Fe-Mg diffusion in olivine I:: experimental determination between 700 and 1,200°C as a function of composition, crystal orientation and oxygen fugacity

We have determined Fe-Mg diffusion coefficients in olivines from different sources (Nanga Parbat, Pakistan and San Carlos, Arizona, USA) at atmospheric pressure as a function of composition, oxygen fugacity (10(-5)-10(-12) Pa) and temperature (700-1200 degrees C) using thin films produced by pulsed laser deposition and RBS to analyze the concentration profiles. We have characterized the nano-scale structure and composition of the thin films annealed at various conditions and shown that the nature of the film (e.g. crystallinity, wetting behavior) depends strongly on the annealing conditions. If these variations are not taken into account in the form of boundary conditions for modeling the diffusion profiles, artifacts would result in the diffusion data. The diffusion coefficients obtained from 75 experiments reveal that (i) between fO(2) of 10(-5) and 10(-10) Pa, diffusion along all three principal crystallographic directions in olivine, [100], [010] and [001], are described by a constant activation energy of -200 kJ/mol, precluding any temperature dependence of diffusion anisotropy and change of mechanism of diffusion at temperatures between 950 and 1200 degrees C, (ii) diffusion coefficients increase with oxygen fugacity at fO(2) > 10(-10) Pa, with an fO(2) exponent that lies between 1/4 and 1/7, and (iii) at fO(2) below 10(-10) Pa, and consequently at temperatures below -900 degrees C, diffusion becomes weakly dependent/independent of fO(2), indicating a change of diffusion mechanism. Activation energy of diffusion at these conditions is slightly higher, -220 kJ/mol. The data, including the change of mechanism, are analyzed in terms of point defect chemistry in Part II of this work to derive an equation that allows calculation of diffusivities in olivine over its entire field of stability. Availability of directly measured data at temperatures down to 700 degrees C imply that for the first time diffusion coefficients can be interpolated, rather than extrapolated, for modeling most natural systems.