Global Lower Mantle Attenuation Model and the Origin of Lower Mantle Seismic Heterogeneities

The details of heterogeneities in the lower mantle have increased considerably during the last decades thanks to seismic imaging revealing ULVZs, D” layer, PERM anomaly and now mega-ULVZs. However, the origin of these anomalies is actively debated, as seismic velocities alone cannot disentangle between thermal or compositional origins. Seismic attenuation can provide an additional perspective to seismic velocity for constraining physical properties of the lower mantle heterogeneities. In this study, we aim to develop the first 3D global model of body wave attenuation (Q) in the lower mantle using various S-phase measurements. To maximize the depth and spatial coverage, we incorporate multiple phases (S, SS, SSS, SSSS), core phases (ScS, ScSScS, ScSScSScS), Sdiff and their depth phases (e.g., sS, sScS, sSdiff). We process > 80, 000 seismic data recorded on more than 2000 global seismic stations from earthquakes occurring during 2009-2023. We measure differential anelastic delay times between the observed S phases and the same phases on 3D synthetics using the instantaneous frequency matching method in period range of 100 ~ 10 seconds. These synthetic seismograms are computed in 3D mantle model S40RTS and crust model CRUST1.0 using SPECFEM3D-globe, which can fully account for the effect of 3D heterogeneities, allowing for reliable attenuation measurements. The differential anelastic time delays exhibit abnormally large variations for all S phases, reflecting the complexity of the data potentially brought by the elastic effects. Despite this, the average differential anelastic time delays for all S phases remain consistently negative across all epicentral distances and generally decrease with increasing epicentral distances, suggesting that the Earth is, on average, less attenuating than the PREM model. We further find that the scattering of differential anelastic time delays can be significantly reduced, and abnormal measurements effectively excluded, when the waveform similarity of observed and synthetic phases is high. This is likely because, in such cases, uncertainties arising from factors like source mechanisms and heterogeneity are substantially minimized. We perform a 1D tomographic inversion using high-similarity data. The preliminary 1D attenuation model we obtain is similar to PREM model, with lower attenuation in the lower mantle and the highest attenuation in the upper mantle, roughly corresponding to the depth range of the low velocity zone. However, the Q values in our model are approximately 1.5 times larger than the Q values in PREM model. Next, we will perform 3D tomographic inversion and subsequently integrate this 3D Q model with global 3D shear wave models to jointly invert for the thermo-chemical state of the lower mantle.