Trouble with fitting with Arthemis

Kleper Oliveira Rocha

13 Jan 2009 13 Jan '09

10:13 p.m.

Hi all, Please, help me. When I try to do any fit in Arthemis, setting all parameters unless one, been this one any of the parameters, the fit gives for the answer -1.0000 +/- 0.000000 like example down. What is happening? ============================================================ Independent points = 8.239257812 Number of variables = 1.000000000 Chi-square = 0.120000000E+37 Reduced Chi-square = 0.165762849E+36 R-factor = NaN Measurement uncertainty (k) = 0.000437667 Measurement uncertainty (R) = 0.000724906 Number of data sets = 1.000000000 Guess parameters +/- uncertainties (initial guess): * amp = -1.0000000 +/- 0.0000000 (1.0000)* Set parameters: enot = 0 delr = 0 ss = 0.003 c3 = 0.0001 c4 = 0.00001 -- Kleper de Oliveira Rocha Chemical Engenieer Doctorated Tels. 55 016 3351-8694

Attachments:

attachment.html (text/html — 2.2 KB)

Show replies by date

Scott Calvin

13 Jan 13 Jan

10:20 p.m.

Hi Kleper, It looks a little to me like you don't actually have any paths included in the fit, or perhaps the path is defective (all zeroed out, or something like that). If you attach the project file, it will probably be easy to tell. --Scott On Jan 13, 2009, at 5:13 PM, Kleper Oliveira Rocha wrote:

...

Hi all,

Please, help me. When I try to do any fit in Arthemis, setting all parameters unless one, been this one any of the parameters, the fit gives for the answer -1.0000 +/- 0.000000 like example down. What is happening?

============================================================ Independent points = 8.239257812 Number of variables = 1.000000000 Chi-square = 0.120000000E+37 Reduced Chi-square = 0.165762849E+36 R-factor = NaN Measurement uncertainty (k) = 0.000437667 Measurement uncertainty (R) = 0.000724906 Number of data sets = 1.000000000

Guess parameters +/- uncertainties (initial guess): amp = -1.0000000 +/- 0.0000000 (1.0000) Set parameters: enot = 0 delr = 0 ss = 0.003 c3 = 0.0001 c4 = 0.00001

-- Kleper de Oliveira Rocha Chemical Engenieer Doctorated

Bruce Ravel

10:23 p.m.

Hi Kleper, But your reduced chi-square is only 10^35. That seems like a pretty good fit... ;-) There is obviously a numerical problem, corrupted data, corrupted project file -- something like that. Can you send me the project file? B

...

Please, help me. When I try to do any fit in Arthemis, setting all parameters unless one, been this one any of the parameters, the fit gives for the answer -1.0000 +/- 0.000000 like example down. What is happening?

============================================================ Independent points = 8.239257812 Number of variables = 1.000000000 Chi-square = 0.120000000E+37 Reduced Chi-square = 0.165762849E+36 R-factor = NaN Measurement uncertainty (k) = 0.000437667 Measurement uncertainty (R) = 0.000724906 Number of data sets = 1.000000000

Guess parameters +/- uncertainties (initial guess): * amp = -1.0000000 +/- 0.0000000 (1.0000)* Set parameters: enot = 0 delr = 0 ss = 0.003 c3 = 0.0001 c4 = 0.00001

-- Bruce Ravel ------------------------------------ bravel@bnl.gov National Institute of Standards and Technology Synchrotron Methods Group at NSLS --- Beamlines U7A, X24A, X23A2 Building 535A Upton NY, 11973 My homepage: http://xafs.org/BruceRavel EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/

Kleper Oliveira Rocha

10:31 p.m.

Hi Bruce, I send to you the project. 2009/1/13 Bruce Ravel

...

Hi Kleper,

But your reduced chi-square is only 10^35. That seems like a pretty good fit... ;-)

There is obviously a numerical problem, corrupted data, corrupted project file -- something like that.

Can you send me the project file?

B

...
Please, help me. When I try to do any fit in Arthemis, setting all parameters unless one, been this one any of the parameters, the fit gives for the answer -1.0000 +/- 0.000000 like example down. What is happening?

============================================================ Independent points = 8.239257812 Number of variables = 1.000000000 Chi-square = 0.120000000E+37 Reduced Chi-square = 0.165762849E+36 R-factor = NaN Measurement uncertainty (k) = 0.000437667 Measurement uncertainty (R) = 0.000724906 Number of data sets = 1.000000000

Guess parameters +/- uncertainties (initial guess): * amp = -1.0000000 +/- 0.0000000 (1.0000)* Set parameters: enot = 0 delr = 0 ss = 0.003 c3 = 0.0001 c4 = 0.00001

--

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

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

My homepage: http://xafs.org/BruceRavel EXAFS software: http://cars9.uchicago.edu/~ravel/software/exafs/ _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

-- Kleper de Oliveira Rocha Doutorando em Engenharia Química Departamento de Engenharia Química/UFSCar Tels. 55 016 3351-8694

Scott Calvin

10:37 p.m.

That's an Athena project, Kleper. I think we need the Artemis project (.apj file). Nothing wrong with the data itself, though, as far as I can see. --Scott Calvin Sarah Lawrence College On Jan 13, 2009, at 5:31 PM, Kleper Oliveira Rocha wrote:

...

Hi Bruce,

I send to you the project.

2009/1/13 Bruce Ravel

Hi Kleper,

But your reduced chi-square is only 10^35. That seems like a pretty good fit... ;-)

There is obviously a numerical problem, corrupted data, corrupted project file -- something like that.

Can you send me the project file?

B

Kleper Oliveira Rocha

10:43 p.m.

But Scott, this data was that I treat in Athena to begin the treatment in Arthemis. I don´t have the Artemis project. 2009/1/13 Scott Calvin

...

That's an Athena project, Kleper. I think we need the Artemis project (.apj file). Nothing wrong with the data itself, though, as far as I can see.

--Scott Calvin Sarah Lawrence College

On Jan 13, 2009, at 5:31 PM, Kleper Oliveira Rocha wrote:

Hi Bruce,

I send to you the project.

2009/1/13 Bruce Ravel

...
Hi Kleper,

But your reduced chi-square is only 10^35. That seems like a pretty good fit... ;-)

There is obviously a numerical problem, corrupted data, corrupted project file -- something like that.

Can you send me the project file?

B

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

-- Kleper de Oliveira Rocha Doutorando em Engenharia Química Departamento de Engenharia Química/UFSCar Tels. 55 016 3351-8694

Scott Calvin

10:48 p.m.

OK, but the problem doesn't seem to be with the data; I'd guess it's with the path somehow. So you should save the Artemis project you used to generate the log file you sent us, and send that. (In other words, just follow the same steps you did before, but use Artemis to save the project.) --Scott Calvin Sarah Lawrence College On Jan 13, 2009, at 5:43 PM, Kleper Oliveira Rocha wrote:

...

But Scott, this data was that I treat in Athena to begin the treatment in Arthemis. I don´t have the Artemis project.

2009/1/13 Scott Calvin That's an Athena project, Kleper. I think we need the Artemis project (.apj file). Nothing wrong with the data itself, though, as far as I can see.

--Scott Calvin Sarah Lawrence College

Kleper Oliveira Rocha

11:32 p.m.

Ok, Scott. I understand. Follow in attachemt. 2009/1/13 Scott Calvin

...

OK, but the problem doesn't seem to be with the data; I'd guess it's with the path somehow. So you should save the Artemis project you used to generate the log file you sent us, and send that. (In other words, just follow the same steps you did before, but use Artemis to save the project.)

--Scott Calvin Sarah Lawrence College

On Jan 13, 2009, at 5:43 PM, Kleper Oliveira Rocha wrote:

But Scott, this data was that I treat in Athena to begin the treatment in Arthemis. I don´t have the Artemis project.

2009/1/13 Scott Calvin

...
That's an Athena project, Kleper. I think we need the Artemis project (.apj file). Nothing wrong with the data itself, though, as far as I can see.

--Scott Calvin Sarah Lawrence College

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

-- Kleper de Oliveira Rocha Doutorando em Engenharia Química Departamento de Engenharia Química/UFSCar Tels. 55 016 3351-8694

Scott Calvin

11:47 p.m.

Thanks. I looked at it, and everything works--there's no corrupted data; it's just having a hard time getting a decent fit. If you guess multiple parameters you'll see it does manage to find some kind of fit, but it's poor. Your data and background subtraction look acceptable, so that leaves the model. Perhaps you don't have platinum metal. After all, the files are named PtAl--if there are a lot of Pt-Al near neighbors, and you're not accounting for them, that could explain why the fit fails. --Scott Calvin Sarah Lawrence College On Jan 13, 2009, at 6:32 PM, Kleper Oliveira Rocha wrote:

...

Ok, Scott. I understand. Follow in attachemt.

Carlo Segre

11:55 p.m.

Hi Kleper: There is a single Pt-Pt path in your fit. This is effectively a fcc Pt distance of 2.77. I see several problems: This spectrum has distances which are much shorter than 2.77 and so I would expect that this single path would not be a very good fit. I am not sure if the peak around 1.1A is real, but it could be given the hydrogen that you indicate is present in the file name. What is clear is that the bigger peaks are not Pt-Pt peaks at all. The Pt-Pt peak will usually appear around 2.5A in Chi(R). The second thing that I noticed, is that you must have a problem with your installation of the software because I just took your file and executed the fit and did not get the strange values for Chi^2 and reduced Chi^2 that you have posted. The value for amp is negative, indicating that the path is not a good model, but the Chi^2 I got was around 1000. I can't tell much form the name of your file but I suspect that you have Pt-Al and possibly Pt-H distances in your sample. Do you have any other information about this sample? 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 segre@iit.edu http://www.iit.edu/~segre segre@debian.org

Kleper Oliveira Rocha

14 Jan 14 Jan

2:02 a.m.

Hi Carlo, You seem to be right. I try to do the fit in my notebook, which has the same Windows and Services Packs, and worked. I want to know why? Is it possible to resolve this? You right too assuming that there is an H2 adsorbed. In reality, this is a sample catalyst of 1%Pt supported in Al2O3 that was treated in H2. The data of EXAFS was colleted at 500ºC in an flux of H2. It is a catalysts prepared by impregnation method. In the fit model I assuming that is possible to has a oxigen neighbor to Pt, therefore, there is a PtO2 model atom to do the fit. They work well with one path unless by the peak at 1A. I already thought to assuming a H atom neighbor to Pt but, I not find a good model to this. Do you have some idea if is possible to do that? Do you belives that I should take more than one path for Pt-Pt distance? 2009/1/13 Carlo Segre

...

Hi Kleper:

There is a single Pt-Pt path in your fit. This is effectively a fcc Pt distance of 2.77. I see several problems:

This spectrum has distances which are much shorter than 2.77 and so I would expect that this single path would not be a very good fit. I am not sure if the peak around 1.1A is real, but it could be given the hydrogen that you indicate is present in the file name. What is clear is that the bigger peaks are not Pt-Pt peaks at all. The Pt-Pt peak will usually appear around 2.5A in Chi(R).

The second thing that I noticed, is that you must have a problem with your installation of the software because I just took your file and executed the fit and did not get the strange values for Chi^2 and reduced Chi^2 that you have posted. The value for amp is negative, indicating that the path is not a good model, but the Chi^2 I got was around 1000.

I can't tell much form the name of your file but I suspect that you have Pt-Al and possibly Pt-H distances in your sample. Do you have any other information about this sample?

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 segre@iit.edu http://www.iit.edu/~segre segre@debian.org _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

-- Kleper de Oliveira Rocha Doutorando em Engenharia Química Departamento de Engenharia Química/UFSCar Tels. 55 016 3351-8694

Carlo Segre

2:59 a.m.

I think you need to start with the shortest paths first and then work your way to the longer pt-Pt paths. It is likely that there is some Pt-Pt there but it is swamped by Pt-O and others as well. The first thing you need to figure out is if the peak at 1.1A is real or not. You might want to look inthe literature for Pt-H distances. I know that Pt-O distances are generally close to 2.0A so that is likely the shoulder below 2.0A It is possible that the peak just above 2.0A is Pt-Al but you might want to look this up in the literature. Once you have worked out the short distances, you can start to put the Pt-Pt distances in. It is certainly likely that there is more than one but the Pt clusters are probably very small. Cheers, Carlo On Tue, 13 Jan 2009, Kleper Oliveira Rocha wrote:

...

Hi Carlo,

You seem to be right. I try to do the fit in my notebook, which has the same Windows and Services Packs, and worked. I want to know why? Is it possible to resolve this? You right too assuming that there is an H2 adsorbed. In reality, this is a sample catalyst of 1%Pt supported in Al2O3 that was treated in H2. The data of EXAFS was colleted at 500ºC in an flux of H2. It is a catalysts prepared by impregnation method. In the fit model I assuming that is possible to has a oxigen neighbor to Pt, therefore, there is a PtO2 model atom to do the fit. They work well with one path unless by the peak at 1A. I already thought to assuming a H atom neighbor to Pt but, I not find a good model to this. Do you have some idea if is possible to do that? Do you belives that I should take more than one path for Pt-Pt distance?

2009/1/13 Carlo Segre

...
Hi Kleper:

There is a single Pt-Pt path in your fit. This is effectively a fcc Pt distance of 2.77. I see several problems:

This spectrum has distances which are much shorter than 2.77 and so I would expect that this single path would not be a very good fit. I am not sure if the peak around 1.1A is real, but it could be given the hydrogen that you indicate is present in the file name. What is clear is that the bigger peaks are not Pt-Pt peaks at all. The Pt-Pt peak will usually appear around 2.5A in Chi(R).

The second thing that I noticed, is that you must have a problem with your installation of the software because I just took your file and executed the fit and did not get the strange values for Chi^2 and reduced Chi^2 that you have posted. The value for amp is negative, indicating that the path is not a good model, but the Chi^2 I got was around 1000.

I can't tell much form the name of your file but I suspect that you have Pt-Al and possibly Pt-H distances in your sample. Do you have any other information about this sample?

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 segre@iit.edu http://www.iit.edu/~segre segre@debian.org _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

-- 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 segre@iit.edu http://www.iit.edu/~segre segre@debian.org

5857

Age (days ago)

5858

Last active (days ago)

List overview

Download

11 comments

4 participants

participants (4)

Bruce Ravel
Carlo Segre
Kleper Oliveira Rocha
Scott Calvin

Trouble with fitting with Arthemis

tags

participants (4)