[Ifeffit] polarization on FEFF calculation

[Matt Newville]

Hi Eugenio,


On Sun, May 24, 2015 at 7:30 AM, Eugenio Paris <eugenio.paris at gmail.com>
wrote:

> Thank you for the answer, now things are more clear.
>
> Now the question is: how to go further on the fitting procedure?
>
> If the Artemis software is weighting the contribution of different paths
> using the "Rank" value it sounds like i'm in trouble.
>



Fortunately, it doesn't do this.   The "Rank" reported by Feff is an
estimate of how important a Path is.   The Rank (or, "Zabinsky Curved Wave
Importance Factor") is *not* actually the calculation.

When you set the POLARIZATION, the path contributions in feffNNNN.dat do
correctly include the effect of polarization on the EXAFS amplitude.  Well,
"correctly" might sound optimistic, and there might be subtleties and there
might be small errors, but polarization really is taken into account at the
quantum mechanical level.



> The Fourier transform of my EXAFS signal shows a wide peak between 2 and 3
> Angstrom containing contributions from the first 3 scattering paths.
> However, it looks like in this geometry i can neglect the 3rd path. That's
> the reason why i'm interested in modeling the first two scatterings.
>
> I'll try as follows,
> I will import the first 2 scattering paths without taking account of
> polarization and try to include such dependency in the S0^2 parameter.
>
> The information that I have now is that the amplitude of the @S1.2@ path
> is the key to give the correct weight to the first two paths.
>
> Performing the calculation of the @S1.2@ path with POLARIZATION 1. 0. 0.
> and 0. 0. 1. gives Rank = 3.67 and 25.80, respectively.
>
> I wish to assign S0^2 = amp for the @S1.1@ path (in-plane contribution)
> and S0^2 = (3.67/25.8)*amp for the @S2.1@ (out-of-plane contribution).
>
> Can this approach be ok?
>
>

I recommend using the feff paths from a calculation with POLARIZATION set.
Basically, if you do the calculation with a POLARIZATION  that corresponds
to the measurement, there is nothing further or special you need to do in
the analysis.   The EXAFS is still the sum of those paths.
Specifically, S02 should not depend on path geometry.

If I understand your problem correctly, the amplitude of the first and
third path (both along the z axis, and the same atomic species) should be
comparable.   With a polarization vector in the xy plane, this should be
smaller than the contribution from the 4 @S1.1@ paths, but probably not
actually zero (the estimate was 4%, right?).    I'm not sure whether you'd
see those or not -- that's generally right around the scale of "not
obviously detectable", and with the two distances  (~2.5 and 3.3A) the
contributions from paths 1 and 3 might wash out.  I'd recommend trying to
look at the contributions from all three paths (prior to a fit) in both k
and R space.   Artemis used to make this very easy (sum paths without
fitting),  but I don't recall how to do this in the latest version.

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20150524/a28ddce2/attachment.html>