[Pw_forum] CuO, LDA+U, Pseudopots, etc.
amigliore at cmm.upenn.edu
Fri Dec 14 06:32:51 CET 2007
I noted that one of my previous emails was misleading. In fact, the system
is Fe2+/Fe3+, but I noticed the problem when performing calculations on a
single solvated ion, with Fe2+ and no fractional occupations. The system
can anyway get a local minimum for some configuration and some mixing_beta
value (let me remark that this occurs rarely), and this can happen both
with a constrained total magnetization and without it. It seems just a
convergence problem connected to the mixing factor.
> Agostino Migliore wrote:
>> 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.
> Did you characterize the local minimum, by any chance ? The system
> is fairly simple (on one side, 5 d electrons, say all up, and
> on the other side, 6 d electrons). Do you have a sense if the
> additional minimum is due to some long range ferro vs. antiferro
> coupling between the two ions (i.e. 5 up on one side, and 5 up and
> 1 down, vs 5 up, 5 down and 1 up), or if the minorit spin on the ferric
> ion splits in two or three orbtials with fractional occupations ?
More information about the users