How to evaluate the e-ph constant using GW bandstructure with change of band ordering related to DFT

General discussion around the EPW software

Moderator: stiwari

Post Reply
hellolori
Posts: 26
Joined: Mon Dec 18, 2017 5:19 am
Affiliation:

How to evaluate the e-ph constant using GW bandstructure with change of band ordering related to DFT

Post by hellolori »

Dear developers,

I recently try to calculate the e-ph constant and phonon linewidth with the correction of QP energy by G0W0 approximation using QE + EPW + Yambo + Wannier90. I also learn that the above calculation can be done by setting the flag "efermi_read, eig_read". What I have managed to do are (i) calculate the e-ph constant and phonon linewidth using DFT bandstructure; (ii) obtain Wannier-interpolated G0W0 bandstructure using Yambo + Wannier90. I have the following questions.
(1) As I understand, currently the EPW code can only take into account the correction of electronic energy by GW/HSE, while the electron-phonon coupling matrix elements are kept the same as the results of DFT. Is that right?
(2) As I learn from the decription of the variable "eig_read" : "If .true. then read a set of eigenvalues from ksdata.fmt. Can be used to read GW (or other) eigenenergies. The code expect a file called “prefix.eig” to be read. One need to provide the same number of bands as in the nscf calculations and all k-points." I know how to prepare the file "prefix.eig", but I have no idea how to prepare the file "ksdata.fmt" which is needed by the calculation according to the above description.
(3) As I understand,the calculated results maybe wrong if the change of the band ordering happens after GW, which is in my case. How can I do to obtain the reasonable results in the case of change of the band ordering? I know that there is a script in Wannier90/utility/gw2wannier90.py, which can modified the $prefix.mmn, $prefix.amn $prefix.eig files according to the change of band ordering. By using the above script, the correct Wannier interpolation of bandstructure can be realized. In my opinion, if the EPW code can read the $prefix.mmn, $prefix.amn $prefix.eig to recalculate the Wannier, the problem maybe solved. However, it seems that the EPW can not read these files.

Bests,
Feipeng

hlee
Posts: 415
Joined: Thu Aug 03, 2017 12:24 pm
Affiliation: The University of Texas at Austin

Re: How to evaluate the e-ph constant using GW bandstructure with change of band ordering related to DFT

Post by hlee »

Dear Feipeng:

My answers to your questions are as follows:
(1) If eig_read=.true., EPW imports the eigenvalues from prefix.eig containing eigenvalues from the external codes. Then these values are applied to the diagonal parts of Hamiltonian, leading to the different rotation matrices. So I think that the e-ph vertex also changes; that is, the e-ph vertex on coarse grids doesn't change, but the interpolated ones change since rotation matrices are multiplied when interpolating the e-ph vertex.

(2) ksdata.fmt is generated internally from prefix.eig by EPW; you don't need to generate it. This ksdata.fmt is used for restart purpose.

(3) I have no experience with Wannier90/utility/gw2wannier90.py. In some cases, if you generate eigenvalues with the corrected band order using gw2wannier90.py, you don't need further action. But, in general situations, there might be some issues.

Sincerely,

H. Lee

Post Reply