The core-mantle boundary illuminated by PKKP diffracted waves

Ultra-low velocity zones (ULVZs) are small and discrete patches of low seismic velocity at the core-mantle boundary (CMB). There are several proposed explanations of ULVZs, including that they may be partially molten. However, any melt component to a ULVZ is expected to be iron rich and dense and would therefore likely drain to the CMB under gravity. If melt occurs in ULVZs then we would also expect an extensive layer of melt at the CMB that has not been observed by seismologists.

PKKP diffracted (PKKPdiff) is a PKKP phase observed at epicentral distances beyond its ray-theoretical cut-off, where for at least some of the path the seismic energy diffracts along the CMB, leading to high sensitivity to CMB structure. We have examined the CMB using an extensive dataset of high-frequency PKKPdiff waves, coupled with the use of short-period synthetic seismograms, in order to place a limit on the thickness of such a layer. We find that within the scatter of our data, a global layer of several kilometres of melt could be present at the CMB and be essentially seismically invisible. This is thick enough to have profound consequences on mantle dynamics and core-mantle interaction.

As well as one dimensional structure, PKKPdiff is also sensitive to three dimensional structure at the CMB and this may be the cause of much of the scatter in our dataset. For some events, we observe trends in the data as function of azimuth indicating lateral variation in structure. Potential structures include ULVZs themselves, the less extreme large low velocity provinces (LLVPs), high velocity subducted slab material and topography, both long and short wavelength. We test how potential three dimensional structures may manifest in the data. Finally we examine what trends in our dataset can be explained by these three dimensional structures. Interpretations in terms of location are complicated due to PKKPdiff having multiple diffraction paths at the CMB.