- Ifeffit - millenia.cars.aps.anl.gov

Re: Amplitude and the EXAFS equation
by Scott Calvin 30 Oct '02

30 Oct '02

Hi all, I thought I'd add my two cents: I agree with everything Bruce said (!), but I think it is worth emphasizing one aspect of this issue. Basically, unless I'm using a mutiple-element fit with some clever constraints (not yet available in Artemis, but soon, I gather), any determination of coordination number is likely to be pretty uncertain. Not only does N correlate completely with S02 for a single shell fit, it is also likely to correlate strongly with sigma2 (or the Debye or Einstein temperature or whatever I'm using as the free parameter). Multiple-shell fits help when I can use them, but if I don't know the coordination number for the first shell, how likely is it that I know the structure of the material well enough to fit outer shells? Although you may see coordination numbers determined by EXAFS with an "uncertainty" of 1 in publications, this is usually something like the uncertainty reported in IFEFFIT--it probabIy doesn't include systematic errors like the effect of assumptions concerning the Debye-Waller factor or S02. In most cases if I'm looking for a first shell coordination number via EXAFS I'm not counting on better than a factor of two accuracy. That may still be useful in some cases, but doesn't help distinguish between, say, tetrahedral and octahedral coordination. So, if I'm only looking for a factor of 2, and if I'm pretty sure my sample is homogenous and of a consistent and appropriate thickness, I'm happy to just constrain S02 to 0.90 or 0.85 or whatever, knowing that the arbitrary choice of S02 is probably introducing an uncertainty of less than 20% in the coordination number. Since this is still considerably less than I generally believe is introduced by other issues (such as the scheme used for the Debye-Waller factors), it's good enough for me. --Scott Calvin Naval Research Lab --

3 2

Amplitude and the EXAFS equation
by Bruce Ravel 29 Oct '02

29 Oct '02

Hi folks, Today I received one of the top-ten most asked questions. I am going to take a stab at answering it here in an archived forum. I am sort of hoping to get a discussion going and get several people's viewpoints. An Artemis user observed in her email that, in the EXAFS literature, people often publish values for coordination number. She asked how one obtains that using Artemis/Ifeffit. If only there were an easy answer to this question! It turns out that, in the EXAFS equation, there are two terms that directly affect the amplitude. One is N, the coordination number (or one might say "the path degeneracy"), and the other is S02, the so-called passive electron reduction factor. S02 is, according to popular wisdom, chemically transferable. This means that if you somehow have determined S02 for a given central atom in one compound, you can use this value for some other compound. While this may be true, particularly if a standard and a sample are measured back-to-back under identical experimental conditions, it is not a safe course of action. The problem is that lots of empirical effects can also change the measured amplitude. These may include things like detector response, sample inhomogeneity, and those annoying gremlins that sneak into the hutch when you're not looking. Assuming, though,that you know S02 and you are confident that empirical problems are a small factor, then the "S02" path parameter can often be interpreted simply as the coordination. If, however, you are not confident of S02 or of the smallness of the empirical effects, the situation is somewhat more complicated. A common assumption is that you should be able to determine S02 *and* coordination simultaneously from data. This is, at best, extremely difficult to do with statistical significance and, at worst, flatly impossible. If you are fitting only one shell of data, then it is certainly impossible. Why do I say that? Well, consider how coordination (N) and S02 enter the EXAFS equation: N * S02 * F(k) chi(k) = ------------------ sin( 2kR + delta(k) ) exp(-2ksigma^2) 2kR This is the generic formula. In Feff, Ifeffit, and Artemis, this is the the expression for chi(k) for a particular path. The data chi(k) is fit by evaluating this equation for one or more paths then summing them up. In that equation, note that N and S02 are both multiplicative factors. If your fit considers only one path, then N and S02 have *exactly the same effect* on the equation. They cannot be measured independently. In the absence of knowledge about either N or S02, the only thing you can do is measure the product (N*S02). How does one then report on coordination in a journal article? You have to somehow interpret the product (N*S02) and determine how much of that is N and how much is S02. There is no generic answer to how that is done -- it depends on the details of the sample and the experiment. To return to last week's silly posts about training monkeys -- it is yet another reason why data analysis requires a thinking human being. If you are fitting many shells or many data sets (the latter is something that Artemis does not yet allow you to do -- it's coming soon!), then you may have enough information in your data to uncouple the correlation between N and S02 and somehow fit them independently. Again, there is no generic way of doing so -- each situation is different. It is common to do as I suggested above and determine S02 from a well known standard that was measured at the same time as your sample. If there are no significant empirical effects in the sample that do not exist in the standard, then the value of S02 is probably useful for your unknown sample. Armed with that, it is much easier to then determine coordination. The discussion up to now has been conceptual. How are the amplitude terms handled in Artemis. Well, start by taking a look at a screenshot of the path page from a recent version of Artemis: http://feff.phys.washington.edu/~ravel/software/exafs/images/artemis/artemi… Note that there is an entry box labeled "N" and another labeled "S02". The one labeled "N is filled with the path degeneracy when the feffNNNN.dat file is imported. The one labeled S02 is intended to be filled with a math expression used to describe all remaining terms that may effect the amplitude. In the example shown, I have left N equal to the value from the feffNNNN.dat file and set S02 to a very simple math expression. The variable "amp" is being floated in this fit to copper foil data and used to evaluate the S02 term in the EXAFS equation for each path. In this case I am using knowledge of the system to uncouple N and S02. This is a copper foil, thus I am confident that the coordination is as expected from crystallography. I am also confident that the empirical effects are negligible. Whatever value amp ends up being, I will interpret that as S02. Now suppose I want to measure a nano-particulate copper oxide. In such a system, I may expect coordination to be reduced relative to the bulk oxide. For that, I may do the following: 1. Run feff using the crystal structure of bulk copper oxide. This will give me a copper/oxygen path to use to fit the first shell of my nanoparticle. 2. Read that feff path into Artemis along with my data 3. Set amp to the value determine from the copper foil (this is done by selecting "set" rather than "guess" from the appropriate optionmenu on the "Guess, Def and Set" page in Artemis. 4. For the first shell path, set N to 1 (rather than what ever value it was in the feffNNNN.dat file) 5. Set S02 for the first path to this math expression: n_first * amp 6. On the "Guess, Def and Set" page, define n_first as a guess parameter and choose a sensible starting value. The N and S02 part of the screenshot above will then look something like this: ___ ___________________________________ N |_1_| * S02 |_ n_first * amp ___________________| Here I have fixed amp to represent the known value of S02, I have set the N term in the EXAFS equation to be 1, and I have made a math expression, "n_first * amp", that will be evaluated and used for the S02 term in the EXAFS equation. n_first will be adjusted in the course of the fit and can be directly interpreted as a coordination number. Is this the only way of dealing with coordination? Certainly not. Is this the "official" way of measuring coordination? No. There isn't one. Every situation is different. Should you apply this example to every one of your analysis problems? No way! Did this example help you understand how N and S02 are used in Artemis and Ifeffit? I hope so. May this example be applicable to some analysis problem you run into at some point in your career? Quite possibly. Do you need to think hard about every analysis problem and figure out the best way to deal with the problem of N and S02 being coupled parameters? Absolutely! Hope that helps, Bruce -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

detaching plot buttons in Athena on Windows
by Bruce Ravel 25 Oct '02

25 Oct '02

Hmmm.... the situation with the detaching plot buttons is actually worse than I thought when I wrote my last (and somewhat sarcastic) message. It turns out that the mechanism I used to handle the detaching and reattaching of the plot buttons will not work on Windows. It seems to be one of the very few things in perl/Tk that does not port from Unix to Windows. Unless I can find a better way, in the next release I will remove this option from the Plot menu on Windows. For now, just don't use it. Thanks for the report, Raul. It's appreciated, as always. B -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

detaching plot buttons in Athena on Windows
by Bruce Ravel 25 Oct '02

25 Oct '02

Raul said: RB> I just downloaded your last exe files of Athena and Artemis. I RB> tried the detach plot buttons feature and I found that the Data RB> group card is gone. I've done this before and after loading some RB> data and it happened in both cases. Then I selected the replace RB> option and nothing happened except the message "this may take a RB> few seconds". So, there is no way to select any data after RB> select this option. Hmmm.... that's a problem.... I guess I forgot to test that little feature on Windows before releasing. Drat! Fortunately, the work-around is really obvious ;-) B -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

Re: Ifeffit digest, Vol 1 #70 - 1 msg
by Raul Barrea 25 Oct '02

25 Oct '02

Hi Bruce, I just downloaded your last exe files of Athena and Artemis. I tried the detach plot buttons feature and I found that the Data group card is gone. I've done this before and after loading some data and it happened in both cases. Then I selected the replace option and nothing happened except the message "this may take a few seconds". So, there is no way to select any data after select this option. Thanks Raul Barrea ----- Original Message ----- From: <ifeffit-request(a)millenia.cars.aps.anl.gov> To: <ifeffit(a)millenia.cars.aps.anl.gov> Sent: Friday, October 25, 2002 12:00 PM Subject: Ifeffit digest, Vol 1 #70 - 1 msg > Send Ifeffit mailing list submissions to > ifeffit(a)millenia.cars.aps.anl.gov > > To subscribe or unsubscribe via the World Wide Web, visit > http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit > or, via email, send a message with subject or body 'help' to > ifeffit-request(a)millenia.cars.aps.anl.gov > > You can reach the person managing the list at > ifeffit-admin(a)millenia.cars.aps.anl.gov > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Ifeffit digest..." > > > Today's Topics: > > 1. Athena 0.8.004 and Artemis 0.5.001 (Bruce Ravel) > > --__--__-- > > Message: 1 > Date: Thu, 24 Oct 2002 15:48:34 -0400 > To: ifeffit mailing list <ifeffit(a)millenia.cars.aps.anl.gov> > Reply-To: ravel(a)phys.washington.edu > From: Bruce Ravel <ravel(a)phys.washington.edu> > Subject: [Ifeffit] Athena 0.8.004 and Artemis 0.5.001 > > > Hi everyone, > > I posted a new source tarball (horae-005.tar.gz) and new windows > executables on my webpage today. Both Athena and Artemis have > significant new features and bug fixes. > > Athena: > The biggest new feature is the many new keyboard shortcuts. Here's > a summary: > Ctrl-t toggle marking of current group > Ctrl-p e plot current in energy > Ctrl-p k plot current in k-space > Ctrl-p r plot current in R-space > Ctrl-p q plot current in q-space > Meta-p e plot marked in energy > Meta-p k plot marked in k-space > Meta-p r plot marked in R-space > Meta-p q plot marked in q-space > Ctrl-k select group above > Ctrl-j select group below > It is now possible to do much more work with Athena without taking > your hands of the keyboard. > > You can now save derivative of mu(E) spectra to files. The > auto-alignment algorithm changed slightly. Let me know if it is > better or worse than before. You can detach the plot buttons from > the main window. That's rather wacky, but people may like it. See > the "Plot" menu. > > Artemis: > I fixed a half dozen really serious bugs that were discovered during > and after the recent analysis course at NSLS. I will certainly not > claim that artemis is bug-free -- you just have to go find new ways > to break it ;-) > > The most significant new feature is on the paths page. There is now > a little box for the degeneracy. It is filled automagically with > the degeneracy from the feffNNNN.dat file. It must be a number and > must not be a variable. I hope this will help clarify the inherent > confusion between N, S02, and other terms which affect the > amplitude. There are menu items for setting all degeneracies to 1 > or resetting them to their feffNNNN.dat values. > > Also I changed the background color of Artemis so that, if you have > both Athena and Artemis open at the same time, they are easier to > tell apart. > > The complete lists of changes and bug fixes are available on the web > site. > > Extra special thanks this time to Mark Jensen (YeS, i fiXeD YouR bUg), > Jia Liu (who was *very* patient in helping me track down some serious > Artemis bugs, thanks!), Shelly Kelly (whose mouse avoidance regimen is > proceeding nicely), and Scott Calvin (for color consultation). > > Have fun! > B > > > > -- > Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu > Code 6134, Building 3, Room 222 > Naval Research Laboratory phone: (1) 202 767 5947 > Washington DC 20375, USA fax: (1) 202 767 1697 > > NRL Synchrotron Radiation Consortium (NRL-SRC) > Beamlines X11a, X11b, X23b, X24c, U4b > 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(a)millenia.cars.aps.anl.gov > http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit > > > End of Ifeffit Digest

1 0

Athena 0.8.004 and Artemis 0.5.001
by Bruce Ravel 24 Oct '02

24 Oct '02

Hi everyone, I posted a new source tarball (horae-005.tar.gz) and new windows executables on my webpage today. Both Athena and Artemis have significant new features and bug fixes. Athena: The biggest new feature is the many new keyboard shortcuts. Here's a summary: Ctrl-t toggle marking of current group Ctrl-p e plot current in energy Ctrl-p k plot current in k-space Ctrl-p r plot current in R-space Ctrl-p q plot current in q-space Meta-p e plot marked in energy Meta-p k plot marked in k-space Meta-p r plot marked in R-space Meta-p q plot marked in q-space Ctrl-k select group above Ctrl-j select group below It is now possible to do much more work with Athena without taking your hands of the keyboard. You can now save derivative of mu(E) spectra to files. The auto-alignment algorithm changed slightly. Let me know if it is better or worse than before. You can detach the plot buttons from the main window. That's rather wacky, but people may like it. See the "Plot" menu. Artemis: I fixed a half dozen really serious bugs that were discovered during and after the recent analysis course at NSLS. I will certainly not claim that artemis is bug-free -- you just have to go find new ways to break it ;-) The most significant new feature is on the paths page. There is now a little box for the degeneracy. It is filled automagically with the degeneracy from the feffNNNN.dat file. It must be a number and must not be a variable. I hope this will help clarify the inherent confusion between N, S02, and other terms which affect the amplitude. There are menu items for setting all degeneracies to 1 or resetting them to their feffNNNN.dat values. Also I changed the background color of Artemis so that, if you have both Athena and Artemis open at the same time, they are easier to tell apart. The complete lists of changes and bug fixes are available on the web site. Extra special thanks this time to Mark Jensen (YeS, i fiXeD YouR bUg), Jia Liu (who was *very* patient in helping me track down some serious Artemis bugs, thanks!), Shelly Kelly (whose mouse avoidance regimen is proceeding nicely), and Scott Calvin (for color consultation). Have fun! B -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

Athena: problems at the Bi L3 edge
by Carlo U. Segre 23 Oct '02

23 Oct '02

Bruce: I have encountered some strange behavior at the Bi L3 edge. I have a file which runs out about 1200eV above the edge (13419eV). While Athena does find the high energy when setting the default normalization and spline ranges, it does not plot any of the data beyond 14220 (about 800eV above the edge). In fact, there seems to be a hard limit of about 800eV above the edge for all data and this translates into the k-space plot as well. I am now unsure of what effect this might have on the back transform if the defaults are not touched but certainly the graphical setting of limits for normalization, spline fit, fourier transform and back transform wil not work since the plot does not show the complete data range. Another problem I have found is that the Auto Align feature is somewhat unreliable and does not always yield a visually correct shift. I haven't characterized when exactly this happens but it has seemed to work once on a file and then not work later on the same two files. Cheers, Carlo -- Carlo U. Segre -- Professor of Physics Associate Dean for Research, Armour College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 Carlo.Segre(a)iit.edu http://www.iit.edu/~segre

2 1

Artemis and templates
by Bruce Ravel 22 Oct '02

22 Oct '02

Hi folks, I received this question today: > The third question is about the fitting templates provided in your > software, such as firstshell, mixedshell, three shell, etc. How > can I predict which one is good for the fitting? What are the > differences among them? A bit of recent history. The templates are a feature of Artemis that I started working on just prior to last month's EXAFS course at NSLS. The concept seemed to be working pretty well for copper metal ($ironic_mode=1), so I decided to put the templates in the official release. The thing that the templates have been most useful for so far is to point out that Artemis remains almost completely undocumented. The templates are but one confusing part of the program. Once a user gets confused, there is no place (other than, of course, this mailing list) for the user to turn for information. Sigh. Let me start by explaining what the templates are not: They are not intended to be a complete description of any fit to any data set. That is, you cannot expect to choose an item from the Templates menu, click the big green fit button, and be ready to publish. The templates are intended to help you start a new fitting project by filling in many of the little entry boxes scattered throughout the program with reasonable first guesses for interesting parameters. Another way to say this is that the templates will hopefully help save you from having to do a certain amount of really tedious typing. As an example, the template called "firstshell" makes a non-stupid guess about what kinds of parameters you might want for a fit to the first shell of your data. You will almost certainly have to edit the guess/set/def parameters as well as the path parameters to get a fit to your data that is statistically valid and physically reasonable. Some of the templates ("threeshell" for example) do nothing at all at this time. So, to answer the questions directly: > How can I predict which one is good for the fitting? In a sense, none of them are. However, if you want to do a fit to the first shell, the "firstshell" template may be helpful. If you want to do a fit for which the correlated Debye model is valid, the "debye" template may be helpful. > What are the differences among them? Their names tend to explain the kind of situation in which they might be helpful. Let me end by saying my main point one more time in different language: Templates are the *starting points* for fits. Use them in that sense. They are NOT complete fitting models. Do not use them as such. Hope that helps, B -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

SUIs
by Bruce Ravel 22 Oct '02

22 Oct '02

Matt said: MN> Bruce also points out (again, correctly) that we have very little MN> experience training monkeys. OK. You caught me in a little, white lie. Perhaps it is possible to train a monkey to do data analysis. In fact, all my experience is in writing GUIs for Ifeffit. I actually have no experience writing SUIs (simian user interfaces). Perhaps one of our eager users would be interested in doing the pioneering work in this field. ;-) B -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

1 0

measurement uncertainty
by Bruce Ravel 22 Oct '02

22 Oct '02

Gustavo de Medeiros Azevedo said: GdMA> I''ve recently started using IFEFFIT, and I'm a bit confused on GdMA> the way it estimates the error bars. In the IFEFFIT Reference GdMA> guide, (page 38) you mention the variables epsilon_k and GdMA> epsilon_r, where the measurement uncertainty is stored. In GdMA> that same page, you remark that changing the values of GdMA> epsilon_k and epsilon_r will change chi_squared, Chi-reduced GdMA> but won't affect R_factor and the error bars. As far as I GdMA> understand, the error bars should given by the diagonal GdMA> elements of the co-variance matrix, multiplied by the square GdMA> root of Chi_reduced. In this way, I would expect that varying GdMA> epsilon_k or epsilon_r should affect the estimated GdMA> uncertainties. GdMA> Could you, please, clarify this point? Gustavo, Well, I live one time zone to the east of Matt, so I guess I'm seeing this before he does ;-) I'll take a stab at your question. epsilon_r is measured from the data by performing the Fourier transform of chi(k) and then computing the root mean square value of the magnitude of chi(R) between 15 and 25 Angstroms. The assumption is that, because of disorder and mean-free-path effects, any finite spectral content in that R range can only be due to white noise in the data. The relationship between epsilon_k and epsilon_r is established by the normalization of the Fourier transform. Because epsilon_R enters into the fitting metric, chi-square, when the fit is done in R-space (or epsilon_k for a fit in k space), you are correct that the value of chi-square and reduced chi-square are affected by the value of epsilon_r. The R_factor is just a percentage misfit and so is independent of epsilon. The error bars are independent of epsilon as reported by ifeffit, but the reason for this requires some explanation. The first question to ask at this point is whether it is valid to presume that the dominant error in the exafs measurement is the shot noise measured from the high-R portion of the Foruier spectrum. In most situations it is not -- that is, in most situations the shot noise is much smaller that detector non-linearity, sample inhomogeneity, errors in the theoretical fitting standards, errors in the empirical fitting standards, or a whole host of other problems you might have in a real experiment. Particularly at 3rd generation light sources, photon flux and the shot noise associated with it is the least of your problems. By a long stretch. Thus, the way ifeffit uses to estimate epsilon is manifestly wrong. It is way too small, thus chi-square is very large. Even for a fit that is clearly a good fit, such as the common example of a copper foil that comes with the Artemis, the reduced chi-sqare is much larger than 1. By the formalism of Gaussian statistics, we expect that a good fit has a reduced chi-square of about 1. So what's going on? Well, the problem is that ifeffit makes no attempt (since it operates within a Gaussian framework and not a Baysian framework) to estimate all the other sources of error besides shot noise. Thus it is almost always doomed to have reduced chi-square much larger than 1 (except in the odd case when the shot noise is quite large). So what about the error bars? The diagonal elements of the covarience matrix (i.e. the error bars) are much too small for the same reason that the reduce chi-square is too large, i.e. because epsilon has been vastly underestimated. Ifeffit then makes the assumption that the fit it just finished was, in fact, a good fit. It assumes that the reason that reduced chi-square is not near one is NOT because it is a bad fit, but rather because epsilon has been significantly underestimated. Multiplying the error bars by the square root of reduced chi-square is mathematically equivalent to re-evaluating the covarience matrix with epsilon set to the value it must be for reduced chi-square to equal 1. Thus, it is the mathematical equivalent of assuming that the fit is, in fact, a good fit. So finally, you asked how setting epsilon to some other value will effect the fit. The answer is "Not much." It will have some small effect in how the fit is evaluated numerically because chi-square may be of a different order of magnitude. However, because Ifeffit rescales the error bars under the assumption that the fit was good, manually changing epsilon will have little effect on the results. One more thing needs to be said. Because Ifeffit *always* assums that the fit is a good fit and reports statistics accordingly, it is the RESPONSIBILITY OF THE USER to actually evaluate the quality of the fit. Are the best fit values physically reasonable? Are the parameters highly correlated? Are the rescaled error bars of a reasonable size? Ifeffit cannot answer those questions. You must. And that, finally, is why a human needs to do data analysis. Sadly we cannot train a computer or a monkey to decide if the fit is, in fact, the one you want to publish. Hope that helps, Bruce -- Bruce Ravel ----------------------------------- ravel(a)phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b 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/

2 1