[Wien] Accuracy (relevance almost) of lstart energies

[Peter Blaha]

lstart is "exact" (for good radial mesh and within a given DFT and for
spherical atoms) !!! You cannot get a lower energy for a spherical atom
in a given DFT functional.

lstart is NOT a basisset method, but integrates the radial Schrödinger
(in fact the dirac !) equation numerically. It finds the eigenvalues by the constraint
that the solution at infinity must go to zero (normalization condition).

It really is fully relativistic, therefore its energies are LOWER than
WIEN2k-supercell energies and not compatible for the calculation of
atomization energies.

To test what I've said, you can switch off relativistic (NREL in case.struct)
and then lstart and wien2k-supercells give (almost) the same energies.

Laurence Marks schrieb:
> I am curious about how good/useful are the energies in lstart. Of
> course if one wants to calculate, for  instance, cohesion energies or
> atomization energies one has to compare apples with apples, i.e. do a
> Wien2k supercell calculation for a single atom. Ignoring this, how
> "good" are the energies in lstart (assuming that one has an
> appropriate NRAD etc)? Is the code using essentially the same method
> as elsewhere in Wien2k but with subtle additions such as
> orthogonalization of the levels? Is it technically better (or worse)?
> 
> N.B., unfortunately the comments are in French, which I don't
> understand as well as I probably should.
> 

-- 

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