Deactivation of ice nuclei due to atmospherically relevant surface coatings

The ice nucleation characteristics of Arizona test dust (ATD) and illite clay, surrogates for atmospheric ice nuclei, have been determined at the Aerosol Interactions and Dynamics in the Atmosphere (AIDA) chamber located at the Research Center Karlsruhe in Germany. The objective of this research was to determine the effect of sulfuric acid and ammonium sulfate coatings on the ability of these mineral dust surrogates to nucleate ice in an environment where particles realistically compete for water vapor. Coated ATD particles required higher saturations at all temperatures considered, from −20 to −45 °C, than did identical uncoated particles. Freezing of coated particles often required saturations approaching those for the homogeneous freezing of aqueous solutions of the coating material alone. Less pronounced effects were found for illite, although the presence of a coating consistently increased the saturation or decreased the temperature required for ice formation. Analysis of ice residue at the single particle level suggests that the first coated particles to freeze had thinner or incomplete coatings when compared to particles that froze later in the expansion. This observation highlights a need to verify coating properties since an assumption of homogeneity of a group of coated aerosols may be incorrect. The increase in saturation ratio for freezing suggests that gas-phase uptake of sulfates, a large fraction of which are due to anthropogenic emissions, will reduce the ice and mixed-phase cloud formation potential of atmospheric ice nuclei.


Introduction
Atmospheric ice formation takes place via two distinct pathways. The majority of atmospheric particles throughout the depth of the troposphere are aqueous mixtures of sulfates, organics and nitrates, neutralized to some extent by ammonia Thomson 1997, Murphy et al 2006). These aqueous particles require temperatures below approximately −38 • C and saturations near that of liquid water before the spontaneous formation of ice occurs (Koop et al 2000). This process is termed homogeneous freezing. Freezing which occurs above this temperature or below this saturation requires the presence of a special ice nucleus (IN) that stabilizes the formation of a water ice embryo (Pruppacher and Klett 1997). This process is termed heterogeneous freezing and it can occur via several distinct sub-pathways. These include the deposition of ice on a surface, freezing within a condensation layer, or by an IN fully immersed in a droplet (Pruppacher and Klett 1997). Less well studied to date is the process of freezing when an IN contacts a droplet (Cantrell and Heymsfield 2005).
The formation of ice in cirrus and mixed-phase clouds is among the most uncertain aspects of global climate change (IPCC 2007). This lack of certainty highlights a need for more comprehensive studies of the conditions at which IN initiate freezing. It is known from laboratory and field studies that mineral dust aerosols are among the most atmospherically common and effective IN (i.e., they require the highest temperature and lowest saturation; Pruppacher and Klett 1997). Other particles, for example carbonaceous cores from biomass burning and fossil fuel combustion, are more atmospherically common (Murphy et al 2006) but less effective as IN (DeMott et al 2003a, Dymarska et al 2006. A few bacteria and some other bioaerosols such as pollen have been shown to be effective IN (Möhler et al 2007, Pratt et al 2009 but their atmospheric abundance and distribution are not known and appears to be extremely low. For example bacterial mass spectrometric patterns have been resolved (Fergenson et al 2004) but they have not been found in great numbers during atmospheric studies.
Given their combined large atmospheric burden (Usher et al 2003) and ice formation potential (Pruppacher andKlett 1997, DeMott et al 2003b), mineral dusts are of considerable interest in cloud formation, precipitation initiation, and the global radiative balance. Gas-phase components such as nitrates, sulfates, and organics are known to adhere to mineral dust surfaces Thomson 1997, Grassian 2002). The effect of these surface layers is to increase the ability of the normally insoluble and hydrophobic mineral dust core to take up water and form cloud droplets (Levin et al 1996, Herich et al 2009. The theoretical effect of the formation of coatings due to the uptake of gas-phase species on ice nucleation as well as due to coagulation with pre-existing particles is summarized by Pruppacher and Klett (1997): whereas the coating process enhances cloud drop formation it has been suggested to 'deactivate' ice nucleation.
More recently, sulfate coating experiments have been conducted by Zuberi et al (2002) who studied kaolinite and montmorillonite and Hung et al (2003) who studied hematite and corundum mineral dust cores immersed within larger ammonium sulfate droplets using optical microscopy and infrared spectroscopy, respectively. Ettner et al (2004) also optically studied freezing of multiple kaolinite and montmorillonite inclusions in large sulfuric acid droplets (>1 mm diameter) levitated by acoustic forces. Zobrist et al (2008) used differential scanning calorimetry and optical techniques to observe several possible IN, including silica and ATD, within several inorganic salts and organics including ammonium sulfate and sulfuric acid. Due to instrumental limitations the majority of these studies were on supermicrometer droplets and were only able to compare the required freezing conditions to pure droplets of the coating material, not to uncoated mineral dust (i.e., immersion versus homogeneous freezing only). In all cases the presence of an immersed nucleus initiated freezing at higher temperatures than required for homogeneous freezing of the coating material alone.
Organic coating experiments have been undertaken by Kanji et al (2008) who compared ice nucleation of silica with silica coated with an octyl chain in a thermal gradient chamber. Möhler et al (2008) undertook experiments in the AIDA chamber with organic coatings from the reaction of ozone with α-pinene on mineral dust cores of ATD and the clay illite. Both research groups concluded that organics 'deactivated' the IN, resulting in a higher saturation and/or a lower required temperature for ice formation. Möhler et al (2008) also found that homogeneous freezing of the pure organic was, in all cases, less effective than material with an immersed dust particle. Eastwood et al (2009) investigated the clay kaolinite both in a bare state and coated with sulfuric acid and ammonium sulfate. Eastwood et al (2009) also found that aqueous coatings in all cases inhibited ice nucleation although crystalline ammonium sulfate had little or no effect on the IN. Archuleta et al (2005) performed experiments in a continuous flow diffusion chamber on variable-sized metal oxides and alumina silicates from 50 to 200 nm diameter both in an uncoated state and coated with sulfate. Archuleta et al found that the effect of the coating depended on the core material. Specifically, sulfate appeared to enhance ice nucleation in about 50% of the studied samples of iron oxide whereas the alumina silicates were predominantly deactivated by a coating. Aluminum oxide was not affected. The coated and uncoated mineral dust particles were in all cases more effective at ice formation than the homogeneous freezing of the coating material alone. Knopf and Koop (2006) investigated bare ATD and particles coated with sulfuric acid. In agreement with the finding of Archuleta et al (2005) for aluminum oxide, but contrary to the findings of Eastwood et al (2009), Knopf and Koop (2006) found little or no coating effect on IN potential. Specific experimental conditions are compared in section 4.
In this study we have undertaken freezing experiments on ATD and illite clay, meant to be representative of atmospheric mineral dust particles. We have determined the saturation required, as a function of temperature, for ice nucleation by uncoated as well as particles coated with sulfuric acid or ammonium sulfate. These results are compared to the Table 1. Coating and aerosol properties. ATD and illite particles were uncoated or coated with sulfuric acid (SA) or ammonium sulfate (AS). A higher saturation temperature (T SA,sat ) in the coating unit resulted in a thicker coating. Also given are the initial gas temperature T g,0 and aerosol number concentration n ae,0 in the AIDA chamber measured just before the start of the expansion. homogeneous freezing threshold of the pure coating material. Both coated and uncoated ATD are more effective at ice formation than the homogeneous freezing of the pure coating material. Bare ATD is more effective than coated ATD from −20 to −45 • C, the range of temperatures investigated in this study. The deposition, immersion, and contact freezing modes that are possible in the atmosphere are all also possible in the cloud chamber.

Experimental methods
Experiments were conducted during fall 2005 as part of the IN08 campaign at the AIDA cloud chamber located at the Karlsruhe Institute of Technology in Germany. AIDA is an 84 m 3 actively cooled aluminum chamber with a diameter of 4 m. The chamber is evacuated before experiments to a pressure <0.1 hPa and is then refilled with particle-free air (background density <0.1 cm −3 ). Particles of a known composition and size can then be added to the chamber. For the ice nucleation experiments described here the chamber was set to a constant temperature between −20 and −30 • C. Mixing with a fan was used to keep the temperature variability within the chamber below ±0.3 • C. The pre-experimental pressure was, for these studies, ∼1000 hPa. Humidity conditions were near the saturation pressure of ice at the wall temperature and were set by a thin coating of ice on the inner surface of the chamber. Ice nucleation was initiated by inducing an expansion within the chamber with vacuum pumping. While the chamber walls remained at an essentially constant temperature the pressure within the vessel was dropped by ∼200 hPa over the course of a few minutes. This can be visualized in figure 1(a) where the gas pressure during a typical expansion, in this Table 2. Experimental conditions. The ice nucleation onset temperature T 0.01 , relative humidity with respect to ice RH i0.01 , and relative humidity with respect to water RH w0.01 correspond to an ice-active aerosol particle number fraction of 1%. The mode of ice nucleation is indicated in the last column. During experiments 26, 27, and 28, only a few ice particles formed at about −26 • C with a number fraction below 0.01. The freezing modes were derived from the RH w and Welas measurements. When RH w remained below 100% and there was no indication of the formation of droplets deposition nucleation is noted. In experiments with RH w at 100% where the formation of ice was only observed after droplets had formed immersion freezing is noted. Formation of ice observed with RH w close to 100% is noted as either deposition or condensation/immersion freezing. case the 17th of the campaign which utilized ATD coated with sulfuric acid at −45 • C, is plotted. The result is a corresponding drop in the gas temperature and increase in the relative humidity (panel (b)). This pressure drop rate mimics the process of uplift in the atmosphere, corresponding to a velocity of ∼0.1-2 m s −1 . A description of aerosol properties is given in table 1 and expansion conditions in table 2. As the relative humidity increased and the temperature decreased, ice nucleation was initiated by some fraction of the aerosol in the chamber. This can be visualized in figure 1(c) where the number fraction f ice = n ice /n ae of particles which nucleated ice is plotted (right axis). The total aerosol number concentration n ae is also plotted in this panel (left axis). In addition to the formation of new ice, post-nucleation crystals deplete water vapor via growth (panel (d)). The initiation of ice formation by a given fraction of the aerosol (i.e., their cloud formation potential) can thus be related to a temperature and relative humidity using data of the type shown in figure 1. Ice nucleation continued until the combination of the expansion ending and ice crystal sedimentation depleted the gas-phase water until it ultimately relaxed to an equilibrium with the ice on the chamber walls. These measurements require the concerted effort of several independent instruments including facility temperature, pressure and a tunable diode laser for water vapor pressure measurements. A comprehensive list of these instruments and a schematic of the AIDA chamber  (a) is the response of the gas-phase pressure to vacuum pumping with respect to time and the corresponding gas and wall temperature profiles. Panel (b) relates the effect on the relative humidity with respect to ice (i.e., RH i ) and liquid water (i.e., RH w ; Murphy and Koop 2005) using a tunable diode laser and MBW frost point hygrometer. Panel (c) is the concentration of aerosol and of ice crystals formed as discerned by a CPC and the Welas and Welas2 instruments, respectively. Panel (d) shows the evolution of crystal growth. Note that Welas2 overestimates ice crystal sizes for d > 30 μm. See text for further details. is given in Möhler et al (2008). A condensation particle counter (CPC; TSI Inc. model CPC3010) for aerosol number density and two optical particle counters (Palas GmbH, Welas type) for ice crystal concentration were also connected to the chamber via sampling tubes. The two versions of the Welas type OPC, located within the cold section surrounding the AIDA chamber, were used with different size ranges and sensitivities for single particle detection. Assuming spherical particles with a refractive index of 1.33, Welas and Welas2 detect particles with diameters from 0.8 μm to 46 μm and from 9.0 μm to 237 μm, respectively. For non-spherical ice particles, both instruments are only used in the counting mode to measure ice particle concentrations. No attempt was made to quantitatively derive crystal sizes from the Welas scatter intensities. Nevertheless, growing ice particles can be detected and distinguished from large aerosol particles or droplets of almost constant size during the AIDA expansion experiments. For concentration measurements of 1.0 cm −3 for Welas and 0.3 cm −3 for Welas2 the counting uncertainty was less than ±50% (i.e., more than 4 counts) for acquisition periods of 10 s.
Aerosol particles were generated from samples which were used and characterized in earlier studies (Möhler et al 2006. Dry samples were first aerosolized with a rotating brush disperser (RBG-1000, Palas) and then further deagglomerated in a dispersion nozzle. The resulting aerosol was passed through a cyclone impactor to remove particles larger than ∼1 μm in diameter and was then added to the aerosol preparation and characterization (APC) chamber, a 3.7 m 3 volume outside AIDA, and uniformly distributed with the help of a mixing fan. Both the brush disperser and the dispersion nozzle were operated with dry and particle-free synthetic air to avoid contamination of the dust aerosol with particles of unknown origin. From the APC chamber a minor fraction of the dust aerosol was passed through a stainless steel connection tube to the large AIDA cloud chamber. Particles were coated with H 2 SO 4 when the flow from the aerosol generator was mixed at a temperature of ∼150 • C with a flow of synthetic air saturated with sulfuric acid vapor. The sulfuric acid condensed to the dust particles upon slow cooling of the mixture. The coating amount was varied by selecting different sulfuric acid saturation temperatures of the synthetic air flow (see table 1). Ion chromatography of bulk filter samples collected during the experiments showed average sulfuric acid concentration of ∼10% by mass and with a range from 5 to 20%. The coated dust aerosol was also added to the APC chamber and from there transferred into the larger AIDA cloud chamber. Coating with ammonium sulfate was achieved inside the AIDA chamber. Ammonia gas was added to the chamber in excess to the amount of sulfuric acid. The particles took up ammonia until the sulfuric acid layers were fully neutralized to ammonium sulfate. Fourier transform infrared spectroscopy was used to verify composition.
Single particle mass spectrometry was performed on the test aerosol before and after, and on the ice residue during, expansions. This technique was used to distinguish the presence or absence, and give a rough indication of the extent of surface coatings. Aerosol was directly removed from the APC chamber or AIDA before or after expansions. During expansions a pumped counterflow virtual impactor (PCVI) was used to inertially separate crystals from aerosol which did not form ice (Boulter et al 2006). The crystals were exposed to a warm and dry counterflow such that condensed-phase water was evaporated. The residual particles were then conducted to the particle analysis by laser mass spectrometry (PALMS) instrument. PALMS draws particles into the instrument and focuses them using an aerodynamic inlet and differential pumping stages. Particles are detected and aerodynamically sized as they scatter light from two visible 532 nm YAG lasers set 34 mm apart. Sizing is accomplished because particles exiting the aerodynamic lens are accelerated to a size-dependent velocity and this is determined by recording the transit time between the visible lasers. A 193 nm excimer laser is then triggered to strike the particle, thereby ablating and ionizing the components. Positive or negative ions are then accelerated in a reflectron mass spectrometer flight tube and detected as a function of their mass. Particle size and monopolar chemical composition are thus recorded on the single particle level in situ and in real time. A comprehensive instrument description of PALMS is provided in Cziczo et al (2006) and the PCVI-PALMS combination has been utilized at the AIDA chamber previously (Gallavardin et al 2008). It is noteworthy that aerodynamic focusing and optical detection restrict the particle size range to 200-3000 nm, but within this range both refractory (e.g., mineral dust) and semivolatile (e.g., sulfates and organics) components are resolved (figure 2). Although not inherently quantitative, single particle mass spectrometer trends in ion signal can be used as an indication of relative abundance (Murphy et al 2006, Spencer andPrather 2006). For ionizable substances particle fractions on the order of 1% (i.e., monolayer-level coatings) are detectable. Figure 2(a) is an exemplary negative polarity mass spectrum of a coated ATD particle to illustrate the ions associated with sulfate coatings. Panel (b) is a histogram which shows the distribution of the coating signal among particles in an experiment where all particles were subject to the aforementioned coating process (see section 3).

Results
Expansions were conducted to determine freezing onset temperature and saturation of uncoated and sulfuric acid and ammonium sulfate coated particles at approximately −20, −30 and −45 • C. The ice nucleation onsets in this work are defined as the threshold condition for an ice-active particle number fraction f ice of 1%, and were obtained from respective scatter plots of f ice versus the temperature and the relative humidity with respect to ice and to water. Results for the various expansions are plotted in figure 3. The homogeneous freezing line for 500 nm diameter particles is plotted for reference (Koop et al 2000). The most important result of this study is that atmospherically relevant coatings decrease the ice formation potential of very efficient ice nuclei. Specifically, coatings of sulfuric acid and ammonium sulfate increased the saturation and/or decreased the temperatures required for the formation of ice by ATD particles. In the extreme case of experiments at ∼−20 • C coated particles were not observed to freeze up to the saturation of liquid water whereas uncoated particles did freeze. At lower temperatures coated particles were observed to freeze but a higher saturation was required. At −50 • C, for example, freezing occurred at 30-40% higher RH i for coated ATD. Furthermore, sulfuric acid and ammonium sulfate coated ATD did not freeze until conditions were near the homogeneous freezing threshold. Particles coated with ammonium sulfate at all temperatures required slightly higher saturations for the onset of freezing (by a few %RH i ) than sulfuric acid. The difference between uncoated and coated particles of both types was in all cases greater than the difference between the coatings. . Ice nucleation onset conditions at the level of 1% of particles from this work for uncoated, sulfuric acid and ammonium sulfate coated ATD particles. Previous studies include uncoated and sulfuric acid coated ATD from Knopf and Koop (2006) for experiments where aerosols were not pre-activated, and sulfuric acid and ammonium sulfate coated ATD from Zobrist et al (2008). Water saturation (dashed) and homogeneous freezing curves for 500 nm particles at 1 min −1 (solid) are shown for reference. ATD points at −51 • C are from Möhler et al (2006). See text for further details.
The sulfuric acid and ammonium sulfate coatings were detected by PALMS primarily as the presence of the m/z 97 (HSO − 4 ) ion ( figure 2(a)). This peak was present for ∼80% of the coated particles larger than 200 nm. A distribution of coating signal was observed as shown in figure 2(b). PALMS does not ablate and ionize the entire particle for dust of the sizes used here. The absence of a sulfate peak is ambiguous as to whether that particle was completely uncoated or the PALMS ion signal came from a bare spot. In either case, the implications for heterogeneous ice nucleation are the same: there were uncoated portions of the particle available for nucleation.
Particles that nucleated relatively early in an expansion had different compositions than those that nucleated later. First, particles that nucleated early were those that had incomplete or thinner coatings than aerosols in the main freezing segment. Figure 4(a) shows the time series of the fraction of ice particle residues with a coating signature at m/z 97 for IN08-17 (ATD coated with sulfuric acid). A lower fraction of the earliest nucleating aerosols had a coating signature, and average coating intensity was relatively low at <160 s. If single particle analysis is used to disregard this early nucleation segment when coating signal was absent or low, the freezing onset conditions for sulfuric acid coated ATD shift by as much as 25%RH i and −2.5 • C. Second, early nucleation of stainless steel contamination particles generated inside the chamber was occasionally observed, as in expansion IN08-18 (figure 4(b); ATD coated with ammonium sulfate). The residue composition began to reflect coated mineral dust at 25% higher RH i and 3.5 • C lower temperature than the initial freezing due to stainless steel. Expansions were also conducted on sulfuric acid and ammonium sulfate coated and uncoated illite particles at ∼−35 • C. Freezing of the uncoated aerosol was slightly lower (by a few %RH i ) than coated particles. The freezing point of the uncoated aerosol was considerable higher, ∼130%RH i , than ATD (100-120%RH i ) at this temperature. This example shows that the characteristics of the nucleus (e.g., composition, surface morphology, etc) as well as the coating state of the aerosol are important determinates in freezing potential.
Mass spectral signals for refractory mineral dust components were also investigated. As in previous studies, ice residues were often enriched in lead (Cziczo et al 2009). Other than lead, a trace element found in mineral dust, little indication of preferential nucleation due to any major mineral component of ATD or illite was found. Occasionally, minor differences in nucleating behavior correlated with ion signals for lithium, barium, and strontium, but the effects were small and no clear trends could be identified. Condensation of gasphase sulfuric acid, nitric acid, and organic species present in the background chamber air onto ice crystals was observed throughout these studies (Gallavardin et al 2008).

Discussion
The trend in ice nucleation behavior in this study is consistent with the results of Eastwood et al (2009) who found that the ice nucleation threshold of the clay kaolinite was increased from RH i ∼ 110% to ∼140% in the temperature range from −27 to −40 • C when coated with aqueous sulfuric acid and ammonium sulfate. In that study particles of ∼15 μm diameter were coated with ∼0.7 μm of sulfate. Möhler et al (2008) similarly found that organic coatings required an increase in RH i from 10 to 50% between −68 and −73 • C depending on thickness. In that study particles were of ATD and the clay illite from 100 nm to 1 μm diameter coated with from 17 to 41 wt% organic material. Archuleta et al (2005) found that the RH i required to nucleate 1% of alumina silicate particles, the major component of ATD, between 50 and 200 nm diameter was increased by ∼15%RH i between −45 and −60 • C. Coatings of 3-7 layers of sulfuric acid were used.
The results of Knopf and Koop (2006) are plotted with the data from this study in figure 3. Knopf and Koop (2006) found that ATD nucleated ice over a broad range, from ∼100 RH i to ∼100 RH w from −13 to −38 • C. From −38 to −78 • C ATD was observed to nucleate ice over the range from ∼100%RH i up to homogeneous freezing threshold. Particles from 700 nm to 10 μm diameter were investigated. Sulfuric acid coatings from 67 to 1100 layers were not observed to affect this range. Note, however, that this wide nucleation range precluded precise identification of difference in freezing behavior. Zobrist et al (2008) determined ice nucleation onset conditions for super-micrometer diameter particles, approximately an order of magnitude larger in diameter than those used in this study, coated with sulfuric acid and ammonium sulfate. These aerosols froze at a slightly lower humidity which is not surprising given their much greater surface. These data are also plotted in figure 3. Uncoated ATD was not investigated by Zobrist et al (2008) due to instrumental limitations.
Archuleta et al (2005) observed that aluminum oxide was unaffected by sulfuric acid coatings and that iron oxides were rendered slightly more effective, by ∼5%RH i , ice nuclei. Coatings, temperatures, and particle sizes were the same as for the previously described experiments on alumina silicates. Our observations were that the effect of a coating was less pronounced on illite than on ATD. We did not observe cases where coatings enhanced or had no effect on ice nucleation as is given by Archuleta et al (2005). As a general conclusion, the nature of the ice nucleus and the coating, in addition to the particle size, are important determinates in the conditions required for freezing. The use of single particle analysis in our study shows that the particle-to-particle variability in coating presence and thickness can have a large effect on the freezing point. Distributions of nominally 'coated' particles actually contained particles with no sulfate signal. Because in any given batch all particles were exposed to sulfuric acid vapor under the same conditions, the lack of a sulfate signal from some particles is somewhat difficult to understand. Electron microscopy of atmospheric mineral dust often shows that materials such as sulfates do not evenly coat a surface but are instead preferentially found in distinct locations (Buseck and Posfai 1999). Ice nucleation has been shown to occur at specific sites on the surface of mineral dust, often steps or other dislocations (Pruppacher and Klett 1997). This may also be true of the deposition of sulfates although a critical amount would ultimately surround the entire particle. In the case of deposition mode, and experiments below the deliquescence point of ammonium sulfates, this may help to explain the presence of bare mineral areas on which ice nucleated. In the case of studies in which droplet formation occurred (i.e., immersion and deposition mode) even an incomplete coating would be expected to activated and ultimately surround the dust particle with an aqueous solution. This suggests that the coating material was completely absent on some particles in these experiments.
The mass spectrometric observation of partially or fully uncoated particles is independently supported by the expansion experiments since those particles are also more effective ice nuclei than particles with large sulfate signals. We find here that up to a 25%RH i difference is apparent between particles with a small or no coating signal and those with a large signal. The heterogeneity of the coating is almost certainly related to the amount of coating material with monolayer coatings much more likely to be heterogeneous than those thicker than the diameter of the core particle. The occurrence of incomplete coatings in previous experiments, as well as the effect on results, is unknown because, to our knowledge, this is the first ice nucleation study that employed single particle mass spectrometry as an analytical technique.
The consensus of recent experiments is that coatings render bare mineral dust IN less effective or, at the least, have little effect. This is in agreement with theoretical arguments summarized by Pruppacher and Klett (1997). It has been shown that sulfate from anthropogenic sources can result in a coating on mineral dust (Kandler et al 2007). Increased emissions of sulfate by anthropogenic sources may therefore increase the coating on aerosol particles that otherwise may have either not been coated or coated to a lesser extent. This may then result in a deactivation of ice nuclei and an increase in the saturation required for cloud formation above pre-industrial levels. The effects of deactivation of dust by coatings is just beginning to be explored with global models (Hoose et al 2008) and, to our knowledge, is not included in current estimates of indirect aerosol forcing.