[Ifeffit] About S02

[Matt Newville]

Damdin,

An XAFS book or theory that says "S02 must be between 0.8 and 1"
overstates the case.

In simplest terms, S02 is a fudge factor.  There is a theoretical
basis for it (passive electron overlap), but much of this can also be
folded into other "loss terms" that are due to inelastic scattering
events for the photo-electron.  These should be part of the EXAFS
calculation from, say, Feff, but are hard to do well to high
precision.  When using Feff, Ifeffit, Artemis, it does turn out that,
in most cases, the calculated amplitudes tend to overestimate the
experimental amplitudes, and that applying a constant S02 between 0.8
and 1.0 works reasonably well.  That is hardly a "must be" between 0.8
and 1.

In addition to the theoretical issues of Feff not-quite-correctly
calculating S02, there are experimental factors that affect the XAFS
amplitude: the estimate of the edge step being the most important, but
also the energy resolution of the incident x-rays and sample-dependent
issues such as pinholes, and for fluorescence measurements detector
deadtime and self-absorption.

There is an understandable desire to turn these observation into a
rule-of-thumb and then into some sort of rule that can be pointed to
as if to say "this analysis is wrong" or "this analysis is correct".
As is normal for such rules-of-thumb, it is sometimes broken.

Of course the main trouble with the uncertainty of which S02 to use is
that for a single-shell analysis, one cannot distinguish S02 from the
coordination number.  That is, the EXAFS amplitude for S02=1, N=4 is
the same as for S02=0.8, N=5, and for  S02=0.67, N=6.  Distinguishing
these cases is hard (don't let anyone tell you they aren't or that
there is a button in Artemis that fixes this problem).   If you need
to distinguish N=5 from N=6, you better be pretty confident that your
data and analysis have a good value for S02.

The classic approach to this is to measure a data on a "standard" is
free from amplitude-reducing artifacts and for which the coordination
number is well-defined as easy to extract from the data, measure this
at the same beamline conditions (energy resolution, harmonic content)
as your unknown sample, and then work to be highly consistent in
setting the edge step.   The S02 applied to the standard to get the
correct coordination number is then applied to the unknown data to get
its coordination number.     There are many ways for such an analysis
to fall short of perfect, but it is generally the best we can do.

--Matt


On Fri, Mar 27, 2009 at 8:19 AM, Zuzaan, Damdinsuren
<zuzaan at uni-mainz.de> wrote:
> Hello everyone,
>
> I am graduate student and doing EXAFS in my research.
>
> My question is about S02. From the XAFS books or theory it says that this value must be between 0.8 to 1, but in the literature some people used different values like 0.25 or 1.4. In my case I fitted experimental RDF of my standard sample by holding the known r and n values for each path, while varying E0, S02 and sigma2. Obtained E0 and S02 values were holded for unknown sample and rest were floated freely. S02 were 0.64. So my question is that this relatively small S02 value could affect the mineralogical interpretations made in my study. Should I refit it using high value? Thanks in advance
>
> Regards,
>
> Damdin
>
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
>
>