You're describing the way many of us, including me, used to analyze data before FEFF became as reliable and easy to use (courtesy of
Artemis and the like) as it is now. As shown in the paper,
you extract amplitude and phase from model compounds, by back-transforming filtered shells, then use those to fit data. Multiple
data sets aren't new, either; I used to do that with data taken on
dilute solid solutions at different temperatures. Another trick I used to use a lot was if I didn't have a standard for the right
pair of atoms, I'd use FEFF to get the difference in scattering factors (phase+amp)
between the pair I wanted and a nearby (in Z) pair I had data for. For instance, you can get CuAl2 as a compound and extract the
phase+amp for Cu as the central atom and Al as the scatterer. You can't do
that for Cu and Si because there's no Si-rich intermetallic in the CuSi system. Therefore, you can do:
Amp(Cu->Si) = Amp_exp(Cu->Al)*Amp_Feff(Cu->Si)/Amp_Feff(Cu->Al)
Phi(Cu->Si) = Phi_exp(Cu->Al)+Phi_Feff(Cu->Si)-Phi_Feff(Cu->Al)
and get a "semi-empirical" Cu->Si "standard". Why the +- for phase and the */ for amp? You can think of it as extrapolating the
log of a complex signal chi~(k) = Amp(k)*exp(i*Phi(k)), of which the measured chi is the Im().
This expresses everything in terms of relatively slowly-varying quantities and treats amp and phi as parts of the same quantity,
which they really are.
One of the programs available from the ALS beamline 10.3.2 website http://xraysweb.lbl.gov/uxas/Beamline/Software/Software.htm is
called EXAFSfit and does this sort of fitting to amp and phase. You feed it amp and phases which
come from the FT program, also available at the website. The amp and phase files are simple 2-column ASCII, thus readable anywhere.
mam
----- Original Message -----
From: "Frenkel, Anatoly"
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit