Introduction

Abstract

The lowermost mantle—also known as D\(^{\prime \prime }\)—comprises the few hundred kilometres above the core–mantle boundary, and is known to show significant seismic anisotropy. In this thesis I attempt to use observations of shear wave splitting to constrain flow in this region. In order to accurately measure splitting in D\(^{\prime \prime }\), it is necessary to constrain that present at the top of the mantle. Using the method of source-side shear wave splitting, over 100 novel measurements of anisotropy in the mantle beneath mid-ocean ridges are made across the globe. Splitting mostly increases away from the axis on ridges themselves, and fast directions become increasingly parallel to the spreading direction. This is consistent with the alignment of olivine a-axes being parallel to flow. However, models based purely on lattice preferred orientation of olivine cannot predict the observed difference between splitting in SKS waves by previous authors and the new S wave measurements; one possible explanation is the presence of horizontal pockets of melt at \({\sim }80{-}150\) km depth. These new measurements allow us to vastly improve our understanding of both ridges processes and the lowermost mantle, as they provide the means to use mid-ocean ridge earthquakes to probe regions of D\(^{\prime \prime }\) inaccessible with only deep events. Over 700 measurements of differential S–ScS splitting are then made, using corrections obtained for the upper mantle, beneath North and Central America. Fast orientations along paths from South American earthquakes are consistent with previous observations (showing horizontally-polarised shear waves travel faster than vertically-polarised ones), but measurements made along paths from mid-ocean ridge earthquakes constrain the possible symmetries of anisotropy in the lowermost mantle. They show that radial anisotropy is not an adequate approximation to the style of anisotropy present in D\(^{\prime \prime }\) under the Caribbean. Using elastic constants obtained from ab initio calculations and deformation experiments for a variety of candidate lowermost mantle phases, the shear planes and directions which are compatible with the observations are shown. Assuming horizontal shear, slip on (001) planes in post-perovskite seems the likeliest mechanism to produce the observed splitting. These and earlier measurements of shear wave splitting are then tested against recent models of mantle flow derived from seismic and other geophysical constraints. The flow model is used to derive elastic constants using three different candidate sets of slip systems in post-perovskite. A method to forward-model the shear wave splitting in an arbitrarily anisotropic Earth is developed and used to show that the plasticity model which favours slip on (010) in post-perovskite produces fast orientations best compatible with observations. However, not all observations can be accurately reproduced, suggesting that deformation-induced texturing in post-perovskite may not be the only mechanism producing anisotropy in D\(^{\prime \prime }\). I suggest possible routes towards addressing our current lack of understanding of lowermost mantle processes, some of which are developed in the current work