Machine Learning Guided Growth of Ultrathin 2D Fe ₃ O ₄ With a Single Magnetic Domain

ABSTRACT 2D non‐layered ferrimagnetic materials have opened up opportunities to explore new physics and develop spintronic devices. Among them, 2D Fe 3 O 4 with a single magnetic domain has attracted great attention due to its high remanence ratio and data retention ability. Thickness reduction is effective to achieve single magnetic domain structure, which is challenging for non‐layered materials. Here, we established the relationship between the thickness of 2D non‐layered Fe 3 O 4 and growth parameters by a Bayesian nonlinear regression model based on machine learning. We studied the effect of concentration, solution volume, temperature, and substrate wettability on the thickness of 2D Fe 3 O 4 and found that the combined effect of concentration and substrate wettability has the greatest impact. Importantly, the converged and accurately predicted model suggests the optimal condition for the required thickness. Guided by this prediction, we grew single‐crystalline 2D Fe 3 O 4 with a thickness down to 4 nm, exhibiting a single magnetic domain structure. This work obtains single magnetic domain 2D ferrimagnetic materials and provides a feasible model to guide growth with the required thickness, promoting the preparation and application of 2D ferrimagnetic materials.