[Ifeffit] Artemis

[Scott Calvin]

Hi Juan,

The same S02 should be OK; that's generally thought to be chemically 
transferable.

Unless I'm misunderstanding something about your system, there is no 
physical basis to justify the same delr. Think about it: why would 
both phases show the same difference in bond length between the model 
and the actual structure?

Using the same sigma2 is also problematic; it seems unlikely that the 
disorder and vibrational spectrum is the same in the two cases.

deltaE0, to first approximation, may be the same. In both phases the 
valence of the absorbing atom is the same (i.e. 0), so assuming the 
same E0 is not unreasonable.

To my mind, therefore, you need at least 6 parameters (plus whatever 
else you're using; e.g. the fraction of each phase present).

So I'll use this opportunity to beat one of my favorite drums: why 
not try a multishell fit? It sounds to me like you have phases that 
are likely to be fairly crystalline. If your data is decent, you may 
be able to extend your fit out to several coordination shells without 
adding many new free parameters. (You have to do this in a physically 
reasonable way: delr might need to be scaled by reff, etc.) This 
might be a net gain in degrees of freedom, and also might reduce 
uncertainties in some of the fitted variables by reducing correlations.

--Scott Calvin
Sarah Lawrence College

At 03:25 AM 11/9/2006, you wrote:


>Hi people!!
>
>I was fitting with Artemis bimetallic catalysts with two different 
>structures:
>platinum structure plus an intermetallic compound structure. As I only use a
>small range of R space (1st shell) for fitting, I have a small number of
>independents points and I used an appropriate number of variables. The
>question is: I used the same values of SO2, delR, sigma^2 and 
>deltaE0 for both
>structures, is that ok? or maybe I have to increase until 8 the number of
>variables?