Questions about velocities in VASP

[gael]

Hi,

I'm not sure this is the right place for that but I'm trying to enhance a CONTCAR- and OUTCAR-file parser.

As a first step, I'd like to get the velocities from the CONTCAR file. Extracting the data is trivial but the units kind of puzzle me.

As the documentation states, in Cartesian coordinates, the units of the velocity is A/fs (vector/timestep in Direct).

1) If the line between positions data block and velocities is empty, does it always imply "Cartesian" as my tests suggest?
2) Does VASP always output positions in the direct coordinate set in CONTCAR?
3) Does VASP always output velocities in the Cartesian coordinate set in CONTCAR?


As a second step, as VASP does not save velocities at each ionic step, I tried to calculate the corresponding velocities from XDATCAR (or OUTCAR). The problem is that the velocity in CONTCAR is given with 8 significative figures while recalculated velocities from positions gives velocities with 2-3 significative figures.

4) Is there a way to get more accurate velocity data?


As a last step, there seems to be no documentation about the predictor-corrector coordinates.

5) Would you mind telling me if such a documentation exists? And where?

My aim here is to be able to rebuild a CONTCAR/POSCAR file from any step of a MD calculation to be able to "branch" the calculations. I could launch the calculation 1 step at a time copying CONTCAR at each step, but this is not what I would call "convenient", nor "clean".

Thanks in advance,
Gael
<span class='smallblacktext'>[ Edited ]</span>

[admin]

1. The line before velocities is always empty.
2. CONTCAR gives direct coordinates.
3. Velocities are in cartesian coordinates.
4. Using your own code you can read velocities from CONTCAR.
5. There is no documentation.