Hi all,

I am trying to do calculations for bulk (fcc) Pd using the van der Waals
functional parametrized by Klimes et al (vk in atom, KBM in siesta, as in
JPCM 22, 022201 - 2009). I have generated my own pseudo and I just can't
obtain the results for the lattice constant described in PRB 83, 195131
(2011). I have tried many different schemes (non-linear core corrections
and inclusion of semicore states down to the 4s channel), but my best
results are overestimated by at least 4% with respect to the lattice
constants reported in PRB. Here are a few remarks I have about the
calculation:

1) With GGA, I get lattice parameters in very good agreement (about 0.1%
diference) from the results of the aforementioned PRB, when using the
Abinit pseudo. However, when I tried generating my own GGA pseudo using
exactly the same rc's as in the Abinit translation, I get a 2% difference
in the final lattice constant. I tried using both ifort and gfortran, to
make sure this wasn't a spurious result coming from an obscure compiler
bug, but they gave me the same results.

2) I notice, in the siesta calculation output:

xc_check: Exchange-correlation functional:
xc_check: VDW KBM
xc_check: WARNING: Pseudopotential generated with Unknown atomic XC code
functional

and I'm not sure if this would be harmless or not. For Platinum, results
for the KBM XC functional were in good agreement with PBE GGA, which is
what I would expect since KBM and PBE seem to be closely related (the
difference would be mostly in the parameters of the enhancement factor -
please correct me if I'm wrong).

I am using a DZP basis set, with an energy shift of 0.15 eV. Has anyone
performed calculations for bulk metals, in particular Pd, using the vk/KBM
XC functionals? If so, would you be so kind as to give me tips and hints on
how to generate a suitable pseudo for Pd (using the version of atom
distributed in the latest trunk version, siesta-3.67)?

Best regards,

Marcos.

Responder a