Visible-to-infrared reflectance spectra of magnetite–chlorite-controlled mixtures

Magnetite is an important source of iron ore in Australia and globally. However, deleterious gangue mineral removal from magnetite ore concentrate is economically and environmentally costly, and so cost-effective delineation of these minerals using visible and infrared reflectance spectra (a.k.a. hyperspectral) data has tremendous potential for process optimisation. Mineral unmixing and predictive modelling methods require reference spectral libraries, which rarely include complete mineral mix series or span the wavelength range of modern hyperspectral sensing instruments. In this study, a reference spectral library containing magnetite mixed with chlorite (n = 34) was created to address this knowledge gap. Hyperspectral data were acquired across the 380–16 669 nm wavelength range, using HyLogger-3 and Fourier transform infrared instruments. Mineral ratios and grainsize variation, important for chemometric modelling applications, were determined using quantitative X-ray diffraction and scanning electron microscopy with energy-dispersive X-ray spectroscopy analysis. In magnetite–chlorite mixtures, the detection limit of chlorite is dependent on grainsize and the wavelength range of diagnostic chlorite absorption features. A chlorite-related absorption feature at 2830 nm could be detected at 10 wt% from a mixture of <30 µm grainsize magnetite and <45 µm grainsize chlorite, and a linear correlation indicated that <10 wt% concentrations could be detected. For chlorite grainsizes between 45 and 180 μm, the detection limit based on 2830 nm feature depth was at around 30–40 wt%. For the chlorite feature centred around 2250 nm, the detection limit was at ∼50 wt%, and for the feature at 9740 nm, the detection was only reliable from the fine-grainsize mix series, where chlorite detection was ∼10 wt%. The magnetite–chlorite spectral dataset resulting from this study can be applied to gangue mineral prediction from most Australian magnetite ores, as these are crushed to the corresponding fine grainsize, enabling a significant improvement in ore characterisation and processing workflows.