Supercell optical calculations

[Inaoton]

Dear ATK team
                                I have calculated optical spectrum for GaAs both for unit cell and super cell and I found both static dielectric constants to be same . But for CH3NH3PBI3, I am not getting the same value, with  lower value in case of super cell, though I am getting band gaps  be equal. How can solve this problem?

[Petr Khomyakov]

It likely means that you have to better converge k-point sampling for the CH3NH3PBI3 unit cell and supercell by increasing the number of k-points (k-point grid density). You should also check if your calculations are converged with respect to the number of bands above and below the Fermi level.

[Inaoton]

You should also check if your calculations are converged with respect to the number of bands above and below the Fermi level.

How to determine the number of bands?
How can I check the convergence w.r.t. number of bands?

[Petr Khomyakov]

You just have to increase the number of bands, do the optical spectrum calculation, and then verify if there is any significant difference with your calculations with less bands. And you do it until the optical spectrum is converged within the tolerance you set yourself. 

[Inaoton]

You just have to increase the number of bands, do the optical spectrum calculation, and then verify if there is any significant difference with your calculations with less bands. And you do it until the optical spectrum is converged within the tolerance you set yourself.

Yes I am getting a value of static dielectric constant which approaches the experimental value by increasing the number of bands. The number of bands has been increased to 500 for both above and below E_F. Can I estimate  the required number of bands from the band structure itself?

[Petr Khomyakov]

It would be rather difficult to estimate it from the band structure itself, because it also depends on the character of the wave functions (Bloch states) corresponding to the bands.  It is more practical to just do convergence tests to be sure that you get converged results.

As a matter of fact, convergence tests with respect to computational setting should be done for any kind of calculations. You may get some idea of what are the reasonable initial settings for a given material type after doing many calculations, developing some intuition. But one should do checks afterwards  in any case.