Does the mobility relate to the energy windows of wannier90?

[joshie]

Hello, everyone,
Recently, I have tested the example of calculating the mobility of bulk silicon using EPW code. I have obtained different results of mobility with various parameters of energy windows
(dis_froz_max/min, dis_win_max) in the process of wannier fitting. I found they were set to be dis_win_max = 18.d0 dis_froz_max = 8.5d0 in the original example, and got good fitting
band structures compared to the bands from the DFT. However, when I used dis_win_max = 18.d0, dis_froz_max = 9.5d0 and dis_froz_max = 3.5d0, I got the similar fitting plot as the
example, but the mobility becamehalf the size now. Besides, I tried to increase the dis_froz_max = 12 to include more wannier funtions, but the fitting bands became a little bad that I
cannot understand. I think that the mobility should only relate to the electrons and holes near the Fermi level. Why I obtained very different mobility, although the fitting results look the same.

Looking forward to your help.

[sponce]

Dear joshie,

If by example you refer to the q-e/test-suite/epw_mob* examples, those are used for CI testing and are absolutely not converged. You need much denser grids.

Regarding the Wannierization, having larger frozen window puts more constrain on the Wannierzation making it more difficult for the code so that you might end up with worst results.

In your question you refer twice to dis_froz_max = 9.5d0 and dis_froz_max = 3.5d0. I guess the latest is dis_froz_min ?

To answer your question, yes the mobility should not change if the bandstructure and interpolated matrix elements are very similar.

If you want more tailored answer, please provide figures of interpolated bands, matrix elements and convergence with fine k/q grids.

Best wishes,
Samuel