GOLDSCHMIDT CONFERENCE TOULOUSE 1998

Adam-Gibbs theory and the prediction of glass transition temperatures: the join Na2Si2Os-NaAISi206 M. J. Toplis CRPG-CNRS, Vandoeuvre-l~s-Nancy, F-54501, France The viscosity of silicate melts is a parameter of extreme importance, not only in the Earth sciences, but also in a number of applied disciplines uch as glass, and steel making. Empirical predictive models of melt viscosity proposed in the geological literature (e.g. Bottinga and Weill 1972), have been shown to be reliable at temperatures well above the glass transition, but these models do not take good account of non-Arrhenain behaviour, and extrapolation to temperatures close to the glass transition can lead to significant errors in estimated viscosity. In contrast to these empirical approaches, various attempts have made to quantitatively link viscosity to some structural property of the melt. Of these, the Adam-Gibbs theory (Adam and Gibbs, 1965) which considers viscosity to be related to the configura-tional entropy of the melt, has been shown to be particularly successful in accounting for non-Arrhenian variations of viscosity of molten silicates with temperature (Richet, 1984; Bottinga et al., 1995). Despite this fact, the Adam-Gibbs (A-G) theory has not yet been used in a predictive sense, due to highly non-linear variations of the A-G parameters (Ae, Be, temperature independent constants, Sr the configurational entropy at the glass transition) as a function of composition, even along simple pseudo-binary joins, (e.g. CaSiO3-MgSiO3, Neuville and Richet, 1991)