN activity expressed by log-log plots of d p vs. s c for sodium-dominated particles are shown in Fig. 2. Pure salt particles are shown in solid curves while salt particles with OxA deposited are shown with dashed curves.For reference, pure sodium oxalate and ammonium sulfate are shown by the thick-dashed red line, and the thick black line, respectively.Following an analysis similar to that of Sullivan et al. (2009), we can calculate intrinsic κ values, κ intr (i.e., hygroscopicity if the material is completely soluble), to aid the interpretation of observed hygroscopicites, where ρ s is the density and M s the molecular weight of the material, ρ w is the density and M w the molecular weight of water, and ν is the number of ions for complete dissociation of the solute.These values are shown in Table 2 for monovalent salts and their hydrates (densities from CRC handbook).While our measurements for the pure mono-valent chloride aerosols (with no oxalate) yield κ values that are consistent with previously measured values, measurements for mono-valent salts containing oxalate are similar to those of salt-hydrates, even after drying.This is consistent with the findings of Ma et al. (2013), which suggest the persistence of hydrates under dry conditions, without heating in addition to drying.Addition of oxalate to sodium (mono-valent) dominated salt particles has only a modest effect, due to their high solubility.From the results of Sullivan et al. (2009) large increases in hygroscopicity may result from mixing small amounts of highly soluble material to lower solubility material that has a high intrinsic hygroscopicity.Added water solubilizes the low-solubility core and induces non-linear response in mixing (Nenes et al., 2002, Kokkola et al., 2008).The orange points in Fig. 2 show mixtures of NaCl with the other main component in sea salt, MgCl 2 .The two mixtures are 8 : 1 and 1 : 1 NaCl/MgCl 2 , and it is clear that deposition of OxA to particles with greater di-valent salt content creates significantly less hygroscopic particles.The measured hygroscopicities for these mixtures in fact agree well with the calculated κ intr , as might be expected for particles with dominant (8 : 1 NaCl/MgCl 2 ) and significant (1 : 1 NaCl/MgCl 2 ) amounts of highly soluble NaCl.
The CCN measurements for particles dominated by divalent salts, both pure and after OxA deposition, are shown in Fig. 3.In all cases, deposition of OxA incurs a very large decrease in hygroscopicity to the salt particles, as indicated by substantial increases of the critical diameter at a given supersaturation.For the pure di-valent salt particles, our experimental values agree closely with the κ intr for the chloride hydrate salts rather than the pure salts.(Table 3).As for the mono-valent salts, this is reasonable because fully removing water from the deliquesced salt particles can require heating in addition to low humidity (Ma et al., 2013).Table 3. Solubilities (sol), intrinsic κ values (κ intr ), and measured κ values (κ meas ) for di-valent metal salts and their hydrates (Cl-Hyd).κ intr were calculated using Eq. ( 5) with densities from the CRC handbook.
Aerosol Species ρ/ρ hydrate (g cm −3 ) Cl-sol (g L −1 ) Cl-κ intr (Cl-Hyd)-κ intr Cl-κ meas  The s c vs. d p curves for the mixtures of di-valent salts and OxA (Fig. 3) are not linear over the full range of measurements.For example, the d p50 for CaCl2 + OxA at an s c of 0.4 is roughly 30 nm below the linear trend of the values at high s c .This dependence on size, and hence surfacearea: / volume ratio, suggests that surface effects are at play.The process of CCN activation begins with water adsorption and is then governed by component volume fractions, so the surface-area / volume ratio of the particle, which is size dependent, may be affecting particle hygroscopicities.
Calculating the κ values of the final mixtures present after OxA deposition is slightly complicated by the volatility of co-formed HCl upon reaction of the salt with oxalic acid.As has been shown by Laskin et al. (2012), drying mixtures of di-acids (malonic, etc.) with chloride salts can cause HCl to be expelled from the particle, as evidenced by a 30 % loss of Cl upon reaction.Given that particles in our study were prepared in a similar manner to the Laskin et al. (2012) approach, chloride depletion is likely to occur for the di-valent salts as has been observed for NaCl.Cl depletion must be taken into account when estimating the fraction of OxA actually deposited to our salt particles, because we measure the particle diameter after evaporation of HCl would have occurred.Thus, the final particle volume can be expressed as the sum of the component volume fractions multiplied by their respective molar volumes.Assuming negligible changes to the individual component molar volumes during reaction where V final is the measured particle volume after OxA deposition and any chemistry that occurs after OxA deposition, V init is the initial, pure-salt particle volume, V react is the volume of the pure salt that reacts, V prod is the volume of the  reaction products, f react is the fraction of oxalic acid reacted, GF is the ratio of the final-to-initial particle diameter, and β tot is the sum of the molar volume of the products divided by molar volume of the pure salt.The volatility of HCl gives rise to two limiting cases in which HCl is treated as completely remaining in the particle and one in which HCl completely evaporates from the particle.The first case is given by and the second case (complete HCl evaporation) is given by where now and β M-Ox is the stoichiometry-weighted (n M-Ox /n salt ) ratio of the molar volume of M−Ox to the initial pure-salt molar volume.
The assumption that the final particle volume was measured after complete reaction and evaporation of HCl results in the lowest prediction of the final particle hygroscopicity, because the particle would then contain the maximum possible fraction of insoluble, low-hygroscopicity material.These results, with the measured growth factors after OxA deposition, are listed in Table 4.The κ calc data in Table 4 are calculated using Eqs.( 8) and ( 9).Growth factors were calculated from particle geometric mean diameters.Some variation in composition (amount of OxA deposition) with size is expected.Three factors suggest that even small particles do not have excessive volumetric fractions of OxA: the linearity of s c vs. d p for the mono-valent salt particles, the fact that their hygroscopicities are well below that of pure OxA, and the agreement between their calculated and measured values of κ.The measured values for d p50 will only be affected by changes in composition for particles with a diameter near d p50 and above.Any size-dependent deposition should be biased towards smaller particles, due to larger surface area to volume ratios.The complete reaction is in fact not likely to occur, as the Cl / Na ratios for di-acid particles with initial 1 : 1 molar ratio of acid to salt observed by Laskin et al. (2012) never reached zero.While the observed particle hygroscopicities (after OxA deposition) for di-valent salts are low (0.02-0.05), the lowest predicted hygroscopicity is 0.30 for MgCl 2 .The measured κ values are close to the values expected/measured for the pure M-Ox compounds, which have fairly high κ intr that are not realized due to their low solubilities (Sullivan et al., 2009).
The discrepancies between the measured and predicted particle hygroscopicities can be attributed, at least in part, to particle morphologies with inhomogeneous composition, such as a coating.A possible explanation is the formation of an insoluble M-Ox coating at the particle surface.This is supported by the similar κ values for our salt particles with OxA deposition and the pure M-Ox particles measured by Sullivan et al. (2009).Recently, species present at the particle surface, such as surfactants, have been shown to strongly affect CCN activation (Sareen et al., 2013).Though M-Ox compounds may reduce particle surface tensions to as low as 32 mN m −1 (Wu and Nancollas, 1999), we find good agreement between our measured and calculated κ values for the NaCl / MgCl 2 mixtures using a surface tension of 72 mN m −1 .Thus a coating of the insoluble M-Ox may create a particle that, in terms of CCN activation, behaves as if the particle were composed entirely of the insoluble M-Ox salt.This potential scenario is also supported by the fact that evolution of HCl from the particle, which leaves behind insoluble M-Ox, must occur at the particle surface.Finally, a recent study has directly shown the surface enhancement of both Ca and Mg, suggesting this behavior is not specific to Mg and may apply to di-valent salts in general (Ault et al.,

2013
).Thus, organic coatings, which may lead to formation of insoluble organic salts, especially in regions with di-valent salts, may be prevalent in the atmosphere.

Volatility of oxalate salts
The TPD aerosol-CIMS was calibrated using oxalic acid with well-known H vap (83 kJ mol −1 at T = 190 • C, Yaws, 2003).In order to determine the H vap of oxalic acid aerosol, the average oxalic acid gas phase signal at three different temperatures (mean inlet temperatures 25, 39, 56 • C) was divided by the average signal at room temperature (25 • C) and its logarithm was plotted against the inverse mean inlet temperature (Fig. 5, upper panel).This way the Clausius-Clapeyron equation could be directly applied (Eq.4) and H vap could be estimated from the slope of the calibration curve.In the temperature range of 25-56 • C, we obtained H vap of (85 ± 12) kJ mol −1 for oxalic acid (Table 1), comparable to the literature value for H vap of oxalic acid (95 kJ mol −1 ) (Yaws, 2003).A residence time of one second may not be sufficient to fully equilibrate the aerosol (Riipinen et al., 2010).However, high-particle concentrations and small-particle sizes yield shorter equilibrium times, and in our case particles concentrations are high and particles relatively small, compared to the conditions of Riipinen et al. (2010).4 and Fig. 5).Hence, the oxalate salt particles were significantly less volatile compared to pure oxalic acid aerosol.In other words, the energy needed for the vaporization of oxalates, H vap , was very large at temperatures below 75 • C, whilst H vap was (85 ± 12) kJ mol −1 for pure OxA.Similar to our study, high onset temperatures of evaporation have been observed also with other atmospheric organic constituents (Tritscher et al., 2011;Meyer et al., 2009).Tritscher et al. (2011) reported onset evaporation temperature of around 70 • C for citric acid and both Tritscher et al. (2011) and Meyer et al. (2009) observed onset evaporation temperature at around 50 • C for secondary organic aerosol from α-pinene oxidation.
At temperatures above 75 • C, where volatilization of oxalates was observed to be significant, the vapor pressure of oxalates did not follow a strictly exponential dependence with T (Fig. 5), in contrast to OxA aerosol (Riipinen et al., 2006(Riipinen et al., , 2007)).It is possible that the thermal decomposition of OxA causes a convolution of two separate signals with slightly different onset temperatures.The complexity of the thermal decomposition was not further studied, because it is not atmospherically relevant.However, we acknowledge that it has some influence on our results at higher temperatures and may be important in thermodenuder experiments.In order to estimate the energy needed for the vaporization of oxalate salts at temperatures > 75 • C, the H vap in this temperature range was determined using a linear fit to only the data above 75 • C (see the lower panel of Fig. 4 for the example linear fit).The average of the obtained H vap values for oxalate salts was (72 ± 12) kJ mol −1 (Table 1).As discussed above, this value should be taken only as an estimate.TPD aerosol-CIMS results show that, in the case of CaOx and Na 2 Ox, OxA/oxalate is sequestered as very lowvolatility organic salt compounds.

Conclusions and atmospheric implications
Deposition of OxA to submicron particles containing di-valent salts creates very low hygroscopicity material.Dramatic increases in the CCN activation diameter, up to 50 nm, for relatively small particle mass fractions of OxA (10-20 %) with di-valent salts (e.g., CaCl 2 ) indicate that standard volume additivity rules for hygroscopicity do not apply.High-oxidation state, soluble organic species can strongly decrease hygroscopicity in particles dominated by di-valent salts.Our results are specific to OxA, but strong binding may occur for other DCA.Highly oxidized, soluble organic species, particularly including carboxylic acid groups, can strongly decrease hygroscopicity in particles dominated by di-valent salts.Given current knowledge of the formation mechanisms of OxA and M-Ox salts, surface enrichment of insoluble M-Ox salts is expected, explaining the greater than stoichiometric effects of M-Ox salts on particle hygroscopicity.An analogous situation does not occur for particles dominated by mono-valent salts (NaCl), as κ values from the standard κ-calculation for a mixture Eq. (3) agrees with our measurements.A threshold of soluble material is likely met in particles dominated by mono-valent salts, such that overall solubility is high enough to reach ideal behavior.Particles containing oxalate salts, whether pure or formed by OxA deposition to pure salt particles, have very low volatility, indicated by a 75 • C "onset-temperature" for detection of OxA using TPD aerosol-CIMS.A coating of insoluble material such as CaOx could result in particles that are hard enough to exhibit bounce on particle impactors and, combined with the low volatility of these salts, may also affect mass transfer properties (evaporation/uptake) of such coated particles.The formation of an insoluble and hard coating of organic salts, supported by the discrepancy between our calculated and observed hygroscopicities, could thus appear in experimental observations (mass transfer limitations and bounce factor) as a glassy, highly viscous state.Because of the potential implications for partitioning/volatility, uptake coefficients, evaporation, and hygroscopicity, the irreversible transformation of volatile, soluble, and high-oxidation-state organic material into non-volatile material may be important in models of organic compound partitioning in the atmosphere.
Edited by: A. Laskin

Fig. 1 . 28 Figure 1 .
Fig. 1.A schematic diagram of the experimental setup for CCN measurements.figure

Fig. 2 .Figure 2 .
Fig. 2. Critical supersaturation, s c , of monovalent salt particles as a function particle diameter.Results for inorganic salts without OxA deposition are shown by solid lines: NaCl (red), Instant Ocean (blue), Na 2 Ox (grey).The corresponding salts with OxA deposition are dashed lines with the same color.The orange, thick-dashed and thin-dashed lines are for the 1 : 1 and 8 : 1 mixtures of NaCl : MgCl 2 with OxA deposition, respectively.Pure OxA and ammonium sulfate are shown in solid magenta, and black lines, respectively.κ values are shown for NaCl (red), Na 2 Ox (gray), and OxA (magenta).

Figure 4 .
Figure 4. Volatility of Na 2 Ox aerosol with TPD aerosol-CIMS (Na 2 Ox (1)).The detected oxalate signal when aerosol was heated stepwise from room temperature to 160 • C as a function of time (minutes from the beginning of the experiment) is illustrated in the figure (black solid line).Also the average oxalate signal at different temperatures is plotted (blue dots).The average background oxalate signal was subtracted from the oxalate signal for further analysis.

Fig. 5 . 34 Figure 5 .
Fig. 5. Evaporation of oxalate particles observed with the TPD aerosol-CIMS -natural logarithm of the oxalate signal normalized with the signal at room temperature as a function of inverse mean inlet temperature.Oxalic acid calibration curve (blue), two independent experiments with Na 2 Ox (green) and CaOx along CCN experiment (red) are illustrated in the upper panel.Also the approximate onset temperature of evaporation of oxalate salts is marked (75 • C).In the lower panel the estimation of the ∆H vap by fitting the data points to Clausius-Clapeyron equation (Eq.4) above this temperature is shown for Na 2 Ox particles (Na 2 Ox (1)).Error bars illustrate the standard deviation of the signal from the mean signal.

Table 1 .
Physical properties of the generated particle size distribution -mean geometric size and mean concentrations of the particles during the TPD aerosol-CIMS experiments are presented.Onset temperatures for evaporation and H vap of oxalate at different temperature ranges; the number of data points used in the linear fits are in parentheses.Experiments were performed using oxalic acid aerosol, sodium oxalate Na 2 Ox and calcium oxalate (CaOx).In this study mean temperatures at the VFT centerline were used in the data analysis.Linear correspondence between the temperature of the flow tube external wall and the actual VFT temperature was determined in the range of 25-114 www.atmos-chem-phys.net/14/5205/2014/Atmos.Chem.Phys., 14, 5205-5215, 2014 • C. See Sareen et al. (2010) for a more detailed description of the instrument used seeSareen et al. (2010).

Table 2 .
κ values for mono-valent salts.κintrcalculatedusingEq.(5).The densities are from CRC handbook, except for Na 2 Ox−(4−H 2 O), which was assigned a density of 1.52 g cm −3 .κreflect a theoretical range of hygroscopicities that can arise from reaction and bulk mixture effects alone.Deviations from this range of values suggests additional compositional effects not accounted for by mixing alone.The CCN activity of the aerosol is defined by the minimum dry diameter, d p50 , at which 50 % of the particles activate at a given s c .d p50 is determined by fitting a sigmoid curve to the measured ratio of CCN to CN concentrations as a function of dry particle diameter, accounting for multiply charged particles as inMoore et al. (2010). for

Table 4 .
Growth factors (V M+OxA / V M ) with uncertainties and κ values for salts and salt mixtures with OxA deposition.Equations (8) and (9) are used to calculate κ calc .Our measured component κ values are used for the individual components' κ i , and a GF of 1.2 was used.