[Pw_forum] EXC (hybride) functionals

Paolo Giannozzi giannozz at nest.sns.it
Wed Apr 20 14:05:33 CEST 2005


On Monday 18 April 2005 09:52, Jose C. Conesa wrote:

> Indeed computing exact exchange (to be able to use an hybrid
> functional) is very computer-intensive.

the straightforward algorithm with plane waves is described in
Chawla and Voth, J. Chem. Phys. 108, 4697 (1998). It is perfectly 
feasible but very slow. I guess this is what is implemented in CPMD

> An alternative to LDA+U seems to be self-interaction correction.
> Normally this is also rather computer-intensive

also rather ill-defined, at least in the original formulation: "the
self-interaction of one KS overlapping with many others is not a 
well-defined quantity" (Harris and Ballone, Chem.Phys.Lett. 303, 
420 (1999)). 

> but there are some proposals on approximations to it which 
> claim that is is possible to do it more tractable (e.g. Filippetti and
> Spaldin, PRB 67, 125109, 2003). Has someone experience or more
> detailed information on their behavior for different types of systems,
> their ease of implementation and the computation effort involved?

disclaimer: I read the above paper but did nothing!

The implementation, both in the pseudopotential generation code
and in the self-consistent code, should be relatively simple. Most 
of the needed stuff (projection on atomic orbitals for instance) should
already be there. It is not clear to me whether the orbitals should
be orthogonal or not, though, and how forces can be calculated 
(since "a physically meaningful energy functional which is also
related to [the KS equations] by a variational principle is not
available"). I would be curious to try, but it is not clear to me that
this approach will yield better results than "traditional" SIC, whose
performances do not seem to be fully convincing (a recente paper 
on this subject: Vydrov and Scuseria, J.Chem.Phys. 121, 8187 (2004))

The SIC for one unpaired electron systems: d'Avezac, Calandra, Mauri,
http://xxx.lanl.gov/abs/cond-mat/0407750, is implemented in FPMD.

P.
-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy



More information about the Pw_forum mailing list