The most important fitting parameters
Dear all, I am trying to use EXAFS, along with other techniques, to attempt to find information about an unknown structure. It is an organic-inorganic hybrid material, EDX reveals Cu1-W1-O4 as the stoichiometry of the inorganic component, which we are most interested in structurally. We have not been able to solve the structure from XRD, although not for want of trying. I started off with CuWO4 as the model compound and constrained the fitting parameters: - take all 1st oxygen shells (x6 at slightly different distances) and give them all amp1, dr1*reff, enot and ss1 - take all 1st copper shells (x2) and give them both amp2, dr2*reff, enot and ss2 - take all 2nd oxygen shells (x6), amp3, dr3*reff, enot, ss3 - take all 1st tungsten shells (x6), amp4, dr4*reff, enot, ss4 This gave me 13 parameters with 14.8 independent points. (Other factors: fitted in kw=2&3, k-range 2-9, used Athena with default background subtraction parameters as the chi(k) from that seemed sensible. The data were collected at SSRL, beamline 11-2, in transmission.) I was happy with the final result (R-factor 0.014), but am just wondering if this is a sensible way to go about doing things - bearing in mind I am not sure of the structure in the first place. Any hints or advice would be much appreciated. I should have collected data at the tungsten edge as well, as it is not that much higher in energy from Cu, but this didn't occur to me until later. Regards, Bridget This electronic transmission and any documents accompanying this electronic transmission contain confidential information belonging to the sender. This information may be legally privileged. The information is intended only for the use of the individual or entity named above. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution or the taking of any action in reliance on or regarding the contents of this electronically transmitted information is strictly prohibited. This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
Hi Bridget, Here are some other things to consider. 1) Is the second oxygen shell really necessary? What happens when you leave it out of the fit and remove 3 parameters? Whenever I "make-up" a structure I always take the minimalist approach. If the reduced-chi-squares for the two different models are not more than ~50% different then use the fewest paths possible. 2) Is the structure really the "same" as CuWO4? What happens when you use the same delr=alpha*reff for all 4 paths. What happens when you use CuWO4 numbers for the amplitudes? 3) Is the structure made of smaller bits of molecules that you can constrain in some way? For example, maybe you know that the second O atom is attached to either a neighboring Cu atom or the neighboring W atom. Then you could have the same delr for both the second O and the other neighbor attached to it (assuming the angle of the molecule isn't another variable). Similarly, for the amplitudes, maybe you know that one additional W neighbor will always bring with it "two" O2 atoms, so that you can constrain the amplitudes. 4) Do you have data from CuWO4? Or can you find it somewhere? Maybe the EXAFS model of that compound could be refined with your unknown and some of the parameters could be constrained to the same values. 5) Look carefully at the fit-values and the uncertainties. If the value +/- the uncertainties overlaps for two different alphas/amps/sigmas you might consider containing the values to the same variable. You need to have some physical intuition that gives you confidence in this type of constraint. 6) Sometimes it isn't possible, but I always feel better about my models if I have 2times the number of independent points as the number of variables. Always work hard to get the number of variables as low as possible. Well, that is all that comes to mind. HTH, Shelly --------------------------- Shelly Kelly Argonne National Laboratory Bldg 203, RM E113 9700 S. Cass Ave Argonne, IL 60439 Molecular Environmental Science Group www.mesg.anl.gov skelly@anl.gov phone: 630-252-7376 -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of Bridget Ingham Sent: Friday, June 16, 2006 8:27 AM To: 'ifeffit@millenia.cars.aps.anl.gov' Subject: [Ifeffit] The most important fitting parameters Dear all, I am trying to use EXAFS, along with other techniques, to attempt to find information about an unknown structure. It is an organic-inorganic hybrid material, EDX reveals Cu1-W1-O4 as the stoichiometry of the inorganic component, which we are most interested in structurally. We have not been able to solve the structure from XRD, although not for want of trying. I started off with CuWO4 as the model compound and constrained the fitting parameters: - take all 1st oxygen shells (x6 at slightly different distances) and give them all amp1, dr1*reff, enot and ss1 - take all 1st copper shells (x2) and give them both amp2, dr2*reff, enot and ss2 - take all 2nd oxygen shells (x6), amp3, dr3*reff, enot, ss3 - take all 1st tungsten shells (x6), amp4, dr4*reff, enot, ss4 This gave me 13 parameters with 14.8 independent points. (Other factors: fitted in kw=2&3, k-range 2-9, used Athena with default background subtraction parameters as the chi(k) from that seemed sensible. The data were collected at SSRL, beamline 11-2, in transmission.) I was happy with the final result (R-factor 0.014), but am just wondering if this is a sensible way to go about doing things - bearing in mind I am not sure of the structure in the first place. Any hints or advice would be much appreciated. I should have collected data at the tungsten edge as well, as it is not that much higher in energy from Cu, but this didn't occur to me until later. Regards, Bridget This electronic transmission and any documents accompanying this electronic transmission contain confidential information belonging to the sender. This information may be legally privileged. The information is intended only for the use of the individual or entity named above. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution or the taking of any action in reliance on or regarding the contents of this electronically transmitted information is strictly prohibited. This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
From what you have said, it would seem that you are on the right
Hi Bridget, Welcome to the list! track. Without knowing what CuWO4 looks like or how it behaves, I have a few suggestions that you might want to think about
I started off with CuWO4 as the model compound and constrained the fitting parameters: - take all 1st oxygen shells (x6 at slightly different distances) and give them all amp1, dr1*reff, enot and ss1 - take all 1st copper shells (x2) and give them both amp2, dr2*reff, enot and ss2 - take all 2nd oxygen shells (x6), amp3, dr3*reff, enot, ss3 - take all 1st tungsten shells (x6), amp4, dr4*reff, enot, ss4
This gave me 13 parameters with 14.8 independent points.
I think you have not quite explained what your constraints are. 4 shells times 4 parameters is 16 variables, not 13. It is hard to justify independent amplitude parameters for the different shells of a crystal. If you have any confidence at all in the crystallography, then it is hard to justify floating the amplitude for each shell. More appropriate would be to constrain each shell to have the coordination indicated by the crystallography and float a single S02 parameter. Similarly, it is hard to justify a priori the use of an independent enot for each shell. If we have faith in the theory, then you should only need one enot parameter to line up the energy grid of the data with the energy grid of the theory. Those two suggestions reduce the number of parameters from 16 to 10. You probably need to float sigma^2 independently for each shell, but you might be able to constrain the deltaR parameters in some way. I don't know what the structure of CuWO4 is, but there might be some clever constraint you could use to reduce the number of deltaR's. That might give you enough freedom to try to determine more about the first oxygen shell than just an overall deltaR. Who knows? It's a shame that you data range only goes out to 9 inverse Angstroms.
(Other factors: fitted in kw=2&3, k-range 2-9, used Athena with default background subtraction parameters as the chi(k) from that seemed sensible. The data were collected at SSRL, beamline 11-2, in transmission.)
It is good that you are using the multiple-k-weight trick. You might consider including k-weight of 1 as well. Using 1 and 3 emphasizes the high and low bits quite differently. Here's some reading on the topic: http://www.google.com/search?q=site:millenia.cars.aps.anl.gov+correlation As you say, getting both edges would be a boon, particularly given that your data range is so limited. Fortunately, it's *only* a 14 hour plane ride from New Zealand to California, so measuring more data will be no trouble at all ;-) 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/exafs/
participants (3)
-
Bridget Ingham
-
Bruce Ravel
-
Kelly, Shelly D.