Reconciling dynamic and seismic models of Earthʹs lower mantle: The dominant role of thermal heterogeneity

Two large low-shear-velocity provinces (LLSVPs) in the deep mantle beneath Africa and the Pacific are generally interpreted as hot but chemically dense ‘piles’, which have remained isolated from mantle circulation for several hundred million years. This interpretation largely hinges on four seismic observations: (i) their shear wave velocity anomalies are considered too large for purely thermal structures; (ii) shear wave velocity gradients at their edges are sharp; (iii) their shear to compressional wave-speed anomaly ratios are high; and (iv) their shear and bulk-sound velocity anomalies appear to be anti-correlated. However, using compressible global mantle circulation models driven by 300 Myr of plate motion history and thermodynamic methods for converting from physical to seismic structure, we show that observed lower mantle shear wave velocity anomalies do not require, and are most likely incompatible with, large-scale chemical ‘piles’. A prescribed core-mantle-boundary temperature of 4000 K, which is consistent with current estimates, combined with anelastic seismic sensitivity to temperature, ensures that purely thermal LLSVPs, strongly focussed beneath Africa and the Pacific by subduction history, can reconcile observed shear wave velocity anomalies and gradients. By contrast, shear wave velocity anomalies from models that include dense chemical ‘piles’ at the base of Earthʹs mantle, where ‘piles’ correspond to only 3% of the mantleʹs volume, are substantially stronger than the tomographic model S40RTS, even after accounting for limited tomographic resolution. Our results also suggest that in the presence of post-perovskite, elevated ratios between shear and compressional wave-speed anomalies and the correlation between shear and bulk-sound velocity anomalies cannot be used to discriminate between thermal and compositional heterogeneity at depth: in all calculations, an anti-correlation only occurs within the post-perovskite stability field. Taken together, this implies that although there must be considerable chemical heterogeneity within Earthʹs mantle, large, coherent ‘piles’ are not required to reconcile the seismic observations examined here. Indeed, our results suggest that if chemical heterogeneity is present in these regions, its dynamical and seismic significance is far less than has previously been inferred.