Hi Timm, You can use an expression that will return only positive values for you variables. sigma2 = abs(ss) guess ss This will force the model to use a positive values for sigma2. You can also use a penalty factor to keep the value inbetween a given range of values. I believe that there is an example of that in the Artemis or Ifeffit manual. -Most often negative values is a strong indication that something in the model is not quite right. Be sure to look at your data and fits showing the individual contribution from each path in the real or imaginary part of the FT. A negative sigma2 value or a really small sigma2 value usually indicates one of two different things. 1) You have included too much noise in the FT which is randomly adding signal all over in R-space hence your sigma2 values are unrealistically small to compensate. To check for this case, reduce the k-max value in the FT of the data by 1 1/angstrom and see what happens in the fit. 2) The path type is not right. sigma2 is small because you actually need more electron density for that scattering path. Hence try a path with larger Z. Negative values for the degeneracy would indicate that the path length and/or atom type is not correct. Set the degeneracy, sigma2, and delr to some reasonable values then look at the real/imaginary part of the FT for the data, fit and the contribution from the path of interest. Look to see if the signal from the path of interest has a positive contribution to the data before you try to optimize the parameters. Here is my rule of thumb for reasonable sigma2 (units= angstroms2) values sigma2 = 0.001 to 0.003 for double bonds that do not participate in binding to another atom. i.e. for the axial oxygen atoms of a uranyl moiety, UO2. sigma2 = 0.003 to 0.008 for "normal" well ordered shells. Like for Cu metal. This is the usual range. sigma2= 0.009 to 0.015 for disordered shells that are rather weekly bound. i.e. for the equatorial oxygen atom of the uranyl. There is usually 4 to 6 of them and they are not well ordered. Some times these shells can be split into two shells given enough data range in k-space. sigma2=0.015 to 0.025 for really soft bonds like solutions or squishy metals (AgCl) Sigma2 values also depend on the distance. The rules given above are for 1st and 2nd shells of atoms. HTH Shelly -----Original Message----- From: Strathmann, Timothy J. Sent: Monday, August 30, 2004 3:42 PM To: ifeffit@millenia.cars.aps.anl.gov Subject: [Ifeffit] Artemis fit results question This may be a question related to Artemis or to EXAFS fitting in general for a relative newcomer. Over the past months, when I've fit EXAFS data using Artemis and feff-derived single-scattering paths, I sometimes end up with best fit values for 2nd-shell coordination numbers or sigma squared values that are negative. Do these values have any meaning, or do they most likely signify that the chosen paths are not appropriate for fitting the spectral region in question? Also, is there a way to constrain the parameters to only consider positive values. Thanks. Timm Strathmann
On Tuesday 31 August 2004 10:06 am, Kelly, Shelly D. wrote:
-Most often negative values is a strong indication that something in the model is not quite right. Be sure to look at your data and fits showing the individual contribution from each path in the real or imaginary part of the FT.
I wanted to expand on this little part of Shelly's excellent response to Timm's question. The ability to quickly and easily plot the data, the fit, and each individual path is one of Artemis' main strengths. You should definately do as Shelly suggests and compare to the data each path that you consider including in the fit. This comparison should be made as the real and/or imaginary parts of chi(R) as well as the magnitude. And this comparison should be done both before and after a fit. Sometimes problems in a fitting model that are unclear when plotting the fit and looking at the log file become quite transparent when looking at the real/imaginary parts of each contributing path. To plot things, they need to be selected. In the language of Artemis "selected" means "highlighted in orange". You can select multiple entries in the Data and Paths List by holding down the control key while clicking the left mouse button on items in the list. Other mouse event combinations that are useful for doing extended selection are shift-click and click-drag. Paths also must be "included" in the fit to be plotted. The included ones are the ones written in black text. The excluded ones are the ones written in brown text. HTH, 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/
Shelly, Bruce Thanks for the advice. I'll try your suggestions. The negative parameters usually only appear when I try fitting the FT using an alternative path to the one that I believe is correct, or when I try improving the fit of broad second-shell peaks by considering 2 paths instead of 1. I've been assuming that the negative values were indicative of the second possibility you indicated, but wanted to verify this. I'll certainly tryout the suggestions you both made. Thanks. Timm ____________________________________________ Timothy J. Strathmann Environmental Engineering and Science Program Department of Civil and Environmental Engineering 3209 Newmark Civil Engineering Laboratory 205 North Mathews Avenue Urbana, IL 61801-2352 phone: 217-244-4679 fax: 217-333-6968 strthmnn@uiuc.edu http://cee.uiuc.edu/people/strthmnn ___________________________________________ -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of Kelly, Shelly D. Sent: Tuesday, August 31, 2004 9:07 AM To: XAFS Analysis using Ifeffit Subject: RE: [Ifeffit] Artemis fit results question Hi Timm, You can use an expression that will return only positive values for you variables. sigma2 = abs(ss) guess ss This will force the model to use a positive values for sigma2. You can also use a penalty factor to keep the value inbetween a given range of values. I believe that there is an example of that in the Artemis or Ifeffit manual. -Most often negative values is a strong indication that something in the model is not quite right. Be sure to look at your data and fits showing the individual contribution from each path in the real or imaginary part of the FT. A negative sigma2 value or a really small sigma2 value usually indicates one of two different things. 1) You have included too much noise in the FT which is randomly adding signal all over in R-space hence your sigma2 values are unrealistically small to compensate. To check for this case, reduce the k-max value in the FT of the data by 1 1/angstrom and see what happens in the fit. 2) The path type is not right. sigma2 is small because you actually need more electron density for that scattering path. Hence try a path with larger Z. Negative values for the degeneracy would indicate that the path length and/or atom type is not correct. Set the degeneracy, sigma2, and delr to some reasonable values then look at the real/imaginary part of the FT for the data, fit and the contribution from the path of interest. Look to see if the signal from the path of interest has a positive contribution to the data before you try to optimize the parameters. Here is my rule of thumb for reasonable sigma2 (units= angstroms2) values sigma2 = 0.001 to 0.003 for double bonds that do not participate in binding to another atom. i.e. for the axial oxygen atoms of a uranyl moiety, UO2. sigma2 = 0.003 to 0.008 for "normal" well ordered shells. Like for Cu metal. This is the usual range. sigma2= 0.009 to 0.015 for disordered shells that are rather weekly bound. i.e. for the equatorial oxygen atom of the uranyl. There is usually 4 to 6 of them and they are not well ordered. Some times these shells can be split into two shells given enough data range in k-space. sigma2=0.015 to 0.025 for really soft bonds like solutions or squishy metals (AgCl) Sigma2 values also depend on the distance. The rules given above are for 1st and 2nd shells of atoms. HTH Shelly -----Original Message----- From: Strathmann, Timothy J. Sent: Monday, August 30, 2004 3:42 PM To: ifeffit@millenia.cars.aps.anl.gov Subject: [Ifeffit] Artemis fit results question This may be a question related to Artemis or to EXAFS fitting in general for a relative newcomer. Over the past months, when I've fit EXAFS data using Artemis and feff-derived single-scattering paths, I sometimes end up with best fit values for 2nd-shell coordination numbers or sigma squared values that are negative. Do these values have any meaning, or do they most likely signify that the chosen paths are not appropriate for fitting the spectral region in question? Also, is there a way to constrain the parameters to only consider positive values. Thanks. Timm Strathmann
participants (3)
-
Bruce Ravel
-
Kelly, Shelly D.
-
Timm Strathmann