BSE results with Coulomb cut off method

Deals with issues related to computation of optical spectra in reciprocal space: RPA, TDDFT, local field effects.

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

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

BSE results with Coulomb cut off method

Post by Laura Caputo » Tue Apr 26, 2022 3:44 pm

Good afternoon,

I am calculating the excitonic effect in some 2D materials by solving the BS equation. Following the tutorial on the website, I've used the RIM method and the Coulomb potential cut off. Using these parameters, I have obtained a spectrum with very low absorbance compared to the examples on the website.
On the other side, when not using the Coulomb cut off method, I obtain different peaks (which I could expect since the vacuum is not converged) but the absorbance is way higher and comparable to the examples in the tutorials. Are these differences (peaks and absorbance) just due to the different physics of the systems or is it something wrong in the files? I have put in the attachment the report/output files of the calculation with and without Coulomb truncation.

Moreover, I have calculated the GW gap also for this system (to do the convergence of vacuum and screening) at this k-point mesh (3x3) and the QP gap is way higher (around 3.7eV) with Coulomb box. I know that with BSE that gap is reduced since it's an optical gap (with box, the first peak is at around 1eV). Is it possible to reduce it that much or should I confront the gaps just at converged kpoints?
You do not have the required permissions to view the files attached to this post.
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

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

Re: BSE results with Coulomb cut off method

Post by Daniele Varsano » Tue Apr 26, 2022 4:27 pm

Dear Laura,

1) low intensity in epsilon, Please have a look to faq on the usage section:
http://www.yambo-code.org/faq/

2)It seems you do not have included qp corrections in the diagonal part of the BSE kernel. This ca be done by reading the qp database or by approximate them with a scissor operator. You can have a look to the tutorials. The corresponding variables KfnQPdb, KfnQP._E are listed when adding "-V qp" in the generation of the input file. Please check the exact name of the variables in the input variables page, I'm writing from a mobile phone and it's possible I'm misspelling them.

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/

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

Re: BSE results with Coulomb cut off method

Post by Laura Caputo » Tue Apr 26, 2022 5:07 pm

Hello,

Thanks for your reply.

1) So, as far as my understanding goes, due to the construction of the Coulomb potential, the vacuum is what makes the epsilon very small. So, if I'm just interested in the excitonic properties and in particular the bright ones (particularly only on their position in energy), it should not be an error or a problem that needs to be corrected, and it should be enough to just consider the sharpest peak, like in the tutorial, right?

2) I've looked a little bit better at this point that I think I've missed. Basically, one should do firstly a GW calculation and check the QP correction from that. Then, read the QP corrections or use the scissor operator. In the last case, for the syntax of KfnQP_E, it should be (Conduction band correction - Valence band correction) | 1.000 | 1.000 ?

3) For one of the calculations for which I have already a GW database, I've used ad input like in the attachment and used as the database for the calculation the one containing the GW calculated screening. The band gap is still quite low compared to the GW case, so I was wondering if everything was ok or I missed some error.
You do not have the required permissions to view the files attached to this post.
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

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

Re: BSE results with Coulomb cut off method

Post by Daniele Varsano » Wed Apr 27, 2022 8:53 am

Dear Laura

1) if you are interested in the absorption you may have a look at the alpha output. If you are interested in only the excitation energy eps and alpha are proportional. What makes the intensity very low is the fact you use a truncated potential that doe not behave as 1/q^2

2) You can either read the QP energy from database or use a scissor/stretching approximation. The stretching parameters can be obtained by a fit plotting Eqp vs Eks, see Step 3 of this tutorial:
http://www.yambo-code.org/wiki/index.ph ... 21_version

3) In this calculation you are using only two bands and they could not be enough. The number of valence and conduction bands in BSE is a convergence parameter. Also, note you calculated the QP correction for just two bands and one k point. What yambo does is then interpolate the missing QP energies. In this case you have just two corrections so the fit could be rather inaccurate. I suggest you calculate the QP correction for few bands around Fermi for all the k points. Also, if you want to apply a scissor stretching you do not have enough data to perform the 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/

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

Re: BSE results with Coulomb cut off method

Post by Laura Caputo » Wed Apr 27, 2022 10:23 am

Dear Daniele,

Thanks for your reply. If I understand correctly, I should modify the %QPkrange input to all the kpoints and using more than only VBM and CBM.
Would you suggest calculating the QP corrections for the same amount of bands around the Fermi energy as used in BSE (or more)?

However, I had another issue,I tried to do a calculation with your suggestion (report in the attachment). It should have read the QP corrections (there was the line indicated in the tutorial) and I've changed the em1s line into ppa. I used, in particular, yambo -F *BSEinput* -J *gwdatabase*, is it okay? This way it doesn't calculate again dipole and screening. I tried to converge BSENGexx, BSENGBlk and BSEbands but as you can see it doesn't change much when changing these parameters, if not for the bands convergence. This is quite strange for me to achieve convergence at 1Ry, so, is there something wrong in the report file? Because I cannot find any error message or maybe I'm missing something.


Also, if I let the code read the QP database, I guess the stretching parameters should be left to 1 due to the non-usage of the scissor operator, right?
You do not have the required permissions to view the files attached to this post.
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

Re: BSE results with Coulomb cut off method

Post by Laura Caputo » Fri Apr 29, 2022 4:26 pm

I post here the report.
You do not have the required permissions to view the files attached to this post.
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

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

Re: BSE results with Coulomb cut off method

Post by Daniele Varsano » Sat Apr 30, 2022 10:29 am

Dear Laura,
Would you suggest calculating the QP corrections for the same amount of bands around the Fermi energy as used in BSE (or more)?
In this case, you provide explicitly all the QP energies and there is no need for scissor/stretching. Scissor/stretching is an approximation that in many case is valid and allow you to save time as you do not need to calculate the QP correction for *all* the kpoints/bands but it needs to be verified if it is valid.
Also, if I let the code read the QP database, I guess the stretching parameters should be left to 1 due to the non-usage of the scissor operator, right?
Yes, if you read the QP you do not need to assign scissor/stretching parameters.

Indeed you obtain convergences with very low parameters, in particular the exchange. I cannot inspect the output as it contains the jdos and not the absorption. Next, you post the report for the first calculation only, it would be useful to look also at the other reports in order to see if all the variable were read correctly.

My suggestion is:
1) Consider more bands than the one you posted (95-99) as in this case you have a very small matrix.
2) Try to converge using a large step for the exchange (20.40-60) it does not cost much computationally
3) The same for the direct terms (5-10-15-..)
4) If nothing change please post a couple of reports/ouputs

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/

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

Re: BSE results with Coulomb cut off method

Post by Laura Caputo » Sun May 01, 2022 10:58 am

Dear Daniele,

Before posting the outputs,
In this case, you provide explicitly all the QP energies and there is no need for scissor/stretching. Scissor/stretching is an approximation that in many case is valid and allow you to save time as you do not need to calculate the QP correction for *all* the kpoints/bands but it needs to be verified if it is valid.
By this, do you mean that if I provide the QP energies I should calculate the corrections for all the k-points and bands? Or all the kpoints and 20/30 occupied/empty bands?
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

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

Re: BSE results with Coulomb cut off method

Post by Daniele Varsano » Sun May 01, 2022 3:34 pm

Dear Laura,

the options are:

1) Calculate the QPs for all the k-points and bands you insert in the BSE. Yambo will read the QP and will insert them in the diagonal part of the kernel
2) Calculate the QPs for a subset of kpoitns or bands. Yambo will read the database and the missing energies are automatically obtained by a fit done on the calculated energies
3) You provide scissor/stretching parameters. Yambo will not read any database and it will apply the QP corrections according to the parameter you inserted.

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/

Laura Caputo
Posts: 88
Joined: Sun Apr 11, 2021 3:02 pm

Re: BSE results with Coulomb cut off method

Post by Laura Caputo » Sun May 01, 2022 3:40 pm

Dear Daniele,

Thanks. So, if I understand correctly, I should always calculate the QP corrections for all the k-points (since BSE uses all of them) and for the bands, the QP corrections of the bands used in BSEbands.

Regarding the strange convergence that I obtain, I noticed that in the automatically generated file in yambo there are, other than em1s to change to ppa if the QP corrections are provided, also all the inputs of the static screening like Blks, BndsRnXs. Should they be changed in 'p' (like BndsRnXp) if the ppa option is used? Could it be a cause of this strange behaviour?
Laura Caputo
Ph.D. Student
Université Catholique de Louvain
https://uclouvain.be/fr/repertoires/laura.caputo

Post Reply