Coercivity influence of nanostructure in SmCo-1:7 magnets: Machine learning of high-throughput micromagnetic data

TL;DR

This study uses machine learning and high-throughput micromagnetic simulations to identify key nanostructural features influencing coercivity in SmCo-1:7 magnets, enabling inverse design of optimized nanostructures.

Contribution

It introduces a combined machine learning framework with extensive simulations to understand and design nanostructures for improved coercivity in SmCo-1:7 magnets.

Findings

1:5-phase enhances coercivity via high anisotropy.

Z-phase improves pinning through domain wall energy fluctuations.

Magnetocrystalline misorientation is the dominant factor.

Abstract

As a pinning-controlled permanent magnet, tailoring the cellular nanostructure of samarium-cobalt-based 1:7-type (SmCo-1:7) magnets remains crucial for improving magnetic performance. Jointing forward and inverse machine learning models with the high-throughput micromagnetic simulations (42,300 runs), we identify the nanostructural and magnetic features that are most effective for coercivity, combining both nucleation and pinning mechanisms. Sensitivity analyses reveal that the 1:5-phase enhances coercivity by providing high anisotropy, and the Z-phase strengthens pinning through fluctuations in domain wall energy. Cu additions in the 1:5-phase significantly reduce coercivity, while Fe substitutions in the 2:17-phase modestly reduce coercivity but improve pinning locally and increase saturation magnetization. Among all examined features, magnetocrystalline misorientation emerges as the dominant factor. Finally, the framework enables the inverse design of nanostructures with prescribed coercivity, demonstrating a computationally cost-effective toolkit for guiding the performance tailoring of SmCo-1:7 magnets.