STRANDED DIFFUSION PROFILES IN PARTIALLY RESORBED GARNETS: A KEY LINK BETWEEN NATURE AND EXPERIMENT

Large uncertainties accompany the long extrapolations in temperature, pressure and composition that are required to apply existing garnet diffusion experiments to natural processes. These uncertainties are markedly reduced by incorporating into the diffusional dataset information from natural occurrences with well-constrained thermal histories.

Polymetamorphism in the Llano Uplift of central Texas resulted in crystallization, variable homogenization, and partial resorption of garnet porphyroblasts across a wide range of composition. During resorption spanning ~625-475 °C at 0.3 GPa, partial re-equilibration of garnet rims with their surroundings produced steep gradients in composition that drove intracrystalline diffusion of Mn, Fe, Mg and Ca. The shapes of the resulting concentration profiles depend upon the relative rates of intracrystalline diffusion for each element and the rate of the resorption reaction. Because the rate of the resorption reaction as a function of temperature can be calculated from the thermal history (constrained by isotopic cooling ages) and the total volume of garnet resorbed (easily measured in coronal resorption textures), quantitative estimates of rates of intracrystalline diffusion can be extracted from the stranded diffusion profiles.

In numerical simulations of this coupled resorption-diffusion process, values for the intracrystalline self-diffusion coefficients can be varied to maximize the congruence between calculated and measured concentration profiles. Best-fit profiles yield diffusion rates that agree well with down-temperature extrapolations of experimental determinations for Mn in almandine-spessartine diffusion couples (Chakraborty and Ganguly, 1992, CMP 111:74), but extrapolated experimental values for Fe, Mg, and Ca are significantly lower than those required to produce congruence in the simulations.

The dependence of diffusion rates on the composition of the host garnet is substantial, but complex and poorly constrained. Self-diffusion coefficients are larger for garnets with larger unit-cell dimensions, but until data are available over a wider range of composition, this dependence cannot be precisely quantified.