[Ifeffit] Athena open data project but not shown the curve

Robert Gordon ragordon at alumni.sfu.ca
Thu Nov 5 15:59:20 CST 2015


Hi Carlos,

My previous exposure to MD for EXAFS was as part of a collaboration here:
J. Synchrotron Rad. (2013). 20, 555–566

There, to get estimates of energy shift and amplitude (So^2), a 
simplified model was used
to fit the experimental data. In your case, that could be done by 
performing a FEFF
calculation for an atom best representing the average.

The FEFF calculation won't include deviations in Eo or So^2 or even
temperature effects by default. DEBYE, s02 (or HOLE or EDGE), and 
CORRECTIONS cards
could be used in generating the FEFF chi.dat.

  If your MD simulation gives you N, R and sig^2
for each shell (and each atom) you wish to include in your comparison, then
you can implement the Eo-shift and amplitude values from the simplified 
fit in each shell
of a simulation (using the FEFFNNNN.dat files) up to whatever shell you 
have the energy to try
to include in the simulation...I would suggest starting with the first shell
and include the second once you get a good agreement with the first.

There may still be deviations in peak position between experiment and 
calculation.
That can be considered by examination of a known standard.

You should consider the same k-space ranges for experiment and theory 
when comparing
your own results...comparison itself would be done in R-space, 
particularly if only using
a few shells.

  For comparison with the results of others, it would depend on 
available data
ranges and quality as to how to proceed. e.g. If I analyse 3 to 18 A-1 
and want to compare to
someone else who analysed 3 to 10 A-1 due to their data limits, then I 
should also
look at my results over their range to give a fair comparison. If the 
reverse situation is present,
then the comparison is only within the limits of available data.

regards,
Robert

On 11/5/2015 10:31 AM, Carlos Triana Estupinan wrote:
> Hello
> ROBERT
>
> Thanks for your kind answer
> I have an amorphous structure with 216 atoms, 72 of them are absorbing 
> atoms. So, I did FEFF calculation for each of those absorber to get 
> individual EXAFS, and then I dis an average of them to get an averaged 
> EXAFS (total EXAFS).
> I don't want to refine a structural model, but I want to compare the 
> experimental EXAFS data with the averaged EXAFS I got for my model.
>
> My question is, if I get the EXAFS just from FEFF calculation (chi(k) 
> or xmu.dat), then I can import it to artemis, and compare with the 
> experimental (but without do any fiting), just to compare EXAFSs. From 
> FEFF I can obtain chi(k), u, u0, and X as a function of energy. Which 
> of those (chi(k), u, u0, and X) should I compare with the experimental 
> data.
>
> Also, if I compare mi experimental EXAFS with other EXAFS spectra 
> already published in papers, should be the EXAFS the same or it 
> depends on the range for the fourier transform.
>
> regards.
>
> 2015-11-05 18:50 GMT+01:00 Robert Gordon <ragordon at alumni.sfu.ca 
> <mailto:ragordon at alumni.sfu.ca>>:
>
>     Hi Carlos,
>
>     Please clarify somethings for me regarding your inquiry.
>     You have a structural model and you have data.
>     You calculated a FEFF simulation of your structural model.
>     You did not use the FEFFNNNN.dat files to try to fit the model to
>     your data,
>     but instead just did a comparison of the simulated chi(k), presumably
>     with just some default settings in FEFF.inp in addition to your
>     structural model.
>     Is that correct?
>
>     Generally, it is the fit to the data that gives one feedback on
>     the validity
>     of the structural model (GIGO rules apply).
>
>     If you have a model, then you have a model coordination number and
>     can use
>     a program (e.g. Artemis) to refine amplitude, distance,
>     mean-square-relative displacement
>     and energy shift. Do you need explanations of what these
>     parameters represent?
>
>     So, to address what you were wondering, I believe the answer is
>     yes...you need to actually
>     do a fit, and the goal isn't to get the obtained spectrum to look
>     like your simulation.
>     Data is sacred. The goal is find a model that best reproduces the
>     data (within a confidence
>     level governed by the goodness of fit).
>
>     regards,
>     Robert
>
>
>     On 11/5/2015 7:06 AM, Carlos Triana Estupinan wrote:
>>     Dear
>>     Bruce
>>
>>     I am doing calculation of EXFAS for a amorphous structure created
>>     by molecular dynamics. I am using FEFE to obtain chi(k) and the
>>     file xmu, however, when comparing with experimental data with the
>>     calculated EXFAS I have the following problems.
>>
>>     1) The Fourier transform of the EXAFS calculated by FEFF has the
>>     same shape that the experiment spectra, however it is higher in
>>     amplitude.
>>
>>     2) The calculated EXFAS by FEFF has the same oscillation in
>>     K-space, but it does not match the oscillation in K (both of them
>>     are shifted to each other).
>>
>>     I am wondering if I need to do an extra correction to the
>>     calculated EXAFS by FEFF in order to match the K-space and the
>>     amplitude in R-sapace. Shall I change the energy shift or the
>>     Edge step to match the experimental spectra. I so, how can I am
>>     sure about the obtained spectra really correspond to my molecular
>>     dynamics simulation.
>>
>>     Thanks for your help.
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20151105/19740483/attachment.html>


More information about the Ifeffit mailing list