[Ifeffit] a basic (?) EXAFS question

Scott Calvin SCalvin at slc.edu
Mon Apr 18 08:31:02 CDT 2005


Hi Juro,

Just a note of caution--if I've followed this discussion, you're 
fitting the amplitude for each shell separately. While there are 
sometimes good reasons to do this (it's the way Anatoly Frenkel 
determines crystallite size and morphology, for example), if you 
think you have fairly ordered, fairly large (> 10 nm, say) 
crystallites, then you need a physical justification. Although the 
match between fit and data will of course usually look marginally 
worse by using the same amplitude for every shell, it often has the 
side benefit of reducing uncertainties in the guessed parameters.

In terms of fitting strategy, when I was an EXAFS neophyte I would 
start my fits by having lots and lots of parameters and then would 
gradually reduce the number. Since then I've switched around to what 
most people recommend, which is starting with very few parameters 
(many recommend starting with none!), and then adding guessed 
parameters as necessary. The latter has the benefit of forcing you to 
think about what's physically reasonable, reducing the incidence of 
bizarre "false minima," giving you a sense of what's going on early 
in the process, and being easier to defend in print.

--Scott Calvin
Sarah Lawrence College

>Matt,
>
>thanks a lot for the few simple questions. One of them -
>the R range - seems to be the problem, since the range
>wasn't big enough. Oh well ... The structure of loellingite
>has very well defined shells, up to the fifth one.
>
>Thanks again,
>
>Juro
>
>  > On Sun, 17 Apr 2005, Juraj Majzlan wrote:
>  >
>  >
>  > > I have a question about fitting an EXAFS dataset. The
>>  > compound is FeAs2, crystalline, and will be used as a
>>  model
>>  > for further study. I was able to fit the first three
>>  shells
>>  > very well. However, when I try to fit the fourth shell,
>>  I
>>  > either get amplitude that is too high (> 2, with large
>>  > errors), or 0.0. There are several single-scattering
>>  and
>>  > several multiple-scattering paths in that region.
>>  > Sometimes, only excluding one of the MS paths leads
>>  > immediately from amplitude of >2 to 0.0.
>  >



More information about the Ifeffit mailing list