I calculated the bandgap for one bulk system in the unit cell and the supercell of 2x2x1 and 3x1x1 and 2x2x2. Although 3x1x1 gives the same bandgap as I got from the unit cell (1.80 eV) but 2x2x1 and 2x2x2 are giving 1.75 eV. I think the bandgap should remain the same in the supercell. I tried with different K points set in 2x2x2 and 2x2x1 supercell but still did 0.05 eV difference I am getting. Is this much difference is acceptable.
Any suggestion to improve this band gap in 2x2x2 and 2x2x1 supercell will be very helpful.
Hi,
could you please mention how did you calculate the bandgap! By Bands or DOS calculations. if by bands, then check if the k-path in Brillouin zone is not changed. Also, for the system the k-points should be optimized. If all done. Please check in literature, it also depends upon what atoms are there in the unit cell.
I hope this would be helpful.
Khagesh
Thank you so much for your response.
Bandgap I calculated from the DOS.
In my uni cell, I have 16 atoms . In supercell of 311 48 atoms, 221 64 atoms, 222 128 atoms.
I tried with different K points set but still, nothing works.
Interesting question. First of all, when you change the size of supercells, the environment of atoms in certain direction changes. The final convergence has the total energy. We accept the total energy/per unit cell as almost the same. Sometimes if the atomic arrangement and environment have not energetically equiprobable, you might get a different bandgap. The electronic structure is associated with the crystal structure. Also, in your question, the information about geometrical optimization is missing. I guess the discrepancy of 50 meV might be due to that also. Even if you optimise the crystal structure using the option of all degrees of freedom (lattice parameter, atomic positions, etc.), this kind of difference may occur as the SCF calculation may have a chance to converge in local minima. The convergence criteria i.e. energy convergence can be set more tightly. Finally, do you really bother about 50 meV difference, when DFT results in itself based on the approximation? In my opinion you should not worry much until and unless the bandgap difference is more significant (> 100 meV). You have to take a close look on the density of states near chemical potential if the DOS feature is changing with different size/configurations of the supercell, then it is the point of big worries, and solutions need to be found out. Relax and proceed for further electronic structure properties as supercell calculations consume too much time. Good luck
Saurabh Singh
Thank you so much for sharing this useful information.
I completely agree with your suggestion also my dos configuration is not changing in the different supercells around chemical potential.
Thanks once again.
- Badges
- Science topic