[CPMD-list] B3LYP implementation

[Juerg Hutter]

Hi
>
> In CPMD3.9.2 I found out some strange thing... B3LYP is implemented with unusual correlation
> part: 0.19EC(LYP,local)+0.81EC(LYP,local+nolocal) while standard form is
> 0.19EC(VWN)+0.81EC(LYP,local+nolocal) (as in GAMESS(VWN5) or Gaussian(VWN3)).
> Maybe, right form is 0.19*VWN+0.81*LYP+0.81*GLYP (now 1.0*LYP+0.81*GLYP)?
> What are your opinions?

Yes, it's true that QC codes use the first formula.
However, I thought it is more in the spirit of the original
proposal of Becke not to touch the LDA part of the functional.
In both cases the correct LDA value for the free electron gas
will not be recovered.
Anyway, in my opinion, the variation of energies is larger
from the different approximations to basis sets and pseudopotentials
than from the difference in the definition of the functional.

regards

Juerg Hutter


>
> Best regards,
> Alex Kargovsky
> Department of Physics,
> Moscow State University
>
>
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
>