This is the final installment in a four-part sequence that examines the nature of mantle layering required by the multiple-ScS phases and internal reflections observed within the reverberative interval of SH-polarized seismograms. In this paper, long-period, SH-polarized, multiple-ScS phases reflected once from a mantle discontinuity (first-order reverberations) are used to search for abrupt shear-wave impedance contrasts in the lower mantle. Beneath the geographic region sampled, the depth interval of 1000--2600 km (Bullen's region D') appears free of any distinct, radial layering, in agreement with the majority of recent seismic models and the notion of near-adiabatic compression of a compositionally homogeneous lower mantle. To pass undetected, discontinuities in D' must be either small (SH-polarized reflection coefficient R0<1.0%), Highly intermittent or subject to lateral depth variations well in excess of 50 km. At greater depths, corresponding to the D' region of the lowermost mantle, we find evidence for a reflector of long-period seismic energy situated an average of 325 km above the core-mantle boundary (CMB), similar to the discontinuity proposed by T. Lay and coworkers. Our results indicates a 4.4% increase in shear wave impedance and a 1.7% increase in density if the 2.75% increase in shear velocity proposed by Lay and Helmberger (1983) holds true for our study area.

The discontinuity is observed beneath less than one-third of the 18 seismic corridors examined and does not appear to be a ubiquitous feature of the lower mantle. This, plus an observed correlation of discontinuity depth and lower-mantle velocity heterogeneity, and the lack of experimental evidence for deep-mantle phase transformations strongly favor a compositional origin. Mantle-side layering nearer the CMB, perhaps associated with a thin chemical boundary layer, would mimic the effects of the surface crust, producing a complex wave train of first- and higher-order reverberations superposed upon all other reverberative interval arrivals. It present and unaccounted for, CMB structure would drive estimates of crustal structure and QScS obtained by multiple-ScS waveform inversion away from their true values. Compared to independent estimates, our results admit little bias. Monte Carlo tests are used to assess the sensitivity of this technique to CMB layering, which decreases with decreasing boundary layer thickness. Nonetheless, the results limit CMB transition zones in excess of 20 km to impedance contrasts less than 12% (8% for a 40-km transition) beneath much of the western Pacific, Australia, Melanesia, and Indonesia. ¿ American Geophysical Union 1991