Laser-Induced Breakdown Spectroscopy (LIBS) has gained interest among analytical techniques the last few years, thanks to its relative simplicity of use: generate compositional information from a material with minimal sample preparation. However, accurate recognition of chemical species in complex mixtures typically requires expert interpretation or extensive calibration datasets, and usually can only be performed on selected chemical elements. To mitigate these limitations and enhance the analytical capability of LIBS, we propose a novel vector space model coupled with singular value decomposition (SVD) and demonstrate its potential for both qualitative and quantitative LIBS analysis. In this vector space based algorithm, each chemical element is encoded as a unit vector; the complete set of unit vectors constitutes a basis that spans the spectral space of any unknown sample. Synthetic spectra generated from the LIBS spectra database of the National Institute of Standards and Technology is used to construct a database of 16 elements and assess the relevance of our approach in the identification and quantification of alloys in LIBS spectra. The algorithm reliably identified all elemental components and estimated their proportions; quantitative precision was highest for simpler binary alloys and decreased with compositional complexity. This approach provides a promising foundation for automated and calibration-light LIBS analysis. The limits of this approach are also discussed and some ideas for improvement are proposed.
Douti et al. (Fri,) studied this question.