Researchers of Tohoku University in Japan have succeeded in creating a device that allows electrical mutual control of both ferromagnets and antiferromagnets. The heterostructure, comprised of a non-collinear antiferromagnet (Mn3Sn), a non-magnetic material (Mo) and a ferromagnet (CoFeB), can switch magnetic states efficiently as well as store and process information with very few energy needs, making it a promising basis for energy-efficient artificial intelligence hardware.
The Tohoku research team used the non-collinear antiferromagnet Mn3Sn as an active magnetic medium, capable of generating a spin current when an electric current is applied. This spin current induces magnetization switching in the CoFeB ferromagnet via the magnetic spin Hall effect. This interaction is reciprocal: CoFeB not only responds to the spin-polarized current but also modulates the magnetic state of Mn3Sn enabling bidirectional switching between the two layers of magnetically ordered materials. This switching mechanism where the polarity of the current can change the sign of the written information, could find unique application in neuromorphic computing based on memristive behaviors.
While conventional devices operate with a fixed binary switching, the team of Shunsuke Fukami achieved electrically programmable switching for multiple magnetic states: “By realizing the electrical mutual switching between a non-collinear antiferromagnet and a ferromagnet, we have opened new possibilities for current-programmable neural networks” said Fukami.
More information: New AI function on the horizon thanks to elec | EurekAlert!
Original article: https://doi.org/10.1038/s41467-025-56157-6