e-e calculation using GW

Forum for discussing issues arising with interfacing the Abinit code (KSS) file with Yambo.

Moderator: Conor Hogan

Post Reply
ajitkv1
Posts: 6
Joined: Sat Feb 18, 2017 9:10 pm

e-e calculation using GW

Post by ajitkv1 » Sun Feb 19, 2017 5:48 pm

Dear users and developers,

I want to calculate e-e scattering lifetimes in a metal based on GW approximation based on abinit. This is my first time working on GW. I was wondering how long it would take for a metal like copper to get the lifetimes starting from the scf calculations in DFT with and without running yambo in parallel. Is using Quantum espresso Pw + yambo faster than Abinit + yambo ? I also would like to add {DFT+U with spin polarization} as well after my DFT calculation. Apologies if the questions are too simple.

Thanks
Ajit
Ajit Kumar
Virginia Technology University
USA

User avatar
Daniele Varsano
Posts: 4047
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: e-e calculation using GW

Post by Daniele Varsano » Sun Feb 19, 2017 6:00 pm

Dear Ajit,
please fill you signature with your affiliation, this is a rule of the forum. You can do that once for all in the settings of your user profile.
I was wondering how long it would take for a metal like copper to get the lifetimes starting from the scf calculations in DFT with and without running yambo in parallel
Unfortunately we do not have benchmarks for this kind of calculations, so I cannot tell you.
Anyway, depending on the convergence parameters, it will depend a lot from the pseudo you use etc. The GW part of the code is totally parallelized with different strategies so running in parallel will be very advantageous both in terms of time and memory distribution.
Is using Quantum espresso Pw + yambo faster than Abinit + yambo ?
Once the ground state is calculates, next the yambo calculation is independent of the code used to generate it. Which is faster between Abinit and QE I do not know, I imagine to find similar performance for a standard ground state calculation, I suggest you to use the one you are more familiar.
I also would like to add {DFT+U with spin polarization} as well after my DFT calculation.
Yambo can support spin polarized calculations, anyway if you plan to do a GW calculation starting from DFT+U, there will be some term misising in the commutator coming form non local potential needed for the evaluation of dipoles, if these are important or not, it is something that need to be verified.

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/

ajitkv1
Posts: 6
Joined: Sat Feb 18, 2017 9:10 pm

Re: e-e calculation using GW

Post by ajitkv1 » Mon Feb 20, 2017 7:36 pm

Daniel,

Thanks a lot for the quick reply. I was able to reproduce the example of Al in yambo with Pwscf. One question i have is does yambo works only with Norm-conserving pp's. I tried with an Ultrasoft PP and during db file generation using p2y, i got an error " ierr: 1, Potential file not norm conserving". Please clarify. Also, what does the -ve values for linewidths in the last column means? (just above & below the Fermi level)
Can i ignore the sign when calculating any transport property using the tau's ?

Thanks
Ajit
Ajit Kumar
Virginia Technology University
USA

User avatar
Daniele Varsano
Posts: 4047
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: e-e calculation using GW

Post by Daniele Varsano » Tue Feb 21, 2017 10:16 am

Dear Ajit,
please fill you signature with your affiliation, this is a rule of the forum. You can do that once for all in the settings of your user profile.
One question i have is does yambo works only with Norm-conserving pp's.
Yes, at the moment Yambo support norm conserving pseudopotential only. Implementations to support ultrasoft PP are envisaged, but I can't tell you when they will be operative.
what does the -ve values for linewidths in the last column means (just above & below the Fermi level) ?
Yes, it is just a convention, positive for electron and negative for holes.

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/

ajitkv1
Posts: 6
Joined: Sat Feb 18, 2017 9:10 pm

Re: e-e calculation using GW

Post by ajitkv1 » Tue Feb 21, 2017 3:10 pm

Hi Daniele,

I updated my signature as per your suggestion. Are you familiar with any other code than can do GW calculations with Ultrasoft potentials? Abinit has a GW tutorial that shows the calculation of electron SE http://www.abinit.org/doc/helpfiles/for ... n_gw2.html for Al. Do you know how accurate and efficient they are ? Please let me know.

Thanks
Ajit
Ajit Kumar
Virginia Technology University
USA

User avatar
Daniele Varsano
Posts: 4047
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: e-e calculation using GW

Post by Daniele Varsano » Tue Feb 21, 2017 3:18 pm

Dear Ajit,

no I can't help you on that, I do not know details of other GW codes.
What I can tell you is that the new class of norm conserving Don Hamann pseudo potentials (ONCVPSP) are very accurate and relatively soft, in the sense that requires smaller energy cutoff with respect other norm conserving pseudos.
You can find here a table of elements:
http://www.pseudo-dojo.org/index.html
and this should be the reference:
http://journals.aps.org/prb/abstract/10 ... .88.085117

Here you can find the code to generate them:
http://www.mat-simresearch.com/

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