Relationships between intensity of deformation induced Cr–Al chemical zoning and geometrical properties of spinel: An approach applying machine learning analyses

Abstract

Chromian spinel [(Mg, Fe2+) (Cr, Al, Fe3+)2O4], a fundamental petrogenetic indicator in Earth's crust and upper mantle systems, carries a unique Cr–Al chemical zoning. It has been postulated that a specific variant of this Cr–Al chemical zoning, particularly when influenced by deformation, can offer considerable insights into the processes of deformation of peridotites, and further advancing our understanding of spinel deformation. In our study, we observed distinct Cr and Al distributions in spinel grains showing the Cr–Al chemical zoning within a dunite sample from the Horoman Peridotite Complex, contingent on their geometric properties such as grain size and aspect ratio. Our observations and previous studies suggest a strong correlation between geometric properties and Cr–Al chemical zoning, but the complexity of these relationships has not been fully explored. To bridge this research gap, we utilized a combination of electron backscatter diffraction (EBSD) and energy-dispersive X-ray spectroscopy (EDS) to provide detailed grain descriptions and analyze origin of the observed Cr–Al zoning. Furthermore, Machine learning algorithms were applied to elucidate the relationships between Cr–Al zoning and geometric properties. This integrated approach illuminated intricate relationships between the geometrical properties of spinel grains and their Cr–Al chemical zoning. The results of this study suggests that lattice diffusion is primarily activated in coarser grains (cluster 1), moderately in intermediate-sized grains (cluster 2), and not active in finer grains (cluster 3). We propose a model where coarser grains undergo deformation predominantly via dislocation-assisted diffusion creep, with lattice diffusion as the primary underlying mechanism. Conversely, finer grains seem to deform mainly through diffusion creep, where grain boundary diffusion plays a significant role. This study offers a novel model, developed via machine learning, for understanding variations in the intensity of Cr–Al chemical zoning and the respective deformation mechanisms in spinel grains.