Page 1 of 1

NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Fri Sep 01, 2023 4:45 am
by Gao Siyuan
Dear developers,
I'm following the tutorial to obtain the quasi-particle band structure with yambo-5.1.2. There are on errors presented in my log files, but the values of (E-Eo) and Sc(Eo) are NaN when I set "DysSolver" to "n" or "s".
The HF quasi-particle correction calculations are normal. But X_Q_1 in ndb.pp_fragment_1 file are all "NaN".
I attached my input and report files. How can I fix it?
Thanks for your generous help!

Best,
Siyuan

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Thu Sep 07, 2023 5:11 pm
by Daniele Varsano
Dear Gao,
we will try to reproduce your problem, I suspect that there could be something related to the compilation, can you post also your config.log?

Best,
Daniele

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Fri Sep 08, 2023 9:39 am
by Davide Sangalli
Dear Siyuan,
what about the ndb.dipoles. Does it contain NaNs as well?

Best,
D.

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Fri Sep 08, 2023 11:43 am
by Daniele Varsano
Dear Gao,
Can you download the new release (5.2) and see if the problem persists.

Best,

Daniele

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Sat Sep 09, 2023 7:07 am
by Gao Siyuan
Daniele Varsano wrote: Thu Sep 07, 2023 5:11 pm Dear Gao,
we will try to reproduce your problem, I suspect that there could be something related to the compilation, can you post also your config.log?

Best,
Daniele
Dear Daniele,
Thanks for reply, here's config.log. And I'll try v5.2.

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Sat Sep 09, 2023 7:17 am
by Gao Siyuan
Davide Sangalli wrote: Fri Sep 08, 2023 9:39 am Dear Siyuan,
what about the ndb.dipoles. Does it contain NaNs as well?

Best,
D.
Dear Davide
Thanks for your reply. The ndb.dipoles does contained several NaNs.

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Thu Oct 05, 2023 9:10 am
by Gao Siyuan
I‘ve run a tutorial hBN again using Yambo5.2.0, but it still cause NaN.

Re: NaN in quasiparticle energies calculation with yambo-5.1.2

Posted: Thu Oct 05, 2023 9:57 am
by Davide Sangalli
The ndb.dipoles does contained several NaNs.
So the issue is surely there.
The operation done to generate the dipoles are simple linear algebra operations.
When NaNs are there, usually it is due to a problem with the linking of blas/lapack libraries.

Can you try to recompile yambo using the flag

Code: Select all

--enable-int-linalg
when running the configure?

It will download and compile the blas and lapack library together with the code.

Best,
D.