Abstract, We examine vertical component short-period teleseismic seismograms from earthquakes in the Izu-Bonin subduction zone recorded by networks in the western United States for phases associated with conversions from mantle discontinuities. The dominant phase in the stacked P coda is the result of near source S-to-P conversions from a subhorizontal discontinuity at a depth ranging from 650 km to 745 km. We previously used 88 timings of this phase, called S•aoP, to determine the topography of the 660-km discontinuity. Employing the 17 best recorded SaaoP phases, we modeled the SaaoP amplitude accounting for attenuation and correcting for three-dimensional discontinuity topography. Just to the east of the Izu-Bonin subduction zone, the 660-km discontinuity is sharp (< 10 km) and the S-wave velocity contrast across the discontinuity is 0.60 4-0.11 km s-1. This value is 60% larger than the preliminary reference Earth model (PREM) value (70% larger than in the iasp91 and ak135 models) and, unlike estimates from reflected SH waves, is independent of the estimated density contrast at 660 km. The large S-wave velocity contrast inferred here is consistent with recent mineral physics experiments and extrapolations if the mantle at 660 km depth and a few hundred kilometers east of the slab is at near normal mantle temperatures and contains between 3% and 5% cation aluminum, as expected in a pyrolitic mantle.
The 660-km discontinuityThe transformation from ringwoodite (7 spinel) to seismicly faster perovskite and magnesiowiistite occurs at the pressures and temperatures near 660 km depth [Liu, 1979]. Seismic observations of the velocity and density increases across this discontinuity are reference points for thermal, dynamical, and compositional models of the mantle. The dearth of estimates of the velocity jumps, and especially error bounds, is therefore surprising and has stimulated work such as that by Shearer and Flanagan