[Sax-project-users] GW - core-valence exchange -correlation

[marsamos at democritos.it]

Dear Ali,
Quoting kazempoor at ph.iut.ac.ir:

> Dear All
> As I know the GW method add the self energy to state of valence 
> band.From the literature we know that core-valence exchange 
> correlation is important in some case like Zn. I mean Zn(3d) and 
> Zn(P) orbital overlap with each other. so ,LDA and GGA remain intact 
> the position of state. But when we combine LDA with GW , fock 
> operator just act on 3d and don't change the p state( treated as 
> semicore). Is  any trick that take into account the core-valence 
> exchange correlation by GW?

For this you need to use a pseudo potential that contains d electrons 
in valence. Then you can do the calculation as always. I strongly 
suggest you to 1) use a pseudo with d electrons in valence
2) perform a LDA+U calculation (this is important because simple LDA is 
completely wrong for localization of d states to do GW you need a 
reasonable starting point.)
3) GW


> another question is that :is it possible to just apply the exchange 
> or correlation part of selfenergy?if yes, How?
>

In the code if you want the band energy only with exchange (i.e. 
Kinetic+ion-electron_hartree+exchange) ask for an HF calculation and 
see HF_QP.

>
> Another question:According to previous email, we know that GW codes 
> don't give tottal energy. If we want to compute the defect formation 
> energy to obtain defect state, How can I get the total energy?
> Thanks a lot


You should read the paper from P. Rinke et al Phys. Rev. Lett. 102, 
026402 (2009). It is really the best way for defect formation energy in 
charge state. I use it for charged interstitials in SiO2 and it really 
works good.

bests

Layla

>
> Ali Kazempour
> Physics Department, Isfahan University of Technology
> _______________________________________________
> Sax-project-users mailing list
> Sax-project-users at qe-forge.org
> http://qe-forge.org/mailman/listinfo/sax-project-users
>



----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.