GW+U and HSE+U

Question on input files/tags, interpreting output, etc.

Please check whether the answer to your question is given in the VASP online manual or has been discussed in this forum previously!

Moderators: Global Moderator, Moderator

Message
Author
yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

GW+U and HSE+U

#1 Post by yuanchang_li » Mon Jun 27, 2022 11:15 am

Dear friend:
I am considering rare earth elements and want to obtain BSE results,so I need to calculate GW. Because f electrons included in my system, maybe I need to consider on-site Coulomb interaction(U). I want to know whether it is feasible to consider the U value when calculating the self energy of GW, or whether it is meaningful to consider the U value when hybrid HSE is carried out, and whether there will be conflict when calculating GW and BSE in the program?
Best wishes

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#2 Post by fabien_tran1 » Mon Jun 27, 2022 1:36 pm

Hi,

In principle, using the LDA+U (or GGA+U) or HSE orbitals/eigenvalues as input to GW (and then BSE) is ok.

However, from the physical point of view it would not make sense to combine HSE with U (i.e., do HSE+U), because the Hartree-Fock exchange in HSE and U are playing essentially the same role (roughly speaking, LDA+U/GGA+U could be considered as a very crude approximation of hybrid functionals). Furthermore, it is not clear if combining HSE with U is programmed correctly.

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#3 Post by yuanchang_li » Tue Jun 28, 2022 3:47 am

Dear Fabien:
I try the GW+U in with the VASP code. However,I find a big ERROR called "LDA+U in combination with GW calculations causes a double counting of unidentifiable many-body perturbation terms" ,which is not a wanning . So I began to doubt the feasibility of this operation in VASP.
Can you give me some help and advice on this question?
Best wishes

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#4 Post by yuanchang_li » Tue Jun 28, 2022 3:55 am

Dear Fabien:
First of all, thank you for your reply !I try to do the GW+U with the VASP code . However ,I got an big ERROR ,which called "LDA+U in combination with GW calculations causes a double counting of unidentifiable many-body perturbation terms. " That is not an WANNING ,So I have doubts about using VASP program. Can you give me some suggestions or help to understand that error?
Best wishes

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#5 Post by fabien_tran1 » Tue Jun 28, 2022 7:53 am

Hi,

Which version of VASP are you using? In the most recent VASP versions the message "LDA+U in combination with GW calculations causes ..." is only a warning.

What are the steps of your GW calculation? Are you using the Single step procedure: GW in one go or not?

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#6 Post by yuanchang_li » Tue Jun 28, 2022 2:42 pm

Dear Fabien:
Thanks for your reply ! I calculate GW+U very smoothly, except for this error, but I still want to find out the reason. Please give me some help.
I used the vasp5.4.4 to the calculation with the 'Single step procedure: GW in one go'.
However ,about the second step 'G0W0 calculation' ,I used the ALGO=G0W0,which referred the 'VASP the GUIDE' ,rather than the ALGO=EvGW0 or the G0W0R.
Is that will be the reason of the error?
Best wishes!

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#7 Post by fabien_tran1 » Tue Jun 28, 2022 3:01 pm

Hi,

The "Single step procedure: GW in one go" works only when a LDA, GGA or meta-GGA functional is used for the DFT calculation. It is not yet implemented for DFT+U and hybrid.

If you want to use DFT+U or a hybrid instead, then you need to follow the procedure which consists of manually doing "First step", "Second step", ... (see https://www.vasp.at/wiki/index.php/Prac ... lculations). In principle the results should be correct. It should also not matter which flavor of GW you use.

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#8 Post by yuanchang_li » Tue Jun 28, 2022 4:44 pm

Dear Fabien:
Thank you for your reply !I'm sorry for my previous reply. I used the first and second steps you said. I read the manual carefully again.
I used the first step and the second step, and added the U value in the DFT calculation. Only G0W0 was selected in the second step, and the other parameters are the same as those on the official website.
About the problem of error, please give me more advice!
Thank you again for your patience!
Best wishes!

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#9 Post by fabien_tran1 » Tue Jun 28, 2022 5:47 pm

Can you please show the INCAR files of the DFT and GW calculations.

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#10 Post by yuanchang_li » Wed Jun 29, 2022 2:25 am

Dear Fabien :
Thanks for your help!There is my INCAR file.
Best wishes

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#11 Post by fabien_tran1 » Wed Jun 29, 2022 7:21 am

You forgot to attach it.

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#12 Post by yuanchang_li » Wed Jun 29, 2022 3:30 pm

Dear Fabien :
Thanks for your help!There is my INCAR file in the Files.
INCAR.rar
Best wishes
You do not have the required permissions to view the files attached to this post.

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#13 Post by fabien_tran1 » Wed Jun 29, 2022 4:07 pm

Hi,

It seems that you used the same single INCAR file for all calculations. What you could try instead is to separate the DFT and GW calculations by doing the first and second steps as explained at https://www.vasp.at/wiki/index.php/Prac ... lculations:
1) Create INCAR for PBE (or PBE+U) and execute VASP. This will generate WAVECAR that is used for the following GW calculation.
2) Create INCAR for GW and execute VASP.

Check if this works and if the results seem ok. If yes, then the choice of the functional chosen at step 1 should have an influence on the GW results.

Besides I see that your INCAR has "MAGMOM = 10 10 0 0" which is nonsense.

yuanchang_li
Newbie
Newbie
Posts: 28
Joined: Tue Mar 17, 2020 2:01 am

Re: GW+U and HSE+U

#14 Post by yuanchang_li » Wed Jun 29, 2022 5:31 pm

Dear Fabien:
Thank you very much for your help!
I think I didn't make it clear before. There are three calculations in total. So I uploaded the file again. Please take a look.
INCAR.rar
The first step is PBE + U calculation, the second step is ALGO=EXACT following the tutorial, and the third step is G0W0 operation. The three INCAR files are not the same ,according to the tutorial, I found that it is generally consistent.
As for your suggestion, I understand that you want to say that I do not need to add parameters about U value in INCAR during G0W0 calculation, such as LDAU = .TRUE.,etc. Is that right ?In other words, the control parameter about U value is removed from INCAR in GW calculation ? Is that right?
Thank you again for your selfless help!
Best wishes!
You do not have the required permissions to view the files attached to this post.

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: GW+U and HSE+U

#15 Post by fabien_tran1 » Wed Jun 29, 2022 7:04 pm

Ok. Yes, try again step 3 but with the whole "LDAU" block in INCAR removed. Of course, WAVECAR and WAVEDER generated at step 2 should be present when starting step 3.

According to my understanding, the GW step 3 needs only WAVECAR and WAVEDER from step 2 and it should not matter which functional (PBE, PBE+U or hybrid) was used to generate them (i.e., the GW step should not crash and produce correct results). As I wrote previously, the choice of the functional chosen at steps 1 and 2 should have an influence on the GW results.

Post Reply