Calculation of a2F and lambda with Hnm(R)

General discussion around the EPW software

Moderator: stiwari

Post Reply
chendix
Posts: 5
Joined: Mon Dec 19, 2022 6:38 pm
Affiliation: Clemson University

Calculation of a2F and lambda with Hnm(R)

Post by chendix »

Dear Friends,

I'm a newcomer to the EPW calculation, so I apologize if my question is too generic. I have a series of Hamiltonians Hnm(R) of a hydride in the same set of MLWF basis under different external fields. Now I'm searching on how to calculate the Eliashberg function a2F and e-ph coupling constant lambda, using these Hamiltonians, or their eigenvalues and eigenvectors, to benchmark the available results in literature. Is this doable in our current EPW system? Or is there an alternative method?

Thank you and wish you all a happy new year!

Best regards,
Chendix

chendix
Posts: 5
Joined: Mon Dec 19, 2022 6:38 pm
Affiliation: Clemson University

Re: Calculation of a2F and lambda with Hnm(R)

Post by chendix »

As a follow up of this question, I'm replacing the hr.dat file and .eig file with my modified Hamiltonians and eigenvalues. However, after I fed the modified hr.dat and .eig files into EPW, it grows stranger to the original band structure when the external potential increases.
0.00.png
0.00.png (154.39 KiB) Viewed 12805 times
1.00.png
1.00.png (174.43 KiB) Viewed 12805 times
From my understandings, to reconstruct the Hamiltonian from Wannier to Bloch space, I'm supposed to have both eigenvalues and eigenvectors. Based on these two graphs, I believe that the problem occurs when I used the original eigenvectors (aka Wannier basis) and the modified eigenvalues to reconstruct the band structure. I'm now reading the codes to find the place to plug in the eigenvectors, or to read hr.dat, but my knowledge on F90 is limited.

I appreciate all comments and helps!

Best regards,
Chendix

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

Re: Calculation of a2F and lambda with Hnm(R)

Post by hlee »

Dear Chendix:
.., I'm replacing the hr.dat file and .eig file with my modified Hamiltonians and eigenvalues. However, after I fed the modified hr.dat and .eig files into EPW ...
How did you feed them into EPW?

The correct way is to (1) perform scf, nscf, phonon calculations with external fields, (2) starting from the obtained results to carry out EPW calculations.

It is not the correct way just to use the Hamiltonian and the eigenvalues under external fields without going through the steps above.

Sincerely,

H. Lee

chendix
Posts: 5
Joined: Mon Dec 19, 2022 6:38 pm
Affiliation: Clemson University

Re: Calculation of a2F and lambda with Hnm(R)

Post by chendix »

hlee wrote:
Wed Feb 08, 2023 4:33 pm
Dear Chendix:
.., I'm replacing the hr.dat file and .eig file with my modified Hamiltonians and eigenvalues. However, after I fed the modified hr.dat and .eig files into EPW ...
How did you feed them into EPW?

The correct way is to (1) perform scf, nscf, phonon calculations with external fields, (2) starting from the obtained results to carry out EPW calculations.

It is not the correct way just to use the Hamiltonian and the eigenvalues under external fields without going through the steps above.

Sincerely,

H. Lee
Hi Dr. Lee,

Thank you for the response.

Actually the external fields are regarded as ultrafast pump fields. Considering the retardedness of phonon in ultrafast timescale, we assume there's no change on relaxation, scf, nscf and phonon. Based on this idea, we ran the Wannierization step of EPW to generate .ukk, .mmn and .bvec files and then we replaced the Hamiltonian and eigenvalues by our modified version on el-ph coupling step to calculate lambda and a2f. (Please correct me if I'm wrong, but I don't think it'll read H_R, since H_R (variable chw(nbndsub, nbndsub, nrr)) is directly delivered from subroutine 'hambloch2wan' to 'hamwan2bloch')

However, with the lack of correct eigenvectors, the band structure shows great difference. To solve this problem once and for all, I'm now learning the EPW source code to find the real space vectors required for my calculation, which can be derived from lattice and orbital indices (first five columns in hr.dat). I'm wondering if you are familiar with subroutine 'hambloch2wan' in bloch2wan.f90, specifically for the variable wslen(nrr), which represents Wigner-Seitz vector length?

Thank you again.

Best regards,
Chendix

Post Reply