[Wien] Pt atom in a box

[Peter Blaha]

> I am trying to calculate a Platinum atom in a box. The problem is, that
> no convergence can be reached even after 100 iterations. The charge
> distance seems to oscilate between values of 0.04-0.01.

As discussed previously, one should use a F-centered cubic cell with a
sufficiently large lattice parameter (test it).

I don't think it is a problem of GGA, not would I use empty spheres!!

I would use a much smaller value of RKMax at the beginning (to make sure
not to get problems from this). You can increase it later...

Most likely it is simply a well known DFT problem (both in LDA and GGA)
that the atomic configurations of late transition metals does not come out
correctly. You have the 5d and 6s orbitals as valece states. Now it
probably happens, that when 6s is occupied with 2 electrons, it's energy
raises and get higher than the 5d. Thus the occupation will change and 6s
is occupied only by ONE electron. However, this changes the potential such
that the 6s orbital will lower it's energy BELOW the 5d and in the next
step it will be fully occupied again.......

I suggest you try a smearing method (TEMP or GAUSS with eg. 0.005).
Together with small mixing a solution (with non-integer 6s occupation)
should be possible.

PS: I've no experience with Pt, but this is from analogy to Fe, as
discussed already in "Janak's" book on the elements.


                                      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/
--------------------------------------------------------------------------