Material for “ Establishing the Impact of Model Surfactants on Cloud Condensation Nuclei Activity of Sea Spray Aerosols ”

Sara Forestieri1,#, Sean M. Staudt2, Thomas M. Kuborn2, Katharine Faber3, Christopher R. Ruehl4, Timothy H. Bertram2, Christopher D. Cappa1, * 1Dept. of Civil and Environmental Engineering, University of California, Davis, CA, USA 2Dept. of Chemistry, University of Wisconsin, Madison, WI, USA 3Dept. of Chemistry, Carleton College, Northfield, MN, USA 4California Air Resources Board, Sacramento, CA, USA * Correspondence to: E-mail: cdcappa@ucdavis.edu

Abstract.Surface active compounds present in aerosols can increase their cloud condensation nuclei (CCN) activation efficiency by reducing the surface tension (σ) in the growing droplets.
However, the importance of this effect is poorly constrained by measurements.Here we present estimates of droplet surface tension near the point of activation derived from direct measurement 15 of droplet diameters using a continuous flow stream-wise thermal gradient chamber (CFSTGC).
The experiments used sea spray aerosol mimics composed of NaCl coated by varying amounts of (i) oleic acid, palmitic acid or myristic acid, (ii) mixtures of palmitic acid and oleic acid, and (iii) oxidized oleic acid.Significant reductions in σ relative to that for pure water were observed for these mimics at relative humidity (RH) near activation (~99.9%) when the coating was 20 sufficiently thick.The calculated surface pressure (π = σH2O -σobserved) values for a given organic compound or mixture collapse onto one curve when plotted as a function of molecular area for different NaCl seed sizes and measured RH.The observed critical molecular area (A0) for oleic acid determined from droplet growth was similar to that from bulk experiments conducted in a Langmuir trough.However, the observations presented here suggest that oleic acid in 25 microscopic droplets may exhibit larger π values during monolayer compression.For myristic acid, the observed A0 compared well to bulk experiments on a fresh subphase, for which dissolution has an important impact.A significant kinetic limitation to water uptake was observed for NaCl particles coated with pure palmitic acid, likely as a result of palmitic acid

Introduction
Surface active organic matter present in atmospheric aerosols has received considerable attention given its important role in heterogeneous chemistry (Knopf et al., 2005;Shaloski et al., 20 2017), aerosol water uptake and evaporation (Davies et al., 2013), and potential impact on the ability of particles to activate into cloud droplets (Ruehl et al., 2016;Ovadnevaite et al., 2017).Amphiphilic, surface active species have the potential to lower surface tension of a growing droplet, relative to pure water, due to their presence at the air-water interface.Through the Kelvin effect, this reduction in surface tension can, in theory, reduce the critical supersaturation, 25 thereby increasing particle cloud condensation nuclei activation efficiency (Farmer et al., 2015).
One challenge in assessing the potential influence of surface tension depression on CCN activity has been the lack of unambiguous evidence for the impact of surface tension on droplet activation.Recent work, using a custom-built continuous flow thermal gradient chamber (CFSTGC) to measure the sizes of droplets around 100% RH, has allowed for more direct 15 estimates of the surface tension depression in microscopic droplets at humidities relevant to cloud droplet formation.The size of droplets around 100% RH is sensitive to deviations in surface tension from that of pure water, much more so than at lower relative humidities.Thus, droplet size measurements can be used to infer surface tension values.Using this method, it was determined that the surface tension of droplets was substantially reduced below that of pure 20 water at RH ~ 100% for binary mixtures of individual surface active compounds and ammonium sulfate and for NaCl coated with secondary organic aerosol when the coatings were sufficiently thick (Ruehl and Wilson, 2014;Ruehl et al., 2012).The reduction in surface tension at a given RH caused the droplets to grow larger than predicted if the surface tension was assumed to be that of pure water (Ruehl et al., 2016).25 One particle type where surface tension impacts on cloud droplet activation might be of particular importance is with sea spray aerosol (SSA) particles.There is indirect evidence that surface tension reductions affect the efficiency of SSA particle activation.For example, Collins et al. (2016) observed high hygroscopicity parameters (κ > 0.7) for small SSA particles (<150 nm) generated during a suite of microcosm phytoplankton bloom, up to very high chlorophyll-a 30 concentrations in the source water (a marker for biological activity), yet no correlation between κ and biological activity was observed.Although Collins et al. (2016) did not measure the particle composition it seems likely that the organic fraction of the particles was large due to the high biological activity (O'Dowd et al., 2004b).Additionally, very large organic fractions were separately observed for small particles (< 200 nm) for a similar microcosm experiment (Deane et 5 al.,Submitted).The potential for substantial surface tension lowering was invoked to explain the high κ values observed.However, Fuentes et al. (2011) performed experiments using seawater enriched in marine exudates and found no evidence that surface tension reduction impacted CCN activation, observing instead a slight decrease in the CCN activation efficiency with increasing organic content, although the measured organic fractions of their particles were relatively low 10 (<40%) and therefore would not to enhance CCN activity if this organic matter formed a compressed film on the droplet surface (Ruehl et al., 2016).
To better understand the impact of surface active organic species on SSA particle activation efficiency, and on droplet activation in general, we report on measurements of droplet sizes at RH values just below activation for NaCl particles coated with varying amounts of 15 marine-relevant organic compounds, specifically long-chain fatty acids.These experiments were done as part of the MadFACTS campaign (Madison Fatty Acid Coating Thickness Study).Longchain fatty acids are an important class of organic compounds found in submicron SSA particles (Cochran et al., 2017;Cochran et al., 2016).Since this class of organic species is surface active (Schwier et al., 2012) they have the potential to enhance observed CCN activation efficiency by 20 depressing surface tension, but the overall effect of fatty acid addition on CCN activation efficiency of salt particles has been shown to be minimal (Nguyen et al., 2017).In this work we will clarify why such highly surface active species have little impact CCN activation efficiency.
Additionally, since particles in the ambient atmosphere are composed of complex mixtures of organic species, the role of mixtures was investigated by comparing single component and binary 25 surfactant systems and by comparing oxidized and unoxidized systems.These measurements were used to estimate the surface tension of the droplets as a function of the relative abundances of NaCl and the organic component(s) at a given RH, or as a function of RH for a fixed dry particle composition.We connect these near-activation surface tension measurements to the surface tension estimated at activation with a traditional CCN counter to assess whether the 30 observed reductions in surface tension significantly affect critical supersaturations for these chemical systems.

Methods
A suite of instrumentation was used to generate sea spray aerosol mimics and monitor their hygroscopic properties and chemical composition.A general schematic of the experimental 5 set-up is shown in Figure 1.

Particle Generation and Processing
Polydisperse NaCl particles were generated from a constant output atomizer (TSI Inc.) containing a 0.05 M solution of NaCl (99% purity).The air stream was dried with a diffusion denuder to < 20% RH.The particles were size selected according to their mobility diameters 10 ranging from 150 nm to 200 nm.The sheath to aerosol flow ratio was 10:1 to provide a relatively narrow distribution.The monodisperse particles were passed through an oven containing an aluminum sample holder with surface-active organic compounds.The organic compounds used were: myristic acid (99% purity; Sigma Aldrich), palmitic acid (99% purity; Sigma Aldrich), and oleic acid (90% or 99% purity; Sigma Aldrich).Myristic acid and palmitic acid are saturated 15 fatty acids containing 14 and 16 carbons, respectively, and exist as waxy solids at room temperatures.Oleic acid is an unsaturated fatty acid that exists as a liquid at room temperature and contains 18 carbons.For mixed surfactant studies, a sample holder containing known amounts of each organic compound (molar ratio 1:1) was added to the oven.In the oven, the organic compounds evaporated into the vapor phase.Upon exiting the oven, the air stream 20 cooled and the organic compounds condensed onto the NaCl particles resulting in surfactantcoated NaCl particles.Oven temperatures ranged from 85°C for thin coatings to 130°C for thick coatings.A charcoal denuder was placed after the oven to remove residual vapor from the air stream.A second DMA (TSI Inc.; Model 3071) was used to select particles of a given size from the coated distribution and served to remove small nucleated particles composed purely of the 25 organic compound(s).Note that for the second DMA the neutralizer (Kr-85) was bypassed, ensuring that only particles containing NaCl were sampled.Flow was then split isokinetically through a flow splitter (TSI Inc.; model 370800) to the various instrumentations.cm below the top of the tube and exited through a side port 77 cm below the inlet.Ozone was injected through a moveable stainless steel injector that passed through the top of the flow tube.
The exit of the injector was positioned between the particle inlet and exit ports, hereafter referred to as the active region.The residence time was varied by moving the injector position and calculated based on the distance between the injector and the flow tube exit.Total flow through 5 the flow tube was 1.8 SLPM, with 0.5 and 1.3 SLPM for the O3 and particle streams, respectively.The injected O3 concentration was 1 ppm resulting in an estimated concentration of 278 ppb in the flow tube, after dilution.Residence times calculated based on the flow rate and flow tube volume varied from 0 to 57 s.However, the air was sampled from the side of the flow tube and so the effective residence time was longer and was never as short as 0 s.There was a 10 bypass channel for the flow tube to monitor particle size and hygroscopicity prior to the flow tube experiments.The estimated ozone exposure ranged from 1.4 x 10 14 to 3.9 x 10 14 molecules cm -3 s for non-zero residence times.These are likely lower limits to the true O3 exposure, given the non-ideal flow within the flow tube resulting from the side sampling.Also note that a stainless steel injector was used, and thus the O3 concentration in the flow tube was likely 15 somewhat lower than the estimated value.

Hygroscopicity Measurements
Wet droplet diameter distributions at RH values near 100% (both above and below) were measured by a CFSTGC, described in detail by Ruehl et al. (2010).Briefly, particles are humidified in a temperature-controlled tube lined with wetted filter paper (102 cm in length, with 20 an effective inner diameter of 2.2 cm).The temperature gradient across the length of the tube could either be positive or negative allowing for both sub-and super-saturated RH values to be achieved (Roberts and Nenes, 2005), with a working range from 99.8% to 100.06%.Before exiting the CFSTGC, wet droplet size distributions and velocity were measured with a phasedoppler interferometer (PDI; Artium Technologies, Inc.).The particular configuration of this 25 CFSTGC instrument allows for accurate and precise determination of droplet diameters at a known RH value prior to activation.The total flow through the chamber was 0.5 SLPM, with a sheath:sample flow ratio of 2.33.The total residence time of the chamber centerline was ~22 seconds.The mode diameter for each measurement was obtained by fitting the number-weighted wet droplet distribution to a Gaussian curve using a function in the data processing program Igor 30 (Wavemetrics, v6.37), with each scan consisting of ~1,000 droplet measurements.Average temperatures in the CFSTGC ranged from 20 °C to 25 °C.The absolute temperatures were varied from day to day due to fluctuations in the room temperature, and were selected to prevent condensation in the detection chamber.However, since the instrument RH depends primarily on ΔT and not absolute T, these variations do not affect instrument performance.
RH in the CFSTGC was calibrated by sampling size-selected salt particles composed of 5 either NaCl or ammonium sulfate.The κ-Köhler equation (Petters and Kreidenweis, 2007) was where Dwet is the measured wet diameter, Ddry is the dry diameter, κ is the known hygroscopicity 10 parameter, σ is surface tension, VH2O is the molar volume of water in the droplet, k is the Boltzmann constant, and T is temperature.The first term corresponds to the Raoult effect, which accounts for reductions in water activity to the dissolution of solutes.The effective solubility of a given species is parameterized by the κ parameter, which ranges here from ~0 (insoluble) to 1.4 (very soluble).The exponential, or Kelvin, term accounts for the enhanced vapor pressure over a 15 curved surface and is proportional to σ.For the calibrations, σ is assumed to be equal to that of water.This is reasonable given the dilute concentrations of salt (~0.05 M) in the aqueous droplets and the lack of surface-active species.The accuracy of RH calibrations depends on the assumed κvalues and shape-correction factors of the calibration salts.The RH values were calculated from Eqn. 1 assuming that κ values were 1.3 and 0.61 and shape-correction factors 20 were 1.08 and 1.04 for NaCl and ammonium sulfate, respectively.As such, the uncertainty for RH in the CFSTGC was characterized by calibrating the chamber with both size-selected NaCl and ammonium sulfate, switching between the two salts every 3-5 minutes.Results of this comparison are shown in Figure S1.The slope of the linear fit was 1.00, with the majority of the data falling within 0.025% of the 1:1 line.During experiments with NaCl particles coated with 25 organic compounds, the RH was calibrated multiple times a day with either never-coated particles or with NaCl particles where the organic coating was completely removed by passing the particles through a thermodenuder at 250 °C.(No difference was found between never-coated and thermally denuded NaCl particles.)By calibrating throughout the day noise due to RH drift was minimized.Most experiments were conducted holding both RH and the NaCl seed size constant while varying the amount of organic coating.For one experiment with oleic acid as a coating, the composition (or coating amount) of the particles was held constant, while RH was varied to characterize how droplet size varied leading up to the point of activation.In this manner, a Köhler curve is mapped out (Ruehl et al., 2016).5 For a subset of experiments, the number of particles that activate into cloud droplets at a given supersaturation was characterized with a CCN counter (CCNC; Droplet Measurement Technologies, Model CCN-100) (Roberts and Nenes, 2005).Total particle concentrations were measured concurrently with a condensation particle counter (CPC; TSI Inc. Model 3787).The combination of these measurements allowed for the calculation of the fraction of activated 10 particles (fact) as a function of supersaturation (s = RH/100 -1), with scanned supersaturation values ranging from 0.03% to 0.1%.The critical supersaturation (sc) was determined by fitting a sigmoidal function to fact versus s.The apparent hygroscopicity parameter (Petters and Kreidenweis, 2007) was then calculated using the following equation: where Sc = sc + 1, MW is the molecular weight of water, Ddry is the dry diameter, σ is surface tension, ρW is the density of water, R is the ideal gas constant and T is temperature.In using Eqn.
2, it is assumed here that σ is equal to that of water (72 mN/m).Since the true value of σ may deviate from 72 mN/m, we refer to the derived κ as the apparent κ (or κapp).The supersaturation as a function of temperature gradient in the instrument was calibrated with NaCl in the scanned 20 range (Figure S2).

Size and Composition
Diameters of the size-selected, dry, coated particles were measured with a scanning mobility particle sizer (SMPS; consisting of a DMA TSI Model 3081 and TSI CPC 3775).Mode diameters were obtained by fitting a lognormal function to the obtained size distributions.The 25 organic volume fraction was calculated as:

=
(3) where Dtot is the total coated diameter and DNaCl is the diameter of the size-selected NaCl particles, adjusted for shape effects.Precision-based uncertainty for the selected diameter was assessed by size-selecting particles with one DMA and measuring the resulting size distribution.
The mode diameter and the size-selected diameter agreed within 1%.For a subset of experiments, a scanning electrical mobility sizer (SEMS; BMI, Inc.) was used to measure size 5 distributions.
The organic composition was monitored with a chemical ionization mass spectrometer (CIMS) coupled with a thermal desorption chamber to vaporize the particle-phase organic coating (McNeill et al., 2008).The CIMS used Cl -, as opposed to I -, as the reagent ion so as to retain the ion-molecule adducts (Cl -•HA) within the mass range of the quadrupole mass analyzer.10 After each experiment, the CIMS sampled particle-free air by setting the second DMA voltage to 0, downstream of the oven, to characterize background gas-phase concentrations.Each compound had a unique spectrum with an identifiable ion corresponding to the unfragmented parent compound.The counts for each compound were used to calculate relative abundances for the binary organic mixtures under the assumption that the sensitivity of the CIMS was the same 15 for each carboxylic acid.When I -is used as the reagent ion, the CIMS method is somewhat more sensitive towards palmitic acid compared to oleic acid (Lee et al., 2014).If this difference in sensitivity similarly applies to the Cl -reagent ion then the relative abundance of palmitic acid will be underestimated by the assumption of equal sensitivities.

Upper-Limit Surface Pressure Calculation 20
The combination of Dwet from the CFSTGC, the mode coated diameter (Ddry), and the calibrated RH allows for the surface tension (σ) of the droplets to be calculated for every individual measurement using Eqn. 1.Since the particles were mixtures of NaCl and organic compounds, the κavg term was calculated, assuming volume mixing, as: The values of κorg and κNaCl were assumed to be 0 (Petters et al., 2016) and 1.3 (Petters and Kreidenweis, 2007), respectively.The use of Eqn. 1 provides a lower limit estimate of σ, since it is assumed that none of the organic component partitions into the bulk droplet and is present only at the surface.These values will therefore be referred to as the lower-limit σ.To facilitate the comparison between bulk surface tension studies and this work, the surface pressure (π) was calculated from the observed σ values as: where σH2O and σobs are the surface tension of water and the observed surface tension, respectively.The π values calculated from Eqn. 5 using the lower-limit σ will thus be referred to 5 as the upper-limit π estimate.A corresponding molecular area (A), which is a measure of the inverse of concentration of surfactant at the interface, is calculated as where MW is the molecular weight, ρ is the density of the organic compound, and NA is Avagadro's number.For experiments where NaCl was coated with two types of organic 10 compounds, a molar weighted average of the molecular weight and density for each component was used to calculate molecular area.

Surface and Bulk Partitioning
Given the large surface area to volume ratios in microscopic droplets, it is necessary to account for surface and bulk partitioning when deriving σ values from the observations.The 15 compressed film model was used for this purpose (Ruehl et al., 2016).The organic compounds at the air-water interface can contribute to σ depression, while organic compounds dissolved into the bulk contribute to droplet growth through the Roault effect.The film model is a 2dimensional (2D) equation of state (EoS) that parameterizes σ as a function of molecular area as: where A0 is the critical molecular area, A is the molecular area, mσ is a term that accounts for the interaction between surfactants at the interface, and σmin is an imposed lower-limit for σ.In this model, reductions in σ relative to water requires the formation of a full monolayer, which occurs at molecular areas smaller than A0.At molecular areas larger than A0, the droplet σ is assumed to be equal to 72 mN/m and the surface phase is said to be in a "gaseous" state where molecules 25 present at the surface do not interact.Strictly speaking, π is non-zero in this state, but we assume this increase in π is negligible and use σ = 72 mN/m.A 2D phase change occurs at A0 and further addition of surfactant molecules to the interface (A<A0) causes compression of the surface film.This leads to a sharp decrease in σ, or increase in π.The corresponding isotherm for this EoS is given as where C0 is the bulk concentration at the 2D phase transition, Cbulk is the bulk concentration, R is the gas constant, NA is Avogadro's number, and T is temperature.The variation of σ with A is 5 solved for from the observations by minimizing the chi-square value for Dwet by varying A0, mσ, and C0 with Dcoat, DNaCl, and RH as inputs.The surface tension is constrained by always be larger than some minimum value (σmin or πmax) that is determined as part of the data fitting.When the system reaches σmin the addition of more molecules or increased compression causes dissolution of surface molecules into the bulk and does not further depress surface tension.An outer iteration 10 uses the Köhler curve to solve for Dwet for a given RH.An inner iteration solves Eqn. 8 for the fraction of organic matter at the surface (fsurf).For each assumed fsurf, the inner iteration calculates a value for Cbulk as and A as 15 The film model as used here assumes that all organic molecules are either dissolved in the bulk or located at the droplet surface.The σ is calculated from the A value using Eqn. 7. Dwet is then calculated from the water activity (i.e. the Raoult term) using the number of moles of organic in the bulk droplet (determined from εorg and fsurf) and the number of moles of NaCl (determined 20 from DNaCl).The Kelvin term is calculated based on the derived σ value.The use of the film model results in a single curve that represents the best-fit π-A relationship, rather than the pointby-point determination that results from the upper-limit method.Uncertainty in the film model curve was estimated by perturbing the input RH values by the average precision-based uncertainty in RH.The data were then fit to the film model using the perturbed (+ and -) RH 25 values, and the uncertainties were calculated as the difference between the original and perturbed cases.Some of the differences between the film model results and the upper-limit results is due to the upper-limit coming from a point-by-point analysis while the film model is a fit across all data points.If the upper-limit values are on the whole equal to film model π values, this indicates that dissolution is minimal and that droplet growth enhancement is caused enhancements in π (or reductions in σ).However, if the upper-limit values are systematically lower than the film model 5 π values, then droplet growth enhancement is due to both enhancements in π and organic compounds dissolving into the bulk.
In addition to the compressed film model, the data were also fit to the Szyszkowski-Langmuir EoS: where C0 is the bulk concentration at which half of the surface sites are occupied and A0 is the maximum surface concentration.The corresponding isotherm is: (10).
In contrast to the film model, the Szyszkowski EoS does not include an interaction parameter.
The Szyszkowski EoS allows for the continuous adsorption and desorption of surfactants from 15 the interface and does not include a 2D phase transition.

Oleic acid coatings
Variation in the observed wet diameters for 200 nm size-selected NaCl particles coated 20 with oleic acid are shown as a function of the coated particle diameter for one RH (~99.94%) in Figure 2A.If it is assumed that π is constant and equal to that of water (i.e.σ = 72 mN m -1 ), the predicted Dwet from Eqn. 1 are much lower than the observed Dwet, with the difference between the two increasing as the coated diameter (and εorg) increases.This observation indicates that the π values in the droplets must be greater than 0 mN m -1 (σobs < σH2O) for these coated particles, or 25 equivalently that σ is lower than that of pure water.The difference between the observations and the assumed pure water curve increases with the dry coated particle diameter (and the εorg), but only once the coating reaches a critical value.This threshold behavior, or a minimum required εorg to observe an enhancement in the droplet diameter, was previously observed by Ruehl and Wilson (2014) for ammonium sulfate particles coated with oleic acid and other fatty acids.
The wet droplet diameters are used to calculate upper-limit π values from Eqn. 1 for all NaCl core sizes, RH, and εorg used (see Figure S3), and are considered as a function of molecular 5 area (Figure 2B).Experiments with NaCl coated with 99% purity oleic acid and 90% purity oleic acid were similar, indicating that this change in oleic acid purity did not impact observed π values (Figure S4).Notably, these upper-limit estimates of π, when plotted versus A, collapse onto a single curve independent of the selected NaCl seed diameter or measurement RH (99.85-100.03%).This occurs because when RH is increased (for a given seed diameter), the droplet is larger and the 10 molecular area increases, which then decreases the calculated π.Similarly, variations in the NaCl seed diameter (with RH held constant) lead to variations in εorg for a constant coated dry diameter, but this in turn leads to corresponding variations in surface concentrations, molecular areas, and ultimately calculated π values.Consequently, the data all collapse onto one curve.
The corresponding compressed film model fit is also shown, both as Dwet versus the dry 15 particle diameter (Figure 2A) and as π versus molecular area (Figure 2B).Below the threshold εorg or above the corresponding A0 the film model gives π = 0.In this low-coverage region the surfactant concentration is below monolayer coverage and the molecules are in a "gaseous" state.
Above the threshold εorg (or below A0), a full monolayer forms and π begins to increase.In this compression region, as more organic molecules are added (i.e. at larger εorg) compression of the 20 monolayer leads initially to a steep increase in π as the molecular area decreases.Eventually π reaches a maximum and remains constant at this value.This could be indicative of monolayer collapse, the formation of a 3-dimensional phase, and/or dissolution.This region of the π isotherm will be referred to as the collapsed region.In contrast, the Szyszkowski EoS does not permit a 2dimensional phase transition and thus cannot reproduce the observed droplet growth behavior 25 (Figure 2A).Instead, the Szyszkowski EoS yields wet diameters that increase continuously with the coated diameter and that do not exhibit the plateau at high coverage.This indicates that the Szyszkowski EoS is not appropriate for use with molecules such as oleic acid.
The best fit from the film model for oleic acid on NaCl gives A0 = 48.5 ± 3.8 Å 2 , mσ = 2.15 ± 0.1 mJ m -2 , and πmax = 37.5 ± 4.8 mN/m (see Table 1).In the compression region, the film model 30 π values are close to upper-limit π values.This indicates that the effect of bulk partitioning on the Raoult term is small (bulk concentrations < 5%vol) and nearly all of the enhanced droplet growth is due to enhancements in π.In the collapsed region, the film model π values are slightly less than the upper-limit estimates, which indicates that bulk partitioning of the organic compounds becomes important following monolayer collapse.5 The π-A isotherm determined here using the film model for microscopic droplets can be compared to similar measurements for bulk systems (see Figure 2).In the bulk measurements, variation in A is induced through physical compression of the surfactant in a Langmuir-Blodgett trough.The A0 values for the wet droplets compare well to those in bulk systems (Voss et al., 2007;Mao et al., 2013;Seoane et al., 2000).However, the film model derived π values seem to be 10 higher than the bulk system, both in the compression and collapsed region.We hypothesize that this might reflect real physical differences between the droplet and the bulk experiments.For one, in typical bulk experiments, the molecular areas are changed slowly (0.2 to 5 Å 2 min -1 ) because they are aimed at measuring under equilibrium conditions.In contrast, in the droplet experiments water uptake and growth occurs rapidly, in just a few seconds.In addition, the droplet growth 15 experiments proceed from a state where the hygroscopic salt core is covered with a purely organic, initially very thick coating (at the highest εorg).This thick coating is stretched out as water uptake by the inorganic core causes the overall particle to increase in size.This is opposite that in typical Langmuir-Blodgett experiments, which start at an expanded state and compress over time.Some bulk studies indicate that higher compression rates lead to higher surface pressures during 20 monolayer compression and higher collapsed π (Rabinovitch et al., 1960;Jeffers and Daen, 1965;Wüstneck et al., 2005), attributed to time-dependent structural changes in the monolayer.Additionally, bulk studies measuring time-dependent changes in π following rapid expansion indicate that relaxation to the equilibrium state can occur with time constants on the order of 10 seconds (Murray and Nelson, 1996) to several minutes (Smith and Berg, 1980), and dynamic 25 surface pressure measurements of atmospheric aerosol extracts indicate that full relaxation can take hundreds of seconds (Noziere et al., 2014).

Myristic acid coatings
Upper-limit estimates of π and the film model fit to myristic acid coated NaCl particles are shown in Figure 3. Like oleic acid, no enhancements in π are observed above a critical 30 threshold at A0, i.e. π = 0 when A > A0.For myristic acid, A0 = 29.2 ± 1.2 Å 2 and is smaller than that for oleic acid.This indicates that myristic acid packs more efficiently than oleic acid and that more surfactant molecules are needed at the surface to enhance π.We suspect this difference arises because oleic acid has a cis double bond, which adopts a bent configuration.This results in less efficient packing at the droplet surface compared to myristic acid, which is a straight chain 5 alkanoic acid (Kanicky and Shah, 2002).The upper-limit π estimates are comparable to the film model.This indicates that bulk solubility played a minimal role.Takahama and Russell (2011) predict from molecular dynamics studies that the mass accommodation coefficient of water on myristic acid coated particles can be suppressed, with the reported range being 0.0-0.04.Here, there is no evidence that such suppression impacted droplet 10 growth, in contrast to experiments with palmitic acid, discussed below.These findings are consistent with Ruehl and Wilson (2014) for ammonium sulfate particles coated with myristic acid.It may be that myristic acid coatings do reduce the accommodation coefficient from unity, but that the ultimate extent of reduction is insufficient to have a substantial impact on the droplet growth on the timescale of these experiments.In general, the experiments here are reasonably 15 insensitive to variations in the accommodation coefficient when greater than 0.01 (Ruehl and Wilson, 2014).
The myristic acid isotherm for droplets determined here is compared to bulk measurements (see Figure 4).Two different isotherms are shown, both taken from Albrecht et al. (1999).One is for myristic acid compressed on a fresh subphase (water), which corresponds to 20 the first in a series of monolayer compression-expansion cycles.The other is for a used subphase, which corresponds to the 8 th compression, and where additional surfactant is added prior to each compression.The two isotherms differ substantially.The difference between sequential compression-expansion cycles becomes smaller as the number of cycles increases, and eventually there is little difference between two sequential cycles.Albrecht et al. (1999) 25 concluded that that myristic acid (as well as other short chain fatty acids) slowly dissolves into the subphase, with the effects of this process declining over time as the subphase becomes saturated.Thus, whether a fresh or a reused subphase has a significant impact on the resulting isotherm for bulk systems.This makes quantitative comparison between the myristic acid isotherm from this study and previous bulk studies difficult.The A0 for the isotherm observed 30 here on the growing droplets is most similar to the fresh subphase (A0 ~ 25 mN m -1 ) and is substantially smaller than the used subphase (A0 ~ 50 mN m -1 ).Dissolution decreases myristic acid molecules from the interface, which means that lower apparent molecular areas (assuming no dissolution) are required to change π.However, the dissolution of myristic acid into the bulk is slow, occurring over timescales of 10's of minutes (Albrecht et al., 1999).These time scales 5 are much longer than the time scale for droplet growth (seconds).Based on these time scales, dissolution should not play a large role the experiments presented here, consistent with the similarity between the upper-limit π and compressed film model.However, observations from Smith and Berg (1980) indicate that larger π can lead to greater rates of dissolution (10% decrease in molecular area over 3 minutes at π ~ 22 mN/m).In droplet experiments, the interface 10 is expanding from a highly compressed state (high π).Therefore, it is possible that dissolution was significant for these experiments.However, this conflicts with film model measurements matching the upper-limit because it is assumed that solubility is negligible (i.e.κorg = 0) for upper-limit calculations.Nevertheless, the film model is able to reproduce general π-A behavior of this system, even if bulk phase concentration is significant.15

Palmitic acid coatings
Figure 4A shows droplet diameters as a function of coating thickness for NaCl particles coated with pure palmitic acid.In the range of temperatures used in these experiments (< 25 °C), palmitic acid exists as a waxy solid (Inoue et al., 2004), though it is possible that supercooling can occur (Hearn and Smith, 2005).When palmitic acid is coated on the NaCl particles with εorg 20 ≥ 0.8, the observed droplet sizes were much lower than what is observed for just the uncoated NaCl core.As coating thickness (and εorg) increases, the extent of suppression in droplet growth decreases.This indicates that the presence of pure palmitic acid as a coating, especially at thick coatings, inhibits droplet growth.Thus, π values could not be determined for these experiments.
This behavior is consistent with the low mass accommodation coefficients (α) calculated from 25 Ruehl and Wilson (2014) for ammonium sulfate particles coated with palmitic acid.The dependence of droplet growth suppression on coating thickness is likely due to the slow initial diffusion of water through the palmitic acid coating.Thicker coatings apparently lead to greater inhibition of water transport and slower water diffusion to the NaCl core, corresponding to an initial period of low α.However, as some water infiltrates to the salt core and growth due to 30 water uptake occurs, palmitic acid may form islands (Davies et al., 2013;Lovrić et al., 2016).
This may lead to the formation of holes that would enhance water transfer, leading to an increase in the effective α.over time, which allows the droplets to grow to the observed sizes.

Binary surfactant systems
To understand how mixing of different organic compounds might impact the ability of 5 compounds such as palmitic acid to inhibit water uptake, experiments were carried out for binary oleic acid-palmitic acid coating.For these experiments, a 1:1 molar mixture of these surfactants was added to the oven.The CIMS composition measurements indicated that the average oleic acid fraction on the particles was 0.33 ± 0.08 (Figure S5), slightly lower than that of the mixture in the oven.This difference likely results from differences in the vapor pressures of these 10 compounds (Cappa et al., 2008), although could result from differences sensitivity of the CIMS to these compounds (Lee et al., 2014).For the mixed palmitic acid and oleic acid coatings on NaCl seeds, the observed droplet sizes were equal to or larger than that of the pure NaCl component, with the droplet size increasing as the total coating thickness increased.This indicates that, unlike for pure palmitic acid, the mixed coating does not inhibit water uptake.15 Thus, there is no evidence of kinetic limitations for this system.Presumably, the mixing of liquid oleic acid (Inoue et al., 2004) with solid/waxy palmitic acid prevented the palmitic acid from packing tightly enough to inhibit initial water uptake.
Since no kinetic limitations were observed for NaCl particles coated with palmitic acid-oleic acid mixture, it is possible to determine π values for these experiments.The π-A isotherms from 20 the upper-limit and film models for the oleic acid-palmitic acid mixture (on NaCl seeds) are shown in Figure 4B.These are compared with both the film model curve for oleic acid determined above and observations from bulk palmitic acid measurements (Tang et al., 2010).
The film model fit yielded πmax = 43.6 ± 6.3 mN/m and A0 = 35.2± 3.3 Å 2 (Table 1).The A0 is smaller than that for pure oleic acid coated particles by 10 Å 2 , indicating that oleic acid packs 25 more efficiently when mixed with palmitic acid.The observed film model isotherm for this system can be compared to the isotherm of an ideal mixture (Adamson and Gast, 1967).The ideal isotherm was calculated from a molar-weighted average of the pure oleic acid isotherm (from the film model fit) and the bulk palmitic acid isotherm (Tang et al., 2010).The observed π at a given molecular area was higher than the ideal prediction, closer to oleic acid than to 30 palmitic acid.This is consistent with behavior observed in previous work for oleic acid-stearic acid mixtures in bulk solutions (Feher et al., 1977) and indicates some difficulty incorporating the bent double bond in oleic acid into the monolayer.(Stearic acid is a saturated fatty acid that is two carbons longer than palmitic acid and has been observed to suppress water uptake, similarly to palmitic acid (Ruehl and Wilson, 2014).)Other work has found that saturated and 5 unsaturated fatty acids do not mix ideally, but mixed unsaturated fatty acid-saturated fatty acid films instead form distinct domains enriched in either component (Ocko et al., 2002).
The isotherm for palmitic acid from the bulk measurements exhibits more complex behavior than is observed for either oleic acid (either in the bulk or in droplets here) or for the mixture here.For palmitic acid in bulk experiments, the π increases as A decreases through the 10 compression region until it reaches a sharp maximum.At even greater compression (smaller A) the π then decreases until reaching a plateau.The behavior of palmitic acid stems from the formation of complex 3D structures following monolayer collapse.If the palmitic acid and oleic acid in the mixture were to form distinct domains there should be two distinct collapse pressures (Ocko et al., 2002;Griffith et al., 2012).Here, the derived upper-limit π values for the mixture do 15 not appear to decrease at small A (after the monolayer collapse), instead continuing to rise steadily with decreasing A. However, some of this continued increase in π likely results from attribution of dissolution of palmitic acid and oleic acid to surface tension in the upper-limit model.In comparison, the π values from the film model plateau at small A for the mixture, rather than peak, decline, and then plateau as with pure palmitic acid, although it should be noted that 20 the film model is not designed to capture such complex behavior.
Previous modelling (Takahama and Russell, 2011) and experimental (Ruehl and Wilson, 2014;Davies et al., 2013) studies, along with our measurements above, indicate that there can be a kinetic limitation to water evaporation and uptake imposed by films composed of singlecomponent surface active organic species, in particular long-chain organic species that form 25 solid films, which pack densely and exhibit long-range order.However, CCN measurements of ambient particles (Raatikainen et al., 2013) suggest that water uptake and droplet growth are not kinetically limited (i.e.α > 0.1) for particles sampled at a variety of locations around the world.
Here, the observations demonstrate that mixing of one component that did not inhibit water uptake (oleic acid) with another that did (palmitic acid) completely removed the kinetic 30 inhibition.Most likely, the mixing of the two components prevented the formation of a tightly packed film and so water uptake was facile.These observations serve as a potential explanation for why low α (< 0.1) water values are not observed for particles in the ambient atmosphere (Raatikainen et al., 2013), since ambient particles are multi-component mixtures.They also support the suggestion by Davies et al. (2013) that kinetic inhibition to water uptake should 5 rarely be observed in ambient particles, although it is possible that kinetic limitations could become more pronounced at lower temperatures, as decreasing temperature leads to increasing packing density (lower molecular area) (Davies et al., 2013).
Measurements were also made for a mixture of myristic acid and oleic acid coated on NaCl particles.For this mixture, the coating composition on the particles was dominated by 10 myristic acid (~90% by mole), even though the mixture composition in the coating apparatus was 1:1 by mole.The upper-limit π estimates as a function of molecular area are shown in Figure S6.At molecular areas > 10 Å 2 , the π estimates for the mixture approximately match the pure myristic acid case, but for molecular areas <10 Å 2 , the surface pressures for the mixed case were slightly larger.The general consistency between the mixed and pure case is consistent with the 15 high myristic acid fraction.

Köhler curves and the surface tension at activation
Above, we showed that the addition of sufficient amounts of fatty acid surfactants to salt particles (here, NaCl) can produce substantial σ depression (or π enhancement) in droplets near the point of activation, that is just above or below 100% RH.These observations are consistent 20 with other similar observations for both fatty acids and other compounds (Ruehl et al., 2012;Ruehl et al., 2016;Ruehl and Wilson, 2014).However, an important question is the extent to which this σ depression ultimately impacts activation into cloud droplets.Here, this is assessed by measuring CCN activation curves for NaCl particles coated with: oleic acid; oleic acid mixed with either palmitic acid or myristic acid; and for oleic acid particles oxidized by O3. 25 These activation curves were used to calculate the apparent κ values from the measured critical supersaturation (sc).The observed κapp values are compared with those predicted assuming volume mixing (κmix; Eqn. 4).Any enhancement in κapp values relative to the volume mixing line is attributed to σ depression.The observed κapp values are generally consistent with volume mixing rules, although there are some small differences at higher εorg (Figure 5A).Our 30 observations are consistent with Nguyen et al. (2017), who found good agreement between observed and calculated κapp values for sea salt particles coated with various long-chain unsaturated fatty acids, including oleic acid.Values of σ at activation reported here were estimated based on the small differences between the κapp and the κmix by calculating the σ required to achieve perfect closure between the observed and calculated sc when it is assumed 5 that κ = κmix.For εorg < 0.90, the calculated values were very close to 72 mN/m.This indicates that the surfactants have little influence on the CCN activation (i.e. on the sc) despite there being substantial depression of σ at RH values slightly lower than the critical value for similar εorg.
Above εorg > 0.90, the average calculated σ was 66.4 ± 2.9 mN/m, slightly less than that for pure water.This indicates that when these surfactants are sufficiently abundant there is a small, but 10 non-negligible influence of σ depression on activation.
To understand in greater detail the influence of these types of surfactants on CCN activation and to examine the robustness of the conclusion that the surfactants have some influence on activation at sufficiently high εorg, experiments were performed in which the composition of the particle was held constant (εorg ~ 0.95, using oleic acid) while RH was varied 15 from just below 100% RH to the point of activation.Figure 5B shows the directly observed Köhler curve for these particles, that is the variation in Dwet with RH.The individual points are colored according to the derived σ calculated from the upper-limit estimates (Eqn.1).As RH is increased and the particles approach activation, the droplets grow.The growth leads to an increase in both the droplet surface area and the corresponding molecular area, since the number 20 of organic molecules is fixed.This leads to a fairly continuous increase in σ over the RH range considered.Upon sufficient growth the molecular area increases beyond A0 and the reduction in σ becomes zero.This underscores the importance of considering surface-area normalized concentrations as opposed to bulk concentrations.Importantly, the Dwet at which the observed Köhler curve intersects the constant-κ Köhler 25 curve (calculated assuming σ remains constant at 72 mN m -1 ) is past the constant-κ maximum (Figure 5B).Consequently, the observed sc (= 0.057%) is slightly lower than that predicted value assuming constant σ (sc = 0.062%), consistent with the comparison between κapp and κmix above.This indicates that σ depression has some impact on activation, but that the impact on sc is relatively small (8%) for the model surfactants used in these experiments.Considered along with 30 the CCN activation measurements (Figure 5A), it is also apparent that the surfactants impact the critical supersaturation only at particularly high εorg.These modest reductions in σ are broadly consistent with results found in Schwier et al. (2011) for NaCl mixed with acidified sodium oleate (where almost all sodium oleate is present as oleic acid) and with Nguyen et al. (2017).
Although the ultimate effects on sc are small for these surfactants, depressions in σ change the 5 trajectory of activation through enhancement of droplet sizes, especially at lower RH (~99.9%), which is consistent with Ruehl et al. (2016).To have a significant impact on activation, organic compounds must substantially reduce the (maximum) critical supersaturation relative to the pure, uncoated salt.

Oxidation experiments 10
To understand how oxidation affects π (or σ), experiments were conducted in which NaCl particles coated with oleic acid were oxidized with O3 in a flow tube.Particle size and coating thickness were kept constant, while the residence time in the flow tube was varied to change the extent of oxidation.The composition of the oxidized particles was characterized by a CIMS (Figure 6A).The average spectra for the most oxidized case and for the pure non-oxidized oleic 15 acid are shown in Figure S7.The major oxidation products measured in this study include nonanoic acid, azelaic acid, and 9-oxononanoic acid.As the O3 residence time increased, the absolute abundance and fractional contribution of oleic acid decreased.The most abundant product was 9-oxononanoid acid (~86%), with minor contributions from azelaic acid (~7%) and nonanoic acid (~7%).The fractional contribution of products is generally consistent with 20 previous work (Katrib et al., 2004;Hearn and Smith, 2004).Note that some oxidation occurred for the zero second case because of the flow tube geometry described above.The diameters of the coated particles decreased relative to the bypass channel upon reaction with O3, likely due to the formation and subsequent evaporation of nonanal.The volume of the particles was reduced by 8% for lightly oxidized particles and 18% for highly oxidized particles.The εorg for the 25 oxidized particles ranged from 0.86 to 0.88, after reaction.
The CCN-derived κapp values were used to assess changes in solubility of the coating material, assuming that the σ of oleic acid coated NaCl is close to that of water at activation, a reasonable assumption given the results shown above.A difference between the observed κapp values and the calculated κmix could indicate an increase in the solubility of the organic 30 by the low κ values (~<0.003) for 9-oxononoic acid and nonanoic acid predicted from a functional group model (Petters et al., 2016).Azelaic acid has a higher observed κapp value of 0.02 (Kuwata et al., 2013), but since this product is only a minor component of the coating material, it does not have a strong impact on the overall κ.
Given that the observations indicate no substantial change in solubility, upper-limit π 10 values for oxidized particles have been determined, as measured before activation at RH ~100%.
The π values for oxidized particles at a given A are compared to the π-A isotherm measured for unoxidized oleic acid particles (Figure 6B).Molecular areas were calculated using molarweighted fractions of oleic acid and the oxidation products.After oxidation, the π values of these particles are still well above that of pure water.There is some indication that the π for oxidized 15 particles are lower than that for unoxidized particles at the same molecular area, although the difference is generally small.These observations are reasonably consistent with previous findings in Schwier et al. (2011) for particles composed of acidified sodium oleate and NaCl reacted with O3 at somewhat higher exposures than here (exposure ~8.8 x 10 14 to 4.4 x 10 15 molecules cm -3 s).However, unlike the observations presented here, the σ for their experiments 20 were estimated at activation based on the observed sc and so the overall influence of reductions in σ should be much smaller.

Sensitivity to film model parameters
To go beyond the specific chemical systems experimentally investigated here, the film 25 model can be used.The film model allows for theoretical exploration of the relationship between the properties of surface-active organic molecules (characterized by A0, mσ, C0 and σmin, and also the molecular volume or vorg) and sc and κapp.This serves to establish under what conditions the addition of organic material should lead to a reduction in the sc (and increase in κapp) and more efficient CCN activation, relative to that expected from volume mixing rules.As a starting point, theoretical Köhler curves have been calculated for three specific sets of film model parameters: (i) those determined here for oleic acid, and those determined by Ruehl et al. (2016) for (ii) glutaric and (iii) pimelic acid (Table 1).Glutaric acid and pimelic acid were selected because, unlike oleic acid, they were observed by Ruehl et al. (2016) to have a substantial impact on sc 5 and κapp through reductions in σ.Glutaric acid is a 5-carbon and pimelic acid a 7-carbon straight-chain dicarboxylic acid.The calculations use 80 nm NaCl seed particles coated to εorg = 0.80, which corresponds to a coated particle diameter of 136.8 nm (Figure 7A).The selected εorg value is similar to that observed for SSA with Dp < 200 nm generated from laboratory breaking waves in seawater during a microcosm (Deane et al., Submitted) to that observed for 10 ambient marine particles in high chlorophyll-a environments (O'Dowd et al., 2004a).The film model results are compared to the Köhler curve calculated assuming the organic component was insoluble and that σ = 72 mN m -1 , referred to as the constant σ or no σ reduction case.All κapp values are calculated from the model-predicted sc values, which account for σ reduction, assuming that there is no σ reduction.15 For the constant σ case the calculated sc = 0.142%, corresponding to κapp = 0.27, for these conditions.For oleic acid, the calculated critical supersaturation and κapp were unchanged from the constant σ case (Figure 7A).This is because the calculated σ in the film model reaches 72 mN m -1 at a Dwet that is substantially smaller than the critical diameter for the constant σ case (Figure 7B).In contrast, the sc for glutaric and pimelic acid coated particles are reduced slightly 20 and the κapp increased slightly, to sc = 0.137 (κapp = 0.29) and 0.133 (κapp = 0.31), respectively.This is because the reduction in σ persists to beyond the constant σ case activation point for these cases.The particles ultimately activate at the point where σ reaches 72 mN m -1 , the constant σ value, as was noted by Ruehl et al. (2016).The difference between oleic acid and the two dicarboxylic acids results in part from oleic acid having a larger vorg.Consequently, the 25 molecular area for oleic acid (at εorg = 0.8) substantially exceeds A0 prior to the constant σ activation point and there is no impact of the reduction in σ observed at slightly lower RH on the actual sc or κapp.This is not the case for the two diacids, and thus surface tension depression impacts sc.For comparison, at the activation diameter of the constant σ case (0.98 μm) the A for oleic acid is ~150 Å 2 but only 95 Å 2 for glutaric acid and 62 Å 2 for pimelic acid (see Figure 7C).30 Given the above, the general sensitivity of κapp to the surfactant properties (and their influence on σ depression) has been more systematically assessed, again using 80 nm NaCl seed particles.The predicted κapp values from the film model were found to be most sensitive to variations in the vorg and the A0 of the organic species (Figure 8).The predicted κapp values were not sensitive to variations in mσ values above 0.1 mJ m -2 and variations in C0 had limited impact 5 for A0 < 100 Å 2 .Additionally, the predicted κapp were not sensitive to σmin for values below 65 mN/m.Thus, we focus on vorg and A0.
The dependence of κapp on the organic species vorg and A0 is examined for two cases: εorg = 0.8 (Figure 8A) and εorg = 0.9 (Figure 8B).(For these calculations, mσ = 1.0 mJ m -2 , C0 = 1 x 10 -6 mol mol -1 and σmin= 40 mN/m.)For a given εorg, species with smaller νorg have smaller molecular 10 areas (higher surface area-normalized concentrations) at a given Dwet.Thus, the extent of σ depression is greater for smaller νorg.This is consistent with the observations in Ruehl and Wilson (2014).We find that the sensitivity of κapp to a change in νorg increases as νorg becomes small.Likewise, species with larger A0 will require less surfactant at the surface to reduce σ and impact activation.The calculated κapp are approximately constant over a wide range of A0 and 15 νorg, in particular when A0 is small and νorg is large, with κapp ~ 0.3 when εorg = 0.8 and κapp ~0.17 when εorg = 0.9 (Figure 8A).These values are very similar to those predicted from volume mixing rules assuming σ = 72 mN/m.However, when A0 is instead relatively large and/or νorg is relatively small the κapp can be very large, with substantial changes in κapp predicted for relatively modest changes in νorg.The exact transition from the nearly constant κapp to the A0 and ϖorg-20 sensitive kapp depends on εorg.In general, for larger εorg the kapp becomes sensitive to variations in νorg and A0 at larger absolute νorg.These calculations demonstrate that it is, at least in theory, possible to observe large κapp values for mixed inorganic-organic particles even when the organic fraction is large.However, substantial increases in κapp are only obtained for a particular range of organic properties, within the film model framework.25

Linking to sea spray and secondary marine aerosol
The chemical systems considered in this study-long-chain fatty acids coated on NaClwere considered in part because even-numbered fatty acids have been identified as substantial components in submicron sea spray aerosol particles (Cochran et al., 2016;Schmitt-Kopplin et al., 2012;Mochida et al., 2007).Our measurements demonstrate that the fatty acids can have a substantial impact on σ near activation, when present at sufficient abundance, but that reductions in σ ultimately have limited impact on activation (i.e. on sc).A lowering of sc due to reductions in σ has been used (or hypothesized) to explain the often large CCN activation efficiency for particles observed in the marine environment, in particular nascent SSA particles.For example, 5 Ovadnevaite et al. (2011) observed high CCN activity concurrent with low water uptake at subsaturated RH for ambient SSA particles sampled during a field study.In general, at RH values well below 100% the Raoult, or solubility, effect primarily controls water uptake, with limited influence of σ (and thus limited sensitivity to variations in σ).Thus, the low water uptake under sub-saturated conditions indicates the particles have limited salt content, and the CCN/sub-10 saturated difference implicates σ as an important factor.As another example, Collins et al.This was even though the SSA with Dp < 200 nm particles were likely highly enriched in organic matter during these mesocosms (Deane et al., Submitted).This again implies an important role 15 for σ in affecting the CCN activity.If σ depression is responsible for the observed high CCN activity of SSA then the average properties of the marine surfactants must differ substantially from the fatty acids systems tested here, given the distinct lack of impact of the fatty acids on sc observed here and by (Nguyen et al., 2017).Our theoretical analysis above using the film model suggests that the complex mixture of marine organic compounds (which includes fatty acids) 20 with salts must interact to have an effective A0 > 100 Å 2 and overall relatively small νorg values.
In Section 4.1, the dependence of κapp on A0 and νorg was examined for particles with constant εorg.Here, drawing on the Collins et al. ( 2016) observations, calculations of κapp have been performed as a function of εorg for select pairs of A0 and vorg.This provides insight into how variation in the relative abundances of salts and organics in (theoretical) SSA particles could 25 impact κapp.Three pairs of A0 and vorg values were chosen to produce κapp values at εorg = 0.8 of κapp = 0.35, 0.50, and 0.70 for an 80 nm NaCl seed particle, which can be compared to κapp = 0.27 if σ = 72 mN/m.There is more than one (A0, vorg) pair that yields the same κapp (Figure 8A).Thus, the specific pair chosen here were selected to test the sensitivity of κapp to each parameter, and the three pairs considered were (i) A0 = 150 Å 2 , νorg = 0.6 x 10 -4 m 3 mol -1 , (ii) A0 = 100 Å 2 , 30 νorg = 0.6 x 10 -4 m 3 mol -1 , and (iii) A0 = 100 Å 2 , νorg = 10 x 10 -5 m 3 mol -1 .For the parameter combinations tested, the κapp decreases monotonically with εorg when εorg < 0.5, following the volume mixing line assuming constant σ = 72 mN/m (Figure 9).The decrease in κapp with εorg in this range results from the increased fraction of insoluble organic material, but with no influence on the σ because the organic is not sufficiently abundant.However, at εorg > 0.5, the organic is 5 sufficiently abundant to reduce σ and this buffers the decline in κapp that results from addition of insoluble material.That is, at εorg > 0.5 the calculated κapp are relatively insensitive to variations in εorg and are larger than that obtained assuming σ = 72 mN/m.Consistent with the above analysis (Figure 8A), for a given A0 a decrease in νorg leads to an increase in κapp, here from 0.35 (νorg = 1 x 10 -4 m 3 mol -1 ) to 0.50 (νorg = 6 x 10 -5 m 3 mol -1 ) when A0 = 100 Å 2 and εorg = 0.8. 10 Correspondingly, at a given νorg an increase in A0 leads to an increase in κapp, here from 0.50 (A0 = 100 Å 2 ) to 0.67 (A0 = 150 Å 2 ) when νorg = 6 x 10 -5 m 3 mol -1 and εorg = 0.8.The exact dependence of κapp on εorg depends on the assumed film model parameters (Figure 9).In certain cases, the κapp can even increase with increasing εorg.Regardless of the exact behavior, this exercise demonstrates that σ depression by mostly insoluble species has the potential to buffer 15 observations of κapp against changes in composition.Relevant to nascent SSA particles, in particular, we find it is theoretically possible for this buffering effect to maintain κapp > 0.7 even at high εorg.It may be that such buffering effects explain the observations of Collins et al. (2016), who observed variability in the κapp values for nascent SSA particles over the range κapp = 0.7 -1.3 for a wide range of chlorophyll-a concentrations, but did not observe κapp < 0.7.20 Beyond SSA, recent observations of the CCN activity of very small, ambient secondary marine aerosol (SMA) particles have also been interpreted as indicating that σ depression has a substantial impact on sc (Ovadnevaite et al., 2017).Ovadnevaite et al. (2017) report the relationship between σ and Dwet obtained from thermodynamic calculations of the droplet phase behavior (in particular, liquid-liquid phase separation) and an assumption that σ can be 25 calculated as the surface-area-weighted mean of the composition-dependent σ for each of the two liquid phases.Here, we have translated the reported σ-Dwet relationship to π versus A (see Figure 7C).It was assumed here that the organic mass fraction = 0.55 (as reported) and that ρorg = 1.6 g cm -3 and MWorg = 332 g mol -1 ; these are the estimated density and molecular weight of the dimer species used as surrogate compounds used by Ovadnevaite et al. (2017), and correspond to a νorg 30 of 2.1 x 10 -4 m 3 mol.The inferred A0 is 1000 Å 2 .This is much larger than the compounds considered here, and might generally be considered as very large.Consistent with our above general analysis, it is evident that large A0, in addition to sufficiently low vorg, are necessary to significantly reduce critical supersaturations.The reason for the particularly large derived A0 (based on their reported σ-Dwet relationship) is not totally clear, but is likely related to their 5 assumption of the minimum coating amount (reported as a minimum thickness) necessary to impact σ.If the minimum thickness for a given compound is low, the buffering effect described above would be significant, even at lower εorg values, because a much lower concentration of surfactants at the surface (higher molecular area) would be required to reduce σ.

Conclusion 10
In this study, surface tension values were estimated for droplets at RH near activation that were grown from NaCl particles coated with the fatty acids oleic acid, myristic acid or palmitic acid, or a palmitic acid-oleic acid mixture.The retrieved σ explicitly on the relative amount of organic coating, the exact RH, and the identity of the fatty acid.For particles with εorg > 0.80, the observed σ were reduced significantly compared to pure water.Observed variability 15 in the relationships between σ, RH and εorg can be explained by the dependence of the molecular area of the organic molecules on these parameters.When s, or equivalently the surface pressure, is considered as a function of molecular area (i.e. in π-molecular area isotherms), organic molecule specific relationships are obtained that are independent of whether RH or εorg is responsible for the variation in the molecular area.The π-molecular area isotherms from the 20 droplets are used to determine critical molecular areas for each fatty acid.The A0 value for oleic acid on droplets compared well to Langmuir trough measurements on bulk solutions, but the magnitude of π values at high compression (i.e.small molecular areas) may be larger in the droplets.For myristic acid, the A0 value on droplets compared best with bulk experiments in which dissolution was significant, i.e. when a fresh subphase was used.25 When NaCl was coated with pure palmitic acid there was substantial suppression in water uptake observed.This was likely due to the formation of densely packed films through which water molecules could not efficiently permeate.The suppression disappeared when palmitic acid was mixed with oleic acid, indicating a decrease in packing density.For oleic acid coated NaCl particles exposed to O3 the σ values remain significantly lower than water at large εorg, but may be slightly higher than pure oleic acid coatings.Overall, chemical changes due to oxidation of oleic acid by O3 had minimal impact on the σ depression.
Values of the critical supersaturation were also quantified by a CCN counter for comparison to the observations made at RH values just below activation.Despite the large reductions in σ 5 observed at RH values just below sc, the measured sc indicate that the fatty acids have minimal impact on the ultimate activation into cloud droplets.This is because the additional growth as the RH increases to sc causes the molecular area to rapidly increase above A0, limiting the impact on activation.However, σ has a large effect on the trajectory of activation by enhancing droplet sizes when RH is < sc. 10 The film model of Ruehl et al. (2016) was used to theoretically explore what properties surface active organic compounds must have to have a substantial impact on CCN activation, and not just on the σ at RH values very close to (yet below) sc.We find that it is theoretically possible for surface active organics to have a substantial impact on CCN activation efficiency, even though this was not the case for the fatty acids here.In particular, the model sc, and 15 consequently κapp, are most sensitive to variations in the organic compound molecular volume (νorg) and A0, for NaCl-coated particles with relatively large organic fractions.Further, we show that surface tension depression from surface active organic molecules can serve to buffer κapp against changes in the organic-to-salt ratio when the εorg > 0.5.Overall, we conclude that surface active organic molecules can have a substantial impact on CCN activation efficiency.However, 20 the extent to which this will occur is strongly dependent upon the specific molecular properties of the organic molecules, and traditional surfactants (such as fatty acids) can actually have a negligible impact on CCN activation.

Acknowledgements
This study was funded by the Center for Aerosol Impacts on Climate and Environment (CAICE), 25 a NSF Center for Chemical Innovation (CHE-1305427).The authors also thank Hansol Lee and Alexei Tivanski (University of Iowa) and Michael Vermeuel (UW Madison) for their input and support for this project.

For
oxidation experiments, oleic acid coated NaCl particles were exposed to O3 in a vertical glass flow tube.Particles were introduced into the flow tube via a side port located 19 30 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-207Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2018 c Author(s) 2018.CC BY 4.0 License.
Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-207Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2018 c Author(s) 2018.CC BY 4.0 License.component after oxidation.Shown in Figure 5A are the measured κapp values for each oxidation condition at the measurement εorg and the theoretical line for the volume mixing assumption (assuming σ = 72 mN/m).The measured κapp values for the oxidized particles are very similar to both the volume mixing line and to the unoxidized oleic acid coated NaCl particles.This indicates that changes in solubility were minimal.This minimal change in solubility is supported 5 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2018-207Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 2 March 2018 c Author(s) 2018.CC BY 4.0 License.

(
2016) observed persistently high CCN activation efficiency (κapp > 0.7) for microcosm studies of nascent SSA with Dp < 200 nm regardless of the biological activity within the seawater used.

Figure 2 .
Figure 2. (A) Wet diameter (at RH = 99.93%) vs. dry diameter of 200 nm NaCl particles coated with varying amounts of oleic acid.The red line is the calculated droplet size, assuming σ = 72 mN/m independent of coating amount and the black and dashed black lines are film model and Szyszkowski model fits to the data, respectively.The error bars correspond to the confidence 5 interval for mode of the wet diameter distribution.The data points are colored by organic volume fraction (εorg).(B) Upper-limit estimates (grey points) and film model estimates (black lines) for surface pressure as a function of molecular area for various NaCl seed sizes (circles = 200 nm, squares = 180 nm, and triangles = 150 nm) and RH.Error bars on individual points and the film model fit (dashed black lines) are based on the precision in RH.Bulk measurements (red lines) 10 adapted from Voss et al. 2007, Mao et al. 2013, and Seoane et al. 2000 are included for reference.

Figure 3 .
Figure 3. Upper-limit and the film model π estimates for NaCl particles coated with myristic acid as a function of molecular area.Uncertainties in surface pressure are estimated from precision in RH.The black line is the film model estimate for pure oleic acid and the red line are bulk measurements of surface pressure adapted from Albrecht et al. (1999).5

Figure 4 .
Figure 4. (A) Measured wet diameter at RH = 99.88% as a function of organic volume fraction for 200 nm NaCl particles coated with pure palmitic acid (diamonds) and a mixture of palmitic acid and oleic acid (triangles).(B) Upper-limit and film model estimates for NaCl particles coated with a mixture of oleic acid and palmitic acid surface pressure as a function of molecular 5 area.The black line is the oleic acid film model estimate and the grey line corresponds to bulk surface pressure measurements from Tang et al. 2010, since surface pressure measurements for NaCl particles coated pure palmitic acid were not possible (see text for details).The purple line is the ideal mixing estimate for surface pressure based on the measured molar fractions of oleic acid and palmitic acid.10

Figure 5 .
Figure 5. (A) Apparent κ values as a function of εorg as calculated from the critical supersaturation for NaCl particles coated with oleic acid (circles), a mixture of myristic acid and oleic acid (squares), a mixture of palmitic acid and oleic acid (crosses) and oxidized oleic acid (triangles).Also shown is the predicted κ based on volume mixing rules assuming that κNaCl = 5 1.3 and κorg = 0.001, with σ = 72 mN/m.Each point is colored by the actual surface tension required to match observations in κ. (B) Relative humidity as a function of measured wet diameter for 180 nm NaCl particles coated with a fixed amount of oleic acid (εorg = 0.95).Points are colored by upper-limit estimates for surface tension and the size of the points corresponds to the wet diameter.Also shown are values predicted from the compressed film model and kappa-10 kohler theory assuming σ = 72 mN/m.

Figure 6 .
Figure 6.(A) Fractional contribution of oleic acid and oxidation products as a function of residence time.(B) Surface pressures as a function of molecular area for oxidized oleic acid coated NaCl particles colored by residence time (triangles).Upper-limit surface pressure estimates (black circles) and film model estimate (dashed black line) for unoxidized oleic acid 5 coated NaCl experiments are provided for comparison.

Figure 7 .
Figure 7. Film model derived (A) relative humidity and (B) surface tension (σ) as a function of wet diameter for uncoated 80 nm NaCl particles (dashed black line) or 80 nm NaCl particles coated with oleic acid (dark blue line), pimelic acid (light blue line), and glutaric acid (purple line) until εorg = 0.80.Film model parameters for oleic acid are from this study, while the 5 parameters for glutaric acid and pimelic acid were taken from Ruehl et al. (2016).(C) Surface pressure (π) as a function of molecular area for the coated-particle model systems in (A) and (B).Also shown is the π curve derived from in Ovadnevaite et al. (2017) based on the reported wet diameter and σ, assuming ρorg = 1.6 g cm -3 and a molecular weight = 332 g mol -1 .

Figure 8 .
Figure 8. Apparent κ values calculated as function of critical molecular area (Å 2 ) and molar volume (m 3 mol -1 ) from the film model, with the color corresponding to κapp.Results are shown for 80 nm NaCl particles coated to organic fractions (εorg) of either (A) 0.80 and (B) 0.90.Also shown are the observed film model fit parameters for oleic acid (circle) from this study.5 Parameters for glutaric acid (square), succinic acid (upside down triangle), glutaric acid (square), pimelic acid (triangle), and suberic acid (diamond) are from Ruehl et al. (2016).The color scales for panels A and B range from 0.25 to 0.8 and 0.15 to 0.8, respectively, with the color scale saturating at 0.8. 10

Table 1 .
Summary of experiments and corresponding compressed film model fit parameters.Uncertainties on the film model fit parameters on the precision in RH (see text for details).