Quantitative assessment of tomographic proxies for lowermost mantle composition and mineralogy

Large low velocity provinces (LLVPs) dominate Earth’s lowermost mantle, but their detailed thermochemical nature remains a topic of discussion. In particular, it is unclear to what extent the bridgmanite to post-perovskite phase transition is able to explain their seismic velocity characteristics. Robust constraints on the origin of these seismic structures would shed light on large-scale mantle dynamics and Earth’s thermal and chemical evolution. Here, we examine the combined effects of temperature, chemical heterogeneity and phase transitions on lowermost mantle tomographic signatures. To investigate this, we calculate synthetic seismic velocities expected from a range of scenarios for the stability of post-perovskite combined with models of different lowermost mantle temperatures and compositions using recent thermodynamic data. These are filtered to account for limited tomographic resolution, allowing for quantitative comparisons between our synthetic seismic velocities and a recent Backus-Gilbert based tomography model. Crucially, this model provides robust ratios and correlations of velocity anomalies derived from nearly identical V p and V s resolution, and includes uncertainty quantification that accounts for both data and theoretical errors. Given the tomographic uncertainties and limited resolution, our comparisons focus on globally and depth averaged seismic characteristics, which capture the effects of lateral compositional and mineralogical variability. By rejecting synthetic models that do not fit within tomographic uncertainties, we quantitatively eliminate the following: (i) models containing LLVPs with an iron-rich primordial composition, as these generate anomalously high root mean square seismic velocity anomalies; and (ii) models without post-perovskite in the lowermost mantle, as these cannot explain observations of elevated ratios of shear-wave to compressional-wave velocity ( R s / p ) and a negative correlation between variations in shear-wave and bulk sound velocity ( r s − c ). Additionally, we demonstrate that observations of R s / p and r s − c in the lowermost mantle cannot be explained by thermochemical LLVPs alone, but require bridgmanite and post-perovskite to co-occur at depth in the mantle. As such, we demonstrate that globally averaged seismic velocity characteristics can distinguish between composition and mineralogy in the lowermost mantle. • We create synthetic seismic models of various LLVP compositions and pPv stability • Synthetic seismic models that fall outside tomographic uncertainties are rejected • Seismic ratios and correlations ( R s / p , r s − c ) are evaluated with equal V s - V p resolution • d ln V p and d ln V s amplitudes cannot explain LLVPs composed of only Fe-rich material. • Lowermost mantle R s / p and r s − c require a co-occurrence of Bdm and pPv.