[Ifeffit] XAFS Analysis using Ifeffit

Matt Newville newville at cars.uchicago.edu
Mon Jul 31 10:41:18 CDT 2006

Hi Francois,

Thanks for the comments about GNXAS and MEE. It's possible that I'm
just not understanding some of your points, but here are some
replies to your comments:

> I've seen, in contrast, papers in which people were
> claiming so much about feff (so-call "good and bad things") but their
> calculations were far from optimum (they are never, I know for sure myself
> !) so their feff "evaluation" was quite very poor in fact.

Does GNXAS prevent one from making mistakes?  It sounds like it's more
likely to prevents novices from doing anything.  It's hard for me to see that
as an advantage.  My experience is that novices tend to be very careful,
and more respectful of the process (or perhaps just have a more difficult
time publishing sloppy analysis), and that it's the so-called experts claiming
to do "state of the art" EXAFS who are most likely to make mistakes and/or
over-interpret their data.

> But basically, a modernized front end of gnxas would be great. But one has
> to get into that programming to make it easy-to-click (like many would
> like too). With Atoms of Bruce Ravel, there are options to prepare these
> so-odd files needed for gnxas. this is already a wonderful first step.
> Any decent model of the EXAFS for high Z cations should be done with that
> program, basically. Sixpack, ifeffit, but mostly FEFF, are missing the
> multi-electronic excitation (MEE) modeling.

Hmm,  if I understand right, GNXAS "models MEE" by allowing the user to
add arc-tangents or lorentzians (maybe other lineshapes??).  Do they
do something more sophisticated than that?

And yet, it seems that you're suggesting that to correctly analyze "high Z
cations", one must include MEE effects, and so that GNXAS does a much
better job at modeling EXAFS than Feff/Ifeffit.   Is that a fair reading?

For what it's worth, Arc-tans for MEE have been in an out of autobk a couple
times -- it has never been advertised and is currently taken out because it is
too easy to abuse, and never made a big difference in the structural results.
>From a few different conversations at XAFS13, I've agreed to put something
like this in to Ifeffit. Personally, I'd like to see more data with
clear MEE to play
with, and would like to model these better than with an ad hoc arc-tanget.  I'd
also like to see a real case where including MEE transitions gave different
structural results than an analysis that did not include them.  Any takers?

> But most users don't even "smell" these MEE's (or AXAFS if you prefer) in
> their data and process them straight away. they cannnot be 100% blamed,
> MEE are also quite vicious features. but more and more is done that is not
> appropriate because of these nice GIUs (athena, sixpack). so it is
> difficult to decide what to do.

Again, you seem to be saying that people are doing inappropriate analysis
because they're ignoring MEE.   It also seems that your concerned about
having GUIs for EXAFS analysis. Is this a fair reading of your statements?
It probably no surprise that I would not agree with either of these points.

> Also, like others, I ve found that MEE are more and more visible as much
> as the metal concentration probed is getting lower and lower (comparative
> study of Mn permanganate solutions at more and more dilutions makes that
> the MEE is increadingly more detected in the normalized EXAFS). And this
> now standard to look at diluted metals, in environmental samples with
> these 3rd generation, highly-stabilized sources (SLS for instance).

Well, I'm pleased to hear you're able to getting good data, but I'm confused
by MEE  in Mn K edge of permanganate.   What transition are you seeing?

When Arcon and Kodre talk about the MEE in Br and Kr K-edges, they see
1s+4p transitions: the 4p levels are empty for Mn.  When D'Angelo sees MEE
in Sr,  it's from the 3d M4/M5 edge energies,  which are at ~5eV for Mn,  and
so too low to matter (especially in permanganate, where these orbitals are
hybridized with O 2p levels).   The M1(3s) transition would be ~80eV for Mn,
but the 3s MEE is hard to see in vapor data, so I'd be skeptical that one could
see this in a system where one had EXAFS to analyze.  The M2/M3  edges are
~45eV for Mn.  I  don't see anything obvious at this energy in permanganate,
but I'd guess it to be at a low enough energy to not effect the EXAFS.

As I understand MEE (not very well!??),  the MEE peaks should be mirrors of
the main absorption edge.  For permanganate,  there should be a strong
"pre-edge MEE peak" as well as a "main MEE peak".  Do you see this?



More information about the Ifeffit mailing list