[Pw_forum] CuO, LDA+U, Pseudopots, etc.

[Matteo Cococcioni]

Hi Agostino

what kind of system are you referring to? it must be one with several 
minima. if this is the case I think you get to a different one changing 
whatever parameter in the input just because of (slightly) different 
initial conditions or numerical noise that may bias the system towards a 
different state. I'm not sure what could be a good way to explore them 
all or to bias the system towards the "right" (lowest energy) one. Maybe 
using fractals..... (just kidding). Seriously, if one can distinguish 
the minima then one can write a constraint but what is a relevant degree 
of freedom to perturb in general?

Thinking of CuO I have the suspect that in this system the hybridization 
between Cu and O is important (although the electron may still be quite 
localized). If this is the case then the lda+U on atomic orbitals may be 
not a good idea because it will force integer occupations on atomic orbitals
while hybridization requires sharing of electrons between different 
atoms. Probably wannier functions (that Paul was going to try) are a 
better option.

Matteo



Agostino Migliore wrote:
> Hello
>
> Also independently of this specific problem, I use the occasion to say
> that sometimes, with finite U, the code gets trapped in local minima for
> some values of the "mixing_beta" parameter.
>
> Agostino Migliore
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>   

-------------- next part --------------
A non-text attachment was scrubbed...
Name: matteo.vcf
Type: text/x-vcard
Size: 294 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20071213/3c09b609/attachment.vcf>