Spin-orbit torque (SOT) devices are redefining the landscape of neuromorphic computing by enabling adaptive and efficient information processing. We introduce the Néel tensor torque, a new symmetry-breaking mechanism that captures hidden spin correlations in polycrystalline antiferromagnets (AFMs) [1]. This framework enables field-free SOT switching in FM/AFM systems and, remarkably, allows the AFM state to be “trained” and memorized—an unprecedented feature that mirrors synaptic plasticity.
Complementing this advance, we demonstrate a PtMn/(Co/Pd)4/Ta multilayer device that unifies binary and multilevel SOT switching within a single geometry [2]. Binary switching provides digital reliability, while multilevel control supports analog weighting for synaptic functions, together achieving high recognition accuracy on benchmark tasks.
These breakthroughs highlight how SOT devices can evolve beyond conventional memory elements into multifunctional building blocks for neuromorphic computing, combining efficiency, adaptability, and scalability.

1. Chao-Yao Yang, Sheng-Huai Chen, Chih-Hsiang Tseng, Hsiu-Hau Lin, Chih-Huang Lai*, “Néel Tensor Torque in Polycrystalline Antiferromagnets”, Adv. Mater. (2025), doi/10.1002/adma.202506462.
2. Abhijeet Ranjan, Tamkeen Farooq, Chong-Chi Chi, Hsin-Ya Sung, Rudis Ismael Salinas Padilla, Po-Hung Lin, Wen-Wei Wu, Ming-Yen Lu, Rahul Mishra, Chih-Huang Lai* “Dual SOT Switching Modes in a Single Device Geometry for Neuromorphic Computing” Nano Letters, 25 (2025)

​​
More information :
​https://www.spintec.fr/seminar-spin-orbit-torque-devices-and-their-applications-for-neuromorphic-computing/​