Mapping small-scale mantle heterogeneities using seismic arrays

In recent years array seismology has been used extensively to detect and locate the small scale (~10 km) structure of the Earth. In the mantle, small scale structure likely represents chemical heterogeneity and is essential in our understanding of mechanical mixing processes within mantle convection. As subducted crust is chemically distinct from the background mantle, imaging the remains of the crust provides a tracer for convectional flow. Evidence for heterogeneities has been found in the lower mantle in previous seismology studies but the arrivals associated with such heterogeneities are difficult to detect in the seismic data as they are typically low amplitude and are often masked by a multitude of larger amplitude arrivals. In this study we find global and regional seismic heterogeneities in the mantle by processing teleseismic earthquake data through array seismology methods. We find global patterns of heterogeneity using a stacking approach. To locate regional heterogeneities, we target the "quiet" window prior to the PP arrival for earthquakes with epicentral distances of 90-110°. Within this time window, we enhance the weak coherent energy that arrives off great circle path by calculating the observed directivity (slowness and backazimuth) and using a semblance weighted beampower measure. We use the directivity and travel times of suitable precursors to back-trace the energy to the origin of P-to-P reflections, using a 1D raytracer. Most of the P-to-P reflections that we observe have reflection origins in the upper/mid mantle. Beneath the western Pacific subduction zones, such reflections show a good correlation with subduction zone contours that are derived from subduction zone seismicity, and correlate well with tomography gradients of 0.01-0.5% per degree, interpreted as the edge of the slab. Deep mantle reflections (>600 km) are also observed to depths of ~1900 km. The locations of these heterogeneities are combined with previous seismological observations and with historical trench locations to discuss subduction and mantle flow scenarios as proposed from earlier geodynamical and numerical modelling of mantle convection.