VASP 6 centre of mass

[victor_robinson]

Dear vasp,

For vasp 6 I notice that the XDATCAR's now have unwrapped fractional coordinates. In vasp 5(-) were they always wrapped?

vasp-MD-COM.png

I also notice the COM for a system (solid bcc here) is moving during MD as picture above. Here I was using MLMD in training mode in vasp 6.3.2. Has this changed since vasp 5? Was the COM previously fixed? Where can I find information on this or changes on this (if there were any)?

Regards, Victor

[victor_robinson]

Actually I wonder if this is limited to the Langevin thermostat.

Why is there no "total drift" reported for MLMD steps? i.e.
"total drift: -0.000000 0.000000 0.000000"

[ferenc_karsai]

The coordinates in VASP can be outside of the box on input, but internally they are placed inside.

The center of mass you observe should only appear for the Langevin thermostat. But it doesn't neccessarily always appear. It's magnitude is fully system dependent and very often it doesn't matter much.

The Nose-Hoover has no drift, but still I would advise in most cases to use the Langevin thermostat, since it's much more ergodic and it's the only thermostat implemented in the NpT ensemble.

[victor_robinson]

Hi, thanks for your reply. I think it was langevin combined with bad forces leading to COM drift.

For XDATCAR coordinates:
I am still not sure as to whether the XDATCAR fractional coordinates are absolute or wrapped in either vasp 5 or 6? Vasp 6 is definitely seeming to be absolute (numbers greater than 1.0).

Thanks, Victor