Hi Matthias,

I don't see a problem with this background removal (using Rbkg=1.0 and kweight=3 as in your project file).  The mu0 spline does shoot up (to perhaps a value of 3 for the normalized XAFS) whereas the data goes to about 1.8 (again, for normalized XAFS).    Mu0 then drops down into the range of the data within 5 eV.

This is fairly normal for the AUTOBK algorithm.  It's not even looking at mu(E), but rather at the Fourier transform.  So to first order it doesn't care (and, I'll argue, neither should you) about what happens at the end points, especially at low k.  You're going to k-weight the chi(k) spectra anyway.  That will multiply the low k portion of the spectra by a very small number over the first few eV -- asserting that chi(k) below 1 or 2 Ang^{-1} doesn't matter.

You can use the "low energy spline clamp" to force the mu0 spline to vary less at low energy.  Even setting this to "weak" in Athena should change mu0 to be much closer to the data value at E0.  If anything, that probably over compensates.

--Matt



On Fri, Aug 29, 2014 at 2:13 AM, Matthias Filez <Matthias.Filez@ugent.be> wrote:
Dear mailing list users,
 
I am trying to fit the spline background function in a proper way to a spectrum (Ga K edge, see attached).
However, it doesn't seem to work out, but I don't see why.
 
I already tried the following:
* Vary the Rbkg value
* Change spline k-weight
* Change k-range of spline fitting
* Alter the strength of spline clamps
... and the combination of these effects.
 
Since there is a strong white line, I suspect this causes the spline misfit. The data quality is OK.
If you have any idea, please post it.
 
Thanks in advance,
Matthias

_____________________

Matthias Filez

Laboratory for Chemical Technology

Department of Chemical Engineering

Ghent University

 

Technologiepark 914, B-9052 Zwijnaarde, Belgium
+32 (0)9 33 11721
www.lct.ugent.be

Disclaimer: http://helpdesk.UGent.be/e-maildisclaimer.php


_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit