EPW+Wannier

[spguo]

Dear EPW,

I am trying to learn the EPW method based on the PWSCF framework. It seems that EPW embeds the Wannier function automatically. During the PWSCF+EPW+Wannier calculations, we have to specifically choose the projections of each atom in the wannierization. Usually, it is pretty hard to guess good initial projections for a complicated system. I notice that the latest Wannier90 code (V3.0) could perform a methodology of automatic generation of initial projections, so-called the selected columns of the density matrix (SCDM) algorithm. However, it seems that the SCDM algorithm is not included in the current EPW code.
1. I am wondering how to implement the algorithm in the EPW code?
2. Or could we separately run Wannier90 and EPW? In other words, could we firstly use Wannier90 (SCDM) to wannierize a system and then use EPW to do the rest work. And how to do that?

Thank you very much!
Cathy

[hpaudya1]

Hi Cathy,

I have an answer for your 2nd question. You can run Wannier90 and EPW separately. Only you need to save "prefix.ukk" file from Wannier90 and use it in the EPW calculations (please make sure that "wannierize = .false." in epw input).

Best,
Hari

[spguo]

Hi Hari,
Thanks for your help. Following your suggestions, I have run Wannier90 and EPW separately. Unfortunately, I encounter the problem to save the prefix.ukk file. I check the codes and find that it is EPW not Wannier90 that could generate the file. However, the write_filukk subroutine is part of the EPW code and it needs inputs from pervious subroutine runs. Would you please tell me how to run the write_filukk subroutine separately, based on the Wannier90 outputs?
Thanks,
Cathy