P and S waves diffracted around the core and the velocity structure at the base of the mantle

Decay spectra and dT/dΔ on 13 great circle paths for P and seven for SH, are used in a joint interpretation of P and SH diffraction data, in terms of a transition zone with linear velocity profile above the core—mantle boundary (CMB). Error analysis reveals that the effect of large-scale lateral variations at the base of the mantle can be significant, and this effect is accounted for in the standard errors as associated with the data and used in the inversion procedure. Employing a full wave theoretical approach discussed previously, synthetic decay spectra and dT/dΔ are calculated for every individual path, to be compared to the real data. Following recommendations of a previous paper, it is illustrated here by examples that thickness zt and velocity gradients dα/dz, dβ/dz of the transition zone have a characteristic effect on the decay spectra of P and SH, and these effects are taken into account in searching the model space for the best fitting model (in the least-squares sense). Subject to the constraint of a linear velocity profile, the combined P and SH data set allows a thickness zt of 50–100 km and the P and S velocities both decrease with depth. The preferred model, PEMC-L01, has zt= 75 km and dα/dz=dβ/dz ⋍ −0.0019; the S velocity gradient is close to critical and the ‘index of inhomogeneities’ η ⋍ 4.5. There is a suggestion of an additional low -Q at the base of the mantle, but this is not absolutely required due to the large standard errors associated with the relevant data. The models are determined primarily from the decay spectra. They also fit the dT/dΔ data reasonably well, but these data do not discriminate between several different models considered here. Allowing for the possibility of lateral variations, previously published data can be reconciled with the present model, possibly with some modifications warranted by short-period information. The physical implications are an anomalous density gradient and/or superadiabatic temperature gradient above CMB. The numerical results with regard to density and/or temperature must be treated with caution, since the seismic parameters dø/dz and η are very sensitive to apparently small changes in the model.