GW+BSE, Plotting exciton wavefunction

[xinwuchn]

Dear all,

I am using the newest version to try to plot exciton wavefunction. https://www.vasp.at/wiki/index.php/Plot ... vefunction

I follow the tutorial with the BSEHOLE keyword in INCAR on the BSE calculation step, but not CHG.xxx file output. The following is my INCAR and POSCAR.

INCAR:
ISTART=1
ICHARG=11
ISMEAR=0
SIGMA=0.05
EDIFF=1E-08
IVDW = 12

NSW=0
NBANDS=66
PRECFOCK=FAST
LOPTICS=T
LORBIT = 11
ENCUT = 420
ALGO=BSE

LREAL=Auto
NELM=1
NEDOS=2000
NOMEGA=50
ENCUTGW=100
LSPECTRAL=T

NBANDSO = 12
NBANDSV = 12
OMEGAMAX = 8

BSEHOLE = 0 2.32455 16

POSCAR:
InSe
1.000000000000000
4.0262388326039815 0.0000000000000001 0.0000000000000000
-2.0131194163019908 3.4868251107744146 0.0000000000000000
0.0000000000000000 0.0000000000000000 30.0000000000000000
In Se
2 2
Direct
0.6666666700027335 0.3333333300097578 0.6918175254852116
0.6666666700027335 0.3333333300097578 0.5986565311814559
0.3333333299951207 0.6666666699902422 0.5553877760915944
0.3333333299951207 0.6666666699902422 0.7350862805750731

How can I output the first bright exciton in real space? If u need other information, pls tell me.

Best,
Xin WU

[alexey.tal]

Dear Xin WU,

I see that you have not specified the number of eigenvectors/wavefunctions that should be written out, i.e., NBSEEIG.
Also, you should provide the coordinates of the hole BSEHOLE in fractional coordinates.
If you would like to plot the first bright exciton you can look at <varray name="opticaltransitions"> block in vasprun.xml and identify the first transition with non-zero strength and make sure that your NBSEEIG is large enough to include this state.

[xinwuchn]

Thanks so much for your reply.

The follow is the
<varray name="opticaltransitions">
block .

<varray name="opticaltransitions" >
<v> 2.5761 0.3043 </v>
<v> 2.6701 28.0418 </v>
<v> 2.6701 28.0417 </v>
<v> 2.7566 0.0151 </v>
<v> 2.7566 0.0151 </v>
<v> 2.9042 0.3921 </v>
<v> 2.9170 0.0001 </v>
<v> 2.9170 0.0001 </v>
<v> 3.0113 0.0000 </v>
<v> 3.0113 0.0000 </v>
<v> 3.0419 0.0000 </v>
<v> 3.1584 37.9590 </v>
<v> 3.1584 37.9589 </v>
<v> 3.1923 0.0000 </v>

What quantities do these two columns represent? If the first column typically represents the energy of the transition (in eV), and the second column represents the oscillator strength or the transition probability.

So, the first bright exciton is 2.576, right? So, the NBSEEIG = 1?

I have another problem if every visual exciton function needs to be judged from vasprun.xml. Then the process is to first perform a BSE level calculation, and then check vasprun.xml, and then perform another BSE level calculation, which will be particularly time-consuming and wasteful. Is there a better way to do this?

Thanks again for your help!

Best,
Xin WU

[alexey.tal]

Code: Select all

What quantities do these two columns represent? If the first column typically represents the energy of the transition (in eV), and the second column represents the oscillator strength or the transition probability.

Exactly. The first column is energy and the second one is the oscillator strength.

Code: Select all

So, the first bright exciton is 2.576, right? So, the NBSEEIG = 1?

In this case I would say that the first bright exciton is at 2.6701 eV because the oscillator strength of the first transition is less than 1% of maximum intensity and such transitions should not be considered as bright. Also, the transition at 2.6701 eV is doubly degenerate and thus one should add the charge densities of both excitonic states #2 and #3 together, i.e., CHG.2 and CHG.3. Hence, NBSEEIG should be set to 3 or higher.

Code: Select all

I have another problem if every visual exciton function needs to be judged from vasprun.xml. Then the process is to first perform a BSE level calculation, and then check vasprun.xml, and then perform another BSE level calculation, which will be particularly time-consuming and wasteful. Is there a better way to do this?

That is a good point. Of course, the transitions which violate the dipole selection rules will be dark, so one can get some idea of how many state are likely to be dark by considering what states constitute the bands at the valence band maximum and the conduction band minimum. Although, in general I would say that one has to perform a BSE calculation to identify the first bright transition, but that can be done in the convergence tests with less dense k-points mesh and fewer bands. Another option would be to set NBSEEIG to some reasonably large value which would certainly include the first bright transition, although that would add some overhead to the calculation.

[xinwuchn]

Thanks so much for your quick response. I totally agree with you.

1. ​The transition at 2.6701 eV is doubly degenerate and thus I should add the charge densities of both excitonic states #2 and #3 together.

So, if NBSEEIG = 3, vasp would output CHG.3? Does this CHG.3 file contain excitonic states #2 and #3 together?

Or if NBSEEIG = 3, vasp will output CHG.1, CHG.2, CHG.3 three files, and then I need to manually add the wave function in VESTA?

2. I couldn't agree more with what you said about doing a low-precision BSE level calculation to judge the first bright transition. It's very useful.

3. I have one more small question about the position of the fixed particle.

In my INCAR, I provided "BSEHOLE = 0 2.32455 16". And then the OUTCAR contains the following contents:

$ grep "hole position" OUTCAR
  hole position is fixed at: 0.000000 0.312500 0.000000

So what do these three coordinates in OUTCAR mean?

If I fix the hole position in the vacuum layer, is that possible? Let's say my hole is supplied by some other way. How do different fixed hole positions affect the exciton function. What I mean is that my holes may be provided by other heterojunction layers in the actual case, but the computation of heterojunction is very difficult due to the huge amount of BSE computation.

It should be noted that I am currently dealing with a monolayer InSe system, and you may need its INCAR.

POSCAR file written by Ovito 3.0.0-dev34
1.000000000000000
4.0262388326039815 0.0000000000000001 0.0000000000000000
-2.0131194163019908 3.4868251107744146 0.0000000000000000
0.0000000000000000 0.0000000000000000 30.0000000000000000
In Se
2 2
Direct
0.6666666700027335 0.3333333300097578 0.6918175254852116
0.6666666700027335 0.3333333300097578 0.5986565311814559
0.3333333299951207 0.6666666699902422 0.5553877760915944
0.3333333299951207 0.6666666699902422 0.7350862805750731

Thank you very much for your help. It really helped me understand a lot.

Best,
Xin WU

[alexey.tal]

Code: Select all

1. ​The transition at 2.6701 eV is doubly degenerate and thus I should add the charge densities of both excitonic states #2 and #3 together.

So, if NBSEEIG = 3, vasp would output CHG.3? Does this CHG.3 file contain excitonic states #2 and #3 together?

Or if NBSEEIG = 3, vasp will output CHG.1, CHG.2, CHG.3 three files, and then I need to manually add the wave function in VESTA?

If NBSEEIG = 3 VASP writes out the first three excitonic states, i.e., CHG.1, CHG.2, CHG.3. You can then visualize the double degenerate exciton charge density either by manually adding two CHG files or in VESTA: Edit->Edit Data->Volumetric data->Import.

Code: Select all

3. I have one more small question about the position of the fixed particle.

[code]In my INCAR, I provided "BSEHOLE = 0 2.32455 16". And then the OUTCAR contains the following contents:

$ grep "hole position" OUTCAR
  hole position is fixed at: 0.000000 0.312500 0.000000

You have to provide the position of the hole in fractional coordinates (see here), i.e., the same way you provide positions of your atoms in POSCAR:

pic.png

As we describe wave functions on a regular grid, we can only fix the position of the hole/electron at a point which is included in our grid. Hence, VASP will fix the hole position at the grid point closest to the coordinate provided in BSEHOLE/BSEELECTRON and report the coordinates of that point in OUTCAR.

If I fix the hole position in the vacuum layer, is that possible? Let's say my hole is supplied by some other way. How do different fixed hole positions affect the exciton function. What I mean is that my holes may be provided by other heterojunction layers in the actual case, but the computation of heterojunction is very difficult due to the huge amount of BSE computation.

Usually you would want to fix the position of the hole somewhere near or at the atom from which the electron would be excited as the hole is typically more localized, but not always. BSE calculations are performed based on the ground state electronic structure, i.e., we consider the process of taking an electron in the ground state structure and exciting it into unoccupied states, so the source of the hole should be a part of the calculation.

[alok_shukla1]

This thread has been incredibly helpful for those working on calculating and plotting excitonic wavefunctions using BSE calculations. Thank you so much for the guidance provided here.
I want to add my query related to the same problem I am facing. I followed the instructions in the manual https://www.vasp.at/wiki/index.php/Plot ... vefunction and carefully included the two essential tags (BSEHOLE & NBSEEIG) required for generating the CHG.XX files. However, after completing my calculations (which ended without any errors), I found that these special CHG.XX files were not generated. Below are my INCAR and POSCAR files for an h-BN system, following the example provided on the VASP wiki. Could you please suggest where I might be going wrong? Is there something important missing in my INCAR that is necessary for generating these special CHG.XX files?"

INCAR :

PREC = Normal
ALGO = BSE
ANTIRES = 0
ENCUT = 550
ISTART = 1
ISMEAR = 0 ; SIGMA = 0.05
NBANDS = 80
NBANDSO = 8
NBANDSV = 8
OMEGAMAX = 15
PRECFOCK = Normal
BSEHOLE = 0.0700 0.0000 0.0000
NBSEEIG = 10
LCHARG = .TRUE.

POSCAR :

h-BN
1.00000000000000
2.5123896950807785 -0.0000000000001883 -0.0000000000000000
-1.2561948475907203 2.1757933001705934 0.0000000000000000
-0.0000000000000000 0.0000000000000000 7.4831498153853175
B N
1 1
Direct
0.0000000000000000 0.0000000000000000 -0.0000000000000000
0.3333333431843499 0.6666666863420758 0.0000000000000000

0.00000000E+00 0.00000000E+00 0.00000000E+00
0.00000000E+00 0.00000000E+00 0.00000000E+00

I tried calculation for various hole positions (BSEHOLE), but nothing worked !!

[alexey.tal]

Could you please provide all the relevant output files? Also, this feature was only introduced in VASP 6.4.3, so if are using an older version that could explain the issue.

[alok_shukla1]

I am performing these calculations in the VASP 5.4.4 version. So, these calculations can not be done using version older than VASP 6.4.3 ?

[alexey.tal]

Unfortunately, this feature is only available in VASP 6.4.3+.