Rolling hills on the core–mantle boundary

Recent results suggest that an iron-rich oxide may have fractionally crystallized from a primordial magma ocean and settled on the core–mantle boundary (CMB). Based on experimental results, the presence of only a few percent of Fe-rich oxide could slow seismic waves down by several percent. This heavy layer can become highly undulating as predicted from dynamic modeling but can remain as a distinct structure with uniform velocity reductions. Here, we use the large USArray seismic network to search for such structures. Strong constraints on D″ are provided by the core-phase SKS where it bifurcates, containing a short segment of P-wave diffractions (Pd) when crossing the CMB, called SKSd. Synthetics from models with moderate velocity drops (less than 10%) involving a layer with variable thickness, perhaps a composite of sharp small structures, with strong variation in thickness can explain both the observed SKSd waveforms and large scatter in differential times between SKKS and SKS. A smooth 3D image is obtained from inverting SKSd waveforms displaying rolling-hills with elongated dome-like structures sitting on the CMB. The most prominent one has an 80-km height, ∼8° length, and ∼4° width, thus adding still more structural complexity to the lower mantle. We suggest that these results can be explained by a dynamically-stabilized material containing small amounts (∼5%) iron-rich (Mg,Fe)O providing a self-consistent physical interpretation.