CCL: Spin density

[Mikael Johansson <mikael.johansson*iki.fi>]

 Sent to CCL by: Mikael Johansson [mikael.johansson(~)iki.fi]
 Hello Bonoit and All!
 On Tue, 5 Apr 2011, Bonoit Bonoit bonoit_10{:}yahoo.fr wrote:
> I'm writing to enquire about the spin density, we need
>  the MPA (Mulliken
>  population analysis) to calculate it. My question is, c'd we use
>  another
>  population analysis to calculate it as NPA forexample?
>  
It depends on what you mean by spin density. In short, you can get the
 spin population with practically any method that can provide a total
 density population.
Note, however, that the "charge" of an atom obtained by any
 of the
 population analyses is an attempt at emulating the difference of the
 negative charge of the electrons surrounding a specific atom and the
 positive charge of the nucleus in question. Thus the "spin
 density" that
 you get will simply be one number, the difference of the number of
 alpha
 and beta electrons that your partitioning scheme of choice assigns to a
 specific atom. This is not really the spin density, if you think of it
 as
 regions in space dominated by alpha/beta electrons, but rather an
 average
 of the differences over a "volume" defined by the
 population/partitioning
 scheme.
That said, the values of Mulliken "spin densitites" do
 correspond quite
 well to what you would expect, giving, for example, roughly the number
 of
 unpaired d-electrons for tranition metals. And the sum of the
 individual
 "spin densities" will (should) of course add up to the total
 number of
 unpaired electrons in your molecule, just like the sum of the atomic
 charges add up to the charge of the molecule.
As a side note, Mulliken spin densities are also much less sensitive
 towards basis set choice, in contrast to the Mulliken atomic charges,
 which can turn out quite droll with, for example, diffuse functions.
 Have a nice day,
     Mikael J.
     http://www.iki.fi/~mpjohans/