[Ifeffit] How to distinguish whether the coordination element is heavy or light

[Matt Newville]

Hi Zhan,


On Mon, Jul 7, 2014 at 3:40 AM, ZHAN Fei <zhanfei at ihep.ac.cn> wrote:

> Hi Matti，
>  Thanks for your attention and patience.
>  The back fourier transform of the specific peak (use window)of Chi(R)
> gives the amplitude,donated by amp_bft。As in previous mails,I ask whether
> the compare the amp_bft and the amplitude of specific Z number element can
> determine the Z.
>

Sorry, I'm not sure I understand this.  I didn't understand the figure you
attached in your earlier mail.   Generally speaking, I find
back-transformed data to be not very useful -- making any sense of these
requires very well isolated shells of atoms.   The amplitude of the
back-transformed chi(k)  (is that what you mean by amp_bft??) has many
contributions, and is not simply f(k).  It will have Z dependence, but it
will have other dependencies too.


> And thanks for telling me the useful trick using the total phase shift of
> the specific element.
>
>
> The discrepancy between R and Rphcor is below,the Zn and Br is close to
> Se(the best fit)'s 0.013,
>

Yes, Ge actually gives the closest match, and Se the second closest match,
and I left Ga out of the test.  I would probably say that anything closer
that 0.015 Ang (and, really, maybe 0.02 Ang) is pretty darn close.   So the
phase-correction approach appears to be (in this case) not as sensitive to
Z as the reduced chi-square, but does provides a check on
self-consistency.    The fits with Zn and Rb are noticeably worse than the
fit with Se... hence Z +/- 3 or perhaps 5 seems like a reasonable rule of
thumb, and sometimes one might be able to do better.


> should the enot also be important criterion in this trick?and dose the
> plus or minus of enot  indicate the lighter or higher element relative to
> the specific coordination shell?
>

Well, an E0 shift > 10 eV probably indicating that something is off ....
but that could be just the selection of E0 for the experimental data.   I
wouldn't put much meaning of the absolute value of E0 though for any single
fit.

  "a bond valence sum can be an independent check on the consistency of N,
> R, and valence",can your give the ref. paper of this method? Thanks
>
>
http://en.wikipedia.org/wiki/Bond_valence_method   The idea is that N, R,
and valence are not independent.

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20140707/76abe41c/attachment.html>