Spatial Distribution of Deep Earthquakes Controlled by Spatially-variable Strain-rate in Subducting Slabs

Consideration of proposed mechanisms for deep (~300-680 km) earthquakes, such as dehydration embrittlement, shear instability, or transformation faulting, has focused on the ability of these mechanisms to explain the global depth-seismicity profile in terms of the thermal structure of the subducted lithosphere (slabs). However, such analysis fails to explain the variability of observed seismicity in slabs. I propose that the location of deep earthquakes is controlled by the rate of deformation in the slab, similar to the strain-rate dependence of seismicity in tectonic plates at the Earth's surface. Analysis of the strain-rate release (Bevis, 1989) from earthquakes occurring during 1964-2014, shows that seismic strain-rate release follows the seismicity pattern, with magnitudes of 5e-15 to 1e-18 1/s. To test if this seismically-released strain-rate is indicative of the overall strain-rate of the slab, I use numerical simulations of subduction with non-linear rheology and compositionally-dependent phase transitions. These models exhibit strong variability in the strain-rate magnitude both in space and time. High strain-rates occur in bending regions of the slab and migrate as the slab buckles and folds at the base of the transition zone. However, in between these strongly-deforming regions the strain rate is low due to the strong temperature-dependence of viscosity and high yield strength of the slab. Time snap-shots of simulated slabs exhibit similar depth-strain-rate profiles as observed, with some profiles exhibiting continuous strain-rate with or without a transition zone peak and other profiles containing large gaps or shallower strain-rate peaks. These results support the hypothesis that the first order control on the distribution of deep earthquakes is high strain-rates regions within a slab that is also rheologically strong enough to accumulate strain that is not accommodated by viscous flow. In these region, multiple mechanisms for triggering strain release as an earthquake are viable and may vary between subducton zones depending on the thermal structure and ability to advect water to depth. At shallower depth, the simulations predict consistently high strain-rates, which decrease down to 250-300 km, consistent with the seismicity being controlled by the triggering mechanism rather than the strain-rate distribution.