Re[2]: a question about background refinement
Normally, you'd want to keep Rbkg/Rmin below the 'first shell peak', which is sometimes difficult to define for the asymmetric peak common with short metal-oxygen distances, but if the peak is around 1.3Ang, the Rbkg/Rmin of 1.5Ang seems like it might be too high. Can you get decent results with a lower Rbkg/Rmin? No, I can not. 1.5A is around the minimum value in Chi(R) magnitude where the first peak can be separated from the others. However, the peak is a result of superposition of background, oxygen path and also
To Matt: the alloy structure. As for the correlation between Ru-O variables and the background parameters, they are somewhat correlated I think If I do not play with the rmin value and keep it equal to 1.0A these correlations are: delr_ru_o and bkg01_05 --> 0.5440 delr_ru_o and bkg01_04 --> -0.3878 dwf_pt_pt and bkg02_11 --> 0.3551 x_o and bkg01_01 --> -0.2856 delr_ru_o and bkg01_03 --> 0.2760 delr_ru_ru and bkg01_05 --> 0.2709 dwf_ru_o and bkg01_01 --> -0.2614 x_o and bkg01_02 --> 0.2600 dwf_ru_o and bkg01_02 --> 0.2595 delr_ru_pt and bkg02_11 --> 0.2562 x_o means the amplitude of Ru-O path e0 for Ru-O path is constrained to be the same as for all paths with Ru central atom To Jeremy:
You mention that the amplitude remains consistent at about 0.3. However, what is happening to R and E0? Are these values also reasonable? I have found that, on occasion, a negative DWF is telling me that oxygen may not be the correct nearest neighbor. In these cases, R and E0 are always anomalous for the metal-oxygen path.
I have two e0 parameters in the model that correspond to ru and pt data sets (they are e0_ru and e0_pt) the same e0_ru I use for Ru-O path Typical values that they take are about: e0_pt = 7.2318013 +/- 1.0171952 (0.0000) e0_ru = -9.1526876 +/- 0.7730603 (0.0000) and this is consistent through all data that I have and also their values almost do not depend on whether I use rmin=1.0A or 1.5A the same I can say about Ru-O distance. Typical value is: delr_ru_o = -0.0899771 +/- 0.0406058 (0.0000) To Bruce:
I have a question for you. Are you trying to determine the oxide content? That is, do you want to know what fraction of the Ru atoms are in an oxide environment? Or are you trying to account for the oxide so that you better measure the metallic component?
My advisor (Carlo Segre) has already answered this question I just want to add that the observed amplitude of Ru-O path (0.3)is approximately in agreement with the number of atoms on the nanoparticle surface. This may indicate that all Ru atoms are on the surface and all they have oxygen bonds. Stanislav
On Wednesday 14 April 2004 04:27 am, Stanislav Stoupin wrote:
As for the correlation between Ru-O variables and the background parameters, they are somewhat correlated I think
Stanislav, I think you will have to live with correlations between the oxide and the background. That is the flip side to my suggestion that you just treat the oxide as the background. (Which you and Carlo clearly explained is an undesirable thing to do.) The oxide has low frequency Fourier components and its spectrum leaks into the very low-R rang. The background spline has very low frequency components and its spectrum leak into the higher-R range. Consequently, it is hard to them apart. Oxides are "fun" that way. ;-( Would it help to impose reasonable but unvalidated prior constraints to the fit? You largest oxide/background correlations involve the oxide delta_R value. Perhaps it would be prudent to fix the R of the oxide to what you expect from bulk ruthenium oxide. That may be slightly wrong, but it might help shore up the fit. B BTW, Does the new release of Artemis seem to address the many problems you reported with your multiple data set, multiple feff calculation fits? -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b 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/
Hi Stanislav,
Normally, you'd want to keep Rbkg/Rmin below the 'first shell peak', which is sometimes difficult to define for the asymmetric peak common with short metal-oxygen distances, but if the peak is around 1.3Ang, the Rbkg/Rmin of 1.5Ang seems like it might be too high. Can you get decent results with a lower Rbkg/Rmin? No, I can not. 1.5A is around the minimum value in Chi(R) magnitude where the first peak can be separated from the others. However, the peak is a result of superposition of background, oxygen path and also the alloy structure.
I think I don't understand this. What are the "others" that the first peak can't be separated from??
I have two e0 parameters in the model that correspond to ru and pt data sets (they are e0_ru and e0_pt) the same e0_ru I use for Ru-O path Typical values that they take are about: e0_pt = 7.2318013 +/- 1.0171952 (0.0000) e0_ru = -9.1526876 +/- 0.7730603 (0.0000)
and this is consistent through all data that I have and also their values almost do not depend on whether I use rmin=1.0A or 1.5A
Two e0's that differ by 16eV for metal-metal bonds seems strange. I'd guess that Pt and Ru would be easily distinguished in the fit, but also that E0 would not be dramatically different for these. Are you sure there's not something weird going on with these being out-of-phase or something?? I could believe it's possible, but if you look at the different fit components look in k- and r-space does it all seem to add up as you would expect? Sorry to be so slow in responding.... --Matt
On Wednesday 14 April 2004 10:32 pm, Matt Newville wrote:
I have two e0 parameters in the model that correspond to ru and pt data sets (they are e0_ru and e0_pt) the same e0_ru I use for Ru-O path Typical values that they take are about: e0_pt = 7.2318013 +/- 1.0171952 (0.0000) e0_ru = -9.1526876 +/- 0.7730603 (0.0000)
and this is consistent through all data that I have and also their values almost do not depend on whether I use rmin=1.0A or 1.5A
Two e0's that differ by 16eV for metal-metal bonds seems strange. I'd guess that Pt and Ru would be easily distinguished in the fit, but also that E0 would not be dramatically different for these. Are you sure there's not something weird going on with these being out-of-phase or something?? I could believe it's possible, but if you look at the different fit components look in k- and r-space does it all seem to add up as you would expect?
I think Stanislav is doing a two-data set fit and these are the e0 values for the two different absorbers. That wouldn't, then, be so unreasonable. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 405 Naval Research Laboratory phone: (1) 202 767 2268 Washington DC 20375, USA fax: (1) 202 767 4642 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b 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/
yes, these are data from two sets of data taken at Ru and Pt edges with no recalibration in between because we needced to switch back and forth during the experiment. Carlo On Thu, 15 Apr 2004, Bruce Ravel wrote:
On Wednesday 14 April 2004 10:32 pm, Matt Newville wrote:
I have two e0 parameters in the model that correspond to ru and pt data sets (they are e0_ru and e0_pt) the same e0_ru I use for Ru-O path Typical values that they take are about: e0_pt = 7.2318013 +/- 1.0171952 (0.0000) e0_ru = -9.1526876 +/- 0.7730603 (0.0000)
and this is consistent through all data that I have and also their values almost do not depend on whether I use rmin=1.0A or 1.5A
Two e0's that differ by 16eV for metal-metal bonds seems strange. I'd guess that Pt and Ru would be easily distinguished in the fit, but also that E0 would not be dramatically different for these. Are you sure there's not something weird going on with these being out-of-phase or something?? I could believe it's possible, but if you look at the different fit components look in k- and r-space does it all seem to add up as you would expect?
I think Stanislav is doing a two-data set fit and these are the e0 values for the two different absorbers. That wouldn't, then, be so unreasonable.
B
-- Carlo U. Segre -- Professor of Physics Associate Dean for Special Projects, Graduate College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 segre@agni.phys.iit.edu http://www.iit.edu/~segre
Hi Matt,
Can you get decent results with a lower Rbkg/Rmin? No, I can not. 1.5A is around the minimum value in Chi(R) magnitude where the first peak can be separated from the others. However, the peak is a result of superposition of background, oxygen path and also the alloy structure.
MN> I think I don't understand this. What are the "others" that the MN> first peak can't be separated from?? Sorry for the confusion that I have introduced. I am doing two-data set simultaneous fit (Bruce already mentioned that in the list) where the two absorbers are Ru and Pt atoms The model structure looks like this: 1. Ru data FEFF0(Ru-Ru):feff0001 FEFF1(Ru-Pt):feff0001 FEFF2(Ru-O):feff0001 2. Pt data FEFF3(Pt-Pt):feff0001 FEFF4(Pt-Ru):feff0001 I was talking about the first part of the model that is related to Ru data. What I meant was that around 1.0-1.5A I saw a peak. According to the model that peak was a result of overlap of background and contributions from different scatterers, such that I could not say that the peak was primarily because of oxygen or any other scatterer. I think the above was what you described as:
'first shell peak', which is sometimes difficult to define for the asymmetric peak common with short metal-oxygen distances
Now if I set rmin=1.5A, the peak is not included in the R-range(it is below rmin). At lower values of rmin a part of the peak is in the range. That could be the reason why I was unable to get decent results with lower rmin. Stanislav
Hi Stanislav, Sorry I missed the point that the different E0's were for two different data sets!
a lower Rbkg/Rmin? No, I can not. 1.5A is around the minimum value in Chi(R) magnitude where the first peak can be separated from the others. However, the peak is a result of superposition of background, oxygen path and also
Can you get decent results with the alloy structure.
MN> I think I don't understand this. What are the "others" that the MN> first peak can't be separated from??
Sorry for the confusion that I have introduced. I am doing two-data set simultaneous fit (Bruce already mentioned that in the list) where the two absorbers are Ru and Pt atoms The model structure looks like this: 1. Ru data FEFF0(Ru-Ru):feff0001 FEFF1(Ru-Pt):feff0001 FEFF2(Ru-O):feff0001 2. Pt data FEFF3(Pt-Pt):feff0001 FEFF4(Pt-Ru):feff0001
I was talking about the first part of the model that is related to Ru data. What I meant was that around 1.0-1.5A I saw a peak. According to the model that peak was a result of overlap of background and contributions from different scatterers, such that I could not say that the peak was primarily because of oxygen or any other scatterer.
I think the above was what you described as:
'first shell peak', which is sometimes difficult to define for the asymmetric peak common with short metal-oxygen distances
Now if I set rmin=1.5A, the peak is not included in the R-range(it is below rmin). At lower values of rmin a part of the peak is in the range. That could be the reason why I was unable to get decent results with lower rmin.
Hmm, it does seems a little troubling, but it might be OK. I guess that's why you asked in the first place. Sorry I can't give a more definitive answer without looking at the data and/or fit. Does the Ru-Ru/Ru-Pt contribution overlap with Ru-O in this 'low-R peak'? Related to that: Is there a good understanding of why the Ru would partially oxidize while the Pt appears to not oxidize? This seems perfectly reasonable to me, but I'll ask anyway: you see Ru-O but not Pt-O right? If you see that the Pt edge background is OK and there's no need to add oxygen to the Pt model, that does make the need to include a 'background+oxygen peak' at the Ru edge more believable. But I'd still suggest looking at the different fit contributions in both k- and R-space. Hope that helps, even though you've probably tried all these things and more.... --Matt
I'll let Stanislav answer some of the other questions but I'll try these. I am sure that Stanislav will correct me if necessary. The particles are in a reducing environment and therefore we did not expect to see any Ru-O bonds at all. In the case of Pt, a model which only includes Pt-Pt and Pt-Ru neighbors works extremely well. The same model is used to fit the metal neighbors of the Ru and for all data that we have thus far, there is consistency in the combined fits. When we take the same particles and measure them at room temperature outside the fuel cell, it is clear that there are Pt-O (or other light element) bonds as well as Ru-O while the metal bonds are the same as in the rest of the data. Therefore, I am quite confident that under operating conditions, there is no Pt-O contribution. The hypothesis is that the Ru in the particles promotes the oxidation of the methanol fuel. It is therefore reasonable to see Ru-O bonds at the surface of the particle. What appears as partial oxidation is simply the ratio of surface Ru to total Ru in the 3.6nm particle. The picture is appealing and we are fairly confident because the near edge data also seems to be a linear combination of metallic Ru and partially oxidized Ru (Ru oxide hydrate fits best) but we have been vexed by the negative Debye Waller factor of the Ru-O bond. I suppose that it is possible that it is not Ru-O at all but Ru-C but certainly there is a light atom there. What Stanislav was trying to do, and we are not sure that it is a valid procedure, is to separate the real Ru-O contribution from what might be just background leakage. The data is taken in transmission and since we must take the electrode the way it is made for real fuel cells, the edge jump is relatively small. This limits the range in k-space that we can use and I think it leads to some background subtraction issues. Probably more than anyone wanted to hear about this... Carlo On Thu, 15 Apr 2004, Matt Newville wrote:
Related to that: Is there a good understanding of why the Ru would partially oxidize while the Pt appears to not oxidize? This seems perfectly reasonable to me, but I'll ask anyway: you see Ru-O but not Pt-O right?
If you see that the Pt edge background is OK and there's no need to add oxygen to the Pt model, that does make the need to include a 'background+oxygen peak' at the Ru edge more believable. But I'd still suggest looking at the different fit contributions in both k- and R-space.
Hope that helps, even though you've probably tried all these things and more....
--Matt
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
-- Carlo U. Segre -- Professor of Physics Associate Dean for Special Projects, Graduate College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 segre@agni.phys.iit.edu http://www.iit.edu/~segre
participants (4)
-
Bruce Ravel -
Carlo U. Segre -
Matt Newville -
Stanislav Stoupin