The Magnetic Mineralogy of Carbonaceous Chondrites: A Microscopy and Machine Learning Study of Meteorite WIS 91600

A central challenge in the paleomagnetic study of meteorites is to characterize their diverse remanence carriers. Here, we develop a machine‐learning assisted workflow that combines multi‐scale (mm to nm) and multi‐dimensional (2D‐3D) microscopy to build a comprehensive picture of ferromagnetic mineralogy in the C2 ungrouped carbonaceous chondrite Wisconsin Range (WIS) 91600. Scanning electron microscopy images are acquired across the entire thin section with sufficient resolution to locate all major remanence carrying ensembles. A deep learning classifier yields the volume fraction and morphometric properties of each ensemble, distinguishing between different textural varieties of the same ferromagnetic mineral. Chemical maps of representative ensembles are analyzed using an interactive machine‐learning tool that provides automated mineralogical segmentation and hyperspectral unmixing of the underlying endmembers. Focused‐ion‐beam nanotomography is used to reconstruct representative volumes with 3D spatial resolution sufficient to characterize all remanence carriers down to the stable single‐domain range. 4D scanning transmission electron microscopy provides access to grains in the superparamagnetic size range and diffraction data to enable phase identification. We conclude that the matrix, composed of phyllosilicates with fine‐scale pyrrhotite, pentlandite and magnetite, has a high potential to carry strong and stable paleomagnetic remanence in carbonaceous chondrites. The magnetic contribution from matrix‐hosted pyrrhotite is lowered in WIS 91600, however, due to transient heating inducing M C rather than 4C vacancy ordering. The remanence carrying potential of matrix‐hosted magnetite is similar to magnetite framboids but is less likely to be adversely affected by strong interactions, making it the preferred target for paleomagnetism.