Determination of Ce and La in REE-rich ores using handheld LIBS and PLS regression

Abstract

In this study, we utilized a handheld LIBS (laser-induced breakdown spectroscopy) analyzer (Z-300, SciAps) to quantitatively determine cerium and lanthanum in ores enriched with rare earth elements (REEs). Due to the complex electronic structure of REEs and their presence in laser plasma in multiple ionization stages (neutral, singly- and doubly-ionized species), determining their concentrations using a handheld instrument with relatively low spectral resolution is a challenging task for on-field quantitative analysis. Another issue is the rich emission spectrum of matrix elements of ores, which can cause strong overlapping between lines of analytes and matrixes, preventing quantitative analysis of ores. To address this issue, we employed Design of Experiment based on Latin Hypercube Sampling to construct an artificial calibration set of samples with varying REEs content. We compared univariate calibration and projection to latent structures (PLS) regression for atomic and ionic lines of REEs, as well as for molecular bands of LaO (740.5 nm). We observed a worsening of calibration due to matrix effects for LaO molecules. To validate our method, we used certified reference material of OREAS natural ore samples to simulate the real in-field application of the handheld analyzer. Our study revealed that Ce can be quantified regardless of concentration in niobium REE-rich ores with a relative error of about ±10% if the Ce II line at 462.82 nm is used, which is close to the level of quantitative accuracy. However, predicted concentrations of La in OREAS 460–465 appear to be biased by matrix effects. We conclude that the accuracy achieved in our study is mostly sufficient for qualitative analysis of uranium ores (OREAS 100a–102a). Furthermore, considering the possible matrix effects in La determination, we suggest that it is possible to use OREAS samples as a calibration set instead of artificial samples to build linear univariate regression models, as the lines employed in this study do not overlap. Our results are expected to optimize the procedure of online in-field monitoring of the mining process and provide rapid preliminary analysis for decision-making, with further, more time-consuming, and accurate analysis (e.g., using ICP-OES/MS) required as necessary.