Dear Matt,
>I'm still confused about the right way to handle these in practice. It
>seems that
>a choice of "arc-tangent", "lorenztian", or "mimic the shape of the main edge"
>should work for most cases -- do you agree? It also seems there ought to be
>a reasonable way for the algorithm to pick which of these lineshapes to use.
Maybe you`ll find the two graphs in the attachment helpful - they are
from our
paper cited at the top and also accessible on Iztok Arcon`s homepage. The
first graph shows the usual range of EXAFS (~ 1000 eV above the edge) on the
horizontal scale, and the atomic number on the vertical scale, with
inserted energies
of the main MPE. You can see that for 3d elements (Ti-Zn) the EXAFS range is
almost free of MPE, the prominent 1s3p still within 100 eV of the edge, and the
weak 1s2p at the end of the usual EXAFS range.
For 4p elements(Ga-Kr) the 1s3p slowly moves into the EXAFS region. And beyond
Kr, the strong 1s3d excitation gets beyond the 100 eV mark, so that two
prominent
MPE are to be taken into account.
The L EXAFS are, as seen from the graph, practically all infested by MPE,
so that
there are no "clean" or "safe" ranges of Z.
In the second graph, the "real" atomic backgrounds, extracted from EXAFS
spectra,
are shown for 4p elements. ALthough very noisy, they show what kind of an
ansatz
would be necessary to approximate at least crudely their shape. I imagine
that two
functional terms with proper energy position and amplitude (these are mostly
transferable, prepared in tables) but with a small common fittable energy
shift (to account for different
energy calibrations or small chemical shifts). To allow the usual sloppines
in the practical
EXAFS analysis regarding the use of proper "matrix" subtraction etc, it
would be
sensible to retain some smooth background function, either a spline or a
polynomial,
but with much less free parameters. My feeling is that a 4-interval spline
or 4th degree
polynomial should be sufficient since the spline would not need to simulate the
sharp jumps of MPE.
Best wishes
Lojz
