[Pw_forum] more about phonon frequencies are different when treated as insulator or metal

[Katalin Gaal-Nagy]

> when trying to describe Si as a metal what kind of smearing are you using ?
> Methfessel and Paxton or gaussian  or something else ?
> MP scheme, although in principle"better" than gaussain  for *real*
> metals, suffers
> of the fact that the corresponding occupation function is not positive
> definite.
> Usually this is not a problem with many k-points but may give spurious
> results
> with few points and a small degauss....
> Try simple gaussian or Marzari-Vanderbilt cold smearing scheme that combines
> a nice (lack of) dependence of the results on the smearing width and a
> positive definite
> occupation.

Are you sure, that the cold smearing of Marzari-Vanderbilt is positive
definite?
We had also some problems with the occupation numbers using
Marzari-Vanderbilt, which have been similar to the use of
Methfessel-Paxton, so we thought that the occupation function is also not
positive definite, but I never checked it out.

Greetings,
Katalin