BSE convergence question

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

Moderators: Davide Sangalli, Daniele Varsano

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

BSE convergence question

Post by arb83@cam.ac.uk » Wed Nov 11, 2020 3:51 pm

Dear Yambo community forum

I realise I am unclear on BSE convergence. Should I converge the .eps spectra, or the relative strength of the exciton eigenvalues (.exc_qpt1_E_sorted)? If I converge on the former then I find there are still small differences in the latter at larger values of BSENGBlk and BSENGexx. My guess is that this is to do with normalising the relative strengths, but I wanted to double check that converging on the .eps spectra is sufficient to say I have a converged parameters?

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: BSE convergence question

Post by Daniele Varsano » Wed Nov 11, 2020 4:05 pm

Dear Alan,
the two are perfectly equivalent as the eps spectrum is built considering the excitation energies and the strength found in the .exc_qpt1_E_sorted file.
Yes, as you argued, in the sorted file the strengths are normalized to its maximum.
You can look at the normalization factor in your output indicated as:

Code: Select all

Maximum Residual Value =
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: BSE convergence question

Post by arb83@cam.ac.uk » Wed Nov 11, 2020 6:13 pm

Dear Daniele

Thanks for the quick response. I attach two files to demonstrate the problem. I'm converging on BSENGexx and while the .eps file looks converged, the energy outputs do not. The integrated value does not seem to change much hence my confusion. I'm probably just missing something obvious but if you could advise it'd be much appreciated.

Best wishes,

Alan

P.S. The integrated strengths go as follows for the different calculations:

10 .24664E+00
20 .26816E+00
30 .27925E+00
40 .27270E+00
60 .27675E+00
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: BSE convergence question

Post by Daniele Varsano » Thu Nov 12, 2020 8:51 am

Dear Alan,
can you please post the outputs (eps, and weight) e.g. for a couple of them (e.g. 10Ry and 60Ry).
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: BSE convergence question

Post by arb83@cam.ac.uk » Thu Nov 12, 2020 11:34 am

Dear Daniele

Please find attached. Note I haven't done convergence with respect to bands yet for this system.

Best wishes,

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

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

Re: BSE convergence question

Post by arb83@cam.ac.uk » Thu Nov 12, 2020 11:35 am

Final attachment...
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: BSE convergence question

Post by Daniele Varsano » Thu Nov 12, 2020 12:22 pm

Dear Alan,
the valures you reported are not the integrated strengths, but the value of the higher peak.
The strengths are normalized to this value (the higher peaks is 1) and not the integral.

If you rescale to this value, for instance, the first excitation you get for the two cases:

0.08120057732 *.27675= .02247
0.09095347673 * .24664= 02243

The spectra are very well converged.

Pay attention, the file you named 60Ry is 46Ry: see the mirror of the input file at the end of the file eps.

Code: Select all

BSENGexx=  46          Ry   
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: BSE convergence question

Post by arb83@cam.ac.uk » Thu Nov 12, 2020 1:04 pm

Dear Daniele

I'm afraid this still doesn't fully make sense. If you multiply the normalised strengths by the maximum residual value not all the strengths are converged - see the attached plot.

Seeing as the .eps file is converged in all cases, is this something to do with oscillator strengths being passed between different excitons with comparable energies?

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: BSE convergence question

Post by Daniele Varsano » Thu Nov 12, 2020 1:19 pm

Dear Alan,
the excitation energies are at the same position.
Probably you have just a small redistribution of the intensity that you cannot perceive once the convolution of the spectra is done.
Probably you can recognize it by decreasing the vale of the damping:

Code: Select all

BDmRange
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: BSE convergence question

Post by arb83@cam.ac.uk » Thu Nov 12, 2020 1:36 pm

Dear Daniele

Perfect, that makes sense, thank you!

Best wishes,

Alan
Alan Bowman
University of Cambridge

Post Reply