Single atom transition metal energy...

[pdalach]

Hi,
I'm attempting to calculate substitutional energies of species in bulk, which require me to perform calculations of the substituted atoms in isolation.

I know that the energy without entropy of an atoms should be equal to zero within VASP's output. Some single atom species have converged (such as oxygen) and are significantly non-zero...order of .5eV. Others, the transition metals, are not converging at all.

I have tried to use, for consistency, the same input parameters as my bulk calculations, so spins are unpolarized. POTCARS are from the standard GGA pseudopotential (PW91 I think). To try and help SCF convergence, I have increased/decreased ENMAX, decreased AMIX and BMIX, increased/decreased KPOINTS and varied ISMEAR. No luck.

The dE in OSZICAR just flip flop between negative and positive no matter how much I increase NELM.

Some of the transition metals converge if spin polarization is added in. For those, could I just subtract off the correction found in the vasp documentation?

Any suggestions?

Thank you

[admin]

please note that for some transition metal atoms the POTCAR is generated for the ground state of the atom in its bulk phase (ground state) , which may not be in accordance with HUND's rule for the free atom (please have a look at the online manual).
--> you HAVE to add spin polarizarion, unless the configuration is s2d10
-- you may also have to add the magnetic moment of the free atom explicitely (NUPDOWN, or FERWE and FERDO) to converge to the correct configuration and degeneracies
-- use ISMEAR = 0 and decrease SIGMA to about 0.001

-- please don't forget that you have to use the Gamma point only for free atoms