Heterogeneous nucleation of ice on surrogates of mineral dust

Authors


Abstract

[1] Field studies have shown that mineral dust particles can act as ice nuclei in cirrus clouds. Here, we present a laboratory investigation of heterogeneous ice nucleation on surrogates of mineral dust particles, in particular pure Arizona test dust (ATD) particles, and ATD particles coated with sulfuric acid. The experiments have been performed using a new apparatus in which ice formation on the particles is determined by optical microscopy at temperatures between 197 and 260 K and relative humidities up to water saturation. The experiments reveal that pure and sulfuric acid coated ATD particles nucleate ice at considerably lower relative humidities than required for homogeneous ice nucleation in liquid aerosols. Nucleation occurred over a broad relative humidity range indicating that the different minerals contained in ATD have different ice nucleation thresholds. No significant difference in the ice nucleation ability of pure and H2SO4 coated ATD particles was observed. Below 240 K, ice nucleated on ATD particles apparently by deposition nucleation. Preactivation of ATD particles, that is, a reduction in supersaturation, required for heterogeneous ice nucleation after a previous ice nucleation event on the same particle, has been observed for temperatures as low as 200 K. Differences of 10–30% in the onset RHice values were obtained for particles with or without preactivation. The results indicate that pure and sulfuric acid coated mineral dust particles may act as efficient ice nuclei in the atmosphere. Preactivation of the particles should be considered when modeling long-range transport of mineral dust particles and their impact on cloud formation.

1. Introduction

[2] Atmospheric aerosol particles scatter solar and terrestrial radiation, thereby influencing the global radiation budget [Ramaswamy, 2001]. In addition, aerosol particles modify the radiative properties of clouds by acting as cloud condensation nuclei (CCN) and ice nuclei (IN) [Penner, 2001; Twomey, 1974, 1991; Albrecht, 1989; Pincus and Baker, 1994; Baker, 1997; Rosenfeld, 2000; Ramanathan et al., 2001; Pruppacher and Klett, 1997]. Global circulation model studies have indicated that ice clouds triggered by preexisting aerosol particles may contribute significantly to the global radiation budget [Senior and Mitchell, 1993; Fowler and Randall, 1996; Penner, 2001]. However, the impact of cirrus clouds on the global radiation budget is still not assessed in a satisfactory manner [Penner, 2001].

[3] Freezing of ice in aerosol particles can occur by homogeneous or heterogeneous nucleation [Pruppacher and Klett, 1997]. In aqueous solution droplets, such as aqueous H2SO4 or (NH4)2SO4 droplets, homogeneous ice nucleation sets in at supersaturations with respect to ice of about 45–75% depending on temperature [Koop et al., 2000b]. However, the presence of IN may decrease the energy barrier for ice nucleation, thus allowing the nucleation of ice at lower supersaturations when compared to homogeneous nucleation. Heterogeneous ice nucleation may occur by various mechanisms such as deposition nucleation, immersion freezing, or contact freezing [Vali, 1985; Pruppacher and Klett, 1997]. Deposition nucleation describes the formation of ice from a supersaturated gas phase environment on the surface of a solid nucleus. Immersion freezing represents the formation of ice at the surface of a nucleus suspended in a supercooled liquid. Contact nucleation describes the formation of ice at the moment of contact between an insoluble particle and a supercooled liquid droplet. All of the above-mentioned nucleation mechanisms may occur in the atmosphere, however, the significance of each process with respect to cirrus cloud formation is not fully evaluated yet.

[4] Heterogeneous nucleation of ice is regarded as a possible formation mechanism of cirrus ice clouds [Heymsfield et al., 1998; DeMott et al., 1998; Seifert et al., 2003; Ström et al., 2003; Haag et al., 2003]. It has been shown that mineral dust particles have a significant impact on cloud formation, cloud properties, and precipitation [Levin et al., 1996; Rosenfeld et al., 2001; Yin et al., 2002; DeMott et al., 2003; Sassen et al., 2003; Toon, 2003; Mahowald and Kiehl, 2003; Sassen, 2005]. Recent field measurements indicate that Saharan mineral dust particles can act as efficient IN far from the source region [DeMott et al., 2003; Sassen et al., 2003; Toon, 2003; Sassen, 2005]. In the course of Saharan dust storms, mineral dust particles smaller than 1 μm have been observed to act as IN at concentrations exceeding 1 particle per cm3 [DeMott et al., 2003; Sassen et al., 2003]. Several studies have shown that mineral dust is often internally mixed with H2SO4 which originates from sources such as soil, oxidation of the mineral by SO2, and scavenging of acidic solution droplets [Levin et al., 1996; Ganor et al., 1998; Falkovich et al., 2001; Kojima et al., 2004, 2005; Desboeufs and Cautenet, 2005]. Therefore it is conceivable that the concomitance of H2SO4 and mineral dust may significantly alter the CCN and IN properties of the particles.

[5] Several laboratory studies have shown that mineral dust particles efficiently nucleate ice in the deposition mode or the immersion mode [Pruppacher and Sänger, 1955; Mossop, 1956; Mason and Maybank, 1958; Mason, 1960; Hallett, 1961; Roberts and Hallett, 1968; Anderson and Hallett, 1976; Schaller and Fukuta, 1979; Bailey and Hallett, 2002; Zuberi et al., 2002; Hung et al., 2003; Archuleta et al., 2005; Möhler et al., 2005]. However, there are only limited studies which investigated formation of ice by deposition nucleation at upper tropospheric temperatures [Bailey and Hallett, 2002; Archuleta et al., 2005; Möhler et al., 2005].

[6] In addition, the nucleation ability of IN may depend on the previously experienced atmospheric relative humidities and temperatures. For example, Fournier D'Albe [1949] observed that cadmium iodide particles became more effective IN after they had been involved in ice formation previously. In other words, a particle that nucleated ice once will nucleate ice subsequently at lower supersaturation when compared to the initial nucleation event. This effect is called preactivation or memory effect. Preactivation has been observed to occur for different mineral and clay mineral particles, however, these studies were mostly qualitative in nature or limited to temperatures above 240 K [Mossop, 1956; Mason and Maybank, 1958; Serpolay, 1959; Mason, 1960; Higuchi and Fukuta, 1966; Roberts and Hallett, 1968].

[7] Here, we present a new experimental setup to investigate heterogeneous ice nucleation. Arizona test dust (ATD) particles composed of various mineral species served as surrogates for natural mineral dust particles in the atmosphere. The formation of ice on these particles was studied for temperatures between 197 and 260 K at relative humidities with respect to ice between ice saturation and water saturation. In contrast to most previous studies the experiment can be performed in a way that allows the observation of repeated ice nucleation of the same particles, thus, changes in the nucleation ability due to preactivation can be studied. Additionally, the effect of a sulfuric acid coating of the ATD particles on ice nucleation was investigated.

2. Experimental Procedure

2.1. Setup

[8] Figure 1 shows a sketch of the experimental setup. The particles are placed on a silanized Herasil quartz plate. The silanization provides a hydrophobic surface by forming a monolayer of silanes that are attached to the SiO2 surface through covalent bonds. The observed sample area is about 1.5 mm in diameter and its temperature is controlled within a range of 197 to 293 K by the upper temperature stage. The temperature gradient over the sample area is smaller than 0.1 K mm−1 and was determined by measuring the melting points of organic substances at different positions on the sample area [Knopf et al., 2002]. The lower-temperature stage, which controls the temperature of a water reservoir, is situated opposite to the sample area. This water reservoir, about 36 mm in diameter, is established by introducing water vapor through an inlet and condensing it at low temperatures. Two capacitive pressure gauges measure the absolute pressure in the chamber within a range of 10−5 to 100 mbar. The vacuum cell shown in Figure 1 is connected to a turbo-molecular pump (TMP) and attached to a confocal Raman microscope.

Figure 1.

Sketch of the experimental vacuum system. “TMP” indicates the connection to a turbo-molecular-pump and “baratron” indicates the outlet to a pressure gauge.

[9] Variation of the temperature of the lower stage changed the water reservoir temperature, Tlow, and thus its water vapor pressure. This, in turn, changed the water partial pressure inside the vacuum cell, image which was monitored with the pressure gauges. (No buffer gas was present, i.e., water vapor was the only gas species inside the cell.) Here image is determined by the vapor pressure of the water reservoir according to the following relationships:

equation image

and

equation image

where image and image are the water vapor pressures of liquid water and ice, respectively [see, e.g., Murphy and Koop 2005]. The relative humidity with respect to liquid water, RH, or with respect to ice, RHice, experienced by the particles on the sample area inside the cell can be obtained from the ratio of the water partial pressure inside the cell and the water vapor pressures of pure water or ice as appropriate:

equation image
equation image

where Tup represents the sample temperature of the upper stage.

[10] Ice nucleation on the particles can occur at RHice values larger than 100%. This can be achieved by providing a sufficient amount of gas-phase water and by avoiding mass transfer of water between ice and nonactivated particles on the sample area. At water reservoir temperatures below 235 K, the experiment is operated in the Knudsen flow regime (i.e., the Knudsen number is larger than 0.01) [Wutz, 1989]. In this operation mode, water partial pressure gradients are assumed to be negligible since the gas flow changes from viscous to molecular flow due to an increase of the molecular mean free path of the water molecules. At higher temperatures (and thus larger absolute pressures) this assumption is no longer valid. Hence results which have been obtained at temperatures higher than 235 K represent the first ice nucleation events that were detected and ice nucleation of other particles thereafter were not used in the analysis due to a possible bias toward higher RHice values.

[11] Phase changes of the investigated particles and ice formation were observed by optical microscopy [Koop et al., 1998] or confocal Raman spectroscopy [Knopf et al., 2002, 2003]. The detailed specification of the confocal Raman microscope is described elsewhere [Knopf et al., 2002, 2003].

2.2. Particle Preparation

[12] Aqueous solutions were prepared from LiCl (99.99%), (NH4)2SO4 (≥99.99%), NaCl (≥99.5%), aqueous 95–98 wt% H2SO4 (all purchased from Aldrich) and Millipore water (resistivity ≥ 18.2 MΩ · cm). Droplets of these solutions were deposited on the silanized quartz plate employing an inkjet cartridge. The inkjet cartridge generates aqueous droplets with diameters of 50–70 μm [Knopf, 2003] which after efflorescence of the particles leads to crystals with diameters of 1–20 μm.

[13] Commercially available ultrafine grade Arizona test dust (ATD) was used in the experiments (Powder Technology Inc.). ATD represents a standardized naturally occurring dust type. Its main components are given in Table 1 and were corroborated by a recent study of Vlasenko et al. [2005]. The elemental composition of the ATD particles closely represents the typical composition of Saharan dust particles [Krueger et al., 2004, 2005] and the composition found within the upper crust of the Earth [Wedepohl, 1995]. This makes ATD a favorable surrogate for atmospheric mineral dust particles in laboratory studies. It should be noted that ATD does not represent a homogeneous mineral dust sample as it consists of a variety of different crystallites such as quartz, illite, kaolinite, and montmorillonite commonly found in dust samples [Glacuum and Prospero, 1980; Caquineau et al., 2002; Reid et al., 2003; Möhler et al., 2005].

Table 1. Composition of Ultrafine Grade Arizona Test Dust According to the Distributor (Powder Technology Inc.)
ComponentPercentage By Weight
SiO268–76
Al2O310–15
Fe2O32–5
Na2O2–4
CaO2–5
K2O2–5
MgO1–2
TiO20.5–1.0
Loss on ignition2–5

[14] In our experiments the dust samples consisted of particles approximately 0.7–10 μm in diameter, with a mean value of about 4.2 μm. Here 5 mg of ATD were suspended in 50 mL H2O. Droplets consisting of this suspension were deposited on the sample area using the inkjet cartridge. The water evaporated instantaneously as the particles were exposed to vacuum leaving behind pure ATD particles. About 20 particles were observed simultaneously during an experiment. Ice nucleation on particles in the size range of 1–10 μm have been considered in the data analysis. This size range corresponds to particle diameters typically observed in atmospheric dust aerosols [Glacuum and Prospero, 1980; Gomes and Gillette, 1993; Falkovich et al., 2001; Reid et al., 2003; Shi et al., 2005].

[15] We have also suspended ATD in dilute H2SO4/H2O solutions to study the effect of a sulfuric acid coating of the ATD particles on ice nucleation. In this case 2 mg ATD were suspended in a H2SO4/H2O solution 5 × 10−3 wt% in concentration. The droplets were deposited on the sample area using the inkjet cartridge. We estimated that the particles are coated by about 67–1100 H2SO4 monolayers assuming a surface density of 1014 molecule cm−2. This estimate is based on BET surface areas commonly found for oxide particles and dust samples [Börensen et al., 2000; Underwood et al., 2001; Hanisch and Crowley, 2003; Adams et al., 2005].

2.3. Calibration of the Setup

[16] The temperature of the upper and lower temperature stages were calibrated prior to the experiments. The temperatures of both stages were measured with identical resistance temperature sensors (Pt 100). The linear response of the sensor attached to the lower-temperature stage was confirmed by measuring the absolute water vapor pressure over ice and water in a temperature range between 200 and 300 K. The experimentally obtained vapor pressures were in agreement with water vapor pressures from the literature [Goff, 1957; Marti and Mauersberger, 1993] within the experimental uncertainty of about 5%.

[17] The linear response of the resistance temperature sensor attached to the upper temperature stage was confirmed by measuring the melting points of ice at 273.16 K, the deliquescence relative humidity (DRH) of NaCl at 255 K, and the ferroelectric phase transition of solid (NH4)2SO4 at 223.1 K.

[18] First, the melting point of small ice particles was observed visually as the temperature of the upper stage was increased continuously. Second, the deliquescence of NaCl particles was performed in the following way. After the NaCl particles had been deposited on the sample area, the temperature of the upper stage was held at a constant temperature. Then, the water partial pressure in the cell, image was increased by raising the temperature of the lower stage until deliquescence of the NaCl particles occurred. At this point, the water vapor pressure over pure liquid water, image is defined by

equation image

Using image from the experiment and the temperature dependence of the DRH of NaCl known from the literature [Tang et al., 1977; Cziczo et al., 1997; Weis and Ewing, 1999; Cziczo and Abbatt, 2000; Koop et al., 2000a] the temperature of the upper stage Tup was obtained by iteration. Third, the ferroelectric phase transition of solid (NH4)2SO4 [Hoshino et al., 1958] was observed using Raman spectroscopy. This structural phase transition is associated with a change in space group from D2h16/Pnam in the paraelectric phase at T > 223.1 K to C2v9/Pna21 in the ferroelectric phase at T < 223.1 K. A larger (NH4)2SO4 solution droplet 0.4 mm in diameter was placed on the sample area using a micropipette. Evaporation of the water led to a solid (NH4)2SO4 crystal. The detection of the ferroelectric phase transition was then performed by recording Raman spectra as a function of temperature in steps of 0.1 K until the phase transition occurred. Table 2 shows the experimentally derived vibration bands for the paraelectric and ferroelectric phases. The vibration bands identified in this study are in agreement with the vibration bands determined by Torrie et al. [1972] also given in Table 2.

Table 2. Raman Shifts of (NH4)2SO4 Crystals in the Paraelectric and Ferroelectric Phase
 Vibration ModePeak Maximum, cm−1
This WorkLiteraturea
Paraelectric phase (T > 223.1 K)ν1(a1)972 ± 4977 ± 20
ν3(f2)1061 ± 41062 ± 10
ν3(f2)1090 ± 41106 ± 10
Ferroelectric phase (T < 223.1 K)ν1(a1)970 ± 4972 ± 20
ν3(f2)1053 ± 41043 ± 10
ν3(f2)1083 ± 41077 ± 10
ν3(f2)1115 ± 41124 ± 10
ν3(f2)1138 ± 41147 ± 10

[19] Once both temperature stages were calibrated, deliquescence experiments of (NH4)2SO4 and LiCl crystals were performed to control the quality of the calibration over a broader temperature and RH range. Table 3 shows the agreement of the experimentally derived DRH values and the corresponding literature DRH values of (NH4)2SO4 and LiCl crystals [Tang and Munkelwitz, 1993; Xu et al., 1998; Cziczo and Abbatt, 1999; Onasch et al., 1999; Greenspan, 1977] for different particle temperatures. These experiments indicate that the apparatus is also suitable to study deliquescence of aerosol particles.

Table 3. Deliquescence Relative Humidity (DRH) of (NH4)2SO4 and LiCl
T, K(NH4)2SO4 DRH, %LiCl DRH, %
This WorkLiteratureaThis WorkLiteratureb
275.480.5 ± 1.781.2 ± 2  
269.583.4 ± 1.481.6 ± 2  
255.182.5 ± 2.682.6 ± 2  
272.3  11.0 ± 0.411.2 ± 0.5
292.6  11.3 ± 0.311.3 ± 0.3

[20] The maximum RHice at which our experimental system could be used to study ice nucleation was dependent on the quality of the silane coating of the quartz plate. We performed an experiment in the absence of any particles on the silanized quartz plate. In such a blank experiment, ice crystallized most likely at minor defects of the coating at RHice values that were about 5% lower than the threshold for homogeneous ice nucleation in a particle 0.2 μm in diameter at a freezing rate of 1 min−1 [Koop et al., 2000b]. We are convinced that the data points for nucleation in the presence of ATD particles described below that fall within 5% of the homogeneous nucleation threshold are not affected significantly by defects of the coating because we have visually confirmed that ice indeed formed at the dust particles, and not on the plain glass slide.

2.4. Experimental Procedure

[21] Observations of ice nucleation for different particles were derived according to the following procedure. After the particles were placed on the sample area, the vacuum system was sealed and pumped to remove any ambient gases in the system. The lower temperature stage was cooled to below 235 K and water vapor was supplied through the inlet into the vacuum system. The generation of an ice surface was verified by measuring the water partial pressure inside the cell. Before an ice nucleation experiment started, the lower temperature stage was set to 20 K below the temperature of the upper stage. Under these conditions, the RHice experienced by the particles on the sample area ranged between 5 and 40% RHice. Subsequently, the water reservoir was heated at 0.2–1 K min−1. This increased RHice at a rate of about 2–20% min−1 and corresponds to relative humidity changes observed in convective updrafts [DeMott et al., 1997, 1998; Kärcher and Lohmann, 2003; Kärcher and Ström, 2003; Gierens, 2003]. RHice was increased until ice formation on the particles was observed. During the experiments, optical images of the particles and the temperature of both stages were recorded on a digital video tape. Between two subsequent ice nucleation experiments the vapor pressure inside the cell was measured in order to detect leaks or any possible malfunction of the apparatus.

3. Results and Discussion

3.1. Ice Nucleation on ATD Particles

[22] Figure 2 shows the results of ice nucleation on ATD particles as a function of RHice and temperature. Additionally, the conditions under which a particle 0.2 μm in diameter nucleates ice homogeneously within one minute (this corresponds to a homogeneous nucleation rate coefficient of ∼5 × 1011 cm−3 s−1) are plotted as a dashed line [Koop et al., 2000b]. Figure 2 indicates that ATD particles nucleated ice over a broad RHice range at each specific temperature. Some particles initiated ice nucleation at saturation ratios as low as 105–110% RHice. Immersion freezing has been observed only at temperatures higher than 240 K when RHice reached water saturation. At this point water droplets formed around the dust particles, which then froze subsequently. At temperatures lower than 240 K no liquid activation of the particles was observed within our experimental resolution prior to ice nucleation, implying that deposition nucleation of ice occurred. We conclude from these observations that the formation of an ice germ on ATD particles from the vapor phase appears to be the favored nucleation pathway for temperatures below 240 K and RHice above ice saturation but below water saturation. In experiments conducted at temperatures below 230 K we observed that some particles did not nucleate ice within the experimental achievable RHice range. This might indicate that a subset of the ATD particles are very inefficient ice nuclei in the deposition mode.

Figure 2.

Ice nucleation events on ATD particles as a function of RHice and particle temperature. The shaded bars indicate the RHice range at which ATD nucleated ice. The dashed line represents the threshold for homogeneous ice nucleation of a solution droplet 0.2 μm in diameter using a freezing rate of 1 min−1 [Koop et al., 2000b]. The dotted line indicates ice saturation; the solid line is water saturation.

[23] Distributions of the number of ice nucleation events as a function of RHice at two given temperatures are shown in Figure 3. These distributions have been obtained from several ice nucleation experiments, either from one individual sample or several samples. Figure 3a shows that ATD particles nucleated ice over a range of 120–160% RHice at a temperature of 208 K. We attribute this behavior to the different nucleation characteristics of individual particles of different mineral types contained in ATD. This interpretation is supported by additional experiments that were performed in a different manner. After detection of the first ice nucleation event on an individual particle, RHice was held constant for several minutes and, during this time period, no further ice nucleation events occurred on other particles. In contrast, when RHice was increased again additional ice nucleation on other ATD particles occurred. This clearly indicates that the various crystallites contained in ATD (and in natural dust samples) have different IN properties leading to the broad range in observed ice nucleation events shown in Figure 3a. Further support for the different ice nucleation abilities of the various crystallites contained in ATD come from experiments shown in Figure 3b. The distribution of ice nucleation events recorded at a temperature of 230 K appears to be bimodal. These data stem from three different sets of experiments each with a newly prepared particle sample (originating from the same ATD suspension) which was processed multiple times. The nucleation events from two of these experiments corresponds to the peak around 100–113% RHice and those of a third experiment to the peak around 130–143% RHice. We interpret the bimodal behavior of this distribution as an inhomogeneous distribution of efficient IN in the different particle samples, in accordance with the experiments discussed above. However, we anticipate that the bimodal distribution would change to a single mode distribution for a large number of individual experiments using samples from the same ATD suspension.

Figure 3.

(a) A distribution of the number of ice nucleation events as a function of RHice obtained from multiple nucleation experiments of the same ATD sample at 208 K. (b) A distribution of the number of ice nucleation events observed at 230 K, however, three different ATD samples from the same ATD suspension have been used.

[24] Figure 4 shows under which conditions ice nucleated on ATD particles that were not involved in ice formation previously. These RHice onset values for ice nucleation are higher than the lowest RHice values shown in Figures 2 and 3, indicating the effect of preactivation. We extracted these data of ice nucleation events on ATD particles that were not preactivated in order to allow a better comparison to the data of Archuleta et al. [2005] and Möhler et al. [2005]. Both these studies investigated ice nucleation on oxide particles and ATD particles that were not preactivated. The RHice onset values of the different particles by Archuleta et al. [2005] are similar to ours. The RHice onset values of ATD particles by Möhler et al. [2005] are significantly lower than our onset values and those of Archuleta et al. [2005] but are in agreement with the lowest RHice values that we observed for ice nucleation on ATD (see in Figure 2). We are not sure about the cause for these differences but offer two possible explanations. First, ATD particle preparation differ between both studies. Here, the particles are suspended in water before they are dispersed on the sample area, whereas in the study of Möhler et al. [2005] dry particles are injected into the AIDA expansion chamber. Second, the investigated ATD particle size fraction differ between both studies. Here, ATD particles between ∼1–10 μm were investigated, while Möhler et al. [2005] studied particles ∼0.1–1.5 μm in diameter. Different ATD size fractions are likely to be different in composition, since a particular mineral type may be enriched in a certain size range while being depleted in another. For example, it has been observed that silica is enriched in the coarse mode of dust particles when compared to the fine mode [Gomes et al., 1990; Reid et al., 2003; Vlasenko et al., 2005]. A size-dependent dust composition may lead to particle samples with different ice nucleation abilities and therefore might be a possible cause for the different activation thresholds for ice nucleation observed by Möhler et al. [2005] and in this study.

Figure 4.

Ice nucleation on ATD particles as a function of RHice and temperature. Ice nucleation on particles that were not preactivated are shown as solid diamonds. For comparison, ice nucleation data from the literature are also shown: 200 nm diameter particles of Al2O3 (open diamonds), Fe2O3 (open squares), 3Al2O3:2SiO2 (open triangles), and Asian dust (open circles) [Archuleta et al., 2005]; ATD particles (stars) [Möhler et al., 2005]. The lines are identical to those in Figure 2.

[25] Because of the observation that multiple processing of the same particles led to decreasing RHice values for ice nucleation, we investigated this phenomenon in more detail. The observation of a particle nucleating ice at lower RHice in subsequent experiments when compared to the first initial ice nucleation event is termed preactivation [Fournier D'Albe, 1949; Mason, 1950; Mossop, 1956; Mason and Maybank, 1958; Mossop, 1963; Higuchi and Fukuta, 1966; Roberts and Hallett, 1968]. The molecular explanation for preactivation is not known but may be attributed to tiny ice structures in confined spaces such as cavities or pores [Volmer, 1939; Turnbull, 1950; Mossop, 1956; Mason and Maybank, 1958; Mossop, 1963; Fukuta, 1966; Roberts and Hallett, 1968]. Ice may survive in subsaturated environments due to a decrease in the equilibrium vapor pressure as a result of a concave curvature (“negative Kelvin effect”) in small hydrophilic capillaries or pores [Fukuta, 1966; Pruppacher and Klett, 1997]. While we cannot investigate such processes on a molecular level with our experimental approach, we were able to study preactivation phenomenologically. This was done by initiating ice nucleation on a freshly prepared ATD sample at a fixed particle temperature. After the detection of ice formation the temperature of the lower stage was decreased to about 20 K below the particle temperature. Under these conditions RHice in the vacuum system ranged between 5 and 40% RHice. A new experiment started by increasing the temperature of the lower stage, thus increasing RHice, until ice formation was observed. In most experiments the particle which nucleated ice first in the initial experiment, nucleated ice also first in the second experiment but at lower RHice. The difference in the observed RHice values at which ice formation was observed was attributed to preactivation. Figure 5 shows results of preactivation experiments on ATD particles. The nucleation properties of the particles changed by up to 30% in RHice before and after preactivation, and even at temperatures as low as 200 K preactivation had a significant effect on the onset of ice nucleation. Also shown are preactivation values obtained from Kaolinite particles [Roberts and Hallett, 1968].

Figure 5.

The solid circles indicate the measured difference between the initial and the subsequent ice nucleation relative humidity of the same ATD particle, ΔRHice, at various temperatures. The shaded area indicates the 95% confidence interval of a linear fit to the data of this study. The solid triangles represent ΔRHice values obtained from Kaolinite particles [Roberts and Hallett, 1968].

[26] We also performed three experiments to study under which conditions preactivation could be impeded. After ice nucleation had occurred at temperatures of 210, 220, and 240 K, the temperature of the particles was increased to about 250, 265, and 300 K, respectively, for about 5 min. In addition, the temperature of the lower stage was decreased to 20 K below the initial particle temperature. This procedure strongly reduced RHice in the vacuum system to values between 0.2 and 3.5% and ceased preactivation of the particles. Subsequent ice nucleation was observed at significantly higher RHice values than those obtained from preactivated particles. Therefore the exposure of the particles to RHice values lower than ∼5% appears to remove preactivation.

3.2. Ice Nucleation on H2SO4 Coated ATD Particles

[27] The influence of a H2SO4 coating of ATD particles on ice nucleation was investigated. Figure 6 shows the conditions at which ice nucleation on H2SO4 coated ATD particles was observed. H2SO4 coated ATD particles nucleated ice at lower RHice values than required for the homogeneous ice nucleation from aqueous H2SO4 droplets. Figure 6 also shows the RHice onset values of the initial ice nucleation of particles that were not preactivated. The RHice values are similar to those observed for H2SO4 coated metal oxides studied by Archuleta et al. [2005]. Regarding the RHice range in which ice nucleation on H2SO4 coated ATD particles occurred, no significant difference to the ice nucleation ability of pure ATD particles has been found. Again, preactivation of the particles was observed, i.e., after two or more repetitions of the experiment RHice onset values decreased to the lowest values shown in Figure 6. For example, at particle temperatures of 219 and 229 K, ice nucleation was observed to occur at RHice values which decreased from 121 to 104% and from 140 to 105%, respectively. As mentioned above, we are not able to observe the particle surface on a molecular level during the formation of ice. Nevertheless, we suggest a possible explanation for preactivation of H2SO4 coated ATD particles. As in the case of pure ATD particles, survival of minute ice crystals in nanometer-sized hydrophilic pores or capillaries might occur at RHice values below 100% also on H2SO4 coated ATD particles. While a direct contact of the capillary ice with a H2SO4/H2O solution layer may influence the magnitude of the effect through changes in wetting behavior and surface energies, we expect the principle behavior to be the same as that of uncovered ATD particles. Therefore we propose that ATD particles which possess nanometer sized hydrophilic cavities or pores may be able to exist in a preactivated state irrespective of whether they are coated with a solution layer or not.

Figure 6.

Ice nucleation on H2SO4 coated ATD particles as a function of RHice and temperature. The shaded bars indicate the RHice range at which the particles nucleated ice. The solid circles represent ice nucleation on particles that were not preactivated. The lines are identical to those in Figure 2.

4. Conclusions and Atmospheric Implications

[28] We have presented a new experimental technique to study heterogenous ice nucleation for temperatures as low as 197 K and at relative humidities with respect to ice which can range between ice and water saturation. Controlled supersaturation with respect to ice was achieved by operating in the Knudsen flow regime. With this apparatus the formation of ice on ATD particles serving as surrogates for atmospheric mineral dust particles was observed by optical microscopy.

[29] The experimental results indicate that desert mineral dust particles may nucleate ice over a broad RHice range due to the different mineral crystallites contained in naturally occurring mineral dust. Both pure and H2SO4 coated ATD particles nucleated ice at significantly lower RHice values than those required for homogeneous ice nucleation in liquid aerosols. A significant difference in the ice nucleation ability of pure and H2SO4 coated ATD particles has not been observed. The experimentally obtained RHice values at which ice nucleation occurred are in a similar range than ice nucleation thresholds observed in the field for cirrus cloud formation in the northern hemisphere [Ström et al., 2003; Haag et al., 2003].

[30] The heterogeneity of mineral dust particles should be considered in future experimental investigations. Ice nucleation studies might be combined with surface sensitive methods such as single-particle scanning electron microscopy with energy dispersive analysis with X rays [Falkovich et al., 2001; Reid et al., 2003; Shi et al., 2005] in order to improve the assessment of the nucleation properties of individual mineral species. This investigation also shows the difficulty to compare ice nucleation studies of mineral particles between different laboratories due to the heterogeneity of the employed mineral samples, different experimental methods for the detection of ice nucleation, and different procedures to prepare dust samples. Our experiments also clearly reveal that preactivation of dust particles can occur for temperatures as low as 200 K and that a coating of sulfuric acid does not impede preactivation. Thus preactivated mineral dust particles transported to the upper troposphere have the potential to strongly impact dehydration processes by acting as efficient IN. Preactivated dust particles may also seed clouds at lower altitudes [Mossop, 1956, 1963]. For example, once a high-altitude cirrus ice cloud is formed heterogeneously and the ice crystals grow to sizes at which sedimentation to lower altitudes becomes fast, the dust particles might survive in a preactivated state after evaporation and induce ice at RHice values as low as about 100–120% RHice at lower altitudes. Therefore our results imply that preactivation should be considered when modeling transport of mineral dust particles and subsequent formation of ice clouds.

Acknowledgments

[31] Technical support of U. G. Weers and helpful discussions with B. P. Luo, U. K. Krieger, T. Peter, A. K. Bertram are gratefully acknowledged. We are grateful to O. Möhler for providing unpublished data. This work was funded by an internal ETH grant and by the European Commission within the Integrated Project “Stratosphere-Climate Links With Emphasis on the UTLS” (SCOUT-O3).

Ancillary