Nicole Gilon, Nadia Baskali
Laser induced breakdown spectroscopy has demonstrated its high potential in material identification in many areas including field exploration, recycling industry and forensic science. A spectral LIBS database was built from more than 500 minerals, alloys, polymers and glass samples of known composition. Using a short and long delay for LIBS analysis, providing an elemental and a molecular fingerprint of each sample a spectral library was built. The unknown spectra are compared with the database. Similarity metrics are evaluated on repetitions and synthetic spectra, a Pearson correlation gave best results to provide accurate identification. Spectral comparison are then tested using mineral, metal (copper, nickel and iron/neodymium alloys) as examples of more and more complex spectra. The accuracy results are compared between similarity test and classification methods (k nearest neighbors, support vector machine and random forest). Results provided are different, for most materials, the developed protocol gave better identification results. The combined use of both molecular and elemental information allowed reducing the matching list to a restricted number of very close sample. It is especially useful in the case of rare earth elements (REE) alloyed with iron, and reliable identification is obtained in every case.