[Ifeffit] C and N NEXAFS normalization

[Jeff Terry]

Hi Peter,

I have used both methods in the past. Both can run you astray, if done  
improperly.

The better thing to do though is to fix the backgrounds. If you look  
at the I0 signal, you should be able to tell if it is C or N on the  
optics or on the I0 grid.

I have had far more trouble normalizing when C or O is on the grid. I  
have not done much N work. This can be fixed by evaporating Au onto  
the grid.

I just had this problem last week at SSRL. I had terrible backgrounds  
that a fresh coat of Au cleaned right up.

After cleaning the grid, both methods gave similar results.

Jeff



On Jun 8, 2010, at 10:52 AM, Peter Nico <psnico at lbl.gov> wrote:

> Hello All,
>
> I am looking for thoughts and opinions regarding the best way to  
> approach post-edge fitting and normalization of C and N NEXAFS  
> data.  I know soft x-rays aren't exactly the focus of the list, but  
> it is still the best access to a lot of people with a lot of XAS  
> knowledge, and I would be very interested in your opinions.
>
> Obviously, linear or quad curve fitting followed by normalization is  
> standard and well accepted for hard x-ray spectra.
>
> However, as far as I can tell there is no such accepted norm of C  
> and N NEXAFS treatment.  I have seen that some groups use a single  
> point normalization, e.g. 310 for the case of C, and simply divide  
> through the entire spectrum with this value.  Other groups take the  
> min and max value of the spectrum and map it onto a 0 to 1 scale.    
> Neither of these approaches seems satisfactory to me for data with  
> obvious backgrounds like those attached.  While these data are not  
> perfect, it seems like they should definitely be useable.
>
> My thought is why not do a 'standard' hard x-ray type normalization  
> like that shown with an e0 set to the theoretical ionization  
> potential for C or N.  I understand that this has problems with the  
> potential of including tailing sigma* transitions or first EXAFS  
> oscillations in the background region because it doesn't extend far  
> enough beyond the edge.  However, it seems to me an approach like  
> this must be 'less-wrong' than the more simplistic methods mentioned  
> above and capable of yielding useful data.
>
> Thoughts, comments, jeers?  --Peter
>
>
> <moz-screenshot-4.jpg>
>
>
> <moz-screenshot-5.jpg>
> <Iffefit_question.ppt>
> _______________________________________________
> Ifeffit mailing list
> Ifeffit at millenia.cars.aps.anl.gov
> http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit