[Ifeffit] S02

Matt Newville newville at cars.uchicago.edu
Fri May 25 08:21:56 CDT 2018


On Fri, May 25, 2018 at 5:11 AM Julian Ehwald <jehwald at gmail.com> wrote:

> Dear Matt,
>
>
>
> I still have questions about Artemis concerning the S0^2. In your
> book/summary called XAFS-Fundamentals there is a section at the beginning
> of chapter six that says:
>
>
>
> “Starting with this structure, we can calculate the scattering amplitude
> and phaseshifts
>
> and theoretically. The complete calculation is beyond the scope of this
> treatment,
>
> but don’t worry too much by that phrase “we’ll calculate these
> theoretically”! These
>
> functions are actually fairly easy to calculate using one of a few
> different computer
>
> programs – for the calculations here, the FEFF program was used. The
> results of the
>
> FEFF calculation are stored in simple files containing the scattering
> factors and meanfree-
>
> path for a given coordination shell. The calculations in these files can
> be used
>
> directly in a number of analysis programs.”
>
>
>
> Are these somewhere listed in Artemis?
>

The calculations stored in the feffNNNN.dat files can be accessed from
Larch, but I think they're not readily exposed by Artemis.  The point is
that you don't have to worry about these scattering factors, but only about
the Path Parameters (S02, E0, deltaR, sigma2, etc) that modify these
scattering terms.
​

> Because if I start fitting the value in GDS is always guessed to be one,
> for any element in the sample. If I go and look at the FEFF calculation it
> tells me:
>
>
>
>
>
> TITLE LiPF6
>
>
>
> HOLE      1   1.0   * FYI: (F K edge @ 696.7 eV, 2nd number is S0^2)
>
> *         mphase,mpath,mfeff,mchi
>
> CONTROL   1      1     1     1
>
> PRINT     1      0     0     0
>
>
>
> I don’t really understand which is the second number and why all S0^2 then
> are either 0 or 1.
>
>
>
> Best, Julian
>

​One can set an S02 value during the Feff calculation.  That is what goes
as the second number in the HOLE entry in feff.inp (The first value is an
integer indicating the edge type -- 1 for 'K', and so on).  Setting that
value will apply the S02 value to all paths calculated by that run of
Feff.  You can also set values for sigma2 from Feff too,

Artemis/Ifeffit/Larch etc use the Feff calculations stored in feffNNNN.dat
(or equivalent) and build a model from a collection of paths from Feff.  If
S02 or sigma2 is set in the original Feff calculation(s), it is difficult
to track during the analysis (neither S02 nor sigma2 are written to the
individual feffNNNN.dat files).

In some sense Artemis/Ifeffit/Larch etc only use part of Feff -- the
calculation of the EXAFS contribution of a single, idealized path.  They do
not use the sum-of-paths module(s) of Feff, but instead replace that with
its own sum-of-paths implementation that is richer (allowing feffNNNN.dat
from different feff runs, allowing parameters like S02, sigma2, E0, deltaR,
etc to be varied, etc).

So, the recommendation (more like an expectation, really) is to NOT set S02
to any value except 1.0 and to NOT set any parameters that will calculate
sigma2 in feff.inp if the intention is to use those paths for fitting.
 Let the fitting program adjust S02 and/or disorder terms.
​
​Hope that helps,​

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20180525/a965315c/attachment.html>


More information about the Ifeffit mailing list