Nicolas Riel, Boris J.P. Kaus , Eleanor Green, Nicolas Berlie

Prediction of stable mineral equilibria in the Earth’s lithosphere is critical to un-
ravel the tectonomagmatic history of exposed geological sections. While the recent ad-
vances in geodynamic modelling allow us to explore the dynamics of magmatic trans-
fer in solid mediums, there is to date no available thermodynamic package that can eas-
ily be linked and efficiently accounts for the computation of phase equilibrium in mag-
matic systems. Moreover, none of the existing tools fully exploit single point calculation
parallelization which strongly hinders their applicability for direct geodynamic coupling
or for thermodynamic database inversions. Here, we present a new Mineral Assemblage
Gibbs Energy Minimizer (magemin). The package is written as a parallel C library, pro-
vides a direct Julia interface and is callable from any petrological/geodynamic tool. For
a given set of pressure, temperature and bulk-rock composition magemin uses a com-
bination of linear programming, extended Partitioning Gibbs Energy and gradient-based
local minimization to compute the stable mineral assemblage. We apply our new min-
imization package to the igneous thermodynamic dataset of (Holland et al., 2018) and
produce several phase diagrams at supra-solidus conditions. The phase diagrams are then
directly benchmarked against thermocalc and exhibit very good agreement. The high
scalability of magemin on parallel computing facilities opens new horizons e.g., for mod-
elling reactive magma flow, for thermodynamic dataset inversion and for petrological/geophysical
applications.