BSE and GW optical and structure post processing

[SKM]

Thank you Marie, for the reply in the thread forum/viewtopic.php?f=4&t=18223#p20768
The query below is continution to the above thread, but different issue
-----------
i tried with the INCAR tags you suggested in above thread (with slight added things based on some error messages in between). However, finally it run for Wannier and i got the Band. [bold] But [/bold] i realized the 'run' before applying the Wannier was based on BSE (i.e. ALGO) not GW. In such case i supposed to get the optical spectrum data. Is int it? As my intention was doing BSE for optical gap and GW for electronic band gap to compute binding energy of exciton.

So i used the example given scripts for picking-up optical data. but the scripts (except extract_optics-diag.sh) after run, gave an empty files. The ...diag file contain data, but i could not make out what is its use of denationalization values.?

I used the following scripts (given in tutorials/examples in this VASP site)
extract_chi.sh
extract_optics-diag.sh
extract_optics.sh

So, my queries are
1. Did i miss some INCAR tags that needed for optical calculations, when ALGO=BSE? (the INCAR i used for BSE run is given below)

System = xxx

PREC = Normal ;

ALGO = BSE
ANTIRES = 0
ISMEAR = 0 ; SIGMA = 0.01
ENCUTGW = 150

EDIFF = 1.E-8 # default 1.E-4
NBANDS = 288
NBANDSO = 4
NBANDSV = 8
OMEGAMAX = 20
ISYM = 0
PRECFOCK = Normal

The last few lines of 'log file' are as below:


performed 48880000 updates of chi_q(r,r)
total number of BLAS operations 8391769.83 Gflops
NQ= 19 0.4444 0.5000 0.0000,
|.........|.........|.........|.........|.........|.........|.........|.........|.........|......
...|.........|.........|.........|.........|.........|.........|.........|.........|.........|...
......|.........|.........|.........|.........|.........|.........|.........|.........|.........|
.........|.........|.........|.........|.........|.........|.........|.........|.........|.......
..|.........|.........|.........
performed 51700000 updates of chi_q(r,r)
total number of BLAS operations 8879170.49 Gflops
resolving degeneracies of screened two electron integrals


calculate QP shifts <psi_nk| G(iteration)W_0 |psi_nk>: iteration 1
writing wavefunctions
entering main loop
----------------------------------------------------


but after that it did not print data. The OUTCAR file finished with error.
as below:

k-point 35 : -0.2222-0.2500 0.0000 plane waves: 3189
k-point 36 : -0.1111-0.2500 0.0000 plane waves: 3198

maximum and minimum number of plane-waves per node : 3239 3136

maximum number of plane-waves: 3239
maximum index in each direction:
IXMAX= 3 IYMAX= 10 IZMAX= 20
IXMIN= -3 IYMIN= -10 IZMIN= -20


-----------------------------------------------------------------------------
| |
| EEEEEEE RRRRRR RRRRRR OOOOOOO RRRRRR ### ### ### |
| E R R R R O O R R ### ### ### |
| E R R R R O O R R ### ### ### |
| EEEEE RRRRRR RRRRRR O O RRRRRR # # # |
| E R R R R O O R R |
| E R R R R O O R R ### ### ### |
| EEEEEEE R R R R OOOOOOO R R ### ### ### |
| |
| The file WXXXX.tmp for XXXX= 1 can not be read. |
| Rerun the GW calculations with LRPA= .TRUE., and do not change |
| ENCUTGW or NBANDS between runs. |
| |
-----------------------------------------------------------------------------

i think i messed-up.
Can you give some hint/suggestion, if the tags and procedure has gone wrong where?
Regards

[marie-therese.huebsch]

Hi SKM,

Great that you are trying so may things. However unfortunately it is not clear to me what you did step by step. In order for me to be able to help, could you please upload the input files and the OUTCAR of each step and include some short instructions. If you cannot remember each step, I am afraid you probably have to start over. Also note that my suggestion was referring to a GW calculation. If you shifted to a BSE calculation, can you please open a new thread referencing this one.

Best regards,
Marie-Therese

[SKM]

Hi Marie,
Thanks for the reply.
the previous response i gave i thought i started a new thread, but unfortunately i responded in the same thread.

by SKM » Sat Sep 04, 2021 10:29 pm

Thank you Marie, for the reply in the thread forum/viewtopic.php?f=4&t=18223#p20768
The query below is continution to the above thread, but different issue

will raise another topic, then.
But unfortunately, i cannot upload POSCAR/OUTCAR files, as they are not published but become open to public once i upload.
In such case, how can i resolve it?
Regards

[ferenc_karsai]

Your input tags look ok.

Most likely the hint lies in the last lines of the error message:
"The file WXXXX.tmp for XXXX= 1 can not be read. |
| Rerun the GW calculations with LRPA= .TRUE., and do not change |
| ENCUTGW or NBANDS between runs"

It looks like you changed some of the parameters between the runs. They have to be compatible.
NBANDS can automatically change if you change the number of cores you run with, regardless if you set it. So best practice is to run the calculation in each step on the same number of cores.

[marie-therese.huebsch]

Hi SKM,

Regarding your other question: In order to discuss calculations, for which you do not want to disclose the results, you can exchange the system/structure that you are interested in with a similar system, that displays the same underlying physics and preferably it should be computationally cheaper. For instance, the replacement should be also a semimetal, if you are interested in semimetals, but the complexity of the system can be reduced. For the purpose of checking the functionality, it is also not necessary to tune all parameters to values such that the results are publishable.

Best regards,
Marie-Therese

[SKM]

Hi Marie,
Thanks for the reply.
OK.
I am attaching input files of exactly same crystal system and with less number of atom types of a published system.
Please let me know what is wrong with these calculations. What needs to be done further.
As my intention is to get GW band structure and Band-gap, and BSE optical band gap.

tar file attached

Regards

[SKM]

hi Admin,

can i have some input, solution from you, based on my attached input files, in previous mail? My work is kept pending for want to resolve this please.

Regards

[SKM]

hi Marie/ Admin,
any help and update on this issue?
Regards

[marie-therese.huebsch]

Dear SKM,

In the attached zip that you provided, there was a readme with the following questions:

1. Are the INCAR tags enough in each step to get the desired files, which will be used as input to second (next step)?

2. The OUTCAR-gw file, at the end, mentions 'entering main loop' but nothing is written after that. Is this normal? or something wrong? form this i get bandgap as 1.47 whereas i supposed to get 2.63 eV (GW bandgap) and the BSE optical band gap of 2.02 eV.

3. The BSE output files, after processing shows ZERO absorption data. Is it because it did not find and file and asks to rerun LOPTICS?

The answers to this is

1. Your input files should be sufficient to produce an optical absorption spectrum and I was able to get one with your script.

2. The OUTCAR-gw file looks fine and shows the fundamental GW band gap of 2.65 eV, the value of 1.47 eV that you mentioned corresponds to the DFT band gap.

3. The files OUTCAR-bse and log-bse also look reasonable and don't request you to rerun LOPTICS=.TRUE.. In principle, this calculation should have written the BSE dielectric function in vasprun.xml. Please provide your vasprun.xml, so that I can see what appears to be a problem, so far this calculation looks fine.

For your calculation, please note:
- the potentials recommended for the GW calculations are labeled `_GW`.
- the unoccupied bands should be calculated in DFT with `ALGO = Exact` (see Practical guide to GW calculations on the VASP Wiki).

For your future post, please note that the questions should be asked in the text and not in the attached file. Also, we ask for your understanding that technical support is offered as a courtesy only. If your question concerns a general workflow or cutting edge research, I advise to address a broader audience, e.g., in From Users for Users. I am sorry your project was pending for such a long time.

Best regards,
Marie-Therese

[SKM]

Thank you Marie, for the reply.
i am replying to your responses, as below:

3. The files OUTCAR-bse and log-bse also look reasonable and don't request you to rerun LOPTICS=.TRUE.. In principle, this calculation should have written the BSE dielectric function in vasprun.xml. Please provide your vasprun.xml, so that I can see what appears to be a problem, so far this calculation looks fine.

1. I was am referring to the ERROR message at initial stages of the run, in same file log-bse as it mentioned an ERROR. with text as " The derivative of the wavefunctions with respect to k (WAVEDER) |
| can not be found. You should redo the groundstate calculations |
| using LOPTICS=.TRUE. in order to write the WAVEDER file. |
| For metals the present setting is however ok.
If i need to check whether the system is 'metal', then do i need to see, if there is NO partial occupancies for all k-points, in simple DFT calculation?
2. Even, if the ERROR can be ignored, if its METALLIC system, but 'Why does it say, WAVEDER file cannot be found, whereas all the files generated from GW run were copied into 'BSE' folder? i am attaching vasprun.xml files of both GW and BSE runs

3. How did you get the fundamental band gap from OUTCAR-gw ? as i understand its difference between its ionization potential (I) and its electron affinity (A),, do we get these two parameters from this file?

Regards

[marie-therese.huebsch]

Hi SKM,

If you have a look in the log-base before the error you will find

Code: Select all

   ----------------------------------------------------------------------------- 
  |                                                                             |    
  |           W    W    AA    RRRRR   N    N  II  N    N   GGGG   !!!           |    
  |           W    W   A  A   R    R  NN   N  II  NN   N  G    G  !!!           |    
  |           W    W  A    A  R    R  N N  N  II  N N  N  G       !!!           |    
  |           W WW W  AAAAAA  RRRRR   N  N N  II  N  N N  G  GGG   !            |    
  |           WW  WW  A    A  R   R   N   NN  II  N   NN  G    G                |    
  |           W    W  A    A  R    R  N    N  II  N    N   GGGG   !!!           |    
  |                                                                             |    
  |      The number of bands has been changed from the values supplied          |    
  |      in the INCAR file. This is a result of running the parallel version.   |
  |      The orbitals not found in the WAVECAR file will be initialized with    |    
  |      random numbers, which is usually adequate. For correlated              |    
  |      calculations, however, you should redo the groundstate calculation.    |    
  |      I found NBANDS    =      120  now  NBANDS  =     288                   |    
  |                                                                             |    
   ----------------------------------------------------------------------------- 

As the warning states, this is a result of how you submit the job. Because I changed the number of k points, I did not see the same issue. So, I am really sorry I overlooked it in your output files!

In the GW calculation, you use the KPAR tag, which is why ncpus=288 does not result in 288 bands. As the KPAR tag is not available for the BSE calculation (and you thus rightly did not specify it in the INCAR_bse), the number of bands is increased. Let me stress that NBANDS needs to be the same throughout the GW and BSE calculation, as I think is clear from the instructions on the Wiki. That is why WAVEDER file written during the GW calculation could not be used in the BSE calculation. Please adjust your settings to get the same value for NBANDS throughout your calculation.

Best regards,
Marie-Therese

[SKM]

Thank you Marie, for the reply.

Regarding KPAR usage is concerned: -
1. Do you mean to say, though i explicitly mentioned NBANDS in GW [INCAR], but because i used KPAR tag as well, the NBANDS got changed during the run?
2. If the answer to above is YES, then i shouldn't have used KPAR in GW as well, like in BSE?
3. also please clarify the following queries i asked.
a) If i need to check whether the system is 'metal', then do i need to see, if there is NO partial occupancies
for all k-points, in simple DFT calculation?

3. How did you get the fundamental band gap from OUTCAR-gw ? as i understand its difference between its ionization potential (I) and its electron affinity (A),, do we get these two parameters from this file?

[SKM]

Hi Marie/Admin,

awaiting your reply please.

Regards

[marie-therese.huebsch]

Dear SKM,

Ok, let me try to paraphrase: You need to run the BSE calculation on fewer CPUs. For your calculation, ncpus=120 is a good choice. My previous response explains the reason.

Ad 2., for GW calculations, you can set the KPAR tag, but not for the BSE calculation.

For a metallic system, there is no finite bandgap. In other words, if there is partial occupancy at any k point, the system is metallic. You can check the density of state written to the DOSCAR file to see whether there are states at the Fermi energy. (If yes, then it is a metal.)

The bandgap can generally be obtained from the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO). Here, molecular orbital refers to the Kohn-Sham (KS) orbitals, i.e., the eigenstates computed by VASP. The eigenenergies and occupancies for each band and each k point is written to the OUTCAR file. Thus, you need to find the energy of the lowest unoccupied state and subtract the energy of the highest occupied KS orbital: LUMO - HOMO. It is probably quicker to write a script than to do this manually.

All the best for your calculation.

Marie-Therese

[SKM]

Fantastic.
Thank you.
i will try to resolve the issues.
Regards