Continuous-wavelet-transform-based automatic curve fitting method for laser-induced breakdown spectroscopy.

Abstract

In this work, an automatic curve fitting method based on a continuous-wavelet transform (CWT) is proposed to resolve overlapped peaks and to adaptively extract the major peaks in laser-induced breakdown spectroscopy (LIBS). From the local minimum of the second derivative of the LIBS spectrum calculated with CWT, the number of individual peaks is determined, and corresponding peak positions are estimated. The full width at half-maximums (FWHMs) of individual peaks are estimated from the separation of two maxima siding the minimum. A threshold is introduced to eliminate the small peaks and therefore reduce the number of fitting parameters and adaptively extract the major peaks with different spectral intensities. The Trust-Region algorithm is used for parameter optimization. The proposed method is used to analyze both simulated LIBS spectra and experimental overlapped peaks. Both simulated and experimental results show that the proposed method can resolve overlapped peaks even with a low separation degree, although the minimum resolvable separation degree depends on the FWHM ratio and strength ratio of individual peaks and the wavelet scale. In a LIBS calibration experiment of N2/SF6 gasses mixture, after resolving the overlapped peaks with the proposed method, better linear correlations between the concentration and intensity of F (with an adjusted R-squared value 0.9972), as well as between the concentration ratio and intensity ratio of nitrogen to fluorine (with adjusted R-squared values >0.98 and 0.99) are obtained.