NaN E-E0 result for Mg

Concerns issues with computing quasiparticle corrections to the DFT eigenvalues - i.e., the self-energy within the GW approximation (-g n), or considering the Hartree-Fock exchange only (-x)

Moderators: Davide Sangalli, andrea marini, Daniele Varsano, andrea.ferretti, myrta gruning

burkzdemir
Posts: 45
Joined: Tue Sep 09, 2014 7:57 pm

Re: NaN E-E0 result for Mg

Post by burkzdemir » Fri Dec 13, 2019 10:21 am

Mo pseudopot attached.
You do not have the required permissions to view the files attached to this post.
Burak Ozdemir
Post-doc,
University of Ostrava, Ostrava, Czech Republic

burkzdemir
Posts: 45
Joined: Tue Sep 09, 2014 7:57 pm

Re: NaN E-E0 result for Mg

Post by burkzdemir » Fri Dec 13, 2019 10:22 am

S pseudopot attached.
You do not have the required permissions to view the files attached to this post.
Burak Ozdemir
Post-doc,
University of Ostrava, Ostrava, Czech Republic

User avatar
Daniele Varsano
Posts: 2149
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: NaN E-E0 result for Mg

Post by Daniele Varsano » Sun Dec 15, 2019 6:52 pm

Dear Burak,
I have reproduced your problem. In the next days, we will investigate what is causing the problem.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

User avatar
Daniele Varsano
Posts: 2149
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: NaN E-E0 result for Mg

Post by Daniele Varsano » Mon Dec 16, 2019 7:06 pm

Dear Burak,
I've inspected a bit your problem, the problem was present for the screening of some q points and then propagated in the whole GW calculation. Even if not 100% sure, the problem was caused by the occupation numbers obtained with the use of very large smearing. Repeating the ground state calculations by using

Code: Select all

occupations='smearing', smearing='gaussian', degauss=0.001,
it seems to run smoothly. Actually I've not performed a full GW calculation but I've seen that the problem disappeared in the screening of one of the q points that were previously failing. I encourage you to repeat the GW calculation after computing the ground state (scf and nscf) with the smearing set as above. If you still encounter problems, please report.

Not related to your problem: Note that in your GW input the size of the screening was set to 1RL, note that this is a parameter that needs to be converged.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

User avatar
andrea marini
Posts: 445
Joined: Mon Mar 16, 2009 4:27 pm
Contact:

Re: NaN E-E0 result for Mg

Post by andrea marini » Tue Dec 17, 2019 8:27 am

Dear Burak,

if, as Daniele reported, the NaN is not a Yambo bug then...that it is a physical NaN.

Are you sure the PPA works for your system? I encourage you to go trough the theory and see where the NaN could come from.

In the early days of Yambo, the code was applied to metallic systems where, indeed, the PPA fails because of the appearance of low energy plasmonic states. Is this your case also?

Andrea
Andrea MARINI
Istituto di Struttura della Materia, CNR, (Italy)

burkzdemir
Posts: 45
Joined: Tue Sep 09, 2014 7:57 pm

Re: NaN E-E0 result for Mg

Post by burkzdemir » Wed Dec 18, 2019 11:45 pm

Dear Andrea,

Yes, indeed the system is metallic with low plasma frequency.

Regards,
Burak Ozdemir
Post-doc,
University of Ostrava, Ostrava, Czech Republic

User avatar
Daniele Varsano
Posts: 2149
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: NaN E-E0 result for Mg

Post by Daniele Varsano » Thu Dec 19, 2019 6:32 pm

Dear Burak,
note anyway that the NaN appeared in the evaluation of the dielectric function, so before evaluating the plasmon-pole coefficients.
You can have a try to calculate QP correction using the plasmon-pole model by considering reducing your smearing parameter in QE, anyway, as warned by Andrea for this kind of system the PP approximation can breakdown.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

Davins
Posts: 1
Joined: Wed Oct 23, 2019 8:18 am

Re: NaN E-E0 result for Mg

Post by Davins » Fri Jan 10, 2020 11:31 am

What's usually the breakdown point of the PP approximation for this kind of system?

Cheers

User avatar
Daniele Varsano
Posts: 2149
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: NaN E-E0 result for Mg

Post by Daniele Varsano » Fri Jan 10, 2020 11:47 am

Dear Davins,
please sign your post with your name and affiliation, this is a rule of the forum and you can do once for all in the user profiele preferences.
As commented by Andrea, in metals you can experience failure of the PP approximations becuase of the appearance of low energy plasmonic peak in the dielectric function. Anyway, there also cases where it works reasonably well.
Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

Post Reply