Atmospheric Ageing of Inorganic Sea Spray Aerosol: Implications for Hygroscopicity and Cloud Activation Potential


 Sea spray aerosol (SSA) contributes significantly to natural aerosol particle concentrations globally, in marine areas even dominantly. The potential changes of the omnipresent inorganic fraction of SSA due to atmospheric ageing is largely unexplored. We demonstrate that ageing of liquid NaCl and artificial sea salt aerosol by exposure to ozone and UV light leads to a substantial decrease in hygroscopicity and cloud activation potential. The results point towards surface reactions that are more crucial for small particles and the formation of salt structures with water bound within the aerosols, termed hydrates. Our findings suggest an increased formation of hydrate forming salts during ageing and the presence of hydrates in dried SSA. Field observations indicate a reduced hygroscopic growth of sub-micrometre SSA in the marine atmosphere compared to pure NaCl which is typically attributed to organic matter or sulphates. Aged inorganic sea salt offers an additional explanation for reduced hygroscopicity and cloud activation potential.


13
The oceans contribute the largest mass of natural aerosol particles to the atmosphere 1 in the form of sea spray aerosol (SSA). 14 SSA is produced due to stress applied by winds on the ocean surface, which causes waves to form and break. In this process, 15 air bubbles are entrained in ocean water. When entrained air bubbles rise to the surface they burst, and sea spray droplets are 16 released to the atmosphere. Super-micrometer SSA dominate the emissions by mass but the largest number of SSA is produced 17 in the sub-micrometer size range peaking at dry diameters near 100 nm 2-5 , which implies the presence of large numbers of 18 particles potentially relevant for cloud formation. The chemical composition of SSA is complex and reflects to some extent the 19 composition of the seawater from which it originates (e.g. 6 ) although enrichment of inorganic 7 as well as organic constituents 20 (e.g. 8 ) have been reported. While several studies have shown that the organic fraction in SSA is variable depending on location 21 and time of the year 9-11 , the inorganic fraction is ever-present and a key ingredient across all size ranges 10 . The hygroscopicity 22 of sea salt has in many years been approximated with that of NaCl but as shown by King et al. 12 the CCN forming ability of 23 sea salt particles differs from that of NaCl and e.g. Zieger et al. 13 showed that the hygroscopic growth of sea salt is reduced 24 compared to that of NaCl. Rasmussen et al. 14 show and Zieger et al. 13 conclude that the difference in hygroscopicity for 25 inorganic sea salt and NaCl is likely due to hydrates. 26 Hydrate forming salts contain water molecules within their structures even after extensive drying and thus fundamentally impact 27 water uptake. MgCl 2 and CaCl 2 are examples of hydrate forming salts present in sea salt aerosol. A recent study by Rosati et 28 al. 15 shows that aerosolized MgCl 2 and CaCl 2 particles exhibit a substantially reduced water uptake potential compared to 29 predicted values assuming anhydrous salts. Experimental and predicted values could be reconciled when taking into account 30 the hydration states of the dried particles. Thus, measurements of sea salt water uptake at sub-and supersaturated water vapour 31 conditions are profoundly affected by hydrate forming salts. Their presence in dried particles leads to a reduced water uptake 32 which could at least in part be falsely ascribed to other causes, such as the presence of organic matter.
hygroscopicity. Based on the hygroscopic growth factor particles are often divided into sub-categories: nearly hydrophobic 48 particles (NHP), less-hygroscopic particles (LHP), more-hygroscopic particles (MHP) and the even more hygroscopic sea-salt 49 particles (SSP) 29 . As pointed out by Swietlicki et al. 29 the MHP, which are significantly less hygroscopic than pure NaCl 50 particles, have been found to be ubiquitous in the marine environment whereas observation of externally mixed sea salt particles 51 with a distinct SSP-group are limited to high wind speeds. This was also highlighted by first successful eddy covariance flux 52 measurements over the sea, demonstrating a clear GF dependence on wind speed 30 . Several propositions exist to explain the 53 nature of the reduced hygroscopicity of particles in the marine environment: sea spray may not be present in large enough 54 numbers over the oceans, the particle distribution may be dominated by sulphate containing particles resulting from dimethyl 55 sulphate oxidation, and particles may contain organic species as a result of atmospheric ageing or combinations of these 56 options 29 . An enrichment in organic matter for particles with diameters smaller than 1 µm has been proposed in several studies 57 to cause a reduction in hygroscopicity 8, 23, 31-37 , although results from different studies vary quite substantially 4, 38 .  the air-water interface 42 . These studies focused on the gas-phase products of such reactions, neglecting the effects on SSA itself.

67
The experiments by Oum et al. 39 were further repeated with pure NaCl, investigating changes in the chemical composition of 68 the aerosols 43 . For particles with diameters below 1 µm it was found that such a reaction gradually depletes the amount of Cl − 69 and consequently replaces it by oxygen (O), hence leading to the formation of NaOH inside the liquid particles 43 . Effects on 70 the physical properties or the use of a complex salt mixture were not investigated.

71
The current study was designed to explore how ageing by exposure to ozone and UV light affects the hygroscopicity and cloud 72 activation potential of inorganic sea spray aerosol. Through a series of experiments in a well-controlled laboratory environment 73 we examined the hygroscopicity and cloud activation potential of NaCl and artificial sea salt particles of different particle sizes.

74
Additionally, the volatility of the particles was investigated and microscopy analysis was performed to yield insight into the 75 elemental composition of fresh and aged particles.

77
Hygroscopicity and cloud activation potential 78 Sea spray aerosols were generated from aqueous solutions of artificial sea salt (termed sea salt from now on; salinity of 3.5%) 79 or NaCl using a circular plunging jet 20 or an atomiser and injected into a 5 m 3 Teflon chamber 44 at 291 K. Typical particle 80 number concentrations were of the order of 5000 and 15000 #/cm 3 at the beginning of the experiments using the plunging jet 81 and atomiser, respectively. The water uptake by the salt particles was examined with a humidified tandem differential mobility 82 analyser, a cloud condensation nucleus counter and a humidified nephelometer. Four different scenarios were investigated: I) 83 crystalline particles in the dry chamber exposed to O 3 , II) crystalline particles in the dry chamber exposed to O 3 and UV lights, 84 III) liquid particles in the humid chamber exposed to O 3 and IV) liquid particles in the humid chamber exposed to O 3 and 85 UV lights. Table 1 provides an overview of experimental conditions. Figure 1  plunging jet and a nebulizer), for NaCl and sea salt, respectively 13 .

98
For experiments performed at dry conditions the NaCl and sea salt particles are expected to be crystalline in the smog chamber 99 (scenarios I and II, Fig. 1a   For particles aged under humid conditions with O 3 (no UV light) some decrease in GF(80%) with time is observed (scenario 108 III, Fig. 1c). There is some debate as to the detailed chemistry taking place when aqueous NaCl or sea salt is exposed to O 3 in 109 the dark, the formation of chlorates was for example proposed 49-51 which might influence hygroscopicity.

110
A remarkable result in Fig. 1 is the large decrease in GF(80%) when small (50 nm) liquid NaCl and sea salt particles are aged in 111 the presence of O 3 and UV lights, which we ascribe to a strong effect of OH chemistry on hygroscopicity (scenario IV, Fig. 1d). according to model simulations can be ascribed to wall losses, including sedimentation, and coagulation (more details on the 126 model are presented below). However, this will not effect the qualitative particle properties investigated in this study. . a) Scattering enhancement factor (f(RH)) measured at 3 different wavelengths (λ =450 nm, λ =525 nm, λ =635 nm) and RH in the nephelometer. b) number (red lines; left y-axis) and surface distributions (blue lines; right y-axis) at the beginning, after 5 hours and mean distributions throughout the experiment as derived from combined scanning mobility particle sizer and optical particle spectrometer measurements.
Simultaneously, the critical super saturation (SS crit ) for dry, monodisperse 50 nm particles was measured (Exp. 2, 16). This 128 diameter falls in the key size range relevant for super saturations typically experienced in cloud systems in oceanic regions 52 .

129
The results in Fig. 3  Hence, the cloud activation potential of such SSA is evidently reduced by the simulated ageing. These findings are in accordance 134 with the hygroscopicity results presented above.

135
Although all experiments were performed with particles generated from inorganic salts, we cannot fully exclude a small 136 contamination by organic compounds. We find that it is highly unlikely that organic particle mass can be responsible for 137 the observed changes in hygroscopicity and cloud activation for to the following reasons: First, the water uptake potential found in scenario IV for 50 nm particles show that the organic fraction has to be at least 70% to account for a lowering of the 141 growth factor from 1.9 to 1.4 (see Fig. S4), which is very unlikely. Third, a contamination by organic compounds in the initial 142 stages of the experiment by e.g. being present in the used salt solution would yield lower GF values right from the beginning.

143
Organic contamination in later stages would be oxidised through ageing typically leading to an increase in hygroscopicity 144 but it would still be much less hygroscopic compared to the salt particles. Forth, simulations with the Aerosol Dynamics gas- non-spherical particles need to be corrected with a dynamic shape factor to account for differences in shape. According to 162 Zieger et al. 13 fresh salt particles generated in the laboratory that appear cubic-like need a dynamic shape correction factor 163 of approximately 1.10 to recalculate volumetric particle diameters from the measured mobility diameters. By applying this 164 correction, the GF values increase by approximately 10%. Therefore, the GF(80%) illustrated in Fig. 1  occurring at the surface of the particles (Fig. 4c; Exp. 2).

173
The elemental composition can also be used to infer a possible contamination by e.g. organic matter or NaNO 3 . Results for 174 fresh NaCl particles (Fig. 4a) do not suggest the presence of either of them at any of the selected sizes as no oxygen was 175 measured. The increased oxygen signal for the aged particles (Fig. 4b)  found as a shell around the particles producing an oxygen signal concentrated around the particles rather than inside. Although 179 some of NaNO 3 could have evaporated during our microscopy measurements, results of aged NaCl in Fig. 4 do not suggest 180 a core-shell structure but an evenly distributed oxygen signal. Finally, Fig. 4b and c show that aged NaCl particles are still 181 predominately composed of Na and Cl, thus suggestion a very limited potential presence of NaNO 3 .   and NaClO 4 as monohydrate 63 . The GF(80%) for 50 nm NaCl particles was shown to continuously decrease nearly linearly 195 after ageing, indicating a change in the chemical composition within the droplets (Fig. 1). The presence of hydrated salts can 196 explain this phenomenon, as when exposed to high RH a smaller amount of water can be taken up due to the fraction of water 197 molecules already trapped within the crystal. To test this hydrate hypothesis, we dried (RH<10%) and subsequently heated the 198 NaCl particles with a thermodenuder, set to a temperature of 300 • C, before exposing them to RH=80%. The results in Fig. 6a 199 (Exp. 2 and 13) illustrate that this procedure leads to an immediate change in GF(80%) for D dry =50 nm. Each time the heating 200 is turned on, aged and heated NaCl (dark red) increases its GF(80%) to values comparable to those for the fresh NaCl particles. be detected (Fig. S7).

203
Sea salt is a complex inorganic mixture containing compounds like MgCl 2 and CaCl 2 , which are known to exist in hydrated 204 forms under various conditions 61 . This was recently suggested to influence the interpretation of thermodenuder data 14 and 205 to be the reason for the slightly lower hygroscopicity of fresh sea salt particles compared to pure NaCl particles 13 . This 206 small difference is also visible in this study (Fig. 6). When sea salt is dried (RH<10%) and additionally heated to 300 • C, the 207 GF(80%) increases rapidly to values slightly higher than those for fresh sea salt particles. This implies that at 300 • C, water 208 bound as hydrates in the compounds formed during ageing, but also hydrates present even in the freshly generated particles are 209 evaporated.

210
The GF(80%) of particles with D dry =200 nm (Fig. 6b) is also affected by the heating process, reaching values comparable to 211 those at the beginning of the experiment in the case of NaCl and slightly higher values for sea salt. In the marine atmosphere,

212
SSA is close to equilibrium with respect to the atmosphere and therefore, it is mostly in its liquid state. In order to investigate 213 its hygroscopicity, the liquid particles have to be dried most certainly leading to several hydrated species. The ageing reactions 214 described in our study increased the amount of hydrate forming compounds over time. There is no real reason to assume that 215 formation of hydrate forming salts would not take place also in the marine boundary layer and hence has an effect on the 216 aerosols' physical and chemical properties and ultimately their role in climate.

218
The unique setup combining an environmental chamber with a sea spray simulation tank directly targets the evolution of SSA in the clean marine boundary layer. Although the particle number concentrations and hence particle surface in the chamber 227 was partly substantially higher compared to the real atmosphere, the water uptake properties of sea salt and NaCl at different will thus overestimate their GF as previously suggested 13 , but our data suggests in addition that this problem will increase over 239 time if the models do not also include the ageing processes of sea salt: Cl 2 losses, oxidation and hydration. explanation for the lower GFs. 257 We also show that slightly larger particles (D dry =200 nm) are not equally affected by the same amount of ageing. The analysis 258 of the chemical composition using microscopy techniques accentuates processes occurring mainly at the surface of the particles, 259 which are more decisive for smaller sizes and could explain the differences found for the two particle sizes after ageing, as 260 particles with D dry =50 nm are four times as easily changed in bulk composition due to surface processes than 200 nm particles.

261
The new oxygen containing compounds formed in aged, pure NaCl particles are most likely NaOH and Na-chlorates. Both are 262 known to form hydrates at low RH, which strongly influence the hygroscopic properties of particles. By additionally heating the 263 particles to 300 • C, the initial GF(80%) was restored, suggesting that this temperature was sufficient to decompose the hydrated 264 compounds leaving a particle predominately containing NaCl behind. Hydrates hinder the water uptake as they already contain 265 water molecules in their crystalline structure reducing the amount of water that can be taken up at elevated RH. In the case of  was utilised in combination with the HTDMA. In this set-up the TD was set to a temperature of 300 • C and placed downstream 311 of the HTDMA drier but upstream of the first DMA. In this way the hygroscopic behaviour of not only dried but also thermally 312 treated particles was analysed.

313
Hygroscopicity was additionally probed with a commercially available humidified nephelometer system employing a three 314 wavelength nephelometer (Ecotech Pty Ltd., Aurora 4000) together with an especially designed humidification system (aerosol 315 conditioning system (ACS1000) by Ecotech Pty Ltd.). The original set-up was altered to contain solely one nephelometer 316 and RH in the humidifier was alternating between 30% (dry) and 80% (wet) every 20 minutes to measure once dry and once 317 humidified scattering coefficients. The scattering enhancement factor f(RH) can be calculated from the scattering coefficient σ s 318 at dry and elevated RH: f(RH)=σ s,wet /σ s,dry . Measurements using the nephelometer were carried out using the polydisperse size 319 distribution directly obtained from the smog chamber. The nephelometer was calibrated using particle-free air and CO 2 as a 320 span gas and the system was additionally calibrated with ammonium sulphate (Sigma Aldrich, purity >99.95%).  OH concentrations in the range of ∼10 6 molecules/cm 3 and hence values comparable to those found in the troposphere 1 .

345
Although the chamber was carefully cleaned after each experiment, a possible influence by Cl radicals generated from leftover 346 salt particles on the walls on the 1-butanol decrease rate cannot be fully excluded which makes the estimate an upper limit.

347
The UV intensity might have varied between experiments due to failing of some of the lamps in the AURA chamber and this 348 might influence the concentration of OH radicals available for reactions. It is, however, difficult to know the UV intensity in 349 each experiment, as they were carried out over several months (April until October 2018). To ensure that this did not have a 350 significant effect on the results, two ageing experiments were repeated with a complete set of new UV lamps (Exp. 12 and 19 in 351 Table 1. An experiment without lamps was also repeated, Exp. 21). The data presented in Fig. S2 and S3, suggest a negligible 352 influence as the results fall in the same range as those from previous experiments. This is also clearly visible in Fig. S8 where 353 the GF decrease rates for particles of D dry =50 nm (also presented in Table 1) are displayed and compare well to those recorded 354 in the other experiments.  Hygroscopicity of larger sized sea salt particles (Exp. 17). a) Scattering enhancement factor (f(RH)) measured at 3 different wavelengths (λ=450 nm,λ=525 nm,λ=635 nm) and RH in the nephelometer. b) number (red lines; left y-axis) and surface distributions (blue lines; right y-axis) at the beginning, after 5 hours and mean distributions throughout the experiment as derived from combined scanning mobility particle sizer and optical particle spectrometer measurements.  Effect of ageing on particle elemental composition. Microscopy images (STEM-EDX) of fresh (a) and aged (b) NaCl particles. Additionally, the elements found in the fresh and aged particles are illustrated. c) overlap of Na and Cl signals in the aged samples.

Figure 5
Effect of ageing on shape and elemental composition. Microscopy images (STEM-EDX) of fresh (a) and aged (b) sea salt particles, illustrating the shape and elemental compositions found in these particles.

Figure 6
Effect of heating the aerosol particles. Growth factors at RH=80% measured for NaCl and sea salt for the two different sizes of Ddry=50 nm (a) and 200 nm (b) are shown. The influence of heating the particles to 300C, additionally to drying them, is illustrated. The arrows indicate the time period when the heating was switched on.