The ability of airborne particles to take up water may be enhanced by surface-active components, but the importance of this effect is controversial because direct measurement of the surface tension of microscopic droplets has not been possible. Here we infer droplet surface tension from water uptake measurements of mixed organic-inorganic particles at relative humidities just below saturation (99.3–99.9%). The surface tension of droplets formed on particles composed of NaCl andα-pinene ozonolysis products was reduced by 50–75%, but only when enough organic material was present to form a film on the droplet surface at least 0.8 nm thick. This study suggests that if atmospheric particles are predominantly (≳80%) composed of surface-active material, their influence on cloud properties and thus climate could be enhanced, and their atmospheric lifetimes could be reduced.
 If two airborne particles with the same dry diameter (Ddry) are in equilibrium with ambient relative humidity (RH), the more hygroscopic of the two will have a greater wet diameter (Dwet). It will therefore scatter more light, act as a greater condensational sink for soluble compounds, and more readily act as a cloud condensation nucleus (CCN). Since wet deposition is the dominant process removing submicron particles from the atmosphere [Textor et al., 2006], and the chemistry of dilute cloud droplets is distinct from that of aerosol particles [Ervens et al., 2011], an aerosol's hygroscopicity largely determines its chemical evolution, its lifetime and atmospheric burden, and hence its influence on air quality and climate.
 Particle hygroscopicity under supersaturated (“CCN”) conditions is often greater than predicted from subsaturated water uptake measurements (typically at RH ≲ 90%). Several explanations for this behavior have been discussed in the literature, including reductions in droplet surface tension (σ [e.g., Dinar et al., 2007; Asa-Awuku et al., 2008; Ovadnevaite et al., 2011]). Currently no known method can directly measure σ of microscopic (Dwet ∼ 1 μm) droplets. Thus while many studies have shown that atmospheric organic matter can reduce σ of macroscopic aqueous solutions [Facchini et al., 1999; Dinar et al., 2006; Asa-Awuku et al., 2008], conclusive evidence of reduced σ in microscopic droplets has been elusive [Abbatt et al., 2005]. The σvalues of microscopic droplets and macroscopic solutions may differ because surface area to volume ratios are several orders of magnitude greater in the former. This tends to reduce surface excess concentrations of surface-active compounds, potentially inhibitingσ reduction relative to macroscopic solutions. Furthermore, for soluble surfactants, surface partitioning reduces bulk solute concentration, raising water activity and thus decreasing hygroscopicity [Sorjamaa et al., 2004]. It may therefore be inappropriate to apply macroscopic σ to microscopic droplets.
 Although σ in microscopic droplets cannot be directly measured, it can be inferred from measurements of droplet hygroscopicity, which is determined by two components: (1) the Kelvin effect, which is the increase in equilibrium water vapor pressure over a curved surface, and is proportional to σ; and (2) the Raoult effect, which is the reduction in water activity associated with solute dissolution, and does not explicitly depend on σ. Such an inference requires that these two effects can be separated, and that the measurements are sensitive to σ. Previously, σ has been inferred from measurements of CCN activity. However, for an aerosol of a given composition, such a measurement involves only a single independent variable – either the critical supersaturation or Ddry at the point of cloud droplet activation. Thus any enhanced CCN activity cannot be unambiguously attributed to reduced σ, as it could also be caused by an increased Raoult effect. Subsaturated hygroscopicity experiments allow for two independent variables (RH and or Ddry), but the Kelvin effect is negligible at RH < 95%, and thus hygroscopicity is dominated by the Raoult effect. Since RH > 95% is difficult to maintain accurately in most experimental setups, most previous observations of subsaturated hygroscopic growth have not been sensitive to σ.
 Here we report measurements of equilibrium water uptake at high but subsaturated RH (99.2 to 99.9%), to unambiguously determine if σ can be reduced in microscopic droplets. This is possible because measurements over this RH range allow for clear separation of the Kelvin and Raoult effects. These measurements are relevant to CCN activity because the water activity of the droplets is similar to that of typical atmospheric CCN at the point of activation. Our experiments utilize particles generated via dark ozonolysis of α-pinene because they mimic the complexity of ambient organic aerosols, and becauseα-pinene is an important precursor of ambient secondary organic aerosol (SOA). We address the following questions:
 1. Do we see evidence of σ reduction in microscopic droplets at RH near 100%? If so, to what extent?
 2. How much surface-active material is required to achieve any such reductions inσ?
 3. Under what conditions might reduced σ increase the CCN activity of aerosols?
 Reagent grade (>99% purity, VWR International) NaCl was dissolved in ultrapure water (resistivity ≥ 18.2 MΩ cm) and used in a constant-output atomizer (TSI, model 3079) followed by a diffusion drier (output RH < 10%) to generate seed particles. SOA particles were generated by dark ozonolysis ofα-pinene, either via homogeneous nucleation or condensation onto NaCl seed particles. No radical scavenger was used. Ozone was produced with a Hg penray lamp and diluted with a dry nitrogen flow to a concentration of either 350 ppb for seed experiments, or 560 ppb for homogeneous nucleation. The larger [O3] in the homogeneous nucleation experiments was necessary to ensure that droplets were large enough to measure. Liquid α-pinene was delivered via a syringe pump into a dry nitrogen flow, at a rate ensuring that O3 was the limiting reagent. The α-pinene and O3reacted to form SOA in a small stainless steel reaction chamber with a residence time of approximately 30 s. For the mixed SOA-NaCl particles, the SOA volume fractions were determined from the measured dry particle diameters of the pure NaCl and mixed particles. For inorganic particles, shape correction factors of 1.08 for NaCl and 1.04 for AS was applied. For pure organic or mixed organic-inorganic particles, no shape correction factor was used [Zelenyuk et al., 2006].
 High RHs just below (99.3–99.9%) and just above (100.2–100.6%) saturation were generated with a continuous-flow streamwise thermal gradient chamber based on the design ofRoberts and Nenes  and described in detail previously [Ruehl et al., 2010]. The flow rate in the chamber was fixed at 0.82 lpm, resulting in a residence time of ∼12 s. Before exiting the chamber, and while still flowing along the chamber centerline, droplets were counted and sized with a phase Doppler interferometer (PDI; Artium, Inc.).
where is the molar volume of the water in the droplet solution (assumed to be equal to that of pure water because droplets are very dilute when RH is near 100%), R is the gas constant, and T is temperature. The largest source of experimental uncertainty is the RH in the chamber, as RH is extremely sensitive to T near saturation. The chamber Tis controlled to within 0.01 K using high-precision thermistors, which corresponds to an uncertainty in RH of ±0.05% (absolute). All quoted uncertainty and error bars depicted in plots are those associated with the 0.01 K uncertainty inT, unless otherwise indicated. For CCN measurements, κ is determined via measurement of the critical supersaturation (Sc), from the following relationship:
where Ddry,c is the dry diameter that activates at a given Sc. Both RH and S(for CCN experiments) were calibrated with pure AS particles, taking into account non-ideality [Rose et al., 2008].
3.1. Observed κ Values
 Under subsaturated conditions, κ of pure secondary organic aerosol (κSOApure) generated by ozonolysis of α-pinene ranged from 0.011 to 0.042 (mean ± 1σ = 0.026 ± 0.010) (Table 1), increasing with RH (Figure S1a in the auxiliary material). This low value of κSOApure is consistent with Wex et al. , who found that for α-pinene ozonolysisκSOApure ∼ 0.02 at RH = 99.6%. Such a low κSOApure is also consistent measured Dwet/Ddry ratios less than 1.1 at 85% < RH < 90% [e.g., Prenni et al., 2007]. Due to the Kelvin effect, the water activity (aw) in these droplets is at least as high as the ambient RH, and so while non-idealities causeκSOApure to increase slightly with aw (and thus RH), κSOApure ≲ 0.04 even when aw ≳ 0.999 (Figure S1b). After the RH dependence of κSOApure is removed, it still increased with Dwet (Figure S1c). Using this technique, Ruehl et al.  also observed an increase in hygroscopicity with Dwetfor sodium dodecyl sulfate, a well-known surfactant, suggesting that the SOA used in these experiments was also surface-active.
Table 1. Summary of High-RH Hygroscopicity Measurements (Excluding Those inFigure 2)
 In the experiments on SOA-NaCl particles, theκ value for the SOA component alone (κSOA) can be determined using the Zadanovkii-Stokes-Robinson (ZSR) mixing rule [Stokes and Robinson, 1966], which assumes that κtotis the volume-weighted average of the individual components. In most of our mixed SOA-NaCl experiments, the inferred value ofκSOA was much greater than κSOApure (Table 1). For example, for mixed particles with a dry SOA volume fraction (fSOA) of 88% had κ = 0.52 ± 0.18, which yields κSOA = 0.41, given κNaCl = 1.26 and using ZSR. This κSOA is a factor of 15 larger than κSOApure, and well beyond what could be explained by variation in RH. Measurements of fSOA = 80% and 90% SOA particles yielded similarly high values of κSOA (Table 1). However, in deriving all of these κ values, we have assumed σ is that of pure water (72 mJ m−2). If, instead, we assume σ is reduced by a factor of ∼2, the calculated κSOA decreases by an order of magnitude, and is thus similar to κSOApure.
 When fSOA was decreased to 67%, however, κ actually decreased to 0.22 ± 0.11 (Figure 1 and Table 1). The ZSR-derived value ofκSOA for these particles was actually lower than κSOApure, and in fact was negative. This suggests that interactions between SOA and NaCl reduced their combined water uptake. Vaden et al. found that in similar SOA-NaCl particles, some NaCl actually dissolved in the SOA coating; such an interaction could be the cause ofκSOA< 0. In summary, enhanced hygroscopicity was observed in mixed SOA-NaCl particles, but only whenfSOA was greater than 67%.
 To further explore the dependence of hygroscopicity on SOA:NaCl ratio, NaCl particles with a constant Ddry = 0.14 μm were coated with a variable amount of SOA, resulting in mixed particles with Ddry between 0.19 to 0.28 m (fSOA from 60 to 88%). The hygroscopicity of the coated particles was measured at RH = 99.92 ± 0.05%. When fSOA ≤ 70%, κSOA was approximately equal to κSOApure. As fSOA increased above 70%, κSOA increased, eventually attaining a maximal value of ∼0.4 at fSOA ∼ 80% (Figure 2a). This observation is consistent with the above finding that κorg values were above the ZSR line only when fSOA was at least 67% (Figure 1).
3.2. Derived σ Values
 To better distinguish between the Kelvin and Raoult effects, it is helpful to examine the measured Dwet distributions. The sharp increase in Dwet over the range 70% < fSOA < 77% (Figure 2b) cannot be adequately explained solely with the Raoult effect. If κSOA is constant (dashed grey lines in Figure 2b), Dwet would increase proportionally to Ddry. Instead, Dwet increases much more rapidly. Since RH is constant, an increase in SOA solubility also cannot explain these results. Therefore the Kelvin effect must contribute to this rapid change in hygroscopicity. Because all other variables in the Kelvin term are constant, droplet σ must be changing in response to an increase in fSOA of the dry particle.
 Using equation (1), we calculate values of σ that fit our experimental data, using observed Dwet, Ddry, and RH, and assuming κNaCl = 1.26 and κSOA = 0.026 (Table 1). When fSOA ≥ 77%, σ is reduced to below 18 mJ m−2, or by at least 75% from that of pure water (Figure 2c). If we use instead κSOA= 0.1, an upper-limit based on previous studies, we find thatσ < 27 mJ m2 (a 62% reduction). This value of κSOA = 0.1 is greater than that observed for pure SOA particles, even at water activities ≳0.999. Thus, for all reasonable values of κSOA and accounting for experimental uncertainties, sufficient amounts of organics reduce σ by at least 50%, and likely up to ∼75%. These reductions are greater than typically observed of macroscopic aqueous surfactants. This could be due to the smaller length scale of our droplets, and/or the lack mechanical manipulation (including creation or destruction) of the aqueous surface in our experiments. The creation of new surface in particular is known to delay the equilibration of σ [Eastoe and Dalton, 2000], resulting in a higher σ value.
 To better understand the decrease in σ as fSOA increases from 70 to 77%, Figure 2b includes lines of constant aqueous SOA film thickness (assuming all SOA partitions to the droplet surface). These lines are much steeper than those corresponding to constant κSOA, going as Ddry1.5 instead of Ddry, and thus better match the observations. The reduction in σ occurs when the droplet is covered in an SOA film of at least 0.8 nm thickness. Although some of the SOA may remain in the droplet bulk, 0.8 nm is on the low end of typical surfactant monolayer thicknesses, suggesting that a large fraction of the SOA partitions to the surface. In the transitional interval (0.21 < Ddry < 0.23 μm), the droplet grows to the point at which this thickness is reached. When fSOA > 80%, this point is not reached before the droplets attain their equilibrium size at the minimum value of σ. Therefore Dwet is limited by whatever minimum value of σ is possible. Overall, this behavior is more consistent with the model of an insoluble surfactant; soluble surfactants would be expected to have a much more gradual change in σ with fSOA.
3.3. Relationship to CCN Activation
 The previous results describe droplets at equilibrium with high RH (i.e., just below the critical supersaturation); next, we consider the minimum value of fSOA required for σ reduction to still be relevant at the point of CCN activation. The film thickness at this point is determined primarily by fSOA, but also varies with σ and the hygroscopicity (κinorg) of the non-surface-active fraction of the particle, as these influenceDwet. If σ and κinorg at activation do not vary with Dwet, then Dwet at activation is proportional to Ddry1.5 [Lewis, 2008]. Therefore, the droplet surface area at activation is proportional to the dry particle volume, and thus the assumed film thickness is independent of Ddry. For reasonable assumptions of σ and κinorg, a film at least 0.8 nm thick will only exist on an activating droplet if the dry particle is composed of at least ∼80% SOA (Figure S2).
 These theoretical predictions are consistent with observations of mixed SOA-NaCl CCN activity.κtotderived from these experiments for both 67% and 88% SOA-NaCl particles was relatively low (0.13 and 0.12, respectively), similar toκtot of 67% SOA:NaCl particles at high but subsaturated RH (Table 1). Because NaCl is so hygroscopic, a film on an organic-NaCl particle will be thicker than 0.8 nm at activation only iffSOA > 0.9 (Figure S2).
 Finally, unlike water uptake measurements made at lower RH, these experiments are directly applicable to supersturated conditions. This is because aw is similar to what would be expected for smaller (dry) particles at the point of CCN activation. Specifically, aw of the solution droplets in Figure 2 ranges from 0.997 to 0.999, and aw of droplets formed on sure pure SOA particles was at least 0.998 (Figure S1b). Because Köhler theory predicts that the Kelvin effect is three times the magnitude of the Raoult effect at activation, an activating CCN with aw = 0.998 (i.e., reduced from pure water by 0.2%) would have a Kelvin effect that increases the equilibrium RH by 0.6%. It would thus have a Sc = 0.4%, a fairly typical value for atmospheric CCN.
 This study presents strong evidence that surface tension reduction can occur in microscopic droplets and augment their hygroscopicity. Low water uptake was observed for pure SOA formed via α-pinene ozonolysis (κSOApure = 0.026 ± 0.010), even at high RH (99.7–99.9%). The SOA was much more hygroscopic (κSOA∼ 0.4) when internally mixed with NaCl, which is attributed to a ∼75% reduction in surface tension. This only occurred, however, if enough SOA was present to form a surface layer with a minimum thickness of about 0.8 nm on the wet droplet. Our results suggest that only particles that are predominantly (i.e., ≳80%) composed of surface-active material will have films sufficiently thick to experience enhanced CCN activity due toσ reduction.
 This work was supported by the National Aeronautics and Space Administration Atmospheric Radiation Program and the National Science Foundation grant ATM-0837913.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.