Name Surname
Machine Learning assisted multi-modal investigations to unravel sulfide accumulation processes in ore deposits of economic interest
Main Supervisor: Giada Iacono-Marziano (Centre National de la Recherche Scientifique)
Co-Supervisors: Kenneth Koga (University of Orleans) | Ciprian Stremtan (Teledyne Photon Machine)
Objectives
The project aims to characterize chemical and spatial correlations recorded in sulfide grains in natural and experimental samples, and to discover indicators of its formation processes, such as fluid or melt-derived origin, growth rate, and number of growth episodes. ML-assisted multimodal spatial analyses are critical for the study because of their range of spatial variations from µm to cm scale, and the variety of analytical data from FIB-SEM, EPMA, LA-ICPMS, and X-ray tomography. That is the innovation and challenge of the project and is an ideal problem to work with ML algorithms to test the sulfide formation models that are already proposed. The target research site is Noril’sk a well-known site with wealth of pre-existing data (the academic supervisor possesses samples). The developed methods will be also investigated for other sulfide deposits in which samples are already available (e.g., Sakatti and Kevitsa magmatic sulfide ores in Finland, Black Swan ores in komatiite (Western Australia), and Bushveld Complex (South Africa).
Epected Results
- The identification of statistically significant workflows identifying the classes of sulfide formation process, ultimately helping economic prospection of sulfides;
- To establish analytical methods required to obtain such criteria (chemical and spatial indicator);
- Improve our knowledge in exploiting multimodal data in sulfide geochemistry.