Laser-Induced Breakdown Spectroscopy and Chemometrics of Diffuser Alloys for Tritium Production and Byproduct Purification.

This study evaluates the potential of laser-induced breakdown spectroscopy for real-time monitoring of palladium and palladium-based alloys used for tritium recovery. Co-evolving gases include residual oxygen, nitrogen, methane, water, carbon monoxide, and carbon dioxide in the manufacturing of tritium at the Savannah River Site from its establishment. Palladium metal and palladium alloys were analyzed with emphasis on the 320-350 nm spectral region, where relative intensity ratios among key emission lines provide compositional insight. Spectral deconvolution was performed using the atomic emission database to resolve overlapping peaks to unveil spectral signatures of Ru (emission at 343.6681 nm) neighboring dominant palladium features. Given the complexity of the resulting spectral data sets, machine learning techniques were applied to enhance interpretability and predictive performance. Principal components analysis and partial least-squares were employed for dimensionality reduction and classification modeling for their ability to capture nonlinear relationships and improve classification accuracy. These approaches were developed as preliminary models to provide an initial spectral library for the highest predictive accuracy for tritium impurity removal and diffuser process upsets. Our advancements and progress aim at employing machine learning as a powerful noninvasive diagnostic for monitoring gas-metal interactions and impurity removal in tritium production systems that integrate chemometrics and adaptive decision-making strategies developed in this study.