[Ifeffit] a cautionary tale

[Jeff Terry]

Peter,

Your observations are correct. I always run the feff calculation 
multiple times, adding atoms to the feff input file each calculation 
until the results converge.

Jeff



On May 20, 2004, at 8:15 PM, Peter Southon wrote:

> Hello fellow Ifeffiters
>
> This a story about making feff.inp input files for FEFF. The more 
> experienced readers of this list will probably be well aware of the 
> issues involved, but I thought that relative newcomers such as myself 
> might not know (which is why I'm posting this on the Ifefft list 
> rather than the Feff list).
> In a nutshell, it is wise to create a large cluster of atoms to put 
> into Feff (ie use a large Rmax in Atoms), but limit the number of 
> paths with a reduced Rmax in Feff.
>
> I used Artemis/Atoms to make up a Feff input file (feff.inp) from 
> crystallographic parameters, setting Rmax in Atoms to a relatively 
> large value of 7.5 Å to create quite a large cluster of atoms. I then 
> used FEFF to calculate scattering paths within the cluster, but first 
> reduced the Rmax in feff.inp to 5.5 Å to limit the number of paths. So 
> far so good.
>
> I then (for reasons too involved to discuss here) edited the feff.inp 
> file to delete all atoms more than 5.5 Å from the centre, and ran Feff 
> again.. The results (judging from the relative amplitude values) were 
> different, and in the case of some paths very different. Some intense 
> paths changed in amplitude by a factor of 2 or greater, although I 
> haven't looked at the finer details of the Feff results. The paths 
> that changed the most were high-order multiple scattering paths, and 
> surprisingly those paths did not involve atoms at the edges of the 5.5 
> Å radius.
>
> Now I'm sure that someone else can explain the reasons for this much 
> better than I can, but I assume that the potentials for the atoms 
> outside the 5.5 Å radius (which were neglected in the second 
> calculations) significantly affect those inside the radius.
> I know that its always been my practice to perform the Feff 
> calculation using only those atoms that I think are contributing to my 
> spectrum, and I'm sure that others do the same. My advice is to use a 
> larger cluster of atoms in Feff if you can - but as to how large I 
> might leave that to someone else to comment.
>
> Regards,
> Peter
>
> PS - and four cheers to Bruce for improvements to Artemis, I can't 
> wait for the Windows update.
>
>
>
> Peter Southon
> Post Doctoral Fellow
> School of Chemistry
> University of Sydney
> NSW 2006, Australia
> +61 2 9351 4425
>
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit