Optimization of laser-induced breakdown spectroscopy for analyzing heterogeneous materials

Rapid, real-time elemental analysis of heterogeneous materials such as agricultural soils and oil sands tailings are globally demanded for environmental resource management and sustainable food production. Although laser-induced breakdown spectroscopy (LIBS) is ideally suited for rapid multi-element analysis, analyzing heterogeneous samples using LIBS is extremely difficult due to signal variability, matrix effects, difficulties in quantifying trace elements, and classifications of complex samples. Our research intends to improve LIBS measurement accuracy by optimizing the experimental setup and data analysis method. Recent experimental findings indicate a trade-off between enhancing the signal-to-noise ratio of emission lines and the contribution of continuum emission from free electrons via the inverse bremsstrahlung process when the laser energy is increased. The contribution of continuum emission can be decreased by collecting plasma emission at a more co-linear angle with the laser beam. For further minimization of the continuum emission contribution, the continuum emission is estimated using least-squares regression and performed continuum normalization and baseline correction. Through data-driven modeling employing machine learning algorithms and proper sample preparation, the LIBS signal variability caused by heterogeneity in the target can be greatly reduced. Our research shows that bitumen concentration can be determined in heterogeneous tailings samples using partial least squares regression (PLSR), and LIBS measurement is significantly improved for wet samples compared to dried tailings samples, confirming the capability of LIBS to perform multi-element analysis of heterogeneous samples in the field without sample preparation.