Spectral X-ray computed micro-tomography: towards 3-dimensional ore characterization


Spectral X-ray computed micro-tomography: towards 3-dimensional ore characterization

Sittner, J.; Guy, B. M.; Da Assuncao Godinho, J. R.; Renno, A.; Cnudde, V.; Merkulova, M.; Boone, M.; de Schryver, T.; van Loo, D.; Roine, A.; Liipo, J.

Most ore characterization studies are based on analytical tools that are limited to a 2-dimensional (2D) surface of a 3-dimensional (3D) sample. This not only limits the number of particles available for analysis but also results in a lack of 3D morphological information. Especially for the characterization of trace phases, 2D analysis is time consuming. X-ray computed micro-tomography (micro-CT) represents an established 3D technique that is used in numerous applications such as medicine, material science, biology, and geoscience. However, a major drawback of micro-CT in the characterization of ores is the absence of chemical information, which makes mineral classification challenging.
Therefore, we present Spectral X-ray computed micro-tomography (Sp-CT). It is an evolving technique in different research fields and is based on a semiconductor detector that provides chemical information of a sample (e.g., CdTe). This detector can be used with a conventional CT scanner (TESCAN CoreTOM in this study) to image a sample and detect its transmitted polychromatic X-ray spectrum. Based on the spectrum, elements in a sample can be identified by an increase in attenuation at specific absorption edge energies. Therefore, chemically different minerals can be distinguished inside a sample from a single CT scan in the micrometer range. The method is able to distinguish elements with absorption edges in the range from 25 to 160 keV, which applies to elements with Z > 48.
We present the workflow of an ore characterization study using a combination of Sp-CT and high-resolution micro-CT with an example of Au-U ore from the Witwatersrand Supergroup. Different drill core samples from the Kalkoenkrans Reef at the Welkom Gold field were investigated. With the chemical information from the Sp-CT, minerals such as gold, U-phases, and galena can be identified based on their K-edge energies in the spectrum. Several Sp-CT scans were used to train a machine learning segmentation model in the software Dragonfly (version 2021.1) to segment the high-resolution CT data into multiple segments, e.g., gold, U-minerals, sulfide minerals and matrix minerals. The segmented data was then used to extract 3D mineral properties of the segments such as 3D volume or 3D surface area. This new non-destructive approach provides 3D information on distribution of chemically different minerals without any sample preparation. This information can be used for mineral processing simulations but also for genetic mineralogical studies.

  • Open Access Logo Lecture (Conference) (Online presentation)
    3rd European Mineralogical Conference, 29.08.-02.09.2021, Cracow, Poland

Permalink: https://www.hzdr.de/publications/Publ-33594
Publ.-Id: 33594