How to select important MS paths in wurtzite?
Dear all,
Do you know if in literature there are studies on the important multiple
scattering (MS) paths in fitting wurtzite XAFS spectra? I mean which are
the important paths to properly fit a wurtzite structure up to 6 A in a
elegant way (<20 paths)?
I'm asking this question thinking at old works by Filipponi et al. on
the fcc and bcc structures explaining very well the importance of some
specific triangular and collinear paths to fit the EXAFS spectrum...
Well, this was my coffee-break question :)
Mauro
--
Mauro Rovezzi
Hi Mauro, I don't have an answer to your question, but I will ask another that I've often wondered about: why do you prefer keeping the number of multiple scattering paths small? Do you actually have enough free parameters and a stable enough fit so that you're assigning individual sigma2's to each path, including the multiple-scattering paths? If so, I understand. But if not, what's the benefit to keeping the number of paths artificially small? Personally, I've usually approached this kind of question in an empirical way. I'm generally using heuristic constraints on my sigma2's anyway. For example, I might try constraining all direct scattering paths to oxygen beyond the nearest neighbor to the same value of sigma2. By trial and error, I can determine if the fit is significantly improved by splitting these up, or significantly degraded if I, for example, constrain the nearest-neighbor and distant oxygens to the same value (usually the answer is yes to the latter!). Given that kind of approach, I've developed some schemes for constraining multiple-scattering paths in oxides in terms of the values for direct scattering paths that seem to work fairly well. Since these schemes add no additional free parameters, I just keep adding groups of multiple-scattering paths of lesser and lesser importance until they stop having an effect on the fit. Incidentally, I believe my constraint schemes are justified because in most cases I've tried if I take a promising fit without multiple-scattering paths included and then add multiple scattering paths constrained in that fashion, the statistical quality of the fit improves. This suggests that while the constraint schemes are doubtless imperfect, they are better than leaving the paths off altogether, and they cost me no free parameters. --Scott Calvin Sarah Lawrence College At 10:47 AM 5/15/2007, you wrote:
Dear all,
Do you know if in literature there are studies on the important multiple scattering (MS) paths in fitting wurtzite XAFS spectra? I mean which are the important paths to properly fit a wurtzite structure up to 6 A in a elegant way (<20 paths)?
I'm asking this question thinking at old works by Filipponi et al. on the fcc and bcc structures explaining very well the importance of some specific triangular and collinear paths to fit the EXAFS spectrum...
On Tuesday 15 May 2007, Mauro Rovezzi wrote:
Dear all,
Do you know if in literature there are studies on the important multiple scattering (MS) paths in fitting wurtzite XAFS spectra? I mean which are the important paths to properly fit a wurtzite structure up to 6 A in a elegant way (<20 paths)?
I'm asking this question thinking at old works by Filipponi et al. on the fcc and bcc structures explaining very well the importance of some specific triangular and collinear paths to fit the EXAFS spectrum...
Well, this was my coffee-break question :)
OK, let me make a cup of coffee for myself .... Mmmmmm ... espresso! Now I'm ready to offer an answer. I would say that one of the best features of Artemis is that it makes it easy to examine the MS paths. You don't have to place faith in the literature -- you can go look at the paths themselves. It is very easy to import data and import a large number of paths. Then you can make useful, interesting plots with relatively little effort. You don't even have to do a fit to get a sense of which paths are big and which are small. Once you start actually doing some fits, it fairly simple to throw paths in the mix, take paths out, and tweak the fitting model. Again, you don't have to take on faith which paths are important. You can evaluate fits with and without certain paths and decide for yourself which ones are important. There is a rule of thumb that you need the SS, the (nearly) collinear MS, and the short triangles. In my experience, that rule of thumb is surprisingly effective. Shelly uses an interesting approach, which is described on page 25 of this PDF file: http://xafs.org/Tutorials?action=AttachFile&do=get&target=Basics_of_XAFS_analysis.pdf It involves filtering paths on the basis of Feff's approximation of the amplitude, another technique that is surprisingly effective. HTH, 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/
By some reason, when I googled "wurtzite structure", the first link with an image was to the following page: http://britneyspears.ac/physics/lipgloss/gloss.htm and they even posted her picture in the bottom, next to the wurtzite structure. Weird. I would add to Bruce's and Scott's comments: If it is an alloy, not a pure material, the importance criteria for multiple-scattering contributions change. For example, due to the size mismatch of randomized atoms, they are displaced from their regular lattice sites, causing lattice buckling which is not taken into account by FEFF which generates the table of "importance factors" - calculated for the perfert and periodic model compound. Thus, what may be important for a model compound with wurtzite structure, may not be important for an actual compound - an alloy, or a nanoparticle, e.g., CdSe. The reason is that buckling changes the bond angles - and the ms contributions are extremely sensitive to that. The change of bond angle that was 0 degrees for a collinear path to 20 degrees reduces may reduce its f(k) by 20-40%, depending on material. That makes such paths comparable with nominally non-collinear triangular ms paths which are, of course, a majority. Thus, they all compete, and if one were able to neglect non-collinear paths (except short triangles, as Bruce pointed out) in favor of collinear paths in pure compounds, or standards, it is often not possible to do it in alloys, or when some other type of disorder is present, for example - in nanoparticles where proximity to the surface or support distorts atomic order. In this case, the short length triangles (equilateral or like) are the most rigid structural units, and they survive buckling - they can be seen even in nanoparticles. Anatoly -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Bruce Ravel Sent: Tuesday, May 15, 2007 11:35 AM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] How to select important MS paths in wurtzite? On Tuesday 15 May 2007, Mauro Rovezzi wrote:
Dear all,
Do you know if in literature there are studies on the important multiple scattering (MS) paths in fitting wurtzite XAFS spectra? I mean which are the important paths to properly fit a wurtzite structure up to 6 A in a elegant way (<20 paths)?
I'm asking this question thinking at old works by Filipponi et al. on the fcc and bcc structures explaining very well the importance of some specific triangular and collinear paths to fit the EXAFS spectrum...
Well, this was my coffee-break question :)
OK, let me make a cup of coffee for myself .... Mmmmmm ... espresso! Now I'm ready to offer an answer. I would say that one of the best features of Artemis is that it makes it easy to examine the MS paths. You don't have to place faith in the literature -- you can go look at the paths themselves. It is very easy to import data and import a large number of paths. Then you can make useful, interesting plots with relatively little effort. You don't even have to do a fit to get a sense of which paths are big and which are small. Once you start actually doing some fits, it fairly simple to throw paths in the mix, take paths out, and tweak the fitting model. Again, you don't have to take on faith which paths are important. You can evaluate fits with and without certain paths and decide for yourself which ones are important. There is a rule of thumb that you need the SS, the (nearly) collinear MS, and the short triangles. In my experience, that rule of thumb is surprisingly effective. Shelly uses an interesting approach, which is described on page 25 of this PDF file: http://xafs.org/Tutorials?action=AttachFile&do=get&target=Basics_of_XAFS_ana lysis.pdf It involves filtering paths on the basis of Feff's approximation of the amplitude, another technique that is surprisingly effective. HTH, 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/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
OK guys thanks to you all for the answers. The scope of my question was mostly to add some knowledge-base on these "selection criteria" and have a survey on your preferred rules of thumb. In my case I also use the amplitude and the ss+collinear+triangle approaches. I also agree with Bruce that Artemis has a great visualization tool for selecting/understanding paths contributions. To answer to Scott question: Scott Calvin wrote:
I don't have an answer to your question, but I will ask another that I've often wondered about: why do you prefer keeping the number of multiple scattering paths small? Do you actually have enough free parameters and a stable enough fit so that you're assigning individual sigma2's to each path, including the multiple-scattering paths? If so, I understand. But if not, what's the benefit to keeping the number of paths artificially small?
I prefer to keep paths number small because moving from model compounds
to unknown samples, theoretical models will sum up and consequently the
respective number of paths/parameters will increase enormously.
M.
--
Mauro Rovezzi
participants (4)
-
Anatoly Frenkel -
Bruce Ravel -
Mauro Rovezzi -
Scott Calvin