Material informatics for functional magnetic material discovery

Abstract

Functional magnetic materials are used in a wide range of “green” applications, from wind turbines to magnetic refrigeration. Often the magnetic materials used contain expensive and/or scarce elements, making them unsuitable for long term solutions. Further, traditional material discovery is a slow and costly process, which can take over 10 years. Material informatics is a growing field, which combines informatics, machine learning (ML) and high-throughput experiments to rapidly discover new materials. To prove this concept, we have devised a material informatics workflow and demonstrated the core components of natural language processing (NLP) to extract data from research papers to create a functional magnetic material database, machine learning with semi-heuristic models to predict compositions of soft magnetic materials, and high-throughput experimental evaluation using combinatorial sputtering and high-throughput magneto-optic Kerr effect (MOKE) magnetometry. This material informatics workflow provides a quicker, cheaper route to functional magnetic materials discovery.