Multivariate Analysis of LIBS Spectra for Geomaterial Identification, Discrimination, and Classification

LIBS is an emerging technology for geochemical analysis that is in laboratory use and which has attractive potential for real-time man-portable and/or standoff analysis in the field. Since the technique is simultaneously sensitive to all elements, a single laser shot produces a broadband LIBS emission spectrum that captures the complete elemental composition of the target sample. Statistical signal processing techniques can be used to identify, discriminate, or classify different geological materials.

A commercial 'close-in' benchtop LIBS system with broadband detection from ~200-965nm and a developmental 25m 'stand-off' LIBS system with broadband detection from ~200-840nm were used to acquire single-shot LIBS spectra for a wide variety of carbonate and silicate minerals, rocks, and soils. Classification has been achieved with a high degree of success using two different chemometric techniques,‘soft independent method of class analogy' (SIMCA) and ‘partial least squares discriminant analysis' (PLS-DA). By using multivariate chemometric techniques, even chemically similar geological materials can be readily discriminated. Because each broadband LIBS spectrum records the complete elemental character of the sample analyzed and therefore can be considered a unique 'geochemical fingerprint', it is possible to discriminate different samples of the same mineral, rock, or soil type.