Database development and LIBS calibration for the LIBS-Raman Sensor for planetary exploration

Laser-Induced Breakdown Spectroscopy (LIBS) has been a valuable technique in planetary exploration, providing rapid, in situ elemental analysis of extraterrestrial materials. The LIBS-Raman Sensor (LiRS), currently being evaluated for its capabilities for future planetary exploration missions at York University, integrates LIBS with complementary spectroscopic techniques to enhance geochemical and astrobiological investigations. To support the calibration and validation of this instrument, we have compiled a comprehensive database of well-characterized geological samples with known compositions, including carbonate, silicate, sulfate, and planetary simulant standards. These samples were analyzed under controlled conditions to create a robust machine learning-based elemental quantification dataset. Various statistical and regression models, including Gaussian Process Regression (GPR), Artificial Neural Networks (ANN), and Support Vector Machines (SVM), were evaluated for their predictive accuracy in determining compositions (as oxides). The results indicate that GPR consistently outperforms other methods for key planetary oxides such as SiO₂, Al₂O₃, and FeO${}_T$, while ANN and SVM offer strong performance for specific oxides like K₂O and MgO. This work demonstrates the potential of data-driven analytical techniques to improve LIBS-based geochemical analysis for future planetary missions. The developed database and models will aid in refining instrument calibration, improving spectral interpretation, and supporting ongoing planetary exploration efforts.