Mismatch in the number of bands.

General discussion around the EPW software

Moderator: stiwari

Post Reply
CHILLL
Posts: 50
Joined: Fri Apr 26, 2019 11:39 am
Affiliation:

Mismatch in the number of bands.

Post by CHILLL »

Dear experts,

I have been doing calculations on a 2D structure.
https://ibb.co/BGss7Kh : Here are the bands I get from Wannier90. I indicated the Fermi level and the window that I used in epw calculations (fsthick=0.4 eV).
I also indicated all the bands that cross the Fermi level in different color, and I get three bands.
When I look at the epw output file I see: 2 bands within the Fermi window.

I do not consider some lowest bands in the calculations by using a corresponding window in Wannierisation. In EPW I account for it by using nbndskip.
The Fermi level from Quantum espresso is 3.2362 eV, and the one I get from EPW is 3.247973 eV, which is almost identical (the same bands cross both levels). So, there shouldn't be a mistake related to the incorrectness of the Fermi level.

Could you please tell me why the code picks only two bands even though it seems that there are three bands that cross the Fermi level? And at least one more band is within the window, but I just do not plot it here.

Best regards, Mikhail

roxana
Posts: 172
Joined: Fri Jan 22, 2016 6:48 pm
Affiliation:

Re: Mismatch in the number of bands.

Post by roxana »

Hi,


Please plot the Wannier band structure obtained with EPW on top of the band structure obtained with QE. Can you also provide a snapshot of the epw.out file that shows the information regarding min and max indexes of the bands within the Fermi window?

Best,
Roxana
Roxana Margine
Associate Professor
Department of Physics, Applied Physics and Astronomy
Binghamton University, State University of New York

CHILLL
Posts: 50
Joined: Fri Apr 26, 2019 11:39 am
Affiliation:

Re: Mismatch in the number of bands.

Post by CHILLL »

Dear Roxana,

Thank you for your quick response.
Here are the band structures: https://ibb.co/q0kmLC8
To get bands from EPW, I followed the instructions described in this lecture: http://epw.org.uk/School2018/School2018?action=downloadman&upname=Wed.4.Verdi.pdf
It seems that after you ask Wannier part of the code to create bands, they get printed. So why do we need this extra step in epw?
I mean the step with:
band_plot = .true.
filqf = 'prefix_band.kpt'
filkf = 'prefix_band.kpt'

Does it produce bands that are different from the Wannier bands?
Here is the snapshot (I hope the right one) : https://ibb.co/ypcvd45

I have also been looking at this variables (ibndmin/mav and ebndmin/max), but I couldn't find a lot about their meaning. Could you please explain what exactly do they represent?

Best, Mikhail

roxana
Posts: 172
Joined: Fri Jan 22, 2016 6:48 pm
Affiliation:

Re: Mismatch in the number of bands.

Post by roxana »

Hi,

There aren't "Wannier bands", since the Wannier space is a real space representation. What happens in Wannier90 (and similarly in EPW) one goes from k-space (the coarse mesh) to Wannier representation (real space) and then back to the k-space (this final k-space can by a uniform or random mesh, or k-path along specific lines).

ibndmin is the band index of the lowest energy band within the Fermi window, e.g. ebndmin
ibndmax is the band index of the lowest energy band within the Fermi window, e.g. ebndmax

ebndmin and ebndmax are printed in Ryd

In your case,

ibndmin = 2, ebndmin = 0.209 Ryd
ibndmax = 7, ebndmax = 0.268 Ryd

The code prints the nr. of bands within the Fermi window, not the nr. of bands crossing the Fermi window. For instance at G-point, there may be 2 bands within the Fermi window, but at M there may be 3 bands. In these case, the code prints 3 bands within the Fermi window.

Best,
Roxana
Roxana Margine
Associate Professor
Department of Physics, Applied Physics and Astronomy
Binghamton University, State University of New York

CHILLL
Posts: 50
Joined: Fri Apr 26, 2019 11:39 am
Affiliation:

Re: Mismatch in the number of bands.

Post by CHILLL »

Dear Roxana,

Thank you for you response.
Just to clarify, now I have only 2 bands considered by the code. Does that mean that for every k point considered there are not more than 2 bands that lie simultaneously within the Fermi window? And for every point these 2 bands can be any of those from ibndmin to ibndmax bands?

If this is true, then in my case all 6 bands (from ibndmin=2 to ibndmax=7) contribute to the final result. May I ask what is the idea behind printing the output (e.g. gap on the FS) per band? Now I have the output printed for 2 bands, but in reality these 2 bands encompass all 6 bands within the window. So is there a physical meaning behind this per band gap printing? When you do calculations do you just sum this band contributions ?

Best, Mikhail

roxana
Posts: 172
Joined: Fri Jan 22, 2016 6:48 pm
Affiliation:

Re: Mismatch in the number of bands.

Post by roxana »

Hi,

Yes, any of the 6 bands with indexes between ibndmin and ibndmax can be within the Fermi window at some k-point. The bands can interchange, but the code doesn't take this into account. At every k-point, the bands are ordered by increasing energy value. For the results of the superconducting gap you can only sum the up if the contributions are on different parts on the Fermi window. You can check the following tutorial on youtube for the case on MgB2.

http://epw.org.uk/Documentation/MgB2

Best,
Roxana
Roxana Margine
Associate Professor
Department of Physics, Applied Physics and Astronomy
Binghamton University, State University of New York

Post Reply