The structure of the lowermost mantle determined by short-period P-wave amplitudes

Abstract

Summary. The seismic velocities in the D″ region (lowermost 200 km of the mantle) are recognized to be anomalously low, though the details of the velocity structure are not known. The structure of D″ is important, in particular whether a smooth velocity model is appropriate or not. A smooth decrease in the seismic velocities would be consistent with a thermal boundary layer at the base of the mantle. We have used the amplitudes of short-period (T= 1s) P- and diffracted P-waves to investigate the internal structure of D″. A short-period amplitude data set is obtained by using underground nuclear events as sources and applying receiver corrections to the amplitudes. Receiver effects are largely responsible for the factor of ∼8 scatter in the amplitudes of the North American WWSSN stations. Applying receiver corrections reduces the scatter to a factor of ∼2, thereby providing a quantitatively useful amplitude profile into the core shadow. Using Soviet events and North American WWSSN stations, the D″ layer beneath the north polar region is well sampled. The core shadow (at T= 1 s) begins sharply at a distance of Δ= 95.5 and the slope of the amplitude decay is well defined. Also, the amplitudes decrease slightly from Δ∼87 to ∼90, then increase to Δ∼95. Synthetic seismograms are used to test various earth models, with the important conclusion that the amplitudes from smooth D″ models with a small velocity gradient in D″ decay too slowly in the shadow. This mismatch cannot be satisfactorily explained by random forward scattering or a thin low Q layer within D″. Anelastic calculations show that a thin low Q layer in D″ decreases the amplitudes before the shadow, with little effect on the decay slope in the shadow. All of the features of the observed amplitude profile can be explained as the interference effects of a model that has a low-velocity zone in the upper part of D″ followed by a normal velocity gradient in the lower part of D″. This type of model (POLAR series) also explains the scatter often observed in dT/dΔ beyond Δ∼90. The interference effects and required velocity changes in D″ are small, and long-period amplitudes will respond only to the averaged velocity gradient in D″. The POLAR models imply a compositional and/or phase change at the top of D″. Thus, the preferred seismological model does not allow the D″ region to be interpreted as a single thermal boundary layer between the mantle and core.