Real time dynamics with Lumen

[kle]

Dear developers,

I have a couple of questions as I am trying to explore the real-time BSE to simulate the transient reflectance/absorption spectra, similar to Fig.7(b) in the PRB https://journals.aps.org/prb/abstract/1 ... .93.195205.

Is this possible with the current "yambo_nl" on lumen gitlab repo? And does it work with QE's LSDA+U?
I have run this tutorial and got the absorbance vs energy at a delayed probe pulse.
https://www.lumen-code.org/wiki/index.p ... _and_Probe

To simulate the real-time dynamics, I supposed I would have to generate the time profile of the external field according to:
https://www.lumen-code.org/wiki/index.p ... _from_file and use that for the "yambo_nl" run. Would this be a correct way to do this?

Best,
Khoa

[Davide]

Dear Khoa,

1) For the results of the PRB you can follow this tutorial:
https://www.lumen-code.org/wiki/index.p ... lk_silicon
This is based on real time propagation with pump + GW-BSE with neq occupations.

It is a good approach for the cases where carriers are injected in the code, and coherences can be neglected. There is also an ongoing project to extract carriers from perturbo dynamics and use them inside the GW+BSE step. It could be also combined with recent developments in pw.x using constraint occupations in my opinion.

The approach is based on the yambo_rt executable, it works with spin-polarized systems as well as with SOC.
If you do that on top of LDA+U, better to use the covariant approach for the dipoles (assuming that you jave enough kpts to converge them). Also, if you compute GW qp correction there will be a double counting of the U term (which is in part present also in the Sigma_GW)


2) On the pump and probe tutorial.
This is based on real time propagation with pump + real time propagation with pump and probe.

It works better instead if you have a laser pulse which generates excitonics with low density, or if you want to explore coherent effects in the transient absorption (for example Dynamical Frantz-Keldish). It neglects the update of the screening, which is also reasonable if the pump inject excitons. This approach is based on the yambo_nl executable, and at present it is coded without spin and with SOC, but not yet for spin-polarized simulations.


3) For the pump profile you can either use one of the "standard" fields provided by the code (QSSIN is a guassian profile convoluted with a sin function), or generate a custmo une as discussed in the tutorial. It depends on your needs.

Best,
D.