[Ifeffit] S02-Interpretation

Kelly, Shelly D. SKelly at anl.gov
Wed Apr 14 10:17:20 CDT 2004

> To my knowledge, S02 accounts for intrinsic as well as 
> extrinsic losses in the 
> sample. Now, if I have two samples and determined the S02 
> value according to 
> the way which Bruce has in his supplement to the FEFFIT 
> course, I find two 
> differing values for electrochemically prepared oxide (0.74) 
> and crystalline 
> gold oxide (0.92). This difference is independent of 
> beamlines I measured and 
> I can exclude quite safely that it is just a measurement 
> error - it comes up 
> with any spectrum I recorded on these systems.
> When discussing this with my colleagues, we interpreted the 
> S02 difference as 
> an indication of more disorder in case of the electrochemical 
> system (which 
> would well fit to my picture of the whole thing). 
> BUT: Is there any physical reason to assign a difference in S02 (for 
> "chemically equivalent" systems, where just the preparation 
> is different) to 
> a structural disorder?
> Regards,
> Norbert

Hi Norbert,

I believe that the value for S02 should be the same in these two
samples.  But S02 is not the only amplitude term in the EXAFS equation.
The degeneracy of the path, sigma2, and the step height normalization
are also amplitude terms.  One needs to be careful in processing the
data so that the change in the amplitude between the two signals really
belongs to the correct parameter.  

Here is a recipe for how I would try to solve the problem:

1st:  Double check the step height normalization in Athena.  1) Make
sure that Ezero is the same for both data set and that the data sets are
aligned in energy.  2) Plot the xmu data along with the pre-edge and
post-edge lines and  check the step height that Athena reports...make
sure it seems reasonable.  3) plot both chi(k) data sets with kweight of
1, and then 2 and then 3.  You are looking to see if the change in
amplitude between the data sets is the same for all k-range or if the
ampliude changes more at high or low k.  4) you could also Fourier
filter the data and do this again.  If the change in amplitude is the
same for all k, then the difference is because of SO2 or degeneracy or
step height normalization.   If the change in amplitude is more at high
k than at low k the change is because of sigma2 term.  If the change is
more at low k than at high k then I would say different bkg curves or
self absorption effects.  

2nd:  Say you think that the difference is due to a sigma2 term.  I
would fit both data sets with the same parameters for SO2 and different
parameters for sigma2 for what ever paths you decide to include in your
fit.  If you want to prove that the change is a sigma2 and not a SO2
parameter then you do the fit using kweight of 1, and 2 and 3 all at the
same time for both data sets, trying two different models, 1) where SO2
is different for both data sets but sigma2 is the same, and 2) another
where sigma2 is different but SO2 is the same.  You should see a
difference in the goodness-of-fit parameters that is greater than one
standard deviation.  Using all three k-weights helps Artemis "see" the
changes that you saw in Athena, by comparing the two data sets with the
different k-weights.

I just read Bruce's comment, If you think that the change should be in
the coordination number then you should repeat the above models with the
coordination numbers allowed to vary.  To decouple SO2 and N you need to
have more than one shell in the fit, otherwise they are exactly the same


More information about the Ifeffit mailing list