Researchers from Tohoku University, the National Institute for Materials Science, and the Japan Atomic Energy Agency have created a novel spintronic device that enables the electrical mutual control of non-collinear antiferromagnets and ferromagnets, according to a study published in Nature Communications on February 5th, 2025.
Schematic illustration of (a) a conventional magnetic memory device and (b) the device for electrical mutual switching developed in this work. Image Credit: Shunsuke Fukami
AI’s revolutionary impact is well underway. However, as AI systems advance, so will their power usage. Further developments will necessitate AI chips that can do AI calculations at high energy efficiency.
This is when spintronic devices enter the picture. Their combined memory and computational capabilities are similar to the human brain and can be used as a foundation for low-power AI chips.
This implies the device can efficiently swap magnetic states, storing and processing information with minimal energy, similar to a brain-like AI chip. The discovery can change AI hardware by increasing efficiency and lowering energy costs.
While spintronic research has made significant strides in controlling magnetic order electrically, most existing spintronic devices separate the role of the magnetic material to be controlled and the material providing the driving force.
Shunsuke Fukami, Researcher, Tohoku University
After fabrication, these devices have a set operating scheme and usually switch data in a binary fashion from “0” to “1”. Nonetheless, a significant advancement in electrically programmed switching of different magnetic states is provided by the current study team’s discovery.
The core magnetic material used by Fukami and his associates was the non-collinear antiferromagnet Mn3Sn. Using an electrical current, Mn3Sn produces a spin current that drives the switching of a nearby ferromagnet, CoFeB, via the magnetic spin Hall effect. In addition to reacting to the spin-polarized current, the ferromagnet also affects Mn3Sn’s magnetic state, allowing electrical mutual switching between the two materials.
The team's proof-of-concept experiment showed that the magnetic state of Mn3Sn can be used to electrically control information written to the ferromagnet. They changed the magnetization of CoFeB in various traces representing various states by varying the set current. Similar to how synaptic weights (analog values) work in AI processing, this analog switching mechanism, where the polarity of the current can alter the sign of the information written, is a crucial operation in neural networks.
Fukami added, “This discovery represents an important step toward the development of more energy-efficient AI chips. By realizing the electrical mutual switching between a non-collinear antiferromagnet and a ferromagnet, we have opened new possibilities for current-programmable neural networks. We are now focusing on further reducing operating currents and increasing readout signals, which will be crucial for practical applications in AI chips.”
The team’s study opens the door for enhancing AI chips’ energy efficiency and reducing their negative environmental effects.
Journal Reference:
Joon, J.-Y. et. al. (2025) Electrical mutual switching in a noncollinear-antiferromagnetic–ferromagnetic heterostructure. Nature Communications. doi.org/10.1038/s41467-025-56157-6