Changes in the ice nucleation properties of mineral dust particles due to soluble coatings are still not well understood. Here we show that the reactivity with soluble materials deposited on the surfaces of kaolinite particles is an important factor affecting the ice nucleation properties of the particles. Using kaolinite particles treated with levoglucosan or H2SO4(i.e., non-reactive and reactive materials, respectively), we investigated the fraction of particles capable of nucleating ice at temperatures ranging from −34°C to −26°C. Below water saturation, both the levoglucosan and H2SO4 coatings similarly reduced the ice nucleating ability of kaolinite particles. Above water saturation, however, only the H2SO4coatings reduced the ice nucleating ability of the particles, particularly at warmer temperatures. We suggest that the absence or presence of surface chemical reactions plays an important role in determining the number concentrations of ice crystals formed from mineral dust ice nuclei under mixed-phase cloud conditions.
 Mineral dust particles are known as one of the sources of the most effective and abundant atmospheric ice nuclei (IN) that initiate ice nucleation at temperatures warmer than about −36°C [e.g., Isono et al., 1959; DeMott et al., 2003]. Two major possible pathways for heterogeneous ice nucleation involving mineral dust particles are deposition ice nucleation (ice formation on the surfaces of insoluble nuclei from ice-supersaturated water vapor) and immersion/condensation freezing (ice formation during or following the condensational growth of aqueous droplets containing insoluble nuclei). Typically, the former mechanism begins to be activated at ice-supersaturated conditions below water saturation, and the latter ensues above water saturation, although it may occur in solution droplets at sufficiently low temperatures to overcome the freezing point depression effects of solutes.
 Mineral dust particles can be coated with soluble materials through coagulation and heterogeneous reactions as they are processed in the atmosphere [e.g., Sullivan et al., 2007; Tobo et al., 2010], and their ice nucleation properties can be modified depending on the conditions of the particle surfaces. Laboratory studies have shown that a wide variety of soluble materials can alter the ice nucleating ability of mineral dust particles below water saturation [e.g., Möhler et al., 2008; Eastwood et al., 2009; Sullivan et al., 2010a]. On the other hand, the impacts of such materials on the immersion/condensation freezing processes above water saturation may be more complicated. For example, recent experiments at −30°C with different preparations of Arizona test dust (ATD) particles have suggested that the uptake of HNO3 has no apparent impact on their ice nucleating ability in the immersion/condensation freezing regime [Sullivan et al., 2010a], while the uptake of H2SO4 can reduce their ice nucleating ability [Sullivan et al., 2010b; Niedermeier et al., 2011; Reitz et al., 2011].
 In this study, we used two highly soluble materials (namely, levoglucosan and H2SO4) to investigate the effects of these coatings on the ice nucleation properties of kaolinite particles at temperatures ranging from −34°C to −26°C, both below and above water saturation. Kaolinite consists of aluminosilicate clays and is a major component of mineral dust particles [Usher et al., 2003]. Levoglucosan is known as a tracer for biomass burning and is derived from combustion of cellulose [Simoneit, 2002]. Atmospheric SO2 emitted from natural/anthropogenic sources can be oxidized to SO42− on particle surfaces during atmospheric processing [Usher et al., 2003]. Both levoglucosan and H2SO4 form aqueous solutions even at a few percent relative humidity [e.g., Mochida and Kawamura, 2004; Eastwood et al., 2009]. Regarding ice nucleation behaviors of kaolinite particles coated with aqueous solutions, our expectation was that deposition ice nucleation would be shut off in the regime below water saturation, but only H2SO4 could additionally react with kaolinite particles [e.g., Panda et al., 2010] to potentially influence freezing of soluble coatings and activated cloud droplets. The levoglucosan and H2SO4coatings used here serve as surrogates for non-reactive and reactive coatings that have impacts on initial ice nucleation involving mineral dust particles in the atmosphere. We do not consider additional physical and chemical impacts on ice nucleation that may exist during actual cloud processing.
 All the experiments were performed at the Leipzig Aerosol Cloud Interaction Simulator (LACIS) facility, and the experimental procedures closely followed those used during the FROST 1 and 2 campaigns [Niedermeier et al., 2010, 2011; Sullivan et al., 2010b]. Kaolinite particles (Fluka) were generated in a fluidized bed generator and then passed through a self-built corona discharger and a85Kr neutralizer. Subsequently, the particles passed through a heated vapor diffusion tube containing a reservoir of levoglucosan or H2SO4. The tube was surrounded by a thermostatically-controlled water jacket. The amounts of soluble materials deposited on the particle surfaces were controlled by the temperature of the water jacket. In this study, we set the temperature to 93°C for levoglucosan and to 70°C for H2SO4. According to soluble mass estimates based on the measured aerosol hygroscopicity data (not reported here) and previously described methods [Stratmann et al., 2010; Sullivan et al., 2010b], it is expected that most of individual kaolinite particles after passing through the tube took up an amount of levoglucosan or H2SO4 enough to form several monolayers on the particle surfaces (H. Wex et al., manuscript in preparation, 2012). Note that these estimates are based on the assumption that the particles are spherical and that the soluble materials are deposited on the particle surfaces uniformly. Water uptake is certain to aid coating by aqueous solutions during the freezing experiments. Particles with an electrostatic mobility diameter of 300 nm or 700 nm were selected in a differential mobility analyzer for the use of various experiments.
 The IN number concentrations at controlled temperatures and humidities were measured using a continuous flow diffusion chamber (CFDC) [Rogers et al., 2001; DeMott et al., 2010]. The CFDC uses a temperature gradient to produce well-defined ice-supersaturated conditions in a focused (by sheath flow) and flowing aerosol lamina between two ice-covered walls. The CFDC has a particle nucleation/growth section, followed by a droplet evaporation section where only ice saturation exists (the wall temperature gradient is almost zero). Since cloud droplets cannot survive through the evaporation section unless the relative humidity with respect to liquid water (RHw) in the particle nucleation/growth section exceeds ∼108% [e.g., Sullivan et al., 2010a, 2010b], only particles that nucleate ice and grow to super-micron sizes are counted as IN with an optical particle counter at the outlet. In this study, the IN number concentrations at temperatures of −34°C, −30°C and −26°C were determined as a function of RHw by scanning the selected experimental conditions from ice saturation to about 107%RHw. Simultaneously, the total particle number concentrations were also measured using a condensation particle counter to determine the fraction of particles serving as IN (i.e., IN fraction = the ratio of the IN number concentrations to the total particle number concentrations). Typical background IN counts were <1.0 L−1. Given typical total particle number concentrations of 10–100 cm−3, the background IN counts lead to IN fractions of <10−4–10−5. In this study, we regard the IN fraction of 10−4 as the quantification limit of our experiments.
3.1. Untreated Kaolinite Particles
 The changes in the IN fractions of 300 nm pure kaolinite particles as a function of RHw measured at −34°C, −30°C and −26°C are shown in Figure 1a. The data show that the IN fractions tended to increase as the RHw increased and as the temperatures decreased. The data also show that ice nucleation on the untreated particles was already detectable below water saturation. For example, the onset conditions (we define here as the relative humidities at which IN fractions exceeded 10−4) were ∼84%RHw, ∼88%RHw and ∼96%RHw at temperatures of −34°C, −30°C and −26°C, respectively. Thus, our results demonstrate that untreated kaolinite particles can act as apparent deposition IN below water saturation, especially at colder temperatures, corresponding to the results reported previously [e.g., Eastwood et al., 2009].
3.2. Kaolinite Particles Treated With Levoglucosan
 The results for 300 nm kaolinite particles treated with levoglucosan are shown in Figure 1b. The data show that ice nucleation on the treated particles was hardly detected below water saturation at temperatures ranging from −34°C to −26°C. This result indicates that kaolinite particles lost their ability to act as deposition ice nuclei after the treatment with levoglucosan. Furthermore, Figure 1b shows that the IN fractions of the treated particles increased rapidly just below water saturation (∼96–98%RHw), and that the values above water saturation were comparable to those of the untreated particles (see also Figure 1a and Table S1 in the auxiliary material). This finding suggests that the ice nucleating ability of the original kaolinite particles can be maintained via the immersion/condensation freezing mechanism after the levoglucosan coatings.
3.3. Kaolinite Particles Treated With H2SO4
 The results for 300 nm kaolinite particles treated with H2SO4 are shown in Figure 1c. After the treatment with H2SO4, ice nucleation was hardly detected below water saturation, similar to the results after the treatment with levoglucosan (Figure 1b). However, Figure 1c shows that the IN fractions of the particles treated with H2SO4 were lowered even above water saturation, particularly at temperatures warmer than −34°C. For example, after the treatment with H2SO4, the IN fractions of kaolinite particles above water saturation (at 104%RHw) at −34°C decreased from about 0.14 to 0.079 at 104%RHw, while those at −26°C dropped from about 3.5 × 10−3 to below quantification limit (see also Figure 1a and Table S1). These results suggest that the H2SO4 coatings caused a large loss of the ice nucleating ability of kaolinite particles in the immersion/condensation freezing regime, especially at warmer temperatures.
3.4. Effects of Particle Size
 In general, larger mineral dust particles are more efficient IN than smaller ones, and require lower RHw to activate [Welti et al., 2009]. We investigated the changes in the IN fractions of 700 nm kaolinite particles before and after the treatment with levoglucosan and H2SO4. The results indicate that both the levoglucosan and H2SO4 coatings reduced the ice nucleating ability of 700 nm kaolinite particles below water saturation, but only the H2SO4 coatings efficiently altered their ice nucleating ability above water saturation, particularly at temperatures warmer than −34°C (see Figure 2). This tendency corresponds to that found for the 300 nm particles, but the IN fractions of the 700 nm particles tended to be higher than those of the 300 nm particles even after the treatment with levoglucosan or H2SO4 (see also Figure 1 and Table S1).
 The present results demonstrate that the impacts of the levoglucosan and H2SO4 coatings on the ice nucleation properties of kaolinite particles are very different depending on the experimental conditions. In the ice nucleation regime below water saturation, both the levoglucosan and H2SO4 coatings similarly reduced the ice nucleating ability of kaolinite particles at temperatures ranging from −34°C to −26°C. This result suggests that the formation of these soluble coatings inhibits deposition ice nucleation on solid particle surfaces, despite the uncertainty in the uniformity of soluble materials deposited. Heterogeneous freezing of the coatings is then possible below water saturation, in dependence on water activity of the solutions, but can occur only for a more restricted RHw range close to 100% at temperatures warmer than −34°C [e.g., Zuberi et al., 2002]. A more detailed discussion concerning the impacts of the hygroscopic behaviors and thicknesses of soluble coatings on ice nucleation will be performed in an accompanying paper (H. Wex et al., manuscript in preparation, 2012). In the immersion/condensation freezing regime of very dilute solutes (above water saturation), only the H2SO4 coatings reduced the ice nucleating ability of kaolinite particles, while the levoglucosan coatings had little or no impact on the ice nucleating ability of the particles. The reduction of the ice nucleating ability in the immersion/condensation freezing regime induced by the uptake of H2SO4 tended to be stronger for the experiments at temperatures warmer than −34°C.
 We consider that the major difference between levoglucosan and H2SO4 is the reactivity with kaolinite. Considering that levoglucosan cannot react with kaolinite effectively, the morphology and composition of kaolinite particle surfaces may not be modified even after the treatment with levoglucosan and hence the ice nucleating ability of kaolinite particles in the immersion/condensation freezing regime may still be maintained. In contrast, H2SO4 can react with kaolinite. For example, a possible reaction between kaolinite and H2SO4 can be expressed as follows [e.g., Panda et al., 2010]:
During this reaction, the fraction of SiO2 in the parent kaolinite clay increases since Al2(SO4)3 can easily be dissolved in water. We speculate that the morphology and composition of kaolinite particle surfaces can be modified as a result of chemical reactions with H2SO4, as for reaction (1). One of the reasons why kaolinite particles are good IN is thought to be related to the structure of water adsorbed on kaolinite surfaces with trench-like defects [Croteau et al., 2010]. Destruction of ordered water structures at the water/mineral interfaces caused by high concentrations of H2SO4 may also act to reduce the ice nucleating ability of kaolinite particles, although this mechanism is assumed to be activated mainly in the ice nucleation regime below water saturation [Yang et al., 2011]. Further work will be necessary to understand how chemical reactions affect the ice nucleation properties of mineral dust particles in the immersion/condensation freezing regime at different temperatures.
 Recent work showing that the uptake of HNO3 does not impair the ice nucleating ability of ATD particles in the immersion/condensation freezing regime suggested that the ice nucleation properties of mineral dust particles in this regime can change depending on the mineralogical compositions and the uptake mechanisms of soluble materials [Sullivan et al., 2010a]. For example, although both HNO3 and H2SO4 can react with alkaline carbonates (e.g., CaCO3), the products of the reactions may have different solubilities (e.g., Ca(NO3)2 and CaSO4 are highly and slightly soluble salts, respectively [Sullivan et al., 2009]). Therefore, such reaction products may cause the observed difference in the ice nucleation properties of ATD particles above water saturation [Sullivan et al., 2010a]. However, it should be noted here that carbonates are likely to be a minor component and hence distributed between individual ATD particles heterogeneously [Vlasenko et al., 2005]. Also, given that HNO3 cannot effectively react with aluminosilicates that are a major component of ATD, only a small impact of the reactions with HNO3 on the ice nucleation properties of ATD particles in the immersion/condensation freezing regime may be expected [Sullivan et al., 2010a]. On the other hand, the uptake of H2SO4 can effectively reduce the ice nucleating ability of ATD particles in the immersion/condensation freezing regime, particularly at temperatures warmer than −35°C [Niedermeier et al., 2011]. In this regard, the present experiments with kaolinite particles support the idea that the reduction of the ice nucleating ability of ATD particles in this regime is mainly due to the reactions between aluminosilicates and H2SO4, such as reaction (1).
 The IN number concentrations measured above water saturation are assumed to reflect the maximum IN number concentrations spontaneously active in mixed-phase clouds at temperatures warmer than about −36°C [DeMott et al., 2010]. Considering that desert dust storms are occasionally mixed with biomass burning smokes during their long-range transport [e.g.,Ansmann et al., 2011], the mixing of levoglucosan and/or other non-reactive soluble organics with mineral dust particles is assumed to occur in mixed dust/smoke plumes. Our results suggest that the ice nucleating ability of mineral dust particles (i.e., the number concentrations of mineral dust IN) can be maintained under mixed-phase cloud conditions even if they are coated with such non-reactive compounds. On the other hand, the mixing processes of desert dust storms with natural and anthropogenic emissions containing reactive compounds (e.g., H2SO4or its precursors) may play an important role in reducing the number concentrations of mineral dust IN under mixed-phase cloud conditions. Further work will be needed to incorporate such considerations into atmospheric models and to validate such effects occurring in the atmosphere.
 We thank Anthony J. Prenni and Gavin R. McMeeking for their roles in preparing the instrumentation, and Paul Herenz and Sandra Kanter for assistance in aerosol sample preparation. This work was partially funded by the NSF (ATM-0841602). Y. Tobo and P. J. DeMott acknowledge additional financial support for this work from the EUROCHAMP II program. Y. Tobo also acknowledges the JSPS Postdoctoral Fellowships for Research Abroad.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.