Oxygen Diffusion in the Perovskite-Dominated Lower Mantle

Perovskite-structured solids are widely known for their tendency to exhibit rapid oxygen diffusion mediated by vacancy hopping. This has important implications for chemical transport in the deep Earth, given that large portions of the lower mantle are composed of perovskite minerals — bridgmanite (MgSiO₃) and davemaoite (CaSiO₃). Here, we present a comparative study of extrinsic oxygen diffusion in both minerals using machine learning molecular dynamics simulations. We show that the extended time scales enabled by machine learning potentials allow oxygen diffusion in these materials to be studied with high accuracy, permitting reliable determination of their Arrhenius parameters, namely the pre-exponential factor and activation enthalpy. We discuss differences in these properties between the two minerals in light of their crystal structures. Finally, we consider the broader implications of our diffusion results for chemical exchange and electrical conductivity across distinct mantle reservoirs.