Self-Absorption Corrections, Wavelet Transforms
Hi Everyone, There were several conversations around the XAFS 12 conference on self-absorption corrections for EXAFS. It would be nice to add such corrections (and others, like for deadtime corrections) to Ifeffit. Corwin Booth (at LBL lab, Berkeley) presented a very nice procedure for making these corrections for EXAFS at the conference, improving the work of L. Troger, et al from the mid 1990's. Corwin just sent me the Fortran source code for his program so that it might be included in Ifeffit, and is letting me pass it on. I put the source code (with a Unix Makefile and some instructions) at http://cars.uchicago.edu/ifeffit/contrib/sabcor.tar.gz with a README at http://cars.uchicago.edu/ifeffit/contrib/README.SABCOR I'd be interested in hearing others opinions on this topic, and whether this should be included in Ifeffit. A self-absorption correction that worked better for XANES would be useful too. If you've thought about self-absorption corrections or have some data affected by self-absorption corrections, please check out Corwin's code and let us know what you think of this. Speaking of things to check out, Francois Farges recently created a web page discussingCauchy Wavelet transforms of EXAFS data at http://www.univ-mlv.fr/~farges/wav/ The site includes MATLAB code and references. These transforms are very interesting and I recommend checking out this stuff. There was also a recent email from Harald Funke sent to many people about a different code for Wavelet transforms. If I get permission from Harald, I'll post these codes too. Related to all this, I added a 'Contributions' page to the Ifeffit Web Pages, for "contributions and/or things to consider for XAFS analysis and Ifeffit". Currently, this has links to Corwin's self-absorption correction code, and Francois' Wavelet web page. If anyone has any suggestions for other things to add, let me know. Thanks, --Matt
Hi all,
Corwin Booth (at LBL lab, Berkeley) presented a very nice procedure for making these corrections for EXAFS at the conference, improving the work of L. Troger, et al from the mid 1990's.
I'd be interested in hearing others opinions on this topic, and whether this should be included in Ifeffit.
Corwin Booths approach to selfabsorption correction seems to be very nice. I think that especially its possibility to give up the "infinite sample thickness" limitation could be an important improvement over the previous approaches. Still, it makes two (more or less implicit) assumptions: * the detector surface has to be parallel to the x-ray beam (phi + theta = 90 deg) * the detector has to have a neglectable solid angle I'm not sure if these two assumptions hold for most fluorescence experiments? When I collected my last fluorescence data a couple of years ago, large solid angle detectors (like Lytle-detectors) were still in use. I have shown that Troegers approach to selfabsorption correction can be generalized for large detector surfaces (Phys. Rev. B 60, 9335 (1999)). In principle this should be also possible for Corwin Booths formula. But when integrating over large solid angles, the exact geometry of the experimental setup plays a crucial role in determining the selfabsorption correction and I doubt that a useful implementation into iFeffit would be possible. If, however, everyone is using solid state detectors now, I would say that implementing Corwin Booths code into iFeffit could be worth the effort. Best, Peter -- -------------------------------------------------------------- Peter Pfalzer Universitaet Augsburg Tel: +49-821-598-3215 Lehrstuhl fuer Experimentalphysik II Fax: +49-821-598-3411 Universitaetsstr. 1 D-86135 Augsburg Germany Peter.Pfalzer@physik.uni-augsburg.de --------------------------------------------------------------
Hi all,
Corwin Booth (at LBL lab, Berkeley) presented a very nice procedure for making these corrections for EXAFS at the conference, improving the work of L. Troger, et al from the mid 1990's.
I'd be interested in hearing others opinions on this topic, and whether this should be included in Ifeffit.
Corwin Booths approach to selfabsorption correction seems to be very nice. I think that especially its possibility to give up the "infinite sample thickness" limitation could be an important improvement over the previous approaches. Still, it makes two (more or less implicit) assumptions: * the detector surface has to be parallel to the x-ray beam (phi + theta = 90 deg) * the detector has to have a neglectable solid angle
I'm not sure if these two assumptions hold for most fluorescence experiments? surely no ! (cf ID21 at ESRF) and most future expeirments won't be that "ideal" for sure.
When I collected my last fluorescence data a couple of years ago, large solid angle detectors (like Lytle-detectors) were still in use. I have shown that Troegers approach to selfabsorption correction can be generalized for large detector surfaces (Phys. Rev. B 60, 9335 (1999)). In principle this should be also possible for Corwin Booths formula.
except for cations above than Zr.
But when integrating over large solid angles, the exact geometry of the experimental setup plays a crucial role in determining the selfabsorption correction and I doubt that a useful implementation into iFeffit would be possible.
yes but if you say so nothing will ever work. one has to start. I would be amazed to see that one day, the perfect self absorption correction code will work.
If, however, everyone is using solid state detectors now, I would say that implementing Corwin Booths code into iFeffit could be worth the effort.
the most important to me is to expand it to XANES (as in FLUO by Haskel) where self-absorption effects are relatively more important as compared to the EXAFS, and going opposite ways depending if you considering the pre-edge or the main edge regions (because of the collapsing mu0 near E0). FF -- Francois FARGES Laboratoire des Géomatériaux Université de Marne la Vallée 5 Bd Descartes-Champs S/Marne 77454 Marne la Vallée cedex 2 TEL: 01 49 32 90 57 from outside France: +33 1 49 32 90 57 FAX: 01 49 32 91 37 from outside France: +33 1 49 32 91 37
Dear François,
But when integrating over large solid angles, the exact geometry of the experimental setup plays a crucial role in determining the selfabsorption correction and I doubt that a useful implementation into iFeffit would be possible.
yes but if you say so nothing will ever work. one has to start. I would be amazed to see that one day, the perfect self absorption correction code will work.
Maybe the point I wanted to make here wasn't clear enough. When you start integrating for a large solid angle detector you have to take into account the distance from the beamspot on the sample to the detector, vertical and horizontal projection of the position of the beamspot on the detector, the shape and size of the detector, cutoff angles, ... I didn't want to say this is impossible or won't work (it does). I'm just not sure how the realization of this complex task could look like, so that the number of satisfied iFeffit users will justify the time Matt spends on writing the code. Of course you have to start somewhere go get something to work. But this is where I would NOT start. An easy to use small solid angle version with only a couple of unambiguous parameters seems a much more favorable starting point to me. Best, Peter -- -------------------------------------------------------------- Peter Pfalzer Universitaet Augsburg Tel: +49-821-598-3215 Lehrstuhl fuer Experimentalphysik II Fax: +49-821-598-3411 Universitaetsstr. 1 D-86135 Augsburg Germany Peter.Pfalzer@physik.uni-augsburg.de --------------------------------------------------------------
Peter, Francois Thanks for the thought on self-absorption corrections. I agree that Corwin's work is an excellent place to start. Incorporating self-absorption corrections that handle XANES (as from Haskel's FLUO or Sam Webb's SixPack) is also important. Peter Pfalzer wrote:
Corwin Booths approach to selfabsorption correction seems to be very nice. I think that especially its possibility to give up the "infinite sample thickness" limitation could be an important improvement over the previous approaches. Still, it makes two (more or less implicit) assumptions: * the detector surface has to be parallel to the x-ray beam (phi + theta = 90 deg) * the detector has to have a neglectable solid angle
I'm not sure if these two assumptions hold for most fluorescence experiments?
Francois Farges wrote:
surely no ! (cf ID21 at ESRF) and most future expeirments won't be that "ideal" for sure.
I would expect that Peter's assumptions do, and will, hold for most measurements. Perhaps I'm misunderstanding Francois, but I thought that ID21 (a micro-fluorescence line) uses solid-state detectors, and nominally at phi+theta=90. Is that not so? Anyway, the 'phi+theta=90' approximation is still the norm for fluorescence ion and solid state detectors, simply because reducing elastic scatter is important. It is not always the case, but for solid-state detectors and ion chambers it is definitely most common. An important counter-example (and possibly one to become more widely used) is when using crystal analyzers (either in Bragg or Laue geometries) to select a fluorescence line. For these, eliminating the elastic scattering with geometry is not so important and other considerations determine the analyzer/detector geometry. The 'small solid angle' argument seems mostly safe to me too. If I understand the papers by you, Corwin, Troger, etal, and Brewe, etal, this is not a huge effect near 'phi ~= theta ~= 45' (where ~= means +/- 15degrees'), and becomes most important near phi~=0 or phi~=90 (grazing incidence or grazing exit). Corwin wrote: '... for detector geometries where phi+theta=90, we find the maximum error in (sin(phi)/sin(beta)) is on the order of 1-2% even for delta_theta=5degrees at theta=80degrees'. I interpret that to mean that even for fairly large opening angle of the detector the effect should be small, except for the grazing incident/exit geometry. Is that your understanding too? If so, I'm willing to neglect the grazing incident/exit geometries (at least for now), and expect that people who use grazing incident or exit usually know what they're doing and how to make these correction themselves. Peter wrote:
When I collected my last fluorescence data a couple of years ago, large solid angle detectors (like Lytle-detectors) were still in use. I have shown that Troegers approach to selfabsorption correction can be generalized for large detector surfaces (Phys. Rev. B 60, 9335 (1999)). In principle this should be also possible for Corwin Booths formula.
Francois wrote:
except for cations above than Zr.
Maybe I'm misunderstanding Francois on this, but I don't see how the self-absorption correction depends on Z, except for implicit (and known) Z dependence of mu. Do Corwin's approximation break down at high Z? I would have guessed self-absorption got worse at lower energies. Am I missing something? Peter wrote:
But when integrating over large solid angles, the exact geometry of the experimental setup plays a crucial role in determining the selfabsorption correction and I doubt that a useful implementation into iFeffit would be possible.
If, however, everyone is using solid state detectors now, I would say that implementing Corwin Booths code into iFeffit could be worth the effort.
How large of a solid angle do you mean? I'd expect a few percent of 4pi to be typical for both ion chambers and solid-state detectors. Everyone is definitely *not* using solid state detectors, but between those and relatively small fluorescence ion chambers (e.g., Lytle chambers), that does seem like most fluorescence work done. Anyway, I agree (I think with both you and Francois??) that the 'large solid angle' correction can be postponed at least until something works reasonably well. I also definitely agree with Francois that correcting XANES is very important. It sure would be nice to have a complete self-absorption correction for both XANES and EXAFS.... Thanks again for the insight! --Matt
Hi Matt,
Maybe I'm misunderstanding Francois on this, but I don't see how the self-absorption correction depends on Z, except for implicit (and known) Z dependence of mu. Do Corwin's approximation break down at high Z? I would have guessed self-absorption got worse at lower energies. Am I missing something?
I don't see the Z dependence either.
The 'small solid angle' argument seems mostly safe to me too. If I understand the papers by you, Corwin, Troger, etal, and Brewe, etal, this is not a huge effect near 'phi ~= theta ~= 45' (where ~= means +/- 15degrees'), and becomes most important near phi~=0 or phi~=90 (grazing incidence or grazing exit). Corwin wrote: '... for detector geometries where phi+theta=90, we find the maximum error in (sin(phi)/sin(beta)) is on the order of 1-2% even for delta_theta=5degrees at theta=80degrees'. I interpret that to mean that even for fairly large opening angle of the detector the effect should be small, except for the grazing incident/exit geometry. Is that your understanding too?
Sorry to say, but I'm a bit less optimistic with the small solid angle approximation. Corwins 1-2% are only true for large theta, for small theta, the error in ( sin(phi) / sin(theta[+-delta_theta]) ) becomes large (but small theta also means a normal incidence/grazing exit geometry where selfabsorption is small and may be almost neglectable). Anyway, for theta = 45 and delta_theta=5 I calculate a reasonable error of >~4% in the correction factor.
How large of a solid angle do you mean? I'd expect a few percent of 4pi to be typical for both ion chambers and solid-state detectors. Everyone is definitely *not* using solid state detectors, but between those and relatively small fluorescence ion chambers (e.g., Lytle chambers), that does seem like most fluorescence work done.
I have worked with an ion chamber of R=40mm at a distance of d~60mm, giving a solid angle of ~ 12% * 4pi, and exit angles of 10 deg < theta < 80 deg. The solid state detectors I know (but I don't know many) have a roughly estimated R~5-10mm, resulting in a solid angle of < 1% * 4pi. In the latter case, no integration is necessary, of course. In my first example, without integration, the correction will be wrong by >~10% (if the sample is located in front of the center of the detector).
Anyway, I agree (I think with both you and Francois??) that the 'large solid angle' correction can be postponed at least until something works reasonably well.
You're right. The "simple" correction with Corwin's approach will give a correction factor of (at least) the right order of magnitude with any detector. And as the selfabsorption doesn't do anything but smoothly reducing the EXAFS amplitude (at least if the sample's not too thin), one should be able to fit the remaining deviation with a slight change in S_0^2 - but keeping in mind that some cases could require more thorough treatment. Peter -- -------------------------------------------------------------- Peter Pfalzer Universitaet Augsburg Tel: +49-821-598-3215 Lehrstuhl fuer Experimentalphysik II Fax: +49-821-598-3411 Universitaetsstr. 1 D-86135 Augsburg Germany Peter.Pfalzer@physik.uni-augsburg.de --------------------------------------------------------------
On Monday 28 July 2003 02:48 pm, Peter Pfalzer wrote:
You're right. The "simple" correction with Corwin's approach will give a correction factor of (at least) the right order of magnitude with any detector. And as the selfabsorption doesn't do anything but smoothly reducing the EXAFS amplitude (at least if the sample's not too thin), one should be able to fit the remaining deviation with a slight change in S_0^2 - but keeping in mind that some cases could require more thorough treatment.
My two cents worth: Peter has hit the nail on the head in this paragraph. The point that the SA correction mostly affects the amplitude is indeed an important point. (The rest of the effect will be in sigma^2, also an amplitude term.) One can continue to measure bond length even *ignoring* SA corrections. It is not hard, though, to come up with a case where SA correction needs to be done well, though. Suppose, to make up an example from th top of my head, you are looking at some effect in the XANES in a series of ternary phase single crystals (i.e. something of the sort A_{1-x} B_x C). And suppose you are measuring the edge of the atom B. As x approaches 1, the SA correction becomes larger. The SA correction must be applied consistently to make meaningful conparisons between the data. So, any of the approaches out there in the literature (including doing nothing!) are possibly "adequate" if it is sufficient to measure the phase terms in the EXAFS equations. However, to try to make sense of a series of data sets, a consistent and stable SA correction is necessary. In that case it would be prudent for the experimenter to measure a toy system in which the SA correction is predictable to understand the limitations of any correction scheme. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b, X24c, U4b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
Peter, Francois
Thanks for the thought on self-absorption corrections. I agree that Corwin's work is an excellent place to start. Incorporating self-absorption corrections that handle XANES (as from Haskel's FLUO or Sam Webb's SixPack) is also important.
Peter Pfalzer wrote:
Corwin Booths approach to selfabsorption correction seems to be very nice. I think that especially its possibility to give up the "infinite sample thickness" limitation could be an important improvement over the previous approaches. Still, it makes two (more or less implicit) assumptions: * the detector surface has to be parallel to the x-ray beam (phi + theta = 90 deg) * the detector has to have a neglectable solid angle
I'm not sure if these two assumptions hold for most fluorescence experiments?
Francois Farges wrote:
surely no ! (cf ID21 at ESRF) and most future expeirments won't be that "ideal" for sure.
I would expect that Peter's assumptions do, and will, hold for most measurements. Perhaps I'm misunderstanding Francois, but I thought that ID21 (a micro-fluorescence line) uses solid-state detectors, and nominally at phi+theta=90. Is that not so?
not always, Matt. that's even rarely because of steric effects. pin-diods are the thing to use at ESRF because that's the rule (...). so pin-diods are used most of the time, on either micro BL (ID 21 and 22). and anyway one needs a code with all the angles allowed to vary. when you use furnaces, cells and so forth you always have some weird designs that are not the "case-study" like in books. future will be full of such devices just because the ideal experiment will never exists.
Anyway, the 'phi+theta=90' approximation is still the norm for fluorescence ion and solid state detectors, simply because reducing elastic scatter is important. It is not always the case, but for solid-state detectors and ion chambers it is definitely most common.
I never get such geometry because, inside the Lytle for instance, even using the transmission setup while collecting the fluo the average solid angle is not 90 degrees but 88 degrees - based on self-absorption correction of the fluo vs. the transmission. and it makes a different when you want some accuracy on pre-edge intensities of number of neighbors. I never had to use sold state detectors because when my self-absorption arises, it's because the local concentration of the studied element is high and then such detectors saturates by a factor of so much so we need to get rid of it. So using pin diods or Lytle with some lighter gas.
An important counter-example (and possibly one to become more widely used) is when using crystal analyzers (either in Bragg or Laue geometries) to select a fluorescence line. For these, eliminating the elastic scattering with geometry is not so important and other considerations determine the analyzer/detector geometry.
The 'small solid angle' argument seems mostly safe to me too. If I understand the papers by you, Corwin, Troger, etal, and Brewe, etal, this is not a huge effect near 'phi ~= theta ~= 45' (where ~= means +/- 15degrees'), and becomes most important near phi~=0 or phi~=90
this does not work for cations above, say, Fe/Ni and when the local concentration (i.e., at the impact of the spot) is above, say 30 mol.%. I've tons of data that I ve collected to try understanding these effects (because no code really works in the detail). Even at 90°, you get significant self-abs. effects, in ZrO2 at the Zr K-edge for instance. Troger et al is highly tuned for light elements (Si, namely).
(grazing incidence or grazing exit). Corwin wrote: '... for detector geometries where phi+theta=90, we find the maximum error in (sin(phi)/sin(beta)) is on the order of 1-2% even for delta_theta=5degrees at theta=80degrees'. I interpret that to mean that even for fairly large opening angle of the detector the effect should be small, except for the grazing incident/exit geometry. Is that your understanding too?
If so, I'm willing to neglect the grazing incident/exit geometries (at least for now), and expect that people who use grazing incident or exit usually know what they're doing and how to make these correction themselves.
Peter wrote:
When I collected my last fluorescence data a couple of years ago, large solid angle detectors (like Lytle-detectors) were still in use. I have shown that Troegers approach to selfabsorption correction can be generalized for large detector surfaces (Phys. Rev. B 60, 9335 (1999)). In principle this should be also possible for Corwin Booths formula.
Francois wrote:
except for cations above than Zr.
Maybe I'm misunderstanding Francois on this, but I don't see how the self-absorption correction depends on Z, except for implicit (and known) Z dependence of mu. Do Corwin's approximation break down at high Z? I would have guessed self-absorption got worse at lower energies. Am I missing something?
I do not do any theory because I tjust don't know it. Also, as a modest scientists, I would always think that we might not fully understands everything. So to skip my ignorances, I just collected Mo's of data at various edges (Ti, Fe, Mn, Zr, Mo, Th, U etc) that I collected in fluo mode at many angles for many compounds with various concentrations and compared to transmission (and collected at different beamlines with different detectors to get a clue on "localized effects" too). and I used (and abused !!) the Troeger trick (as it used to call it when I did all these test in the early 90's for our high temperature furnace). So I can really tell you that the Troeger trick does not work anymore above Zr (at mpderate concentrations such as in ZrSiO4). at Ni it fails for higher conc. of Ni (such as in NiO, but Ni2SiO4 is fine). etc. if this is not due to self-absorption (despite it was clearly affected by the angles and FLUO did an fairly good job to correct but not perfect, esp. in the pre-edge region), I would be pleased to learn the origin ! anyway I always had a very empirical approach to it (because Goulon's paper is just indigestable to me at this was before Troeger's work). but I could never really detect a single theory that works (just too much for me !). And i think we should all be open-minded as we might not have understood everything. But we will agree, I think, to say that what matters for the user is not the theory (or the theories) but something that can correct efficiently and correctly. and any attempt is great. but the task might be huge. and I would say that experiments should drive and help theoreticians (and not vice-versa).
Peter wrote:
But when integrating over large solid angles, the exact geometry of the experimental setup plays a crucial role in determining the selfabsorption correction and I doubt that a useful implementation into iFeffit would be possible.
If, however, everyone is using solid state detectors now, I would say that implementing Corwin Booths code into iFeffit could be worth the effort.
How large of a solid angle do you mean? I'd expect a few percent of 4pi to be typical for both ion chambers and solid-state detectors. Everyone is definitely *not* using solid state detectors, but between those and relatively small fluorescence ion chambers (e.g., Lytle chambers), that does seem like most fluorescence work done.
Anyway, I agree (I think with both you and Francois??) that the 'large solid angle' correction can be postponed at least until something works reasonably well.
I also definitely agree with Francois that correcting XANES is very important. It sure would be nice to have a complete self-absorption correction for both XANES and EXAFS....
thanks Matt ! -- Francois FARGES Laboratoire des Géomatériaux Université de Marne la Vallée 5 Bd Descartes-Champs S/Marne 77454 Marne la Vallée cedex 2 TEL: 01 49 32 90 57 from outside France: +33 1 49 32 90 57 FAX: 01 49 32 91 37 from outside France: +33 1 49 32 91 37
participants (4)
-
Bruce Ravel -
Francois Farges -
Matt Newville -
Peter Pfalzer