hybrid DFT for bulk

Axel akoh... at gmail.com
Fri Jul 18 23:25:20 CEST 2008



On Jul 18, 5:01 pm, molecule <san... at gmail.com> wrote:
> 1) The executable did pass the tests. I compiled cp2k using both pgf90
> and ifort10.1 and the results are pretty close to each other.
> 2) What's wrong with metals in cp2k and why the Ex is so abnormally
> huge?

the second part was addressed by juerg already.

please explain why it _should_ work?

for a metal you'd need (lots) of k-points (less with smearing),
fractional occupations, and a diagonalization and mixing scheme
that minimizes charge sloshing and (probably) a well tuned basis
set that is maximally diffuse without linear dependencies.

all of that is independent of using hybrid DFT, of course, and
means that even if you'd get a number out of your calculation
it would be a very expensive random number.

cheers,
    axel.


>
> -molecule
>
> On Jul 17, 3:10 pm, Axel <akoh... at gmail.com> wrote:
>
> > On Jul 17, 3:04 pm, "sh ke" <san... at gmail.com> wrote:
>
> > > Hello, Dear All,
>
> > > Has anyone done a hybrid DFT, like pbe0, calculation for bulks, like Au?
> > > I did pbe0 and b3lyp calculations for bulk Au with a 4-atom unit cell
> > > but the result
> > > is a mess. Can anyone help?
>
> > two questions:
>
> > 1) does your cp2k executable pass the regression tests?
> > 2) have you considered the fact that gold is a metal?
>
> > axel.


More information about the CP2K-user mailing list