Implementation of trace element behaviour in the numerical modelling of magmatic processes

Trace element analysis of rocks and minerals can provide valuable insight in the petrogenesis of the crust and therefore may offer fundamental constraints into geodynamical processes. But crustal building involves a multitude of magmatic/metamorphic processes, and the trace element behaviour during each of these processes is complex, which makes the interpretation of trace element composition of the final crustal product non-unique. Therefore, a thorough investigation of trace element systematics between mineral and melt is required to trace the implication of important stable phases in crustal building processes. This can be given using forward numerical modelling and trace element modelling. However, while numerical models simulating thermodynamic reactions are available, a numerical tool for the prediction of trace element behaviour is missing. Such tool would be of key interest to capture the subtle exchanges during magmatic reactions. Therefore, this thesis presents the building of a novel numerical tool able to predict the variation in trace element partition coefficients between 6 mafic minerals and melt (e.g., garnet, olivine, pyroxene, plagioclase and amphibole). To develop such tool, the state-of-the-art predictive models available in the literature are compiled. These predictive models are of two different kinds. First, the lattice strain model (LSM) is based on the energy exchange of trace elements between mineral and melt. Second, multiple regressions analyses are based on the most accurate fit against various chemical/physical parameters and natural data. In addition, this thesis builds predictive models by applying statistical regressions on large dataset of trace element partition coefficients in the case no models are available in the literature. While the stand-alone version of this tool reproduces well the experimental data, coupling with the numerical code Perple_X, which calculates stable mineral assemblages using Gibbs free energy minimization, allows calculation of a full P-T catalogue of trace element behaviour in the partial melting of an amphibolite. The catalogue gives insights in the assimilation processes in a lower crust in arc settings. This coupling offers a new perspective for future, more complex coupling with numerical codes of melt genesis, transfer and emplacement.