Earth and Planetary Science Letters Volcanic ash ice-nucleating activity can be enhanced or depressed by ash-gas interaction in the eruption plume

Volcanic ash can trigger ice nucleation when immersed in supercooled water. This will impact several processes (e.g., electriﬁcation, aggregation, precipitation) in the eruption plume and cloud and in the wider atmosphere upon ash dispersal. Previous studies show that ash bulk properties, reﬂecting the chemistry and phase state of the source magma, likely contribute to the ice-nucleating activity (INA) of ash. However, it remains unexplored how interaction with magmatic gases in the hot eruption plume, which inevitably leads to altered ash surface properties, affects the ash INA. Here we demonstrate that the INA of tephra is raised by exposure to H 2 O (g) mixed with SO 2(g) at both 800 and 400 ◦ C, but is substantially reduced by exposure to H 2 O (g) alone or mixed with HCl (g) at the same temperatures. In contrast, the INA of K-feldspar and quartz is reduced by all three eruption plume processing treatments. The decrease in INA of all silicates after heating with H 2 O (g) might relate to a loss of ice-active sites by surface dehydroxylation and/or oxidation. In the presence of HCl (g) or SO 2(g) , respectively, metal chloride or sulphate salts form on the tephra surfaces only. While NaCl and CaCl 2 seem to have no effect on the tephra INA, CaSO 4 is inferred to create ice-active sites, potentially through a particular combination of surface chemistry and topography. Overall, our ﬁndings suggest a complex interplay of bulk mineralogy and surface alteration in inﬂuencing ice nucleation by volcanic ash, and highlight the general sensitivity (enhancement or depression) of ash INA to interaction with magmatic gases in the eruption plume.


Introduction
Explosive volcanic eruptions generate mixed glassy and crystalline silicate ash that can serve as ice-nucleating particles (INPs). In the vertical eruption plume and the laterally dispersed eruption cloud, the freezing of supercooled water may impact a range of processes including plume/cloud electrification, gas scavenging, and ash aggregation (Van Eaton et al., 2015;Prata et al., 2020). In the wider atmosphere, ice formation on airborne ash may modify the properties and lifetime of clouds, thereby affecting precipitation and the Earth's radiation balance (Seifert et al., 2011). Volcanic activity worldwide produces a recurrent flux of ash to the atmosphere (176-256 Tg a −1 ) (Durant et al., 2010), while sporadic large eruptions release substantial quantities of ash at one time (e.g., ∼500 Tg on 18 May 1980 from Mt. St. Helens volcano, USA) Fig. 1. Schematic of the eruption plume and cloud, meteorological clouds, and altitude-temperature profile of the international standard atmosphere (ISA). Note that the adsorption and condensation zones in the eruption plume (Óskarsson, 1980) are not shown to scale with respect to the ISA profile; the altitudes at which they occur depend on factors such as the volcano summit elevation, eruption magnitude, and air entrainment rate. The hot core of the plume, where ash-gas interaction takes place, is the context of 'eruption plume processing' in this study.
surface properties of ash differ from the bulk properties, due in part to interactions with magmatic gases including H 2 O (g) , SO 2(g) , HCl (g) and HF (g) at high temperatures (∼200-800 • C) in the eruption plume (Óskarsson, 1980;Delmelle et al., 2007). This 'eruption plume processing' emplaces surficial metal sulphate and halide salts and modifies the surface chemical reactivity of ash (Óskarsson, 1980;Delmelle et al., 2007;Maters et al., 2016), but its influence on the INA of ash in particular is not known (Fig. 1).
Based on field measurements showing no increase in background INP concentrations following several volcanic eruptions, it has been suggested that exposure to magmatic SO 2(g) deactivates the INA of ash particles (Langer et al., 1974;Schnell et al., 1982). Numerous experimental studies on mineral dusts have investigated the effects of gaseous and aqueous acids on INA. Treatment of montmorillonite with SO 2(g) had no impact on its ability to nucleate ice from water in the vapour phase (i.e. in the deposition mode) (Salam et al., 2008), while exposure of Arizona test dust (ATD) to HNO 3 vapour reduced, promoted, or had no effect on INA below, near, or above water saturation, respectively (Sullivan et al., 2010a). Treatment of kaolinite, illite, K-feldspar and ATD with H 2 SO 4 vapour and heat (70-250 • C) reduced deposition and/or immersion mode INA (Sullivan et al., 2010b;Wex et al., 2014;Augustin-Bauditz et al., 2014). Additionally, contact with dilute alkali salt solutions (K 2 SO 4 , Na 2 SO 4 , KCl, NaCl; 10 −4 to 1 M) has been shown to decrease the INA of some silicates including K-and Na/Ca-feldspars compared to when they are immersed in pure water (Whale et al., 2018;Kumar et al., 2018Kumar et al., , 2019a. This has been proposed to reflect a suppressed exchange, in the presence of dissolved K + and Na + , of alkali cations from the silicate surfaces with H + /H 3 O + from solution, a process that may be important in ice nucleation by these materials (Kumar et al., 2018(Kumar et al., , 2019a. Further, dissolved SO 2− 4 may form complexes with Al and thereby promote Al dissolution and -OH group removal from the silicates, poten-tially blocking or destroying surface sites at which ice nucleation occurs (ice-active sites) (Kumar et al., 2018(Kumar et al., , 2019a. Collectively, these findings suggest that in-plume exposure to acidic gases and emplacement of soluble metal sulphate and halide salts might depress the INA of volcanic ash. However, the conditions simulated in various 'atmospheric processing' studies on dust differ substantially from the extreme thermal and chemical conditions to which ash is exposed in an eruption plume, calling for dedicated investigation of the effects of high temperature ash-gas interaction on the INA of ash. Heating of silicates (e.g., silica and soda-lime-silica glass, Na/Cafeldspar, quartz, kaolinite) >400 • C can result in condensation of adjacent Si-OH groups to form Si-O-Si bridges, producing a more hydrophobic and compact silica-like surface (Schaeffer et al., 1986;Temuujin et al., 1999;D'Souza and Pantano, 2002). This removal of surface -OH groups (also known as dehydroxylation) can be enhanced by the presence of H 2 O vapour (Temuujin et al., 1999) and can become irreversible as the heating temperature is raised >400 • C (D' Souza and Pantano, 2002). Powdered K-feldspar can also be converted to an amorphous state when 'highly heated,' although no temperature is stated in the patent describing this effect (Swayze, 1907). Additionally, Fe-containing silicates undergo oxidation of Fe 2+ to Fe 3+ at high temperatures (550-1100 • C), accompanied by an outward diffusion of divalent cations (e.g., Mg 2+ , Fe 2+ , Ca 2+ ) (Wu et al., 1988;Cooper et al., 1996). Oxidation of Fe 2+ likely reduces the mobility of alkali cations such as Na + and K + , as these cations may become fixed in a charge-compensating role to accommodate newly-formed Fe 3+ in the silicate network (Cooper et al., 1996;Pelte et al., 2000).
Exposure of silicates (e.g., multi-oxide silicate glasses, Na/Cafeldspar, pyroxene, Kilauea basalt, Mono Craters obsidian) to SO 2(g) and HCl (g) at high temperatures can lead to the formation of metal sulphate and chloride salts as reaction products (Delmelle et al., 2018;King et al., 2018;Renggli et al., 2019). For example, SO 2(g) interacts with soda-lime-silica glass and volcanic glasses at 400 to 800 • C to produce primarily Na 2 SO 4 and CaSO 4 , respectively (Douglas and Isard, 1949;Ayris et al., 2013;Casas et al., 2019). These sulphate compounds are argued to reflect the reaction of adsorbed SO 2 with alkali or alkaline earth cations (e.g., Na + , K + , Ca 2+ , Mg 2+ ) at the silicate surface. Further, HCl (g) interacts with soda-lime-silica, mixed-alkali borosilicate, and volcanic tephrite and phonolite glasses at 300 to 700 • C to produce primarily NaCl (Schaeffer et al., 1986;Sung et al., 2009;Ayris et al., 2014). This is proposed to involve the reaction of Cl − from adsorbed HCl with Na + at the glass surface. Differences in cation diffusivities, which vary as a complex function of factors such as silicate composition and heating temperature, are thought to dictate which particular salts form Renggli et al. (2019). Cation extraction by reaction with SO 2(g) or HCl (g) at high temperatures generates a compacted, cation-depleted and silica-enriched surface beneath the salts (Renggli et al., 2019;Ayris et al., 2014;Li et al., 2010). This process of 'dealkalisation' is often exploited in the glass industry to enhance surface resistance to chemical weathering, for instance by blocking further diffusive exchange of cations (Schaeffer et al., 1986). While these studies shed light on how various high temperature gas-solid interactions might alter the surface properties of volcanic ash, the influence of such interactions on the ability of ash to nucleate ice has yet to be explored. Here we expose three compositionally distinct tephra samples, a K-feldspar sample, and a quartz sample to gas mixtures comprising H 2 O (g) , SO 2(g) , and/or HCl (g) under a heating sequence of 800 and then 400 • C, and assess the INA of the non-treated and treated samples in the immersion mode using cold stage droplet assays. By uniquely combining in-plume gassolid interaction simulations with ice nucleation measurements of various silicate materials, this experimental study contributes new insights on the potential impacts of eruption plume processing on the INA of volcanic ash before it disperses in the eruption cloud and wider atmosphere.

Silicate samples
Three volcanic tephra samples (from eruptions of Tungurahua, Astroni and Etna volcanoes), a K-feldspar (BCS 376 microcline), and a crystalline silica (smoky quartz) were studied (Table 1). These silicate materials have been characterised in terms of chemical and/or mineralogical composition as well as INA in previous studies (Maters et al., 2019;Atkinson et al., 2013;Harrison et al., 2016Harrison et al., , 2019. As in those studies, the materials here were crushed to fine powders in a ball mill using a zirconia ceramic ball and vial, but were not otherwise subjected to physical or chemical alteration, prior to eruption plume processing treatments (see section 2.2). Milling of the materials exposed fresh surfaces to test the influence of gas-solid interaction on INA under controlled conditions in the laboratory. Any pre-existing salts on the tephra would have remained in the powders but only on a small fraction of surfaces following milling. The specific surface area (SSA BET ) of the non-treated and treated samples was obtained, after overnight degassing, from a 10-point N 2 adsorption isotherm at -196 • C based on the Brunauer, Emmet, and Teller model.

Eruption plume processing treatments
An Advanced Gas-Ash Reactor (AGAR), capable of simulating thermal and chemical properties found within the hot core of a volcanic eruption plume (Ayris et al., 2015), was used to subject the five silicate materials to high temperature gas-solid interaction. Specifically, sub-samples of these materials were heated under three distinct gas compositions in the AGAR, and the influence of thermochemical treatment on the materials' INA was subsequently assessed by ice nucleation experiments (see section 2.3). Note that the treatments are not designed to reproduce the actual experience of ash particles in an eruption plume -being unable to replicate the precise conditions and timescales of in-plume processing (Ayris et al., 2015)  were chosen for this investigation, as these species are among the main gases co-erupted with volcanic ash. Sequential heating at 800 and then 400 • C was applied to cover the in-plume temperature ranges in which SO 2(g) and HCl (g) have been reported to react most efficiently with an array of silicate glasses and minerals (King et al., 2018;Ayris et al., 2013Ayris et al., , 2014. Gas delivery to the AGAR consisted of a 50 sccm flow of Ar (g) containing H 2 O (0.4 ml min −1 inlet flow) aerosolised by a quartz nebuliser, and a 100 sccm flow of Ar (g) on its own or mixed (75:25 sccm) with SO 2(g) (1 mol.% in Ar (g) ) or HCl (g) (1 mol.% in Ar (g) ). Excluding the contribution of the inert Ar (g) carrier, the mixed gas flows produced compositions of 62 to 100 mol.% H 2 O (g) and 1 mol.% SO 2(g) or 1 mol.% HCl (g) , consistent with the ranges of measured near-vent volcanic gas concentrations (Ayris et al., 2015). Each sub-sample (∼2 g) of volcanic tephra, K-feldspar, or quartz was loaded into the sample bulb, inserted horizontally into the working tube of the AGAR's three-stage furnace, and rotated continuously (7 rpm) to overturn the powder during exposure to heat and gases. A fast insertion protocol was applied in all experiments. This consisted of sliding the sample bulb into the furnace region at ∼800 • C; turning on the nebuliser, gas flows, and rotation for 600 s; then sliding the sample bulb into the furnace region at ∼400 • C for 600 s; and finally turning off the nebuliser, gas flows, and rotation and removing the sample bulb from the reactor. The treated sample was allowed to cool to room temperature in air and then stored in a sealed vessel until further experiments and analyses (e.g., see sections 2.3 and 2.4).

Ice nucleation experiments
A microlitre Nucleation by Immersed Particles Instrument (μL-NIPI) was used to assess the INA of the non-treated and treated samples. This instrument has been detailed by Whale et al. (2015) and used in various studies of ice nucleation by silicate materials (Maters et al., 2019;Atkinson et al., 2013;Harrison et al., 2016Harrison et al., , 2019. Briefly, a 1 wt.% suspension of sub-sample in Milli-Q water (18.2 M ·cm) was shaken for a few minutes by a vortex mixer, and then pipetted in an array of 1 μL droplets (30-45) onto a silanised glass cover slip on a temperature-controlled stage (Grant-Asymptote EF600 Stirling Cryocooler). The stage was cooled at a rate of -1 • C min −1 below 0 • C until all droplets were frozen. A dry N 2(g) flow (∼0.2 L min −1 ) was applied over the droplets to prevent condensation and frost accumulation on the cover slip and, thus, to avoid frozen droplets affecting neighbouring liquid droplets. A digital camera was used to monitor the droplets during the experiment and to determine the fraction of droplets frozen as a function of temperature f ice (T ), used to calculate the ice nucleation active site density n s (T ), i.e., the number of ice-active sites per unit surface area of a solid sample on cooling from 0 • C down to temperature (T ) (Connolly et al., 2009): where n ice (T ) is the cumulative number of droplets frozen at temperature (T ), n is the total number of droplets in the experiment, and A is the total surface area of the solid sample per droplet. The parameter n s (T ) allows us to empirically define the INA of a range of materials, even though the fundamental nature of ice-active sites remains unclear. The uncertainty in n s (T ) was calculated using Monte Carlo simulations of possible ice-active site distributions across droplets, and incorporating Poisson counting statistics to account for sampling errors, as in Harrison et al. (2019). Experiments were conducted at least in triplicate for each sample. Additionally, ice nucleation of the background water due to impurities in the water or effects of the cover slip was assessed by baseline droplet freezing measurements of water containing no added sample. The n s (T ) data for each sample experiment were corrected to account for the potential background contribution to ice nucleation as in Vali (2019).

Water leachate analyses
Water leachate analysis can contribute to identifying and quantifying the soluble surface species emplaced during gas-solid interactions, including those occurring within the hot core of a volcanic eruption plume or within a high temperature reactor such as the AGAR (Delmelle et al., 2007;Ayris et al., 2013;Casas et al., 2019;Ayris et al., 2014;Witham et al., 2005). Specifically, leachates of the non-treated and treated samples were acquired here to gain insight on potential soluble surface products arising from the different eruption plume processing treatments (see section 2.2), and to assess the concentrations of dissolved elements and ions likely to be present in the 1 wt.% suspensions of sample in water utilised for ice nucleation experiments (see section 2.3).
Sub-samples of the non-treated and treated materials were rinsed in Milli-Q water at a 1 wt.% concentration and under constant gentle rotation for 30 min, i.e., reflecting a typical duration of an ice nucleation experiment. The leachates were then filtered through 0.2 μm cellulose acetate membrane filters and divided into three aliquots; one directly for measurement of SO 2− 4 and Cl − concentrations by ion chromatography (Thermo Scientific Dionex ICS-5000 + ), another acidified to 1.5 vol.% HNO 3 for measurement of Na, K, Ca, Mg, and Fe concentrations by inductively coupled plasma -optical emission spectrometry (Thermo Scientific iCAP 7400 Radial ICP-OES), and the remainder stored in a sealed vessel under refrigeration (∼4 • C). Quantification limits were 0.2 μM for SO 2− 4 , 0.8 μM for Cl − , 7 μM for Na, 3 μM for K, 0.3 μM for Ca, 0.05 μM for Mg, and 0.1 μM for Fe. Additionally, water-rinsed solid sub-samples retained from the leachates by filtration were left to dry in air overnight (under weighing paper) and then stored in a sealed vessel. The leachates and the water-rinsed sub-samples of the treated tephra were used later to investigate the influence of soluble surface species on the INA of the tephra (see sections 4.2 and 4.3).

Ice-nucleating activity
All high temperature gas-solid interactions were expected to deactivate the INA of the silicate materials based on the literature surveyed in the introduction. However, the behaviour of these materials following the eruption plume processing treatments varied considerably, with an enhancement in INA observed in some cases. Fig. 2 shows the n s (T ) values of the non-treated and treated samples from replicate ice nucleation experiments. The difference in INA between the treated and non-treated samples is summarised in Fig. 3 (with the temperature at which n s ≈ 5 cm −2 , T ns≈5cm −2 , chosen as a single-number proxy for INA; Table S1 in the SI).
For all materials, the H 2 O treatment decreased the INA relative to that of the non-treated sample (T ns≈5cm −2 lowered by 1.7 ± 3.0 to 15.0 ± 1.8 • C), although this decrease is within the error for ETN tephra. Similarly, the H 2 O − HCl treatment decreased the INA relative to that of the non-treated sample (T ns≈5cm −2 lowered by 4.0 ± 3.7 to 16.2 ± 1.9 • C), to the same extent as the decrease induced by the H 2 O treatment (tephra and quartz) or more (K-feldspar). In  Overall, while observations for the K-feldspar and quartz are consistent with our hypothesis of deactivation of ice nucleation by heat and/or acids, the increased INA of the tephra exposed to H 2 O (g) mixed with SO 2(g) at high temperatures is striking and is contrary to expectations.

Water leachate chemistry
The chemical composition of the water leachates, in terms of the elements and ions released per unit surface area of the silicate samples, is shown in Fig. 4. For all five silicates, the non-treated and H 2 O -treated sample leachates contain variable amounts of Na, K, Ca, Mg, Fe, SO 2− 4 and Cl − in concentrations below ∼5 μmol m −2 (Fig. 4a, b). For the non-treated tephra, some of these dissolved species may originate from small amounts of pre-existing salt compounds present in the milled powders. Similar quantities (<5 μmol m −2 ) of elements and ions are seen in the leachates of  (Table S1). Sample codes are listed in Table 1. ( Fig. 4c, d), indicating no substantial difference in the amount of soluble material on these silicates irrespective of gas-solid interaction. Conversely, for the three tephra, considerably more soluble material is present after exposure to SO 2(g) or HCl (g) in addition to H 2 O (g) at high temperatures (Fig. 4c,

Discussion
Based on previous studies on mineral dust, it was expected that the INA of the silicate materials here would be universally deactivated by eruption plume processing, but instead these materials exhibited highly variable responses to the three thermochemical treatments. These responses are grouped in the following discussion around trying to relate the observed changes in INA to the suspected modifications of the silicate surfaces induced by the different high temperature gas-solid interactions.

H 2 O depresses the INA of the silicate materials
Exposure to H 2 O reduced the INA of all five silicate materials (Fig. 2), however, the dissolved element and ion concentrations in the H 2 O -treated sample leachates do not differ substantially from those in the non-treated sample leachates (Fig. 4a, b). This indicates that interaction between the materials and H 2 O (g) at high temperatures did not lead to the formation of readily soluble compounds on their surfaces, and so the reduced INA is inferred to relate to some other chemical and/or physical change to the silicate surfaces during H 2 O treatment.
Partial surface dehydroxylation of the H 2 O -treated samples at temperatures >400 • C might explain the decrease in their INA relative to the non-treated samples. Specifically, surface -OH groups are thought to play an important role in heterogeneous ice nucleation as they can engage in hydrogen bonding with H 2 O molecules to facilitate ice formation (Hu and Michaelides, 2007;Pedevilla et al., 2017). Hence, irreversible condensation of some surface Si-OH groups to Si-O-Si bridges during heat treatment likely reduced the abundance of ice-active sites on the silicate materials, since O atoms in Si-O-Si lack sufficient electronegativity for hydrogen bonding. Variation in the extent of INA reduction observed for the H 2 O -treated samples (Fig. 3) probably reflects differences in individual properties of the silicates, such as the abundance and arrangement of -OH groups, which render their surfaces variably susceptible to the effects of high temperatures in the presence of H 2 O (g) .
Further, surface compaction (by dehydroxylation) of the H 2 Otreated samples and/or oxidation of Fe 2+ to Fe 3+ (where present) might have inhibited the exchange of alkali cations such as Na + and K + with aqueous H + /H 3 O + on input of the samples to water during ice nucleation experiments. In other words, these processes might have hindered the surface protonation of immersed solid particles, which has been suggested to be a key step in ice nucleation by silicates bearing exchangeable cations (Kumar et al., 2018(Kumar et al., , 2019a. In some cases, if exposure to high temperatures resulted in the formation of a silica-like surface on the H 2 O -treated samples that is more amorphous (less crystalline) than the original surface (e.g., for K-feldspar) (Swayze, 1907), this might also have contributed to the reduction in INA observed after the H 2 O treatment. Amorphous silicates have been shown, in general, to be less ice-active than crystalline silicates (Maters et al., 2019;Kumar et al., 2019b).
Ice nucleation on silicate materials is thought to be a sitespecific phenomenon, i.e., nucleation occurs at specific sites on the solid surface (Holden et al., 2019). Therefore, the effects of eruption plume processing on INA presumably reflect changes to the properties of these specific sites, rather than changes to the bulk surface. It has been suggested that microscopic patches of highenergy faces, exposed at surface defects (e.g., edges, cracks) and containing a high density of -OH groups, give rise to effective sites for ice nucleation (Harrison et al., 2019;Pedevilla et al., 2017;Kiselev et al., 2017). These patches are typically unstable and tend to undergo thermochemical alteration more readily than the bulk surface (Holdren and Speyer, 1985). By analogy, dehydroxylation and/or oxidation during H 2 O treatment might have occurred preferentially at high-energy sites on the silicate surfaces, potentially where elevated hydrogen bonding and/or protonation capabilities play a role in ice nucleation.

H 2 O-HCl depresses the INA of tephra
As observed for H 2 O , exposure to H 2 O − HCl reduced the ability of the silicate materials to nucleate ice (Fig. 2), with the depression in INA elicited by the H 2 O − HCl treatment either comparable to (for tephra and quartz) or greater than (for K-feldspar) the depression in INA elicited by the H 2 O treatment alone (Fig. 3).
The three tephra display significantly higher concentrations of several dissolved elements and ions (Na, Ca, Cl − , SO 2− 4 ) in the H 2 O − HCl-treated sample leachates than in the non-treated sample leachates (Fig. 4a, d), suggesting that interaction with H 2 O (g) and HCl (g) at high temperatures led to the emplacement of readily soluble compounds on the tephra surfaces. In particular, the relative molar abundances of various dissolved species, with Na:Cl − ratios ranging from 0.7:1.0 to 0.9:1.0 and Ca:SO 2− 4 ratios ranging  (Ayris et al., 2014). Despite reaction of the tephra material with HCl (g) to produce readily soluble surface salts (Fig. 4d), however, there appears to be no greater adverse effect of treatment with HCl (g) in addition to H 2 O (g) on the INA of the tephra (Figs. 2, 3). Presumably, the H 2 O − HCl-treated tephra surfaces underlying such salts, once immersed in water, exhibit similar icenucleating properties to the H 2 O -treated tephra surfaces; with both sets likely featuring a more oxidised, compact and silica-rich layer with fewer sites for hydrogen bonding and/or ion exchange.
Dissolved alkali chlorides and sulphates have been shown to lower the INA of various silicates (Whale et al., 2018;Kumar et al., 2018Kumar et al., , 2019a. Hence, one might have expected a solute-related suppression of INA following dissolution of metal chloride and sulphate salts from the H 2 O − HCl-treated tephra in water (Fig. 4d) during ice nucleation experiments. To investigate whether the soluble salts produced by eruption plume processing could be responsible for modifying the volcanic tephra's ability to nucleate ice, supplementary experiments were carried out on TUN-, ASTand ETN tephra; measuring the INA of the non-treated tephra suspended in the H 2 O − HCl-treated tephra leachates in addition to those suspended in Milli-Q water (Fig. 5). The similarity in INA of the non-treated tephra suspended in Na/Ca/Cl − /SO 2− 4 -rich leachate (10 −2 to 10 −1 M; Fig. 4d) and in pure water suggests that solute ions originating from metal chloride and sulphate salts are unlikely to explain the depressed INA of the H 2 O − HCl-treated tephra. Moreover, the water-rinsed H 2 O − HCl-treated tephra recovered after leaching and filtration (see section 2.4) also display a reduced INA relative to the non-treated tephra (Fig. 6). These tests show that the decreased ability of tephra exposed to H 2 O (g) and HCl (g) at high temperatures to nucleate ice does not relate simply to its soluble surface salts, but rather to irreversible changes in the tephra surface underlying these salts.

H 2 O-SO 2 enhances the INA of tephra
It is striking that the H 2 O − SO 2 -treated tephra displayed an enhanced INA relative to the non-treated and/or H 2 O -and H 2 O − HCl-treated tephra (Figs. 2, 3). From these observations, it is inferred that reaction of the tephra with SO 2(g) altered these materials in such a way that either partially offsets or entirely overcomes the negative influence of exposure to H 2 O (g) alone at high temperatures on their ability to nucleate ice (see section 4.1).
We first address the possibility that this reflects a positive influence of solute ions on INA following dissolution of metal sulphate salts from the H 2 O − SO 2 -treated tephra in water (Fig. 4c) during ice nucleation experiments. The three tephra display significantly higher concentrations of most dissolved elements and ions (Na, K, Ca, Mg, SO 2− 4 ) in the H 2 O − SO 2 -treated sample leachates than in the non-treated sample leachates (Fig. 4a, c), implying that interaction with H 2 O (g) and SO 2(g) at high temperatures resulted in the emplacement of readily soluble compounds on the tephra surfaces. Specifically, the relative molar abundances of various dissolved species, with (Na/2+K/2+Ca+Mg):SO 2− 4 ratios of 1.1:1.0, may indicate their occurrence in the form of metal sulphate salts on the H 2 O − SO 2 -treated tephra. For illustration, surficial deposits rich in Ca and S, likely corresponding to anhydrite (anhydrous  Table 1. CaSO 4 ) (Renggli et al., 2019;Ayris et al., 2013), can be seen on the H 2 O − SO 2 -treated AST tephra by scanning electron microscopyenergy dispersive X-ray spectroscopy (SEM-EDX; Fig. 7; see the SI for details). Such deposits are absent on the surface of the nontreated AST tephra.
Dilute solutions of ammonium salts (e.g., (NH 4 ) 2 SO 4 , NH 4 Cl; 10 −4 to 1 M) have been shown to increase the INA of some silicates including kaolinite, muscovite, biotite, K-feldspar, Na/Cafeldspar, and quartz relative to their INA in pure water (Whale et al., 2018;Kumar et al., 2018Kumar et al., , 2019a. This has been hypothesised to relate to hydrogen bonding of NH + 4 or NH 3 with -OH groups on the silicate surfaces, in turn promoting hydrogen bonding with H 2 O molecules in preferred orientations for ice formation (Kumar et Table 1. evaluate whether a solute effect could be responsible for modifying the tephra's ability to nucleate ice, supplementary experiments were performed with TUN-, AST-, and ETN tephra; measuring the INA of the non-treated tephra suspended in the H 2 O − SO 2 -treated tephra leachates in addition to those suspended in Milli-Q water (Fig. 5). The similarity in INA of the non-treated tephra suspended in Na/K/Ca/Mg/SO 2− 4 -rich leachate (10 −1 to 1 M; Fig. 4c) and in pure water implies that solute ions on their own are unlikely to explain the altered INA of the H 2 O − SO 2 -treated tephra. Moreover, the H 2 O − SO 2 -treated tephra do not bear ammonium salts as reaction products, and so a positive influence of dissolved NH + 4 or NH 3 on INA is discounted here. Interestingly, the water-rinsed H 2 O − SO 2 -treated tephra recovered after leaching and filtration (see section 2.4) continue to display a higher INA relative to the other treated tephra, although rinsing did reduce the INA of the (Fig. 6). Therefore, the H 2 O − SO 2treated tephra presumably exhibit some new or altered surface feature that makes them more ice-active than the other treated tephra, despite the likelihood that they all have in common a more oxidised, compact and silica-rich surface layer.
Some surficial CaSO 4 was most likely retained on the H 2 O − SO 2 -treated tephra even when immersed in water, as the solubility of this compound (∼0.2 g 100 mL −1 at 20 • C and 1 atm) is low compared to other alkali and alkaline earth salts (e.g., Na 2 SO 4 , K 2 SO 4 , MgSO 4 , NaCl, KCl, CaCl 2 solubilities are ∼two orders of magnitude higher; Weast and Lide, 1989) and it might not have dissolved completely on the timescale of the immersion freezing experiments. Therefore, we speculate that surficial CaSO 4 contributed to enhancing the INA of the tephra, with this effect partially offset in AST tephra by strong deactivation of its K-feldspar component by the H 2 O − SO 2 treatment (Table 1, Fig. 2). In order to test the INA of anhydrite, a natural anhydrite specimen from Staffordshire, UK was chosen and found to exhibit a high INA; inducing freezing of water droplets from -4.3 to -10.8 • C ( Figure S1 in the SI). There is also strong evidence for coal fly ash that anhydrite, prior to its dissolution in water, promotes ice nucleation at the ash-water interface (Havlíček et al., 1993;Grawe et al., 2018). A correlation has been observed between the amount of Ca-and S-containing compounds and the immersion freezing efficiency of dry-generated coal fly ash samples (Grawe et al., 2018). However, three H 2 O − SO 2 -treated volcanic glasses additionally tested here, and which similarly appear to bear surficial CaSO 4 , do not display a clear increase in INA compared to non-treated glasses ( Figure S2 in the SI). This might indicate that the role of CaSO 4 in triggering ice nucleation depends on the INA of the original surface, as the non-treated glasses are poorly ice-active to begin with ( Figure  S2 in the SI) (Maters et al., 2019). Perhaps a combination of differing surface mineralogical and morphological properties of pure glass and mixed glassy/crystalline tephra determines if and how CaSO 4 formed on the H 2 O − SO 2 -treated materials acts to influence ice nucleation. Havlíček et al. (1993) suggest that the activity of CaSO 4 (and other sulphate salts) in ice nucleation by coal fly ash is affected by the morphology of the ash particles, in particular by the configurations of insoluble and soluble components of the ash surfaces. This conforms to the notion that it is a certain combination of surface chemistry and topography that gives rise to ice-active sites on silicate materials (

H 2 O-HCl and H 2 O-SO 2 depress the INA of K-feldspar and quartz
For K-feldspar, the adverse effects of H 2 O − HCl and H 2 O − SO 2 exposure on INA exceed that of H 2 O exposure alone (Figs. 2, 3), revealing that contact with HCl (g) or SO 2(g) has some additional negative influence on the ability of this material to nucleate ice. In contrast to the tephra, the K-feldspar shows no substantial difference in dissolved element or ion concentrations in leachates of the H 2 O − HCl-treated or H 2 O − SO 2 -treated samples and that of the non-treated sample (Fig. 4a, c, d). This suggests that interaction of K-feldspar with H 2 O (g) and HCl (g) or SO 2(g) at high temperatures did not involve reaction to produce readily soluble compounds (e.g., KCl, NaCl, K 2 SO 4 , Na 2 SO 4 ) on its surface. In the case of HCl (g) , the reason for this is not known and warrants investigation, as we might have expected Na + and K + in the feldspar to exchange with H + and react with Cl − . In the case of SO 2(g) , this observation is consistent with thermodynamic modelling by King et al. (2018), which predicted that no sulphate salts formed on K-feldspar exposed to SO 2(g) at 600 • C, although the solid-state controls behind this finding are poorly understood and require further research. The modelling predicted instead a significant amount of SiO 2 formed on the treated K-feldspar (10 mol SiO 2 :1 mol K-feldspar), which might have acted as a diffusion barrier against cation outfluxing, thereby hindering reaction with surface adsorbed SO 2 (King et al., 2018). As under the H 2 O treatment, heating of the K-feldspar during the H 2 O − HCl and H 2 O − SO 2 treatments might have caused a loss of ice-active sites as its surface becomes less crystalline (Swayze, 1907), with this effect perhaps enhanced by the acidic HCl (g) and SO 2(g) (King et al., 2018). If so, the development of a more extensive amorphous silica-like surface on K-feldspar under HCl (g) Kumar et al. (2019a) noted an inverse correlation between the thickness of the amorphous silica-enriched surface layer on Kfeldspar and its INA (defined by the onset freezing temperature) following exposure of this material to the aqueous acid H 2 SO 4 (Kumar et al., 2018).
In contrast to the tephra and K-feldspar, exposure of quartz to HCl (g) or SO 2(g) appears to have no additional bearing on the ability of this material to nucleate ice compared to when exposed only to H 2 O (g) at high temperatures (Figs. 2, 3). Moreover, it appears that some ice-active surface sites on quartz are not affected whatsoever by treatment with H 2 O − SO 2 , i.e., n s (T ) values of the H 2 O − SO 2treated quartz are comparable to those of the non-treated quartz at temperatures above ∼-15 • C (Fig. 2). The apparent insensitiv-ity of the INA of quartz to contact with HCl (g) or SO 2(g) in addition to H 2 O (g) might relate to the lack of exchangeable cations in this material. In contrast to the tephra, the quartz exhibits similar dissolved element and ion concentrations in leachates of the H 2 O − HCl-treated, H 2 O − SO 2 -treated, and non-treated samples (Fig. 4a,c,d), indicating that quartz interaction with H 2 O (g) and HCl (g) or SO 2(g) at high temperatures did not involve reaction to produce readily soluble compounds on its surface. The absence of soluble chloride or sulphate salts on quartz exposed to HCl (g) or SO 2(g) is not surprising, as this mineral is practically devoid of alkali or alkaline earth cations with which to react. Such metal cations are hypothesised to act as Lewis acid surface sites involved in HCl (g) chemisorption (through the Cl − end) and to be associated with Lewis base O 2− or OH − surface sites involved in SO 2(g) chemisorption on multi-oxide silicates including volcanic ash and glass (Delmelle et al., 2018;Maters et al., 2017). In fact, fused quartz is considered relatively inert and thermally stable, and for these reasons makes up components of the AGAR device and other high temperature reactors (Ayris et al., 2013;Sung et al., 2009;Li et al., 2010;Ayris et al., 2015). Therefore, as under H 2 O treatment, the adverse effect of H 2 O − HCl and H 2 O − SO 2 treatments on the INA of the quartz material here is inferred to relate to the creation of a partially dehydroxylated and more compact surface, which bears fewer sites for hydrogen bonding with water and/or taking up protons in aqueous solution.

Remarks on potential physical effects
In addition to chemical differences between the treated and non-treated silicate materials, the eruption plume processing experiments might have modified physical properties of the solid samples affecting ice nucleation.
The SSA BET (Table 1) is consistently lower for the treated samples than for the non-treated samples, revealing that the gas-solid interactions at high temperatures reduced solid surface area. The treatments may have caused loss of some features associated with surface roughness such as steps, kinks and cracks, as unstable high-energy patches can disappear due to structural relaxation of silicate surfaces at high temperatures under H 2 O (g) (Agarwal and Tomozawa, 1997). These high-energy features may also be more susceptible (than flat surfaces) to alteration by H 2 O (g) and SO 2(g) or HCl (g) . Moreover, gas-solid reactions can alter physical properties of particles in various ways, with porosity and permeability for instance being created or destroyed as reaction products modify particle architecture (King et al., 2018). In the case of the tephra treated to H 2 O − SO 2 and H 2 O − HCl, the reduction in surface area could partly reflect a masking of internal surface by newlyformed metal sulphates and chlorides, which have a higher molar volume than their corresponding metal oxides (King et al., 2018). Microscopic analysis of silicate glasses treated with SO 2(g) at high temperatures reveals CaSO 4 deposits of various morphologies (e.g., wrinkled, buckled, hillocked, ringed, rounded) (Renggli et al., 2019;Casas et al., 2019), such deposits are similarly seen here on the AST tephra treated to H 2 O − SO 2 (Fig. 7). Reaction products might conceivably also increase surface area by introducing additional topography, but this is not evidenced by the SSA BET data. Given that specific combinations of topographical and chemical features are thought to give rise to ice-active surface sites on silicate materials (Harrison et al., 2019;Holden et al., 2019;Whale et al., 2017), changes in surface texture during the eruption plume processing treatments could contribute to the changes in INA observed (Figs. 3, 4).
It is likely that a combination of effects determines the ability of the silicate materials investigated here to nucleate ice, in a manner that is difficult to disentangle given the widely ranging sample physicochemical properties, and the potentially competing influences of various gas-solid interactions on ice-active surface sites. Future research, for example using simple model systems (e.g., soda-lime-silica beads) to further minimise the number of variables being tested at a time, and surface sensitive analytical techniques to elucidate the changes induced by different plume processing treatments at the scale relevant for ice nucleation, will likely be essential to unravel these effects.

Conclusions and implications
Here we combined high temperature gas-solid interaction experiments with ice nucleation measurements to investigate the effects of eruption plume processing on the INA of volcanic ash. The INA of all five silicate materials was depressed by exposure to H 2 O or H 2 O − HCl. Conversely, the INA of the tephra was enhanced by exposure to H 2 O − SO 2 , or else depressed to a lesser degree than by exposure to H 2 O or H 2 O − HCl. Any reduction in INA observed following gas-solid interaction is proposed to reflect a loss of ice-active sites on the silicates potentially due to dehydroxylation, oxidation and/or cation extraction, producing a more hydrophobic, compact and/or amorphous silica-like surface. In contrast, the increase in INA observed following tephra interaction with H 2 O − SO 2 is hypothesised to reflect the creation of ice-active sites potentially relating to CaSO 4 formation. We speculate that such sites arise from a particular combination of surface chemical and topographical features accompanying CaSO 4 on the tephra, consistent with previous suggestions on the role of CaSO 4 in promoting ice nucleation by coal fly ash (Havlíček et al., 1993;Grawe et al., 2018).
While treatments with H 2 O (g) and SO 2(g) or HCl (g) were performed separately here, eruption plumes contain mixtures of these and other gases (e.g., CO 2(g) ), which might interact in affecting ash surface properties and INA. However, this is not expected to change the conclusion that exposure to H 2 O (g) and SO 2(g) at high temperatures can enhance the ash INA, since CaSO 4 forms even in mixed gas atmospheres, with SO 2(g) , HCl (g) , and CO 2(g) apparently not competing for reactive uptake on ash surfaces (Delmelle et al., 2018;Ayris et al., 2014). Water leachates of ash samples whose natural surfaces have been preserved following collection provide evidence that CaSO 4 is often a dominant salt found on ash from volcanic eruptions worldwide (Witham et al., 2005). Metal salts including CaSO 4 may also form by precipitation following leaching and dissolution of ash surfaces by H 2 O (l) , H 2 SO 4(aq) , and HCl (aq) at lower temperatures (<190 • C) in the condensation zone of the eruption plume (see Fig. 1) (Óskarsson, 1980;Delmelle et al., 2007). Such ash-condensate interaction could additionally modify ash physicochemical properties and INA in ways not explored here, although the window of opportunity might be limited before the ash is propelled to temperatures at which ice nucleates. How ash properties are further altered in this interval between high temperature processing and ice nucleation is a target for future study. The dynamics between solid, gas, and liquid phases in an eruption plume remain poorly understood, largely due to the danger and difficulty of direct observations within this environment (Delmelle et al., 2018). Laboratory and modelling studies have an important role to play in shedding more light on ash-gas/condensate interactions and their potential impacts (e.g., on ice nucleation, gas scavenging, ash aggregation), both in the vicinity of a volcano and further afield.
The observations in this study suggest that a combination of factors drives the overall INA of ash following transit through the eruption plume. Ash bulk properties such as mineralogy derived from the source magma might affect INA directly (Maters et al., 2019;Jahn et al., 2019), but they also influence the changes in ash surface properties elicited by ash-gas/condensate interactions in the plume (e.g., salt formation) (Delmelle et al., 2007(Delmelle et al., , 2018, which likely have an overriding effect on the reactivity of airborne ash including its INA (Maters et al., 2016(Maters et al., , 2017. For example, the observed reduction in INA of AST tephra following H 2 O − SO 2 treatment probably reflects deactivation of its K-feldspar, while no such reduction in INA of TUN-and ETN tephra might suggest that the Na/Ca-feldspar and pyroxene in these samples were not deactivated by H 2 O − SO 2 (Table 1, Fig. 2a-c) (Maters et al., 2019). Bulk mineralogy has similarly been shown to influence the response of dust INA to thermochemical treatment, with illite and kaolinite powders containing K-feldspar being most susceptible to a loss of ice-active sites upon heating with H 2 SO 4 vapour (Wex et al., 2014;Augustin-Bauditz et al., 2014). However, on all three tephra tested here, we infer that surficial CaSO 4 formed by reaction with H 2 O (g) and SO 2(g) promoted ice nucleation. Additional thermochemical treatments of a range of minerals found in ash (e.g., Na/Ca-feldspars, pyroxenes, amphiboles, Fe(-Ti) oxides) are recommended to explore more fully this potential interplay of bulk and surface properties in determining the INA of ash emerging from an eruption plume.
Finally, future experimental research applying surface sensitive analytical techniques (e.g., second harmonic and sum frequency generation spectroscopy) (Anim-Danso et al., 2016) will be valuable to reveal the mechanistic relationships between chemical and physical modifications of silicate surfaces and changes in their INA. The impact of eruption plume processing on condensation freezing and deposition ice nucleation on ash is also worth investigating, especially since surficial salts might differentially affect the INA of ash exposed to H 2 O-subsaturated versus H 2 O-supersaturated regimes (Sullivan et al., 2010a;Wex et al., 2014). This could be important for predicting the potential of airborne ash to nucleate ice under different conditions of temperature and relative humidity encountered in the troposphere or tropopause following an eruption. The INA of ash bearing CaSO 4 is likely to be diminished with time immersed in water (e.g., in cloud droplets), as the anhydrite dissolves and/or develops an adsorbed layer of hydrated ions (Grawe et al., 2018). Lastly, transdisciplinary field studies collecting ash particles and complementary volcanic and atmospheric measurements at different distances from the crater could help to improve understanding of how the INA of ash evolves as it disperses away from the eruption plume into the wider atmosphere, where it is susceptible to further physicochemical processing involving low temperature heterogeneous interactions.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
The data associated with this paper is available from the University of Leeds at https://doi .org /10 .5518 /890.