Machine learning-driven dissection of lithium distribution and enrichment mechanisms in bauxite-bearing rock series: insights from northern Guizhou, south China

Le résumé fourni par la source

• A machine learning framework enables rapid lithological classification and resource assessment. • Machine learning (K-means++ and PCA) reveals Li enrichment mechanisms in Guizhou bauxite-bearing rocks. • lg( αAl/Si ) and lg( βTi/K ) are critical Li-enrichment indicators, and their ratio reveals bauxite weathering intensity. • Moderate weathering intensity optimizes Li accumulation in clay and low-grade bauxite stages. Lithium (Li), a critical material for clean energy technologies, faces escalating global demand driven by the rapid growth of battery-powered industries. Bauxite-associated Li resources are extensively distributed and hold immense potential, offering a promising alternative for countries lacking conventional Li reserves. This study employs unsupervised machine learning—integrating the K-means++ clustering algorithm (KCA) and principal component analysis (PCA)—to investigate the Li distribution and enrichment mechanisms in the Permian karst bauxite-bearing rocks of northern Guizhou, China. KCA-based lithological classification categorized 435 samples into eight geochemically distinct clusters, achieving 65.8∼100% consistency with manual identification. PCA dimensionality reduction revealed that the first three principal components explain 65.9% of the geochemical variance, with weathering processes (governed by SiO 2 , TiO 2 , Al 2 O 3 and K 2 O) identified as the primary control on Li enrichment. Our findings confirm that Li predominantly accumulates in overlying/underlying aluminous rocks, clastic bauxites, and compact bauxites, peaking during weak-to-moderate lateritization. Additionally, the Al/Si and Ti/K ratios are identified as crucial indicators for Li enrichment, with the optimal intervals associated with significant Li enrichment ranging from lg( αAl/Si ) = -0.08 to 0.75 and lg( βTi/K ) = 0.5 to 1.5. Overlapping these ranges exhibits the highest Li concentrations (up to 5, 404 × 10 -6 ). It is further confirmed that moderate clay mineralization and low-grade bauxite formation, coupled with intermediate weathering intensity, create favorable conditions for Li accumulation. This work establishes a machine learning-driven framework for rapid lithological discrimination and resource assessment in complex bauxite systems, providing critical insights for exploration of clay-type Li deposits.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.