[Wien] quasiparticles

[Stefaan Cottenier]

> Could somebody explain or point out further reading, why such a comparison
> is obviously valid?


I'm not at all an expert in "DFT philosophy", but this is my understanding
of the problem:

The Kohn-Sham scheme guarantees you that it yields the correct total energy
and the correct density (if you would use the exact XC-functional). On the
way to these 2 objects, the Kohn-Shame scheme introduces some wave-function
like objects of which the scheme itself provides no physical interpretation.
One should just throw away these wave functions once you have obtained E_tot
and the density.

Of course, we want to know more than just E_tot and the charge density. We
want to know e.g. the fermi energy, or a magnetic moment. DFT tells that
this information is contained in the charge density (which we have just
calculated), but it doesn't tell us how the functional for the fermi energy
or the magnetic moment would look like -- so that doesn't help us. Now comes
the magic: close your eyes and do as if those wave-function like things are
true single electron wave functions, and calculate the expectation value for
these wave functions using the normal magnetic moment operator (which we
know, in contrast to the corresponding functional). It's absolutely illegal
and unjustified, but comparison with experiment shows that it works... So
everybody does it.

Your question was *why* this works. I don't know whether this is known or
not, I suppose it isn't. I remember something is written about this in the
book of C. Pisani (QM ab initio calc. of the properties of cryst. mat.,
Springer, 1996), but I cannot find the relevant page right away. For sure
there must be more and/or recent discussions on this. I'll be glad to get
references from other users as well...

Stefaan