Can nanolites enhance eruption explosivity?

A key question faced by volcanologists is whether the ascent of magma in the shallow crust will result in an explosive eruption of ash and gas or a potentially less-dangerous effusive eruption of lava. Indeed, these styles of eruption can switch during a single eruption (e.g., Rust and Cashman, 2004; Colombier et al., 2017a; Cassidy et al., 2018). Magma ascent is driven by bubble growth, and a first-order control on whether an eruption will be explosive or not is the rate at which bubbles grow in the melt (e.g., McBirney and Murase, 1970; Forte and Castro, 2019) and the propensity for permeable outgassing during bubble growth (e.g., Degruyter et al., 2012). Both of these factors—bubble growth rates and permeable outgassing efficiency—are influenced strongly by the number density of bubbles that form in nucleation events, because high bubble number densities shorten the distances for water to diffuse between bubbles, reduce the viscous resistance to growth, and result in smaller bubble-bubble pairs at the onset of permeability (e.g., Gonnermann and Manga, 2007; Degruyter et al., 2012).

Bubble number densities are dominantly controlled by the local decompression rate, volatile diffusivity, volatile concentration, temperature, and surface tension (e.g., Toramaru, 2006; Shea, 2017). Crystals, especially those at micron to millimeter scale, impart the magma with a distributed internal interface to serve as a surface for heterogeneous nucleation, reducing the supersaturation conditions needed to nucleate bubbles (Hurwitz and Navon, 1994; Mangan et al., 2004; Gardner and Denis, 2004; Cluzel et al., 2008; Larsen, 2008; Shea, 2017; Pleše et al., 2018) and to facilitate bubble coalescence (e.g., Lindoo et al., 2017). In turn, crystals “stiffen” the viscous response of the magma such that bulk expansion of the bubbles is resisted (Oppenheimer et al., 2015). Thus, the observation that nanolites, or nanometer-scale crystalline phases, are extant in products of explosive volcanism raises the general question as to what their physical role might be in bubble dynamics during magma ascent.

Nanolites have been identified in both natural volcanic products (e.g., Sharp et al., 1996; Mujin and Nakamura, 2014; Barone et al., 2016; Zellmer et al., 2016; Colombier et al., 2017b) and experimental products (e.g., Liebske et al., 2003; Di Muro et al., 2006; Di Genova et al., 2017). They are inferred to crystallize in magmas at low pressure (Mujin and Nakamura, 2014) and during slow cooling (Liebske et al., 2003; Di Muro et al., 2006). The formation of nanolites has been suggested to favor heterogeneous bubble nucleation (e.g., Shea, 2017), and they have been invoked as a potential control on the eruptive style of volcanoes (Mujin and Nakamura, 2014; Mujin et al., 2017; Di Genova et al., 2017). However, the mechanism behind this control on the eruptive style remains unclear. Here, we present results of experiments designed to explore the degassing dynamics of rhyolitic magma with and without nanolites, and we find that the presence of nanolites can amplify bubble number densities by many orders of magnitude. We use this finding to argue that nanolites may be capable of inducing explosivity in otherwise effusive eruptions.

We used two volcanic materials to conduct our bubble-growth experiments: (1) a natural iron-rich nanolite-free rhyolite obsidian (Fig. 1A) with 0.15 wt% dissolved H₂O, and (2) a synthesized nanolite-bearing sample (Fig. 1B) of the same rhyolitic composition with 0.32 wt% dissolved H₂O (see the Supplemental Material¹). To form nanolites, we re-sintered a powder of the natural obsidian in a sealed container with a small amount of water in a hot isostatic press at 800°C and 23 MPa. Nanolites are ubiquitous in the synthesized sample, with an average nanolite number density N_n of ~3.5 × 10¹² mm⁻³ and an average size of 10–50 nm, corresponding to ~2 vol% of the magma. The crystals and bubbles were investigated via backscattered electron images (Fig. 1) of polished samples using a Hitachi SU5000 scanning electron microscope. A confocal Raman spectrometer was used to identify nanolites (see the Supplemental Material). The nanolites in the groundmass of the synthesized sample correspond to Fe-Ti oxides based on a peak in Raman spectra at 670–690 cm⁻¹. The same peak is absent from the spectra of the glass in the natural obsidian, confirming the absence of Fe-Ti oxide nanolites (Fig. 1G).

We heated these two sample types in an optical dilatometer at 1–60°C min⁻¹ to isothermal dwells at 820–1000°C. We recorded their real-time sample volume and shape as they crossed the glass transition, followed by rapid expansion (see the Supplemental Material). The experimental products for the two sample types were then compared.

When heated at the same heating rates, the nanolite-bearing sample always began expanding significantly earlier and with a higher expansion rate than the nanolite-free sample (Fig. 2). While the two sample types (nanolite-free and nanolite-bearing) had subtly different initial water concentrations (0.15 versus 0.32 wt%), we assume that the discrepancy in onset time and expansion rate (bubble growth) cannot be attributed to this small difference in starting water content. We note that even when comparing conditions where the calculated melt viscosity is higher and water diffusivity is lower in the nanolite-bearing sample (Fig. 2), the growth rate in the nanolite-bearing samples were higher. Thus, this gap between the two sample types is attributed to the difference in their nucleated bubble number density.

Using run products for experiments that reached equilibrium porosity and bubble size, we determined that for the natural obsidian, the bubble number density (N_b) produced during the experiments was 10⁻¹ ≤ N_b ≤ 10¹ mm⁻³ and final bubble sizes (bubble diameter, R_d) were 0.2 ≤ R_d ≤ 1.5 mm (Fig. 1C), while for the re-melted, nanolite-bearing sample, N_b was much higher, at 10⁶ ≤ N_b ≤ 10⁷ mm⁻³, with significantly smaller final bubble sizes of 10⁻³ ≤ R_d ≤ 10⁻¹ mm (Fig. 1D). The values obtained here for the natural, nanolite-free obsidian are consistent with experiments carried out with same composition and at similar conditions (Ryan et al., 2015).

The number density of homogeneously nucleated bubbles during degassing of an ascending magma is dominantly controlled by the decompression rate (e.g., Toramaru, 2006; Shea, 2017). However, heterogeneous nucleation may decrease the decompression rate needed to produce the same N_b by orders of magnitude (e.g., Cluzel et al., 2008; Hamada et al., 2010; Shea, 2017). High N_b (~10⁷ mm⁻³) in natural silicic explosive products (e.g., the 7.7 ka Mount Mazama eruption, Oregon, USA; Klug et al., 2002) have been attributed to a combination of fast decompression rates and the presence of relatively high (10⁰–10² mm⁻³) crystal number densities during degassing (e.g., Hamada et al., 2010), with an emphasis on the additional role played by decompression (or ascent) rates. Our data suggest that the nucleation of nanolites in magmas could result in high bubble number densities and associated high bubble growth rates in situations where nanolites grow below or within the kinetic window of bubble nucleation. This implies a possibility of reaching high bubble number densities by heterogeneous nucleation of bubbles on nanolites even at relatively low decompression rates.

To put our results in context, we compiled published estimates of N_b in natural products of explosive eruptions with a range of magmatic compositions (Fig. 3A). These data show that across the range from intermediate to silicic magmas, mean N_b tend to be ~10⁵–10⁶ mm⁻³, with maxima at >10⁸ mm⁻³ (Fig. 3C; similar to data compiled by Shea [2017]). We note that the bubble number densities produced in our nanolite-bearing samples are in the higher end of the range of bubble number densities (≥10⁶ mm⁻³) found in natural explosive products. For comparison, data for N_b in intermediate to silicic effusive eruptions are sparse, but typical values are <10³ mm⁻³ (Fig. 3B).

Nanolites are not widely reported in erupted products, and therefore the subset of this compilation for which nanolites have been identified is low. Fe-Ti-oxide nanolites have been observed in tephra from Kilian volcano, France (Colombier et al., 2017b; see the Supplemental Material). These nanolite-bearing products have N_b as much as two orders of magnitude higher than the nanolite-free products of the same eruption (Fig. 3A). In the case of the 2011 Shinmoedake eruption, Japan (Mujin and Nakamura, 2014; Mujin et al., 2017), nanolites of low-Ca pyroxenes, plagioclase, and Fe-Ti oxides are present throughout the transition sequence from sub-Plinian to effusive eruptive style. Even though the nature of these nanolites differs from those in this study, due to the different composition of the magma involved, nanolites are present in the products of the first and most explosive phase of the eruption with N_n of ~10¹³ mm⁻³ comparable to those in this study (~10¹² mm⁻³).

This compilation of naturally occurring bubble number densities (Fig. 3) along with the few observations that exist for the natural occurrence of nanolites suggest that nanolites may play a role in increasing the bubble number density relative to nanolite-free magma. This is consistent with the findings of our experimental work (Figs. 1 and 3). While we do not claim that this is the only cause of high bubble number densities, we propose that this process can operate in shallow magma ascent.

Here, we have established that silicic magmas that form nanolites prior to bubble nucleation events can result in high bubble number densities compared with nanolite-free magmas. In a second step, we show that the high experimental N_b values in nanolite-bearing melts are at the upper end of naturally occurring ones, and that in the few available cases, nanolites have been found in high-N_b products of explosive eruptions. As a final step, we employ a reference magma ascent model to show what the effect of such an increase in N_b may be in ascending magmas. In doing so, we do not aim to model heterogeneous nucleation of bubbles on nanolites explicitly, but rather use existing tools to build the case for the critical effect nanolites may have in an ascending magma.

We simulated magma ascent in a conduit using the one-dimensional two-phase model developed by Degruyter et al. (2012) that considers magma ascent and permeable outgassing at high porosities. Although this model works on the basis of equilibrium degassing and does not consider the mechanism (i.e., homogeneous or heterogeneous) and kinetics of bubble nucleation, it is suitable for investigating the effect that existing bubbles exert on the eruptive behavior of ascending magma in a volcanic conduit, because it predicts whether a magma will fragment during ascent or not, effectively showing whether an eruption could be effusive or explosive at the Earth’s surface. This model requires an input of N_b and initial bubble size, as well as conditions regarding the conduit dimensions, initial magma storage depth, temperature, and initial water content. We used three different realistic conduit diameters (10, 16, and 20 m) and magma starting conditions, such as those of the well-studied case of the 2008 Chaitén (Chile; Castro and Dingwell, 2009) eruption (Fig. 4; see the Supplemental Material). We solved this model by varying N_b, using the same range found in natural explosive products and in our experiments (Fig. 3A).

The simulations (Fig. 4) show that for a given range of common conduit dimensions, the range of N_b found in natural magmas (10⁴–10⁸ mm⁻³) can straddle effusive and explosive eruptive style. Although the conduit diameter plays a first-order role, we find that for a given geometry, the increase in N_b associated with bubble nucleation on nanolites could result in a transition from effusive to explosive eruption (Fig. 4). These results are also consistent with low and high N_b inferred for effusive and explosive eruptions respectively (e.g., Cluzel et al., 2008). Moreover, for the conduit diameters investigated, the range of N_b that appears to be associated with nanolites (Fig. 3) is sufficiently high to result in explosive volcanism in most of the studied scenarios.

We conclude from our experiments that nanolites can be efficient sites for bubble nucleation leading to high bubble number densities and growth rates. We observe that high bubble number densities such as those produced by nanolite-bearing samples are found in natural explosive volcanic products. We use these observations and numerical model results to propose that nanolites may play a role in enhancing the propensity for eruptions to be explosive. The conditions of undercooling required for nanolite crystallization remain poorly understood, and further investigation is required to link the nanolite nucleation window with the vesiculation window for natural magma ascent conditions.

We acknowledge funding from the German Research Foundation (DFG) through the projects SCHE1634/1-1 and DI431/33-1. Dingwell acknowledges the support of ERC 2018 ADV Grant 834225 (EAVESDROP). We thank J. Vasseur, J. Coumans, and E. Llewellin for helpful discussion; W. Degruyter for providing the code associated with Degruyter et al. (2012); and A. Ryan and two anonymous reviewers for valuable comments.