GW convergence

Please post here any issue related to convergence parameters in linear response GW and BSE calculations.

Moderators: Davide Sangalli, Daniele Varsano

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

GW convergence

Post by arb83@cam.ac.uk » Thu Oct 01, 2020 5:18 pm

Dear Yambo community forum

I’m carrying out convergence of GW calculations with respect to NGsBlkXp and BndsRnXp. I realise I am unclear on whether convergence means when the bandgap is converged, or when the valence and conduction band energies are converged. In my system the bandgap is converged but the valence and conduction band energies are increasing by the same amount as I increase BndsRnXp (for a long while after the bandgap has converged). I would like to use the parameters I obtain for BSE calculations.

Any advice would be much appreciated.

With best wishes,

Alan
Alan Bowman
University of Cambridge

User avatar
Daniele Varsano
Posts: 2992
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: GW convergence

Post by Daniele Varsano » Fri Oct 02, 2020 8:09 am

Dear Alan,
what you are observing is a rather common behaviour.
It depends now what are interested in. If you need the GW correction for optical properties (BSE) it is enough to have well-converged bandgap and relative transitions energies. Different is the case where you want to calculate ionization potentials or electron affinities, as in this case you need to converge the single GW band energies.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

Re: GW convergence

Post by arb83@cam.ac.uk » Fri Oct 02, 2020 10:35 am

Dear Daniele

Thank you for the quick reply. In addition to needing the GW correction for optical properties, I would like to use the converged values of NGsBlkXp and BndsRnXp to set NGsBlkXs and BndsRnXs for the static screening in BSE calculations. Is it reasonable to use values of NGsBlkXp and BndsRnXp where the bandgap is converged to set these static screening parameters, or do I need to converge the VB and CB with respect to NGsBlkXp and BndsRnXp?

Many thanks for your help.

Best wishes,

Alan
Alan Bowman
University of Cambridge

User avatar
Daniele Varsano
Posts: 2992
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: GW convergence

Post by Daniele Varsano » Fri Oct 02, 2020 11:29 am

Dear Alan,

it is in general enough to use the same parameter. Moreover, if you have calculated the pp databases for the GW you do not need to recalculate the static screening as the static value it is already contained in that database and yambo is able to read them. Just instruct the input file to read them:
e.g.

Code: Select all

yambo -o b -k sex -p p -y d 
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

Re: GW convergence

Post by arb83@cam.ac.uk » Fri Oct 02, 2020 11:37 am

Dear Daniele

I don't quite understand your answer, when you say "it is in general enough to use the same parameter.", is that the parameter that has been converged with respect to the bandgap, or valence and conduction bands?

Many thanks for the advice on calculating the screening, that's very helpful!

Best wishes,

Alan
Alan Bowman
University of Cambridge

User avatar
Daniele Varsano
Posts: 2992
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: GW convergence

Post by Daniele Varsano » Fri Oct 02, 2020 11:39 am

Dear Alan,
as in the BSE you are interested in converged transition energies and not absolute value of each of them, it is enough to use the parameters you obtained to converge the gap.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

Re: GW convergence

Post by arb83@cam.ac.uk » Fri Oct 02, 2020 12:06 pm

Dear Daniele

That's perfect, many thanks!

Best wishes,

Alan
Alan Bowman
University of Cambridge

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

Re: GW convergence

Post by arb83@cam.ac.uk » Wed Nov 25, 2020 1:36 pm

Dear Daniele

I have been trying to do GW convergence of a 2D system (with vacuum spacing) but am finding some odd results as I increase the number of bands included in the simulation. Specifically, for bands up to 1200 the system seems well converged, but above this the results seem to diverge (results below with NGsBlkXp=6Ry):

Bands VB CB Eg
600 0.5178 4.133 3.6152
800 0.6161 4.251 3.6349
1000 0.7042 4.269 3.5648
1200 0.7626 4.299 3.5364
1400 0.8964 4.159 3.2626
1600 0.6378 4.476 3.8382

I attach an example input and run file, along with r_setup. Is there something obvious I am doing wrong? I haven't used the box cutoff before is the only thing i can really think of. Note I didn't use a box cutoff in the QE calculations but I found the vacuum sufficiently large that there wasn't interaction between neighbouring blocks.

Any suggestions would be much appreciated.

Best wishes,

Alan
You do not have the required permissions to view the files attached to this post.
Alan Bowman
University of Cambridge

User avatar
Daniele Varsano
Posts: 2992
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: GW convergence

Post by Daniele Varsano » Wed Nov 25, 2020 2:09 pm

Dear Alan,

your cell in z-direction has a size of nearly 60 Bohr (62.33173).
Does not make sense to set a cutoff Box side of 90 bohr (larger than the cell).

The box size should be slightly smaller than your cell side e.g. 60 bohr in your case.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

arb83@cam.ac.uk
Posts: 98
Joined: Thu Jul 02, 2020 3:56 pm

Re: GW convergence

Post by arb83@cam.ac.uk » Wed Nov 25, 2020 3:04 pm

Dear Daniele

Sorry, where did the value of 62.33 bohr come from? In the r_setup file the z direction is given as 94.49 (under 02 the unit cell section).

Best wishes,

Alan
Alan Bowman
University of Cambridge

Post Reply