No abstract

The Earth’s lowermost mantle large low velocity provinces are accompanied by small-scale ultralow velocity zones in localized regions on the core-mantle boundary. Large low velocity provinces are hypothesized to be caused by large-scale compositional heterogeneity (i.e., thermochemical piles). The origin of ultralow velocity zones, however, remains elusive. Here we perform three-dimensional geodynamical calculations to show that the current locations and shapes of ultralow velocity zones are related to their cause. We find that the hottest lowermost mantle regions are commonly located well within the interiors of thermochemical piles. In contrast, accumulations of ultradense compositionally distinct material occur as discontinuous patches along the margins of thermochemical piles and have asymmetrical cross-sectional shape. Furthermore, the lateral morphology of these patches provides insight into mantle flow directions and long-term stability. The global distribution and large variations of morphology of ultralow velocity zones validate a compositionally distinct origin for most ultralow velocity zones.

Thin patches with ultralow velocities have been proposed to exist at the core‐mantle boundary (CMB). The detection and mapping of ultralow velocity zones (ULVZs) are difficult, in part, because of limited source‐receiver geometries of seismic phases used in ULVZ modeling. Here we develop a new approach that simultaneously utilizes ScS precursor and postcursor energies to investigate the CMB region for ULVZ structure. We stacked source‐deconvolved ScS waveforms within 1.5° geographic bins to extract ScS precursor and postcursor energies, if present, with ScS effectively removed from waveforms. We investigate the CMB beneath the central Pacific Ocean, and evidence for ULVZs is clearly apparent. Geographic bin stacks possessing similar ScS precursor‐plus‐postcursor behavior are grouped by using cluster analysis to produce more robust waveforms by enhancing the signal‐to‐noise ratios. Synthetic seismograms that demonstrate the amplitude and timing of the ULVZ arrivals are sensitive to ULVZ thickness and internal velocities. To pursue local ULVZ properties we processed 13,850 1‐D synthetic models with various ULVZ thicknesses and internal properties, using the identical ScS‐stripping method as with the data. A best fitting model was found for each geographical bin cluster by using an amplitude‐sensitive cross‐correlation algorithm. While limitations exist due to 1‐D modeling, strong lateral variations are clearly apparent in ULVZ thickness and properties across the large low shear velocity province (LLSVP) margin in our study area. Inside hypothesized LLSVP edges, ULVZs appear to distribute unevenly, suggesting 3‐D variations of convection currents.

No abstract