GoKawiilDespite their importance, estimates of the pressure, temperature, saturation state, geometry and location of magmatic storage regions vary widely for even the most-studied individual volcanic systems. Geothermobarometers using mineral–mineral chemistry or phase equilibria have been applied to volcanic materials to constrain magmatic origins, but these methods have large uncertainties (about 50–200 MPa or about 2.5–10.0 km) arising from sparse laboratory constraints, limits on analytical precision, assumptions of local equilibrium and interdependence between temperature and pressure2,3. Furthermore, magmatic reservoirs are challenging to image with geophysical methods owing to limitations in resolution and poorly constrained relationships between lithology and geophysical signals, resulting in typical uncertainties on magma depth of approximately 0.5–10.0 km (ref. 4). As a result, the scientific community lacks consensus on even the fundamentals of the spatial distribution of melt in the crust5.
Deep drilling in hydrothermal fields offers the unique potential for placing tight constraints on the location, temperature, pressure and chemistry of melt stored in natural volcanic systems. Hydrothermal drilling has occasionally intersected magma: dacite in the Puna geothermal field (Hawaii), trachyte at Menengai (Kenya) and rhyolite at Krafla (Iceland). At Krafla, the KJ-39 and Iceland Deep Drilling Project-1 (IDDP-1) boreholes directly intersected magma at roughly 2,500 and at 2,104 m depth, respectively6,7, which was not anticipated on the basis of coarse geophysical imaging before drilling. Retrospectively, a magma body was recognized during reanalysis of magnetotelluric data8 and from seismic imaging9,10, which have still been unable to resolve questions about its lateral and vertical extent.
Silicic glass fragments were recovered from the IDDP-1 borehole, which were quenched through interaction with drilling fluids7,11. The precisely known recovery depth, and temporally and spatially constrained ascent and quench, makes them ideally suited to resolve the unknowns of magmatic storage, as the melt has not been subject to the complex ascent processes that afflict the products of volcanic eruptions. Indeed, the glass chemistry has sparked discussion about the origins of the magma from (1) partially molten, hydrothermally altered basaltic crust12 or (2) mantle-derived basalts evolved by means of fractional crystallization13 and about the degree of crustal assimilation1,12,14,15. Despite the direct sampling of the glass, established equilibrium-based methods to determine pressure and temperature have yielded wide constraints for these magmas (Fig. 1a): various geothermometers yield different estimates and uncertainties from two-pyroxene equilibration between 920 °C and 940 °C (refs. 1,16) and 890 °C and 910 °C (refs. 12,17), whereas the phase assemblage places the weak constraint between 800 °C and 950 °C (ref. 1). Pressure estimates from projection on the haplogranitic ternary suggest <50–100 MPa, whereas Rhyolite-MELTS yields 44 ± 11 to 47 ± 32 MPa depending on oxidation state, all of which require an assumption of volatile content and fail to model this system in that they produce quartz + plagioclase ± orthopyroxene, when the observed phases are plagioclase + two clinopyroxenes18. The pressure of magma is often, instead, determined by the volatile (H 2 O and CO 2 ) concentrations of crystal-hosted melt inclusions, which are sensitive to pressure and temperature and less susceptible to decompression-induced changes than the residual melt (quenched to glass). Melt inclusions are arguably inappropriate to investigate the pressure and saturation state of the IDDP-1 magma for three reasons: (1) the few crystals present in the IDDP-1 chips frequently show dissolution (melt embayments and rounding)1, so a meaningful, reliable population has not been isolated; (2) the pressure/depth of melt inclusions are usually determined on the basis of the assumption of volatile saturation and are, by their very nature, contained within individual crystals and thus separated from mineral pairs, which could provide an independent pressure determination; and (3) are produced preferentially during disequilibrium crystal growth and so systematically oversample non-equilibrated conditions during transport or perturbations from background3 and therefore are poorly suited to investigate equilibrium storage. Instead, we are left with the measured volatile contents of the glass, which correspond to saturation pressures between about 35 MPa and 45 MPa (Fig. 1a and Extended Data Fig. 1), below the lithostatic pressure (about 50–57 MPa) and above the hydrostatic pressure of the well (about 16 MPa)1. These measurements have been interpreted to suggest that the magma is either (1) stored and degassed to equilibrium at a pressure less than lithostatic owing to interaction with the hydrothermal system1,12 or (2) originally undersaturated14,19.
Fig. 1: Conflicting petrological constraints on Krafla magma storage conditions. Full size image a, Temperature (red) with depth in the borehole6 versus geothermometric constraints from augite–pigeonite and clinopyroxene–orthopyroxene–plagioclase–magnetite–ilmenite and Rhyolite-MELTS1 plotted at the recovery depth, with a secondary axis showing lithostatic pressure at the corresponding depth and saturation pressure (blue, from VolatileCalc1,19; error bars indicate 1σ uncertainty). b,c, Although a trans-crustal mush arrangement of magma storage (b) is typically favoured5, classic views of a basalt-underplated continuous, large rhyolite source (c) have also been proposed at Krafla12,13,41 and are not distinguishable based on present geophysical observations8,9,10. d,e, Backscattered electron images of IDDP-1 glass chips that show the variability in vesicularity. Scale bars, 1 km (b); 500 μm (d,e).
Although the drilling fluids rapidly cooled the magma, it was still subject to decompression following intersection by the drill string14, resulting in remobilization; the magma flowed 8 m up the well in 9 min (ref. 20). So we ask, can the IDDP-1 glass be used as a direct record of storage conditions? Here we expand a bubble growth model for H 2 O (ref. 21) to include CO 2 , coupled to a model for water species interconversion22. As well as the dynamics of bubble growth, a direct output of this model is the residual volatile content in the melt/glass between bubbles from which we constrain the magma storage conditions of pressure, temperature and volatile composition against the measured glass chips.
The glass fragments contain total water of 1.3–2.0 wt%, with an average of 1.8 wt%, consistent across several studies1,11,13,19,23,24, and 50–200 ppm CO 2 (refs. 1,19). Vesicularity in the quenched glasses is low, with most chips having <6 vol%, although occasionally up to about 15 vol% (ref. 14). Bubble sizes are 1.5–75.0 μm, with an increase in bubble size with drilling time14. Bubble number densities range between 1011.7 and 1015 m−3, which are inferred to have nucleated during drilling-induced decompression at rates of 105–107 Pa s−1 (refs. 14,25). OH/H 2 O m ratios are between 1.68 ± 0.45 and 2.19 ± 0.37, increasing over time19, and are slightly lower in more vesicular, 2.07 ± 0.20, than in less vesicular, 2.13 ± 0.38, fragments1.
We explore which pressure–temperature (P–T) paths best reproduce the volatile chemistry and vesicularity of the IDDP-1 glass. Although the cooling from the drilling fluid should be rapid, we use a thermal model to seek paths consistent with measurements of the natural glass; geospeedometry through differential scanning calorimetry indicates that the IDDP-1 magma cooled through the glass transition regime (T g ) at 7–80 °C min−1, with a T g of about 480 °C (ref. 26). We begin with one-dimensional cooling in a planar geometry from an initial temperature of 900 °C by means of conduction and forced convection at the surface with steam (400 °C and 16 MPa), in which drilling fluid is in direct contact with the melt. At a planar interface, cooling rates far exceed those measured in the glass. However, at larger length scales, thermal diffusion is inefficient, such that we can only reproduce the cooling rates in a narrow region between 0.3 mm and 1.3 mm from the interface (Extended Data Fig. 2). To produce a nearly homogeneous glass, similar cooling rates must be sustained across a large portion of the sampled material. This is consistent with fracturing or fragmentation of the material during cooling (Fig. 1d,e) to enhance the surface area and enable cooling from several directions.
If we assume that magma fragments (by any mechanism) during the onset or early progression of cooling, we should model the thermal evolution instead using a spherical geometry. Under these conditions, we reproduce the measured cooling rates at T g in the interior of melt fragments of radius 9–20 mm (Extended Data Fig. 3), a spacing common in perlitic glass27,28. This results in a timescale for cooling from storage to T g (at 480 °C) in about 4 min, with a decelerating cooling rate.
Now we turn to the pressure path experienced by the magma. During drilling, the magma should experience rapid decompression, reflected by high bubble number densities14,25. However, the volatile composition and vesicularity of the glass changed only modestly over 9 h of drilling14, suggesting that decompression of the magma was also progressive. We suggest that the decompression could be offset at increasing depth by the pressure drop generated by viscous resistance to flow up the borehole. We estimate the relevant length scale using the Hagen–Poiseuille equation:
$$\Delta P=\frac{8\mu {H}^{2}}{{R}^{2}}\frac{1}{{(1+\phi )}^{2}}\frac{{\rm{\partial }}\phi }{{\rm{\partial }}t}$$ (1)
... continue reading