Doubt with phonon dispersion

ASE
1

Hi everyone,

I was trying to get the phonon dispersion of BaO following Phonon calculations — ASE documentation. However, when I change the supercell size in the

Phonons(atoms, EMT(), supercell=(N, N, N), delta=0.05)

my GM point frequency was changing. I had to go to N=12 for stable acoustic branches. So is phonon dispersion a function of supercell size in ASE?

Thanks
Dominic

2

Screenshot from 2025-10-19 19-35-20
Screenshot from 2025-10-19 19-35-201046×240 55.7 KB

This is how the phonon dispersion is varying in my case. With changing the supercell size, the acoustic branches vary differently. Is this behavior expected?

1 Like
3

You can only resolve phonons with wavevectors that are commensurate to the supercell size (i.e, proportional to 1/k), see e.g., this work. Phonons near the gamma point are the hardest to resolve accurately because the wavevector at Gamma is (0,0,0), and thus a huge supercell is needed

1 Like
4

Any clue so far?

5

Hi Aaron,

Thanks for the reply. So, the dispersion gets more accurate with the supercell size which is a result of having the finite q frequencies at 1/N values (for N - supercell size).

6

Yep exactly! Typically you’ll need to do a convergence test of the supercell size against the lowest frequency modes that you want to study to figure out what a reasonable supercell size is

2 Likes
7

If the frequency at Gamma is changing with supercell size, this is anomalous behaviour and suggests a bug or some very noisy input data. But the improvement of the bands near Gamma in those plots does appear to be a typical supercell size convergence effect.

1 Like