Hi Scott, Feff uses the placement of the atoms to determine the size of the muffin tin radius and the norman radius, which define where the atom is verse where the atom is not. If there are no neighboring atom then these radii become unphysically large and then the scattering amplitudes will not be accurate. -Copper is kinda special. Copper has 12 neighboring atoms, so the muffin tin radii are pretty well constrained even with one shell. If you run a feff calculation with all the Cu atoms in a cluster to 7 Angstroms the top of the feff0001.dat file shows the radii for the atoms and it looks something like this: For large cluster Rmax=7.0 Angstroms: #Copper, Kittel, ISSP Feff 7.02 # metal, fcc # Abs Z=29 Rmt= 1.305 Rnm= 1.454 K shell # Pot 1 Z=29 Rmt= 1.273 Rnm= 1.391 # Gam_ch=1.761E+00 H-L exch # Mu=-5.472E+00 kf=1.724E+00 Vint=-1.679E+01 Rs_int= 2.104 Now do it again for N=12 and Rmax=3.0 For small cluster Rmax=3.0 Angstroms, N=12 # Copper, Kittel, ISSP Feff 7.02 # metal, fcc # Abs Z=29 Rmt= 1.233 Rnm= 1.467 K shell # Pot 1 Z=29 Rmt= 1.282 Rnm= 1.570 # Gam_ch=1.761E+00 H-L exch # Mu=-4.594E+00 kf=1.474E+00 Vint=-1.287E+01 Rs_int= 2.461 And again for N=2 and Rmax =3.0 For small cluster Rmax=3.0 Angstroms, N=2 #Copper, Kittel, ISSP Feff 7.02 # metal, fcc # Abs Z=29 Rmt= 1.256 Rnm= 1.815 K shell # Pot 1 Z=29 Rmt= 1.296 Rnm= 1.872 # Gam_ch=1.761E+00 H-L exch # Mu=-3.892E+00 kf=1.204E+00 Vint=-9.411E+00 Rs_int= 3.013 You can see that the norman radius gets really big for the last case, 1.391 to 1.570 to 1.872 and as a result the scattering amplitude changes. I show a comparison of the chi(k)xk for each of these inputs with a N=1.0, DE=0, SS=0, S02=1.0, DR=0 in the attached clustersize.jpg file. Shelly
-----Original Message----- From: Scott Calvin [mailto:SCalvin@slc.edu] Sent: Wednesday, July 21, 2004 9:53 AM To: XAFS Analysis using Ifeffit Subject: RE: [Ifeffit] Re: Artemis Question
Shelly said:
Feff is truly amazing.
Amen!
1) Feff needs to have a list of atoms that extends well beyond the paths that are actually used in the fit. Always pad the feff.inp file by at least an Angstrom. If you are using feff calculation for the first time. Pad the feff list by one angstrom and then by two angstroms and check that the paths that you are going to use are the same. You do not want to use paths from the edges of the cluster that you give to feff in the feff.inp file. -Hence you need to use a structure like a metal-oxide to model a hydrated metal.
Under what circumstances does FEFF actually need these extra atoms? When I take the copper example included by Artemis, for example, I can rerun the FEFF calculation with a cluster size of 3 angstroms, so that FEFF only has the nearest-neighbors to deal with. The path generated is identical to when a cluster size of 7 angstroms. Certainly if someone is using FEFF8 with a self-consistent field calculation it is necessary to have the cluster extend beyond the furthers paths, but with FEFF6L is this true?
(Incidentally, for the novices out there, I do NOT recommend being satisfied with a single-shell fit when you have a reasonable guess as to the structure beyond the first-shell. One reason can be seen easily with Artemis: paths beyond the first-shell often have low-R "tails" that contribute to the signal even at the first-shell peak. Although the FEFF calculation for the first path may be the same whether or not you also calculated paths further out, the Ifeffit fit will be different depending on whether you include those "outer" paths or not.)
--Scott Calvin Sarah Lawrence College
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listi> nfo/ifeffit
Hi Scott,
Feff uses the placement of the atoms to determine the size of the muffin tin radius and the norman radius, which define where the atom is verse where the atom is not. If there are no neighboring atom then these radii become unphysically large and then the scattering amplitudes will not be accurate.
Hi Shelly, Thanks! It turns out I misunderstood what Artemis was doing--it's smarter than I gave it credit for. :) I didn't realize that if I highlighted a FEFF calculation, changed some of the FEFF.INP parameters and reran FEFF, that it would retroactively change the paths it already had to reflect the new calculation. So when I thought I was comparing the "old" feff0001.dat to the "new" one, I was really just comparing two copies of the new one! In any case, so much for my failure to use Artemis correctly. I appreciate your explanation of why FEFF needs atoms outside of the range we're planning to use; I'm clear on it now. --Scott Calvin Sarah Lawrence College
On Wednesday 21 July 2004 01:45 pm, Scott Calvin wrote:
Thanks! It turns out I misunderstood what Artemis was doing--it's smarter than I gave it credit for. :)
Which is remarkable, given that it's author is much dumber than most people give him credit for! B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
Hi folks, 1) when removing background using athena, there is an option of using a k-weight different than 1 (I am NOT talking about k-weight for FT). I noticed that changing this parameter the resulting chi(k) does not change significantly. Actually, going from 1 to 4 as a value for k-weight for Bkgd removal, the only effect I see is a significant reduction of the peak below the first coordination shell in chi(R) while the other features of the spectrum are essentially the same in terms of absolute and relative intensity. A drastic change in both chi(k) and chi(R) is seen if I use k-weight = 5 and above. What is the exact meaning of it? I know what the meaning is in the FT process, but I do not see the reason why I should use a weight different than 1 (let's say 3 for my noisy data) for background removal. How should I behave? SHould I take the largest value of k-weight so that the essential features of the chi(k) and chi(R) do not change? 2) In artemis, is it a bug or a wanted feature to have the possibility to activate more than one button for k-weight for the fit and FT? I suppose it is a wanted feature, but then I wonder what is the meaning of using more than one weight for the fit and FT. I am sure some of you can unveil this enigma. Ciao, Stefano -- ____________________________________________ Stefano Ciurli Professor of Chemistry Department of Agro-Environmental Science and Technology University of Bologna Viale Giuseppe Fanin, 40 I-40127 Bologna Italy Phone: +39-051-209-6204 Fax: +39-051-209-6203 "Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza" Dante Alighieri - Inferno - Canto XXVI "Ihr seid bestimmt, nicht Tieren gleich zu leben, Nein, Tugend zu erringen und Erkenntnis" "Ye were not form'd to live the life of brutes, But virtue to pursue and knowledge high"
At 01:30 PM 7/23/2004 +0200, Stefano wrote:
Hi folks,
1) when removing background using athena, there is an option of using a k-weight different than 1 (I am NOT talking about k-weight for FT). I noticed that changing this parameter the resulting chi(k) does not change significantly. Actually, going from 1 to 4 as a value for k-weight for Bkgd removal, the only effect I see is a significant reduction of the peak below the first coordination shell in chi(R) while the other features of the spectrum are essentially the same in terms of absolute and relative intensity. A drastic change in both chi(k) and chi(R) is seen if I use k-weight = 5 and above. What is the exact meaning of it? I know what the meaning is in the FT process, but I do not see the reason why I should use a weight different than 1 (let's say 3 for my noisy data) for background removal. How should I behave? SHould I take the largest value of k-weight so that the essential features of the chi(k) and chi(R) do not change?
Ifeffit is choosing a spline for the post-edge background that fits the low-R part of the spectrum as well as possible (or, if you're using a background standard, makes the low-R part of the spectrum as close to the standard as possible). Since that means it's using a Fourier transform, it can also be assigned a k-weight, just like when doing EXAFS fits. A high k-weight means your placing a higher priority on making the spline match the background at high-k; i.e., at energies far above the edge. The default of 1 is reasonable, because it slightly deemphasizes the XANES region which you're not going to use for an EXAFS fit anyway. I've also used 0, although I then generally raise the low end of the energy range for the background spline so it doesn't get hung up trying to figure out the edge itself. If you use very high weights, the spline becomes obsessed with features very far above the edge, which is probably not useful. In other words, this parameter should not be that important. I'd just leave it at 1 unless you have reason to believe that weighting it more toward the behavior far above the edge is important. I wouldn't go as high as you can, since then you may distort the low-k peaks where your data quality should be best! I also don't think "noisy" data argues for a higher k-weight here; I think then you risk the spline fixating on the noise at energies far above the edge, which is probably not helpful.
2) In artemis, is it a bug or a wanted feature to have the possibility to activate more than one button for k-weight for the fit and FT? I suppose it is a wanted feature, but then I wonder what is the meaning of using more than one weight for the fit and FT. I am sure some of you can unveil this enigma.
A very useful feature! For example, the effect of S02 and sigma2 are a bit difficult to distinguish in a typical moderate-quality EXAFS spectrum, since both effect amplitude. But S02 is an overall amplitude factor, while the effect of sigma2 is weighted by k^2. Therefore changing k-weight will often change the values Ifeffit finds for these two parameters. By fitting more than one k-weight at once, Ifeffit is forced to choose a compromise that finds a reasonable S02 and sigma2 that works at all k-weights; reducing uncertainties and correlations. Note that E0 and delr suffer from a similar problem: they both shift phase, but the effect of delr is weighted by k while E0 is not (and if you use a third cumulant in your fits it is weighted by k^3). I personally avoid using multiple k-weights until I have a good "feel" for how my fits are behaving, but if the fits are clearly having trouble distinguishing between the sets of parameters I have mentioned (as evidenced, for example, by high uncertainties and correlations), then I'll try it. --Scott Calvin Sarah Lawrence College
Stefano, Just to follow up on Scott's explanation on multiple k-weights (I have nothing to add to his discussion on k-weights in spline()):
2) In artemis, is it a bug or a wanted feature to have the possibility to activate more than one button for k-weight for the fit and FT? I suppose it is a wanted feature, but then I wonder what is the meaning of using more than one weight for the fit and FT. I am sure some of you can unveil this enigma.
When multiple k-weights are selected, they are _ALL_ used. That is, for each k-weight an FT with that k-weight is done, and the resulting chi(R) spectra are appended together. The refinement is then done on the appended spectra, using all k-weights. This is essentially the same as refining multiple data sets. In fact, until very recently, the only way to use multiple k-weights was to use multiple data set fits where the only thing that changed between data sets was the k-weight. The new way (and checking multiple boxes in artemis) makes the move from 'possible, in principle' to 'easy, in practice'. As for why you'd want to do this, Scott's explanation is right on. I might go so far as to recommend using k-weight = 1 _and_ 3 as the normal way to analyze XAFS data. For what it's worth, the wavelet transforms being used in some XAFS programs (see, for example, http://www.esrf.fr/exp_facilities/BM20/Software/Wavelets/) is a more sophisticated way of doing multiple k-weightings, and may give you a good visualization about why you'd want to do this. You might even view multiple k-weight fitting as a poor approximation of wavelet transforms. Hope that helps, --Matt
I wanted to add a coupe more comments on Stafano's questions regarding k-weighting in Athena and Artemis. Regarding k-weights in Athena: Athena's default for the k-weight using background removal is 1. I chose that pretty much for the reason that Scott gave. Ifeffit's default is 0, so this is a situation where Athena and Ifeffit have different default behavior. As for the insensitivity of your data to this value, I would say that it is situational. Sometimes, the k-weight can have a dramatic influence on the background removal. Stability of the background removal and its impact on subsequent analysis has been a frequent topic of discussion in this forum. k-weight, then, is yet another parameter that can impact the stability of your spline. Rgearding k-weights in Artemis: Scott and Matt covered the "why" of k-weights. How k-weights are actually used by Artemis when fitting and plotting is a bit complicated. See section 5.8.4 of the (incomplete) artemis document for a detailed explanation of how k-weighting is used in Artemis. http://leonardo.phys.washington.edu/~ravel/software/exafs/doc/artemis.pdf B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
participants (6)
-
Bruce Ravel
-
Kelly, Shelly D.
-
Matt Newville
-
Scott Calvin
-
Scott Calvin
-
Stefano Ciurli