Multispectral Infrared and Machine Learning Methods for Assessing Critical Raw Material Potential in Mining Residuals

Critical Raw Materials (CRMs) are vital to modern technologies and key sectors such as renewable energy, electronics, and aerospace. Growing geopolitical, environmental, and market risks make supply diversification essential. Mining residuals, including tailings and waste rock, often retain significant CRM concentrations due to past processing inefficiencies, ore grade changes, and advances in extraction technologies. Exploring and recovering CRMs from these residual resources can contribute to resource security and support circular economy objectives. 
This study evaluates an integrated multispectral infrared spectroscopy approach, combined with machine learning, to identify and map CRM-hosting mineral phases in mining residuals. 
Reflectance spectra in the visible–near infrared (VNIR) and shortwave infrared (SWIR) ranges (0.35–2.5 µm) were acquired using an ASD FieldSpec instrument. Mid- to long-wave infrared spectra (2.5–15 µm) were collected using a Fourier Transform Infrared (FTIR) 4300 spectrometer. Together, these data provide complementary mineralogical information across a broad infrared spectral range. Spectral interpretation was conducted to identify the different mineral phases. The spectral datasets were analysed using supervised machine learning techniques, specifically support vector machines (SVM) and partial least squares – discriminant analysis (PLS-DA). These methods were used to classify materials into relatively high- and low-CRM concentration classes, supported by mineralogical and geochemical reference data.
Integrating VNIR–SWIR and FTIR spectral data enhances discrimination of CRM-hosting mineral assemblages and supports spatial mapping in heterogeneous mining residual deposits. When combined with machine learning, infrared spectroscopy offers an efficient tool for rapid assessment of secondary CRM resources. This scalable method can be applied to three-dimensional modelling to quantify CRM distributions within tailings volumes.
Overall, this integrated methodology enhances the mineralogical and geochemical characterization of mining residuals, supporting informed decisions for secondary resource exploration and recovery.