Question on polar optical phonon frequency

@lycheelee54 asks on GitHub:

I am also tangling with the “pop_frequency” tag in the setting.yaml these days. In AMSET example, it happens that both GaAs and Si system only have one LO-TO splitting phonon band, so it seems that LO happens to be the phonon band with the highest phonon frequency. But in SnS system I calculate, there seems to be more than one optic phonon band that are splitting. So should I still use “the frequency of the highest phonon band” as the input in “pop_frequency” tag?

Also I have read the part about using command
amset phonon-frequency
to extract the weighted-sum polar optic phonon frequency. Is this method helps in solving the problems I met in SnS? Anyway ,I ran the command but only get a effective frequency equals 0 THz. I don’t know if this means my DFPT calculation goes wrong?

FYI, I have attached the phonon dispersion of SnS I calculated here and the output after I entered the command amset phonon-frequency in SnS DFPT calculation directory:

Frequency Weight
--------- ------
  -0.01      0.00
  -0.01      0.00
  -0.00      0.00
  -0.00      0.00
  -0.00      0.00
  -0.00      0.00
   0.00      0.06
   0.00      0.94
   0.00      0.00
   0.00      0.00
   1.11      0.00
   1.92      0.00
   2.00      0.00
   2.34      0.00
   2.76      0.00
   2.86      0.00
   5.12      0.00
   5.43      0.00
   5.56      0.00
   6.20      0.00
   6.47      0.00
   6.61      0.00
   6.63      0.00
   8.25      0.00
effective frequency: 0.00 THz

Response: Something appears to have gone terribly wrong with your DFPT calculation. You can see there are 10 Gamma-point phonon modes with a frequency of zero when there should only be 3 (the acoustic modes). You should check your calculation settings to ensure that

  1. You are running DFPT on a very tight relaxed structure. Example calculation settings are described here. The forces on the atoms and lattice should be close to zero.
  2. That you are using tight energy convergence for the DFPT calculation itself. Example VASP input settings are here.
  3. Check your DFPT calculation reached energy convergence for each perturbation.

If you try these things and this does not fix your problem, please upload your INCAR and OUTCAR files and I will try and troubleshoot further.

Hi Alex,

I have followed your suggestion to use a very tight relaxed structure and also a tight energy convergence for the DFPT. I also checked OSZICAR to see if every step is converged. But the problem is not solved.

It seems that this forum doesn’t allow to upload any files except pictures. I have to upload the INCAR/KPOINTS/POSCAR/OSZICAR and OUTCAR of my DFPT calculation to the original post on Github.

I only used VASP DFPT funtion for several times, and this is quite new and confusing to my. I’m really appreciated that you can help me solve this problem.


Hi Zhi,

I think the issue must be your VASP compilation/version. I tried running the DFPT input files you sent using VASP 5.4.4 and the output was fine. I’ve uploaded the OUTCAR and vasprun.xml files here.
The output of amset phonon-frequency was:

Frequency Weight
----------- --------
-0.01 0.00
-0.01 0.00
-0.00 0.00
1.11 0.00
1.44 0.00
1.62 0.00
1.92 0.03
1.99 0.00
2.01 0.00
2.34 0.00
2.76 0.00
2.87 0.11
4.10 0.33
4.74 0.00
5.13 0.21
5.42 0.23
5.57 0.00
5.62 0.00
6.06 0.00
6.21 0.04
6.49 0.00
6.62 0.00
6.64 0.05
8.24 0.00

effective frequency: 4.61 THz