Yambo Community Forum

Dear Davide,

I am trying to calculate the BSE absorption spectrum of a 2D system (with spin-orbit coupling giving a gap of 0.2 eV) on top of the PBE orbitals, skipping the time-consuming GW part.
However, YAMBO gives me no absorption onset! In other words the first set of excitation energies are NEGATIVE, when you diagonalize the BS matrix!

I checked this for other smaller systems having small gaps with the ELK code; and got a nice absorption onset.
I further checked this system with the VASP code, and again got a nice absorption onset.

I further mention that this issue has nothing to do with broadening.

Please find attached the report/log and QE/YAMBO input files in case you want to run it yourself.
The run should be fast, if you reduce the number of k-points and other parameters.

Best wishes,
Martin

Dear Martin,
I can't see the input/report files, just the plot.
Best,
Daniele

I am sorry,

attached the files.

Best,
Martin

And here are attached the PPs.

Best,
Martin

Dear Martin,
Not sure this is the reason of your problem but when dealing with 1 or 2D sampling, integrations over the BZ are incorrect unless you use the RIM (Monte Carlo integration of the Coulomb integrals in the BZ). The problem is not in the divergence in q=0 which is anyway analytically evaluated, but for the small q near zero. The bigger the sampling the bigger is the error, now I do not know with your sampling if the error you are doing is relevant or not. For my experience, this kind of problem produces redshifted spectra.
You can try if activating the RIM will solve the problem.
Best,
Daniele

Dear Daniele,

many thanks for your reply.

I checked it with RIM and still get NO absorption onset. I think the problem should be somewhere else.

In VASP there is no such random integration method as much as I know; and I am still able to get an absorption onset in contrary to YAMBO.

Best wishes,
Martin

Dear Martin,
sincerely I have no idea how VASP treat the BZ integration in case of non-3D sampling.
I'm just saying that how Yambo works if you do not use the RIM you make an error, that may be small or big depending on the sampling itself.
Did you check if the IP spectrum has the right onset? I also imagine you checked convergence parameter.

Daniele

Dear Daniele,

Yes, I checked the IP spectrum, and even the IP spectrum has no absorption onset.

I think absorption onset is something that should be observed even if parameter are not converged.
In VASP I also had no converged parameter, and still was able to see an absorption onset.

Bests,
Martin

Dear Martin,
this is weird! Are you sure that this is not a problem of broadening at least for the IP spectrum. Did you try do reduce it to 0.01eV?
Try also to change the BSEmod from resonant to retarded. In the resonant mode you force the imaginary part to be symmetric with respect the origin (0 energy) and having the first peak very close to zero you are not able to see the onset as you have a smearing of 0.1eV and the imaginary part is forced to raise to reproduce the same peak at negative energies.
Differently, if the IP sepctrum has negative eigenvalues, then something weird is really happening.

Best,

Daniele

Dear Daniele,

many thanks.

The combination of BSE ="retarded" and 0.01 eV broadening indeed helps; however if you look at the plot, part of the onset of the IP spectrum (RED LINE) is still negative!!!
I have no idea how to get rid of this tiny negative absorption onset.

Bests,
Martin