Normal-mode density tomography of the Earth's mantle and outermost core

Abstract

Advancing our understanding of the dynamics and evolution of Earth's deep interior requires the interpretation of tomographic models of the current seismic properties of the mantle and core in terms of thermochemical structure. The larger the number of seismic parameters that may be constrained together, the more feasible such interpretations become. In this thesis we use measurements of Earth's free oscillations, or normal modes, to simultaneously infer 3-D variations in global-scale seismic velocities and density, as well as topography on global mantle discontinuities. To this end, we use an extensive dataset of normal-mode splitting-function measurements. Several regions are considered in this thesis: the lowermost mantle, the mantle transition zone and the outermost core. In addition, we explore the feasibility of a future splitting-function inversion for global-scale mantle anisotropy.

Regarding the lowermost mantle, we show that our used splitting-function dataset improves constraints on the density of the Large Low Shear Velocity Provinces (LLSVPs), providing important insights into their origin. We find that our inferred LLSVP density structure, consisting of a light top overlying a partially dense base, reconciles conflicting models of dense and light LLSVPs. Our combined tomographic models of variations in shear-wave velocity, bulk-sound velocity and density indicate that the LLSVPs must be compositionally distinct. A subsequent thermochemical inversion of our tomographic models shows that our inferred LLSVPs are consistent with an enrichment in SiO2 and FeO, which may be indicative of the presence of primordial material or recycled oceanic crust inside the LLSVPs.

Our following region of interest is the mantle transition zone. We show that, with our used splitting-function dataset, it is now possible to constrain topography variations on both global discontinuities using normal-mode data alone. We interpret the correlations between these topography models and the simultaneously inferred velocity and density models and find that the 660-km discontinuity may be largely explained by temperature variations, while the 410-km discontinuity requires additional variations in composition.

Next, we explore the feasibility of constraining whole-mantle anisotropy in a future inversion of splitting-function measurements. For individual normal modes and cross-coupled normal modes, we perform a principal component analysis on the sensitivity kernels to density and all 21 elastic parameters describing seismic anisotropy. We find that the sensitivity to mantle anisotropy of each normal mode is captured by only three principal components. This is a promising result, as a reduction in the number of model parameters increases the feasibility of a splitting-function inversion for whole-mantle anisotropy.

Finally, we consider the region directly below the mantle: the outermost core.Through forward modelling of normal-mode center frequencies and SmKS differential travel times we first show that the generally observed low outermost-core Vp is a robust result, even when potential trade-offs with lowermost mantle structure are taken into account. We then perform a center-frequency inversion that allows for velocity and density variations throughout the entire mantle and core, which results in a tomographic model with an outermost core that is easier to reconcile with body-wave constraints, but remains difficult to reconcile with scenarios of outermost-core stratification.