My Community

I have a system with a Ge-Sb-Te slab and water molecules around it in a big vacuum box using vasp 6.3.1 and the Gamma point. I am attempting to run molecular dynamics on the system to calculate some linear response calculations. The calculation starts out fine, but by the second ionic step, there is a warning issued that "The initial velocities result in a center-of-mass drift but there must be no drift. The drift will be removed! ". After the 3rd time step, Vasp seems to be running, but no further output is produced (over 12 hours) implying Vasp is hanging up in some manner. As this implies that the forces are potentially not accurate enough, I tried reducing the time step from 1 fs to 0.5 fs only to see the same warning. I also tried increasing NELMIN from 4 to 6 and this did not stop the error as well. I am not sure of what to try next and am open to suggestions. I have attached the relevant input files (and the OUTCAR/OSZICAR fiels for reference). As this is a simulation that has to run for long periods of time, I would like to come up with input cards that lead to accurate results, but waste a minimal amount of cpu time.

You are setting IBRION=0 and ISIF=3 without setting MDALGO=3.
IBRION=0 without an MDALGO tag selects Nose-Hoover thermostat, which is incompatible with ISIF=3.

Please either specify MDALGO=3 and possibly PMASS, LANGEVIN_GAMMA and LANGEVIN_GAMMA_L or run with ISIF=2.

Unfortunately the incopatible settings are not caught by VASP, so we will have to fix that. Thank you for pointing us to that problem.

I also saw again that you used a possibly too large energy criterion EDIFF = 0.00428.
I would not suggest to use that, I guess it was copied over from another calculation because I also saw it in your other posts.
I can only advise to order the INCAR file and keep it as minimal as possible. That means only set tags that are different from the defaults.
Also it is good to group tags that belong together (MD tags, ML tags, etc.). All this reduces errors in the INCAR file.

Thank you for your help. You are completely correct about the thermostat mistake. I had copied and (partially) edited my machine learning file and left by accident the Langevin thermostat setting in the file. I shall be more careful in the future. The simulation is running fine now.