[Ifeffit] third cumulant

Matt Newville newville at cars.uchicago.edu
Mon May 7 06:33:27 CDT 2012

On May 6, 2012 11:01 PM, "JeongEunSuk" <eunsuk1986 at hotmail.com> wrote:
> Hello
> I measured temperature-dependent EXAFS at Pt L3 edge with Pt
nanoparticles in room and high temperature(400 C). I have some questions
about thermal vibration in EXAFS fit. I read that third and fourth
culmulants related with phase and amplitude in anharmonic term,
respectively. especially, I concerned third culmulant to relate with phase.
> As you know, the phase also relates with bonding length. So that, the
bonding length between Pt-Pt pair considerably correlated with third
culmulant. So I can't decide exact bonding length and third culmulant
because their correlation. I think that the relation of both bonding length
and third culmulant is similar to that of number and debye-waller factor.
> Is it right to find  bonding length and third culmulant like finding
number and debye-waller factor using k-weight?

Yes, c3 and R are correlated.  And, as it turns out, in almost exactly the
same way that N and sigma2 are.   That doesn't mean they can't be
determined accurately, though.   It will tend to make for larger
uncertainties, but this is taken into account in ifeffit (at least to first
order, ie assuming that the errors are normally distributed and a map of
chi-square would be ellipsoidal).

Many people vary the k-weighting to try to "break" these correlations.   I
think it doesn't really reduce the correlation that much (certainly not
below 50%), but it can't hurt.

You might also consider asserting some mathematical relationship for  R and
c3 with temperature, and fit the parameters of those relationships.
FWIW, my experience is that c3 is rarely significant (ie different from 0),
until you get to very high temperature. For 400C for Pt, I'd expect that it
would start to be noticeable.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20120507/9594fc71/attachment.html>

More information about the Ifeffit mailing list