Numerical tests on the seismic visibility of metastable minerals in subduction zones

Abstract

In current views on the origin of deep-focus earthquakes the presence of metastable olivine in subducting slabs plays a central role. In this study we construct synthetic slab models to investigate the expression of metastable minerals in seismic velocities. In particular, we investigate the visibility of different slab mineralogy for seismic traveltime studies. From forward modeling we construct a range of mineralogical models which include phases with olivine and pyroxene structures. Differentiation of subducted lithosphere with respect to the surrounding mantle is included, and critical temperatures are used to bound the eventual wedge of metastable minerals. A case with pure olivine mantle and slab is also considered. From the mineralogical slab models we compute seismic P-wave velocities including uncertainty bounds due to a variation range in mineralogical composition, elastic data and slab temperature. As model predictions we compute the traveltime signatures of different slab models by 2-D ray tracing for a number of earthquakes located at the top and bottom side of slabs. We infer that upper and lower bounds on computed mineralogy lead to significantly distinct seismic velocities for slabs with equilibrium or metastable mineralogy. The predicted traveltime signature (including uncertainty bounds) of different slab models suggests the possibility to discriminate between metastable and equilibrium slab models. Models with a pure olivine mineralogy lead to differences in traveltime reaching 1 s. The consideration of non-olivine phases is thus crucial when looking for a weak seismological signal using forward modeling. Further, we conclude from our results that the traveltime signature of slab mineralogy strongly depends on earthquake position. Specifically, events at the top side of slabs yield significantly different traveltime signatures from events located at the bottom side. Top-side events are most discriminating in traveltimes directly up-dip of the slab, whereas bottom-side events are most discriminating in traveltimes observed directly above the slab, i.e. above the mantle wedge. This leads to the important conclusion that highly accurate event locations and accurate knowledge of 3-D structure outside the slab region are prerequisites for studying slab mineralogy with traveltimes.