Page 1 of 1

Oscillator strength from inversion

Posted: Fri Nov 05, 2021 11:00 pm
by Bruno
Hello,

I'm aware that it's possible to extract the oscillator strength of excitons solving the BSE via diagonalization. I would like to know with there is an implemented way of extracting these oscillator str. from the inversion method. If it is not, is it possible to use an alternative version of the double grid method that is compatible with the direct diagonalization ?

Regards,

Re: Oscillator strength from inversion

Posted: Sat Nov 06, 2021 10:25 pm
by Davide Sangalli
Dear Bruno,
for the double grid, we are currently working on an implementation via the Haydock solver.
It is not yet released, but it will be soon, hopefully:
https://arxiv.org/abs/2108.11174
The oscillator strength can be extracted from the plot of the dielectric function. If you need to resolve a specific exciton you can try to use a very small smearing parameter.

Instead, the excitonic wave-function can only be extracted via diagonalization.
You may consider using the slepc solver to this end: http://www.yambo-code.org/wiki/index.ph ... ver:_SLEPC
Please notice that yambo needs to be compiled with

Code: Select all

--enable-slepc-linalg
Best,
D

Re: Oscillator strength from inversion

Posted: Sun Nov 07, 2021 11:25 am
by Bruno
Hello,

Thank you very much for your reply, I'll look into the slepc solver. Just a small question, how is it compared to the diagonalization solver? In terms of memory consumption for example, is it similar?

Regards,

Re: Oscillator strength from inversion

Posted: Sun Nov 07, 2021 11:45 am
by Daniele Varsano
Dear Bruno,
the slepc solver being an iterative method should be faster and less memory consuming than the full diagonalization.
the drawback is that you have access to a limited number of eigenvectors which also affect the resulting spectrum for energies larger than the energy of the last calculated excited state, that can be anyway calculated using the haydock solver.

Best,
Daniele

Re: Oscillator strength from inversion

Posted: Sun Nov 07, 2021 11:58 am
by Bruno
Very interesting! Thank you very much. I'll look into it.

Regards,