AbstractNumerical models of diffusion‐controlled nucleation and growth of garnet crystals, which successfully replicate diverse textures in 13 porphyroblastic rocks, yield quantitative estimates of the magnitudes of departures from equilibrium during crystallization. These estimates are derived from differences in chemical potential between subvolumes containing stable product assemblages and those containing persistent but metastable reactant assemblages. The magnitude of disequilibrium is evaluated in terms of the thermal overstepping, which is commonly referenced to the garnet‐in isograd; the reaction affinity in the intergranular fluid at the site and time of each nucleation event, and on average throughout the rock, and the ‘latent energy of reaction’ per unit volume, a measure of the average unreacted capacity of the bulk rock, which describes its overall metastability. Across all of the models, the first crystals nucleate after 5–67 °C of thermal overstepping (correspondingly, 0.7–5.8 kJ mol−1 of 12‐oxygen garnet); the maximum reaction affinity averaged across the intergranular fluid is between 4.7 and 16.0 kJ mol−1 of 12‐oxygen garnet; and the maximum latent energy of reaction ranges from 7.3 to 51.7 J cm−3. These results demonstrate that impediments to crystallization significantly delay nucleation and retard reaction, with the consequence that nucleation of new crystals extends throughout nearly the entire crystallization interval. This potential for protracted reaction during prograde metamorphism, with reactions continuing to temperatures and pressures well beyond equilibrium conditions, suggests the likelihood of overstepping of multiple – possibly competing – reactions that can progress simultaneously. Isograds and ranges of stability for metamorphic assemblages along a metamorphic field gradient may therefore be significantly offset from the positions predicted from calculations based on equilibrium assumptions, which poses a substantial challenge to accurate interpretations of metamorphic conditions and processes.
Read full abstract