[Pw_forum] Problem with nbnd in clusters

Matteo Cococcioni matteo at umn.edu
Thu Nov 22 21:22:47 CET 2007


Dear Cyrille

let me add last comments too and ask few questions.


On Nov 22, 2007, at 7:30 AM, Cyrille Barreteau wrote:

>
> A final comment concerning this tricky problem.
> I think I got the reason of my troubles.
> I have two scf solutions:
>
> 1) one that (almost) satisfies the symmetry of the real dimer
> and for which xy and x^2-y^2 eigenfunctions are degenerate.
> However this solution has the bad idea of having an unfilled homo
> state.

I think this is the same bad idea of e.g. transition metal oxides  
that turn out to be metals (instead of insulators) in DFT.
The difference is that in your case you don't have a particular  
reason to suppose that the electron should instead
be localized on one of the degenerate states, while in a crystal you  
may appeal to crystal field, symmetry etc.
Are the degenerate xy and x^2-y^2 orbitals part of the unfilled homo?  
if so, do you get one of them above the "Fermi energy" and one below
when you get the other, lower energy solution? If this is true I  
think that you would obtain the same energy with the two states
inverted (the one that was above ef goes below and vice versa). Can  
you check this?

I agree with Stefano in general and prefer the broken symmetry  
solution. I think this gives you the "right" (of course in a DFT  
sense) energy.

However I don't feel completely confident about the fact that you can  
obtain it just with nosym = true and placing the molecule in a generic
orientation w r t the cell. Your ground will still have the symmetry  
of the fake crystal that you construct (even f you don't impose it)
unless a broken symmetry solution exists.
I guess my question is: is the numerical noise produced in the  
diagonalization (and of course not washed out by the imposing of the  
symmetry)
always enough for the code to recognize a broken symmetry solution  
and to choose it? I'm not sure.
In your case I would feel more confident simulating a "supercell"  
with few molecules randomly oriented wrt each other. This would be
a more faithful simulation of the gas phase you are probably thinking  
of than the fake crystal you are using. Of course this will be more  
expensive though...
Maybe the answer to this doubt of mine is trivial. Does anybody have  
more clear ideas than mine on how to avoid symmetry?

Regards,

Matteo


>
> 2) an other that do not satisfy the symmetry of the real dimer
> (but of course satisfy the symmetry of the crystal) since a
> rather large spitting of xy and x^2-y^2 is observed.
>
> The second solution is the most stable for reasons related
> to the so called Jahn-Teller effect. There is no true Jahn
> Teller distorsion but the dissymetrization of xy and x^2-y^2
> leads to a splitting of levels that decreases energy....
> bad luck!
>
> I guess this problem will disappear if I add spin-orbit coupling
> since it will induce splittings and the (bad) chance to have
> an unfilled homo state is smaller.
>
> that's all folks.
>
>    cyrille
>
> -- 
> ==================================================================
> Cyrille Barreteau  | phone : +33 (0)1 69 08 29 51
> CEA Saclay         | fax   : +33 (0)1 69 08 84 46
> DSM/DRECAM/SPCSI   | email cyrille.barreteau at cea.fr
> Batiment 462       |
> 91191 Gif sur Yvette Cedex  FRANCE
>           ~~~~~~~~~~~~~~~~~~~~~~~~
> http://www-drecam.cea.fr/Images/Pisp/cbarreteau/cbarreteau_fr.html
> ==================================================================
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> Matteo Cococcioni
> Department of Chemical Engineering and Materials Science
> University of Minnesota
> 151 Amundson Hall
> 421 Washington Av. SE
> Minneapolis, MN 55455
> Tel. +1 612 624 9056    Fax +1 612 626 7246

> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20071122/af69c595/attachment-0001.htm 


More information about the Pw_forum mailing list