[Ifeffit] Fwd: average first shell position and width

[Matt Newville]

Hi Moneeb,

I'll bite at this one (and promise to carefully inspect the Send-To line!)

> Dear Dr. Ravel,
>  Hello,
>  I am Moneeb Shatnawi from physics department at Michigan State University,
>
> I have a PDF and XAFS data on GeSe system, and I would like to check if
> the PDF and XAFS results are consistent with each other in terms of the
> first PDF peak position and width.
>
> The first PDF peak in this system is a mixed shell of Se-Se bonds as
> well as Ge-Se bonds.
>
> The PDF data gave a width of the first PDF peak of about 0.06 A, and the
> XAFS data gave sigma^2 for Ge-Se bond is 0.006 A^2, and for Se-Se bond
> is 0.003 A^2.
>
> How is sigma in XAFS results (I was using Artemis program) related to
> the width of the first PDF peak?

It sounds like you're asking what the width of the PDF peak means....
As with any "width of the peak" measurement, this will depend on
details of both the measurement the physics. Since the PDF "G(r)"
(which is what I assume your "peaks" are in) I normally see is both
positive and negative,  I think defining exactly what a "peak" is
might be subtle.

So: as far as I understand, the g(R) that PDF measurements sample are
broadened by both the k-range of the measurement and the width of the
distribution atomic distances.  For a system (perhaps such as yours)
with a mixed distribution of atoms, that would include both the
distribution of all atoms.      Whether the total of the partial
g(r)'s or the k-range dominates the width of the PDF peak probably
depends on several experimental details.

The sigma2 gives the second moment of the Gaussian distribution that
best matches the partial g(R) to match the data.   The width of the
XAFS peak has essentially no physical meaning.

But, perhaps I misunderstood your question?

> Also, when obtaining the XAFS fits to the data, I was using: Theory
> --->  first shell fit.
> In this, I was giving r_eff for Se-Se bond to be 2.32 A and that for
> Ge-Se to be 2.368 A. Through the XAFS refinement, I obtained r, dr as
> well as r_eff.
> which of these should be considered as the actual bond length (r or r_eff)?
> and how are these related to the average position that I get from
> fitting the first PDF peak in this system?

The bond length is r_eff + dr.  That's the centroid of the
distribution that best matches the data.

How that relates to the g(R) measured by PDF is much a more
interesting question.  At some level, it will depend on how you model
the g(R) to best match the PDF data.

--Matt