Band ordering in GW eigenvalues

[ivona]

Dear developers,

I started convergence tests for the band gap of my systems to calculate the GW corrected eigenvalues that will enter the BSE calculation. After having converged the optical band gap I calculated the remaining bands for other k-points in my system.
I then realised following issue:
At some high symmetry points some bands seem to experience quite some large destabilisation when i account for quasiparticle corrections while the valence band does not change as much. This then changes my band ordering at parts of the BZ. However yambo does not change the band labeling of the system.
This makes it particularly difficult to converge the results when I look at the band gap. Is this a known issue or did I make a mistake in my calculation (I would not expect this strong change in the bands anyways).

I have made a screenshot of the bands and I have highlighted the position where the bands invert. Have you ever accounted this issue before?

Thanks,
Ivona

[Daniele Varsano]

Dear Ivona,
the fact that the labelling it is not changed it is not a problem as far as all the relevant bands are included in the BSE calculations, anyway I agree with you that these huge corrections are somehow unexpected. It seems to me that it happens in correspondence and near degenerations. It is not easy to understand if this is reasonable or an artefact. It may happens that you experience very different corrections if states have very different character (let's say localisation) and maybe you can inspect the KS degenerate wfs at high symmetry points if this is the case. If you post here your report files we can have a look if something odd is happened in your calculations, also to look at more carefully at other occupied bands, in the plot t seems you show more GW(4?) bands than KS (2?), but I can also see some red spot on the blue one so the picture it is not totally clear.

Finally is this plot obtained by interpolation of bands?, could be a problem of the interpolator?

Best,
Daniele

[ivona]

Dear Daniele,

thanks for your response. This strange behaviour indeed happens at high symmetry-points (where I would expect the wf character not to be to be too different to the band edges), however I am just really surprised that the lower lying states are being modified so much. I have decided to use the PBE eigenvalues with a scissor correction for now (funnily the optical gap with qp correction and el-hole correlation is in superb agreement with the low-temperature transmission experiment), however I will also like to try to converge the low lying bands at the weird points with much finer convergence criteria. I will let you know if it arises from too coarse convergence parameters. Perhaps the convergence criteria are strongly k-point dependent.
I did use the interpolator but I can already see in the eigenvalue output (before I interpolate) that the quasiparticle corrections for the low-lying bands are very large at the given k-points.
Also, the plot only shows GW bands, I only highlighted the low-lying bands that are being destabilized and "slide" above the valence band of the system (red). I have posted the KS band structure below as a comparison.

The report file is too large to be posted here. Can I send you a personal message instead?

Best,
Ivona

[Daniele Varsano]

Dear Ivona,
you can upload the report file using the attachments button here below, what you need to do is to rename it with an allowed suffix e.g. .txt or .zip

Best,
Daniele

[ivona]

I tried this but it seems to be too large regardless.
So sorry for the inconvenience

Best,
Ivona

[Daniele Varsano]

Dear Ivona, can you split it in different files, or zip it?

Daniele

[ivona]

Dear Daniele,

Unfortunately the zip was too large as well. Here the split into different files. Sorry for the numerous messages.

[ivona]

More to come

[ivona]

This is all. Thank you so much for your help

[Daniele Varsano]

Dear Ivona,
I had just a quick look and seems to me that the corrections for the top valence bands are ok, the degeneracy it is not broken (see attached plot).
Where do these wiggles come from? deeper bands?
I had a look also to the bands 74 wrt 76, here I can see some inversion, but not that large as you are observing in the band plot. Probably it needs a deeper investigation, but I cannot see nothing serious from the report. Also the Z factor seems to be reasonable. I have the impression that the interpolation didn't work correctly, but I can be wrong as it needs a deeper inspection.

Best,
Daniele

ks_vs_gw.png

ks_vs_gw2.png