Hi Stefano,
1) I understand that energy calibration should be performed, and I did so by using the atomic edge energy. I also understand that this parameter could be fluctuating in the subsequent fitting. Any comment on this procedure?
I agree that all the energy shift stuff can be confusing. The important end result for Extended X-ray Absorption Fine Structure Analysis is to get the energy-grid of the data to match the energy-grid of the theory with the zero energy of the photo-electron somewhere on the absorption edge. In the process of doing this you may have more than one data set from the same sample with different energy-grids... Furthermore you may have more than one sample which may have different energy-grid from the other samples and the theory. Then to make things all the more complicated X-rays can change the oxidation state of metals in organic materials. So that the energy grid is supposed to be different and the data is useless. After all that, here are some answers. 1) Always measure a foil reference at the same time that you measure any EXAFS data. It is easy to do. Here is a reference for how to do it when your sample is too thick and you can not simply put the foil behind the It ion chamber. J L Cross and A I Frenkel. "Use of scattered radiation for absolute energy calibration." Rev. Sci. Instrum. 70: pp 38-40, 1998 2) Align each data scan: 2a) Without an energy reference. Align the data to one of the data sets (remember which one you used) so that the edges are all right on top of each other. Each sample has it's own energy-grid. 2b) By using the energy reference of the foil measured simultaneously with each scan. Align all the reference spectra to one of the reference spectra. Then shift each of the data sets that belong to the reference spectra by the same amount. Now each scan is independently aligned and you have an absolute energy reference for each scan. This is useful for XANES comparisons or if you suspect changes in the data while the measurement was taking place. The difference in energy from the reference to the data or from one data set to another is a real change and can be used to "say" stuff about the sample. 3) Check the alignment: Plot all the normalized xmu(E) data from the same sample in Athena, zoom in on the edge region -20 to 20 eV and take a good look. Hopefully they all line up. If there is a systematic change in the edge features? If so how much...less than 10 % (which is about the accuracy of the measurement). This means that you can most likely still use the data but that you have changes happening to your sample while the measurement was taking place and you should fix that next time. 4) Now the problem of alignment is somewhat smaller. All the data sets have the same energy grid if you used 2b...or all the data sets have there own energy grid if you used 2a. Now we need to align the data and the theory. To do this there are two steps. Choosing a zero energy of the photo-electron to define the k-grid in converting the data to chi(k), and then aligning the chi(k) data with the theory chi(k). 5) Choosing a Ezero in Athena: Pick any point that you like on the edge for Ezero in Athena. Plot the xmu data with the background and pre-edge and post-edge range. A circle on the plot shows the position of Ezero. Then right click The Ezero to set this energy to be the same value for all the data sets from this sample. Then plot the chi(k) data and see that the curves all look the same at low k-values. If not adjust Ezero a bit for the data sets that are different to get the chi(k) data to look the same. Then average the chi(k) data and write it out so that you can align it with the theory in Artemis. 6) Aligning data to theory: Read the data into Artemis. Read in what you think should work for a first shell feffxxxxx.dat file. Fit the data to the first shell. Plot the data and the theory in chi(k), take a look at the deltaE value from the fit. This value for deltaE is telling you how much the theory had to move to match-up with your choice of Ezero that you picked in step 5. If you think that the theory is correct and deltaE is larger than a few eV then you need to adjust your chi(k) data. But first write out what the theory should look like if you choose Ezero just right. 6b) Making a standard: In Artemis, set all the parameters to their best-fit values, except for deltaE. Set deltaE to zero and run the fit again. This is not really a fit. There are no parameters to optimize. Now the theory is showing you what the data should look like if you choose the perfect value for Ezero and the perfect background function in step 5. 7) Redo the background and energy choice: In Athena open up the chi(k).fit that was produced by step 6b. This is the "standard" that you can use to help get Ezero and the background correct. Then select individual scans taken from the sample and then open the standard menu and choose the theory chi(k) data. Right click on the standard to use the same standard for all data sets. Redo the background removal. Plot the chi(k) data and the standard. If they are different at low k you need to adjust Ezero so that they start at the same place. When you are done adjusting your choice for Ezero, plot the xmu(E) data and make sure that the circle that is your value for Ezero is on the edge. If it isn't, then the theory that you used to align the data doesn't work..and you need to start over in step 6. If the value is on the edge some where, right click on the Ezero choice and set its value for Ezero of all the scans. Then merge all the data in chi(k) and write it out for use in Artemis...as you did the first time making sure that all the scans have a similar bkg. 8) Redo the fit: Now the data and the theory are aligned and you can redo the fit varying all the parameters...deltaE in Artemis should be small..and the background should give a nice match at low-r values. Often this entire procedure is used at the end of your analysis to make a nice figure for publication. 9) Bonus of using an energy reference: If you used an energy reference measured simultaneously with the data sets then in the end you can simultaneously fit the different samples with only one deltaE in the fit for all of the samples. If you aligned each one independently then each data set should have it's own deltaE value..unless you have more information and can prove that they should be the same. HTH Shelly