In numerical modeling, we use the philosophy of Occam’s Razor, or the law of parsimony, which states that the simplest construction capable of explaining a phenomenon is the preferred explanation. But in nature, even simple processes always act in concert with a multitude of compounding factors. The writing of this paper is a story of asking, and continually expanding, what is the minimal working model required to interpret a deceptively “simple” rock sample.
Drilling into magma
In 2009, geothermal drilling in the Krafla Geothermal Field by the Iceland Deep Drilling Project (IDDP) was intended to reach 4 km in search of supercritical (and therefore super-hot) hydrothermal fluids. However, at 2100 m depth, drilling hit a snag when the string was repeatedly stuck and experienced loss of cooling fluid. When drill cuttings were retrieved from depth, they contained chips of fresh, glassy material, indicating the presence of still-molten magma stored in the crust.
Drilling into magma was an exciting prospect for both volcano scientists and geothermal production companies who saw the opportunities for in situ monitoring of volcanic systems and high energy-density reservoirs for enhanced geothermal energy production. At the time of the IDDP-1 well, drilling into magma had only been reported once previously, at the Puna Geothermal Venture on Kilauea volcano, Hawai’i. Since then, magma intersection has also been recognized at the Krafla KJ-39 well, and two wells at Menengai volcano, Kenya.
The drill cuttings from IDDP-1 gave an unusual opportunity to ask (1) under what conditions is magma stored in the crust in active volcanic/geothermal settings? (2) What happens to the magma when we drill into it? (3) If magma ascends within the borehole, is drilling into magma safe? What do we need to ensure safety in future drilling efforts?
Is glass a time capsule for magma storage conditions?
When the drill cuttings were first analyzed, the simplest interpretation was that cooling fluids from the well had rapidly quenched the magma. That would imply that the glass, and the crystals and vesicles (bubbles) that it contains, (1) give us direct information about the pressure, temperature, and volatile content of the magma beneath Krafla volcano, and that (2) the magma is quenched before significant changes can occur.
Because volatiles such as H2O, CO2, sulfur, and halogens have solubilities in silicate melts that are highly sensitive to pressure and temperature, their behavior can give insight into the conditions of magmatic systems. The glass chips still contain a moderate amount of water: 1.77 wt.% which is more than would be typical for a basaltic rifting system like Krafla, but lower than most arc volcanoes above subduction zones, and a small amount of CO2, up to 150-180 ppm. The glass chips also contain a large number of very small vesicles, which would have been vapor bubbles growing in the magma before and as it solidified. If we assume that this vapor phase was mostly H2O, then a basic volume conservation suggests that a measurable, but small, amount of water (about 0.03 wt.%) should be in the bubbles.
This presents a conundrum, though. Empirical models of solubility, constrained from laboratory experiments, suggest that at 150 ppm of CO2, a likely temperature of 900 °C, and lithostatic pressure from the weight of the rock above (55 MPa), the magma should be saturated with 1.87 wt.% H2O, significantly higher than the about 1.80 wt.% water we can account for.
What’s missing?
Does that mean this magma wasn’t saturated at lithostatic pressure? If so, this presents a big issue for the volcanological community. When we look at a rock, or a mineral in that rock, or a tiny pocket of melt and vapor trapped as an inclusion in that mineral, we see the entire integrated history that sample has experienced. To learn about magmatic systems, we want to trace back to where that sample came from. We often assign the origin depth on the basis of volatile saturation pressure. And our understanding of what drives volcanic eruptions suggests that the way magma moves through the crust to the surface is intimately connected to the presence of these volatile phases, both dissolved in the melt and in a co-existing vapor, which we often believe to be present. So, if we have new evidence of a non-erupting body of magma being stored undersaturated, that seemingly contradicts a lot of other evidence and mechanistic understanding of volcanic systems.
A simpler solution is that our interpretation of the IDDP-1 chips is missing a key process. So, we went back to look at these small vesicles in the glass. The assumption had been that in a water-rich system, these bubbles should contain only a small fraction of the total water. But H2O and CO2 compete in mixed-volatile systems, and only small amounts of CO2 need be present to reduce the amount of H2O soluble in the magma.
Placing tighter constraints on magma response to drilling
To quantify this effect, we took an existing model for H2O diffusion into vapor bubbles, and added the capability to jointly solve for the diffusion of CO2. We found that not only does adding a reasonable amount CO2 to the starting melt – now lost to the bubbles – restore the possibility of lithostatic vapor saturation, but it also places tighter constraints on the transformation from stored magma to a quenched glass. In order to match the number and size of vesicles, the volatiles left in the glass, and the glass transition temperature (an independent measure of cooling rate at the instant in time when the volatiles stop diffusing), the magma could only have responded to drilling along a small family of paths of 5-10 min duration with simultaneous rapid decompression and cooling.
Figure 3: Plausible degassing and cooling paths (blue) and the resulting melt/glass composition (blue markers) from lithostatic magma storage (red circle), compared to the glass measurements (gray) and the solubility model of Liu et al. (2005) (black curves) with red annotations for the hydrostatic (16 MPa) and lithostatic (55 MPa) pressures.
These results provide a robust observation of magma stored in the crust, saturated with volatiles at lithostatic pressure, validating previous assumptions that show magmas may be volatile saturated, even when they are not erupting. Furthermore, the model we developed provides a tool for interpreting the glass quenching process.
Now that we have an explanatory model for what happened in past magma drilling events, we can look to the future and plan for the next time. As we move towards a green energy transition, we will need more high energy-density geothermal resources, which will mean more drilling close to and into magma. With a forward model for decompression and cooling of magma, we can engineer optimal drilling conditions to control the magma response and prevent unintended magma movement, ensuring safe access.