Numerical modeling is frequently used in, amongst other fields, geoscience to simulate a wealth of different geological mechanisms. Numerical models have the advantage that the model can be fully controlled and no unknown material properties or physical laws can influence the simulation.Therefore, numerical models are a superb method to test and enhance mechanisms or laws that have been deduced from observation of, or experiments with, natural rocks. In experiments or observations of natural rocks it is often difficult to recognize and/or estimate the influence of parameters that cannot be directly measured or controlled (such as stress/strain, heating/cooling, surface energies etc.).

In this paper we will present a method to simulate melting processes using numerical modeling. The simulations were performed to test the predicted behavior of solid-liquid systems with given surface energies and grain fabrics. Numerical modeling of partial melt microstructures is still at its beginning, but first results are very promising. A full control on the starting grain fabric and parameters such as the surface energies, and hence, the wetting angle or the mobility of boundaries is possible and all of these parameters can be easily changed. Furthermore, no unknown parameters or mechanisms can influence the simulation so the observed results must be directly related to the used algorithms. A set of simulations with varying surface energies, mobilities of boundaries and grain fabrics were performed that all yielded results that are in good agreement with observation on natural rocks, theoretical predictions and analogue experiments. One of the most important improvements on current theory, however, is that the wetting angle cannot be treated as constant, but rather changes over time to accommodate the gain or loss of melt in equilibrium and disequilibrium shaped melt pockets which influences the rheology of the solid-liquid system.