Fitting Problems via Artemis
Dear All, Greeting.This is Yu-Chuan. After struggling a long time, I just felt a little bit progress in the fitting program. However, there are still some questions confused me. 1. What's the meaning and purpose of "Fit Background" in Artemis? I know the backgound as "bare atom" in EXAFS theory. So, what kind of situation we need to use this function? 2. Is that possible for the Debye-Waller factor(sigma^2) be negative? Actually, I am not so sure about the physical meaning for that. Just know its a kind of thermal vibration. 3. Sometimes, The results obtained from Artemis seems not the same. Says, the guess parameters and uncertainties may be varied even if you input the same data for different time.Has anyone met this problem before? Sincerely yours, Yu-Chuan Lin Graduate Student and Reseach Assistant Kansas State University
Norbert answered the last two but I think that I can give a bit of an answer to the first one. On Fri, 24 Sep 2004, hsnuboy@ksu.edu wrote:
1. What's the meaning and purpose of "Fit Background" in Artemis? I know the backgound as "bare atom" in EXAFS theory. So, what kind of situation we need to use this function?
The background fitting in Artemis is a sort of second chance to remove background that was not completely removed in the initial processing (Athena). It attempts to fit a background function to the residual below the lower R fitting limit (for R-space fitting). I like it because it allows you to be somewhat less concerned about the initial data processing (it is still a good idea to be careful and thorough), particularly with difficult samples with noisy data. HTH, Carlo -- Carlo U. Segre -- Professor of Physics Associate Dean for Special Projects, Graduate College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 Carlo.Segre@iit.edu http://www.iit.edu/~segre
On Friday 24 September 2004 03:52 pm, hsnuboy@ksu.edu wrote:
3. Sometimes, The results obtained from Artemis seems not the same. Says, the guess parameters and uncertainties may be varied even if you input the same data for different time.Has anyone met this problem before?
Yu-Chuan, I have seen this happen, but the situation was somewhat unusual. I would like to see a project file for which this happens reproducibly. 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/
On Friday 24 September 2004 05:32 pm, Carlo U. Segre wrote:
Norbert answered the last two but I think that I can give a bit of an answer to the first one.
On Fri, 24 Sep 2004, hsnuboy@ksu.edu wrote:
1. What's the meaning and purpose of "Fit Background" in Artemis? I know the backgound as "bare atom" in EXAFS theory. So, what kind of situation we need to use this function?
The background fitting in Artemis is a sort of second chance to remove background that was not completely removed in the initial processing (Athena). It attempts to fit a background function to the residual below the lower R fitting limit (for R-space fitting). I like it because it allows you to be somewhat less concerned about the initial data processing (it is still a good idea to be careful and thorough), particularly with difficult samples with noisy data.
What Carlo says is completely correct and is the reason I tend to use that feature as well. It is important to understand why it works and why it is an OK thing to do. Let's first go back to Athena and the initial back ground removal. In Athena the most important background removal parameter is usually Rbkg. That parameter marks the cutoff in the space of the Fourier transform between the background and the chi(k). To say that another way, when we choose a value of Rbkg, we are stating that we believe that the Fourier components below Rbkg are the Fourier components associated with the background and the components above Rbkg are the components associated with the chi(k). The Autok algorithm then (essentially) tries to remove the background components while leaving the chi(k) components. Doing this background then, in principle, leaves a spectrum that only has the Fourier components associated with the data and none of the components associated with the background. In practice, this rarely happens -- particularly in the case of materials with white lines. Pretty typical is to have a certain amount of spectral weight below your original Rbkg. Thus Artemis has the option of doing an additional refinement of the background. Between Rmin and Rmax, the Feff calculation is used to fit those Fourier components. Between 0 and Rmin, a spline is used to fit those Fourier components. Thus Rmin acts rather like Rbkg in Athena. The advantage of doing this is that it may reveal correlations between the background spline and the parameters that you are trying to fit. It has the additional cosmetic advantage of often making the fit look pretty. HTH, 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/
Dear Bruce,
I attached the apj file of magnesium orthovanadate. If you change the
k-range, says, from 4~10 to 3~10. The fitting results will vary. Or you
can change other stuffs. Somtimes the results are consistantly,
somtimes not.
By the way, I wonder if my first sell DW factor is about -0.0001 and the
uncertainties is +/-0.0015. Can I say that it is possible for positive
DW factor for my first sell fitting? Thus, my fitting is reasonable?
Sincerely,
Yu-Chuan
Quoting Bruce Ravel
On Friday 24 September 2004 03:52 pm, hsnuboy@ksu.edu wrote:
3. Sometimes, The results obtained from Artemis seems not the same. Says, the guess parameters and uncertainties may be varied even if you input the same data for different time.Has anyone met this problem before?
Yu-Chuan,
I have seen this happen, but the situation was somewhat unusual. I would like to see a project file for which this happens reproducibly.
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/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
Hi Yu-Chuan, I have not looked at the project file, but I have a few comments anyway:
I attached the apj file of magnesium orthovanadate. If you change the k-range, says, from 4~10 to 3~10. The fitting results will vary. Or you can change other stuffs. Somtimes the results are consistantly, somtimes not.
If you change the k range (or other FT parameters), the fitting results can definitely change. How much the results change for small changes in FT parameters is a good indication of the reliability of the fit results. Of course, saying that the results change needs to include not only the best-fit values, but also the uncertainties.
By the way, I wonder if my first sell DW factor is about -0.0001 and the uncertainties is +/-0.0015. Can I say that it is possible for positive DW factor for my first sell fitting? Thus, my fitting is reasonable?
Well, -0.0001 +/- 0.0015 is consistent with zero. This is also a very small absolute value for sigma2. At k=10, the Debye-Waller factor, exp(-2*k*k*sigma2) = 1.02 for sigma2=-0.0001. There are many ways to get a 2% error in amplitude. One possibility (there are many) for why you could get a sigma2 that's smaller than "it should be" is that the fitting model puts in more static disorder than is really there. --Matt
participants (4)
-
Bruce Ravel -
Carlo U. Segre -
hsnuboy＠ksu.edu -
Matt Newville