Tracing the evolution of a fertile REE granite by modelling amphibole-melt partitioning, the Strange Lake story

Abstract

In principle, potentially economic (fertile) granites can be distinguished from uneconomic (barren) granites based on the composition of their magmas. Ideally, this composition would be evaluated by analysing melt inclusions, but such inclusions are rarely preserved and difficult to analyse. Information on the composition of magmas can also be obtained from the composition of the minerals crystallising from these magmas, if the mineral-melt partition coefficients are known. In some cases, where these partition coefficients are not known, they have been calculated from lattice strain theory. Here we report arfvedsonite-melt partition coefficients for the rare earth elements (REE) calculated from the results of analyses of arfvedsonite, bulk rock and melt inclusions in granites and pegmatites in the Strange Lake peralkaline rare metal pluton. These coefficients were fitted to the lattice strain model from which values of the ideal partition coefficient (D0), the ideal radius (r0) and the elastic response (EM) were determined. The light REE were shown to partition preferentially into the M4 site and the heavy REE into the M2 site, requiring that partition coefficients be calculated for each site. The partition coefficients for both sites were shown to decrease systematically with increasing evolution of the magma, i.e., from the early hypersolvus granite to the late pegmatites. In the case of the M4 site, D0, r0 and EM vary linearly with the Ca content of the arfvedsonite. By contrast, these parameters for the M2 site vary linearly with temperature. The two relationships form the basis of a predictive model, in which the arfvedsonite-melt REE partition coefficients for other peralkaline granites can be estimated, provided that the Ca content of the amphibole and its crystallisation temperature are known. This model was applied to a peralkaline granitic pegmatite from the Amis complex (Namibia) for which data on the composition of the amphibole and corresponding magma (melt inclusions) have been reported. The model successfully predicts the concentrations of the various REE in this magma, thereby providing confidence that it can be used to discriminate between granites that are potentially fertile and those that are barren in respect to the REE.