Need of supercell in phonon calculation using the DFPT method

[payal_saha]

Hello all,
I want to do calculations for phonons using the DFPT method. I have confusion about whether I could use a unit cell, where a=23.9 angstrom, b=4.8 angstrom, and c=5.6 angstrom, for my DFPT phonon calculation, or do I need to consider a supercell for that?

Again, it is mentioned in the vasp wiki page for "LPHON_DISPERSION" that "After the computation of the force constants using finite differences (IBRION=5,6) or density-functional perturbation theory (IBRION=7,8) on a supercell it is possible to compute the phonon dispersion for the equivalent primitive cell determined by VASP by setting LPHON_DISPERSION=.TRUE."

That is why I am confused regarding the consideration of unit cell in the DFPT calculation. It will be of great help if anyone can clear up the confusion.

Thanks in advance.

[pedro_melo]

Dear payal_saha,

As you can read here, with DFPT VASP only computes phonons at q=0. If you want other momenta you must use different supercells where those momenta are folded into q=0.

If you want the full phonon dispersion, I would suggest that you would read on how to compute phonons from finite differences, which should be a faster method, in general.

Kind regards,
Pedro Melo

[pedro_melo]

A small addendum to my previous reply: by default, both DFPT and finite differences only compute phonons at q=0. It is by activating the interpolation procedure with

Code: Select all

lphon_disperion=true

that you compute phonons at other q-points. However, DFPT is currently just worse in many aspects, so finite differences should be the go-to method.

If you wish to skip interpolation, then you must use supercells which are commensurate with the que q-point you want, but this is in most cases too costly.

Best,
Pedro Melo

[payal_saha]

Dear pedro_melo,
Thank you for the reply .

Does that mean if I use the interpolation method (lphon_disperion=true), then I can do the phonon calculations in a unit cell?