Oxide nanolitisation-induced melt iron extraction causes viscosity jumps and enhanced explosivity in silicic magma

Abstract

Explosivity in erupting volcanoes is controlled by the degassing dynamics and the viscosity of the ascending magma in the conduit. Magma crystallisation enhances both heterogeneous bubble nucleation and increases in magma bulk viscosity. Nanolite crystallisation has been suggested to enhance such processes too, but in a noticeably higher extent. Yet the precise causes of the resultant strong viscosity increase remain unclear. Here we report experimental results for rapid nanolite crystallisation in natural silicic magma and the extent of the subsequent viscosity increase. Nanolite-free and nanolite-bearing rhyolite magmas were subjected to heat treatments, where magmas crystallised or re-crystallised oxide nanolites depending on their initial state, showing an increase of one order of magnitude as oxide nanolites formed. We thus demonstrate that oxide nanolites crystallisation increases magma bulk viscosity mainly by increasing the viscosity of its melt phase due to the chemical extraction of iron, whereas the physical effect of particle suspension is minor, almost negligible. Importantly, we further observe that this increase is sufficient for driving magma fragmentation depending on magma degassing and ascent dynamics.

Fig. 1. Magnetic hysteresis loops and Raman spectra.

Hysteresis loops are normalised by sample weight and adjusted for dia-/paramagnetic contributions from: A, B controlled cooling experiments (#1 Cooling) with rates as indicated, C differential scanning calorimetry (#2 DSC), and D micro-penetration (#3 MP) viscosity analyses. E Raman spectra after each analysis. Note that analyses were performed in the same sequence and both magnetisation as well as the 670–690 cm⁻¹ Raman peak are consistently growing, indicating an increase in crystallinity. All individual normalised magnetic hysteresis curves are shown in Fig. S2. arb. units arbitrary units. Source data are provided as a Source Data file.

Fig. 2. Differential scanning calorimetry (DSC) analyses for the previously cooled (controlled cooling experiments) samples.

According to initial cooling rate, analyses correspond to the samples: A 0.1 °C min⁻¹, B 0.2 °C min⁻¹, C 0.3 °C min⁻¹, D 0.4 °C min⁻¹, E 0.5 °C min⁻¹, and F rapid-quenched (rq). Analyses were performed in the same sequence for each sample: (1) heating to 900 °C at 25 °C min⁻¹, (2) cooling at 25 °C min⁻¹ and then heating to 850 °C at 25 °C min⁻¹, (3) cooling at 15 °C min⁻¹ and then heating to 850 °C at 15 °C min⁻¹, (4) cooling at 10 °C min⁻¹ and then heating to 850 °C at 10 °C min⁻¹, (5) cooling at 25 °C min⁻¹ and then heating to 850 °C at 25 °C min⁻¹. Note that the first analyses performed with initially known slow cooling rates show a remarkably higher glass transition peaks, followed by a crystallisation (exothermic) peak. Second analyses (25/25 °C min⁻¹) show a shift of the glass transition peak to higher temperatures and insignificant crystallisation peaks, similar to those shown by the last analyses conducted at the same cooling-heating rates. This indicates that no crystallisation occurred in a significant extant in these samples after the first crystallisation event during the analyses. arb. units arbitrary units. Source data are provided as a Source Data file.

Fig. 3. Imaging, quantification and determination of oxide nanolites.

A–C Field emission-scanning electron microscope (FE-SEM) images of the nanolite-bearing samples. Shown are the samples after micro-penetration analyses for the initially slowly cooled samples (cc) at rates of 0.1 and 0.5 °C min⁻¹, as well as the sample subjected to rapid quenching (rq) and that was originally nanolite-free. For this sample (rq), nanolites nucleated during heating of the last calorimetry analysis (DSC) and grew during micro-penetration analysis, evidenced by both magnetic and Raman analyses (Fig. 1). All images acquired after micro-penetration (MP) analyses. D Crystallinities at each step of analyses, determined by magnetic analyses and calibrated by crystallinities determined from SEM images (see “Methods”). Crystallinity data can be found in Table S1. E Curie temperatures of the samples above with 0.61 vol.% (rq) and 1.12 vol.% (cc) crystals. Curie temperatures were determined after micro-penetration analyses, showing no change between heating and cooling curves, indicating no major change in the stoichiometry of the Fe-Ti oxides. Curie temperatures confirm that nanocrystals present correspond to titanomagnetite crystals, with slightly higher titanium content for the crystals in the samples with 1.12 vol.% (cooling-controlled) compared to that with 0.61 vol.% (rapid-quenched). Source data are provided as a Source Data file.

Fig. 4. Viscosities of nanocrystal-free melt and nanocrystal-bearing magma.

A The viscosity of the nanolite-free melt (blue), the melt viscosity of the nanolite-bearing magma (purple) and the bulk magma viscosities measured with micro-penetration for the nanolite-bearing magmas (triangles): 0.61 vol.% nanolites in blue and 1.12 vol.% nanolites in purple. Dotted lines show the melt viscosities calculated by Giordano et al. for nanolite-free and nanolite-bearing melts. Nanolite-bearing melt was calculated from a simulation in rhyolite-MELTS when reaching a 1.12 vol.% nanocrystals. Note the shifts between the viscosity of the nanolite-free melt and the modelled viscosity, as well as the concordance between the modelled viscosity and the measure bulk viscosity of the nanolite-bearing magma. The ±0.1 log-units uncertainty associated to the viscosity measurements is shown within the symbols size (see “Methods” section for details). B Non-Arrhenian fits for the viscosities of synthesised rhyolite melts with variable iron concentrations. Total iron as Fe₂O₃ according to normalised values in Table 1. Variable iron concentrations represent approximate percentages with respect to the original natural rhyolite composition (3.34 wt%): 0, 25, 50, 75 and 100%. Complete curves including high-temperature and low-temperature viscosity data points can be found in Supplementary Material. C Bulk magma viscosity vs. nanocrystal content at 875 °C. Blue dot represents the extrapolated viscosity obtained from the DSC-derived viscosity. The ±0.1 log-units uncertainty associated to the viscosity measurements is shown by the error bars. D Melt viscosity vs. iron content. Note the exponential increase in magma viscosity as the melt iron content decreases. Dotted line represents the best fitting line of the data. Uncertainty of the total iron contents represents the standard deviation of the chemical (EPMA) analyses shown in Table 1. Source data are provided as a Source Data file.

Fig. 5. Fragmentation criterion for strain-induced magma fragmentation.

Line shown considers a fragmentation threshold where strain rate equals to $\kappa G_{\infty }/{\mu }_m$, where κ and G_∞ are an experimental constant and the elastic modulus at infinite frequency with values of 0.01 and 10 GPa respectively^,, and µ_m is the melt viscosity. Considering common strain rates for silicic magma ascent^,, an event of oxide nanolites crystallisation can move a magma into the conditions necessary for fragmentation and explosive behaviour by increasing mainly its melt viscosity. Dotted grey line shows the strain and viscosity conditions of a magma that may potentially fragment if oxide nanolites crystallise in it, reaching the fragmentation criterion (black solid line) by increasing its viscosity by one order of magnitude. Dotted horizontal grey lines show the explosive and effusive behaviour from Gonnermann and Manga. Blue circle and purple triangle show the actual viscosities of the nanolite-free melt and nanolite-bearing magma of this study plotted at the strain rates at which their jump from one viscosity to the other one would drive fragmentation and are for reference only. Such a shift in viscosity will cause fragmentation dependent on the actual strain rate and degassing dynamics of the ascending magma.