Converting RL to Ry - EXXRLvcs tag

Concerns issues with computing quasiparticle corrections to the DFT eigenvalues - i.e., the self-energy within the GW approximation (-g n), or considering the Hartree-Fock exchange only (-x)

Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano

Post Reply
Jagjit Kaur
Posts: 20
Joined: Fri Apr 16, 2021 12:51 pm

Converting RL to Ry - EXXRLvcs tag

Post by Jagjit Kaur » Wed Aug 03, 2022 12:13 pm

Hello,

I am interested in doing GW and BSE calculations in YAMBO. In the tutorials,
gw_ppa.txt
it was said that the EXXRLvcs should be converted from units of RL to Ry. My questions are:
1. How to convert the units (r_setup is attached)?
2. Is it necessary to convert it to Ry? If yes, why is it so?

I have also attached the GW input file. How many bands are preferred to be taken in QPkrange? I want to top this calculation with BSE. So, should we take more number of valence and conduction bands to get accurate absorption spectra?
gw_ppa.txt
r_setup.txt
Thank you.
You do not have the required permissions to view the files attached to this post.
Jagjit Kaur
Harish-Chandra Research Institute
Allahabad
India

User avatar
Daniele Varsano
Posts: 3773
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: Converting RL to Ry - EXXRLvcs tag

Post by Daniele Varsano » Wed Aug 03, 2022 1:25 pm

Dear Jagjit,

1) It is not necessary at all to convert reciprocal lattice vectors, you can provide them as a number (RL) or as energy cutoff, all G such that G^2/2 < Ecut.
2) It depends on your GW correction behavoiur. In line of principle you should include all the energy to be corrected in the BSE, so the bands included in the BSE calculation over the whole IBZ. You can anyway caluculate a protion of that and inspect if a scissor/strechting approximation is valid (see e.g. step 3 in https://yambo-code.eu/wiki/index.php/Ho ... rial:_h-BN)

By plotting Eqp vs Elda, you can perform a linear fit on valence and conduction bands to calculate strethcing parameters to be assigned in the BSE runlevel using the KfnQP_E namelist.

Anyway, if you need to provide explicit qp energies from database (KfnQPdb), and some of them are missing, yambo will perform automatically a linear fit.

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/

Jagjit Kaur
Posts: 20
Joined: Fri Apr 16, 2021 12:51 pm

Re: Converting RL to Ry - EXXRLvcs tag

Post by Jagjit Kaur » Thu Aug 04, 2022 11:52 am

Thank you Daniele for clearing my doubts.

Regarding the link to the tutorial that you have posted here, it is about interpolating bandstructure by gw calculations. For the optics, I could find a tutorial for gw0+BSE calculation. I was also interested in optics with just gw0. But when I am generating the input file by the command,

Code: Select all

yambo -hf -gw0 p -dyson n -optics c -F gw_optics.in
, I don't get the "optics" tag in the input file.

I am a bit confused regarding that. So, do we need to go through similar steps as BSE+GW with screening, kernel etc? Or if we have the databases from the band structure at gw0 level, we can use that and directly get the dielectrics written in the output like at the IPA level.

If the question confuses you, the simpler question is "How to get optical spectra at GW0 level?"
Jagjit Kaur
Harish-Chandra Research Institute
Allahabad
India

User avatar
Daniele Varsano
Posts: 3773
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: Converting RL to Ry - EXXRLvcs tag

Post by Daniele Varsano » Thu Aug 04, 2022 6:04 pm

Dear jagit,
In the tutorial, besides the interpolation, there is also the procedure to find scissor and stretching to be used in bse, or ip. This is an alternative to the reading of the database.

Optics in gw is just an ip calculation where you provide gw energies.
You need to calculate QP energy first and generate the QP database, or alternatively find scissor/stething parameters.

Next you can do the IP calculation in G space:

Code: Select all

 yambo -o c -V qp
setting XfnQPdb or XfnQP_E to account for the GW corrections.

or equivalent in transition space,

Code: Select all

yambo -o b -V qp 
setting KfnQPdb or KfnQP_E

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/

Jagjit Kaur
Posts: 20
Joined: Fri Apr 16, 2021 12:51 pm

Re: Converting RL to Ry - EXXRLvcs tag

Post by Jagjit Kaur » Tue Aug 09, 2022 10:04 am

Hello,

Thank you for guiding me through the GW spectra calculations. Now I am done with the GW and GW+BSE spectra calculations. But, when I plot them, the amplitudes for both the imaginary dielectric come out to be way different. I have attached the report files for both the calculations.

Regards,
Jagjit
You do not have the required permissions to view the files attached to this post.
Jagjit Kaur
Harish-Chandra Research Institute
Allahabad
India

User avatar
Daniele Varsano
Posts: 3773
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: Converting RL to Ry - EXXRLvcs tag

Post by Daniele Varsano » Tue Aug 16, 2022 9:10 am

Dear Jagjit,

the BSE report file you attached is truncated, so I cannot see all the parameters.
Anyway, the difference in the spectra most probably is due to the different number of bands included in the calculations:
1-260 in the IP
205-215 in the BSE.

Please note that in the BSE output in columns 4-5 you also have the IP spectrum calculated with the same number of bands of the BSE. You can compare the two of them. If it is not at convergence (you need to look at the lower energy part of the spectrum, as usually, it is the interesting one) you will need to add more bands in the BSE spectrum.

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/

Post Reply