Artificial intelligence-driven magnetic property prediction and materials discovery for next-generation spintronics

Abstract

The rapid advancement of spintronic technologies has intensified the demand for magnetic materials with precisely engineered properties, including high spin polarization, large magnetic moments, tunable exchange interactions, and robust thermal stability. However, the simultaneous optimization of these properties remains challenging due to the vast compositional, structural, and interfacial design space governing magnetic systems. In this context, artificial intelligence (AI) and machine learning (ML) have emerged as powerful tools for accelerating materials discovery and property prediction. This review provides an overview of recent progress in ML-driven spintronic materials research, focusing on how data-driven models, integrated with high-throughput first-principles calculations and experimental databases, are transforming the prediction of key magnetic properties. Advances in descriptor engineering, spanning compositional, structural, and electronic features, are discussed alongside emerging approaches such as graph neural networks and physics-informed learning. Key material classes, including Heusler alloys, topological spin systems, and two-dimensional magnets, are highlighted in the context of AI-assisted screening and inverse design. The review concludes by outlining future directions toward physics-guided, uncertainty-aware, and autonomous discovery frameworks that may enable closed-loop optimization of next-generation spintronic materials.