We have published a paper in the Journal of Petrology describing a novel approach to quantifying the distribution of atoll garnet in a metamorphic rock using micro-computed tomography and deep learning algorithms capable of automated textural analysis of crystal shapes in 3D. We show that this method can be applied to study the kinetics of fluid-driven metamorphic reactions such as atoll garnet formation, and propose a kinetic model to calculate the reactivity of metastable garnet to form atolls at different pressure and temperature conditions.

Garnet grain shapes and distribution obtained using micro-computed tomography (μCT) and grain shape analysis (a) Selection of garnet grain shapes. Three shape-based classes capture the gradual progression from whole garnets, with no evidence of internal resorption, to atoll garnets, with complete core dissolution. The μCT imaging and segmentation method results in other artefactual grain geometries; these are filtered from the dataset by classifying them into separate classes. (b) Microspatial distribution of garnet. Circles represent the centroid position of all whole, pitted, or atoll garnet grains (r > 0.28 mm) in ZS-21-02. Circle size is proportional to the grain size, and the colors differentiate the shape-based subpopulations. (c) CSD of the entire garnet population (gray), atoll garnet (red), and pitted garnet (blue) in ZS-21-02. Crystal size before resorption is estimated as a
convex hull that fits the grain. The x-axis shows the grain size as spherical equivalent radius on a logarithmic scale. A second axis shows the fraction of grains in each size category that are atoll (red) and pitted (blue). Segmentation cut-off of 0.28 mm is highlighted by the vertical line.

Find out more

Hartmeier, P., Lanari, P., Forshaw, J., Markmann, T.A. (2024). Tracking garnet dissolution kinetics in 3D using deep learning grain shape classification. Journal of Petrology, 65, 1-9. Download pdf | Visit the journal webpage

Pin It on Pinterest

Share This