Search

In modern geoscience, mounting evidence in support of large-scale lateral variations in the composition of Earth's deep mantle has fueled the debate on the structure and evolution of mantle convection. Seismological data supports the view that material transport across the mantle transition zone is significant. On the other hand, complete homogenisation of the entire mantle by convective overturn is argued against by geochemical data suggestive of long-lived compositionally distinct domains and seismological evidence for large-scale lateral variations in the composition of the lowermost mantle. It has become difficult to integrate the observational data with models of "layered convection" and "whole mantle convection", which are end-member modelscharacterised by (piece-wise) iso-chemical layers that are well-mixed internally by convective overturn. More successful in reconciling the observational data is a model of thermo-chemical convection, with large-scale lateral variation in the deep mantle imposed by compositionally distinct reservoirs (e.g. Kellogg et al. (1999); Tackley (2002)). Nevertheless, validation of this model is difficult because of uncertainties in the parameters that determine deep mantle convection. In this thesis, the hypothesis is tested that the mantle has evolved as a compositionally heterogeneous entity on the billion-year time-scale. Relevant sub-problems are: Can the existence of compositionally distinct domains in the present-day deep mantle be reconciled with geochemical, gravity, and seismological data? How can a compositionally distinct reservoir survive on a billion year time-scale in the deep mantle? How is the convective evolution affected by material transport at the mantle discontinuity near 660 km depth? Is it possible to discriminate between previously proposed processes forc the formation of large-scale compositionally distinct reservoirs in the deep mantle? To address these research questions, we perform a series of numerical modelling studies of thermo-chemical convection of the Earth's mantle. Synthetic data is then extracted from the modelling results for comparison with available observational data. The different "observables" considered this way are: excess temperatures in and chemical heterogeneity of the upper mantle, long wave-length gravity signals, and the distribution of lateral variations in seismic wave velocity. The conclusions from this thesis are the following. Large-scale compositional heterogeneity that formed early in Earth's evolution can survive for billions of years in a mantle subject to convective overturn. Thermo-chemical models evolved this way are more successful than iso-chemical models in fitting observational data from different research disciplines. The shallow lower mantle is identified as a key region for future testing, because the counterpart of deep mantle compositional heterogeneity residing in this region serves as an indicator for the degree of convective layering at the transition zone --an important parameter to understand the thermal and compositional evolution of the Earth. The preferred mantle convection model that emerges from this thesis is characterised by compositionally distinct domains that occupy a moderate fraction of the mantle volume and which are configured as critically stable piles residing on top of the core-mantle boundary, with a counterpart underneath the upper-lower mantle boundary at which convective upwellings become deflected and are modulated both thermally and chemically, before reaching the upper mantle.