[Ifeffit] How to refine the float

Matt Newville newville at cars.uchicago.edu
Thu Dec 20 06:57:54 CST 2012

Hi Zhaomo,

On Thu, Dec 20, 2012 at 1:15 AM, Zhaomo Tian <zhaomo1989 at postech.ac.kr> wrote:
> Dear XAFS community,
> I start my first EXAFS data analysis practice with Cu first shell fitting.
> I have 3 questions:
> 1.Usually what are the stratages I can use to refine the fitting?The
> floated parameters for each path will be automatically changed by
> some modifications,usually what are these modifications?

I think I don't fully understand the question.  The general approach
is that the Feff calculations have a close but imperfect model of the
EXAFS due to incorrect structural parameters, and the fit we do
refines the structural parameters to better match the experimental
data.  The most common things to modify (ie what Feff gets wrong
and/or what we put into the starting structural model) are N*S02, E0,
R, and sigma2.

> 2.I guess changing the value of fit variables or some experiment data
> parameters would be 2 stratages for fitting refinement, but the values of
>  them are usually limited in a very short range(cannot be changed much)
>  and when changing the values of them, I am also confused about what kind
> of changes can generate better fitting, I understand there are maybe no
>  routines for different cases, but are there some trends which can be
> helpful for getting an improved fitting? I fit the Cu first shell using
>  the first path which is generated by FEFF, the curve of first shell
>  fitting seems ok but the guessed parameter N (initiall guess N=12) is
>  never closed to12 but resulted to be some values around 7, I think there
>  must be some problems, could you give me a help?

Are you setting S02 to something other than 1?  How are you refining sigma2?

> 3.If there are more than 2 paths that give the same shell the comparable
>  contributions that both cannot be ignored, then how to deal with the
>  resulted different path parameters? For example, if we have to consider 2
>  major paths for the single shell, then after fitting we get one path
>  parameter Dr1=0.5A, and another path parameter Dr2=0.2A, and the Reff of
>  both paths are different too, then how to identify the final path
>  length(bond length) R for this shell?

The concept of "shell" is not very precise.  If you have two or more
paths making up a shell, it's generally better to include all the
paths in your description.

I know those are fairly general responses, but hope they are a good start.


More information about the Ifeffit mailing list