A GALES--MELTS framework for thermodynamically constrained forward modelling of crystallisation kinetics in magmas

Abstract Crystallisation controls magma differentiation, rheology and texture, yet its evolution depends on the coupled effects of thermodynamic disequilibrium, nucleation, crystal growth and melt depletion. We present a forward crystallisation framework coupling GALES with alphaMELTS, in which phase affinity and incipient crystal composition are used directly to calculate homogeneous and heterogeneous nucleation and crystal growth. Nucleation follows classical nucleation theory with stochastic crystal births and a kinetic attachment prefactor governed by mixed diffusion- and interface-controlled processes; growth uses an analogous mixed kinetic formulation. Crystallised material is removed explicitly from the liquid, so that residual-melt evolution continuously modifies the thermodynamic driving force. The framework is tested against two contrasting experiments. For the initially crystal-free Iezzi case (Iezzi et al., 2011), homogeneous nucleation precedes heterogeneous nucleation, and the model reproduces the observed crystallinity, plagioclase composition and main residual-melt differentiation trend. In the slowly cooled Billon G2-2 experiment (Billon et al., 2025) inherited plagioclase provides the substrate for heterogeneous nucleation and no homogeneous nucleation occurs along the modelled cooling path. Retaining the kinetic parameters constrained by Iezzi and recalibrating only the effective heterogeneous nucleation parameters reproduces plagioclase abundance, crystal number density, nucleation rate and residual-glass composition, although the modelled plagioclase remains more sodic than the measured new-growth rims. In both benchmarks, modelled crystal-size distributions remain narrower than observed. Rather than constituting only a model limitation, this mismatch provides an additional constraint on crystal-population evolution, showing that bulk crystallinity, number density and melt chemistry do not fully determine the textural record. The framework therefore offers a physically based platform for identifying and testing additional processes governing crystal growth and post-nucleation evolution.