Report

An interfacial mechanism for cloud droplet formation on organic aerosols

See allHide authors and affiliations

Science  25 Mar 2016:
Vol. 351, Issue 6280, pp. 1447-1450
DOI: 10.1126/science.aad4889

Organic contributions to cloud theory

Current theories about the formation of cloud droplets from aerosol particles containing organic components assume that the organic molecules are distributed throughout the droplet. Ruehl et al. show that this assumption is not always correct (see the Perspective by Noziere). During droplet nucleation, droplet diameters were 50% larger than predicted by the standard model. This suggests that the organic particles reside in a surface layer rather than in the bulk of the droplet. Models that neglect organic surface activity will thus underestimate how well organic-rich particles seed clouds.

Science, this issue p. 1447; see also p. 1396

Abstract

Accurate predictions of aerosol/cloud interactions require simple, physically accurate parameterizations of the cloud condensation nuclei (CCN) activity of aerosols. Current models assume that organic aerosol species contribute to CCN activity by lowering water activity. We measured droplet diameters at the point of CCN activation for particles composed of dicarboxylic acids or secondary organic aerosol and ammonium sulfate. Droplet activation diameters were 40 to 60% larger than predicted if the organic was assumed to be dissolved within the bulk droplet, suggesting that a new mechanism is needed to explain cloud droplet formation. A compressed film model explains how surface tension depression by interfacial organic molecules can alter the relationship between water vapor supersaturation and droplet size (i.e., the Köhler curve), leading to the larger diameters observed at activation.

Accurate predictions of the impact of aerosols on cloud properties, and thus the radiative balance of the atmosphere, rely on simple parameterizations of cloud droplet formation. Despite the simplicity required of such parameterizations, they must be based on robust chemistry and physics to ensure the validity of climate predictions. From Köhler theory (1), the cloud condensation nuclei (CCN) activity of an aerosol is governed both by its size and its molecular constituents that can lower the water activity and/or surface tension of aqueous droplets below that of pure water. There are a number of extensively used empirical parameterizations of Köhler theory that neglect surface tension depression and assume that cloud droplets form on particles that contain sufficient solute (2, 3). In our study, droplet sizes measured up to and including the point of cloud droplet activation reveal that most organic aerosol (OA) contributes to CCN by adsorbing to the air/droplet interface.

One popular parameterization of Köhler theory, known as κ-Köhler theory (3), describes the lowering of water activity by a solute using a single parameter, κ: a dimensionless ratio of the molar volume of water to the average osmolar volume of the aerosol. κ-Köhler theory is used to interpret field observations and laboratory experiments in an effort to relate aerosol composition to hygroscopicity (4). Although many field studies, including those in Colorado (5), Ontario (6), coastal California (7), and the Amazon (8), yield reasonable values for the hygroscopicity of the organic component of the aerosol (κorg) when interpreted via κ-Köhler theory, there are observations that suggest more complex behavior (non-ideal solution and surface tension effects) not captured in current water activity–based parameterizations of CCN activity (9, 10).

Neglecting surface activity in CCN parameterizations appears consistent with predictions of surface-bulk partitioning models (e.g., Szyszkowski-Langmuir adsorption theory). When using parameters obtained for macroscopic solutions, the bulk concentrations of surface-active solutes are predicted to be strongly depleted in microscopic droplets, thus increasing water activity and negating any increase in CCN activity caused by a reduction in surface tension (11). This has led many to conclude that accounting for surface activity is not necessary to accurately predict the CCN activity of OA, despite measurements of reduced surface tension in macroscopic aqueous solutions of atmospheric OA and relevant model compounds (12, 13). Furthermore, partitioning models fail to predict concentrations of surfactants in submicrometer droplets, suggesting significant limitations of current approaches to accurately describe surface activity at that length scale (14).

Using a continuous-flow streamwise thermal gradient chamber [section S1 of (15)], we measured the diameter of droplets (Dwet) that form on mixed organic-ammonium sulfate (AS) particles at water vapor supersaturation [S, or relative humidity (RH) – 100%] approaching and including the point of CCN activation. Although it is common to compute Köhler curves to predict the critical supersaturation (Sc) of aerosols, direct measurements of Dwet as a function of S (15) allow OA CCN activity to be additionally constrained by droplet activation diameter (Dwet,c).

As shown in Fig. 1A, pure 200-nm-diameter (Ddry) AS particles activate at Dwet,c ~ 2.5 μm, consistent with κ-Köhler theory. Mixed aerosols of AS and sucrose (known to be surface-inactive) activate at Dwet,c = 0.8 μm, consistent with a water activity–based parameterization (Fig. 1B). These results are in contrast to AS aerosol coated with a series of small dicarboxylic acid (Fig. 1, C and D, and figs. S5 to S7) and α-pinene secondary OA (SOA, Fig. 2). The functional form of Dwet versus S deviates substantially from κ-Köhler predictions, with a Dwet,c that is ~50% larger than predicted by constant κorg for a given Sc. For succinic acid–coated AS (Fig. 1D), Dwet,c is observed to be 1.8 μm, which is substantially larger than predicted (Dwet,c = 1.4 μm), assuming κorg = 0.31. A similar difference was observed for malonic acid (Fig. 1C). For SOA-coated AS in Fig. 2, Dwet,c = 1.3 to 1.4 μm, which is much larger than the κorg predictions of Dwet,c = 0.9 μm.

Fig. 1 Köhler curve observations for (A) pure AS, (B) sucrose + AS, (C) malonic acid + AS, and (D) succinic acid + AS particles.

D = 200 nm for pure AS particles, and for all other particles D = 150 nm (a 50-nm AS seed + a 50-nm radial coating thickness, corresponding to 89% organic by volume). As described in section S1 of (15), Dwet is measured by phase Doppler interferometry, along the centerline of a thermal gradient chamber (29) after ~10 s of exposure to RH ~ 100% and is therefore not sensitive to decreases in surface tension that might occur over longer time scales as recently observed for aerosol and biological surfactants (14). Solid and open circles represent unactivated and activated droplets, respectively. Because it is not always apparent when activation occurs solely from measurements of Dwet, Sc (horizontal solid black lines) is measured using a separate CCN counter (CCNC, Droplet Measurement Technologies). Dashed red lines are the Köhler curves predicted using a water activity parameterization (κorg). Solid blue lines are those predicted by the compressed film model [section S2 of (15)], and dashed black lines are those for the AS seed particles. The solid red line in (A) is the Köhler curve obtained with a parameterization of the water activity of dilute AS (30). Also shown are the points of CCN activation predicted by κorg (red diamonds) and δorg (blue diamonds).

Fig. 2 Köhler curve observations for AS + α-pinene SOA particles (D = 175 nm) prepared in a flow tube reactor.

As described in section S1 of (15), SOA was generated by ozonolysis in a flow tube reactor and coated onto 85-nm-diameter AS seeds (91% SOA by volume). (A) Dashed red lines are the Köhler curves predicted using a water activity parameterization; solid blue lines are those predicted with the compressed film model [details in section S2 of (15)]; and dotted black lines are those for the AS seed particles. The horizontal black line indicates the Sc observed with a conventional CCN instrument (CCNC, Droplet Measurement Technologies). Also indicated is the point of CCN activation predicted by κorg (red) and δorg (blue) symbols. Included are the (B) surface tension (σ) and (C) fraction of SOA at the surface (fsurf), as predicted by the compressed film model.

For pure AS aerosol, the evolution of Dwet with S, and the size of Dwet,c, are exactly what is predicted by κ-Köhler theory and serve to validate the experimental approach. The mixed AS/sucrose observations show that κ-Köhler theory correctly predicts Dwet versus S and Dwet,c in the case of a highly water-soluble organic compound that does not depress surface tension (i.e., is surface-inactive). In contrast, AS aerosols coated with a series of dicarboxylic acids, some of which are known to be surface-active, exhibit much larger activation diameters and a different functional form of Dwet versus S than is predicted by κ-Köhler theory (Figs. 1 and 2 and figs. S3 to S7). Although limited organic solubility can alter the shape of the Köhler curve (16), it cannot explain the consistently large droplet activation diameters observed for these dicarboxylic acids, whose bulk solubility varies from near that of sucrose (i.e., malonic acid) to over two orders of magnitude smaller (table S1). The mixed SOA/AS aerosols exhibit the same deviation from κ-Köhler theory as the dicarboxylic acids. Collectively, the observed differences between Dwet,c for the organic acids (also SOA) and sucrose suggest that the discrepancies with κ-Köhler theory originate from surface effects rather than non-ideal behavior of mixtures.

To account for the surface activity of organics within Köhler theory, an equation of state is required to relate the bulk and surface concentrations of the various solution components. Associated with the equation of state is an isotherm that relates organic surface concentration to surface tension. Previously, the Szyszkowsky-Langmuir equation (13, 17) was used to compute Köhler curves (i.e., to predict Sc) for model organic compounds (16). This particular treatment of bulk-surface partitioning [section S2 of (15)] cannot fully explain the observed Dwet as a function of S. Instead, an equation of state that allows for a two-dimensional (2D) phase transition is required. To provide a self-consistent model description of the observations shown in Figs. 1 and 2, a compressed film model (18) was used to describe the relationship between surface tension depression and organic surface coverage or thickness on the droplet (figs. S3 to S7). Model details can be found in section S2 of (15) and are described conceptually here. The model contains a 2D phase transition between “gaseous” and “compressed” surface states, which depends on surface concentration (i.e., molecular packing). Because the quantity of organic material is fixed by the composition of the original dry aerosol, changes in surface tension occur when S increases and the droplet grows, decreasing the surface concentration by providing a larger surface area per molecule. At low S (below Sc), the surface organic concentration is high, and the molecules adopt a compressed state, which lowers the droplet’s surface tension below that of liquid water. At higher S (near Sc), the droplets are larger, the surface concentration is lower, and the molecules at the interface are non-interacting (i.e., in a gaseous surface state), with a droplet surface tension nearly equal to that of pure water. The data shown in Figs. 1 and 2 (and figs. S3 to S7) are best replicated if it is assumed that the surface tension increases linearly with decreasing surface concentration (18). As shown in Figs. 1 and 2 and figs. S3 to S7, the compressed film model can reasonably account for the functional form of S versus Dwet. Although the compressed film model shows that some of the organic material is dissolved in the droplet bulk, the model reveals that for SOA and most of the model compounds [section S3 of (15)] most of the organic material is at the droplet surface (Fig. 2B and figs. S3 to S7), in stark contrast with the underlying assumption of κ-Köhler theory.

The compressed film model predicts that a particle will reach Sc when the surface film decreases in thickness to the point that individual molecules begin to separate; i.e., a 2D phase transition occurs, and the surface tension no longer varies with increasing Dwet (15). This occurs at an S and Dwet that lies on the postactivation portion of the Köhler curve for the bare inorganic AS seed. At this intersection point, all of the hygroscopicity can be attributed to the reduction of water activity by AS. The compressed film model explains this by predicting that the organic material is adsorbed to the interface (hence no decrease in water activity by the organic), but Dwet is large enough to lower the organic surface concentration to a point where the resulting surface tension is near that of pure water. Although there are several parameters needed to constrain the model, a single-parameter approximation to the compressed film model can be used if it is assumed that all the organic material resides at the droplet surface. This parameter, termed δorg, corresponds to a film thickness (in nanometers) on the droplet surface where the 2D phase transition occurs (i.e., where the surface tension depression goes to zero). Although surface concentration is used more often than film thickness for monolayers of known composition, film thickness is preferable for discussions of particle hygroscopicity, because aerosol composition is often complex and poorly constrained molecularly, and because sizes (particle and droplet diameters) are measured in CCN experiments.

The Sc for a particle is computed from δorg if both the organic fraction (forg) and κinorg (inorganic hygroscopicity) are known [section S2 of (15)], which is the same set of parameters required to predict Sc given κorg. In Fig. 3, the surface (δorg) and bulk activity (κorg) parameterizations are compared with additional measurements of Sc versus dry organic coating thickness on AS seed particles for the same series of dicarboxylic acids and SOA. The measured critical supersaturation (Sc) decreases with increasing dry organic coating thickness for all particles measured. The data for each compound/SOA in Fig. 3 are fit with both a constant organic osmolar volume (related to κorg) and film thickness (δorg).

Fig. 3 Sc as a function of coated dry diameter for (A to F) a series of dicarboxylic acids or (G) SOA generated via ozonolysis of α-pinene coated onto AS seed particles (D = 35 nm).

Both seed and coated diameters were size-selected by a differential mobility analyzer (TSI, Model 3080). Dashed and solid lines are the best fits for each substance, using a single parameter for organic hygroscopicity (κorg and δorg, respectively), and dotted lines indicate predictions assuming an ideal solution (κorg,ideal). These values are listed in table S1 of (15).

In most cases, constant κorg does not replicate the observed curvature in Sc versus coated diameter (Fig. 3), generally predicting a more shallow slope than is observed. There are two exceptions: malonic and suberic acid. However, the molar volumes required to replicate these data are much smaller than what is reported in the literature: malonic acid by 36%, and suberic acid by 3.7 times. Such a dramatic increase in hygroscopicity cannot be explained by the dissociation of these weak acids. In contrast, the δorg approximation does capture the curvature observed in most plots of Sc versus forg shown in Fig. 3. Thus, the δorg approximation to the compressed film model correctly accounts for both the droplet size at activation (Figs. 1 and 2 and figs. S3 to S7) and the evolution of Sc with dry OA fraction (Fig. 3).

The compressed film model offers an explanation for some recent ambient CCN observations. In CCN closure studies, agreement between observations and predictions is often best when the OA fraction is assumed to be insoluble, as recently reported in California (19). In particles that have similar amounts of organic and inorganic material, or are predominantly inorganic, there is not enough organic material to form a compressed film on the droplet at or near the point of activation. Although this organic material will be adsorbed to the droplet surface, there is an insufficient concentration at the interface to reduce surface tension; instead, the organic will effectively behave as insoluble material with respect to CCN activation. Recently, activation diameters inferred from particle and droplet size distributions in urban fog were unexpectedly large (20), similar to what was observed in this study. Finally, observations of enhanced OA hygroscopicity that were recently reported for a coastal site on Vancouver Island (9) do not seem reasonable for marine organic material, suggesting, in light of our results, that a major contribution to CCN activity from surface activity could be likely.

The compressed film model also offers an explanation for some unresolved questions arising from laboratory CCN studies, including several single-component OAs that exhibit anomalously large CCN activity despite limited bulk solubility (2124). For example, pimelic acid is CCN-active (κ = 0.14 to 0.16), despite its low solubility that suggests that κ should be five times smaller than observed (25). This model predicts that most OA (especially OA of limited solubility) is adsorbed to the surface of microscopic droplets. Thus, the relatively small bulk concentrations may not, in fact, exceed solubility limits. Although Raoult (water activity) effects may also be important for aerosol with unknown or more complex composition (26), it is an unlikely explanation for the anomalously high CCN activity of individual compounds. The compressed film model also resolves the “gap” between κorg values derived from CCN activity experiments, which are often much larger than those derived from subsaturated measurements of hygroscopic growth (27). The surface tension reduction by a compressed film will only increase hygroscopicity at high RH (near 100%) (28), whereas at lower RH, where the water activity term dominates, the effect of the film will be to lower hygroscopicity relative to the fully soluble assumption.

These results point to an alternative mechanism for cloud droplet formation in mixed organic/inorganic aerosol. Although a water activity–based parameterization correctly predicts the droplets’ sizes under supersaturated conditions for pure AS and sucrose (a non–surface-active compound), it fails to correctly account for the larger cloud droplets that form on AS coated with a series of dicarboxylic acids (with a broad range of water solubility) and SOA. Thus it is unlikely that our results can be attributed to bulk solubility effects, but rather can be explained if most of the organic material exists as an interfacial compressed film, which reduces surface tension, allowing larger droplets to form before activation. At activation, the compressed film transitions to a gaseous state, and the surface tension of the droplet is nearly equal to that of pure water. Although the assumption that organic material dissolved in the droplet bulk may yield reasonable CCN predictions, in field measurements and in the laboratory these results can help explain several outstanding questions and highlight the potential importance of interfacial organics in the formation of cloud droplets on OA.

Supplementary Materials

www.sciencemag.org/content/351/6280/1447/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S7

Tables S1 to S3

References (3139)

References and Notes

Acknowledgments: This work is supported by the Office of Science Early Career Research Program, through the Office of Energy Research, Office of Basic Energy Science of the U.S. Department of Energy under contract no. DE-AC02-05CH11231. The continuous-flow streamwise thermal gradient chamber was originally developed by Patrick Chuang and Anthanasios Nenes with support from NASA’s Atmospheric Radiation Measurement program.
View Abstract

Navigate This Article