Questions about the XPS spectra analysis:
The XPS spectra is deconvoluted into several components, my questions are
1) For one spectra, should the FWHM (full width at half maxima) of each component be set with similar width?
2) For two spectra, should the same component be set with similar width?
lets take an example: natural Oxide on Si and fit of the Si2p, The FWHM of the Si bulk is smaller than for the Oxide. The same is true for Al(met) and Al (ox). In some cases the different FWHM are due to different lifetimes of the core hole, or in other cases due to inhomogenities of the stoichiometry. A good starting point is to look in data bases to spectra of the clean substances you suggest to be present in your sample. Then you construct a start fit with all spin orbit and satteilte structures (e.g. Cu2p for CuII) of the measured core level. If you have data with FWHM, asymmetries etc. you can fix the values for some parts or components like FWHM, intensity ratios, asyymmetries...
And at last try to use a minimum of peaks first and add only a component if you have a clear evidence that this component is really present, otherwise you now you can fit with four peaks an elephant...
lets take an example: natural Oxide on Si and fit of the Si2p, The FWHM of the Si bulk is smaller than for the Oxide. The same is true for Al(met) and Al (ox). In some cases the different FWHM are due to different lifetimes of the core hole, or in other cases due to inhomogenities of the stoichiometry. A good starting point is to look in data bases to spectra of the clean substances you suggest to be present in your sample. Then you construct a start fit with all spin orbit and satteilte structures (e.g. Cu2p for CuII) of the measured core level. If you have data with FWHM, asymmetries etc. you can fix the values for some parts or components like FWHM, intensity ratios, asyymmetries...
And at last try to use a minimum of peaks first and add only a component if you have a clear evidence that this component is really present, otherwise you now you can fit with four peaks an elephant...
I am fully agree with Christian Pettenkofer's answer. You first go for minimum number of peaks and use the data base of clean substance in your sample. Many researcher also use same FWHM for all the components with justification.
To add to C Pettenkofer's excellent answer I would suggest only these two notes during peak fitting (not deconvolution) of XPS spectra. First, the peaks of different elements can have different half-widths and peak shapes. For example, metals can be Doniach-Sunjic while C 1s is mostly gaussian. Secondly, when fitting one spectrum versus another, the peak shapes should be kept predominately the same, but the peak positions and half-widths may vary within some reasonable bounds. For example, on TiO2, the presence of sub-oxides can cause what appears to be a minor shift in the position or or broadening of the Ti+4 peak. The one exception is when you would be convolving a line shape with an analyzer transmission function via a Voigt and/or a Gauss + Lorentz. The shape factor in the Voigt or the fraction Gauss could be allowed to vary within minor limits from one spectrum to another. All of these changes should be second order though.
In summary, when peak fitting from spectrum to spectrum ...
* Keep the shapes consistent for each element
* Keep the positions and half-widths constrained to be the same within reasonable limits
Finally, if you really are doing peak deconvolution (rather than peak fitting), then you probably have a different set of questions to ask.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Material Characterization