[DFTB-Plus-User] Electronic temperature

[Bálint Aradi]

Dear Reinaldo,

> However, a new system under study is becoming really hard to converge.
> I'm able to achieve convergence with small electronic temperatures but
> at the cost of having some non-integer occupation numbers. If the
> temperature is decreased a little, then no convergence is achieved at
> all. Thus, my question is: what limiting values are accepted for
> non-integer occupation numbers when electron smearing or electron
> temperature is used for scf? That is, is a distribution like 0.70-0.30,
> for example, acceptable or not?

 The short answer would be: depends on your system, but basically
everything is allowed. In theory, in DFT you could even make the
occupation numbers part of the optimization process and optimize them to
yield the lowest possible energy. Of course, no one really does that...

 In metallic systems or more general in systems with degenerate or
nearly degenerate levels, you would naturally have non-integer
occupancies. It wouldn't be even physical to have only integer ones, as
through that (provided your system converges at all), you would pick out
some of those degenerated orbitals and occupy them, while the others of
them you would leave unoccupied. But that would create a sort of
symmetry breaking, by treating theoretically equivalent orbitals
differently.

 It is important, however, that when temperature causes non-integer
occupations, it is the free energy, which is the relevant energy, not
the total energy any more. You'll see, that in those cases the two
quantities differ somewhat.

  Best regards,

   Bálint

-- 
Dr. Bálint Aradi
Bremen Center for Computational Materials Science, University of Bremen
http://www.bccms.uni-bremen.de/cms/people/b_aradi/


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 259 bytes
Desc: OpenPGP digital signature
URL: <http://www.dftb-plus.info/pipermail/dftb-plus-user/attachments/20120903/c0136be6/attachment.pgp>