band energies precision
Posted: Tue Mar 27, 2018 6:33 pm
Dear VASP users and developpers,
After having converged the electronic structure for some systems, I normally do a NSCF calculation (ICHARG = 11) to get the band structure. We are interested in studiying some bands close to the Fermi level.
The problem is that, sometimes, only by changing the k-mesh for the band structure (ex: G-X to G-X/3 to get a denser mesh), we get different band energies, with energy differences of the order of some 10e-2 eV. All other parameters are kept fixed. These differences are important as they can make some bands to appear partially occupied or fully unoccupied.
Normally, for fixed density and potential, the band energies are calculated (and can be calculated) independently for each k-point, and then the result should not depend on the k-mesh (if we compare the energies at the G point for example), isn't it?
We have mostly encountered this problem for huge slab calculations.
Do you have any idea about the origin of this problem and any suggestion which could help to cure it?
May it be due to a lack of precision by defining the charge density in the SCF run? Is there some critical parameters that should be considered?
Best regards,
Rémi
After having converged the electronic structure for some systems, I normally do a NSCF calculation (ICHARG = 11) to get the band structure. We are interested in studiying some bands close to the Fermi level.
The problem is that, sometimes, only by changing the k-mesh for the band structure (ex: G-X to G-X/3 to get a denser mesh), we get different band energies, with energy differences of the order of some 10e-2 eV. All other parameters are kept fixed. These differences are important as they can make some bands to appear partially occupied or fully unoccupied.
Normally, for fixed density and potential, the band energies are calculated (and can be calculated) independently for each k-point, and then the result should not depend on the k-mesh (if we compare the energies at the G point for example), isn't it?
We have mostly encountered this problem for huge slab calculations.
Do you have any idea about the origin of this problem and any suggestion which could help to cure it?
May it be due to a lack of precision by defining the charge density in the SCF run? Is there some critical parameters that should be considered?
Best regards,
Rémi