Zn NCPP for Yambo calculations

[giacsport]

Dear All,
I honestly hope this request will not sound "weird" and/or inappropriate.....

Since this is the PW section of the list and since the PWscf release does not provide it, I wonder if someone can share a tested Zn pseudopotential NCPP/PBE for PW.

(the idea is then to us it for Yambo calculations)

Very best,
Giacomo


Giacomo Giorgi, Ph. D.,
Department of Chemical System Engineering,
School of Engineering, The University of Tokyo,
Japan

[claudio]

Dear Giacomo

you can have a look in the Rappe Group:

http://www.sas.upenn.edu/rappegroup/htd ... ga.html#Zn

where they have the parameters to generate a pseudopotential for Zn (GGA) with Opium,
and if you just change the output format you can get a UPF pseudo.

Or have a look to this post in the Yambo forum:
viewtopic.php?f=14&t=172

in the attached tar.gz file they provide a NCPP pseudopotential for Zn, but you should
ask them how they generated it and so on.
Moreover in the same post there is a discussion on the fact that a pseudo with a small core
is necessary to get reasonable results, try to contact them.

Cla

[giacsport]

Ciao Claudio e grazie!
Yes, indeed the idea was to start from the Rappe one, but you know I am starting from scratch in using Yambo, PW and also in generating PP .... thus I am a little bit doubtful.

Let me stress one point about the post you indicated me.
If I correctly understand the point is that GW gives rise to an underestimation of the experimental bandgap of ZnO of about 0.8eV.

I think this is quite obvious since the DFT-LDA calculations on such systems are affected by a bad positioning of 3d Zn orbitals.
I tell this because it is my plan to start from an LDA+U guess in order to do GW calculations on it.

It should be the only viable way to solve this underestimation.
(1. Aligning the 3d Zn with "+U" to the experimental value 2. Open the bandgap with GW)

Maybe it helps (or maybe not...)

thanks a lot!
Best,
Giacomo

Giacomo Giorgi, Ph. D.,
Department of Chemical System Engineering,
School of Engineering, The University of Tokyo,
Japan

[andrea marini]

I tell this because it is my plan to start from an LDA+U guess in order to do GW calculations on it.

Dear Giacomo, be careful with this approach. To perform GW calculations you need to remove the Kohn-Sham xc term first. At the moment Yambo can handle only LOCAL LDA xc-correlations, no LDA+U.

If you want to use LDA+U you should code this term in yambo somehow. If you are interested in doing it please contact any of us.

[giacsport]

Thanks Andrea!

And indeed this is my doubt.....
Is it possible to make for example an initial LDA+U guess with PWscf and then create with /p2y a Yambo input for GW calculation on the previously PWscf calculated guess?

Thanks once more for the hint!
Giac

P.S.: I know for sure that PWscf can "correct" the bandgap through a GW approach on a converged LDA+U guess (./pw4gww )



Giacomo Giorgi, Ph. D.,
Department of Chemical System Engineering,
School of Engineering, The University of Tokyo,
Japan

[andrea marini]

Is it possible to make for example an initial LDA+U guess with PWscf and then create with /p2y a Yambo input for GW calculation on the previously PWscf calculated guess?

Sure. It would a matter of saving in the core databases something to make Yambo able to evalute the matrix elements of the xc-potential LDA+U. Alternatively one could write (trough p2y) the matrix elements themselves.

However, keep in mind that this can be done by knowing quite well yambo (from this POV we can help you) AND PWscf (I have no idea where the m.e. of LDA+U are written/store in PW).

[claudio]

Ciao Giacomo

alternatively you can also think to use other exchange correlation functionals as B3LYP, that is now implemented in PWSCF,
but also in this case you have remove the Kohn-Sham xc term in GW, perhaps it is easier than LDA+U.

cla

[giacsport]

....a horrible suspect, does yambo ONLY works with LDA pseudo (not GGA/PBE)?.....
I knew indeed it works with NCPP but this message highly alarmed me


<01s> Symmetries...[ID yes]...[INV no]...[TR yes]...[SI no]
<01s> [WARNING]GGA not supported: swicthing to PZ. PW XC string is: SLA PW PBE PBE
<01s> XC functional...Perdew & Zunger (xc)
<01s> K-points mesh...done


best,
G

Giacomo Giorgi, Ph. D.,
Department of Chemical System Engineering,
School of Engineering, The University of Tokyo,
Japan

[claudio]

Dear Giacomo

your suspect is right
anyway it should not be difficult to add GGA functional to yambo,
because the last version of YAMBO uses LibXC (http://www.tddft.org/programs/octopus/w ... .php/Libxc).
I think the error that one makes using a LDA functional instead of GGA one in YAMBO should be small, but I never checked it,
buy anyway it is not correct.

Cla

[andrea marini]

anyway it should not be difficult to add GGA functional to yambo,
because the last version of YAMBO uses LibXC

... except that we need to code the density gradient to be passed to LibXC ! Any volunteer ?