OrderN
Improve NPT stability
The NPT dynamics are very sensitive to settings and need improvement to make them more stable
I think the stability is more an artifact of the Parinello-Rahman formalism, and the best way to improve it is to carefully choose the barostat and thermostat relaxation times, MD.tauP and MD.tauT respectively. Increasing the barostat mass by increasing the coupling time period will damp out the "ringing" oscillations in the pressure, but increasing it too much will lead to an overdamped situation where the system takes a very long time to reach the target pressure. This is difficult in practice, since there is a strong coupling between the thermostat and barostat, and the thermostat time period which is relatively unimportant in the NVT ensemble has a big effect on the NPT stability, and it can take a lot of effort to find the right combination, since it also seems to be somewhat dependent on system size.
I implemented an ad hoc drag factor MD.PDrag (borrowed from LAMMPS) which will reduce the velocity of the barostat by a small amount (around 5%) every MD step, which technically breaks the conservation, but in practice significantly stabilises the system. This is detailed in the manual in the MD section, but perhaps could be more obvious.
