The strange position of exciton absortion peak

[Yuan Pang]

Dear developer,

Nearly, I caculate the absorption spectrum of 2D semiconductor by GW+BSE. However, I find a strange result which exhibits the absorption peak in the area of negitive energy, e.g. the exciton excitation energy is negitive. Therefore, I don't know how to explain this calculation. Please see attachment about the detials of the cacluation.

Best,
Pang

[Daniele Varsano]

Dear Yuan,
can you post the report file of your calculation? otherwise, it is hard to understand if there is a problem.

Some notes:
*you are using quite an old version of the code. I suggest you to update to a newer release.
*RandQpts= 100000 These are really few points for Montecarlo calculations, usual values are 1 million-3 million

Best,

Daniele

[Yuan Pang]

Dear Daniele,

Thank you for your reply, the report files have been uploaded as attachments.

Best,

Pang

[Daniele Varsano]

Dear Pang,
you have a redshifted spectrum because the QP particles have not been applied to the BSE Hamiltonian.
From the GW report file I can see that the ndb.QP file has been written in gwbisseteout directory, while the BSE input points to a different directory where the ndb.QP database is not present:

KfnQPdb= "E < ./bse_gwout/ndb.QP"

Modifying it in:

Code: Select all

 KfnQPdb= "E < ./gwbisseteout/ndb.QP"

should fix the problem.

Best,
Daniele

[Yuan Pang]

Dear Daniele,

Thank for you reply.
I get it.
I have copied the ndb.QP file into the bse_gwout folder before doing the BSE calculation.

Best,

Pang

[Yuan Pang]

Dear Daniele,

I have obtain the abosorption spectum of the system mentioned last time by considered QP particles. As a result,I found that the first four excitons are approximately considered to be energy degenerate. I calculated the exciton wave functions of the first four excitons and found that they are uniformly distributed in real space rather than local, as shown in figure of attachment. Is it normal that the exciton binding energy of about 1 eV corresponds to such a discrete wave function?

The related files have been unloaded as attachments.


Best,

Pang

[Daniele Varsano]

Dear Jinan,
it is possible if the excitation is very localized in k space. You can check it by looking at the weights of the single-particle transitions.
You can find in this tutorial how to do that:
http://www.yambo-code.org/wiki/index.ph ... 021_school

Anyway, looking at your input/report I noticed that you are using a cutoff box size (60au) larger than your supercell (59.6qu).
I suggest you check it as it can cause severe problems as the cutoff coulomb potential is not reliable. Essentially you should repeat the calculations for both GW quasiparticles and BSE recalculating the cutoff potential and the screening potential.

Net I would also check the convergence wrt the k points, I have the impression you are using a very coarse grid. Convergence wrt k point grids in 2D is systems is very delicate (both for GW and BSE) and it has a large impact on the energy and localization of the exciton.

Best,
Daniele

[Yuan Pang]

Dear Prof. Daniele


Thank you for your reply!


The same weights of the single-particle transitions for first four excitons as following:

# Electron-Hole pairs that contribute to Excitonic State 1 for iq=1 more than 5.000000%
#
# K-point [iku] Weight
# : 0.00000000 0.00000000 0.00000000 1.00000000
#
# Band_V Band_C Kv-q ibz Symm_kv Kc ibz Symm_kc Weight Energy
#
68.0000000 69.0000000 1.00000000 1.00000000 1.00000000 1.00000000 0.997540998 0.220522875
67.0000000 70.0000000 1.00000000 1.00000000 1.00000000 1.00000000 0.997540484 0.220522875
67.0000000 69.0000000 1.00000000 1.00000000 1.00000000 1.00000000 0.997514959 0.220522875
68.0000000 70.0000000 1.00000000 1.00000000 1.00000000 1.00000000 0.997514476 0.220522875

The localization of the exciton weights in k space corresponds to the delocalization of exciton wave function in the real space. Can I understand that more precise k-grid will not make this result better? In other words, even if the k-grid is added, the exciton wave function in real space is still delocalized.

Best,

Pang

[Daniele Varsano]

Dear Yuan,

Can I understand that more precise k-grid will not make this result better?

I would say the opposite. With the present sampling, only the gamma point contribute to the exciton (thus the exc-wf is extremely delocalized in real space), and it is possible you do not have enough resolution in the BZ to include transitions from other k points.
I suggest you check your convergence wrt the k point grid.

Best,
Daniele

[Yuan Pang]

Dear Prof. Daniele


Thank you for your reply!

I will check and modify all the issues mentioned above.

Best,

Pang