[Ifeffit] Residual EXAFS spectra

[Bruce Ravel]

On Monday 30 August 2010 09:57:09 am Bruce Ravel wrote:
> > Hi all,
> > I am doing an EXAFS spectra analysis of Eu(III) sorbed on mineral oxides.
> > In  course of this, I came across a lierature article(Geochimica et
> > cosmochimica Acta 64(16),2737,2000.) where a residual spectra (denoting a
> > particular shell) was generated after subtracting four other different
> > paths contribution from the EXAFS spectra(k3.chi(k)). This procedure was
> > adopted to identify and isolate different paths/shells informations
> > available in the EXAFS spectra. As the paper doesnot explain about the
> > procedure, I am little confused whether the author has adopted just the
> > linear subtraction methodology(i.e, sum of few selected paths minus the
> > original spectra) in creating the residual spectra or some other
> > mathematical approach. 
> 
> The caption to figure 5 in the article you cite says:
> 
>  "Residual EXAFS (produced by subtraction of U=Oax, U-Oeq, U-C, and
>   U-U shells) ..."
> 
> Seems pretty clear to me that what they show is the Fourier transform of 
> 
>     data_chi(k) - sum_of_specified_paths(k)
> 
> B
> 

Ummm... of course, I meant to say that the subtraction is done after
performing some kind of a fit to determine values of the parameters
affecting the paths included in the sum.

Now I'll editorialize a bit ... the advantage of analyzing a
difference spectrum of this sort is that a small part of the data that
might get "lost" in the full fit gets emphasized.  The disadvantage --
and a big one at that -- is that you loose the ability to evaluate
correlations between the parameters of the fit to the residual and the
parameters that were used to determine the sum of paths used to make
the residual.  Consequently, doing an analysis of the residual
spectrum requires a high level of confidence in the parameters used
for the subtracted paths.  In my experience, that level of confidence
is rarely defensible.

B

-- 

 Bruce Ravel  ------------------------------------ bravel at bnl.gov

 National Institute of Standards and Technology
 Synchrotron Methods Group at NSLS --- Beamlines U7A, X24A, X23A2
 Building 535A
 Upton NY, 11973

 My homepage:    http://xafs.org/BruceRavel
 EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/