RE: [Ifeffit] a question about the white line
Hi Bruce, I have worked with tough background removals. They are particularly difficult the first time you encounter each one. My methodology goes something like this. It isn't easy and sometimes very frustrating. 1) Determine how broad the rise in the adsorption edge. How many eV is it from the bottom to the flat part above the edge. If you edge is broad > 10 eV then Ezero will be somewhere high on the adsorption edge. If you edge is sharp then Ezero will be somewhere low on the adsorption edge. 2) Remove the background. Look at the FT with all three k-weights. Determine if Rbkg is reasonable. 3) You must have a good idea about the first shell. Run Feff and look at the theory with all variables set to zero, Ezero =0, sigma2=0 and delr=0. If you have a split shell be sure to include all the paths that make up the first peak in the FT of the data. 4) Compare theory to data using a well padded k-range. Leave out the first two oscillations in k, starting around 4 to 5 inverse angstroms. Set Ezero equal to zero. Run a fit, determine delr for the paths and sigma2. Look at the fit and the data in k-space with all three k-weights. Check to see if delr and sigma2 are reasonable. See if you think that moving Ezero around in Athena will help line things up at low k. 5) Import the "fit" into Athena and use it as the theory. Remove the background moving Ezero to line up the data and the theory at low k. And then start all over. 6) This procedure will not work if you use the wrong theory hence the frustrating part. HTH - If someone on the list is willing to try this procedure out and gets stuck and you are willing to share all. I will continue to help on the mailing list. Shelly Shelly Kelly ---------------------------- skelly@anl.gov Bldg 203, RM E113 phone: (1) 630-252-7376 Argonne National Laboratory fax: (1) 630-252-9793 Argonne, IL 60439 www.mesg.anl.gov
-----Original Message----- From: Bruce Ravel [mailto:ravel@phys.washington.edu] Sent: Tuesday, December 07, 2004 3:00 PM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] a question about the white line
On Tuesday 07 December 2004 03:47 pm, you wrote:
Hi Bruce,
Feel free to have them contact me. I have done a lot of Re compounds.
Jeff
Hmmm.... The point of a mailing list is to have a public, searchable archive where useful information can be discussed and shared in an open manner.
I could have just answered the question myself, but I thought that it was one of broad interest. Thus I had hoped to solicit discussion on the mailing list.
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/
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
Shelly, That seems like a lot of steps to me, and it almost seems like you're mimicking by hand the use of a standard chi(k) with autobk/spline(). Do you find very different results using a standard chi(k) with autobk/spline()? If so, do you have any suggestions for how the more automated version could be improved? My experience with strong white lines is to adjust E0, kweight (=1, 0, or 2), now the low-k clamp, and sometimes kmin (I know others regularly use kmin >1, but I prefer to have kmin<0.5 and adjust E0). I adjust these until the background mu0(E) goes through the mu(E) in a way that "looks right" (yep, it's subjective!). This isn't a particularly strong white line (only going to 1.45 of the normalized mu(E), not 3 or 4 that you can get with Re L-edges), but this is the kind of result I'm looking for: http://cars.uchicago.edu/ifeffit/misc/bkg_xanes.png I know it's completely unfair to show one favorable example and say "it should be like this", so I'll say that this is meant as one favorable example, and it will not always be like this. What I'm looking for is that the background goes smoothly and fairly "straight" to the main absorption edge without swooping up or down too much. This may not always be the best background, but it's the best I can come up with for a 'rule of thumb'. This kind of background isn't always easy to get: I don't mind swooping up to part way up the white line (say to 1.2 or so on the above plot or up a little higher on stronger white lines), but I try to avoid swooping down or swooping too far up above the white line. I should emphasize that I usually start with the simplest background possible and plow ahead to the firs shell analysis, coming back to background subtraction only after I have a sense of what the first shell is doing. And then I often use a rough first shell standard chi(k) with spline(), or just refining the background with feffit(). That's probably consistent with your (Shelly's) advice, but I think it's easy for beginners to get trapped into spending way too much time on the background removal. Hope that helps, --Matt
We too have experienced difficulties in fitting backgrounds to systems with strong edge features. Playing with the edge parameters sometimes works ok, but it's not a robust procedure. Thus we have attempted to use the theoretical mu_0 from FEFF as an a priori, and spline corrections on top of that. Thus the form of mu_0 would be: mu_0 = mu^thy_0(E,E_0,Gamma)[1 + lambda(E)] where the theoretical mu^thy_0 has an adjustable edge position (E_0) and broadening (Gamma), and lambda(E) is the spline correction which includes both instrumental variations with E and theoretical errors. In our experience, the FEFF8 mu^thy_0 can often give a good approximation to mu_0, even near the edge where there are large white lines. We've tested this on several systems, and the approach does seem promising, though it requires the extra step of running FEFF in the XANES region. In fact one has to run FEFF more than once if one tries to fit the edge parameters. A brief description is in a paper which should be out soon: ``Bayes-Turchin Approach to XAS Analysis," J. J. Rehr, J. Kozdon, J. Kas, H. J. Krappe, and H. H. Rossner, J. Synchrotron Radiation (in press, 2004). On the other hand the approach can be automated and can give a simultaneous fit of both EXAFS and XANES. We'd be interested in comments/suggestions on this approach. For example, how best to represent small corrections to the broadened edge step (e.g., arctangent corrections). John Rehr Joshua Kas
Hi John,
Thus we have attempted to use the theoretical mu_0 from FEFF as an a priori, and spline corrections on top of that. Thus the form of mu_0 would be:
mu_0 = mu^thy_0(E,E_0,Gamma)[1 + lambda(E)]
where the theoretical mu^thy_0 has an adjustable edge position (E_0) and broadening (Gamma), and lambda(E) is the spline correction which includes both instrumental variations with E and theoretical errors.
In our experience, the FEFF8 mu^thy_0 can often give a good approximation to mu_0, even near the edge where there are large white lines.
That sounds interesting. But if you include a spline with the calculated mu0(E), how important is the mu^thy_0(E)? Like, how much work do you need to put into mu^thy_0 if you have lambda to pick up the slack? On the one hand, since a spline is needed, it might imply that you don't really gain much. On the other hand, it might also imply that you could calculate and tabulate a reasonable 'universal background function' for any given absorber-scatterern pair as a starting background curve. Otherwise, it seems to me that the prior information that goes into getting mu0(E) might be roughly equivalent to using Feff to generate a standard chi(k) for autobk.
We'd be interested in comments/suggestions on this approach. For example, how best to represent small corrections to the broadened edge step (e.g., arctangent corrections).
I think the experimental broadening should be adjustable, probably defaulting to a Lorenztian with Delta E/E = 1.e-4. Anyway, I'd be interested in comparisons of this with autobk with/without a standard chi(k). And, of course, code donations for improved algorithms are always welcome. ;). --Matt
Dear All, Has anyboby managed to install the Tk version Bruce provides in his homepage using Perl 5.8.5 that comes with Fedora 3? I am trying but I am getting an error! I also got an error with Perl 5.8.3 but that I managed to fix! Thanks, Igor
On Wednesday 08 December 2004 01:40 pm, Igor Frota de Vasconcelos wrote:
Dear All,
Has anyboby managed to install the Tk version Bruce provides in his homepage using Perl 5.8.5 that comes with Fedora 3? I am trying but I am getting an error! I also got an error with Perl 5.8.3 but that I managed to fix!
Hi Igor, I have not tried the old perl/Tk with perl 5.8.5. But I have been pretty happy lately with Tk 804. Earlier releases of 804 had a very serious bug in the hierarchical list widget, which is used all over the place in Artemis. That seems to have been fixed in more recent releases. That is, I have gotten 804 to work with my codes without problem. My advice would be to try installing the perl and perl/Tk that come with Fedora 3, then install ifeffit and my stuff and see if it all works. I think there is a decent chance that it will. My next piece of advice would be to try building the latest perl/Tk from CPAN. After that, let me know what the error is. Maybe it'll make sense to me. Good luck, 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/
Thank you Bruce for the reply! I installed Tk-804.027. Let's see how it goes now! If it really works for me as it is working for you, then I don't see any reason to mess around with the old Tk version! Anyway I will reproduce the error and send the details to you later! Igor Bruce Ravel wrote:
On Wednesday 08 December 2004 01:40 pm, Igor Frota de Vasconcelos wrote:
Dear All,
Has anyboby managed to install the Tk version Bruce provides in his homepage using Perl 5.8.5 that comes with Fedora 3? I am trying but I am getting an error! I also got an error with Perl 5.8.3 but that I managed to fix!
Hi Igor,
I have not tried the old perl/Tk with perl 5.8.5. But I have been pretty happy lately with Tk 804. Earlier releases of 804 had a very serious bug in the hierarchical list widget, which is used all over the place in Artemis. That seems to have been fixed in more recent releases. That is, I have gotten 804 to work with my codes without problem.
My advice would be to try installing the perl and perl/Tk that come with Fedora 3, then install ifeffit and my stuff and see if it all works. I think there is a decent chance that it will.
My next piece of advice would be to try building the latest perl/Tk from CPAN.
After that, let me know what the error is. Maybe it'll make sense to me.
Good luck, B
-- IGOR F. VASCONCELOS Department of Physics University of Notre Dame 225 Nieuwland Science Hall Notre Dame, IN 46556 USA phone: (1 574) 631 5650 fax: (1 574) 631 5952 e-mail: idevasco@nd.edu www.nd.edu/~idevasco
participants (5)
-
Bruce Ravel
-
Igor Frota de Vasconcelos
-
John J. Rehr
-
Kelly, Shelly D.
-
Matt Newville