No Absorption with BSE calculation on monolayer MoS2

[Daniele Varsano]

Dear Reza,

as you can see the IP spectrum onset correspond to the gap you posted before, it seems to me perfectly reasonable.
About the q=0 limit, this is the optical spectrum and if you are interested in optic you can set in input:

Code: Select all

% BSEQptR
   1 | 1 |                           # [BSK] Transferred momenta range
%

and this will save you a lot of computer time.
You calculated instead the response for all the 26 q points, and you should have 26 outputs, besides iq=1 (q0 limit) the others does not describe optical excitations but excitations with transferred finite momentum, which are observable e.g. in EELS experiments.

Best,
Daniele

[Reza_Reza]

Dear Daniele,

Thank you very much for your response.
I only have ambiguity. I don't understand the concept of iq=1 (q0 limit) here.
Also, what is the difference between the spectra when I plot u 1:2 and u 1:4 ?

Best,

Reza

[Daniele Varsano]

Dear Reza,

In the BSE output file, you have the BSE results (column 2 and 3) and also the Independent Particle spectra (column 3 and 4).
In Yambo each q points is associated to an index, the list of q points and corresponding indexes are reported in the report file. The q=0,0,0 is always index 1.
If you want to calculate optics, then you want to calculate the response at q=0 limit (vertical excitations) so you need to set iq=1.
You can have a look at the Yambo cheatsheet (slide 11a):
https://www.yambo-code.eu/wiki/images/c ... et-5.0.pdf

If these concepts are not crystal clear, I suggest you to go step-by-step looking to the tutorials. In the Yambo wiki you can find several og them.
For instance, here: https://www.yambo-code.eu/wiki/index.ph ... #Tutorials
you can find a tour from Independent particle calculations, GW and Bethe Salpeter.

Best,
Daniele

[Reza_Reza]

Dear Daniele,

I am trying to compute the optical absorption for a small band gap system (~0.3 eV). However, I have observed strong fluctuations in the system when attempting to converge the k-points. I have attached the optical absorption data with different k-points. Additionally, when I computed the exciton wave function, it appears that it has not converged, as the exciton wave function is expanded and delocalized.

Do you have any ideas on how to deal with this problem?

Best regards,

Reza

[Daniele Varsano]

Dear Reza,

actually, it is not easy to spot what's happening here.
Here two suggestions to look at:

1) How does compare the IP spectrum for different sampling (4th column in the output)
2) Can you repeat the calculations (including the screening part) without setting the coulomb cutoff, e.g. CUTGeo="none" ?

Best,

Daniele

[Reza_Reza]

Dear Daniele,

1. I plotted the second column since it is a low-dimensional system.
2. Yes, I can repeat the calculation without setting the Coulomb cutoff. Because I am computing the BSE on top of GW, Do I need to repeat the GW calculation without setting the Coulomb cutoff too?

Best,

Reza

[Daniele Varsano]

Dear Reza,

1) Independently of the dimension, the structure of the output file is
Energy, Im (eps), Re(eps), Im(eps0), Re(Eps0).

My suggestion was to compare the Im(eps0) from the different sampling.

2) You can avoid recomputing the GW, but you should to recompute the screening (ndb.em1s), i.e. do not use the ndb.pp calculated for the GW step as it was computed using the cutoff.

Best,
Daniele

[Reza_Reza]

Dear Daniele,

1) I plotted the optical absorption (Energy (column 1) and Im(alpha_o) column 4) in my output files for three different k-points= 120, 150, and 200

Best,

Reza

[Daniele Varsano]

Dear Reza,

the IP are reasonable, and looking at the BSE results according to your output also they seem to be ok (see attachment) and different from what you reported in your previous post.

Best,
Daniele

plot.png

[Reza_Reza]

Dear Daniele,

Thanks a lot for the help. I think I found the problem of optical absorption. I used two machines to compute the optical absorption with different k-points. To ensure the consistency of my results, I have used the same input file and database to compute the band structure for the same system on two different machines (one of them is IRENE). However, the results are different, particularly in the size of the band gap.

Best regards,
Reza