The geochemistry of global mantle melts suggests that both mid-ocean ridge basalts (MORB) and ocean island basalts (OIB) sample lithological heterogeneities originating in both the upper and lower mantle, with recycled crustal material accounting for a significant part of this variability. Recently, heavy stable isotopes have been suggested as a new tool to complement existing tracers of mantle heterogeneity and lithology (e.g., major and trace elements, radiogenic isotopes), because mineral- and redox-specific equilibrium stable isotope fractionation effects can link the stable isotope ratios of melts to their source mineralogy and melting degree. In this thesis, I present a unique `bottom-up' approach to understanding how mantle lithology, such as recycled crust (pyroxenite), could be reflected in the stable isotope composition of the erupted melts, and the insights that heavy stable isotope data from basalts could provide into mantle source and process.

Throughout this thesis, I investigate five stable isotope systems (Mg-Ca-Fe-V-Cr) that have shown promise in models or natural samples as tracers of mantle lithology. I develop a quantitative model, combining thermodynamically self-consistent mantle melting and equilibrium isotope fractionation models, to explore the behaviour of the stable isotope ratios of these elements during melting of three mantle lithologies (peridotite, and silica-excess and silica-deficient pyroxenites). I also present new Fe isotope data for Samoan shield and Azores volcanoes, and for a suite of samples from 90 million years of evolution of the Galápagos mantle plume system. These OIB allow me to study the role of recycled mantle components in generating Fe isotope variability in melts, to compare to my mantle melting and isotope fractionation model.

I find that single-stage melting of a MORB-like eclogitic pyroxenite cannot generate the high δ⁵⁷Fe seen in some OIB, notably Pitcairn, the Azores and rejuvenated Samoan lavas. Instead, the generation of high δ⁵⁷Fe melts in OIB requires: (1) processes that make subducted eclogite isotopically heavier than its pristine precursor MORB (e.g., hydrothermal alteration, metamorphism, sediment input); (2) lithospheric processing, such as remobilisation of previously frozen small-degree melts, or a contribution from lithospheric material metasomatised by silicate melts; and/or (3) melting conditions that limit the dilution of melts with high δ⁵⁷Fe by ambient lower δ⁵⁷Fe materials. Therefore, it cannot be assumed that a pyroxenite lithology derived from recycled crustal material is the sole producer of high δ⁵⁷Fe melts in OIB, as has sometimes been assumed in the literature. Instead, the observation of high δ⁵⁷Fe OIB melts cannot be ascribed to a unique source or process. This ambiguity reflects the multitude of processes operating from the generation of recycled lithologies through to their mantle melting: from MORB generation, its low temperature alteration, through mantle heterogeneity development and lithospheric processing, to eruption at ocean islands.

I also find that, given current analytical precision, the five stable isotope systems examined here are not predicted to be sensitive to mantle potential temperature variations through equilibrium isotope fractionation processes, for the melting of peridotite. By contrast, source lithological heterogeneity is predicted to be detectable in some cases in the stable isotope ratios of erupted basalts, although generally only at proportions of > 10% MORB-like pyroxenite in the mantle source, given current analytical precision. However, even when considering analytical uncertainty on natural sample measurements, the range in stable isotope compositions seen across the global MORB and OIB datasets suggests that kinetic isotope fractionation, or processes modifying the isotopic composition of recycled crustal material such that it is distinct from MORB, may be required to explain all the natural data.

Finally, I combine the insight into and modelling of Fe stable isotope behaviour presented throughout the thesis to highlight the potential of heavy stable isotopes to constrain mantle dynamics, in the Galápagos plume system. I show that although the proportion of pyroxenite-derived melt has increased through time as the plume has cooled by 400°C over its lifetime, these results are consistent with a cooling plume containing a small and approximately constant proportion of pyroxenite. This result is consistent with geodynamic models of entrainment of dense material, such as from a lower mantle low velocity superstructure underlying the plume. The small proportion of pyroxenite throughout plume evolution also suggests that geochemical signatures of primordial mantle may be diluted approximately uniformly by recycled components throughout plume evolution and therefore could be identified in early plume localities.

From my combined mantle melting and isotope fractionation model, and comparison to natural datasets, I conclude that the five stable isotope systems considered in this thesis have potential to be powerful tracers of the source lithology of erupted basalts, complementary to other geochemical tools. However, continued improvements in analytical precision in conjunction with experimental and theoretical predictions of isotopic fractionation between mantle minerals and melts are required before these heavy stable isotopes can be unambiguously used to understand source heterogeneity in erupted basalts.