Machine learning to learn new technologies
Getty
This week I attended the 2023 IEEE Intermag Conference in Sendai, Japan. This is a conference organized by the IEEE Magnetics Society (my first IEEE Society and a member for 45 years). I was invited as the next president of IEEE. The conference had over 1,700 physical and virtual attendees, nearly 1,500 of whom attended the conference in person. I believe this is the largest magnetics conference since the start of the COVID-19 pandemic in 2020.
I attended a session where a paper on the application of artificial intelligence to magnetic materials research was published. This is just one example of the debate taking place in the scientific and engineering community about how people can effectively use new AI tools to advance and aid our understanding of the physical world and its real-world applications. These include making better magnetic memory devices, more efficient motors, and many other practical activities.
The session was also joined by Mingda Li from MIT, who said, “Data fitting is one of the many applications that can benefit from machine learning. Another is exploring hidden data, or structure and property relationships.” For this latter application, the papers in this session made use of a large materials database. Mingda in this paper he mentions a database of 146,000 materials.
Yuya Iwasaki of the National Institute for Materials Science in Tsukuba, Ibaraki Prefecture, uses an autonomous materials search system that combines machine learning and non-empirical calculations to find multi-element alloys with magnetization higher than Fe. found the composition.3Co (the material at the peak of the Slater-Pauling curve). The image below shows the results of this materials search over nine weeks, gradually finding ways to increase the intrinsic magnetization of the modeled alloy.
This study showed that adding a small amount of Ir and a small amount of Pt could increase the magnetization of iron-cobalt alloys. After making and measuring several physical iron-cobalt-iridium and iron-cobalt-platinum alloys, it was found that about 4% Ir actually increased the magnetization of FeCo alloys. Similarly, a small amount of Pt in the FeCo alloy also increased the magnetization. Although the alloy composition has higher magnetization than Fe,3It has been discovered before, but this study provides an example of how AI can be used as a tool for new material discovery.
Max Planck Institute for Solid State Chemistry and Physics and Claudia Felser from Spain, USA, China and others discussed using AI techniques to develop new materials, so-called topological magnetic materials. They exploit chiral electronic states on the bulk, surface and edges of solid objects. In physics, a chiral phenomenon refers to a phenomenon that is not identical to its mirror image. The electron spin gives the electron chirality. She showed how materials with very high anomalous Hall effects and large anomalous nearest neighbor effects can be identified. An interesting element of this work relates to gravitational interactions in light matter interactions with magnetotopological matter. Perhaps these phenomena could offer new ways to detect and understand gravity?
Masafumi Shirai and colleagues at Tohoku University used a large database of magnetic properties of so-called Heusler alloys that interact with the MgO tunnel layer of magnetic tunnel junctions (MTJs). Using machine learning and this database, they were able to determine the Curie temperature (the temperature at which the magnetization becomes zero) and the so-called exchange stiffness (the exchange stiffness is the strength of the so-called exchange interaction between the alloys) of the four component alloys. ) was able to be predicted. adjacent magnetic spins) he exists at the interface with MgO. Note that MTJ is used as a read sensor in hard disk drives and magnetic tape heads, and as a commonly used magnetic sensor.
The final paper of this session, presented by Alexander Kovacs with co-authors from Austria and Japan, uses a combination of machine learning and finite element analysis of crystalline grains in permanent magnetic materials to develop more efficient motors. and reduce the use of rare earths in things like windmills. . They used machine learning models developed through the assimilation of data from experiments and simulations to optimize the magnet’s chemical composition and microstructure. These show how high performance He Nd lean magnets can be created using machine learning techniques.
Machine learning is increasingly being used in the development of new materials, such as magnetic materials used in digital storage. Various approaches are possible, but by using databases of known materials, these models can predict the properties of new materials, create combinations and evaluate them virtually much faster than humans can. increase. Although not foolproof, these approaches can accelerate scientific and engineering discovery.
follow me twitter Or LinkedIn. check out my website.
