The simulations of melt processes on the grain scale have shown, that the most important parameter in the simulations is the relative surface energy and hence the wetting angle. If the surface energies in the simulations were chosen so that the wetting angle is high, the melt pockets tend to get straight or even concave boundaries before they quickly disappear, their melt being redistributed into large melt pockets. In simulation3 (wetting angle 120°) the large melt pockets are not triangular but resemble a distorted rectangle. The same holds true for the melt pockets in simulation 2 (wetting angle 60°), however, redistribution of melt into the larger melt pockets is much slower and the triangular shapes of melt pockets seems to be stable for a much longer period. Nevertheless, towards the end of the simulation all triangular melt pockets disappear and the melt is redistributed into the larger, rectangular shaped melt pockets towards the end of the simulation.
The triangular shaped melt pockets in simulation1a quickly adjust their ratio of circumference to area until they are stable (Figure 3). This includes a constant adjustment of the wetting angle at any given triple point because changes of the surrounding grains have to be accommodated. This change of the wetting angle also influences the mean curvature of boundaries between two triple points. The changes of the wetting angle can be quite large in case of melt being redistributed into D-shaped melt pockets. Only when the D-shaped melt pocket splits into two E-shaped melt pockets (in this case triangular) and melt is redistributed, the wetting angle is readjusted to approximately 10°. The same holds true for wetting angles in simulations 2 and 3. Therefore, the wetting angle (and hence the mean curvature of the melt pocket boundaries) during the simulations is not static but interactively adjusts to the loss or gain of melt and/or to the shapes of the surrounding grains.