This week I was at the 2023 IEEE Intermag Conference in Sendai, Japan. This is a conference organized by the IEEE Magnetics Society (my first IEEE society, member for 45 years). I was invited to attend as IEEE President-Elect. In total there were over 1700 physical and virtual participants with close to 1500 people at the conference in person. I believe this is the biggest magnet conference since the beginning of the Covid 2020 pandemic.

I attended a session that had papers on the application of artificial intelligence to magnetic materials research. This is an example of the discussions taking place in the science and engineering community about how humans can effectively use new AI tools to accelerate and aid our understanding of the physical world and its application to real-world applications. This includes making better magnetic memory devices, more efficient motors, and many other practical activities.

This session included Mingda Li, from MIT, who said that “data fitting is one of many other uses that can benefit from machine learning. Another is a focus on exploring hidden data or building structure-property relationships.” For this second application, the papers in this session used large material databases. Mingda mentions a database of 146,000 materials in this paper.

Y. Iwasaki of the National Institute of Materials Science, Tsukuba, Ibaraki, Japan used an autonomous material search system combining machine learning and ab initio calculation to find multi-element compositions that can find alloy magnetizations larger than Fe₃Co (material at the peak of the Slater-Pauling curve). The image below shows the results of searching this material over a period of 9 weeks, gradually finding ways to increase the internal magnetization of the modeled alloy.

This research showed that adding a little Ir and a little Pt could increase the magnetization of the iron-cobalt alloy. When some physical iron cobalt iridium and iron cobalt platinum were made and measured, it was found that about 4% Ir actually increased the magnetization of the FeCo alloy. Likewise, a little Pt in the FeCo alloy also increased the magnetization. Although alloys with magnetization higher than Fe₃If they have been found before, this research has shown an example of how AI can be used as a tool for new material discoveries.

Claudia Felser and colleagues from the Max Planck Institute for Chemical Physics of Solids, as well as from Spain, the USA and China, talked about using artificial intelligence methods to develop new materials for what are called topological magnetic materials. These exploit the chiral states of electrons on the bulk, surfaces and edges of solid objects. In physics, a chiral phenomenon is one that is not identical to its mirror image. Electron spins give chirality to an electron. She showed how materials with a very high anomalous Hall effect and a large anomalous Nearst effect were identified. An interesting element of this work relates to the interaction of gravity in the interactions of light matter with magnetic topological materials. Perhaps these phenomena could provide new ways to detect and understand gravity?

Masafumi Shirai and co-workers at Tohoku University used a large database of magnetic properties for what are called Heusler alloys interacting with an MgO tunnel layer for magnetic tunnel junctions (MTJs). Using machine learning and this database, they were able to predict the Curie temperature of the four-component alloys (the temperature at which the magnetization goes to zero) and what is called the exchange stiffness (the exchange stiffness represents the strength of what are called exchange interactions between adjacent magnetic spins) at the interface with MgO. Note that MTJs are used as read sensors in hard disk and magnetic tape heads and in the most commonly used magnetic sensors.

The final paper of this session, given by Alexander Kovacs with co-authors from Austria and Japan, discussed the use of machine learning combined with finite element analysis of crystalline grains of permanent magnet material to create more efficient motors and use less rare earths, for example, for windmills. .They optimized the chemical composition and microstructure of the magnet using machine learning models developed by assimilating data from experiments and simulations. They demonstrate how high performance magnets can be created using machine learning methods.

Machine learning is increasingly used in the development of new materials, including magnetic materials used for digital storage. Different approaches can be used, but using databases of known materials these models can predict the properties of new materials, making and virtually evaluating combinations much faster than a human could. Although not infallible, these approaches can accelerate scientific and engineering discoveries.