NSCF (non-self-consistent field) calculations are done on a dense mesh. I did it on a unit cell. Then since I have to try using a supercell for alloying, I am doing calculations to see the results of band gaps, and band structures.

But I got an increased band gap when using a supercell. Please help me in this, and also let me know if it's because of something wrong I am doing here.

  • $\begingroup$ Provided both calculations are sufficiently converged in all computational parameters you should get the same band gap. It is the same substance after all. (caveat, I don't know what you mean by a "nscf calculation", I'm assuming you mean a standard SCF - please avoid code specific terms if at all possible) $\endgroup$
    – Ian Bush
    Jan 2, 2022 at 9:48
  • $\begingroup$ @IanBush It's non self consistent field calculations with tetrahedra smearing to get highest occupied and lowest occupied energy levels. I will edit it in the question. ok thank you, the calculations are converged. So there must be something wrong, I should do it again. $\endgroup$ Jan 2, 2022 at 10:04
  • $\begingroup$ @epsilon02fft Try comparing SCF calculation of both instead, or better their band structures. NSCF calculations are kind of like extrapolating the wavefunction from SCF to a denser mesh. Therefore, I don't think changing any inputs between SCF and its subsequent NSCF will give you any meaningful data. $\endgroup$ Jan 3, 2022 at 11:14
  • $\begingroup$ Your non-self-consistent calculation has to have a sufficiently dense grid. You have to check this by using a convergence method and systematically increasing the sampling (kpoints or reciprocal sampling if you are using a plane-wave code). Then when you are having the supercell are you running the scf again? In a supercell depending on the ordering of the alloying your lattice parameters might be different which can result in a different bandgap. $\endgroup$ Jan 5, 2022 at 4:14
  • $\begingroup$ @TaraMishra Yes. I will perform it again. Then I will update here. $\endgroup$ Jan 8, 2022 at 11:46

1 Answer 1


If the band gap increased when you moved to the supercell, its likely that your kpoint grid doesn't hit the ideal valence max/conduction minima. This problem could occur in the SCF step, even if you use the same exact path for the primitive and supercell for the NSCF step. This is especially likely if the primitive is gamma centered and the supercell is not, or vice versa.

You should also check if you get a distortion in the supercell, this could also change the valence band position or conductance band position and increase or decrease the bandgap. If you did not do a geometry optimization after making the supercell, this is unlikely to be the case though.

  • $\begingroup$ I will do the calculation for supercell again for the same optimization of unit cell and use denser grid. Should I do optimization again for supercell? what parameters do I consider checking on both unit cell and supercell to make sure the calculations are going fine? $\endgroup$ Jan 8, 2022 at 11:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .