normalization methods

newer
JFEFF/FEFF9 "Save&Run" no longer...

older
Re: [Ifeffit] normalization methods

Matt Newville

15 May 2013 15 May '13

2:35 p.m.

Hi Folks, Over on the github pages for larch, Mauro and Bruce raised an issue about the "flattening" in Athena. See https://github.com/xraypy/xraylarch/issues/44 I've added a "flattened output" from Larch's pre_edge() function, but the question has been raised of whether this is "better" than the simpler normalized spectra, especially for doing PCA and/or LCF for XANES. Currently, the "normalized" spectra is just "(mu - pre_edge_line)/edge_step". Clearly, a line fitted to the pre-edge of the spectra is not sufficient to remove all instrumental backgrounds. In some sense, flattening attempts to do a better job, fitting the post-edge spectra to a quadratic function. As Mauro, Bruce, and Carmelo have pointed out, it is less clear that it is actually better for XANES analysis. I think the main concerns are that a) it is so spectra-specific, and b) it turns on at E0 with a step function. Bruce suggested doing something more like MBACK or Ifeffit's bkg_cl(). It would certainly be possible to do some sort of "flattening" so that mu follows the expected energy dependence from tabularized mu(E). Does anyone else have suggestions, opinions, etc? Feel free to give them here or at the github page.... --Matt

Show replies by date

Matthew Marcus

15 May 15 May

2:57 p.m.

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]*muNorm[ref](E)}. A sign that this degree of freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam On 5/15/2013 7:35 AM, Matt Newville wrote:

...

Hi Folks,

Over on the github pages for larch, Mauro and Bruce raised an issue about the "flattening" in Athena. See https://github.com/xraypy/xraylarch/issues/44

I've added a "flattened output" from Larch's pre_edge() function, but the question has been raised of whether this is "better" than the simpler normalized spectra, especially for doing PCA and/or LCF for XANES.

Currently, the "normalized" spectra is just "(mu - pre_edge_line)/edge_step". Clearly, a line fitted to the pre-edge of the spectra is not sufficient to remove all instrumental backgrounds. In some sense, flattening attempts to do a better job, fitting the post-edge spectra to a quadratic function. As Mauro, Bruce, and Carmelo have pointed out, it is less clear that it is actually better for XANES analysis. I think the main concerns are that a) it is so spectra-specific, and b) it turns on at E0 with a step function.

Bruce suggested doing something more like MBACK or Ifeffit's bkg_cl(). It would certainly be possible to do some sort of "flattening" so that mu follows the expected energy dependence from tabularized mu(E).

Does anyone else have suggestions, opinions, etc? Feel free to give them here or at the github page....

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

Matt Newville

3:25 p.m.

Hi Matthew, On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus wrote:

...

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]*muNorm[ref](E)}. A sign that this degree of freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value). Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible. And, of course, over-absorption is next on the list! --Matt

Matthew Marcus

3:41 p.m.

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam On 5/15/2013 8:25 AM, Matt Newville wrote:

...

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus wrote:

...
What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]*muNorm[ref](E)}. A sign that this degree of freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

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

George Sterbinsky

4:58 p.m.

The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of the XAS background that is not accounted for by linear subtraction and multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background? Thanks, George On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus wrote:

...

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

...
Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus wrote:

...
What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]***muNorm[ref](E)}. A sign that this degree of freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt ______________________________**_________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.**gov http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Matthew Marcus

5:20 p.m.

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent. I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves. A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions? If that miracle occurs, then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities. There are people here at ALS, such as Elke Arenholz , who do this sort of spectroscopy. I suggest asking her. mam On 5/15/2013 9:58 AM, George Sterbinsky wrote:

...

The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of t

...

XAS background that is not accounted for by linear subtraction and multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus mailto:mamarcus@lbl.gov> wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus mailto:mamarcus@lbl.gov> wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]__*muNorm[ref](E)}. A sign that this degree of freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt _________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.__gov mailto:Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

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

George Sterbinsky

6:13 p.m.

Hi Matthew, On Wed, May 15, 2013 at 1:20 PM, Matthew Marcus wrote:

...

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent.

Sorry, I think my wording wasn't particularly clear here. What I should have said is: "The goal then is to subtract the *normalized* XAS measured in a positive field (p-XAS) from *normalized* XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. *However, standard normalization does not give this result*" Italics indicate new text.

...

I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves.

I would agree, I think the effect of the magnetic field on the electrons is the likely source of the differences in background.

...

A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions?

No, after division of the p-XAS by the n-XAS, both the pre and post-edge regions are smooth, but one would need a step-like function to connect them. I've attached a plot showing the result of division. If that miracle occurs,

...

then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities.

There are people here at ALS, such as Elke Arenholz , who do this sort of spectroscopy. I suggest asking her. mam

Thanks for the suggestion and your reply. George

...

On 5/15/2013 9:58 AM, George Sterbinsky wrote:

...
The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of the

XAS background that is not accounted for by linear subtraction and

...
multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus > wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus > wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]**__*muNorm[ref](E)}. A sign that this degree of

freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt ______________________________**___________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.**__gov > http://millenia.cars.aps.anl._**_gov/mailman/listinfo/ifeffit < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 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 http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

George Sterbinsky

6:52 p.m.

Hi Matthew, On Wed, May 15, 2013 at 1:20 PM, Matthew Marcus wrote:

...

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent.

...

I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves.

I would agree, I think the effect of the magnetic field on the electrons is the likely source of the differences in background.

...

A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions?

No, after division of the p-XAS by the n-XAS (before any normalization), both the pre and post-edge regions are smooth, but one would need a step-like function to connect them. I've attached a plot showing the result of division. If that miracle occurs,

...

then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities.

There are people here at ALS, such as Elke Arenholz , who do this sort of spectroscopy. I suggest asking her. mam

Thanks for the suggestion and your reply. George

...

...
On 5/15/2013 9:58 AM, George Sterbinsky wrote:

...
The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of the

XAS background that is not accounted for by linear subtraction and

...
multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus > wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus < mamarcus@lbl.gov mailto:mamarcus@lbl.gov> wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]**__*muNorm[ref](E)}. A sign that this degree of

freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt ______________________________**___________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.**__gov > http://millenia.cars.aps.anl._**_gov/mailman/listinfo/ifeffit < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 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 http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Matthew Marcus

7:04 p.m.

OK, I guess I don't know what 'standard normalization' is. It looks from the quotient that you'll need some sort of curved post-edge. I guess the division didn't work because the electron energy distribution is different pre- and post-edge, so the magnetic effects are different and vary across the edge. Thus, the shapes of the MCD peaks will be at least a little corrupted even if the pre- and post-edge spectra are taken into account. I don't know what to do about this. Did you try asking Elke? mam On 5/15/2013 11:52 AM, George Sterbinsky wrote:

...

Hi Matthew,

On Wed, May 15, 2013 at 1:20 PM, Matthew Marcus mailto:mamarcus@lbl.gov> wrote:

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent.

Sorry, I think my wording wasn't particularly clear here. What I should have said is:

"The goal then is to subtract the /normalized/ XAS measured in a positive field (p-XAS) from /normalized/ XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. /However, standard normalization does not give this result/"

Italics indicate new text.

I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves.

I would agree, I think the effect of the magnetic field on the electrons is the likely source of the differences in background.

A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions?

No, after division of the p-XAS by the n-XAS (before any normalization), both the pre and post-edge regions are smooth, but one would need a step-like function to connect them. I've attached a plot showing the result of division.

If that miracle occurs, then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities.

There are people here at ALS, such as Elke Arenholz mailto:earenholz@lbl.gov>, who do this sort of spectroscopy. I suggest asking her. mam

Thanks for the suggestion and your reply.

George

On 5/15/2013 9:58 AM, George Sterbinsky wrote:

The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of the

XAS background that is not accounted for by linear subtraction and multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov>> wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov>> wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]____*muNorm[ref](E)}. A sign that this degree of

freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt ___________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.____gov mailto:Ifeffit@millenia.cars.aps.anl.gov> http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>

___________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.____gov mailto:Ifeffit@millenia.cars.aps.anl.gov> http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>

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

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

George Sterbinsky

8:43 p.m.

By standard normalization, I meant subtraction of a linear pre-edge and multiplication by a constant. If this treatment is applied to the XAS spectra before subtraction, one does not obtain an XMCD spectrum that goes to zero in the post edge region for the data I described. As you noted, that is what would be expected given the p-XAS and n-XAS have different slopes in the post-edge region. On the other hand, standard normalization + flattening does result in pre and post-edge regions that go to zero, again as one might expect. So perhaps, the background modeled by standard normalization + flattening is an accurate representation of the real background in some cases and can be used in quantitative analysis. Is there reason to believe that cannot be the case? Thanks, George On Wed, May 15, 2013 at 3:04 PM, Matthew Marcus wrote:

...

OK, I guess I don't know what 'standard normalization' is. It looks from the quotient that you'll need some sort of curved post-edge. I guess the division didn't work because the electron energy distribution is different pre- and post-edge, so the magnetic effects are different and vary across the edge. Thus, the shapes of the MCD peaks will be at least a little corrupted even if the pre- and post-edge spectra are taken into account. I don't know what to do about this. Did you try asking Elke? mam

On 5/15/2013 11:52 AM, George Sterbinsky wrote:

...
Hi Matthew,

On Wed, May 15, 2013 at 1:20 PM, Matthew Marcus > wrote:

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent.

Sorry, I think my wording wasn't particularly clear here. What I should have said is:

"The goal then is to subtract the /normalized/ XAS measured in a positive field (p-XAS) from /normalized/ XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. /However, standard normalization does not give this result/"

Italics indicate new text.

I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves.

I would agree, I think the effect of the magnetic field on the electrons is the likely source of the differences in background.

A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions?

No, after division of the p-XAS by the n-XAS (before any normalization), both the pre and post-edge regions are smooth, but one would need a step-like function to connect them. I've attached a plot showing the result of division.

If that miracle occurs, then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities.

There are people here at ALS, such as Elke Arenholz < earenholz@lbl.gov mailto:earenholz@lbl.gov>, who do this sort of spectroscopy. I suggest asking her.

mam

Thanks for the suggestion and your reply.

George

On 5/15/2013 9:58 AM, George Sterbinsky wrote:

The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is then some component of the

XAS background that is not accounted for by linear subtraction and multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus < mamarcus@lbl.gov mailto:mamarcus@lbl.gov >> wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus < mamarcus@lbl.gov mailto:mamarcus@lbl.gov >> wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]**____*muNorm[ref](E)}. A sign that this degree of

freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt ______________________________** _____________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.**____gov >> http://millenia.cars.aps.anl._**___gov/mailman/listinfo/**ifeffit <http://millenia.cars.aps.anl.**__gov/mailman/listinfo/ifeffit < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 <http://millenia.cars.aps.anl.**__gov/mailman/listinfo/ifeffit < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 <http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 < http://millenia.cars.aps.anl.**gov/mailman/listinfo/ifeffit 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 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 http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

Matthew Marcus

10:41 p.m.

I'm not sure what 'flattening' means. Does that mean dividing by a linear or other polynomial function, fitted to the post-edge? mam On 5/15/2013 1:43 PM, George Sterbinsky wrote:

...

By standard normalization, I meant subtraction of a linear pre-edge and multiplication by a constant. If this treatment is applied to the XAS spectra before subtraction, one does not obtain an XMCD spectrum that goes to zero in the post edge region for the data I described. As you noted, that is what would be expected given the p-XAS and n-XAS have different slopes in the post-edge region.

On the other hand, standard normalization + flattening does result in pre and post-edge regions that go to zero, again as one might expect. So perhaps, the background modeled by standard normalization + flattening is an accurate representation of the real background in some cases and can be used in quantitative analysis. Is there reason to believe that cannot be the case?

Thanks, George

On Wed, May 15, 2013 at 3:04 PM, Matthew Marcus mailto:mamarcus@lbl.gov> wrote:

OK, I guess I don't know what 'standard normalization' is. It looks from the quotient that you'll need some sort of curved post-edge. I guess the division didn't work because the electron energy distribution is different pre- and post-edge, so the magnetic effects are different and vary across the edge. Thus, the shapes of the MCD peaks will be at least a little corrupted even if the pre- and post-edge spectra are taken into account. I don't know what to do about this. Did you try asking Elke? mam

On 5/15/2013 11:52 AM, George Sterbinsky wrote:

Hi Matthew,

On Wed, May 15, 2013 at 1:20 PM, Matthew Marcus mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov>> wrote:

You say that the flipping difference (p - n) is 0 in pre-edge and far post-edge regions, which is as it should be, but then say that the slopes of p- and n- post-edges, considered separately, are different. I must be misunderstanding because those two statements would seem to be inconsistent.

Sorry, I think my wording wasn't particularly clear here. What I should have said is:

"The goal then is to subtract the /normalized/ XAS measured in a positive field (p-XAS) from /normalized/ XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. /However, standard normalization does not give this result/"

Italics indicate new text.

I wonder if the sensitivity of the TEY changes with magnetic field because of the effect of the field on the trajectories of the outgoing electrons, which would explain the differing curves.

I would agree, I think the effect of the magnetic field on the electrons is the likely source of the differences in background.

A possibility - if you divide the p-XAS by n-XAS, do you get something which is a smooth curve everywhere but where MCD is expected? Does that curve match in pre- and far post-edge regions?

No, after division of the p-XAS by the n-XAS (before any normalization), both the pre and post-edge regions are smooth, but one would need a step-like function to connect them. I've attached a plot showing the result of division.

If that miracle occurs, then perhaps you could fit that to a polynomial, except in the MCD region, then divide the p-XAS by that polynomial, to remove the effect of the differing sensitivities.

There are people here at ALS, such as Elke Arenholz mailto:earenholz@lbl.gov mailto:earenholz@lbl.gov>>, who do this sort of spectroscopy. I suggest asking her.

mam

Thanks for the suggestion and your reply.

George

On 5/15/2013 9:58 AM, George Sterbinsky wrote:

The question of whether it is appropriate to use flattened data for quantitative analysis is something I've been thinking about a lot recently. In my specific case, I am analyzing XMCD data at the Co L-edge. To obtain the XMCD, I measure XAS with total electron yield detection using a ~70% left or right circularly polarized beam and flip the magnetic field on the sample at every data point. The goal then, is to subtract the XAS measured in a positive field (p-XAS) from XAS measured in a negative field (n-XAS) and get something (the XMCD) that is zero in the pre-edge and post-edge regions. I often find that after removal of a linear pre-edge, the spectra still have a linearly increasing post edge (with EXAFS oscillations superimposed on it), and the slope of the n-XAS and p-XAS post-edge lines are different. In this case simply multiplying the n-XAS and p-XAS by constants will never give an XMCD spectrum that is zero in the post edge region. There is the

n some

...

component of the

XAS background that is not accounted for by linear subtraction and multiplication by a constant. It seems to me that flattening could be a good way to account for such a background. So is flattening a reasonable thing to do in a case such as this, or is there a better way to account for such a background?

Thanks, George

On Wed, May 15, 2013 at 11:41 AM, Matthew Marcus mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov> mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov>>> wrote:

The way I commonly do pre-edge is to fit with some form plus a power-law singularity representing the initial rise of the edge, then subtract out that "some form". Now, that form can be either linear, linear+E^(-2.7) (for transmission), or linear+ another power-law singularity centered at the center passband energy of the fluorescence detector. That latter is for fluorescence data which is affected by the tail of the elastic/Compton peak from the incident energy. Whichever form is taken gets subtraccted from the whole data range, resulting in data which is pre-edge-subtracted but not yet post-edge normalized. The path then splits; for EXAFS, the usual conversion to k-space, spline fitting in the post-edge, subtraction and division is done, all interactively. Tensioned spline is also available due to request of a prominent user. For XANES, the post-edge is fit as previously described. Thus, there's no distinction made between data above and below E0 in XANES, whereas there is such a distinction in EXAFS. mam

On 5/15/2013 8:25 AM, Matt Newville wrote:

Hi Matthew,

On Wed, May 15, 2013 at 9:57 AM, Matthew Marcus mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov> mailto:mamarcus@lbl.gov mailto:mamarcus@lbl.gov>>> wrote:

What I typically do for XANES is divide mu-mu_pre_edge_line by a linear function which goes through the post-edge oscillations. This division goes over the whole data range, including pre-edge. If the data has obvious curvature in the post-edge, I'll use a higher-order polynomial. For transmission data, what sometimes linearizes the background is to change the abscissa to 1/E^2.7 (the rule-of-thumb absorption shape) and change it back afterward. All this is, of course, highly subjective and one of the reasons for taking extended XANES data (300eV, for instance). For short-range XANES, there isn't enough info to do more than divide by a constant. Once this is done, my LCF programs allow a slope adjustment as a free parameter, thus muNorm(E) = (1+a*(E-E0))*Sum_on_ref{x[ref]______*muNorm[ref](E)}. A sign that this degree of

freedom may be being abused is if the sum of the x[ref] is far from 1 or if a*(Emax-E0) is large. Don't get me started on overabsorption :-) mam

Thanks -- I should have said that pre_edge() can now do a victoreen-ish fit, regressing a line to mu*E^nvict (nvict can be any real value).

Still, it seems that the current flattening is somewhere between "better" and "worse", which is unsettling... Applying the "flattening" polynomial to the pre-edge range definitely seems to give poor results, but maybe some energy-dependent compromise is possible.

And, of course, over-absorption is next on the list!

--Matt _____________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.______gov mailto:Ifeffit@millenia.cars.____aps.anl.gov http://aps.anl.gov mailto:Ifeffit@millenia.cars.aps.anl.gov>> http://millenia.cars.aps.anl.______gov/mailman/listinfo/__ifeffit <http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>>

_____________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.______gov mailto:Ifeffit@millenia.cars.____aps.anl.gov http://aps.anl.gov mailto:Ifeffit@millenia.cars.aps.anl.gov>> http://millenia.cars.aps.anl.______gov/mailman/listinfo/__ifeffit <http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>>

___________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.____gov mailto:Ifeffit@millenia.cars.aps.anl.gov> http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>

___________________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.____gov mailto:Ifeffit@millenia.cars.aps.anl.gov> http://millenia.cars.aps.anl.____gov/mailman/listinfo/ifeffit <http://millenia.cars.aps.anl.__gov/mailman/listinfo/ifeffit http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit>

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

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

Matt Newville

16 May 16 May

11:47 a.m.

Hi Matthew, George, Zach, Thanks for the discussion! On Wed, May 15, 2013 at 5:41 PM, Matthew Marcus wrote:

...

I'm not sure what 'flattening' means. Does that mean dividing by a linear or other polynomial function, fitted to the post-edge? mam

...

From my point of view, the question is: what's the best way to do

Sorry, I should have been clearer. "Standard Athena/Ifeffit" is to a) regress a pre-edge line to mu(E) (no power laws) b) regress a post-edge quadratic c) set edge_step = post_edge_quadratic(E0) - pre_edge_line(E0) b) set norm(E) = (mu(E) - pre_edge_line(E)) / edge_step. Flattening (Athena only, now backported to larch) fits a quadratic to the post-edge range (typcically E0+100 to end of data) of norm(E), and then sets flattened(E) = norm(E) for E<= E0 = norm(E) - quadratic(E) + quadratic(E0) for E > E0 I think this was originally meant for display purposes only. Hopefully Bruce can correct me if I'm wrong on any of the details here. I think it's fair to say that the "Standard Athena/Ifeffit" approach to normalization is simple-minded. It was designed for EXAFS in an era when accessing databases seemed like a challenge, so even for EXAFS it is simple-minded. Flattening might be better at removing instrumental backgrounds, and be better for linear analysis of XANES. The main concern I would have is the potential for a slight discontinuity at E0, or the potential strong dependence from the choice of E0. Using something like bkg_cl() (which matched mu(E) to the data from the Cromer-Liberman tables) or MBACK (which I believe is similar, but also accounts for "elastic/Compton leakage" into the pre-edge part of fluorescence spectra). this? The pre_edge() function in Larch does include an energy exponent term, and now writes out the "flattened" array, as above. It does not include the scaling MAM described, but that would not be hard. Reimplementing bkg_cl() would not be too hard, but perhaps trying to port MBACK would be better. Perhaps all of the above is best? --Matt

Matthew Marcus

4:37 p.m.

That 'flattening' function seems over-complicated to me and makes an artificial discontinuity, at least in slope, at E0 which is a somewhat arbitrary quantity. Why not simply divide by the post-edge quadratic (norm(E) = (mu(E)-pre_edge_line(E))/quadratic(E))? In some cases, where there's a big curvature, it may make sense to divide mu(E) by the quadratic, then subtract a pre-edge. What I've never solved satisfactorily is the case in which the extrapolation of the pre-edge line crosses the post-edge, so mu(E)-pre_edge_line(E)<0 for some part of the range. I've never understood why this happens. mam On 5/16/2013 4:47 AM, Matt Newville wrote:

...

Hi Matthew, George, Zach,

Thanks for the discussion!

On Wed, May 15, 2013 at 5:41 PM, Matthew Marcus wrote:

...
I'm not sure what 'flattening' means. Does that mean dividing by a linear or other polynomial function, fitted to the post-edge? mam

Sorry, I should have been clearer. "Standard Athena/Ifeffit" is to

a) regress a pre-edge line to mu(E) (no power laws) b) regress a post-edge quadratic c) set edge_step = post_edge_quadratic(E0) - pre_edge_line(E0) b) set norm(E) = (mu(E) - pre_edge_line(E)) / edge_step.

Flattening (Athena only, now backported to larch) fits a quadratic to the post-edge range (typcically E0+100 to end of data) of norm(E), and then sets

flattened(E) = norm(E) for E<= E0 = norm(E) - quadratic(E) + quadratic(E0) for E > E0

I think this was originally meant for display purposes only.

Hopefully Bruce can correct me if I'm wrong on any of the details here.

I think it's fair to say that the "Standard Athena/Ifeffit" approach to normalization is simple-minded. It was designed for EXAFS in an era when accessing databases seemed like a challenge, so even for EXAFS it is simple-minded.

Flattening might be better at removing instrumental backgrounds, and be better for linear analysis of XANES. The main concern I would have is the potential for a slight discontinuity at E0, or the potential strong dependence from the choice of E0.

Using something like bkg_cl() (which matched mu(E) to the data from the Cromer-Liberman tables) or MBACK (which I believe is similar, but also accounts for "elastic/Compton leakage" into the pre-edge part of fluorescence spectra).

...
From my point of view, the question is: what's the best way to do this? The pre_edge() function in Larch does include an energy exponent term, and now writes out the "flattened" array, as above. It does not include the scaling MAM described, but that would not be hard. Reimplementing bkg_cl() would not be too hard, but perhaps trying to port MBACK would be better. Perhaps all of the above is best?

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

4207

Age (days ago)

4208

Last active (days ago)

List overview

Download

12 comments

3 participants

participants (3)

George Sterbinsky
Matt Newville
Matthew Marcus

normalization methods

tags

participants (3)