Thanks Andrei,

I was trying to use the silver system as a simple example.  What I'm truly
interested in calculating is the polarization in a polymer-metal
nanocomposite; and what I've found is that with a matrix of 400 atoms and a
metal (i.e. silver) inclusion with an odd number of atoms (I tried with 13)
then SIESTA won't do the polarization calculation: odd total number of
electrons.  But if the inclusion is only 12 atoms, it performs the
calculation!  I also tried giving the system a single positive charge but to
no avail.

Chris Rowan
M.Sc. Candidate
University of Victoria
Canada

2009/11/29 <[email protected]>

> > Hi,
> >
> > I'm wondering why polarization calculations cannot be performed when the
> > system has an odd number of electrons.  I've read through the derivations
> > by
> > Resta and the Berry phase approach, and while I follow along in a
> moderate
> > sense I don't see how the requirement for an even number of electrons
> > enters
> > the picture.
>
> Dear Chris,
> I think this requirement enters via the demand for a system
> to be non-metal. (Speaking about bulk materials,
> the polarization is a property of dielectrics...)
>
> > For example, the polarization of a single silver atom in
> > void
> > cannot be calculated, but a system of two silver atoms works fine.
>
> I think you can polarize any isolated atom, even a silver one,
> but then the Berry phase approach won't be appropriate, because
> it treats the polarization current flowing through a given cross section,
> in a periodic system. The periodicity is important: it permits you
> to access the polarization as a bulk property, without bothering
> about charge distribution at the boundaries. This does not make much
> sense if your entire system is confined in your cell, so that the "surface"
> is explicitly included. If your silver atom is isolated, you won't
> (or at least you shouldn't) have any current flowing from it to the
> adjacent cell. In this case, you can calculate polarization
> straightforwardly, via dipole moment, and will have no need of Berry phase.
> The same about a dimer: even if the calculation "works fine"
> I'd suggest that you check what is really calculated; in my opinion
> the Berry phase approach doesn't make much sense for this system either.
>
> Best regards
>
> Andrei Postnikov
>
> >
> > Any comments much appreciated,
> >
> > Chris Rowan
> > M.Sc. Candidate
> > University of Victoria
> > Canada
> >
>
>

Responder a