The importance of crystalline phases in ice nucleation by volcanic ash

Volcanic ash is known to nucleate ice when immersed in supercooled water droplets. This process may impact the properties and dynamics of the eruption plume and cloud, as well as those of meteorological clouds once the ash is dispersed in the atmosphere. However, knowledge of what controls the ice-nucleating effectiveness (INE) of ash remains limited, 10 although it has been suggested that crystalline components in ash may play an important role. Here we adopted a novel approach using nine pairs of tephra and their remelted and quenched glass equivalents to investigate the influence of chemical composition, crystallinity and mineralogy on ash INE in the immersion mode. For all nine pairs studied, the crystal-bearing tephra nucleated ice at higher temperatures than the corresponding crystal-free glass, demonstrating that crystalline phases are key to ash INE. Similar to findings for desert dust from arid and semi-arid regions, the presence of feldspar minerals 15 characterises the four most ice-active tephra samples, although a high INE is observed even in the absence of alkali feldspar in samples bearing plagioclase feldspar and orthopyroxene. There is evidence of a potential indirect relationship between chemical composition and ash INE, whereby a magma of felsic to intermediate composition may generate ash containing iceactive feldspar minerals. This complex interplay between chemical composition, crystallinity, and mineralogy could help partly to explain the variability in volcanic ash INE reported in the literature. Overall, by categorically demonstrating the importance 20 of crystalline phases in the INE of volcanic ash, our study contributes insights essential for better appraising the role of airborne ash in ice formation. Among these is the inference that glass-dominated ash emitted by the largest explosive eruptions may be less effective at impacting ice-nucleating particle populations than crystalline ash generated by smaller, more frequent eruptions.


35
By definition, volcanic ash consists of pyroclastic particles <2 mm in diameter usually dominated by aluminosilicate glass derived from the melt and/or aluminosilicate and Fe(-Ti) oxide minerals in the form of crystals suspended within the original melt (Heiken and Wohletz, 1992). Field and laboratory measurements present conflicting evidence as to the ice-nucleating effectiveness (INE) of ash, even for samples from the same volcano, and it is far from clear what drives this variation (Durant Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1326 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 January 2019 c Author(s) 2019. CC BY 4.0 License. et al., 2008, and references therein;Mangan et al., 2017). Studies on desert dust from arid and semi-arid regions (1000-3000 Tg a -1 emitted; Penner et al., 2001) -considered one of the most important INP types globally (Hoose et al., 2010;Vergara-Temprado et al., 2017) -suggest that chemical composition, crystallinity and mineralogy can influence the abundance of iceactive surface sites on the solid particles (Murray et al., 2012, and references therein). Specifically, the presence of K-rich feldspar is thought to dominate the INE of dust (Atkinson et al., 2013;Yakobi-Hancock et al., 2013;Kaufmann et al., 2016).
To date however, the roles and potential interplay of differing physicochemical attributes in determining a solid particle's INE remain poorly understood, having rarely been systematically investigated for any ice-nucleating material.
Here we examine the influence of three properties dictated primarily by the state of the erupted source magma -chemical composition, crystallinity, and mineralogy -on the INE of volcanic ash in the immersion mode which is likely relevant to ash 10 particles in the water-rich eruption plume/cloud and in mixed-phase atmospheric clouds (Textor et al., 2006;McNutt and Williams, 2010;Pruppacher and Klett, 2010;Murray et al., 2012). To assist in disentangling the individual effects of these properties on ice nucleation, we have adopted a novel approach of sample selection by using a wide range of natural tephra and their remelted and quenched glass equivalents. In this manner, we aim to contribute to improving understanding of the potential for airborne ash from different eruptions to impact ice formation above the volcanic vent and/or once dispersed in 15 the ambient atmosphere.

Volcanic tephra and glass samples
Eighteen tephra and (remelted and quenched) glass powders spanning a range of compositions associated with volcanic activity were studied (Table 1). The powders are represented by an abbreviated sample code with the subscripts 'teph' or 'glass' used 20 to designate tephra or glass material, respectively. The tephra samples correspond to ash or pumice originating from different eruptions. The glass samples were synthesised by melting a portion of the tephra at 1400 to 1600 °C, homogenising the melts by stirring for 12 to 72 h, and quenching the melts at room temperature to form glasses. This technique for generating volcanic glass has been used previously in studies of volcanic ash reactivity (e.g., Ayris et al., 2013;2014;Maters et al., 2016). All samples were crushed to fine powders in a ball mill using a zirconia ceramic ball and vial, to ensure consistent treatment of 25 the tephra and glass materials prior to ice nucleation experiments. The specific surface area (SSABET ; Table 1) of the samples after overnight degassing was obtained from a ten-point N2 adsorption isotherm at -196 °C based on the Brunauer, Emmet and Teller model (Brunauer et al., 1938) using a Micromeritics TriStar 3000 instrument.
The chemical composition of volcanic ash predominantly reflects the state of the source magma at the point of eruption but can also be influenced by lithic material (pre-existing country rock) entrained during the explosive eruption (Heiken and 30 Wohletz, 1992). The composition of the tephra and glass samples studied here (Table S1) was measured by Nora Groschopf (Institute of Geosciences, Johannes Gutenberg University Mainz, Germany) by X-ray fluorescence at 3.2 kW ionisation energy using a Philips Analytical MagiX PRO spectrometer. The conventional classification of these materials according to total alkali (Na2O + K2O) versus silica (SiO2) content is shown on a Total Alkali versus Silica diagram in Fig. 1a (Le Maitre et al., 2002).
As magma ascends to the surface prior to a volcanic eruption, the aluminosilicate melt typically carries a cargo of mineral 35 species in the form of crystals predominantly originating from the melt. Upon magma fragmentation, the ash generated comprises a corresponding mixture of glass and crystal components (Heiken and Wohletz, 1992). The crystallinity refers to the relative abundance of crystals in ash (i.e., crystal mass/total mass) and typically ranges from 0 to 65 wt.%, depending on factors such as the prior state of the magma (e.g., chemical composition, temperature) and even the dynamics of the conduit such as the ascent rate of the magma (Heiken and Wohletz, 1992;Wright et al., 2012). The incorporation of lithic material 40 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1326 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 January 2019 c Author(s) 2019. CC BY 4.0 License. from the conduit or vent and/or the interaction with ground or surface water can also influence the crystallinity of the bulk erupted ash.
The mineralogy refers to the identities and abundances of crystalline phases in ash. Among the factors that influence crystallisation from the melt, the chemical composition of the magma is a key determinant of the mineral phases that can form (Heiken and Wohletz, 1992). Common minerals in basaltic ash include pyroxene, olivine, amphibole, and (Ca-rich) plagioclase 5 feldspar, whereas those in rhyolitic ash include quartz, mica, amphibole, (Na-rich) plagioclase feldspar, and (K-rich) alkali feldspar ( Fig. 1b; Rogers, 2015). Note that in this study, we use the terms 'plagioclase' and 'alkali' to refer specifically to Na-/Ca-rich and K-rich feldspars, respectively.
The proportions of various crystalline phases in the tephra studied here (Table 2) were determined by X-ray diffraction (XRD) with a CuKɑ1 X-ray beam using a GE diffractometer (XRD 3003 TT) and the Profex software program for Rietveld refinement 10 (Döbelin and Kleeberg, 2015). Briefly, this involved crushing the tephra and spiking it with a known mass (~17 wt.%) of pure Si powder, as crystalline internal standard, to quantify the crystallinity and mineralogy of the samples. While the LIPteph and CIDteph crystallinities cannot be quantified below the ~2 wt.% detection limit of the technique, this does not rule out the possibility of smaller amounts of crystals and/or nanoscale crystallites being present in these samples. The glasses were also analysed by XRD to confirm their amorphous nature (i.e., the absence of crystalline minerals within the ~2 wt.% detection 15 limit).

Immersion mode ice nucleation experiments
The INE of the tephra and glass samples was assessed using a microlitre Nucleation by Immersed Particles Instrument (µL-NIPI). This instrument has been described in detail elsewhere (Whale et al., 2015) and has been used previously to study heterogeneous ice nucleation by mineral and ash material (e.g., Atkinson et al., 2013;Harrison et al., 2016;Mangan et al., 20 2017;Whale et al., 2017). Briefly, a 1 wt.% sample suspension in Milli-Q water (18.2 MΩ·cm) was shaken for a few minutes by a vortex mixer, and then pipetted in an array of 30 to 40 1 µL droplets onto a hydrophobic silanised glass cover slip placed on a temperature-controlled stage (Grant-Asymptote EF600 Stirling Cryocooler). The stage was cooled from room temperature at a rate of -5 °C min -1 down to 0 °C, and subsequently at a rate of -1 °C min -1 until all droplets were frozen. A dry nitrogen flow (~0.2 L min -1 ) over the droplets prevented condensation and frost accumulation on the cover slip and hence served to 25 avoid frozen droplets affecting neighbouring liquid droplets. A digital camera was used to observe the droplets throughout the experiment and determine the fraction of droplets frozen as a function of temperature fice(T) according to: where nice(T) is the cumulative number of droplets frozen at temperature (T) and n is the total number of droplets in the experiment. At least three replicate experiments were conducted for each sample. In addition, ice nucleation of the background 30 water at higher temperatures than those predicted by classical nucleation theory (Murray et al., 2010a;Koop and Murray, 2016), due to impurities in the water and/or effects of the cover slip (Polen et al., 2018), was assessed by acquiring baseline droplet freezing measurements of water containing no added particles.
To facilitate comparison of different materials including across literature studies, their ability to nucleate ice is often expressed in terms of the ice nucleation active site density ns(T), which represents the number of active sites per unit surface area of a 35 solid sample on cooling from 0 °C down to temperature (T) (Connolly et al., 2009): where A is the total surface area of the solid sample per droplet. Although the fundamental nature of ice-active sites remains unclear, and may vary across different materials, ns(T) allows us to empirically define the INE of a range of solid substrates (Vali, 2014 Harrison et al., (2016). The uncertainty in temperature for the µL-NIPI is estimated to be ±0.4 °C (Whale et al., 2015).

Results
The fice(T) and ns(T) values for the eighteen samples studied are shown in Fig. 2. For the sake of clarity, droplet freezing events across replicate experiments have been combined here to generate a single dataset for each sample. The samples display wide For the tephra samples, the fice(T) curves are separate from those of the background water (Fig. 2a). This gives confidence to attribution of the observed freezing to ice nucleation by tephra particles. In terms of ice nucleation active site densities ( Fig.   2c), taking the temperature at which ns ≈ 1 cm -2 as a simple single-number proxy for INE (Tn s ≈ 1 cm -2 ), the most active tephra Consideration of the compositionally analogous tephra-glass pairs clearly illustrates the observation of tephra nucleating ice more effectively than the equivalent remelted and quenched glass (Fig. 3). The ns(T) curves of each tephra sample fall at higher temperatures compared to those of its counterpart glass sample, albeit displaying varying temperature differences between them.

Discussion
The eighteen samples were chosen to encompass a variety of chemical compositions, crystallinities and mineralogies encountered in volcanic ash, with the aim of investigating the influence of these physicochemical properties on ash INE. The results of our ice nucleation experiments are examined in relation to each of these properties below.

30
As noted above, our sample pairs of crystal-bearing tephra and crystal-free glass of nearly identical chemical composition (Table S1) constitute a novel approach to studying controls on volcanic ash ice nucleation. These pairs were chosen to disentangle variation in crystallinity from that in composition, which together might complicate interpretation of INE trends in natural ash collections. For all nine pairs studied, the tephra nucleates ice at higher temperatures than the corresponding glass ( Fig. 3 Fig. S1a). Additionally, a plot of Tn s ≈ 1 cm -2 values versus crystallinity of the tephra samples shows no correlation between ice nucleation and crystalline content (Fig. S1b). The compositionally quite similar NUOteph and ASTteph display the highest INE (Tn s ≈ 1 cm -2 values of -5.0 and -6.4 °C, respectively) and are characterised by markedly contrasting crystallinities (60 and 28 wt.%, respectively). A difference in crystallinity was proposed as one of a number of potential 5 explanations for the variable INE of two ash samples from Soufrière Hills volcano, Montserrat, with the 100% crystalline material produced by a dome collapse showing a higher INE than the 89% crystalline material produced by a magmatic eruption (Schill et al., 2015;Mangan et al., 2017). However, in light of our findings, it seems unlikely that this slight difference in crystallinity of two Soufrière Hills ash samples can adequately explain the large disparity in their INE.
A study comparing ice nucleation by crystalline and glassy anorthite (CaAl2Si2O8), with the former displaying a ns(T) curve 10 reaching higher freezing temperatures than the latter, suggested that crystals may introduce rarer, more ice-active surface sites but are not required for ice nucleation (Harrison et al., 2016). In addition, aqueous organic solutions in a glassy state have been shown to nucleate ice (Murray et al., 2010b;Wagner et al., 2012;Ignatius et al., 2016). In contrast, our observations strongly suggest that the presence of crystals is crucial in making volcanic ash an effective ice nucleant, although the abundance of crystalline phases in ash may be less important than the mere presence and specific properties of those phases in determining 15 the INE.

Mineralogy
Consideration of the nine tephra samples might provide insight into the influence of mineralogy on the INE of volcanic ash.
As noted above, comparison of the compositionally analogous tephra-glass pairs points to a role of crystalline phases in promoting freezing, and we infer that the properties of those phases have a strong effect on a sample's INE. A study of ash 20 from Fuego, Soufrière Hills, and Taupo volcanoes attributed differences in their INE to their contrasting mineralogies (Schill et al., 2015). Overall, when the Tn s ≈ 1 cm -2 values are plotted against the content of various minerals in the tephra samples studied here, no clear correlations are evident (Fig. 4). However, certain features do stand out; the two most ice-active NUOteph and ASTteph have the highest contents of alkali feldspar (60 and 19 wt.%, respectively; Fig. 4a), while the next most ice-active COLteph and TUNteph are characterised by an abundance of plagioclase feldspar (55 and 43 wt.%, respectively; Fig. 4b). 25 The ns(T) curves of our tephra samples are compared with the ice-nucleating activity of the two feldspar mineral groups in Other tephra samples containing feldspar are comparatively less effective at nucleating ice, in particular; the intermediately 35 ice-active ETNteph (44 wt.% plagioclase feldspar; Fig. 4b) and the least ice-active KILteph (3 wt.% plagioclase feldspar; Fig.   4b) and LACteph (9 wt.% alkali feldspar; Fig. 4a). Such differences may relate to the specific chemistry of the mineral phases present in the tephra. Harrison et al. (2016) showed that the INE of feldspar generally decreases from the K end-member (KAlSi3O8) to the Na end-member (NaAlSi3O8) to the Ca end-member (CaAl2Si2O8), with the exception of a hyper-active NaAlSi3O8 specimen (Amelia albite). Based on electron microprobe analysis (Text S1, proportion of the more ice-active NaAlSi3O8 relative to CaAl2Si2O8 in the former samples. On the other hand, the K2O/Na2O ratio in alkali feldspar in NUOteph, ASTteph and LACteph (1.2, 5.0, 1.9, respectively) does not support a link between these samples' INE and the proportion of KAlSi3O8 relative to NaAlSi3O8. This is consistent with the results of Whale et al. (2017), who found that the INE of alkali feldspar does not relate to K content, but rather to the presence of perthitic intergrowth 5 microtexture arising from phase separation (exsolution) into Na-and K-rich regions. Strain at the boundary of these regions gives rise to nanoscale topographic features that are suggested to be important in generating sites for ice nucleation (Whale et al., 2017).
Perthite in alkali feldspar develops in metamorphic and plutonic contexts during slow cooling at temperatures <700 °C (Parsons, 2010), and is generally not expected in volcanic ash which cools rapidly from magmatic down to ambient 10 temperatures during eruption (Parsons et al., 2015). The low INE of LACteph could hence reflect a lack of perthitic microtexture, as noted for the alkali feldspar mineral sanidine (KAlSi3O8) sourced from the same geological setting (Eifel volcanic field), and similarly showing a poor ability to nucleate ice (Whale et al., 2017). In contrast, the high INE of NUOteph and ASTteph is surprising in the absence of perthitic microtexture, and perhaps some other textural feature underlies their ability to nucleate ice as effectively as alkali feldspar of non-pyroclastic origin (Fig. 5). Pyroclastic material from both the 1538 Monte Nuovo 15 eruption (i.e., the origin of NUOteph) and the ~4 ka Astroni eruption (i.e., the origin of ASTteph) has been found to contain antirapakivi overgrowth texture characterised by plagioclase feldspar cores rimmed by alkali feldspar (D'Oriano et al., 2005;Astbury et al., 2016;. Such textures are challenging to resolve optically in powdered samples. Further optical and microanalytical (Scanning-and Transmission Electron Microscopy) observations (e.g., Whale et al., 2017;Holden et al., in press) will be needed to explore whether the boundary between Na-and K-rich regions in anti-rapakivi texture may give rise 20 to nanoscale topography that induces effects analogous to perthitic microtexture in promoting ice nucleation.
In addition to feldspar, several of the tephra samples contain pyroxene (Table 2). Specifically, the presence of orthopyroxene distinguishes COLteph and TUNteph from the other tephra (Fig. 4d), raising the question of whether it may underlie the high INE of these two samples (i.e., rather than plagioclase feldspar). We are aware of only a few early studies on ice nucleation by ortho-and clinopyroxene minerals (hypersthene, augite; Hama and Itoo, 1956;Isono and Ikebe, 1960), but these studies have 25 only reported semi-quantitative onset freezing temperatures (between -8 and -15 °C). Therefore, at present we cannot rule out a potential influence of pyroxene minerals on ice nucleation by volcanic ash. Additional research is needed to quantify the INE of a range of pyroxene minerals in order to better inform this assessment.

Chemical composition
To explore a potential link between chemical composition and INE, the Tn s ≈ 1 cm -2 values of the tephra and glass samples are

35
The glass samples, in lacking crystalline minerals, are well-suited to assess for any direct relationships between INE and specific element oxide abundances. However, due to the overlap of droplet freezing temperatures of the glass suspensions and the background water (Fig. 2b), our ability to distinguish differences in INE across the nine samples is impeded. While CIDglass and COLglass are the most ice-active glass samples, with signals clearly above the background (Fig. 2b, d) In contrast, the tephra samples are characterised by variations in crystallinity and mineralogy as well as composition, which convolutes the assessment of relationships between INE and specific element oxide abundances. However, if the crystallinity and mineralogy of the tephra samples are taken into consideration, a broad pattern emerges in the plots of Tn s ≈ 1 cm -2 versus Fe2O3, MgO, and CaO contents (Fig. S2c-e). Excluding a cluster of three samples with comparatively low crystallinities (LIPteph, CIDteph, LACteph), the Tn s ≈ 1 cm -2 decreases with increasing Fe2O3, MgO, and CaO contents for NUOteph, ASTteph,

5
COLteph, TUNteph, ETNteph, and KILteph (Fig. 6), in an order consistent with interpretations relating to their feldspar contents and chemistries (see discussion Sect. 4.2). This conforms to the notion of an indirect relationship between chemical composition and volcanic ash INE, whereby FeO/Fe2O3, MgO, and CaO contents increase from felsic to mafic magma, influencing the mineral phases that can crystallise from the magma and hence exist in the resultant ash (Fig. 1b).

10
Here we used nine compositionally analogous pairs of natural tephra and remelted and quenched glass to investigate the influence of chemical composition, crystallinity and mineralogy on the INE of volcanic ash. The higher INE of the tephra relative to the glass strongly suggests that the presence of crystalline phases promotes ice nucleation. The large variability in INE of the tephra is inferred to reflect an influence of mineralogy -and hence an indirect influence of magma compositionon ice nucleation. As in desert dust, alkali feldspar is probably the most ice-active component in volcanic ash, conferring the 15 highest INE to NUOteph and ASTteph in this study. However, the ability of alkali feldspar in ash to nucleate ice likely cannot unequivocally be attributed to perthitic microtexture, as has been done for alkali feldspar of non-pyroclastic origin. Additional research is needed to explore whether other textural features in ash may elicit a similar effect in promoting ice nucleation.
Further, the presence of alkali feldspar is neither always sufficient nor necessary for effective ice nucleation by ash. The high INE of COLteph and TUNteph may alternatively reflect very ice-active plagioclase feldspar, or possibly orthopyroxene which is 20 present exclusively in these studied tephra. Previous studies on Soufrière Hills ash, also lacking alkali feldspar and containing plagioclase feldspar and orthopyroxene, have reported both a low and high INE of this ash (Schill et al., 2015;Mangan et al., 2017). Future studies quantifying the INE of individual crystalline phases found in ash will be necessary to unravel the precise role of mineralogy in volcanic ash ice nucleation.
An improved knowledge of the link between particular ash properties and ash INE may ultimately enhance predictive capability 25 regarding volcanic eruptions likely to generate ice-active material. For example, as crystalline phases are primarily controlled by magma composition and storage/ascent conditions (Rogers, 2015), we speculate that highly ice-active ash particles might be erupted by volcanoes with intermediate to felsic alkaline magmas and possessing a history of magma mixing leading to feldspar crystals featuring overgrowth textures (e.g., Astbury et al., 2016;. In addition, an eruption producing an abundance of crystal-bearing particles is expected to elicit a greater impact on heterogeneous ice nucleation than an eruption 30 producing an abundance of crystal-free glass particles, all else being equal. Accordingly, massive outputs from the largest and most explosive eruptions, corresponding to violent caldera-forming ignimbrite events that generate ash clouds dominated by the glassy component (Sparks et al., 1997;Cather et al., 2009), may be less efficient in affecting INP populations than ash emissions from smaller eruptions. Further, since airborne ash typically becomes enriched in glassy fragments during longrange transport due to earlier gravitational settling of crystalline fragments (Hinkley et al., 1982), the INE of a suspended ash

35
population is expected to decrease over time and distance from the volcano.
Lastly, it must be noted that once ash particles are generated, their surface properties can be altered by interactions with gases and condensates (e.g., H2O, SO2, H2SO4, HCl, HF) at variable temperatures in the eruption plume/cloud and ambient atmosphere (Delmelle et al., 2007;Ayris et al., 2013;2014;Maters et al., 2016;. The effects of such interactions on ash INE are not known, although it has been suggested from field measurements that volcanic gases may deactivate INPs (Schnell 40 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1326 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 January 2019 c Author(s) 2019. CC BY 4.0 License. and Delany, 1976;Schnell et al., 1982). Laboratory studies on desert dust show that 'aging' of dust particle surfaces by exposure to H2SO4 vapours at elevated temperatures reduces dust INE (Sullivan et al., 2010;Niedermeier et al., 2011).
Therefore, exploring potential eruptive and atmospheric controls on ash INE is an important next step towards developing a better understanding of the capacity of volcanic ash emissions to affect heterogeneous ice nucleation during their airborne lifetime.

5
Data availability: e.c.maters@leeds.ac.uk   (Fig. 1) based on chemical composition (Table S1). c Uncertainty is in the range of 0.5-1.2 %. . Droplet fraction frozen (fice) as a function of temperature for 1 wt.% suspensions of (a) tephra or (b) glass in water. The grey bands represent the spread of fice(T) measurements (mean values ± standard deviation) of the background water (i.e., containing no added sample). Ice nucleation active site density (ns) as a function of temperature for 1 wt.% suspensions of (c) tephra or (d) glass in water. For the sake of comparison, the grey curves represent theoretical upper limit ns(T) values of the 5 background water, calculated using the upper limit fice(T) measurements (mean values + standard deviation) of the background water, and assuming it contains particles with SSABET values equal to the lowest from the tephra and glass sets (1.4 and 0.9 m 2 g -1 , respectively). The tephra ns(T) values are well above this background but most of the glass ns(T) values should be regarded as upper limits. The uncertainty in ns(T) is shown as error bars for a subset of data points (of NUOteph and CIDglass) and omitted from remaining data points for clarity, but is typical of all tephra and glass samples. Sample codes are listed in Table 1. 10 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1326 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 January 2019 c Author(s) 2019. CC BY 4.0 License. Figure 3. Ice nucleation active site density (ns) as a function of temperature for 1 wt.% suspensions of tephra or glass in water. Each plot shows replicates of a compositionally analogous pair of tephra (red triangles) and glass (blue circles). The grey curves with triangle and circle symbols represent the detection limit for ns(T) based on the background water runs accompanying the tephra and glass experiments, respectively (see Fig. 2 caption for details). Sample codes are listed in Table   5 1.
Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-1326 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 2 January 2019 c Author(s) 2019. CC BY 4.0 License. Figure 5. Ice nucleation active site density (ns) as a function of temperature for 1 wt.% suspensions of tephra in water. The blue and red lines represent parameterisations for, respectively, alkali and plagioclase feldspar of non-pyroclastic origin reported in Harrison et al. (in prep.)