Gaussian, Lorentzian, and Pure Voigt Peak Fitting for UV/VIS, IR/FTIR, NIR/FTNIR Spectra
Spectral absorbances often involve heavily convoluted overlapping peaks. PeakLab’s full precision analytical Voigt function is one of 46 built-in spectroscopy functions which is used in the above peak fit to separate individual IR spectral components.
True Voigt Deconvolution
Fitting a true Voigt function will automatically deconvolve the separate Lorentzian and Gaussian components of each spectral peak.
In the above optimum UV-VIS fit, a nine-Voigt model fits the data with a variance error of just 4.5 ppm and an F-statistic of 1.6 million. The deconvolved Lorentzian spectral peaks are those that would be seen in a perfect instrument with no signal smearing from the Gaussian instrument response function, shown in red for each peak. An optimized Voigt fit such as this reveals the principal wavelengths that will be seen by an effective chemometric modeling system.
Voigt Models with IRFs for Asymmetric XPS and Raman Peaks
XPS and Raman spectral peaks generally have an intrinsic asymmetry that can be effectively fit with Voigt models containing instrument response functions.
The XPS peak shape above is fitted to a Voigt with an IRF consisting of a sum of the simplest probability and kinetic distortions. In the above figure, the red points are the raw data, the blue curve is the peak fit, and the green peak is the deconvolved Voigt with the asymmetric distortion removed.
Fluorescence excitation and emission spectra can be complex with highly convoluted absorbances. Often a simple multiple-Gaussian model suffices to capture the principal excitation and emission wavelengths.
The above are Coumarin 6 excitation and emission spectra. Note the unusual shapes of the overall excitation and emission curves.
Chemometric Predictive Modeling of Spectroscopy Data
PeakLab introduces a new and innovative chemometric modeling solution that is an attractive alternative to the traditional PLS and PCR modeling. These are direct spectral models that outperform the PLS and PCR ‘unscrambling’ algorithms for predictive accuracy and web-based computational performance. These predictive models for spectroscopy are far simpler to understand and exceedingly easy to code and evaluate.