Per-species COM velocities for multicomponent Green-Kubo from VASP AIMD

[tanmoy_paul]

I’m running NVE AIMD at 300 K on a ternary system(Li3YCl6) with on-the-fly MLFF and already have per-step heat flux from ML_HEAT. For multicomponent cepstral GK I also need element-wise center-of-mass velocities(mass fluxes or mass current) at each step. Is there a VASP-native way to output species-resolved COM velocities, or should I post-process vasprun.xml/OUTCAR by grouping atoms by element and mass-averaging per step (with optional drift removal)? Please also confirm that the per-atom velocities there are in Å/ps and that the time spacing is set by POTIM in fs. Thanks!

[manuel_engel1]

Hello,

Thanks for posting on the VASP forum. We do not offer a way to compute these species-resolved COM velocities directly in VASP, so you will have to do the post processing yourself based on the output we provide. However, the ion velocities are not written by default to the output for each ionic step. Since VASP 6.4., it possible to set VELOCITY=True in the INCAR file (not currently documented on the Wiki, but we will update it soon), which will write the velocities to the vaspout.h5. This should be the easiest way to do post processing. You can find the important datasets at these HDF5 paths:

Code: Select all

vaspout.h5/intermediate/ion_dynamics/position_ions
vaspout.h5/intermediate/ion_dynamics/ion_velocities

The positions and velocities of the ions are also available via py4vasp:

Code: Select all

from py4vasp import Calculation
calc = Calculation.from_path(".")
positions = calc.structure[:].positions()
velocities = calc.velocity[:].read()["velocities"]

Regarding the units, I can confirm that the units of POTIM are indeed fs. However, the units for the velocities in the vaspout.h5 are in Å/fs, not Å/ps.