[howto] configurational average in EXAFS with FEFF
Dear all,
I would like your point of view on how to approach configurational
average in EXAFS analysis with FEFF (let's not consider GNXAS for the
moment).
In particular, I'm trying to combine ab-initio DFT numerical simulations
with EXAFS for doped semiconductors where in the minimized cluster there
are N absorbing atoms.
Usually I don't use the CFAVERAGE card because fractional degeneracies
are not taken into account and, to have a qualitative analysis, I
calculate an "extended"-XANES (up to 8.0 A^-1) for each cluster - where
each absorbing atom is shifted to (0,0,0) - and then I average over all
single-cluster signals.
Well, at this point my question is: how to proceed in the quantitative
analysis with Ifeffit? In fact, considering only single-scattering
expansion for all the clusters the result could be seen as a radial
distribution for each scattering pair where the integral gives the
coordination number and the FWHM a structural Debye-Waller factor. How
to combine these informations?
Thanks for the answers,
Mauro
--
Mauro Rovezzi
Mauro,
If I do understand your question correctly, you may be asking how to handle a multiple of inequivalent sites in atomic clusters with N atoms, where each atom generates a unique sequence of single- and multiple-scattering paths. A possible solution is given in this article:
D. Glasner and A. I. Frenkel
Geometrical characteristics of regular polyhedra: Application to EXAFS studies of nanoclusters
AIP Conf. Proc. 882 , 746-748 (2007). https://exchange2000.bnl.gov/exchange/frenkel/Drafts/RE:%20%5BIfeffit%5D%20%...
The PDF is here:
http://pubweb.bnl.gov/users/frenkel/www/XAFS13-Proc/clusters-geometry.pdf
It is the same type of averaging and radial distribution analysis approach that you are proposing unless I misunderstood your question.
Anatoly
________________________________
From: ifeffit-bounces@millenia.cars.aps.anl.gov on behalf of Mauro Rovezzi
Sent: Tue 3/27/2007 9:38 AM
To: Ifeffit-ML
Subject: [Ifeffit] [howto] configurational average in EXAFS with FEFF
Dear all,
I would like your point of view on how to approach configurational
average in EXAFS analysis with FEFF (let's not consider GNXAS for the
moment).
In particular, I'm trying to combine ab-initio DFT numerical simulations
with EXAFS for doped semiconductors where in the minimized cluster there
are N absorbing atoms.
Usually I don't use the CFAVERAGE card because fractional degeneracies
are not taken into account and, to have a qualitative analysis, I
calculate an "extended"-XANES (up to 8.0 A^-1) for each cluster - where
each absorbing atom is shifted to (0,0,0) - and then I average over all
single-cluster signals.
Well, at this point my question is: how to proceed in the quantitative
analysis with Ifeffit? In fact, considering only single-scattering
expansion for all the clusters the result could be seen as a radial
distribution for each scattering pair where the integral gives the
coordination number and the FWHM a structural Debye-Waller factor. How
to combine these informations?
Thanks for the answers,
Mauro
--
Mauro Rovezzi
Anatoly, I don't quote your answer because I would like to add some more elements to specify what I'm searching for. In my previous post I have used the term "cluster" in a wrong way, that is I was thinking at the crystal-box (normally less than 216 atoms) used in the ab-initio density functional theory calculations and "doped" with few percent (< 10%) of an absorbing atoms. For example, if we put 5 absorbing atoms in a box of 64 (that represent the crystal host) we have 5 possible configurations for our FEFF calculation and is impossible in an EXAFS fit to include all the important paths (too many!). My idea to take into account the numerical simulation and combine it with the experimental EXAFS data could be to use the pair radial distribution function extracted from the simulation, average the configurations, recalculate paths and proceed in the fit. Well, I hope now is more clear... In any case, thank you for the useful article! Mauro Frenkel, Anatoly wrote:
Mauro,
If I do understand your question correctly, you may be asking how to handle a multiple of inequivalent sites in atomic clusters with N atoms, where each atom generates a unique sequence of single- and multiple-scattering paths. A possible solution is given in this article:
D. Glasner and A. I. Frenkel Geometrical characteristics of regular polyhedra: Application to EXAFS studies of nanoclusters AIP Conf. Proc. 882 , 746-748 (2007). https://exchange2000.bnl.gov/exchange/frenkel/Drafts/RE:%20%5BIfeffit%5D%20%...
The PDF is here:
http://pubweb.bnl.gov/users/frenkel/www/XAFS13-Proc/clusters-geometry.pdf
It is the same type of averaging and radial distribution analysis approach that you are proposing unless I misunderstood your question.
Anatoly
--
Mauro Rovezzi
On Tuesday 27 March 2007 14:34, Mauro Rovezzi wrote:
Anatoly,
I don't quote your answer because I would like to add some more elements to specify what I'm searching for.
In my previous post I have used the term "cluster" in a wrong way, that is I was thinking at the crystal-box (normally less than 216 atoms) used in the ab-initio density functional theory calculations and "doped" with few percent (< 10%) of an absorbing atoms. For example, if we put 5 absorbing atoms in a box of 64 (that represent the crystal host) we have 5 possible configurations for our FEFF calculation and is impossible in an EXAFS fit to include all the important paths (too many!).
My idea to take into account the numerical simulation and combine it with the experimental EXAFS data could be to use the pair radial distribution function extracted from the simulation, average the configurations, recalculate paths and proceed in the fit.
Mauro, If I understand what you are trying to do, this paper by me might be relevant: Role of local disorder in the dielectric response of BaTaO2N, B. Ravel, Y-I. Kim, P.M. Woodward, and C.M. Fang, Physical Review B, 73, p. 184121 (2006) I used a DFT result as the basis of my fitting model in that paper. The naive approach was to make a Feff path from every absorber/ scatterer pair in the box. That's certainly a good idea, but Ifeffit has some compiled-in limitations that made that impractical. My solution was to make an evenly spaced grid in R and bin together paths with similar path lengths. That reduced the problem to a tractable number of Feff paths at the expense of writing a one-off program as a processing step between the DFT and setting up the EXAFS fit. If my answer is relevant to your problem, I have a few other ideas about how to approach the problem that we can discuss futher, if you'd like. B -- Bruce Ravel ---------------------------------------------- bravel@anl.gov Molecular Environmental Science Group, Building 203, Room E-165 MRCAT, Sector 10, Advanced Photon Source, Building 433, Room B007 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://cars9.uchicago.edu/~ravel EXAFS software: http://cars9.uchicago.edu/~ravel/software/
Bruce Ravel wrote:
If I understand what you are trying to do, this paper by me might be relevant:
Role of local disorder in the dielectric response of BaTaO2N, B. Ravel, Y-I. Kim, P.M. Woodward, and C.M. Fang, Physical Review B, 73, p. 184121 (2006)
I used a DFT result as the basis of my fitting model in that paper. The naive approach was to make a Feff path from every absorber/ scatterer pair in the box. That's certainly a good idea, but Ifeffit has some compiled-in limitations that made that impractical. My solution was to make an evenly spaced grid in R and bin together paths with similar path lengths. That reduced the problem to a tractable number of Feff paths at the expense of writing a one-off program as a processing step between the DFT and setting up the EXAFS fit.
If my answer is relevant to your problem, I have a few other ideas about how to approach the problem that we can discuss futher, if you'd like.
Bruce,
Your article is exactly what I was searching for. Also in my case I have
to take into account the local disorder found by DFT calculations and
the idea to bin together paths with similar length is what I supposed
but I had difficulties to realize it in practice.
What I have done is to plot a pair radial distribution function and find
out averaged distances to put in a FEFF calculation. Reading your
article seems to me that you have done this average-step after a full
calculation of paths, isn't it?
It would be great to discuss further this subject and also to open the
discussion on how to introduce some EXAFS-knowledge in DFT calculations,
as you write in the last paragraph of the article.
Mauro
--
Mauro Rovezzi
participants (3)
-
Bruce Ravel -
Frenkel, Anatoly -
Mauro Rovezzi