Inconsistent EPW and PWSCF bands

Running EPW
1

Deal all
My wannier interpolated band structure is different from the pwscf band structure, but phonon bands interpolate reasonably well. The band structure and phonon bands are attached. epw input file is given below.

I increased num_iter and changed projections but could not solve the problem.

Kindly help how to interpolate the electronic band structure well.



epw.in

&inputepw
prefix = ‘pts2’,
amass(1) = 195.08
amass(2) = 32.07
outdir = ‘./’
dvscf_dir = ‘../phonon/save’

elph = .true.
epwwrite = .true.
epwread = .false.
! use_ws = .true.

etf_mem = 0

wannierize = .true.
nbndsub = 14
bands_skipped = ‘exclude_bands = 1-6’
num_iter = 700
dis_win_max = 8
dis_froz_max= 2
! use_bloch_phases = true
proj(1) = ‘Pt:s;d’
proj(2) = ‘S:sp3’
wannier_plot= .true.

band_plot = .true.
filkf = ‘path2.dat’
filqf = ‘path2.dat’

nk1 = 16
nk2 = 16
nk3 = 1
nq1 = 8
nq2 = 8
nq3 = 1
/

2

Hi mirshowkat07,

There are few things you can play with during wannierization:

  1. number of Wannier bands,
    “nbndsub = 14”

  2. initial projections,
    "proj(1) = ‘Pt:s;d’
    proj(2) = ‘S:sp3’ "

  3. frozen window
    “dis_froz_max= 2”

One last thing, to check the interpolation, you can plot “prefix_bands.dat” from the Wannier90 calculation on the top of the pwscf band before going to an EPW calculation.

I think you will get the best answer from Wannier90 community, here is the link
http://www.wannier.org/support/

Best,
Hari

3

Dear mirshowkat07:

In addition to Hari’s answer, I would like to ask you to check the wout files generated after Wannierization; you can get the hints on what is going on.

Sincerely,

H. Lee

4

Dear H. Lee and Hari

Thank you for the reply.

I think kpoints, nk1, nk2 nk3 also play a critical role in interpolation.
First, I increased the kpoints from 16 16 1 to 24 24, there was no noticeable change, but
for kpoints 32 32 1, the wannier interpolated bands have improved greatly.



with best regards
Showkat

5

Dear Showkat:

It is natural that you can achieve the good interpolation quality when using large number of k points in nscf calculations.

You have to achieve the good interpolation quality even when using coarse grids and it can be done if you construct good (well localized, etc.) Wannier functions.
For this purpose, you need to follow our previous suggestions.

Sincerely,

H. Lee

6

Dear H. Lee

Thanks a ton for your suggestion.
I will try to get converged results for the coarse grid.


with best regards
Showkat