Main

Volcanic eruptions result from magma buoyancy, largely powered by volatile exsolution. In standard models of magma ascent this exsolution is triggered by decompression8,9. Upon ascent, gas bubbles (vesicles) expand and pressure build-up may precipitate fragmentation and explosive eruption1. Yet the solubility, which sets the thermodynamic driving force for saturation and vesiculation in a volatile component has long been known to be a function of temperature as well10. Thus temperature changes may also generate magma vesiculation. Despite this, to our knowledge, no models of volcanic eruptions have explored the role of temperature in generating magmatic vesicularity.

The thermal evolution of magma in volcanic conduits has received increased attention in recent years. First, petrological studies have demonstrated that crystallizing magmas can heat up considerably (up to about 100 °C) owing to the latent heat liberated4—a process acting across the entire magmatic column. Second, zones in which magma undergoes strain localization during ascent also exhibit evidence of considerable heating (up to about 250 °C) resulting from viscous energy dissipation5,6,11,12. Third, the discovery of pseudotachylytes (caused by frictional melting during faulting) in erupted dome rocks13 and at the margin of lava spines7 indicates that fault friction can be an important contributor to the thermal budget of magma (locally up to about 1,000 °C), thus strongly affecting volcanic eruption dynamics14.

Evidence is mounting that magma ascent may often be controlled by strain localization near conduit margins15. Such strain localization in magmas has been proposed as a scenario leading to failure and potentially serving as a trigger for explosive eruptions16,17. Careful examination of shallow volcanic conduit structures lends support to these proposals18. Magmatic conduits or dykes are relatively narrow (tens of centimetres to a few metres) at depths of a few kilometres19, so regions of strain localization may represent an important mass fraction of ascending magma. At shallow depths, where conduits can be wider (metres to a few tens of metres), areas of strain localization may not appear to be inevitable, yet the observation that shallow magma bodies are heavily fractured20, and the influence of such fractures on surficial magma behaviour21 suggest that strain localization and its associated heat may play a large part throughout the length of the magmatic column.

Estimates of magma ascent rates vary widely. In general, explosive eruptions have been associated with high ascent rates, reaching up to a few metres per second before fragmentation9. During such rapid ascent, magma decompresses (one metre per second corresponds to a decompression of 0.02 MPa per second) and simultaneously heat is generated in all areas where strain is localized, either by fault friction14 or viscous dissipation12. The material record of such heat may or may not be documented in the products of the subsequent volcanic explosions. The mineralogical assemblage can often preserve information related to such heating7,13, but if sufficient time passes then the assemblage will recover in response to the mean temperature and pressure conditions and evidence of fluctuations may be lost. The glassy state itself does not provide direct information from above the glass transition temperature, yet, indirectly, evidence of energy dissipation has been inferred from the morphology of the porous network preserved in glassy volcanic products5,6. The difficulty of preservation of evidence of heating in ascending magma, or of the temperature history, is probably a major reason for its neglect in eruption models.

Temperature and pressure both affect the solubility of water22,23,24 (the dominant volatile component of volcanic activity) in magma. For a calc-alkaline rhyolitic melt, the temperature- and pressure-dependence of water solubility can be estimated by24:

where (H2O)total is the total dissolved H2O content (in weight per cent, wt%), T is temperature (in K), and P is pressure (in MPa). Figure 1a shows that owing to the strongly retrograde nature of the H2O solubility curve at low pressures, an increase in temperature is a driving force for vesiculation in this pressure range. This temperature dependence is clearly large enough to have a substantial effect on water saturation during magma ascent in conduits.

Figure 1: Water concentration in rhyolitic magmas.
figure 1

a, Thermobarometric limits on water concentration24 show that the heat induced by mechanical work (orange arrows) during magma ascent causes a decrease in water solubility, Δ(H2O). This decrease in concentration may be related to an equivalent decompression, ~ΔP. At Santiaguito, thermal input of, for example, ~600 °C owing to short-lived faulting events may reduce water solubility by 0.28 wt%. b, Water exsolution, Δ(H2O), driven by thermal input (red curves) versus decompression events (blue curves) for an ascending magma at a nominal temperature of 850 °C. These heating and decompression events are computed as a function of melt pressure at which the event initiates in the magmatic column. c, Fraction of the total water concentration exsolved from the action of heat (left y axis), versus that of decompression (right y axis) for different decompression and heating events. The data shows that thermal input (which acts on a timescale of seconds) generally induces more water exsolution than decompression.

PowerPoint slide

Full size image

We analysed the potential magnitude of water exsolution ΔH2O that is due to (1) decompression and (2) heating at a magmatic temperature of 850 °C (Fig. 1b). The comparison of the individual effects of decompression versus heating yields striking results. We found that events that heat magma by hundreds of degrees, as described above, strongly drive substantial exsolution and vesiculation. For an ascent rate of one metre per second (that is, 0.02 MPa per second, which is capable of triggering explosive events), 1 K of heating has the potential to generate more water exsolution than 0.02 MPa of decompression from initial pressures greater than 13 MPa (Fig. 1c), and further heating can be the main driving force for vesiculation. Expressing it in a different way, a decompression event exceeding 0.1 MPa (>5 m of ascent) would be required to exsolve more water than that exsolved by 1 °C of heating. We therefore conclude from this analysis that the thermal path of decompressing magma can greatly influence volatile exsolution. It is thus easy to envisage scenarios of heating-dominated or ‘thermal’ vesiculation during magma ascent at moderate pressures, and below we provide evidence to support the assertion that such thermal exsolution is also dominant during strain localization in magma at shallower depths.

We have examined eruptive products at the Santiaguito dome complex. The active Caliente lava dome offers one of the most spectacular displays of cyclic, piston-like eruptive activity ever recorded, often climaxing in gas-and-ash explosions along concentric fractures21,25,26 (Fig. 2a). Proximal monitoring of this dome has revealed a regular (~26 min) periodicity in ground inflation–deflation cycles27. At the expansion maxima, the propagation of arcuate faults across the dome’s surface is observed and the dome’s centre thrusts upward and collapses back, followed by dome deflation21. Gas-and-ash explosions occur episodically along the faults, coincident with very-long-period seismic events, which have been interpreted to be associated with gas flow in fractures at the inflation maximum (Fig. 2b)27. In the analysis that follows, the rates of inflation and deflation during ash release and the magnitude and rate of slip are of central importance. Ash ejection occurs only during the fastest inflation–deflation cycles (Fig. 2b)27. In these cases, the arcuate faults undergo a metre of uplift and collapse within one second, corresponding to a slip rate of <2 m s−1 (ref. 21). Importantly, these lava dome dynamics leave striation and slickensides (frictional marks) on the blocks forming the dome carapace.

Figure 2: Explosive eruptions caused by superheated vesiculation.
figure 2

a, Gas-and-ash explosion occurring along an arcuate fault on 10 November 2012 at Caliente dome, Santiaguito. b, Seismic signals (upper) and tilt data (lower) for explosive (red) and non-explosive (blue) inflation–deflation cycles associated with piston-like dynamics at Santiaguito27. The solid lines display the average of the 26-min cycles over the five-day-long data set, whereas the shaded areas exhibit the spread in the data. Gas-and-ash explosion cycles differ markedly from non-explosive cycles and are characterized by faster and stronger inflation or deflation as well as very-long-period (VLP) seismic events. Note that the exact timing of the seismic and tilt signals may be offset slightly. c, BSE image showing heterogeneous protomelt filaments (yellow arrows) with different greyscale values (a proxy for chemical composition; darker grey indicates lighter elements and vice versa) extruded from crystals present in a volcanic ash particle sampled on 12 November 2012; some protomelts host vesicles (blue arrows). d, BSE image showing the shearing of protomelts near the main frictional melt zone (FMZ), produced experimentally by fault slip. e, BSE image showing vesiculation of the interstitial melt near the experimental fault zone, caused by high local temperature.

PowerPoint slide

Full size image

Textural examination of volcanic ash collected upon deposition in November 2012 and November 2014 provides several examples of the material consequences of such frictional processes (Extended Data Figs 1, 2, 3, 4). The interstitial glass phase reveals a juxtaposition of chemically distinct mingled filaments with different shades of grey on back-scattered electron (BSE) images obtained by scanning electron microscopy (SEM; Fig. 2c; Extended Data Figs 3 and 4). The contacts between the light- and darker-toned filaments are diffuse and fluid (unlike crystals with sharp and angular boundaries). The very fine nature of these filaments and the diffuse boundaries prevent us from accurately using standard geochemical analysis techniques, but the greyscale values observed (which reflect the atomic number and thus chemical variations within and between phases) provide clear evidence of chemical heterogeneity (Fig. 2c; Extended Data Fig. 4). These melt phases have evidently mingled with the original interstitial melt on timescales insufficient for homogenization, presumably immediately before the fragmentation and eruption that locked in these dynamic features.

The mingling textures exhibited by the Caliente ash mirror those of protomelts resulting from selective melting of individual crystals that have been observed in the products of frictional melting experiments28,29. Such experiments involve an extremely rapid heating rate (more than tens to hundreds of degrees Celsius per second) and therefore highly disequilibrium melting induced by fault friction28,29,30. We propose here that the Caliente ash samples contain volcanic pseudotachlyte; evidence of the syn-eruptive operation of frictional heating sufficient to generate melting in the piston-like events at Caliente dome. Notably, the protomelts present in the ash contain vesicles (as indicated by blue arrows on Fig. 2c). The crystalline phases present are anhydrous and thus cannot serve as a source of water for vesiculation, so we suggest that vesiculation took place in the interstitial melt. If so, these frictional melts contain clear evidence of thermal vesiculation in volcanic products.

As an experimental demonstration of the feasibility of thermal vesiculation, we have performed fault friction experiments under conditions designed to simulate the piston-like gas-and-ash explosion events at Caliente21. During the experiments the flat ends of two hollow, cylindrical cores of a Caliente dome rock were pushed together at an applied normal stress of 6 MPa (representative of the depth of tilt and seismic sources27) and one core was rotated (against the other) at an equivalent velocity of 1 m s−1 (see Methods and Extended Data Fig. 5). Friction experiments on magmas have shown that under such conditions frictional melting takes place within as little as about 10 cm of slip13,14,28 confirming the feasibility of this process.

As noted above, microscopic inspection of the fault products experimentally generated in the Caliente dome rock reveals the presence of multiple, chemically heterogeneous melt filaments extruded from crystals adjacent to the fault zones (Fig. 2d; Extended Data Fig. 6). In addition, the interstitial glass of the host rock in the first 0.3–0.4 mm near the fault zone has partially vesiculated (Fig. 2e; Extended Data Fig. 7). To ensure that vesiculation resulted from substantial heat near the fault zone, we have tested the stability of dissolved water in this dome rock at background magmatic temperature by subjecting two small cores to 850 °C for 30 min and 15 h, respectively. We observe that no water exsolved to form vesicles, even after a 15-h dwell (Extended Data Fig. 8). We conclude from these experiments that both the generation of crystal protomelts and the surrounding vesiculation result directly from the frictional work converted to substantial heat during faulting events, and are not due to residence at magmatic temperature. From the similarity of these experimental products of frictional melting to the natural samples of Caliente (described above) we deduce that the cyclic phenomena observed during dome extrusion and explosions at Caliente occur in the presence of strain localization, accompanied by thermal vesiculation.

The occurrence of superheated vesiculation at Caliente can be assessed by modelling the conversion of mechanical work to heat (ΔT) during friction, using31:

Using Byerlee’s friction coefficient μ of 0.85 (at static conditions), a normal stress σn of 6 MPa (ref. 27), a slip velocity V of 1 m s−1 for a duration t of 0.5 s (ref. 21), a density ρ of 2,630 kg m−3 (determined by helium pycnometry), a specific heat capacity Cp of 900 J kg−1 K−1, and a thermal diffusivity k of 10−6 m2 s−1, uplift of the dome would generate a local temperature increase of 860 °C along the arcuate faults. Given that the magma already resides at ~850 °C (ref. 32), and that experimental work has shown that only moderate temperature increase occurs once frictional melt lubricates a slip zone13,14,28, the temperature would not be expected to greatly exceed the melting temperatures of the main rock-forming minerals in the Caliente lava (labradorite and enstatite, which melt at >1,300 °C and >1,400 °C, respectively30). This magnitude of heating would induce water exsolution from the melt in zones of strain localization. Owing to the current eruptive cycles and outgassing activity at Caliente, we consider the system to be open to an extent that allows for exsolution of any oversaturated volatile fraction; thus a total of 0.83 wt% would be expected to remain in the magma at the point of fragmentation at 6 MPa (Fig. 1a). Heating of ~550–860 °C would induce a dramatic oversaturation in water of 0.26–0.35 wt%. Faulting, creation of new surface area, and forced convection during frictional melting would all serve to minimize effective diffusion path lengths and enhance the completion of water exsolution. With such overheating, and thus heightened H2O diffusivity, the kinetic limitation to vesiculation (nucleation and growth) should also be easily overcome, promoting foaming. At a depth of about 300 m such vesiculation would, in turn, reduce the strength of magma and thereby trigger fragmentation33. We therefore conclude that vesiculation can be induced by rapid heating in the conduit.

Water is central to magma ascent dynamics and its contribution to magmatic and volcanic processes results from a combination of both pressure and temperature. Decompression is inevitable and acts throughout magma ascent. Here we argue that heating via both crystallization and shearing processes are equally inevitable. More specifically, the magnitude of viscous and frictional heating may be prodigious, and thus exert a primary control on volatile exsolution. At the rates and magnitudes of heating discussed here, the solubility of water in a melt should be affected before heat loss by thermal conductivity to the cooler surroundings—whether in the core of the magmatic column (where further water may exsolve) or in the country rock—could serve to counteract local heating. Heating during magma ascent deserves adequate consideration in conduit transport and eruption models.

The idea that temperature may dominate the dynamics of water saturation and vesiculation during magma transport in volcanic conduits means that the thermal path experienced by magmas during ascent need to be better constrained. A thorough reassessment of strain localization across deep dykes and shallow conduits should lead to the quantification of shear heating during magma transport. In light of the demonstration that heating may supercede decompression as a driving force for degassing, we call for this concept to be included in the simulation and analysis of magma ascent and eruption.

Methods

Volcanic ash sampling and analysis

The ash samples were collected after each explosion from a location (14° 44′ 35.11′′ N, 91° 33′ 40.69′′ W) approximately 275 m east-northeast from the active Caliente vent. The ash was collected by spreading a clean, 1.4 m × 1.4 m synthetic sheet. We used a paintbrush to carefully brush deposited ash into sample bags. The sheets were thoroughly cleaned after each sample collection and laid out to collect the ash of subsequent events. Owing to the proximity of the sampling location, we are very confident of the source and timing of the ash employed in this study.

The grain size of the sampled volcanic ash was measured using a laser diffraction particle size analyser from Coulter. The density was determined on 25-mm-diameter and 50-mm-long rock cores using a 100-cm3 helium pycnometer from Micromeritics.

SEM analysis and energy-dispersive X-ray spectroscopy

Geochemical mapping across the natural samples and the experimental products was conducted in a Phillips XL 30 SEM using BSE and energy-dispersive X-ray spectroscopy (EDS) run on the Oxford Instruments INCA software. BSE images provide an excellent means of identifying frictional melting textures, because the grey value of each phase relates to the atomic number, or the density of major elements representing the geochemical composition34. A dense phase consisting of heavy elements elastically reflects more electrons and thus shows up in light grey on a BSE image; conversely, an elementally light phase shows up in dark grey.

EDS was used to map the chemical concentration of major elements present in the different phases observed by BSE imaging. The EDS system allows mapping of the distribution of these elements across the main phases. We used an electron beam of 5.5 μm at 20 keV and 8 nA. For the purpose of this study, we monitored the distribution of Si, Mg, Fe, Ti, Na and Al. Comparison of BSE and EDS images verify that the filaments have different chemical compositions.

Electron probe micro-analysis

Geochemical analysis of the phases present in the natural samples and the experimental products was performed in a CAMECA SX 100 Electron Probe Micro Analyser at the Ludwig Maximilian University of Munich in Germany. Probing of the glass and mineral phases was done using a focused electron beam of 15 keV and 20 nA (Extended Data Fig. 6). Note that because we used a focused beam on glass, the measured concentrations of the alkalis, namely Na and K, are reduced by some 0.1–0.3 wt% from what is likely to be present; however, the filaments were too thin to be measured with a defocused beam, which would yield higher inaccuracy. Despite this, the results reveal the chemical distinction between the different phases.

In Extended Data Fig. 6, we present the chemical composition of only the primary minerals and glass, and the protomelts and main frictional melt from the experiments, because the phases were large enough to be analysed. In the natural ash, the filaments are rarely larger than 1 μm (see Extended Data Figs 2,3,4) and so electron microprobe analysis was impracticable without a large degree of contamination from surrounding phases; hence, we used the greyscale in BSE images as well as EDS elemental maps to verify the occurrence of the same processes as observed in the experimental samples.

Fault slip experiments

The friction experiment was conducted in a low- to high-velocity rotary shear apparatus at the University of Liverpool, designed by T. Shimamoto and built by Marui, Japan. The experiment was conducted on two hollow, cylindrical samples with outer and inner diameters of 24.99 mm and 15.86 mm, respectively (Extended Data Fig. 5). The samples were axially loaded using an air actuator at a normal stress of 6.0 MPa, as constrained by the depth of seismicity, and slip was applied on one rotating sample via a servo motor operated at 1,200 rotations per minute, to induce an equivalent slip rate of 1 m s−1, while the other sample was held stationary (see Hirose and Shimamoto35 for further detail of apparatus and method). After the test, the sample was cut and a thin section was prepared.

Testing the stability of volatiles in the dome rock at eruptive temperature

We conducted complementary experiments to test the ability of the rock to vesiculate at high temperature to ensure that foaming observed in the friction experiments results from the very high temperatures achieved during fault slip, instead of simply because the rock used contains a concentration of water (quenched-in at high pressure) higher than that which is stable at atmospheric pressure. For this purpose, two small 8 mm × 8 mm cylindrical samples were heated to a magmatic temperature of 850 °C and one was allowed to dwell for 30 min while the other was allowed to dwell for 15 h. After the experiment, the samples were cut, polished and carbon-coated for SEM analysis.