For RPA and the related Exact calculations, vasp examples (SiC) suggests a change of KPOINTS: a change from Monkhorst-Pack to Gamma k-sampling.
Are there any issues in the results, when Monkhorst-Pack is used for optical or RPA?
In addition, Monkhorst-Pack yields issues in the results for standard DOS calculations?
KPOINTS for RPA &DOS
Moderators: Global Moderator, Moderator
-
henrique_miranda
- Global Moderator
- Posts: 506
- Joined: Mon Nov 04, 2019 12:41 pm
- Contact:
Re: KPOINTS for RPA &DOS
Hi,
If you are refering to this tutorial:
wiki/index.php/Dielectric_properties_of_SiC
There can be two distinct issues:
1. If you are doing an RPA calculation, Monkhorst-Pack might not be supported by the code.
If you are using the low scaling version ALGO=RPAR there is the contraint that if k1 and k2 are in the grid, then k1-k2=q must be in the grid too. This is guaranteed only with gamma-centered grid.
The same constraint might not apply for ALGO=RPA
2. If you are computing dielectric properties, by using a shifted grid you might miss high-symmetry points where sometimes the band-gap is located and get an incorrect description of the absorption edge for underconverged k-point grids.
This depends also on whether the transitions at these high-symmetry points are allowed by the selection rules.
The difference between Gamma centered or Monkhorst-Pack should disapear at sufficiently converged k-point meshes but the path to convergence is different.
The same applies for standard DOS calculations.
I recommend that you do your own testing.
Kind regards,
Henrique Miranda
If you are refering to this tutorial:
wiki/index.php/Dielectric_properties_of_SiC
There can be two distinct issues:
1. If you are doing an RPA calculation, Monkhorst-Pack might not be supported by the code.
If you are using the low scaling version ALGO=RPAR there is the contraint that if k1 and k2 are in the grid, then k1-k2=q must be in the grid too. This is guaranteed only with gamma-centered grid.
The same constraint might not apply for ALGO=RPA
2. If you are computing dielectric properties, by using a shifted grid you might miss high-symmetry points where sometimes the band-gap is located and get an incorrect description of the absorption edge for underconverged k-point grids.
This depends also on whether the transitions at these high-symmetry points are allowed by the selection rules.
The difference between Gamma centered or Monkhorst-Pack should disapear at sufficiently converged k-point meshes but the path to convergence is different.
The same applies for standard DOS calculations.
I recommend that you do your own testing.
Kind regards,
Henrique Miranda