Joshua Martin Richard Muir , Michael Jollands, Zhang Feiwu, Andrew Walker 

Experimentally silica activity (aSiO2) has been shown to have an effect on Mg diffusion in forsterite but without any obvious mechanism. We calculated the effects of aSiO2 and aluminium content (the main contaminant in the experimental studies), and their co-effect, on Mg diffusion in forsterite, using thermodynamic minimisations of defect formation energies (calculated using Density Functional Theory (DFT)) and a Monte-Carlo diffusion model. These two variables, in isolation, do not appreciably change the defect concentrations of forsterite and thus do not affect the diffusivity of Mg. However, when elevated together, they cause large increases in the Mg vacancy content and thus can increase the Mg diffusivity by 1-6 orders of magnitude depending on temperature, with little pressure dependence. This effect is largely independent of Al2O3 concentration above ~1 wt. ppm, and thus, for all practical purposes, should occur wherever forsterite is in the presence of enstatite. It is also largely is dependent upon configurational entropy and is thus highly sensitive to the chemistry of the crystal. Small amounts of structurally bound hydroxyl groups at low temperatures (1000 K) suppresses this effect in perfect forsterite but it is likely robust in the presence of water when alternative water sinks (such as Ti or Fe) are present or at high temperatures (2000 K). This effect is also robust in the presence of ferrous iron (or other substitutional Mg defects) at all temperatures. Fe2O3 can operate like Al2O3 in this reaction and should enhance its effect. These findings explain the experimentally-observed dependency of diffusion of aSiO2, and elucidate how chemical activity variations in both experiments and natural settings could affect not only the diffusivity of Mg in forsterite, but of olivine-hosted cations in general.