Work

Closed-loop active-inverse learning for constrained discovery of high-resistivity Fe-based amorphous alloys

Year: 2026

Type: article

Abstract: High-resistivity Fe-based amorphous alloys are difficult to design because electrical transport and glass-forming ability must be optimized simultaneously in a sparse composition space. Here, we develop a closed-loop active-inverse learning framework for accelerated discovery of high-resistivity Fe-Si-B-Ni-Cr amorphous alloys. Using a literature-derived dataset, the framework combines resistivity regression, amorphous-feasibility classification, constrained inverse search, and iterative experimental feedback. After three closed-loop cycles, GMM-based local enrichment was used to densify experimentally supported high-potential regions and improve subsequent model refinement. This stage yielded additional amorphous alloys with resistivities up to 202.85 μΩ·cm. A final experimental validation round identified Fe 65.11 Si 14.96 B 10.04 Ni 6.9 Cr 2.99 (D3), which exhibited a resistivity of 215.27 μΩ·cm. Structural and electronic analyses suggest that the resistivity increase is associated with chemical disorder ( σ ) and hopping-related transport descriptors( t ¯ ). The framework provides a practical strategy for constrained inverse design of amorphous alloys under sparse-data conditions.

Source: Materials & Design

DOI: 10.1016/j.matdes.2026.116895