[Ifeffit] Increase in uncertainty of Eo and fitting mismatch

Matt Newville newville at cars.uchicago.edu
Tue Mar 10 12:08:24 CDT 2015


Some comments and suggestions.

On Tue, Mar 10, 2015 at 11:11 AM, Raj kumar <rajrk37 at gmail.com> wrote:
> Dear All,
> Recently, I recorded Copper foil EXAFS as a reference in (Dispersive EXAFS)
> transmission mode at Soleil, Paris. For understanding the data quality (of
> my sample), i started fitting copper EXAFS through Artemis. After the DATA
> extraction, i fitted copper in Artemis and found a shift in energy. For this
> reason, i have modeled with and without corrected Copper EXAFS in Artemis.

What does "with and without corrected Copper EXAFS" mean?

> Initially, i started with first shell and progressed to three shell fit
> through step by step. During the course of fitting, some of the following
> troubles have been faced. Please help me to rectify it.
> For first shell fit: Everything goes well and converged to reasonable
> physical values of copper foil. Whereas for further increment in copper
> shells leads to increase in Debye-waller factor for all three shells and

By itself would not  worry me too mcuh.

> increase in uncertainty value of E0 than the actual value.

What actual value?  You mean the uncertainty in  E0 increases when you
go from 1 to 3 shells?   That increase in uncertainty could easily be
due to a worse overall fit.

> Moreover, i have noticed that the (increase of shells in the fit) not only
> changes the physical parameters and also modifies the Fourier transform.

The Fourier transform of the data depends only on the k-space data,
and the weighting and windowing applied.  It doesn't depend on the
values of the fitting parameters -- those affect only the Fourier
transform of the model.

> The inference are the following: A shift in first and third shell of Cu-Cu
> are observed, respectively, to larger and shorter distance than the
> experimental spectrum with diminished second shell amplitude.

Not sure what you mean.... larger and shorter than what?

I would recommend fitting one spectrum first -- I don't quite
understand why you have 2 spectra with an obvious E0 shift between
them, and no (unless I missed something) shared parameters.


More information about the Ifeffit mailing list