[Wien] LDA+U

[Stefaan Cottenier]

> I've got a question on LDA+U.
> I would like to get different spin state configurations (High,
> Intermediate  and low spin state) for a some compound in LDA+U and 
> after that comare their  energies. 
> 
> Thus, I need to change appropriate occupation or orbital potential
> matrices.  It is obvious that changing of the orbital potential 
> matrices is an easiest  way, because in this case I should not think 
> about a total number of electrons conservation.

We have recently done similar things with lanthanides, but by changing 
the density matrices. That worked quite nicely, although it is not 
guaranteed that the calculation will converge towards the solution 
that you propose in the density matrix. It might converge to a 
different one, or (rather frequently) not converge at all.

The procedure:

1) converge a normal lda+u calculation
2) change the density matrices
3) take a rather small mixing factor
4) converge again, and keep your fingers crossed

Two considerations:

*) because of the normalization in wien2k, I don't think you should 
worry about electron conservation

**) as far as I know, comparing total energies with lda+u is a 
dangerous story. I'm not sure whether your total energies for 
different spin states have any physical meaning.

Stefaan