VASP Forum

Dear VASP Team,

I hope this message finds you well.

I am currently performing phonon dispersion calculations for the first time using both the density functional perturbation theory (DFPT) and finite displacement (FD) methods at the PBE level. I would like to confirm whether my workflow is appropriate and clarify a few points.
First, I optimized the unit cell structure. Then, I generated supercell displacements (2 × 2 × 1) for my tetragonal system (Ca₄Bi₂O) using Phonopy. After that, I created separate directories for each displaced structure and performed force calculations in parallel. In addition, I am conducting DFPT calculations to compare the results with the FD method.
y main questions are as follows:

Is this workflow correct for comparing DFPT and FD phonon calculations?
If I already have an optimized supercell structure (e.g., for a doped system), can I directly use it for phonon calculations without further enlargement?
What are the main criteria for selecting an appropriate supercell size for phonon calculations?

I would greatly appreciate your guidance on these points.

Thank you very much for your time and support.

Best regards,

Hello,

Welcome to the VASP forum. For phonons, we have lots of great resources available on the Phonons Category Page on the VASP wiki. For your particular case, I would recommend to go through the how-to pages and tutorials.

Here are some general comments that tie into your questions:

• To obtain the phonon dispersion relation (interpolation to finite \(q\)), you always have to converge your calculation (phonon frequencies) with respect to the supercell size. So you keep increasing the cell size until nothing changes any more.

• If you do not wish to use phonopy, you can use LPHON_DISPERSION=True together with the QPOINTS file. That will compute everything inside VASP. With the new checkpoint functionality (CHECKPOINT_FD), it is even possible to split the calculation, like with phonopy.

• DFPT in VASP is only implemented for \(q=0\). This is sometimes a point of confusion for people coming from other codes. However, you can still compute the phonon-dispersion at finite \(q\) using the supercell approach, like with finite differences.

• The required supercell size depends on the system and is influenced by many things (metal/insulator/dielectric, heavy/light atoms, symmetry/cell-shape, ...)