[Wien] isolated atomic calculation

[Peter Blaha]

> To calculate the isolated atomic energy of an element, in Wien2k homepage it
> has been mentioned that a large fcc unitcell with one kpoint should be used.
> But for Si it seems that with one kpoint the system can not be converged so I
> have used a 2x2x2 mesh to can converge the system and the obtained cohesive
> energy agree very well with experiment. Does it physically possible to use more
> than one kpoint for isolated atomic calculation?
> Also for Cobalt I can not converge the system of isolated atom even by
> increasing the number of kpoints to a mesh of 6x6x6. I will appreciate if any
> body give me some advise for calculating the isolated atomic energy of cobalt.

I think we had these questions already several times before.

a) An ISOLATED atom should be done with ONE k-point. (Guess why ?)

b) Some atoms with partly filled shells, in particular transition metal atoms
have the problem that the charge will oszillate between two orbitals
(eg. between 4s and 3d) when you force "integer" occupation numbers. To
overcome this, use a braodening method (eg. TEMP) with sufficient broadening.

c) Don't forget to run those atoms spinpolarized!

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------