[Wien] A general question on LDA+U

[Victor Pardo Castro]

> Inspired by a previous question I came up with this general question on
> LDA+U calculations. When working with an LDA+U case with a large U, say 8
> or 9 eV, what is the best approach to reach a converged state? Either you
> can start from
> (1) a calculation whis has converged without the LDSA+U orbital potential,
> then slowly turning up U until the desired value is reached, or
> (2) a just initialzed (not converged) case with the starting density as
> provide by lstart and the (large) desired U-value applied from the
> beginning.
>
> Could these two approaches lead to different converged states?

Yes. Within LDA+U you usually can converge several different solutions with
different properties and total energies depending on the starting density.

> Is it
> possible to say which of these two approaches is the most reasonable one?

The one giving the lowest energy (at the same U value). But I would say that no
general rule can be applied.



-- 
Dr. Victor Pardo
Electromagnetism - Applied Physics Department
University Santiago de Compostela
Facultad de Fisica, Campus Sur s/n, E-15782 Santiago de Compostela (Spain)
Phone: +34 981 563100 Ext:13960     email: vpardo at usc.es