Unsupervised fitting of emission lines generated from laser-induced breakdown spectroscopy

Abstract This work addresses a bottleneck usually encountered in the processing of large amounts of LIBS data. It shows how the fitting of emission lines in a typical LIBS spectrum can be automated. This is done by exploiting the second derivative of the spectrum to obtain reasonable fitting parameter estimates that, together with a divide and conquer approach, can be used to fit large numbers of lines in a crowded spectrum without user intervention and executed within a reasonable timeframe. As a demonstration, we have successfully fit Voigt profiles to 625 lines in a Martian ChemCam spectrum.