University of Cambridge
Yusuf Hamied Department of Chemistry

Using machine learning to discover more sustainable magnets

Permanent magnets are essential components in many day-to-day technologies from electric vehicle motors and wind turbine generators to data storage devices. There are only a limited amount of commercially viable magnet materials available and these vary greatly in their performance. Furthermore, the strongest performing magnets often rely on rare-earth elements, which can be difficult to source and raise environmental, economic and geopolitical concerns. The range of commercially viable magnet materials is also limited, leaving important gaps in performance for some applications.

Christopher’s research will develop computational tools to predict and discover new magnetic materials. Combining high-performance computing and machine learning, he will create a new type of model that describes atomic-scale interactions while explicitly accounting for magnetism. Designed to work across different material chemistries, the model will first be tested against known magnetic compounds before being used to screen large numbers of potential materials and identify promising candidates for experimental synthesis and validation.

By accelerating the discovery of new permanent magnets, the research could expand the materials available for future energy technologies. The modelling approach could also support research into other magnetic materials, including structural steels and metal oxides used in battery electrodes.