A grain boundary model of textural evolution of a garnet replacement reaction

Garnet crystals in metapelites from the Goshen Formation, western Massachusetts, have experienced replacement by muscovite + biotite + quartz + plagioclase following decompression from ca. 1.0 to 0.35 GPa. The whole-rock reaction was garnet + muscovite = biotite + plagioclase + quartz; however, the phases replacing garnet (herein called the replacement mantles) always include muscovite as well as biotite, plagioclase, and quartz even though muscovite is a reactant. Numerical models are presented in which reactions only occur on a very local scale with adjacent phases and material in the grain boundaries and the progress of each local reaction is determined by the amount of available chemical affinity. Local reactions change the grain boundary composition, which sets up chemical potential gradients across the grid driving diffusive flux through the grain boundaries. This diffusion changes the grain boundary compositions elsewhere in the rock which changes the local chemical affinity and drives additional reactions in these localities. Thus, the local grain boundary composition drives muscovite dissolution in the rock matrix and precipitation in the replacement mantle surrounding garnet. Models are presented in which the amount of diffusion is varied from very little at one extreme to the other extreme with very long diffusion times such that the grain boundary composition remains homogeneous. The model results reveal that for very short diffusion times, the replacement mantle surrounding garnet is comprised largely of muscovite whereas with very long diffusion times the mantle is mostly biotite. Therefore, the ratio of muscovite to biotite in the replacement mantles may be interpreted to reflect the relative efficacy of grain boundary diffusion. This texture in which muscovite locally grows even though the whole rock reaction consumes muscovite is thus not the result of K-metasomatism but rather the consequence of grain boundary diffusion-limited metamorphic recrystallization.