Shear wave splitting and waveform complexity for lowermost mantle structures with low-velocity lamellae and transverse isotropy

[1] Shear waves that traverse the lowermost mantle exhibit polarization anomalies and waveform complexities that indicate the presence of complex velocity structure above the core-mantle boundary. Synthetic seismograms for horizontally and vertically polarized shear waves (SH and SV, respectively) are computed using the reflectivity method for structures with low-velocity sheets (lamellae''), and for comb-like models approximating long wavelength vertical transverse isotropy (VTI). Motivated by evidence for partial melt in the deep mantle, lamella parameter ranges include ( 1) deltaV(P) from - 5 to - 10%, deltaV(S) = 3deltaV(P); ( 2) 100 to 300 km thickness of vertical stacks of lamella; ( 3) lamella spacing and thickness varying from 0.5 to 20 km; and ( 4) lamellae concentrated near the top, bottom, or throughout the D region at the base of the mantle. Such lamellae represent, in effect, horizontally emplaced dikes within D. Excessively complex waveforms are produced when more than similar to 20% of D volume is comprised of low-velocity lamellae. Many lamellae models can match observed S-diff splitting ( 1 - 10 s delays of SVdiff), but typically underpredict ScS splitting ( 1 - 4 s delays of ScSV). VTI model parameters are selected to address D observations, and include ( 1) 0.5 to 3% anisotropy; ( 2) discontinuous D shear velocity increases up to 3%; ( 3) D thicknesses from 100 to 300 km; and ( 4) VTI concentrated at the top, bottom, or throughout D. VTI models readily match observed splits of ScS and S-diff. We discuss lamellae and VTI model attributes in relationship to waveform complexities, splitting magnitude, triplications from a high-velocity D discontinuity, and apparently reversed polarity SVdiff onsets. The possible presence of melt-filled lamellae indicates that local chemical or thermal perturbations can produce regions that exceed the solidus within D. Such melt could occur in the bulk of D because the melt is either close to neutral buoyancy, advective velocities exceed percolative velocities, or both.