Vanadium pre-edge peak fitting
Hi everyone, I've recently collected some V K-edge XANES spectra and am in the process of analysing the data. The pre-edge region has been shown to be quite useful for exploring the oxidation state and coordination geometry of vanadium in various samples. This approach requires the pre-edge peak area and centroid of unknowns to be compared with those of a suite of standards, with these parameters typically determined by fitting a peak (or peaks) to the pre-edge feature. The attached paper by Chaurand et al shows an example of how this is often done, but it raised a question in my mind that I wanted to clarify before proceeding to analyse my data in this way... Is it necessary to justify a particular selection of peak fitting parameters based on physical attributes? For example, Chaurand et al fits multiple pseudo-voigt peaks to the V pre-edge feature, with the Lorentzian contribution to this pseudo-voigt function constrained to be the same as the core-hole lifetime width at the V K-edge (i.e. 0.8 eV). The Gaussian contribution is allowed to vary, but the Lorentzian to Gaussian intensity ratio is set at 1:1. After justifying these parameters, the authors go on to say that *"It should be noted that width and height of the modeled pseudo-Voigt functions have little physical significance, being a convolution of two functions with significantly different width."* If this is the case, does it really matter what specific peak fitting parameters are used? If the goal is to simply obtain an accurate peak area and centroid of the pre-edge feature, wouldn't an empirical peak fitting approach provide comparable data? I'm not against fitting a series of pseudo-voigt functions constrained in a similar way, to my data, but I'd like to have a good justification for doing so. I'm quite new to the world of peak fitting in the context of XAS data analysis, so I'm looking forward to hearing the views of the mailing list. Thanks in advance. Will
Hi Will, I did a lot of work on vanadium edge data analyzing pre-edge and Exafs. Our papers showed very similar results to Chaurand and we used a flexible fitting function mostly because we had vanadium in variable oxidation states as well as variable geometry across our samples. We tried to establish comparison to some standards but would it be helpful to compare internally? One other suggestion is to also fit some type of function to account for changes in the main edge shape. Check some papers by Banerjee for buffalo or TEXAS AM as far back as 2008 Hope that helps, Chris Patridge Sent from my iPhone
On Dec 5, 2017, at 9:52 AM, Will Bennett
wrote: Hi everyone,
I've recently collected some V K-edge XANES spectra and am in the process of analysing the data. The pre-edge region has been shown to be quite useful for exploring the oxidation state and coordination geometry of vanadium in various samples. This approach requires the pre-edge peak area and centroid of unknowns to be compared with those of a suite of standards, with these parameters typically determined by fitting a peak (or peaks) to the pre-edge feature. The attached paper by Chaurand et al shows an example of how this is often done, but it raised a question in my mind that I wanted to clarify before proceeding to analyse my data in this way...
Is it necessary to justify a particular selection of peak fitting parameters based on physical attributes? For example, Chaurand et al fits multiple pseudo-voigt peaks to the V pre-edge feature, with the Lorentzian contribution to this pseudo-voigt function constrained to be the same as the core-hole lifetime width at the V K-edge (i.e. 0.8 eV). The Gaussian contribution is allowed to vary, but the Lorentzian to Gaussian intensity ratio is set at 1:1. After justifying these parameters, the authors go on to say that "It should be noted that width and height of the modeled pseudo-Voigt functions have little physical significance, being a convolution of two functions with significantly different width." If this is the case, does it really matter what specific peak fitting parameters are used? If the goal is to simply obtain an accurate peak area and centroid of the pre-edge feature, wouldn't an empirical peak fitting approach provide comparable data? I'm not against fitting a series of pseudo-voigt functions constrained in a similar way, to my data, but I'd like to have a good justification for doing so.
I'm quite new to the world of peak fitting in the context of XAS data analysis, so I'm looking forward to hearing the views of the mailing list.
Thanks in advance.
Will
_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit Unsubscribe: http://millenia.cars.aps.anl.gov/mailman/options/ifeffit
Hi Will, I worked with the Fe K-pre-edge, following the method of Wilke et al. (e.g. ref 20 in the Chaurand paper). Letting the peaks fit freely (but with shared FWHM and Gaussian/Lorentzian ratio) gave the best results. A few reasons to rely on the fitting program to give you the numbers: - The best number of peaks to fit can be affected by the resolution of your measurements. Crystal field theory might predict a specific number of transitions, but you might have non-local transitions, impurities, etc. that add extra peaks. Also, predicted transitions can be too close together to be fitted with individual peaks. - The Gaussian broadening and contribution (coming from the beamline resolution) is not always well-known. - The Lorentzian broadening and contribution (coming from core-hole lifetime-broadening, traditionally taken from theoretical values of Krause and Oliver in 1979) is also not easy to apply together with all unknowns mentioned above. As Chris mentioned, it is worth to find a good function to model the main edge. You can safely do that before fitting the pre-edge (ignoring the pre-edge peak). An arctangent + low-slope straight line + possibly a peak function should do the job. All the best. Alexey
participants (3)
-
Alexey Boubnov
-
Chris Patridge
-
Will Bennett