Dear Marcos, Alexander,
To be honest, I tried with bsse correction, and no
serious change have been found.
Even with Spinpolarized T, I get the same crude
cohesive energy of -10. eV/atom for bcc Fe
(Ec_exp ~ 4.6 eV/atom).
PBE/DZP
a (Ang) E_KS
2.87 -1569.6594
30. -1550.6534
Ec = 9.5 eV/atom
RPBE/DZP
2.87 -1570.3247
30. -1555.1728
Ec = 7.5 eV/atom !!!! (this is the best)
RPBE/DZP+BSSE
2.87 -1672.518952
30. -1553.344579
Ec = 9.6
PW92/LDA
2.87 -1557.232591
30. -1537.053983
Ec = 10.1
Therefore RPBE performs the best among various xc
functionals, however, still is far from being acceptable.
For me this poor performance for Fe cohesive energy is due
to the weakness of available xc functionals in SIESTA.
In recent publications authors have reported better results
using GGA/PW91 functionals (Ec=5-6 eV/atom), unfortunatelly
this "classic" functional is not available in SIESTA.
Regards, Peter Sule
On Fri, 12 Mar 2010, Alexander Vozny wrote:
No, BSSE would save maybe 0.5 eV but not 5eV.
My guess, is that you have to do SpinPolarized calculation for single atom.
