[Wien] GGA+U

[Stefaan Cottenier]

> I actually wonder whether there is some specific reason why there is no
such
> kind of " GGA+U" calculation in the literature but only LDA+U.....is this
> maybe due to the fact that GGA is already an improvement upon LDA and
> therefore LDA+U should take already into account  the feature of GGA
alone?

Other may correct me if I'm wrong, but I guess this is the reason: the
"U-trick" is a rough approximation, that is not really ab-initio (you have
to tune a free parameter). At best, one can hope for semi-quantitative
agreement with experiment. In such circumstances, it is best to agree to
start from the only XC-functional that can be uniquely defined: LDA (GGA's
come in different flavours). If one would start using GGA+U, then there
might be a correlation between the U and the type of GGA: the question
"which U should I use" would become even more difficult to answer...

Stefaan