Sillimanite mullitization: ATEM investigation and point defect model

A polycrystal sample of Fe-bearing prismatic sillimanite was annealed exper imentally at T = 1675 degrees C and P = 20 kbar for 12 min. ATEM investigat ion of the sample reveals that the starting material partly transformed int o mullite during the annealing, and that this process was assisted by parti al melting. The exsolved partial melt (now a glass), observable at triple e dges and under the form small (< 100 nm) precipitates that exsolved within the primary sillimanite grains, consists of about 80 wt.% SiO2, 12 wt.% Al2 O3, 5 wt.% Fe2O3 with some amount of K2O Composition profiles throughout th e sillimanite residual matrix reveal that sillimanite mullitization was sti ll going on at the end of the run. The equilibrium sillimanite composition corresponding to the annealing conditions, and measured at the contact with the exsolved SiO2-rich melt, corresponds to the formula (Al, Fe)(4.33)Si1. 67O9.83. From the equilibrium melt and matrix compositions and the size of the largest isolated precipitates, a rough estimate of the Si-Al interdiffu sion coefficient in sillimanite (D) (associated to the mullitization at run conditions) is D approximate to 2 x 10(-17) m(2)/s. From the ATEM results, a point defect model is proposed to explain sillimanite mullitization. In the framework of this model, the majority point defects are the oxygen vaca ncies and the aluminum cations substituted for silicon (on T* sites). No di stinction is made between sillimanite and mullite which are considered as o ne unique phase with different nonstoichiometry (i.e. different oxygen vaca ncy concentrations). The model, which accounts for the compositions of nons toichiometric sillimanite measured at various temperatures and pressures, q uantifies explicitly (with respect to P and T) the solid-solution compositi on of the sillimanite-mullite joint.