About Nondiagonal Supercell and DFPT

[sitangshu]

Dear Sir,

I came across a very recent paper by Professor Claudio A. performing DFPT using a non-diagonal supercell method. The section where the fractional k-points are mapped seems very interesting and appealing. I thus took some time to understand first Professor Monserrat's earlier work on non-diagonal supercell. Using the corresponding .f90 program I was able to get the fractional IBZ (I hope I did correct).
Since there is no tutorial on this, I was wondering how to map now this supercell to get the DFPT results. I would greatly appreciate if you make me understand this process flow.

Thanks and regards,
Sitangshu

[claudio]

Dear Sitangshu

it is not completely clear to me your question. You can generate non-diagonal supercell that contains a particular phonon mode,
and the paper of B Monserrat presents a strategy to generate the smallest non-diagonal supercell for a given q-vector.

Usually, after I generate the non-diagonal supercell, I'm interested in, I use standard DFTP of QuantumEspresso
with only gamma-point to get the phonon modes.
I know that in principle you can map phonon modes from the unit cell to the non-diagonal supercell, but I never wrote
a code to do this, so I prefer to recalculate them.

If it can be useful to you, you can use my python code to generate non-diagonal supercells,
https://github.com/attacc/QEplayground

best regards
Claudio Attaccalite

[sitangshu]

Dear Sir, Thank you for your git link. Actually, I was referring to your paper :https://arxiv.org/pdf/1807.11797.pdf. The para just before the result section seems to be very appealing to us in order to study BSE for fictionalized materials.
In-fact we did some work on T-dependent spectra of ML hBN located here: https://arxiv.org/pdf/1811.01695.pdf. I hope the results should be correct.
I am bit confused, or you may say apprehend to solve supercell DFPT+BSE... I don't know if the double-grid part will map to supercell DFPT calculation (using "iu") or should I not use double-grid at all for such cases... This is the reason, I never tried this task.
By going through your paper it seems that this could be doable now. That's why I want to learn your methodology.
Let me take some time to go through the link.

Regards,
Sitangshu

[claudio]

Dear Sitangshu

the methodology we use is based on finite differences calculations of electron-phonon coupling.
In practice we generate supercells that contain the phonon modes we are interested in, then we move atoms
and calculate the dielectric constant, at the end we average on the different configurations.

Please have a look at these papers:
https://arxiv.org/abs/1512.06377
http://iopscience.iop.org/article/10.10 ... aa737/meta

and also this other one on h-BN, where there are much more details, and results similar to ours.
https://arxiv.org/abs/1810.08976

Regarding your results, I think they are correct. The method we use it is different, we calculate electron-phonon coupling
by moving atoms and not by direct calculation with QuantumEspresso.

This allows us to include effects not included in standard perturbation theory as for example the phonon replica,
but the price to pay it is to use large supercells

best
Claudio

[sitangshu]

Thank you Sir for your suggestions. I am going through the methodology. Let me see...

Regards,
Sitangshu