Conclusions

The presented numerical modeling system Elle is capable of simulating melt processes at the grain scale. The results of the simulations comply with theoretical predictions on the behavior of melt in a (poly)crystalline aggregate. The most important parameters of the simulations are, apart from the starting grain fabric, the relative surface energies and hence the wetting angle. Different relative surface energies greatly influence the shape of the melt pockets during the simulation and, in contrast to the theoretical predictions, neither the wetting angle nor the mean curvature of a solid-liquid interface is static when observed over short periods of time. The change of the total energy of all surfaces and the accompanying redistribution of melt influences the wetting angle and the mean curvature at ss-sl interfaces. Only over geological time periods these parameters can be viewed as static.

If the starting grain fabric is chosen so that the grains approximate a sphere with melt located in the gussets between three grains, the surface energies are minimized and no change of the grain fabric occurs. In this case, small distortions of the shape of the grains leads to a small adjustment of the grain fabric. Whether this behavior influences the rheology of the rock by counteracting imposed changes due to deformation mechanisms is not yet clear and has to be investigated with further simulations.

The reason why the theoretical considerations fail to predict the disequilibrium features observed in the simulations as well as in the analogue experiments is the flawed assumption of a static equilibrium within a polycrystalline system. A solid–liquid system consisting of more than one crystal cannot be in real equilibrium and will always strive towards a lower total energy by surface energy driven grain coarsening, i.e. melt assisted recrystallization or Ostwald ripening at low and high melt fraction respectively. The observed formation and destruction of disequilibrium features in the analogue experiments and the simulations therefore show the complicated interaction of grain coarsening, which creates the disequilibrium features, and liquid dispersion, which counteracts disequilibrium liquid distribution.