Spatially detailed analysis of drill core samples with Laser-Induced Breakdown Spectroscopy: Detection, classification, and quantification of rare earth elements and lithium

Abstract

In the transformation towards climate neutral consumption, electric alternatives rise in favour of fossil energy sources in a variety of different fields. Lithium and several elements from the group of Rare Earth Elements (REEs) are of particular importance for modern battery production and the supply of green energy, and therefore play a crucial role for this transformation. Their demand has increased constantly over the last years and an ongoing trend is expected for the future. New instruments and analytical methods for the geochemical investigation of drill cores can support mineral exploration and active mining and thereby help to cope with the growing demand. Laser-Induced Breakdown Spectroscopy (LIBS) is an analytical technique with many advantages for the analysis of drill core material. It has a high measurement speed, no sample preparation is needed, and major, minor as well as trace elements can be detected in a single spectrum under atmospheric conditions. Nevertheless, physical and chemical matrix effects prevent a straightforward analysis of heterogeneous material, which is especially relevant for spatially resolved investigations of drill core samples. This work displays novel methods that enable the analysis of LIBS mappings of large REE- and Li-bearing drill core samples by overcoming the problematic matrix effects with different un- semi- and supervised machine learning algorithms. In the first application, drill core samples of brecciated carbonatites were spatially investigated with LIBS to establish an intensity limit for La using the k-means clustering algorithm. Based on this intensity limit, REE enrichments were detected in the investigated sample. Afterwards, the REE content of the sample was estimated with mass balance calculations. For the second application, different Li-bearing drill core samples were mapped in high resolution with LIBS and a new classification model was developed. It combines Linear Discriminant Analysis (LDA) and One-Class Support Vector Machines (OC-SVM) to enable the classification of minerals that were covered by a train set, while also identifying LIBS matrices that are unknown to the model. The third application combined Laser Ablation – Inductively Coupled Plasma – Time of Flight Mass Spectrometry (LA-ICP-TOFMS) with LIBS measurements of the same sample. After image registration, this reference sample was used to create a Least-Square Support Vector Machine (LS-SVM) quantification model, which can be employed to convert LIBS intensities of similar material into element concentrations. The model allows a pixel-specific, spatially resolved quantification of multiple minerals with a single model. Each application displays possible solutions to minimize the influence of physical and chemical matrix effects on the spatial analysis of LIBS mappings of large drill core samples, which enables different kinds of analysis. Thereby, the great potential but also the challenges of LIBS as an analytical tool in geology and mining are highlighted.