Reassessment of the garnet-clinopyroxene Fe?Mg exchange thermometer: I. Evaluation of the Pattison and Newton (1989) experiments

This is Part 1 of a two-part study on the garnet(Grt)-clinopyroxene(Cpx) Fe−Mg exchange equilibrium widely used in geothermometry of amphibolites, granulites and eclogites. The experimental data set previously published by Pattison and Newton (PN) on this equilibrium comprises potentially the largest source of data used for extraction of thermodynamic properties for Grt and Cpx and for formulation of a new thermometer expression (Part 2 of this study). However, the experiments suggested unusual Grt and Cpx solution properties and resulted in a geothermometer expression that generally returns low temperatures for granulites. To verify the reliability of the experiments, and to test the experimental technique of Grt buffering used in the PN study, 69 of the 125 PN runs were reanalysed using wavelength dispersive spectroscopic analysis guided by backscattered electron images. Special attention was given to analysis of compositional changes in Grt in addition to Cpx. With minor exceptions, the reanalysis supports the analytical results of the original PN study. The mechanism of equilibration in the runs was a multicomponent solution-precipitation process involving dissolution of metastable Grt and Cpx starting materials and precipitation of newly formed Cpx and Grt, the latter assumed to represent equilibrim compositions. The experiments cannot be regarded either as reversed or as strictly bracketing. Nevertheless, Cpx compositional parameters, including Mg#, Na, Ca and Al, show the same smooth variations with temperature, pressure and composition as found by PN, suggestive of a close approach to equilibrium. Grt in run products retained the moderate heterogeneity of the starting Grt, although there are subtle compositional trends indicating small changes in Grt Mg# consistent with mass balance constraints. These results uphold the essential validity of the technique of Grt buffering used in the experiments. Nevertheless, it is possible that the final Grt compositional may not coincide with the nominal starting Grt compositions, requiring assignment of compositional uncertainties to Grt for purposes of thermodynamic modelling of the data. Cpx analyses also require assignment of non-zero compositional uncertainties.