[Ifeffit] Basic issues in EXAFS curve-fitting

[Bruce Ravel]

On Tuesday 21 June 2005 10:08, Stefano Ciurli wrote:
> Dealing with the old question of background correction, I have
> several DIFFERENT recipes from SK and others. Especially the
> alignment of theory and experiment. 

> Is there anyone willing to give me a definitive protocol? 

I should hope not!  What works for a metallic alloy probably won't work for a
uranyl.  What works when you have 18 inverse angstroms of great data probably
won't work when you have 9 of crappy data.

To my mind, the important thing is to understand that Ifeffit treats the
background in a Fourier sense.  The background is that part of the mu(E)
spectrum that contributes the low frequency components.  What does "low" mean?
That is the point of the Rbkg parameter in Athena and Autobk.  Components
below Rbkg are low and are the ones to be removed by the spline.  Components
above Rbkg are data and should not be removed by the spline.  However, since
the Fourier transform is finite, the separation between the frequency regimes
is not strict.  High components bleed into the low region and vice versa,
resulting in correlations between the spline and the parameters that will be
used to describe the data in Artemis.

Similarly, when you do a background corefinement in Artemis, the spline is
used to fit the Foruier components below Rmin and the feff paths are used to
fit the components between Rmin and Rmax.  The advantage offered by the
background corefinement is that it allows you to approximately quantify the
corellations bewteen the fitting parameters and the spline itself.

So, I would say there is no definative protocol.  There are certainly tricks
that you can try in certain situations, even recipes you can follow in certain
situations.  But in every case, the bottom line is that you should understand
that Ifeffit defines the background in a Fourier sense.  If you understand
that, there is very little else that should be mysterious in that part of the
analysis.


> In particular, SK writes that one should
> "read the data into Artemis": now, Artemis does not read the crude
> data and I have to go through Athena, as far as I understand. But
> when I read the athena.prj into Artemis, will the imported data be
> corrected already using the default values for the spline as provided
> in Athena? If so, there is already an initial bias, I think.

Disclaimer: I am going to speak to how things are done in the development
version of Artemis.  This is relevant since the next release of my codes will
drop the 0.7 branch and the development branch and since the development
branch will become the main, 0.8 branch.

Artemis reads data and all the background removal parameters from the Athena
project file.  It then performs the background removal in the exact same way
as in Athena and inserts that chi(k) into the project.  So whatever decisions
you made in Athena, those decisions get transplanted to Artemis.

In the near future, there will be the possibility of tweaking the background
removal from within Artemis.  That will serve to close the loop that Shelly
often describes in which she returns to Athena with new knowledge about the
data, re-performs the background removal, then returns to the data analysis in
Artemis.  With this soon-to-come feature, the same loop will be possible, but
it will be possible to do it entirely within Artemis.

B


-- 
 Bruce Ravel  ----------------------------------- bravel at anl.gov  -or-
						  ravel at phys.washington.edu
 Environmental Research Division, Building 203, Room E-165
 Argonne National Laboratory                        phone: (1) 630 252 5033
 Argonne IL 60439, USA                                fax: (1) 630 252 9793

 My homepage:    http://feff.phys.washington.edu/~ravel 
 EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/