Projecting into the Third Dimension: 3D Ore Mineralogy Via Machine Learning of Automated Mineralogy and X-Ray Microscopy

Analytical capabilities of scanning electron microscopy (SEM) and light microscopy (LM) allow accurate phase identification in two dimensions (2D), using a diverse set of characteristics. Chemistry can be accessed directly from energy (or wavelength) dispersive X-ray spectroscopy (EDS/WDS) and automated mineralogy (AM), or indirectly from backscatter electron (BSE) intensity. Particle shape and morphology can be quantified using image analysis tools on BSE, secondary electron (SE) or reflected light micrographs. Crystallographic data is now also widely available through electron backscatter detector (EBSD) data, which can provide data on stress and strain within a sample. However, with the exception of highly destructive ‘slice and view’ experiments [1], these analytical techniques have so far been limited to 2D. Here we aim to combine analytical data, in this case ZEISS Mineralogic Mining, a form of AM, with three-dimensinal (3D) X-ray microscopy (XRM) data, collected using the ZEISS VERSA, to display the full 3D mineralogy and texture of an ore body.