RPA calculation hangs at dipoles

[vjhalani]

Hi all,

Using Yambo 4.4, I am trying to do an RPA calculation (yambo -o c) on GaAs on pretty fine grids, and after a certain density, I'm finding that the calculation gets stuck after reading the kb_pp* files and is unable to write the dipole databases. This happens at a 24^3 grid and 30^3 grids, but curiously, this problem does not happen when doing BSE (yambo -b -o b -k sex) and I've done full BSE calculations up to 30^3. We can't do BSE on any finer grids, but we'd like to check the RPA spectrum (which should be much less expensive) on finer grids to estimate convergence of the excitons. I've attached the input file and one of the log files (as you can see it ends at about 2 minutes, but the calculation will go until the walltime of one hour). The report file is too large to attach in its entirety, but I attach a version with the k-points and occupations removed.

Thanks you!

[Davide Sangalli]

Thank you Vatsal for reporting this issue.

Can you also attach the reports and logs of the succesfull BSE run ?
Recently I've seen similar problems and were due to missing or incomplete ns.kb_pwscf_fragment_* files.

Moreover, if the problem are the dipoles you can try one of the alternative approaches to compute them as described in our recent yambo paper:
https://iopscience.iop.org/article/10.1 ... 48X/ab15d0

I'd suggest the covariant dipoles.
Just add to the input file

Code: Select all

DipApproach="Covariant"

Best,
D.

[vjhalani]

Ciao Davide,

Sure, I've attached them (I had to compress them since the l-* file is too large). It seems that it gets stuck after kb_pp_pwscf_fragment_412, while the last one in the BSE is *_413. I checked the SAVE/ and ns.kb_pp_pwscf_fragment_413 is there.

I don't know if this could have an affect, but at one point while troubleshooting this, I switched to the official 4.3 release and tried an RPA calculation, and had the same issue. Is it possible that it affected the SAVE/ and there is some compatibility problem?

Grazie,
Vatsal

[vjhalani]

Ciao,

As an added note, the covariant dipole approach resolves the issue!

Grazie,
Vatsal