Phonon Mode Atomic Displacements

[michael_walkup]

I'm trying to fully understand what VASP provides when performing phonon dispersion calculations (IBRION=6, finite differences w/ symmetry). If I understand correctly, the eigenvectors in the OUTCAR file are the Cartesian displacements of the atoms at the zero-point energy with a relaxed structure, in my case. The wiki only states they're "normalized eigenvectors of the eigenmodes in Cartesian coordinates", but could I simply mass-unweight them (i.e. divide by the square root of the mass in amu) to get the displacement vectors of each atom in Angstroms? I'm ultimately trying to get the displacements of each atom at the ZPE.

[ferenc_karsai]

To get the positions of the atoms at ZPE you could run the Monte-Carlo sampling methods described here. These methods use IBRION=6 and create POSCARs with random displacements among the eigenvectors of the harmonic oscillator. If you sample them large enough and average over the structures you get the mean displaced structure. Alternatively you can use the ZG(one-shot method) which gives you approximately the same.

[michael_walkup]

If 0K isn't the default temperature for a finite-differences calculation, then what is it normally producing if an MC method isn't utilized? The unit cell I'm using has 324 atoms in it, so I'm concerned using MC that needs a large supercell would be quite computationally expensive. I'm trying to obtain the classical turning points (which I believe are the amplitudes of the atomic oscillations at the lowest energy level) to conduct further coupling constant calculations.

[ferenc_karsai]

0K is the default for finite diff.
The temperature in MC enters via probability distributions (e.g. Bose-Einstein). You can choose the temperature in the MC too.

Your original question was to run IBRION=6 calculations, that is the computatively expensive part. MC calculations use the same and the writing of the POSCARs is just a very small computational overhead. So if 324 atoms is too big for you, then your original idea has the same problem.