Hi Dr. Newville, I had similar situation and questions as with choosing Path Parameters. My goal is to identify the occupancy sites for a doped element in my sample. We have a rough guess of two possible sites, part of the doped atoms are filling a void site of the unit cell, the other part is to substitute an existing atom sites. I was being able to writing two ATOM files and run FEFF twice to calculate all possible PATH. But when I was going to define/guess the parameters ('N','amp', 'enot', 'delr','ss')for EACH path, I totally lost idea how to group all those paths, or how to guess all those five parameters for each path. Of course, I couldn't get a good fit as I don't know how to build the model and to refine parameters in this step. Here I post my understanding of the five parameters: 'N', path degeneracy, or coordination number; 'amp', usually close to 1, but as I have doping elements, I am expecting small partial numbers in some of the 'amp' to adjust for partial occupancy; 'enot', I think it is just the energy shift of your spectra, and as I believe my spectra are well aligned, I will SET 'enot' to zero. 'delr' is the variation of the real atom-atom distances from the calculated ones (I think I need to group paths to refine 'delr' but I don't know what's the criteria to group paths). 'ss' is to consider about the disorder effect and typically I use the Debye model. Let me know if I am wrong in any way. Could you please give me some suggestion how did you get started? I am looking forward to hearing from you. Best regards, Yanyun Quoting Matt Newville :

On Mon, Aug 18, 2014 at 7:45 PM, 이국승 wrote:

...

Dr Newville,

Thank you for your reply.

My first question is that I did not find the atoms.inp file in the demeter installer. If that is true, where can I get the atoms.inp file?

You might find what you're looking for at http://cars9.uchicago.edu/~newville/adb/search.html Metallic Au (fcc) is there.

...

Anyway, I did copy and paste the atoms.inp file of an old version of Artemis from my colleague's PC. I followed the Artemis intruction video for fitting Au foil data.

Hmm, it seems likely that you missed a step.

...

I filled the crystal lattice information, run atoms, and Feff calculation.

Did you import and define a Path after doing this calculation? Did you define Path Parameters and variables for this fit?

...

Finally I clicked the Fit button and I got the error message that I attached. Now here is my second question. What can I do to successfully Fit without errors? Do you have any idea from the error message?

It indicates that you have defined variables to refine in the fit ('amp', 'enot', and so forth) but not explained how those variables should alter the EXAFS signal. You'll probably want to use these in the Path Parameters for the first shell Au-Au scattering. But, you are doing the analysis, so the most important thing is that you understand how the variables you are asking to be refined actually effect the EXAFS signal. These error messages are a feature of the program to remind you that you need to understand the model you are using for the EXAFS signal.

So, what do you intend 'amp', 'enot', 'delr', and 'ss' to do? If you don't know, go back and review the tutorial information. The goal is not to get a fit, the goal is to understand what the fit results mean.

Another question (it might seem random, but it is not): what is your fitting range?

--Matt

PS: to be clear, these are not rhetorical questions. If you feel the need to respond, please actually think about and answer all questions above.

Hi Matt, Thank you for your quick response. Your explanations of the Path Parameter and Fit Parameter are very clear. Basically we have to guess Fit parameters in GDS page and to know what a Fit Parameter does to your Path Parameters by using math expression. But my question is, for each path there are a set of Path Parameters, does that mean we have to fit this set of path parameters independent from those used in other paths? For instance, if I am going to include 30 paths in my fitting, do I need to guess about 150 (30*5=150) parameters (except guess the same 'enot' for all paths)? As shown on the fit Log file, the Correlation value between two fit parameters is the bigger the better, or the smaller the better? Thanks, Yanyun Quoting Matt Newville :

On Tue, Aug 19, 2014 at 9:55 AM, wrote:

...

Hi Dr. Newville,

I had similar situation and questions as with choosing Path Parameters. My goal is to identify the occupancy sites for a doped element in my sample. We have a rough guess of two possible sites, part of the doped atoms are filling a void site of the unit cell, the other part is to substitute an existing atom sites.

I was being able to writing two ATOM files and run FEFF twice to calculate all possible PATH. But when I was going to define/guess the parameters ('N','amp', 'enot', 'delr','ss')for EACH path, I totally lost idea how to group all those paths, or how to guess all those five parameters for each path. Of course, I couldn't get a good fit as I don't know how to build the model and to refine parameters in this step.

Here I post my understanding of the five parameters: 'N', path degeneracy, or coordination number; 'amp', usually close to 1, but as I have doping elements, I am expecting small partial numbers in some of the 'amp' to adjust for partial occupancy; 'enot', I think it is just the energy shift of your spectra, and as I believe my spectra are well aligned, I will SET 'enot' to zero. 'delr' is the variation of the real atom-atom distances from the calculated ones (I think I need to group paths to refine 'delr' but I don't know what's the criteria to group paths). 'ss' is to consider about the disorder effect and typically I use the Debye model. Let me know if I am wrong in any way.

Could you please give me some suggestion how did you get started? I am looking forward to hearing from you.

For each Path in a model, you need a) a Feff.dat file -- this can be read from any Feff calculation, and those can be generated from a crystal structure b) a set of Path Parameters -- the physically meaningful quantities N*S02, E0, deltaR, sigma2, etc from the EXAFS Equation.

Each of the Path Parameters is given as an equation in terms of the set of Fit Parameters. Often times, the equations are very simple. That is, sigma2 for Path1 may simply be defined to be evaluated from 'ss'. You could make it 'ss*2', or something else.

There is also a set of Fit Parameters -- named quantities that may be adjusted in the fit to make the sum of Paths best match your data. These can be "set", in which case they are frozen -- not varied in the fit. In Artemis, the Fit Parameters are shown in the Guess/Set/Def page. The quantities like "enot", "delr", "ss", etc are Fit Parameters. You need to fill out an expression for each Path Parameter you want changed in the fit in terms of the set of Fit Parameters.

Artemis will help you by complaining if some of the Fit Parameters defined are not used in any Path Parameters -- changing these values won't alter the fit, so you should not have them defined as a variable Fit Parameter.

The system is designed to be very flexible. But this means that before you do a fit you need to understand what each of the Fit Parameters does to the Path Parameters for the Paths you are using in the fit. We could guess that by "enot", you mean the value for the Path Parameter 'E0', and that you want to make sure that the E0 Path Parameter is given as "enot" for all Paths. But this would be only a guess. You are setting up the model, so you have to ask yourself why you have a variable 'enot', and what that means for your model.

--Matt

Hi Scott, Thank you for giving me detailed examples. I will go with your suggestion and try to reduce free parameters. Best, Yanyun Quoting Scott Calvin :

Hi Yanyun,

On Aug 19, 2014, at 12:05 PM, huyanyun@physics.utoronto.ca wrote:

...

But my question is, for each path there are a set of Path Parameters, does that mean we have to fit this set of path parameters independent from those used in other paths? For instance, if I am going to include 30 paths in my fitting, do I need to guess about 150 (30*5=150) parameters (except guess the same 'enot' for all paths)?

No. One of the great principles of Ifeffit (and thus Artemis) is that the path parameters do not have to be the same as the fit parameters.

As a simple example, it's good to start with the model that all paths have the same value for E0. So you can guess a single parameter and then use that parameter for the E0 for all paths.

As another simple example, a cubic crystal might be modelled as having a uniform thermal (or Vegard's law) expansion. Then there could be a single guessed parameter indicating the fractional expansion (called, for example, alpha) while the delr for each path could then be entered as alpha*reff.

Much more complicated constraint schemes are possible, but for fits with large number of paths, the number of guessed parameters is almost always much much lower than the number of path parameters.

...

As shown on the fit Log file, the Correlation value between two fit parameters is the bigger the better, or the smaller the better?

Not necessarily. Correlations are provided to help understand the relationship between guessed parameters, but there's nothing inherently wrong with a high correlation...the uncertainty associated with that relationship is already represented in the uncertainties reported with the fit. In other words, it is more direct to focus on getting the uncertainties to be lower, rather than the correlations.

--Scott Calvin Sarah Lawrence College

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Hi Yanyun, This topic has been dealt with multiple times on the mailing list. Please search the archive: http://www.mail-archive.com/ifeffit@millenia.cars.aps.anl.gov/ For example, searching for "sites" will give you several discussions of the topic. --Scott Calvin Sarah Lawrence College On Aug 19, 2014, at 2:49 PM, wrote:

Hi all,

How would you deal with the case that the absorber atoms are occupying two different crystallography sites? Should I still use/guess the same 'S02' for paths calculated from two sites?

Best, Yanyun

Quoting huyanyun@physics.utoronto.ca:

...

Hi Scott,

Thank you for giving me detailed examples. I will go with your suggestion and try to reduce free parameters.

Best, Yanyun Quoting Scott Calvin :

...

Hi Yanyun,

On Aug 19, 2014, at 12:05 PM, huyanyun@physics.utoronto.ca wrote:

...

But my question is, for each path there are a set of Path Parameters, does that mean we have to fit this set of path parameters independent from those used in other paths? For instance, if I am going to include 30 paths in my fitting, do I need to guess about 150 (30*5=150) parameters (except guess the same 'enot' for all paths)?

No. One of the great principles of Ifeffit (and thus Artemis) is that the path parameters do not have to be the same as the fit parameters.

As a simple example, it's good to start with the model that all paths have the same value for E0. So you can guess a single parameter and then use that parameter for the E0 for all paths.

As another simple example, a cubic crystal might be modelled as having a uniform thermal (or Vegard's law) expansion. Then there could be a single guessed parameter indicating the fractional expansion (called, for example, alpha) while the delr for each path could then be entered as alpha*reff.

Much more complicated constraint schemes are possible, but for fits with large number of paths, the number of guessed parameters is almost always much much lower than the number of path parameters.

...

As shown on the fit Log file, the Correlation value between two fit parameters is the bigger the better, or the smaller the better?

Not necessarily. Correlations are provided to help understand the relationship between guessed parameters, but there's nothing inherently wrong with a high correlation...the uncertainty associated with that relationship is already represented in the uncertainties reported with the fit. In other words, it is more direct to focus on getting the uncertainties to be lower, rather than the correlations.

--Scott Calvin Sarah Lawrence College

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Hi Yanyun, The shift observed in R-space is due to the effect of the potential of the absorbing and scattering atoms. You are asking a series of questions that, while thoughtful, are covered in basic tutorials and XAFS textbooks. I suggest you search for introductions to XAFS analysis, many of which are freely available and which come in many forms (videos, tutorial files, wikis, PowerPoints, chapters/books, etc.) and work through one or more of them. Beginners are encouraged to use this list, but beginners and advanced users alike are also urged to see if the answer to their questions is readily and clearly available elsewhere first. --Scott Calvin Sarah Lawrence College On Aug 20, 2014, at 1:11 AM, huyanyun@physics.utoronto.ca wrote:

Hi all,

Does Feff take into account of the radius of the each atom when it calculates atom-pair distance? I am wondering if the calculated atom-pair distance (Reff) in each SS path has already excluded the radius size of absorber and scatterer atoms? In the R-space spectra, I always have a big peak at R smaller than any of the calculated path distances, in other words, I feel there is no possible path contributing to this peak.

My goal is trying to understand where the filler atom is in a unit cell, so if the above question is 'yes', the true position of the filler atoms is very likely to be far away from the site I input in the ATOM file ?

Best regards, Yanyun

Quoting Scott Calvin :

...

Hi Yanyun,

This topic has been dealt with multiple times on the mailing list. Please search the archive:

http://www.mail-archive.com/ifeffit@millenia.cars.aps.anl.gov/

For example, searching for "sites" will give you several discussions of the topic.

--Scott Calvin Sarah Lawrence College

On Aug 19, 2014, at 2:49 PM, wrote:

...

Hi all,

How would you deal with the case that the absorber atoms are occupying two different crystallography sites? Should I still use/guess the same 'S02' for paths calculated from two sites?

Best, Yanyun

Quoting huyanyun@physics.utoronto.ca:

...

Hi Scott,

Thank you for giving me detailed examples. I will go with your suggestion and try to reduce free parameters.

Best, Yanyun Quoting Scott Calvin :

...

Hi Yanyun,

On Aug 19, 2014, at 12:05 PM, huyanyun@physics.utoronto.ca wrote:

...

But my question is, for each path there are a set of Path Parameters, does that mean we have to fit this set of path parameters independent from those used in other paths? For instance, if I am going to include 30 paths in my fitting, do I need to guess about 150 (30*5=150) parameters (except guess the same 'enot' for all paths)?

No. One of the great principles of Ifeffit (and thus Artemis) is that the path parameters do not have to be the same as the fit parameters.

As a simple example, it's good to start with the model that all paths have the same value for E0. So you can guess a single parameter and then use that parameter for the E0 for all paths.

As another simple example, a cubic crystal might be modelled as having a uniform thermal (or Vegard's law) expansion. Then there could be a single guessed parameter indicating the fractional expansion (called, for example, alpha) while the delr for each path could then be entered as alpha*reff.

Much more complicated constraint schemes are possible, but for fits with large number of paths, the number of guessed parameters is almost always much much lower than the number of path parameters.

...

As shown on the fit Log file, the Correlation value between two fit parameters is the bigger the better, or the smaller the better?

Not necessarily. Correlations are provided to help understand the relationship between guessed parameters, but there's nothing inherently wrong with a high correlation...the uncertainty associated with that relationship is already represented in the uncertainties reported with the fit. In other words, it is more direct to focus on getting the uncertainties to be lower, rather than the correlations.

--Scott Calvin Sarah Lawrence College

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Thank you Bruce. I found the videos are very helpful now. Best, Yanyun Quoting Bruce Ravel :

Yanyun, Kug-Seung,

Have you watched the videos that the folks at Diamond kindly host at

http://www.diamond.ac.uk/Beamlines/Spectroscopy/Techniques/XAS.html

The ones titled "Introduction to Artemis and EXAFS fitting" and "More Artemis and EXAFS fitting" go a long way towards answering the questions you both are asking. Granted, you have to look at my ugly face for a long time, which is pretty horrifying. But I think you might find them helpful.

B

-- Bruce Ravel ------------------------------------ bravel@bnl.gov

National Institute of Standards and Technology Synchrotron Science Group at NSLS --- Beamlines U7A, X24A, X23A2 Building 535A Upton NY, 11973

Homepage: http://bruceravel.github.io/home/ Software: https://github.com/bruceravel Demeter: http://bruceravel.github.io/demeter/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit