[Ifeffit] XEAFS fitting problems

[Bruce Ravel]

On 01/30/2017 04:31 PM, Fuxiang Zhang wrote:
> Dear all,
> I am new for XAS analysis and now working on solid solution alloys with
> EXAFS measurements. Here are some general questions about EXAFS fitting
> that may need your help.
>
> 1. In NiCoCr alloys, I always get obvious smaller ss for the pairs of
> Ni-Cr or Co-Cr than Ni-Co (0.002 vs 0.008-0.011), when measuring the
> K-edge of Ni and Co, does it suggest that Cr is favorably bonded with Ni
> and Co?

Possibly, but I'd be skeptical.  That sort of alloy is not one of the 
strengths of XAFS.  The backscattering amplitudes of elements close in Z 
number are pretty similar, making it difficult in a numerical sense to 
distinguish them.  Z-numbers that differ by only 3 or 4 are very hard to 
distinguish in XAFS.

As a result, I would expect uncertainties on your sigma2 values to be 
rather large.  I would also expect correlations between the various 
amplitude and sigma2 parameters to be high.

As a result, I think you would have trouble making a strong and 
defensible argument.  I would certainly be a skeptical reviewer of a 
manuscript claiming this unless it had a very compelling argument backed 
up by strong data from another measurement or from theory.

> 2. In order to get the bond length of different pairs in binary AB
> system, can I use formula alpha1*Reff and alpha2*Reff to refine the
> length of pairs of A-A, A-B (simple cubic symmetry)?

The "alpha*reff" thing is a way of expressing all the DeltaR parameters 
of a purely cubic crystals using a single parameter.  That is the only 
situation in which that formula is, strictly speaking, applicable.

https://speakerdeck.com/bruceravel/discussion-of-the-fes2-exafs-analysis-example?slide=9

It is a mistake to think that there are magical formulas that can be 
memorized to make EXAFS analysis easy.  No such things exist.

Every way of parameterizing the terms in your EXAFS analysis in Artemis 
must be defensible in the context of the model you are using to try to 
understand your data.

What I mean by defensible is that you can explain to the reviewer of a 
paper or to a member of your thesis defense jury why the fitting model 
is reasonable, why you chose to use that parameterization, and how that 
model helps to answer the scientific question without introducing an 
unsupportable approximation or some other form of systematic uncertainty.

So, the answer to your specific question is "Maybe".

> 3. The selection of k range in Fourier transformation Parameters seems
> to have some effect on the fitting results, how to select the range in
> order to get reasonable results.

This has been discussed plenty in recent months on this mailing list and 
Scott discussed it at some length in his book.

Executive summary: include /data/ in the FT range, not noise.  If the 
measurement is dominated by statistical or systematic uncertainty after 
some value of k, don't use that part of the measurement.

HTH,
B

> Many thanks
>
> Fuxiang Zhang
>
>
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
> Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
>


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

  National Institute of Standards and Technology
  Synchrotron Science Group at NSLS-II
  Building 743, Room 114
  Upton NY, 11973

  Homepage:    http://bruceravel.github.io/home/
  Software:    https://github.com/bruceravel
  Demeter:     http://bruceravel.github.io/demeter/