Partial melt in the lowermost mantle near the base of a plume

Abstract

Precursors to the seismic core phase PKP have long been used to study small-scale heterogeneities at the base of the mantle. They are preferred to PKP coda waves (post-cursors) because the latter are more biased by shallow structure, and so are cruder probes of the deep Earth. In this work, however, we present an array-based analysis of PKP coda waves that provides a unique opportunity to image small-scale structure in the deep mantle. Seismograms of a Peru earthquake recorded by an array of broad-band seismometers in Tibet show strong coda waves following PKPCdiff. The coda waves are strongest for distances of 154°-157° at frequencies above 0.5 Hz. No such strong coda waves exist after PKPDF and PKPAB, and an empirical source time function of the earthquake is likewise simple. This implies that the coda waves are being created by anomalous structures deep within the Earth. Using slant stack analysis we find that the ray parameter of the PKPCdiff coda waves is similar to that of the main PKPCdiff phase, implying that the coda waves are not simple precursors to the minimax PKPAB phase. Combined modelling of the differential traveltimes, differential slownesses (PKPDF - PKPCdiff), and envelopes of the coda suggests that the waves are created by strong heterogeneity at the source side of the lowermost mantle that scatters incident P energy into post-critical reflections from the inner core boundary (P.PKPCD). Although the precise geometry is unconstrained, if we use single scattering theory the data are consistent with a region of at least 15 per cent P-wave velocity reduction. The most likely cause for such a strong velocity anomaly is the presence of partial melt. The geographical location of the anomalous region, beneath the Amazon River in South America, roughly corresponds with the expected base of a newly discovered plume. © 2006 The Authors Journal compilation © 2006 RAS.