Variability in P-T paths in subducting mantle and crust, and its control on the locations of intraslab earthquakes (Invited)

Intraplate earthquakes in subduction zones occur over wide ranges of depths, as shallow as interplate thrusting if not shallower, and well past the volcanic arc. The mechanism by which these earthquakes occur remains unclear, but at least at intermediate or sub-arc depths, much evidence suggests an important role for dehydration reactions within the downgoing plate. As laboratory data and thermal models improve, it has become clear that the locations of major dehydration reactions should vary considerably between “hot” and “cold” subduction zones (as described by plate age and convergence rate), and with subduction geometry. Recent advances the quality and analysis of regional seismicity data sets, and collocated imaging, have made it possible to tell where earthquakes occur within subducted crust and mantle, to depths well past the volcanic front (to 100-150 km depth). These comparisons require high-accuracy absolute hypocenters, and an independent means to identify the top and bottom of the subducted crust, typically through mode conversions (from high-frequency P-S/S-P conversions or teleseismic P coda receiver functions), and also through travel-time tomography. We compare results of four such studies, in two cold to middle-age subduction zones (central Alaska, northern Honshu) and two young subduction zones (Nankai, Cascadia). In each case, the subducting crust could be identified independently from seismicity, and common velocity models were used to minimize systematic biases between data. Although methodologies differ, a clear pattern emerges: in Nankai and Cascadia, intraplate seismicity occurs at and below the subducting Moho, with very little within subducting crust, while in Alaska and northern Honshu most earthquakes occur within subducted crust. Comparison with published studies elsewhere confirms the overall pattern: conditions for seismogenesis occur within subducting crust in most cases, but not in hot subduction zones, where conditions for seismogenesis occur largely within subducting mantle. Thermal models predict dehydration of crust throughout the seismogenic depth range in all of these subduction zones, so a simple correlation between seismicity and dehydration cannot explain this pattern. However, the operative reactions differ; in most subduction zones seismicity persists through depths where major dehydration reactions are mostly a function of temperature, whereas crust in hot subduction zones follow hotter P-T paths, intersecting dehydration reactions with negative but small Clapeyron slopes. We speculate that the difference in earthquake behavior results from different volume changes associated with the different reactions, or differences in the resulting permeability structure of the region generating earthquakes.