Phase corrected Fourier transforms

older
Re: [Ifeffit] Artemis error message

Juan Antonio Maciá Agulló

22 Sep 2006 22 Sep '06

7:16 p.m.

Hi all, I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication? I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why. I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right? I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV. I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong? Sorry for these easy questions but I am a novice in XAFS. Thank you very much. Best regards, JA

Attachments:

Juan.Macia.vcf (text/x-vcard — 284 bytes)

Show replies by date

Scott Calvin

23 Sep 23 Sep

12:27 a.m.

Hi JA, I'll put in my two cents on some of these questions, although in some cases it's a matter of personal taste: At 03:16 PM 9/22/2006, you wrote:

...

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication?

I vote not. :) Most of the time, phase corrections are desired for cosmetic reasons, to make the FT look more like a radial distribution function. But the FT is never a RDF, and so an attempt to make it look like one may mislead some readers without much experience in EXAFS. Also, adding a phase correction necessitates a discussion of what method for applying the phase correction was used, which complicates the publication. On the other hand, I'm less opposed to using phase-correction for presentations and posters intended for a non-EXAFS audience, although I still don't do it. Like I say, this is a matter of personal tatse.

...

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

No, this does not mean the fit is wrong! It is hard to eliminate the correlations you describe. And the algorithm used by Ifeffit to calculate the uncertainties works in such a way that you don't have to worry about correlations adding extra uncertainty beyond what's reported. High correlations do help you, as the person doing the fits, diagnose why an uncertainty might be high. If you find the uncertainty for deltaR is too high to say anything useful about a bond length, for example, then check the correlation with deltaE0. If the correlation is high, there are a number of things you can try to reduce the correlation, and, hopefully, the uncertainties: fitting multiple k-weights, fitting more coordination shells, etc..

...

Sorry for these easy questions but I am a novice in XAFS.

That's one of the things this list is for... --Scott Calvin Sarah Lawrence College

John J. Rehr

2:18 p.m.

Dear Juan Antonio, I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following: 1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted. Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT. J. Rehr On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maci� Agull� wrote:

...

Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, �phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

Matthew Marcus

8:13 p.m.

OK, I have to put in my US$0.02/e0.02. The phase correction can't mean much if the scattering atoms are different, which is usually the case. If the scattering atoms are very heavy, then the phase correction has a kink in it which could cause strange shapes if applied to light-atom shells. I've played with this phase correction and amplitude correction, while I was at it, and not been very impressed. One place where it might be useful is in the aforementioned case of heavy and light scatterers. If you use the correction for one of these, then the corresponding shells sharpen and the other ones blur out, so you can get a rough idea of who's who. I suspect this works better if you do the amplitude as well. A related technique is to use model compounds+FEFF to get 'semi-empirical' amps and phases which include all the artifacts of the experiment. Suppose, for instance, that you're looking at Cu in a matrix of Fe, and you have data for Cu metal but not for any known Cu->Fe scattering pair. You can synthesize a 'semi-empirical' Cu->Fe phase and amp like this: phi(Cu->Fe) = phi_exp(Cu->Cu)+(phi_theor(Cu->Fe)-phi_theor(Cu->Cu)) amp(Cu->Fe) = amp_exp(Cu->Cu)*amp_theor(Cu->Fe)/amp_theor(Cu->Cu) If you don't want to do this for modeling, you can use this method for comparing two spectra which you think might be alike except for the central atom, by "correcting" one spectrum with the difference to make it comparable to the other. Another aspect to this whole thing is that people are very used to uncorrected FT's and are aware that you have to add 0.3-0.4A to the distances. If you show only corrected FT's, I wonder if that will be satisfying to the audience. mam ----- Original Message ----- From: "John J. Rehr" To: "XAFS Analysis using Ifeffit" Cc: "John J. Rehr" Sent: Saturday, September 23, 2006 7:18 AM Subject: Re: [Ifeffit] Phase corrected Fourier transforms Dear Juan Antonio, I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following: 1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted. Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT. J. Rehr On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maciá Agulló wrote:

...

Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

--------------------------------------------------------------------------------

...

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

Anatoly Frenkel

10:15 p.m.

My 0.533 Rouble: In my experience with some Mn oxides, the Mn-O FT magnitude peak's position is 0.5 A lower its corresponding 1NN bond length, while the Mn-Mn are 0.3 A lower than their bond lenghts. Thus, it would be misleading for this and other similar compounds to apply theoretical phase correction of the 1NN to the entire data, as it will shift only one peak correctly. Anatoly -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Matthew Marcus Sent: Saturday, September 23, 2006 4:14 PM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] Phase corrected Fourier transforms OK, I have to put in my US$0.02/e0.02. The phase correction can't mean much if the scattering atoms are different, which is usually the case. If the scattering atoms are very heavy, then the phase correction has a kink in it which could cause strange shapes if applied to light-atom shells. I've played with this phase correction and amplitude correction, while I was at it, and not been very impressed. One place where it might be useful is in the aforementioned case of heavy and light scatterers. If you use the correction for one of these, then the corresponding shells sharpen and the other ones blur out, so you can get a rough idea of who's who. I suspect this works better if you do the amplitude as well. A related technique is to use model compounds+FEFF to get 'semi-empirical' amps and phases which include all the artifacts of the experiment. Suppose, for instance, that you're looking at Cu in a matrix of Fe, and you have data for Cu metal but not for any known Cu->Fe scattering pair. You can synthesize a 'semi-empirical' Cu->Fe phase and amp like this: phi(Cu->Fe) = phi_exp(Cu->Cu)+(phi_theor(Cu->Fe)-phi_theor(Cu->Cu)) amp(Cu->Fe) = amp_exp(Cu->Cu)*amp_theor(Cu->Fe)/amp_theor(Cu->Cu) If you don't want to do this for modeling, you can use this method for comparing two spectra which you think might be alike except for the central atom, by "correcting" one spectrum with the difference to make it comparable to the other. Another aspect to this whole thing is that people are very used to uncorrected FT's and are aware that you have to add 0.3-0.4A to the distances. If you show only corrected FT's, I wonder if that will be satisfying to the audience. mam ----- Original Message ----- From: "John J. Rehr" To: "XAFS Analysis using Ifeffit" Cc: "John J. Rehr" Sent: Saturday, September 23, 2006 7:18 AM Subject: Re: [Ifeffit] Phase corrected Fourier transforms Dear Juan Antonio, I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following: 1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted. Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT. J. Rehr On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maciá Agulló wrote:

...

Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

---------------------------------------------------------------------------- ----

...

_______________________________________________ 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

John J. Rehr

24 Sep 24 Sep

4 a.m.

Anatoly is correct that the phase correction cannot work for all shells in inhomogeneous systems. However, it's *always* possible to remove the central atom phase shift 2 delta_c which often dominates the phase correction. Second, if one corrects by the 1st nn phase, delta_1, the dominant near neighbor shell will appear at the correct distance. The correction is not right for subsequent shells but the error is given by a phase difference delta R = <(1/2)(d/dk) [delta_n-delta_1]> This correction is typically order 0.1 Ang and usually smaller than the typical shift of about 0.3 - 0.4 Ang in uncorrected FT peaks. The upshot is that the phase corrected FT can be correct for the first shell which often dominates the FT and generally better than uncorrected FTs. Corrective lenses for vision aren't perfect either at all distances, but it's hard to argue that it's preferable not to use them for that reason. J. Rehr On Sat, 23 Sep 2006, Anatoly Frenkel wrote:

...

My 0.533 Rouble: In my experience with some Mn oxides, the Mn-O FT magnitude peak's position is 0.5 A lower its corresponding 1NN bond length, while the Mn-Mn are 0.3 A lower than their bond lenghts. Thus, it would be misleading for this and other similar compounds to apply theoretical phase correction of the 1NN to the entire data, as it will shift only one peak correctly.

Anatoly

-----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Matthew Marcus Sent: Saturday, September 23, 2006 4:14 PM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] Phase corrected Fourier transforms

OK, I have to put in my US$0.02/e0.02. The phase correction can't mean much if the scattering atoms are different, which is usually the case. If the scattering atoms are very heavy, then the phase correction has a kink in it which could cause strange shapes if applied to light-atom shells. I've played with this phase correction and amplitude correction, while I was at it, and not been very impressed. One place where it might be useful is in the aforementioned case of heavy and light scatterers. If you use the correction for one of these, then the corresponding shells sharpen and the other ones blur out, so you can get a rough idea of who's who. I suspect this works better if you do the amplitude as well.

A related technique is to use model compounds+FEFF to get 'semi-empirical' amps and phases which include all the artifacts of the experiment. Suppose, for instance, that you're looking at Cu in a matrix of Fe, and you have data for Cu metal but not for any known Cu->Fe scattering pair. You can synthesize a 'semi-empirical' Cu->Fe phase and amp like this:

phi(Cu->Fe) = phi_exp(Cu->Cu)+(phi_theor(Cu->Fe)-phi_theor(Cu->Cu)) amp(Cu->Fe) = amp_exp(Cu->Cu)*amp_theor(Cu->Fe)/amp_theor(Cu->Cu)

If you don't want to do this for modeling, you can use this method for comparing two spectra which you think might be alike except for the central atom, by "correcting" one spectrum with the difference to make it comparable to the other.

Another aspect to this whole thing is that people are very used to uncorrected FT's and are aware that you have to add 0.3-0.4A to the distances. If you show only corrected FT's, I wonder if that will be satisfying to the audience. mam

----- Original Message ----- From: "John J. Rehr" To: "XAFS Analysis using Ifeffit" Cc: "John J. Rehr" Sent: Saturday, September 23, 2006 7:18 AM Subject: Re: [Ifeffit] Phase corrected Fourier transforms

Dear Juan Antonio,

I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following:

1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted.

Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT.

J. Rehr

On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maci� Agull� wrote:

...
Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, �phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

---------------------------------------------------------------------------- ----

...
_______________________________________________ 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

Kelly, Shelly D.

2:15 p.m.

Hi Juan and John, Here us my 2cents worth... Unfortunately the magnitude of the FT is not as simple as a RDF. I agree with John that the ultimate goal of the analysis is to come up with a RDF. But I disagree about the worth of adding a bit more processing to the spectra that makes the spectra just a bit more like an RDF but leaves the data so far from it, such that true analysis is still needed. If the phase shift correction actually allowed the spectra to be inverted, we would all do it! I started a page the XAFS.org page under tutorials about the information that should be reported with EXAFS spectra. On that page I added a link to some of my favorite rants, including a section about why the Magnitude of the FT is not a radial distribution function. http://xafs.org/Common_Mistakes I encourage everyone on the list to read and change these pages and add your own pages. I think that this is a really great way to come up with a community voice. Here are a few thoughts about why I don't favor adding the phase shit. (1) Often 2nd and 3rd neighboring peaks don't turn-up as peaks at all in the FT because of the interference between themselves and the strong signal from the first shell. In this case the "peak" is a shoulder or sometimes even the opposite of a shoulder were the first peak is shaved off on the side. The phase correction isn't going to fix that. As John mentioned (2) multiple scattering paths are not supposed to be in a radial distribution function and they often show up as peaks in the Mag of the FT and (3) some signals are not a single peak, but a doublet in the magnitude of the FT. The phase correction doesn't "fix" multiple scattering paths or the doublets. I'm not sure that I want the multiple scattering paths fixed, they often have a lot of information in them. (4) A Fourier transform will never be able to tell the difference between neighboring atom types as a radial distribution function. (5) In addition the phase correction in Athena only corrects for half of the phase shift, the part from the central atom. The other half of the phase shift comes from the scattering atom type, which can only be corrected in Artemis. I do not recommend using the phase shit option in Athena because it does not turn the data into a radial distribution function. The magnitude of the FT is only an envelope function. A lot of information within the spectra is not displayed in the magnitude of the FT. It is the chi(k), and the real and imaginary parts of the FT which show the complexity of the spectroscopy. I think that this complexity should be embraced not hidden. In John's language, I think that the picture is more blurred by adding a phase correction to the spectra mainly because it promises so much but actually adds so little. It seems to me that adding the phase shit, indicates that there is something inherently "wrong" with the Magnitude of the Fourier transform of the spectra. I don't endorse that point of view. Truly, a radial distribution function is the goal of most analysis, so I encourage the display of the individual contributions/paths to the total model. I think that this display of all the paths particularly in the chi(k), real and/or imaginary parts of the FT adds a lot of transparency to the data analysis and interpretation. Well, Juan you walked right into that one. Welcome to the mailing list...:) Shelly -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of John J. Rehr Sent: Saturday, September 23, 2006 11:01 PM To: XAFS Analysis using Ifeffit Cc: John J. Rehr; anatoly.frenkel@yu.eduA Subject: RE: [Ifeffit] Phase corrected Fourier transforms Anatoly is correct that the phase correction cannot work for all shells in inhomogeneous systems. However, it's *always* possible to remove the central atom phase shift 2 delta_c which often dominates the phase correction. Second, if one corrects by the 1st nn phase, delta_1, the dominant near neighbor shell will appear at the correct distance. The correction is not right for subsequent shells but the error is given by a phase difference delta R = <(1/2)(d/dk) [delta_n-delta_1]> This correction is typically order 0.1 Ang and usually smaller than the typical shift of about 0.3 - 0.4 Ang in uncorrected FT peaks. The upshot is that the phase corrected FT can be correct for the first shell which often dominates the FT and generally better than uncorrected FTs. Corrective lenses for vision aren't perfect either at all distances, but it's hard to argue that it's preferable not to use them for that reason. J. Rehr On Sat, 23 Sep 2006, Anatoly Frenkel wrote:

...

My 0.533 Rouble: In my experience with some Mn oxides, the Mn-O FT magnitude peak's position is 0.5 A lower its corresponding 1NN bond length, while the Mn-Mn are 0.3 A lower than their bond lenghts. Thus, it would be misleading for this and other similar compounds to apply theoretical phase correction of the 1NN to the entire data, as it will shift only one peak correctly.

Anatoly

-----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Matthew Marcus Sent: Saturday, September 23, 2006 4:14 PM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] Phase corrected Fourier transforms

OK, I have to put in my US$0.02/e0.02. The phase correction can't mean much if the scattering atoms are different, which is usually the case. If the scattering atoms are very heavy, then the phase correction has a kink in it which could cause strange shapes if applied to light-atom shells. I've played with this phase correction and amplitude correction, while I was at it, and not been very impressed. One place where it might be useful is in the aforementioned case of heavy and light scatterers. If you use the correction for one of these, then the corresponding shells sharpen and the other ones blur out, so you can get a rough idea of who's who. I suspect this works better if you do the amplitude as well.

A related technique is to use model compounds+FEFF to get 'semi-empirical' amps and phases which include all the artifacts of the experiment. Suppose, for instance, that you're looking at Cu in a matrix of Fe, and you have data for Cu metal but not for any known Cu->Fe scattering pair. You can synthesize a 'semi-empirical' Cu->Fe phase and amp like this:

phi(Cu->Fe) = phi_exp(Cu->Cu)+(phi_theor(Cu->Fe)-phi_theor(Cu->Cu)) amp(Cu->Fe) = amp_exp(Cu->Cu)*amp_theor(Cu->Fe)/amp_theor(Cu->Cu)

If you don't want to do this for modeling, you can use this method for comparing two spectra which you think might be alike except for the central atom, by "correcting" one spectrum with the difference to make it comparable to the other.

Another aspect to this whole thing is that people are very used to uncorrected FT's and are aware that you have to add 0.3-0.4A to the distances. If you show only corrected FT's, I wonder if that will be satisfying to the audience. mam

----- Original Message ----- From: "John J. Rehr" To: "XAFS Analysis using Ifeffit" Cc: "John J. Rehr" Sent: Saturday, September 23, 2006 7:18 AM Subject: Re: [Ifeffit] Phase corrected Fourier transforms

Dear Juan Antonio,

I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following:

1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted.

Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT.

J. Rehr

On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maciá Agulló wrote:

...
Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

---------------------------------------------------------------------------- ----

...
_______________________________________________ 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

Scott Calvin

2:34 p.m.

I use Eudora for email, and with the options right it marks with "hot peppers" messages that might be offensive. I leave that option on, because I find it amusing. Well, sure enough, it marked Shelly's post with 3 peppers, the highest rating! When I looked more closely, I discovered why: At 10:15 AM 9/24/2006, you wrote:

...

I do not recommend using the phase shit option

Now THAT is a funny typo! At least, I'm assuming it's a typo...Shelly did it twice. :D --Scott Calvin Sarah Lawrence College

Anatoly Frenkel

2:39 p.m.

I counted it 3 times. Anatoly

Anatoly Frenkel

2:42 p.m.

In the future, I recommend "phase s..t" format for those who really really hates it. It will be less offensive. Anatoly

Kelly, Shelly D.

2:47 p.m.

Oh crap, I didn't mean it like that. SK -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of Anatoly Frenkel Sent: Sunday, September 24, 2006 9:42 AM To: XAFS Analysis using Ifeffit Subject: RE: [Ifeffit] Phase corrected Fourier transforms In the future, I recommend "phase s..t" format for those who really really hates it. It will be less offensive. Anatoly _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Bruce Ravel

25 Sep 25 Sep

3:23 a.m.

On Sunday 24 September 2006 09:34, Scott Calvin wrote:

...

I use Eudora for email, and with the options right it marks with "hot peppers" messages that might be offensive. I leave that option on, because I find it amusing. Well, sure enough, it marked Shelly's post with 3 peppers, the highest rating! When I looked more closely, I discovered why:

At 10:15 AM 9/24/2006, you wrote:

...
I do not recommend using the phase shit option

Now THAT is a funny typo! At least, I'm assuming it's a typo...Shelly did it twice. :D

This is quite possibly the funniest email ever on the mailing list! B -- Bruce Ravel ---------------------------------------------- bravel@anl.gov Molecular Environmental Science Group, Building 203, Room E-165 MRCAT, Sector 10, Advance 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/

Juan Antonio Maciá Agulló

8:36 a.m.

Thank you very much for your answers. It seems that phase correction is a hot topic (not hot peppers) and everybody is partially right. I will study yours answers to decide which is my favorite "team". I agree with Bruce that this is quite possibly the funniest email, better than Scott Calvin's rule. Sorry to be so tedious, but I still have too easy questions:

...

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

...

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

Thank you very much again. Best regards, JA

Bruce Ravel

3:40 p.m.

On Monday 25 September 2006 03:36, Juan Antonio Maciá Agulló wrote:

...

Sorry to be so tedious, but I still have too easy questions:

Juan, As other have said, the purpose of the mailing list is to have a place to ask questions at any level about spectroscopy and data analysis. You seem to be quite adept at starting very interesting discussions. I think that's cool! 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/

Juan Antonio Maciá Agulló

5:56 p.m.

Thank Bruce, but I am only a beginner full of big doubts. JA

...

As other have said, the purpose of the mailing list is to have a place to ask questions at any level about spectroscopy and data analysis. You seem to be quite adept at starting very interesting discussions. I think that's cool!

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/

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

Juan Antonio Maciá Agulló

26 Sep 26 Sep

7:52 p.m.

New subject: k-weight in Artemis

Hi all, I read in Artemis manual that a good fit must be consistent for different k-weights. When two fits (one with k1 and other with k3) are consistent? I mean, which is maximum allowed difference, 1%, 5%...between these two fits? I have done different fits changing fit k-weights and k-weights in the plotting options. Usually, for k-weights = 1 or 2, I have obtained very strange spectra (why?) and then I have decided to change only k-weight in the plotting options to 3. After that, spectra looked more like before. I am not sure if this gimmick is correct. I mean, can I use different k-weights for fit and for plotting? I have obtained "consistent" fits with fit k-weights of 1 and 3, and both plotted in k-weight 3, that's ok? Sometimes when I have changed fit k-weight from 3 to 1, I have observed that two high correlations (between ss and S02, and delR and delE0) have dissapeared (<0.85). What does it mean? Parameters obtained with fit k-weight = 1 are more reliable? And another question, I have seen an option to fit background in Artemis, is it obligatory to fit the background to publish EXAFS fits? Best regards, Thanks a lot JA

Scott Calvin

8:45 p.m.

New subject: k-weight in Artemis

Hi Juan, Excellent questions! At 03:52 PM 9/26/2006, Juan Antonio Maciá Agulló wrote:

...

I read in Artemis manual that a good fit must be consistent for different k-weights. When two fits (one with k1 and other with k3) are consistent? I mean, which is maximum allowed difference, 1%, 5%...between these two fits?

Ideally, the fitted parameters should have uncertainty ranges that overlap. For example, one bond length might be 2.03 +/- 0.04 angstroms, and the equivalent bond length using a different k-weight might be 2.09 +/- 0.03 angstroms.

...

I have done different fits changing fit k-weights and k-weights in the plotting options. Usually, for k-weights = 1 or 2, I have obtained very strange spectra (why?) and then I have decided to change only k-weight in the plotting options to 3. After that, spectra looked more like before. I am not sure if this gimmick is correct. I mean, can I use different k-weights for fit and for plotting? I have obtained "consistent" fits with fit k-weights of 1 and 3, and both plotted in k-weight 3, that's ok?

What do you mean, "strange" spectra? They will look different for different k-weights, of course. That's not a problem. In general, it's OK to fit in one k-weight and plot in another, and there are a variety of good reasons for doing that: for example, you're comparing to published data with a certain k-weight.

...

Sometimes when I have changed fit k-weight from 3 to 1, I have observed that two high correlations (between ss and S02, and delR and delE0) have dissapeared (<0.85). What does it mean? Parameters obtained with fit k-weight = 1 are more reliable?

Umm...not more reliable. Just less correlated. :) You can look at the uncertainties on the parameters to see if that also was reduced when you changed k-weights. Sometimes one k-weight manages to break correlations more effectively than another, but there isn't a simple pattern to it. Different parameters are weighted by different powers of k in the EXAFS equations, and different scattering elements have large amplitudes at different values of k, so the overall effect is sometimes difficult to predict: it's best to just try it and see what happens. Ideally, the different k-weights will yield similar values for the parameters, although the graphs may look quite different.

...

And another question, I have seen an option to fit background in Artemis, is it obligatory to fit the background to publish EXAFS fits?

No. This is another one of these issues where people can reasonably have different preferences. Personally, I prefer not to use the Artemis background refinement in my published fits, partly because it's one more processing step to explain and justify. But I do use the ability of Artemis to do background refinement to test if my Athena background is OK: if fitted parameters correlated highly with Artemis background parameters, I know there's a problem, and I go back and look at the background subtraction process in Athena again. If they don't correlate highly, then the Artemis background wasn't doing anything except improving the visual aesthetics of the fit anyway, and I turn it back off. --Scott Calvin Sarah Lawrence College

Bruce Ravel

9:53 p.m.

New subject: k-weight in Artemis

On Tuesday 26 September 2006 15:45, Scott Calvin wrote:

...

...
And another question, I have seen an option to fit background in Artemis, is it obligatory to fit the background to publish EXAFS fits?

No. This is another one of these issues where people can reasonably have different preferences. Personally, I prefer not to use the Artemis background refinement in my published fits, partly because it's one more processing step to explain and justify. But I do use the ability of Artemis to do background refinement to test if my Athena background is OK: if fitted parameters correlated highly with Artemis background parameters, I know there's a problem, and I go back and look at the background subtraction process in Athena again. If they don't correlate highly, then the Artemis background wasn't doing anything except improving the visual aesthetics of the fit anyway, and I turn it back off.

I'd like to expand on this a bit. At the level of Artemis, users are often concerned that the background corefinement is somehow a cheat. It isn't. (Well, not any more than anything else Artemis does -- but that's a separate story ;-) The way to think about it is in terms of Fourier components. Artemis uses Feff calculations to fit the Fourier components between Rmin and Rmax and, if the background corefinement is turned on, it uses a spline to fit the components between 0 and Rmin. Considered in that light, background corefinement is very handy indeed. If you set Rmin to be the same as Rbkg, this gives you a chance to evaluate the correlations between your background removal and the parameters you are actually interested in. As Scott says, this can give you a certain confidence in your results, or, I suppose, encourage you to go back to Athena and do the background removal differently. So, the background corefinement is the use of splines to optimize low-frequency Fourier components. In the Autobk algorithm, we use splines to optimize low frequency Fourier components. It is, therefore, reasonable to consider background corefinement as as additional round of playing with Autobk while simultaneously trying to figure out something about your chi(k) data. 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/

Juan Antonio Maciá Agulló

27 Sep 27 Sep

3:42 p.m.

New subject: k-weight in Artemis

Thank you very much Scott and Bruce for your answers. Scott, when I said "strange" spectra, I mean spectra very different (compared to k3) and with sharp valley between peaks. Best regards, JA

Scott Calvin

4:10 p.m.

New subject: k-weight in Artemis

At 11:42 AM 9/27/2006, Juan wrote:

...

Scott, when I said "strange" spectra, I mean spectra very different (compared to k3) and with sharp valley between peaks.

Juan, I teach at Sarah Lawrence College, where we like to insist that our students aren't strange, just different. :) In fact, that goes for the faculty too...our recruiting slogan is (no kidding): "You're different, so are we." So the Fourier transform of a kwt 1 spectrum may look very different from the Fourier transform of a kwt 3 spectrum. There's nothing wrong or unusual about that. Remember: EXAFS Fourier transforms are not radial distribution functions, although they do correlate to them. You can't, for instance, equate the relative height of two peaks in the Fourier transform to the relative coordination number at two distances from the absorbing atom. On the other hand, if as a function of some external variable (time, temperature, concentration...) you see one peak go up while another goes down, you might hypothesize that there are changes in the average coordination number at two distances. But that hypothesis would have to be tested by modeling (e.g. Artemis/Ifeffit) or comparison to known standards--there are other phenomena that could be causing the change in the peak heights. --Scott Calvin Sarah Lawrence College

Juan Antonio Maciá Agulló

4:39 p.m.

New subject: k-weight in Artemis

Ok Scott, thanks. JA

Juan Antonio Maciá Agulló

7:15 p.m.

New subject: k-weight in Artemis

Scott, how did you calculate bond lenth? delR + Reff? Scott Calvin wrote:

...

Ideally, the fitted parameters should have uncertainty ranges that overlap. For example, one bond length might be 2.03 +/- 0.04 angstroms, and the equivalent bond length using a different k-weight might be 2.09 +/- 0.03 angstroms.

Thank you very much JA

Kelly, Shelly D.

24 Sep 24 Sep

3:14 p.m.

New subject: Phase corrected Fourier transforms -edited!

Here is an edited version. I apologize. Shelly Hi Juan and John, Here us my 2cents worth... Unfortunately the magnitude of the FT is not as simple as a RDF. I agree with John that the ultimate goal of the analysis is to come up with a RDF. But I disagree about the worth of adding a bit more processing to the spectra that makes the spectra just a bit more like an RDF but leaves the data so far from it, such that true analysis is still needed. If the phase shift correction actually allowed the spectra to be inverted, we would all do it! I started a page the XAFS.org page under tutorials about the information that should be reported with EXAFS spectra. On that page I added a link to some of my favorite rants, including a section about why the Magnitude of the FT is not a radial distribution function. http://xafs.org/Common_Mistakes I encourage everyone on the list to read and change these pages and add your own pages. I think that this is a really great way to come up with a community voice. Here are a few thoughts about why I don't favor adding the phase shift. (1) Often 2nd and 3rd neighboring peaks don't turn-up as peaks at all in the FT because of the interference between themselves and the strong signal from the first shell. In this case the "peak" is a shoulder or sometimes even the opposite of a shoulder were the first peak is shaved off on the side. The phase correction isn't going to fix that. As John mentioned (2) multiple scattering paths are not supposed to be in a radial distribution function and they often show up as peaks in the Mag of the FT and (3) some signals are not a single peak, but a doublet in the magnitude of the FT. The phase correction doesn't "fix" multiple scattering paths or the doublets. I'm not sure that I want the multiple scattering paths fixed, they often have a lot of information in them. (4) A Fourier transform will never be able to tell the difference between neighboring atom types as a radial distribution function. (5) In addition the phase correction in Athena only corrects for half of t! he phase shift, the part from the central atom. The other half of the phase shift comes from the scattering atom type, which can only be corrected in Artemis. I do not recommend using the phase shift option in Athena because it does not turn the data into a radial distribution function. The magnitude of the FT is only an envelope function. A lot of information within the spectra is not displayed in the magnitude of the FT. It is the chi(k), and the real and imaginary parts of the FT which show the complexity of the spectroscopy. I think that this complexity should be embraced not hidden. In John's language, I think that the picture is more blurred by adding a phase correction to the spectra mainly because it promises so much but actually adds so little. It seems to me that adding the phase shift, indicates that there is something inherently "wrong" with the Magnitude of the Fourier transform of the spectra. I don't endorse that point of view. Truly, a radial distribution function is the goal of most analysis, so I encourage the display of the individual contributions/paths to the total model. I think that this display of all the paths particularly in the chi(k), real and/or imaginary parts of the FT adds a lot of transparency to the data analysis and interpretation. Well, Juan you walked right into that one. Welcome to the mailing list...:) Shelly -----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov] On Behalf Of John J. Rehr Sent: Saturday, September 23, 2006 11:01 PM To: XAFS Analysis using Ifeffit Cc: John J. Rehr; anatoly.frenkel@yu.eduA Subject: RE: [Ifeffit] Phase corrected Fourier transforms Anatoly is correct that the phase correction cannot work for all shells in inhomogeneous systems. However, it's *always* possible to remove the central atom phase shift 2 delta_c which often dominates the phase correction. Second, if one corrects by the 1st nn phase, delta_1, the dominant near neighbor shell will appear at the correct distance. The correction is not right for subsequent shells but the error is given by a phase difference delta R = <(1/2)(d/dk) [delta_n-delta_1]> This correction is typically order 0.1 Ang and usually smaller than the typical shift of about 0.3 - 0.4 Ang in uncorrected FT peaks. The upshot is that the phase corrected FT can be correct for the first shell which often dominates the FT and generally better than uncorrected FTs. Corrective lenses for vision aren't perfect either at all distances, but it's hard to argue that it's preferable not to use them for that reason. J. Rehr On Sat, 23 Sep 2006, Anatoly Frenkel wrote:

...

My 0.533 Rouble: In my experience with some Mn oxides, the Mn-O FT magnitude peak's position is 0.5 A lower its corresponding 1NN bond length, while the Mn-Mn are 0.3 A lower than their bond lenghts. Thus, it would be misleading for this and other similar compounds to apply theoretical phase correction of the 1NN to the entire data, as it will shift only one peak correctly.

Anatoly

-----Original Message----- From: ifeffit-bounces@millenia.cars.aps.anl.gov [mailto:ifeffit-bounces@millenia.cars.aps.anl.gov]On Behalf Of Matthew Marcus Sent: Saturday, September 23, 2006 4:14 PM To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] Phase corrected Fourier transforms

OK, I have to put in my US$0.02/e0.02. The phase correction can't mean much if the scattering atoms are different, which is usually the case. If the scattering atoms are very heavy, then the phase correction has a kink in it which could cause strange shapes if applied to light-atom shells. I've played with this phase correction and amplitude correction, while I was at it, and not been very impressed. One place where it might be useful is in the aforementioned case of heavy and light scatterers. If you use the correction for one of these, then the corresponding shells sharpen and the other ones blur out, so you can get a rough idea of who's who. I suspect this works better if you do the amplitude as well.

A related technique is to use model compounds+FEFF to get 'semi-empirical' amps and phases which include all the artifacts of the experiment. Suppose, for instance, that you're looking at Cu in a matrix of Fe, and you have data for Cu metal but not for any known Cu->Fe scattering pair. You can synthesize a 'semi-empirical' Cu->Fe phase and amp like this:

phi(Cu->Fe) = phi_exp(Cu->Cu)+(phi_theor(Cu->Fe)-phi_theor(Cu->Cu)) amp(Cu->Fe) = amp_exp(Cu->Cu)*amp_theor(Cu->Fe)/amp_theor(Cu->Cu)

If you don't want to do this for modeling, you can use this method for comparing two spectra which you think might be alike except for the central atom, by "correcting" one spectrum with the difference to make it comparable to the other.

Another aspect to this whole thing is that people are very used to uncorrected FT's and are aware that you have to add 0.3-0.4A to the distances. If you show only corrected FT's, I wonder if that will be satisfying to the audience. mam

----- Original Message ----- From: "John J. Rehr" To: "XAFS Analysis using Ifeffit" Cc: "John J. Rehr" Sent: Saturday, September 23, 2006 7:18 AM Subject: Re: [Ifeffit] Phase corrected Fourier transforms

Dear Juan Antonio,

I personally feel that adding phase correction to the XAFS FT is highly desirable, and I encouraged its implementation in Athena. The reasons are the following:

1) Peaks in non-phase corrected FT are substantially in error. 2) The non-linearity of the phase shifts in high-Z materials leads to multiple-peaks, thus blurring the FT. 3) Theoretical phase shifts are good enough that adding phase correction tends to correct the peak positions and the problems due to non-linearities. 4) Adding phase correction does no-harm to the fits. That is, one gets the same results whether or not phase correction is included. 5) Adding phase correction gives a FT which can be more easily interpreted "by eye", that is the peaks have a more physical interpretation. On the contrary, non-phase corrected FTs can be mis-interpreted.

Overall, my view is that the phase correction is like a prescription lens which gives a sharper image. While the image may not be perfect, at least it's generally much superior to the non-phase corrected FT.

J. Rehr

On Fri, 22 Sep 2006, Juan Antonio [iso-8859-1] Maciá Agulló wrote:

...
Hi all,

I have read Phase corrected Fourier transforms in Athena manual and now I have a big doubt, ¿phase correction or not in a publication?

I have read also that this correction is different (more complete) in Artemis and I am not sure if I should correct also in Artemis and which path should I use and why.

I saw many papers dealing with EXAFS fits and they showed a "calculated" bond distance, I think it is: d = Reff + deltaR, right?

I also ask for a paper where I can find that deltaE is ok (even for high-Z backscatterers) if deltaE < 10eV.

I have high correlations between ss and SO2, and deltaR and deltaE. I tried different fits but I can not eliminate them, then...is the fit wrong?

Sorry for these easy questions but I am a novice in XAFS.

Thank you very much.

Best regards, JA

---------------------------------------------------------------------------- ----

...
_______________________________________________ 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

6630

Age (days ago)

6635

Last active (days ago)

List overview

Download

22 comments

7 participants

participants (7)

Anatoly Frenkel
Bruce Ravel
John J. Rehr
Juan Antonio Maciá Agulló
Kelly, Shelly D.
Matthew Marcus
Scott Calvin