A mineral buried 630 kilometres beneath the Mariana subduction zone can transform into a denser crystal structure in under 100 seconds, releasing a heat spike exceeding 1,500 degrees Celsius above background temperature, and that single physical event is what drives some of the most powerful deep earthquakes on Earth. Findings published in Science Advances in April 2026 quantify the complete chain of events inside a rupturing olivine fault at pressures between 15 and 20 gigapascals, the equivalent of conditions inside a cold slab plunging through the mantle transition zone at depths of 400 to 700 kilometres.
The mineral at the centre of this is olivine, the most abundant rock in the upper mantle and the primary material inside the cold interior of a subducting tectonic plate. When a tectonic plate bends and descends into the Earth at a subduction zone, its outer edges heat up and olivine transforms into denser mineral forms at the pressures encountered below 400 kilometres. The cold inner core of the slab, called the metastable olivine wedge, stays frozen in its original crystal structure well past the depth where it should have transformed. That frozen core is packed with olivine sitting in a state it physically should not occupy at those pressures, held there only because the temperature is too low for the transformation to proceed through normal diffusion, the slow migration of atoms through a crystal lattice. Scientists had long assumed that the coldest, deepest part of this wedge was essentially earthquake-proof because the standard transformation mechanism requires warmth to operate. The Mariana subduction zone, which reaches depths of around 630 kilometres and carries some of the coldest slab material on the planet at temperatures near 820 Kelvin, was producing deep earthquakes in exactly the zone the models said should be silent.
To understand how rupture fires inside cold olivine, researchers at Ehime University and the Japan Synchrotron Radiation Research Institute compressed real mantle olivine samples to pressures of 15 to 20 gigapascals at temperatures between 870 and 1,320 Kelvin, replicating conditions inside a deeply subducted slab, then deformed them under load while monitoring stress, strain, pressure, and acoustic signals in real time using synchrotron x-ray equipment at the SPring-8 facility in Hyogo, Japan. Three of the experimental runs produced full throughgoing faults, meaning ruptures that crossed the entire sample from end to end, at temperatures as low as 970 Kelvin, well below the threshold at which diffusion-driven phase transformation becomes significant.
What the experiments captured inside those fault zones changes the accepted picture of how deep earthquakes start. The olivine grains did not simply transform into the expected high-pressure mineral forms. Instead, stress concentrated into narrow bands called kink bands, regions where the crystal lattice of individual olivine grains buckled and folded along specific internal planes under the applied load. Kink bands of this type are familiar in metals and industrial alloys deformed at high strain rates, but their role in mantle olivine under subduction-zone conditions had not been directly observed before. Once the kink band formed, the crystal structure of olivine shifted into an intermediate mineral called poirierite, a dense metastable form of the same magnesium iron silicate that sits at a higher energy state than either olivine or the stable high-pressure phases. Poirierite has been identified previously only in shocked meteorites, material that experienced sudden extreme compression during impact events. Its presence inside a laboratory fault zone replicating subduction conditions is a direct measurement of a transition pathway that bypasses slow atomic diffusion entirely.
Poirierite does not stay poirierite for long under these conditions. It carries the highest stored energy of any olivine polymorph across the full pressure range examined, at least up to 20 gigapascals. When it completes its transition into ringwoodite, the stable high-pressure mineral, it releases that stored energy as latent heat in a burst. The direct olivine-to-ringwoodite transition releases enough heat to raise the local fault zone temperature by approximately 200 Kelvin at 20 gigapascals. The poirierite-to-ringwoodite step releases significantly more, because the energy accumulated during the olivine-to-poirierite stage is also discharged in the same event. In one experimental run conducted at a background temperature of 970 Kelvin, platinum marker foils embedded in the fault gouge melted completely. Platinum melts above approximately 2,500 Kelvin at those pressures. The background temperature was 970 Kelvin. The fault zone flash-heated by more than 1,500 Kelvin in the window of the slip event, a temperature spike localised to a gouge layer measuring between 1 and 10 micrometres thick.
That flash heating is not just a temperature reading. It is the physical mechanism that collapses the mechanical strength of the fault zone and allows the rupture to propagate. The gouge layer, already weakened by grain size reduction and pulverisation during kink band formation, becomes effectively frictionless in the window of the heat spike. Fault slip rates in the experiments reached above 0.05 per second in the fastest runs, consistent with the Peierls creep regime, a deformation mechanism that becomes dramatically more efficient as temperature rises and that accelerates the slip rather than arresting it. Grain-size-sensitive creep, the mechanism that previous models required to explain deep fault weakening, plays no necessary role in this process. The fault ruptures and propagates entirely through the heat spike and the mechanical instability it creates in the kink-band damage zone.
The geometry of the damage zone itself matters. In the experiments, olivine grains buckled and accumulated microcracks along their kink band boundaries over a period of 12 to 40 minutes of progressive deformation before the main fault formed. Those microcracks, each smaller than one micrometre in length, coalesced into larger shear fractures once crack density crossed a threshold, even at confining pressures between 15 and 20 gigapascals that would normally suppress brittle failure entirely. The stress concentration at kink band tips, generated by the pileup of dislocations inside the buckled crystal, drove that coalescence without any requirement for elevated temperature or fluid involvement. The entire sequence, kinking, microcracking, coalescence, phase transition, heat spike, and fault propagation, operates inside cold, dry olivine under nothing more than the pressure and strain accumulation that any deeply subducting slab experiences.
That sequence operates most efficiently in precisely the conditions that prior models excluded. The cold core of a subducting slab, where temperatures are lowest and diffusion slowest, is where kinking and pseudo-martensitic phase transition carry the highest relative advantage over competing mechanisms. In the Tonga, Izu-Bonin, and southwest Japan subduction zones, deep earthquake hypocentres cluster in the warmer outer rim of the metastable olivine wedge, where grain-boundary nucleation of high-pressure phases is feasible, and the cold core is largely aseismic, consistent with the old diffusion-controlled model. In the Mariana subduction zone, that pattern breaks down. Deep earthquakes occur throughout the cold core at depths around 630 kilometres, where slab temperatures sit near 820 Kelvin. The strain rate in the Mariana slab at those depths is independently estimated to be elevated relative to other subduction zones, and elevated strain rate directly increases the probability of kink band formation and shear localisation in metastable olivine. The mechanism identified in this laboratory work accounts for the Mariana anomaly without requiring any additional heat source, fluid, or process outside what a cold, fast-deforming slab already provides.
The transition can proceed even at near-room temperature under sufficient stress. Shock compression experiments on single crystals of mantle olivine at room temperature have confirmed the olivine-to-ringwoodite transition via shear mechanism completing within tens of nanoseconds under impact conditions. The pseudo-martensitic pathway is insensitive to temperature in a way that diffusion-driven transformation is not, because it does not require atoms to migrate through the crystal lattice. It requires only the right stress orientation on the right crystal plane. Inside the metastable olivine wedge, where the crystal preferred orientations are controlled by the deformation history of the slab during subduction, those conditions are met whenever strain accumulation reaches a threshold level.
No seismic network currently in operation directly monitors the mantle transition zone at 630 kilometres depth beneath the Mariana arc at the spatial resolution that would resolve individual fault initiation sites within the metastable olivine wedge. The earthquake hypocentre locations that identify the Mariana cold-core seismicity come from regional and global seismic array inversions, which carry positional uncertainties of several kilometres at those depths. The internal structure of the metastable olivine wedge, its precise thermal profile, thickness, and strain rate distribution, is constrained by seismic tomography and thermal modelling rather than direct measurement. What is now on record is the physical mechanism at grain scale, confirmed in real mantle olivine at verified pressures and temperatures, producing full fault rupture in the cold slab regime that theory said was locked.
Source:
“Faulting Triggered by a Quasi-Diffusionless Shear Transition of Olivine in Deep Subducted Slabs” by Kohei Matsuda, Tomohiro Ohuchi, Sayako Inoué, Yuji Higo, Noriyoshi Tsujino, Sho Kakizawa, and Takeshi Sakai. Published in Science Advances, April 8 2026.
