On 08/31/2016 09:05 AM, Sebastian George wrote:

I hope that I am writing this in the correct place, and if I am not, I apologize. The problem I'm having is pretty much summed up in the Subject line. I'm analyzing some experimental EXAFS data in Athena, and it seems that ticking the "phase correction" box has no effect on my Forward Fourier transform. Has anybody run into this problem, and if so, were you able to fix it? The material I'm studying is amorphous SmCo. Could it be that having multiple backscattering species or a lack of crystalline structure causing problems for the phase correction?

You didn't think to say so, but I am assuming you are using a Windows computer. The phase correction has been broken for a while in the Windows installer. I noticed that last week. Happily, I noticed it in time for the recently posted 0.9.25 release of the Windows installer. If you got to the Demeter homepage and grab the latest, you should find that phase correction now works. If any of my assumptions were wrong (understandable given that you provided little information), let me know and I will continue to look into the problem.

More generally, I'm wondering how exactly this phase correction is determined anyway. If used, is it the sort of thing that can put the first peak in my FT fairly close to the true nearest neighbor distance?

In Athena, nothing is known about the scatterers, so the phase correction is only for the central atom part of the phase shift in the EXAFS equation. The central atom phase shift data used by Athena is simply a static tabulation computed using Feff6.

Perhaps a better phrased question is: when can the phase transformation be expected to give a quantitatively accurate radial distribution (in terms of peak positions),

Basically never. In rare situations, it might be possible that the first peak of a phase-corrected FT might approximate an RDF for the first neighbor in some sense. But I caution you against thinking about it that way. You are more likely to be wrong than right. Here's a dose of wisdom from Shelly from many moons ago: http://www.mail-archive.com/ifeffit%40millenia.cars.aps.anl.gov/msg00171.htm...

and when is it more likely to only be qualitatively useful?

Always. The Fourier transform of chi(k) is not a radial distribution function. I am going to say that again all caps because this point is so important that I want to yell it at you. THE FOURIER TRANSFORM OF CHI(K) IS NOT A RADIAL DISTRIBUTION FUNCTION. * chi(k) includes multiple scattering * chi(k) includes a complex scattering factor * For heavy elements, the magnitude of the complex scattering factor has a lot of structure, resulting in a sort of Ramsauer-Townsend effect -- see https://speakerdeck.com/bruceravel/the-ramsauer-townsend-effect-in-x-ray-abs... * chi(k) can have scattering atoms at very close distances and the contributions from those scatterers can overlap in surprising ways * chi(k) has a sigma^2 term * chi(k) has a mean-free-path term For all of these reasons, the FT of chi(k) is just chi(R). It is not a radial distribution function. That said, when we analyze chi(R) in Artemis, we extract information from the fit that can be used to reconstruct partial pair distribution functions.

Also, I would have assumed that a backscattering species would need to be given in order to calculate the phase shift, but I haven't found anywhere to do that. Have I missed something?

Nope. But in Athena, nothing is yet known about the scatterers. Correcting the FT for the full phase shift is a chore that cannot be done until something is known about them. B -- Bruce Ravel ------------------------------------ bravel@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/