Chemical Characterisation of Deep Earth Reservoirs

Seismic imaging of the internal Earth has emphasized the inhomogeneity of its mantle. This interface layer of nearly 3000 km surrounding the central metallic core conducts heat released from the latter, which played a significant role in shaping the actual mantle as it currently is. The present state of the deep Earth (mainly the lower mantle and the core) is also a consequence of the physical properties of the material composing them. Large-low shear velocity provinces atop the core-mantle boundary, for example, are composed of a denser material than their surrounding environment but their origin, formation and survival over billions of years to mantle convection, is still largely debated. Here we use chemical partitioning to characterise the current composition of those deep Earth reservoirs and better understand the processes leading to their formation. In the first part, experiments were conducted in the laser-heated diamond anvil cell and the latest advances in microanalysis were used to refine iron partitioning between minerals of the lower mantle. We reconcile long-standing discrepancies in previous partitioning data and show that significant variations of silicate iron content are associated to changing ferric iron incorporation. We also note that such chemical variations might significantly affect the physical properties of the silicate, which might in turn provoke mantle rheology variations. Our results and observations thus provide a mineral physics based argument for the stabilization of deep Earth reservoirs and especially large structures like large low-shear-velocity provinces (LLSVPs). In the second and numerical part of the thesis, previous results on siderophile (iron-loving) element partitioning between metal and silicate were used to constrain some aspects of core formation. We constrain the chemical and physical state of the magma ocean during core formation explaining both the siderophile composition of the mantle and the light element composition of the core. We simulate the Moon-Forming Giant Impact in the last steps of Earth accretion, where we constrain the size of the impactor to be smaller than 15% of Earth'€™s total mass, a roughly Mars-sized object, and further show that melting of the large fractions of the mantle upon impact by the giant impactor is consistent with siderophile composition of the bulk silicate Earth, which can have great implications for the geodynamical evolution of the mantle. In conclusions, given the fact that the mantle may have been almost fully molten, further crystallization may have isolated a layer of liquid primitive material at the base of the mantle possibly explaining the seismic signature of the LLSVPs associated with a strong thermochemical contrast with the surrounding mantle. The layered composition in iron of the actual mantle also provides a chemical argument for a viscously layered mantle, which could have helped sustain LLSVPs for billions of years from efficient convection.