MSD

[emika143]

MSD, seems to be periodic without any notable slope until very high temperatures, for example Silver gets a linear MSD at approx. 12400K, the expected way (I think) would be a small slop in the solid state, bigger slope in the liquid state and even bigger slope in the gas state. (Maybe we can't see any difference between liquid and gas due to the chosen potential)...

[krikru]

We need to check all constants (they're located in definitions.h) and in parameters again, maybe some of them are wrong. We should also pick the lattice constant from a table instead of calculating it from sigma, although this is probably not what causes the problem (the current formula for the lattice constant is very simple and only takes the nearest neighbors into account and should hence give a lattice constant that is slightly too large).

If that doesn't work, maybe the forces or the temperature is calculated incorrectly. We have to check everything that may affect the behavior of the simulation. This problem is an indication that the simulation isn't going alright.

Another issue that Emil detected, that may be related but is probably not, was in the energy plots in a long simulation. While the temperature and the potential energy stayed pretty constant, the total energy increased over time. The most probable is that one of these three parameters are calculated incorrectly or that one of them is written to the wrong file. It could also be that the total energy is so close to zero that the deviations that exist get magnified very much, but this is not so likely.

[krikru]

Another thing I came to think of is that a lot of the initial kinetic energy the system has is required to break the bonds that exist between the atoms when the element is being melt. That is the reason why the temperature drops so much in the beginning of the simulation, but as Emil said, it only drops to like half the temperature or something like that, so it should still be above the melting point. We always have to consider that the temperature will drop considerably from the initial temperature in the start of the simulation, though.