Re: [Ifeffit] Normalization in Athena
Hi everybody, I am sure that Athena was tested a lot, but as Jeremy pointed out, which I think now is the answer: If the background is substracted before the fitting of the post edge. That is actually how we are normalizing XANES data in our group and that is what I tried in Athena. It is actually working much better with 2. order norm and flattening. Since I do not want to do it by hand for all spectra and import them back into Athena for LCF. This time I attached a the example Athena project file. But with this approach I rarely get the spectrum oscillating around one. I will think about it. Thanks for the hint, Matt, about "no points" in the range. But Athena is acutally complaining, if there are no points to fit the line. Of course, I read the chapter on Normalization in Athena. And what is giving me trouble, why I try the "1-point" approach, is that the small differences in the first post-edge normalization point make a great deal of difference in the normalization of the spectra respective to each other and thus LCF. Finally, I believe it is a very interesting question of taste how to determine the edge step for XANES data or when to substract the background. Or is there any specific reason for a special approach? Thank you all for your help!! Greetings, Lisa
Hi Lisa,
Be sure that Athena has been rigorously tested and she does know how to normalize data. It is much more likely that the default values are not the "best" ones for your data.
My book chapter goes over how Athena normalizes data in detail. If you send me your email address I'll send you a copy. In particular see figure 14-16.
Cheers, Shelly
Shelly,
I agree that Athena applies its (her) algorithm correctly. Whether that algorithm is correct may be a matter of opinion. The apparent discrepency Lisa sees is due to Athena's algorithm:
i.e., fit preedge line, fit postedge function to normalization range, extrapolate both functions to edge energy to find edge step.
Another option would be to fit predge line and subtract it. After that, fit the post edge normalization range.
With Athena, if there is a non-zero slope to the preedge line and you choose norm order 1, you get Lisa's result. With normalization order 2 or 3, the results should either be the same, or at least, much closer.
Jeremy
It is a little hard to evaluate your results, Lisa, without an example. It may help to attach a small project file which demonstrates your question.
Without an actual example, I would have to guess that the bottom line is an odd choice of parameters. This page from the document might be helpful:
http://cars9.uchicago.edu/~ravel/software/doc/Athena/html/bkg/norm.html
B
Lisa,
I think the issue you're seeing is most likely due to telling athena to fit the data between E0+199 and E0+200 with a constant. I'd recommend expanding that range some to give Athena a chance -- if there isn't any data between E=E0+199 and E0+200, Athena won't be able to figure out what the edge step should be. If you really want the edge step to be the value of pre-edge subtracted mu(E) at E=E0+200eV, you'll have to do that by hand.
--Matt
Hi Lisa,
On Mon, Oct 26, 2009 at 9:12 PM, Gudrun Lisa Bovenkamp
I am sure that Athena was tested a lot, but as Jeremy pointed out, which I think now is the answer: If the background is substracted before the fitting of the post edge. That is actually how we are normalizing XANES data in our group and that is what I tried in Athena. It is actually working much better with 2. order norm and flattening.
If you're concerned about normalization, do not flatten the data!!!
This time I attached a the example Athena project file.
I did not look closely at all the data in your project file, but the first group listed "Pb113 norm in Athena" has E0 = 13037.76. Extracting the mu(E) data, and looking in the region around E0 + 200 eV, I see these energy values: 13233.508 13235.121 13236.630 13238.452 13240.066 so that, in fact. there are no data between E0+199 = 13236.76 eV and E0+200 = 13237.76 eV.
But with this approach I rarely get the spectrum oscillating around one.
Actually, you're not supposed to get a spectrum that oscillates around one. Again, you will get a better normalization if you use a line (Normalization order=2) and a range of 100 to 250eV. When I do this with the "Pb113 norm Athena" and "Pb125 norm in Athena" data in your project, I get a mu(E) that starts out ~1 and decays away with energy, just as it should. The "norm in Orign" groups you have in your project all have about the same value for mu(E) at E=E0+200eV, but are not well normalized. --Matt
participants (2)
-
Gudrun Lisa Bovenkamp
-
Matt Newville