[Ifeffit] Right way of choosing E0 in Athena

Matt Newville newville at cars.uchicago.edu
Thu Jun 29 17:06:28 CDT 2006

Vadim, Gerrit,

For Vadim's question:
> > I have the following question: I have a set of spectra collected in
> > transmission with 3 ionization chambers and a standard between 2nd and 3rd
> > ion chambers for calibration, so I know my spectra are nicely calibrated
> > relative to each other. Should I then force Athena(/autobk) to use the same
> > E0 for all of them (e.g. use the atomic value) or let it find E0
> > independently for each sample? Thanks!
> > Vadim.

Since you know the spectra are well calibrated relative to one another, I would
use a single E0 for all background subtractions.  In such a situation,
I tend to
play with the background parameters for one of the spectra and then apply these
background parameters to all other spectra.  The most important parameters are
typically: E0, rbkg, kmin, kweight, and the high-k clamp.

For Gerrit's questions:

> The initial question was how to choose a value for E0 when comparing
> samples with mixtures and the measurements of the pure compounds (e.g.
> iron, iron-oxide and a sample containing a mixture of both). Usually in
> this case I set E0 to the atomic value for all samples (Assuming I can
> be sure that the monochromator was calibrated to return a correct
> absolute energy. Otherwise I do an appropriate energy-calibration with
> reference to a standard measured at the same time before entering
> Athena). I do that to ensure a uniform background subtraction for all
> samples. However, for a system of e.g. Se, RuSe2 and a mixture of both,
> the atomic value E0 can be quite far away from the inflection point of
> the RuSe2- and mixture-samples. So the first question would be: is it
> better to use the same E0 for all samples or to somehow adjust the E0 to
> the inflection point of the specific sample?

I think there are two different answers (yes and no) depending on how you
are treating the data:

For linear combination fitting (XANES or EXAFS), I would choose a consistent
absolute E0 for all data sets (after appropriate shifts to align the data).

For EXAFS analysis with Feff, I would not worry about choosing a consistent
E0 -- each Feff calculation puts in an assumption about E0 that has to be
refined away anyway.

> In this context my idea of the algorithm of background subtraction was
> that a shift of E0 could be completely compensated for by appropriatly
> shifting all other parameters in the backward direction. E.g. when
> shifting E0 by 4 eV I should end with the same FT when shifting the
> values for pre-edge range, normalization range, spline range and k-range
> by the same amount. However, a test revealed that only the pre-edge
> lines and post-edge lines are identical. The background splines are
> different and the FTs differ quite significantly from each other. Is my
> general assumption wrong?

I think so (that your general assumption is wrong).   E0 sets the definition of
k=0, so can have a large impact on the FT and the background subtraction.

> Another question (more cosmetic) concerns the use of 'E0' and 'E shift'
> as it is irrelevant which parameter to adjust to achieve an energy
> shift. Is it a good idea to set E0 always to the atomic value and use 'E
> shift' for entering the offset? Actually this is the way I prefer to do
> it although Athena initially does not set 'E shift'. I also have a kind
> of feature request in this context and would be interested if others
> would also benefit from it: for some samples the initial value Athena
> picks for E0 is obviously inadequat, e.g. when it chooses a prepeak as
> inflection point. As a consequence the parameters for background
> subtraction are also inadequate. After choosing a right value for E0 a
> new feature could be: "automatically adjust background parameters with
> regard to the current E0", so one could still benefit from Athenas
> auto-function. Usually I adjust these parameters anyway but it would
> simplify doing a quick first look at the data.

I'm not  sure I understand all that.  For non-metals, I'd ignore the
"atomic value",
since it's wrong  (in some sense, that's why we're doing XAFS).   For what it's
worth, I only use the Energy Shift to align spectra.

And, yes, for spectra with large pre-edge peaks, the guessed e0 can be wrong:
if you have ideas for how to improve this situation, let me know!
But once e0
is set, the rest of athena's background subtraction should still be
Is it not, or am I not understanding something?


More information about the Ifeffit mailing list