Measurements of cloud condensation nuclei activity and droplet activation kinetics of fresh unprocessed regional dust samples and minerals

This study reports laboratory measurements of cloud condensation nuclei (CCN) activity and droplet activation kinetics of aerosols dry generated from clays, calcite, quartz, and desert soil samples from Northern Africa, East Asia/China, and Northern America. Based on the observed dependence of critical supersaturation, sc, with particle dry diameter,Ddry, we found that FHH (Frenkel, Halsey and Hill) adsorption activation theory is a far more suitable framework for describing fresh dust CCN activity than Köhler theory. One set of FHH parameters ( AFHH ∼ 2.25±0.75, BFHH ∼ 1.20±0.10) can adequately reproduce the measured CCN activity for all species considered, and also explains the large range of hygroscopicities reported in the literature. Based on a threshold droplet growth analysis, mineral dust aerosols were found to display retarded activation kinetics compared to ammonium sulfate. Comprehensive simulations of mineral dust activation and growth in the CCN instrument suggest that this retardation is equivalent to a reduction of the water vapor uptake coefficient (relative to that for calibration ammonium sulfate aerosol) by 30–80%. These results suggest that dust particles do not require deliquescent material to act as CCN in the atmosphere.


Introduction
Clouds are an important component of the Earth's radiation budget and hydrological cycle.Even small changes in cloud properties may have significant impacts on climate (Collins et al., 1994).Perturbations in aerosol loadings can alter cloud properties, giving rise to the aerosol indirect effect on climate.Aerosol effects on clouds constitute one of the Correspondence to: A. Nenes (athanasios.nenes@gatech.edu)most uncertain components of anthropogenic climate change (Forster et al., 2007).Mineral aerosol (or dust) is one of the lesser understood of aerosol species in the study of aerosolcloud-climate interactions.It has been well recognized that dust plays an important role in cold cloud processes because of its effectiveness as Ice Nuclei (IN) (DeMott et al., 2003;Field et al., 2006).Dust can also affect warm clouds by acting as Cloud Condensation Nuclei (CCN), changes of which affect their radiative (Twomey, 1974) and precipitation properties (Rosenfeld et al., 2001).
In general, the ability of dust particles to serve as CCN depends on their mineralogy, size, morphology, and atmospheric processing.Quantitative understanding of the interactions of dust with water vapor is complex because of its varying source-dependent mineralogical composition and aging during its atmospheric residence.Mineral aerosol may constitute of iron oxides (e.g., hematite, goethite), carbonates (e.g., calcite, dolomite), quartz, and clays (e.g., kaolinite, illite, and montmorillonite) (Lafon et al., 2006;Chou et al., 2008;Coz et al., 2009;Twohy et al., 2009).Dust particles mainly originate from arid and semi-arid regions, with an annual emission of approximately 1000-5000 Tg (Schuttlefield et al., 2007).Differences in parent soils, and emission and transport processes cause substantial variability in size-resolved composition and morphology of dust particles (Sokolik et al., 2001;Jeong and Sokolik, 2007).Dust particles can remain suspended in the atmosphere for up to several weeks and can be transported over large distances downwind from source regions.During their transport, dust particles (especially the carbonate fraction which can comprise up to 30% of the total mass), provides reaction sites for heterogeneous chemical reactions with atmospheric trace gases and pollutants (Levin et al., 1996), resulting in modified dust properties, such as enhanced hygroscopicity (Hatch et al., 2008).However, not all dust particles undergo aging (Prospero, 1999;Ganor and Mamane, 1982;Ganor and Foner, P. Kumar et al.: Measurements of cloud condensation nuclei activity and droplet activation kinetics 1996).Depending on transport routes of dust plumes and environmental conditions, dust particles can remain unprocessed and have the same properties as freshly emitted dust in source regions.Thus, it is important to understand the CCN activity of fresh dust particles as well as aged dust.
To describe the CCN activity of freshly emitted dust, two phenomena must be accounted for: (i) the effect of solute (which may be present in freshly emitted dust or formed during atmospheric aging), and (ii) the adsorption of water on the insoluble component of the dust particles.The former can be accounted for by using Köhler theory (KT) (Köhler, 1936) and the latter with adsorption activation theory (AT) (Henson, 2007;Sorjamaa and Laaksonen, 2007;Kumar et al., 2009a).The formulation of Henson (2007) used the BET (Brunauer et al., 1938) adsorption isotherm, while Sorjamaa and Laaksonen (2007) used the multilayer FHH (Frenkel, Halsey and Hill) adsorption isotherm with two adjustable parameters (A FHH and B FHH ).Based on analysis of published data on dust-water interactions, Kumar et al. (2009b) showed the importance of including water adsorption effects when describing the hygroscopic and CCN behavior of mineral aerosol.The same study found that FHH particles require less water to activate to cloud droplets than particles activating by KT; this implies that the competition for water vapor by FHH particles is less intense that KT particles to form droplets with implications for parcel maximum supersaturation, s max , and cloud droplet number, N d .Kumar et al. (2009a) addressed the need to account for adsorption activation in atmospheric models by developing a cloud droplet formation parameterization where the CCN constitutes an external mixture of soluble aerosol (that follow KT) and insoluble aerosol (that follow FHH adsorption activation theory, FHH-AT).Here, we report new measurements to further support the dust-CCN parameterization developed by Kumar et al. (2009a).
Past studies have already demonstrated that both regional dusts as well as individual clays can interact with water and act as effective CCN.For example, Koehler et al. (2009) and Herich et al. (2009) measured CCN activation of two types of regional dust samples (Northern Africa and Arizona Test Dust) and several clays (kaolinite, illite, and montmorillonite), respectively, at water vapor supersaturation relevant to atmospheric conditions.These studies, however, parameterized the observed hygroscopicity using a KT framework in terms of a hygroscopicity parameter, κ (Petters and Kreidenweis, 2007).This approach was evaluated by Kumar et al. (2009b), who, after examining the relationship between s c and D dry for the published dust samples suggested that FHH-AT is a better description of fresh dust CCN activity as the s c -D dry exponents determined from FHH-AT were closer to observations than from KT. Furthermore, no study to date has accounted for non-sphericity effects in the CCN activity relationships, even when it is well known that dust particles are non-spherical (e.g., Okada et al., 2001;Chou et al., 2008).Further, the effect of multiple-charged particles in the electri-cal mobility classification for measurements of size-resolved CCN activity (required for determining s c and D dry ) is often addressed by removal of the secondary peaks in the activation curves (e.g., Lance et al., 2006;Rose et al., 2008).If multiple-charged particles are present in significant enough numbers (such as for dust CCN), this approach may not suffice causing biases in measured CCN activity towards higher hygroscopicity (Petters et al., 2007).A comprehensive analysis of charging efficiency (e.g., Moore et al., 2010) needs to be considered to avoid such biases in observed hygroscopicity.
In this study, we investigate the CCN-relevant properties of clays and several dust samples representative of major regional dust sources.Measurements were carried out with a Droplet Measurement Technologies Continuous-Flow Streamwise Thermal Gradient CCN (CFSTGC) counter (Roberts and Nenes, 2005;Lance et al., 2006).The CCN activation behavior of mineral aerosols generated from Northern American, African, and East Asian desert soils as well as individual clays (illite and montmorillonite), calcite (CaCO 3 ), and quartz (SiO 2 ) are studied.The effects of multiple charging and shape (non-sphericity) on the electrical mobility sizing of particles and activation curves are examined.The experimental results are used to infer the dominant activation physics (KT or FHH-AT) and determine the appropriate adsorption parameters (e.g., A FHH and B FHH ) that describe the hygroscopicity of fresh dust for the use in droplet activation parameterizations of Kumar et al. (2009a).Finally, using the method of threshold droplet growth analysis (TDGA, e.g., Asa-Awuku et al., 2010;Padró et al., 2010), potential retardations in the activation kinetics of dust (compared to calibration aerosol) are identified.A comprehensive simulation of dust activation in the CCN instrument is then performed to parameterize these kinetic delays in terms of changes in the effective water vapor uptake coefficient.

Measurements and data analysis
2.1 Regional dust samples and individual minerals Aerosols from regional soil samples and individual minerals/clays were generated and analyzed in this study.Table 1 provides a summary of the analyzed samples, including information on the location of sample collection.The soil samples were collected in source regions of Northern Africa and East Asia.Commercially available Arizona Test Dust (ATD) was used as representative of North America soil.Individual minerals/clays used to generate aerosol were analyzed as purchased, with no physical and chemical treatments to resemble atmospheric behaviors.

Measurements of CCN activity
The measurement setup consists of three sections: aerosol generation, particle size selection, and CCN measurement (Fig. 1).To generate aerosol, approximately 3 g of the desired sample were placed in a 1000 ml sealed Erlenmeyer flask which is connected to a Burrell-Wrist Action Shaker (Model 75).Compressed filtered air is introduced into the flask that generates polydisperse fine aerosols by mechanical disintegration ("saltation") with a distribution that resembles the size distributions of dust plumes generated in the natural source regions (Lafon et al., 2006).
The dry aerosol is then sent to the electrostatic classifier for particle size selection (TSI Model 3080) with a Differential Mobility Analyzer (DMA, TSI Model 3081).Before entering the classifier, aerosols are passed through an impactor to remove supermicro-meter size particles (i.e., size greater than 1 µm) and then charged with a series of Kr-85 neutralizers.The particles are then classified in the DMA by their electrical mobility set by the voltage applied to the DMA.The Sheath flow rate in the DMA is set to 2.3 l min −1 , and the monodisperse flow is set to 0.45 l min −1 .The classified aerosol flow is mixed with filtered air and then sampled by a Condensation Particle Counter (CPC, TSI Model 3010), and a Droplet Measurement Technologies Continuous Flow Streamwise Thermal Gradient CCN (CFSTGC) chamber.
The CPC measures the total concentration of aerosol, or condensation nuclei (CN) present in the monodisperse stream.The fraction of aerosol acting as CCN is measured by exposing particles to a constant water vapor supersaturation within the CFSTGC.This is done by flowing the aerosol in a cylindrical column with wetted walls upon which a thermal gradient, T , is applied in the axial direction.The difference in diffusivity between water vapor and heat is exploited for the generation of water vapor supersaturation, s, which reaches maximum at the column centerline.CCN flowing along the column centerline are activated to cloud droplets and are counted at the exit with an optical particle counter (OPC).Each value of T generates a unique supersaturation value, which in this study varied between 0.15% and 1%.CCN activity is characterized by the dry activation diameter, D dry , which corresponds to the minimum dry particle diameter that activates at the certain supersaturation of interest, s c .D dry is found by expressing the ratio of CCN to CN concentration as a function of dry particle diameter, and, determining the diameter for which 50% of the classified aerosol acts as CCN.
The calibration of the instrument supersaturation is determined from the D dry of (NH 4 ) 2 SO 4 calibration aerosol at a given T .(NH 4 ) 2 SO 4 aerosol was generated by atomizing an aqueous solution and subsequently drying the droplet stream with a series of silica-gel diffusion dryers.D dry of (NH 4 ) 2 SO 4 is then related to s by applying Köhler theory, assuming that (NH 4 ) 2 SO 4 has a shape factor of 1.04 in the DMA (Kuwata and Kondo, 2009), density of 1760 kg m −3 , surface tension of water (calculated at the average column temperature), molar mass of 0.132 kg mol −1 , and osmotic coefficients calculated with the Pitzer activity coefficient model (Pitzer and Mayorga, 1973).A relationship between T vs. instrument supersaturation is determined by repeating the above calibration procedure over a range of T .This relationship is then used in all dust activation experiments.Calibration is repeated throughout the measurements, and exhibits little variability (about 5% relative uncertainty in instrument supersaturation).
Size-resolved CCN activity is carried out using the Scanning Mobility CCN Analysis (SMCA) (Moore et al., 2010), where the DMA used for aerosol classification is operated in scanning voltage mode.This allows the concurrent determination of aerosol size distribution and size-resolved CCN activity over a voltage scan cycle.In this study, the complete range of dry particle size (20-850 nm) is scanned over three minutes.The CFSTGC was operated at a flowrate of 0.50 l min −1 and a sheath-to-aerosol ratio of 10:1 (or 7.5:1).SMCA also provides the droplet distribution of activated CCN (measured in the optical particle counter of the CCN P. Kumar et al.: Measurements of cloud condensation nuclei activity and droplet activation kinetics instrument) as a function of their dry diameter.The dependence of droplet size on the supersaturation profile and dust dry particle size is used to study the dust activation kinetics.

Data analysis methodology
The measurement of D dry , and corresponding s c , is fitted with a power law function s c = CD x dry .The experimental exponent, x exp (Kumar et al., 2009b), is then compared against the exponent determined from fits of KT and FHH-AT to the data.The appropriateness of each theory is evaluated based on its ability to reproduce x exp .According to KT, particles with appreciable hygroscopicity exhibit x = −3/2.In FHH-AT, x depends on the value of A FHH and B FHH but generally ranges between −0.80 and −1.20 (Kumar et al., 2009a).The same fitting procedure also determines the adsorption parameters A FHH and B FHH (for FHH-AT), and the hygroscopicity parameter κ (for KT).
B FHH strongly affects the shape of the equilibrium curve and largely determines the existence and value of s c and critical wet diameter, D c (described as the wet diameter of the aerosol particle at the maximum of the equilibrium curve) (Kumar et al., 2009a).A FHH also affects these parameters, but to a lesser extent than B FHH .Figure 2 shows the relationship between D dry and s c for a range of B FHH values computed at surface tension of water equal to 0.072 J m −2 , temperature equal to 298.15 K, and A FHH = 2.50.Lower B FHH values correspond to more hydrophilic dust.As B FHH approaches 3.0, particles become less hydrophilic (with x → −1), which corresponds to insoluble but wettable particles that follow the Kelvin equation.Similarly for KT, as κ decreases from κ = 0.05 to κ = 0 (Fig. 2) particles becomes less hygroscopic causing a decrease in the exponent from x = −1.5 to x → −1.0.It can be seen from Fig. 2 that the slopes determined from KT (expressed in terms of κ) are much steeper than those determined from FHH-AT (expressed in terms of B FHH ).This suggests that the same particle type can exhibit two different s c -D dry exponent values if described by KT or FHH-AT.
Droplet activation kinetics of aerosol inside CFSTGC depends on the supersaturation profiles, residence time, water vapor uptake coefficient, dry particle size (Nenes et al., 2001;Roberts and Nenes, 2005;Lance et al., 2006), as well as the theory used to describe the equilibrium vapor pressure for the particle (KT or FHH-AT).Activation kinetics can be characterized by the difference in droplet size, D w , between dust CCN and (NH 4 ) 2 SO 4 CCN with same s c .A negative D w implies that mineral aerosol exhibits retarded activation kinetics (the converse is typically not observed).This technique is called threshold droplet growth analysis (TDGA) and has been successfully used by a number of insitu and laboratory studies (Asa-Awuku et al., 2010;Padró et al., 2010).
We quantitatively describe the growth of dust by simulating the process of droplet nucleation and growth within the CCN instrument using the comprehensive computational fluid dynamics model.We use the Lance et al. (2006) model, which numerically simulates the temporal and spatial distributions of velocity, pressure, temperature, and water vapor concentration throughout the growth chamber, considering the coupling of particle and gas phases through the release of latent heat and condensation/evaporation of water vapor onto the droplets.The kinetics of dust activation is then parameterized in terms of an effective uptake coefficient, which influences the mass transfer coefficient of water onto the dust CCN.Condensational growth of aerosol is computed based on a size-dependent mass transfer coefficient multiplied by the difference between gas-phase and equilibrium water vapor pressure (Nenes et al., 2001): where D p is the droplet diameter, s is the local instrument supersaturation, ρ w is the water density, M w is the molar mass of water, R is the universal gas constant, T is the average column temperature, P • H 2 O is the equilibrium water vapor pressure, H v is the enthalpy of vaporization of water, D v is the diffusivity of water vapor in air modified for non continuum effects, and k a is the thermal conductivity of air modified for non continuum effects.Here D v is defined by Fukuta and Walter (1970) as: where D v is the diffusivity of water vapor in air, α c is the water vapor uptake coefficient.k a is given by: where M a is the mean molar mass of air, k a is the thermal conductivity of air, ρ a is the air density, c p is the heat capacity of air, and α T is thermal accommodation coefficient (equal to 1.0).For insoluble CCN activating according to FHH-AT, the equilibrium supersaturation of the droplet, s eq , is given by Kumar et al. (2009a): where σ is the CCN surface tension at the point of activation (Pruppacher and Klett, 1997), D dry is the dry CCN diameter, D H 2 O is the diameter of water molecule equal to 2.75 Å (Kumar et al., 2009a), and A FHH and B FHH are adsorption parameters constrained from the activation experiments.
The instrument model was initialized using the appropriate geometric dimensions and operating conditions of DMT CFSTGC (Lance et al., 2006).A computational grid of 200 cells in the radial and 200 cells in the axial direction were used in each simulation.A Lagrangian approach is used to determine CCN growth in the CFSTGC by Eq. ( 1), assuming the particles flow along streamlines occupied by the aerosol region of the chamber (determined from the sheath-aerosol ratio) and grow according to the local water vapor saturation ratio and temperature (Roberts and Nenes, 2005;Lance et al., 2006).The droplet diameter at the exit of the flow chamber is then compared against the measured droplet size distribution, following the binning scheme used in the optical detection of the instrument.The value of uptake coefficient is then inferred by minimizing the discrepancy between predicted and observed droplet distributions in OPC.

Correction for multiply charged particles in SMCA
To account for the effect of multiply charged particles in the activation curves and observed D dry , we assume an equilibrium charge distribution for the particles entering the DMA and apply a correction algorithm as described in Moore et al. (2010).The correction algorithm determines the contribution from the multiply charged (+2, +3, +4, +5 and +6 charges) particles to the total particle counts in each size bin for the CN time series and rebins respective contributions to its "true" size bin.The same procedure is applied to the CCN time series.The inversion of the CN, CCN time series determines the activation fraction and hence D dry .To ensure sufficient residence time for attaining equilibrium charge distribution inside the Kr-85 neutralizers, we determine the number of neutralizers beyond which the inverted size distribution does not change.Test results indicate that 3 Kr-85 neutralizers in series (with a total nominal activity of 10 mCi) were sufficient to completely neutralize the surface charges and attain the Boltzmann equilibrium distribution.The impact of multiply charged particles is shown in Fig. 3, which presents the activation curves (with and without multiple charging corrections) at 0.30% supersaturation for aerosol generated from the Soil 2 sample.D dry increases from ∼170 nm to ∼247 nm upon application of the multiple charge correction.The effect on D dry is further enhanced at lower supersaturations (e.g., s c = 0.15% and s c = 0.20%) that correspond to large particles with a pronounced probability of multiple charging.The uncertainty in the activation efficiency due to counting statistics uncertainty and flow rate variability (expressed as error bars in Fig. 3) were accounted for using the procedure of Moore et al. (2010).

Accounting for dust non-sphericity
Dust particles exhibit a variety of complex shapes that are difficult to measure or express in terms of a unique set of parameters or functions.Characterization of dust nonsphericity is often done by either (i) introducing a dynamic shape factor, χ, (defined as the ratio of drag force, F D , experienced by the non-spherical particle to that experienced by a volume equivalent sphere when both move at the same velocity in the gas; e.g., DeCarlo et al., 2004), or (ii) providing an Aspect Ratio (AR), defined as the ratio of the longest dimension of particles to the orthogonal shortest length (width).Commonly, χ is obtained by tandem electrical mobility and aerodynamic particle sizing (e.g., DeCarlo et al., 2004;Kuwata and Kondo, 2009) and is an integrated measure of the three-dimensional particle shape.AR is measured with electron microscopy that reports two dimensional image projections of particles from which the longest dimension and width are determined (e.g., Kalashnikova and Sokolik, 2004).
Here we assess the effect of dust non-sphericity in CCN activity measurements by considering the range of values of AR or χ reported in the literature for different types of mineral aerosol.A number of recent studies have reported measurements of AR values for species considered in this study.For instance, Chou et al. (2008) report a mean AR equal to 1.7 for Niger dust collected during the AMMA campaign, Kandler et al. (2009) report AR equal to 1.64 for Saharan dust collected over Spain, and Coz et al. (2009) report AR equal to 1.81 for African dust.The AR values for African soils are slightly higher compared to AR of 1.3-1.4 reported by Okada et al. (2001) for East Asian dust.AR can also vary with particle size (Wiegner et al., 2009).To account for this, we considered those values of AR that are most relevant for this study (i.e., particles less than 1 µm).Furthermore, the extent of non-sphericity can be affected by the aerosol generation method (Sullivan et al., 2010) which can give rise to very different particle morphologies from those generated with the dry soft-saltation technique used in this study.Some of the other techniques of aerosol generation include a fluidized bed (Koehler et al., 2009), dry dust generator (Herich et al., 2009), and atomization of a dust aqueous suspension (Koehler et al., 2009;Herich et al., 2009).All above factors can contribute to uncertainty in χ for similar aerosol types.For instances for ATD, Möhler et al. (2008) reported χ = 1.3 while Endo et al. (1998) reported χ = 1.5.Similarly for illite, Hudson et al. (2008) and Möhler et al. (2008) reported χ = 1.3 ± 0.02 and χ = 1.3 respectively.For other clays and minerals analyzed in this study such as montmorillonite, Hudson et al. (2008) reported χ = 1.11 ± 0.03, while Hinds (1999) report a value of χ = 1.36 for quartz.
In this study, we use the published range of dust nonsphericity.As most of the recent studies on African and Asian mineral dust aerosol have quantified dust nonsphericity based on AR, we initially started with the Fuchs (1964) approach to convert from AR to χ assuming mineral aerosol as a spheroid.However, we found that using the Fuchs (1964) approach results in much lower values of χ (∼1.007-1.034)than those determined from direct measurements of χ (∼1.11-1.50).According to Davies (1979), sand particles composed of a mixture of different minerals have a dynamic shape factor of 1.3-1.6.Therefore, in this study non-sphericity corrections are performed for all species considering χ = 1.3 ± 0.2 as this covers as possible values of measured χ.Further, we examine the importance of this uncertainty in χ (between χ = 1.1 and χ = 1.5) for dust-CCN measurements by evaluating its effect on x exp and FHH parameters, A FHH and B FHH .
Size selection in this study is performed using the DMA that classifies a particle by its electric mobility.Electrical mobility can then be related to the physical diameter if the number of elementary charges per particle and χ are known (together with the strength of the electric field and other operational parameters in the DMA).Often χ is assumed unity.For mineral dust, however, χ > 1 which translates into a larger drag force than expected for spherical particles; when neglected, the nonsphericity would eventually lead to underestimation of the particle surface area.In this study, we account for dust non-sphericity by correcting for the surface area of the particle available for water vapor adsorption.This is performed by converting from electrical mobility diameter, D m , to surface-area equivalent diameter, D se , so that the CCN activity data is expressed in terms of the s c -D se relationship.By converting from D dry to D se , the aerosol physical size is expressed in terms of the characteristic length responsible for controlling surface water vapor adsorption.
where λ is the mean free path of the gas molecules and D i corresponds to either of D m or D ve .Application of Eq. ( 5) to determine D ve requires knowledge of χ .
Determining D se from D ve When χ is known (or estimated), the correlation of Leith (1987) is used to relate χ , D se , and D ve : where D n is the diameter of the sphere whose projected area is equal to that of the particle normal to the direction of flow.
For the DMA, D n = D m , hence Eq. ( 7) can be rearranged to express D se as: Despite involved uncertainties, accounting for dust nonsphericity provides more realistic representation of dust particles as well as enables us to determine non-sphericity effects on the physics controlling the activation of insoluble dust particles.showing the effect of including charge and shape corrections on the raw data.Blue shows curve with no correction, brown shows the results with charging corrections and green shows curve after including both charge and shape corrections.Error bars represent experimental uncertainty and numerical uncertainty in D dry at same instrument supersaturation.

Effect of charge and shape corrections on dust CCN activation
The largest change in dry critical activation diameter from multiple charging corrections is observed at the point of lowest supersaturation (corresponding to the largest activation diameters with highest probability of multiple charging).For all species considered in this study (regional dusts and clays/minerals), χ = 1.3 was used to convert from charge corrected electrical mobility diameter, D m to the shape corrected surface-area equivalent diameter, D se .The error bars on D se represent the range using χ = 1.1 as the lower limit and χ = 1.5 as the upper limit.We found that accounting for non-sphericity using χ = 1.3, can result in an increase in activation diameters by up to 18-20% when converting from D m to D se using the procedure outlined above.Based on Fig. 4, as the final activation diameters (after including both charge and shape corrections) lie outside the experimental error bars (or region of experimental uncertainty) determined from the raw data, the essence of including the charge and shape corrections is justified and hence performed for all the samples studied here.We also found that introducing both multiple charging and shape correction changes the dry activation diameters significantly and hence the exponents determined from the s c -D dry relationship.For example in the case of ATD, x exp = −1.23 for the uncorrected data; after applying charge corrections, x exp = −0.78;with shape and charge corrections, x exp = −0.82.This is a very large and important difference, enough to shift the implied activation mechanism from a regime www.atmos-chem-phys.net/11/3527/2011/Atmos.Chem.Phys., 11, 3527-3541, 2011 where both FHH-AT and KT may be active (x exp = −1.23) to a regime where FHH-AT dominates (x exp = −0.82).This example emphasizes the importance of applying corrections (especially for multiple charges) for adequate interpretation of the activation data.
It is also noted that the effect of particle non-sphericity must be incorporated into the diameter used in the Kelvin term at the point of activation if only a few monolayers are adsorbed at activation (Romakkaniemi et al., 2001).We find that for all regional samples considered in our study, the number of adsorbed water vapor monolayers layers range from 100-500 at the point of activation.This implies that at the point of activation, non-spherical dust aerosol has sufficiently high water coverage so that the droplet shape is spherical, and, the molar volume and gas-liquid surface tension of the adsorbed H 2 O approaches that for the bulk water.

Results of dust CCN activation measurements
The CCN activation curves for dry generated dust and mineral/clay samples are shown in Figs. 5 and 6, respectively.CCN activity is presented in terms of dry activation diameter (D se , given by Eq. 8) against instrument supersaturation.The CCN activity data (points) are fit to a power law expression from which the experimental exponent, x exp , is determined.The A FHH , B FHH , and corresponding exponent, x FHH , were determined from fitting the FHH-AT model (lines) to the experimental data via least squares minimization.The dry generation method used in this study did not produce sufficient number concentrations of particles with sizes smaller than 100 nm.Hence the CCN activity is restricted to supersaturations 0.7% and below (corresponding to D dry ∼ 100 nm and above).
Figure 5 clearly demonstrates that dust aerosols are CCN at atmospherically relevant supersaturations.It also indicates that soft saltation technique can generate mineral dust in the fine mode (with D dry between 100 nm and 500 nm) which may contribute to CCN.The measured D dry for different dust samples are much larger than expected for (NH 4 ) 2 SO 4 , suggesting that dust has a lower CCN activation potential than what is expected for soluble aerosol like (NH 4 ) 2 SO 4 .Figure 5 suggests that dust aerosols collected from different regions of the globe can have different activation properties which are attributed to the physical properties, morphology and the chemical composition of the parent soils.The CCN activity comparisons amongst different regional dust samples indicate that East Asian soils have a range of CCN activity potentials with B FHH ∼ 1.1 − 1.3.In comparison, Niger Soil (representative of North African dust) and ATD (representative of North American dust) were found to exhibit less variability, with B FHH = 1.25 − 1.28.The range in CCN activity of East Asian soils is most likely reflective of the compositional variability.Differences in CCN activity amongst samples collected in the same region likely reflect the chemical heterogeneity within the dust samples.We 5. CCN activation curves for different dust types presented in Table 1.Symbols show experimentally determined CCN activity and lines show FHH adsorption activation fits.Error bars represent measurement uncertainty in D dry .Also shown in black thick line is the κ = 0, Kelvin curve.Black dashed line corresponds to κ = 0.05.
found that the experimental data (points) can be described by FHH-AT fits (lines) well, with A FHH ∼ 2.25 ± 0.75 and B FHH ∼ 1.20±0.10for all dust types considered in this study.A direct comparison of CCN activity against data published in the literature is done by expressing our results (for particles of a given dry diameter) in terms of a hygroscopicity parameter, κ.CCN activity results for regional soils, and minerals and clays indicate a κ ≤ 0.05 for all samples considered in this study.It is also noted that the differences in values of the adsorption parameters determined in this study with those determined by Kumar et al. (2009b) for ATD, and likely arise from the application of multiple charge and dust nonsphericity corrections.Furthermore, the ATD experimental data used by Kumar et al. (2009b) in the predictions of adsorption parameters were taken from Koehler et al. (2009) that used a fluidized bed to generate aerosols, while in this study measurements were performed using a dry generation technique.
Figure 6 presents the CCN activity of all the minerals and clays considered.The activation diameters obtained for different dusts (Fig. 5) are within the range of those observed for different clays and minerals (Fig. 6).This suggests that dust CCN activity is controlled by adsorption of water onto the clay and mineral components in the dust samples.Comparison between CCN activities for different clays indicates montmorillonite (both Na and Ca rich) is more hydrophilic than illite, which agrees with the findings of Herich et al. (2009).Higher CCN activation potential for montmorillonite can be attributed to the mineralogy of the sample; the presence of unbounded Na and Ca cations allows water to penetrate the interlayer molecular space, which together with adsorption results in the clay swelling to several times its original volume.In the case of illite, the interlayer space Fig. 6.CCN activation curves for different mineral types presented in Table 1.Symbols (filled) are experimentally determined CCN activity, and lines represent FHH adsorption activation fits.Open symbols represent data obtained from Sullivan et al., 2009, andHerich et al., 2009.Color scheme of open symbols identical to CCN activity observed with measurements in this study.Error bars represent measurement uncertainty in D dry .Also shown in black thick line is the κ = 0, Kelvin curve.Black dashed line corresponds to κ = 0.05. is mainly occupied by poorly hydrated potassium cations that prevent these clay types from expanding, thus reducing the amount of water that can adsorb on the surface and its CCN activation potential.
The CCN activity of SiO 2 and CaCO 3 was also measured (Fig. 6).As expected, SiO 2 was the least CCN active of species considered with B FHH = 1.36 versus B FHH < 1.30 for the other clays and minerals (Table 2).This is because the majority of the silica surface does not interact strongly with water vapor since physisorption occurs primarily on the limited number of silanol sites (Young, 1958).We also find that the charged-corrected activation curves in our study differ from published CCN activation data for CaCO 3 (Sullivan et al., 2009), montmorillonite, andillite (Herich et al., 2009).For example, charge-corrected activation curves for clays (illite and Na-montmorillonite) exhibited κ = 0.02 − 0.04, versus 0.002-0.003 in Herich et al. (2009).In addition, (OMYA) CaCO 3 in this study was found to exhibit multiple κ values, 0.02 at s c = 0.4% −0.5% and 0.003-0.007at s c = 0.2%-0.3%,higher than found by Sullivan et al. (2009) for (Solvay) CaCO 3 (κ = 0.0011).Our results for (OMYA) CaCO 3 at low s c (κ = 0.003 − 0.007) compare well with results obtained for (Baker) CaCO 3 (κ = 0.008) (Sullivan et al., 2010).Similarly, we find a good comparison in CCN activity measurements based on κ values for regional dust samples considered in this study (for activation curves with and without multiple charging corrections) and past studies.For dry generated ATD (non-corrected), κ = 0.04 determined in this study compares well with non-corrected κ = 0.025 found by Koehler et al. (2009).Similarly, a good comparison for African dust samples was found with charge-corrected κ = 0.023 (Herich et al., 2009) and non-corrected κ = 0.054 (Koehler et al., 2009) determined for Saharan Dust, and charge-corrected and non-corrected κ = 0.02 − 0.04 for Niger dust data.
The differences cited above for CaCO 3 can be attributed to factors such as sample-to-sample variability (as confirmed by Sullivan et al., 2010) and method of aerosol generation.The soft saltation technique may yield very different particles from studies using a custom-built dry dust generator (Herich et al., 2009) or fluidized bed (Koehler et al., 2009).Furthermore, the lack of charge correction (in previous studies) will provide activation curves sensitive to the aerosol size distribution (as it determines the fraction of multiply charged particles with same mobility diameter), so differences in the dust size distribution will lead to variable biases in D dry .Unfortunately, absence of number size distributions of the CCN in the published studies precludes a conclusive attribution of these differences to multiple charging biases.
Table 2 shows the values of the experimental exponent, determined from the s c -D dry data for all dust samples and individual minerals/clays.The x exp values determined in this study are much lower than those reported by Kumar et al. (2009b) that were determined from the experimental data of Koehler et al. (2009) and Sullivan et al. (2009).This is a result of experimental measurements performed in this study at much lower supersaturations, as well as the application of multiple charge and shape factor corrections to the activation curves that tend to further shift D dry and x exp (as illustrated in Fig. 4).
In Fig. 7, x FHH is plotted against x exp for all dust samples and individual minerals/clays.For CaCO 3 , the value of x exp from the uncorrected s c -D dry data equals −0.81.The value of x exp after charge and shape correction reduces further to −0.75.As x exp for CaCO 3 is outside the range of exponents that can be predicted by FHH-AT, x FHH deviates from x exp by more than 10%.For Na-montmorillonite, Ca-montmorillonite, Soil 1 and Soil 3, x FHH is in excellent agreement with x exp suggesting that the above can be parameterized using FHH-AT.This suggests that the CCN activity of clays is consistent with multilayer adsorption activation theory.In the case of illite, SiO 2 , ATD, Niger, Soil 2, Soil 4, and Soil 5, x FHH lies within the variability of x exp , suggesting that FHH-AT also is an excellent description of CCN activity.Considering the uncertainty observed in experimental exponents (Fig. 7), it can be argued that the dust samples considered in this study are in excellent agreement with FHH-AT.Furthermore, x exp for all samples are found to be between −0.80 and −1.20 (range relevant for adsorption activation as given by Kumar et al., 2009a) as well as between −0.75 (determined for CaCO 3 ) and −0.93 (determined for Na-montmorillonite) providing support that nucleation of freshly generated regional dust aerosols is controlled by water vapor adsorption on clays and minerals.This confirms the conclusions of Kumar et al. (2009b)   is inappropriate, and, (ii) adsorption effects must be included when describing the hygroscopic and CCN behavior of mineral aerosol.
It can also be seen from the insert in Fig. 2, that for KT to predict the correct exponent determined from the experimental s c -D dry relationships on dust and clays (shown as shaded region), the values of κ must be very low (less than 0.0005), much lower than those determined in previous studies (Koehler et al., 2009;Herich et al., 2009;Sullivan et al., 2009).On the contrary, FHH-AT can predict experimental exponents obtained from dust and clays s c -D dry relation-ships (Table 2) using a single set of values for A FHH and B FHH .Furthermore, the predicted water vapor uptake under sub-saturated conditions is very low, and can explain the very low apparent hygroscopicity measured (using the hygroscopic tandem DMA technique) for dust aerosol (Herich et al., 2009).This strongly supports that FHH-AT describes fresh dust-CCN interactions better than KT for the samples considered in this study.
While the application of the shape factor corrections to CCN activation data changes the dry activation diameters considerably, it does so uniformly so that the exponent derived from the s c -D dry relationship (hence the implied activation physics) is not substantially affected.This can be seen from Table 2. Applying χ = 1.3±0.2,changes x exp by as little as 5% from charge corrected x exp .Using χ = 1.3±0.2 has a minor effect on dust hydrophilicity (indicated by a small range of B FHH ; Table 2).The omission of multiple-charging corrections to the activation curves, however, has a profound effect on the implied activation physics, as the dust appears significantly more CCN active than it really is.

Droplet growth kinetics
In addition to CCN activity, the optical particle counter of CFSTGC measures droplet sizes that can be used to explore CCN activation kinetics of mineral dust.This is carried out using TDGA, by comparing droplet diameter, D w , from the sample CCN against that of (NH 4 ) 2 SO 4 calibration aerosol with same critical supersaturation and maintaining identical instrument conditions (flow rates, pressure, and inlet temperature).If the droplet sizes from mineral aerosol are smaller than that observed from calibration aerosol (for conditions of identical instrument supersaturation, i.e., with same s c ), the activation kinetics of mineral dust is likely slower than the calibration aerosol.However, if activated droplet sizes are indistinguishable (to within experimental uncertainty) from (NH 4 ) 2 SO 4 droplet size, mineral dust exhibits the same activation kinetics as the reference aerosol.
Figure 8 presents the droplet diameters observed at OPC that are activated from regional dust aerosols as a function of instrument supersaturation.For comparison, droplet sizes are presented for pure (NH 4 ) 2 SO 4 aerosol with s c equal to the instrument supersaturation.It is evident that droplet growth for mineral aerosol at same s c is lower than that determined for (NH 4 ) 2 SO 4 calibration aerosol.The difference in outlet size suggests a delay in activation kinetics as both particles are exposed to the same supersaturation profile during their transit through the CFSTGC.This behavior is consistent with the slower time scales associated with water vapor adsorption (Kumar et al., 2009b).A similar behavior of reduced growth is also observed for different clays and minerals (Fig. 9).
Reduced growth at same s c observed for the mineral aerosol inside the CFTSGC can be attributed to three potential factors: (i) different shape of the equilibrium curve (FHH-AT vs. KT), (ii) different mass transfer coefficient (or α c ) of water vapor to the growing droplet, and (iii) dry particle size.To compare the effect of theory (KT or FHH-AT) used to describe equilibrium vapor pressure, we simulated droplet sizes at the exit of CFSTGC column for different α c and s c .Simulations suggest that the size of activated droplets at the exit of the growth column, originating from particles activating at same s c and with same α c are almost identical, suggesting that the activation theory has an almost negligible effect on the final droplet size (not shown).Simulations (not shown) indicate that dry CCN size has a negligible effect  on final droplet sizes.Based on the above, the droplet size difference between dust CCN and (NH 4 ) 2 SO 4 calibration aerosol is primarily driven by the intrinsic activation kinetics of the aerosol (which here is parameterized as difference in water vapor mass transfer coefficients (hence α c )).This is consistent with a slower timescale associated with adsorption of additional multi-layers of water vapor than absorption of water from deliquesced aerosol (Seinfeld and Pandis, 2006;Pruppacher and Klett, 1997 8 can be used to infer the water vapor uptake coefficient for dust by simulating dust CCN growth within the CFSTGC.Figure 10 shows values of α c determined for different regional dust aerosols relative to (NH 4 ) 2 SO 4 .Compared to (NH 4 ) 2 SO 4 calibration aerosol (that activates according to classical KT), mineral dust CCN grows to smaller droplet sizes that implies slower growth rate.When expressed in terms of α c , it corresponds to an average 50% reduction in α c .In absolute terms, if α c of water upon deliquesced (NH 4 ) 2 SO 4 aerosol is of order 0.2 (Davidovits et al., 2006), a 50% reduction would give α c of water upon dust ∼0.1.The α c tends to decrease as instrument supersaturation increases; at the highest supersaturation, the amount of water adsorbed at D c is much lower than for larger particles (low critical supersaturation).The kinetics of adsorption accelerates as the amount of adsorbed water increases (Pruppacher and Klett, 1997), so it is expected that α c would decrease with particle size.The literature value of (6.3 ± 0.7) × 10 −2 determined for the water vapor uptake coefficient on mineral dust (Seisel et al., 2005) is in agreement with the inferred α c from the highest critical supersaturation (α c ∼ 0.065).The diversity of inferred uptake coefficients could also be related to the chemical heterogeneity between samples.
Retarded activation kinetics may have an impact on the activated droplet number in clouds that contain significant concentrations of dust CCN.It is shown by Nenes et al. (2002) that a reduced α c affects the water uptake in the early stages of cloud formation (since droplets do not grow as rapidly); this leads to a higher parcel maximum supersaturation and hence a higher cloud droplet number.The extent of the impact depends on the vertical velocity, CCN concentration and the relative proportion of KT to FHH-AT particles.A thorough assessment will be the focus of a future study.

Summary
In this study, the CCN properties and droplet activation kinetics of aerosol generated from regional dust samples and individual minerals (clays, calcite, and quartz) were measured.The aerosols were generated dry in the lab, and properties were measured using the Scanning Mobility CCN Analysis (Moore et al., 2010).Including multiple charge corrections significantly increased D dry and decreased x exp .Dust nonsphericity was accounted for by converting from electrical mobility diameter, D m , to surface area equivalent diameter such that the surface area available for adsorption can be accounted for.Non-sphericity corrections were accounted for by using the dynamic shape factor, χ = 1.3±0.2 as this range covered published data for species considered in this study.It was found that while the application of the shape factor corrections to CCN activation data changes the dry activation diameters, it does so uniformly so that the magnitude of the exponent derived from the s c -D dry relationship (hence the implied activation physics) is not substantially affected with a deviation of as low as 5%.
The x exp for regional dust samples and mineral aerosols investigated in this study was found to be in excellent agreement with FHH-AT (mostly agreeing to within 10%) and one set of adsorption parameters (A FHH ∼2.25 ± 0.75, B FHH ∼ 1.20 ± 0.10).In contrast, KT cannot capture x exp without a hygroscopicity parameter that exhibits very strong sizedependence.This confirms the assessment of Kumar et al. (2009b) and further supports that FHH-AT provides more realistic representation of fresh dust CCN activity than KT.
Using threshold droplet growth analysis, dust CCN was found to have a reduced growth compared to (NH 4 ) 2 SO 4 calibration aerosol at the same instrument supersaturation.This implies slower activation kinetics of dust relative to (NH 4 ) 2 SO 4 aerosol.These delays in activation by dust CCN, when parameterized in terms of the water vapor uptake coefficient, α c , translates to a 30-80% (average = 50%) reduction in α c (relative to the (NH 4 ) 2 SO 4 aerosol).
The samples studied here are representative of major regional dust sources, and the adsorption activation parameters determined can be used to express their CCN potential in cloud droplet formation parameterizations developed by Kumar et al. (2009a).These parameterizations are valid for fresh dust in the dust source regions and for transported dust if it will not undergo significant atmospheric processing.A combined KT and FHH-AT framework, however, may be needed to accurately describe the CCN activity of aged dust, dry lakebed dust mixed with salts (e.g., Owens Lake, Texcoco, and Aral Sea), and more generally dust particles with significant amounts of soluble materials.
A major implication of this study is that freshly-emitted dust and mineral aerosols can act as CCN through the effects of water adsorption alone.In some cases, 100 nm dust particles can exhibit comparable hygroscopicity to an organic species with κ ∼ 0.05 or a particle with (NH 4 ) 2 SO 4 volume fraction of 10%.Dust particles in the Giant CCN (GCCN) size range will exhibit much lower apparent hygroscopicity because of their lower surface-to-volume ratio.Whether the effects of adsorption is sufficient to make freshly emitted dust GCCN act as a good collector drop is an open question left for a future study.Nevertheless, this study reshapes the conceptual notion of dust CCN activity to one where freshly emitted insoluble dust particles can have an appreciable hygroscopicity (that depends on their surface-to-volume ratio) which can be augmented through atmospheric processing.

Fig. 1 .
Fig. 1.Schematic of the experimental set-up used for size resolved CCN activation and droplet growth kinetics measurements.

Fig. 2 .
Fig. 2. s c -D dry lines for different values of B FHH computed at σ = 0.072 J m −2 , T = 298.15K and A FHH = 2.50.Dashed lines indicate κ isolines determined at above conditions.Also shown in black thick line is the κ = 0, Kelvin curve.The inset figure shows experimental exponent as function of B FHH and κ.

Fig. 3 .
Fig. 3. Activation curves for Soil 2 at s c = 0.3%.Shown are inversions without (blue) and with multiple charge corrections (brown).Error bars represent uncertainty of activation efficiency as a result of counting efficiency and flow rate uncertainty at different diameters.
D se is determined by converting the electrical mobility diameter (D m ), to particle volume equivalent diameter (D ve ), and from there to D se .Determining D ve from D m D ve is determined from D m by iterative solution of the dynamic shape factor equation (DeCarlo et al., 2004): χ = D m C(D ve ) D ve C(D m ) (5) where C(D m ) and C(D ve ) are the slip correction factors for D m and D ve , respectively.C(D m ) and C(D ve ) can be approximated from the correlation of Willeke and Baron, (2001):

Fig. 4 .
Fig. 4. CCN activation curves (s c -D dry ) for ATD (χ = 1.3 ± 0.2)showing the effect of including charge and shape corrections on the raw data.Blue shows curve with no correction, brown shows the results with charging corrections and green shows curve after including both charge and shape corrections.Error bars represent experimental uncertainty and numerical uncertainty in D dry at same instrument supersaturation.

Fig. 8 .
Fig. 8. Activated droplet sizes of mineral dust CCN with s c equal to the instrument supersaturation shown as symbols.Error bars represent experimental uncertainty in droplet size as observed by the OPC at same instrument supersaturation.

Fig. 9 .
Fig. 9. Activated droplet sizes of different minerals and clay CCN with s c equal to the instrument supersaturation shown as symbols.Error bars represent experimental uncertainty in droplet size as observed by the OPC at same instrument supersaturation.

Fig. 10 .
Fig. 10.Inferred water vapor uptake coefficients for the growth kinetics data of Fig. 8 normalized to that of (NH 4 ) 2 SO 4 calibration aerosol as a function of different instrument supersaturation.Error bars represent experimental uncertainty in determination of water vapor uptake coefficients arising due to differences in droplet sizes measured by the OPC.

Table 1 .
Summary of Regional Dust Samples and Clays/Minerals analyzed in this study.

Table 2 .
FHH parameters and exponent comparisons for different regional dusts and individual clays/minerals.Values in parentheses indicate change in magnitude of x exp from change in χ between 1.1 and 1.5.Comparison of x exp and x FHH for dust and clay/mineral types presented in Table1.Dashed line represents +10% deviation from the 1:1 line.Error bars represent deviation in x exp due to the uncertainty in D dry .
Diameter of the sphere whose projected area is equal to that of the particle normal to the direction of flow