Spintronic devices have emerged as promising candidates for neuromorphic computing, owing to their nonvolatility and plasticity that enable the emulation of synaptic and neuronal functions. Compared with conventional ferromagnets, synthetic antiferromagnets (SAFs) exhibit unique and fascinating properties, including negligible stray field, ultrafast spin dynamics, and chiral interlayer coupling. However, the potential applications of SAF, particularly those with chiral coupling, in neuromorphic computing remain relatively unexplored, which motivates the present experimental study. Here, using a chirally coupled $\mathrm{Co}/\mathrm{Ru}/{\mathrm{Fe}}_{0.60}{\mathrm{Tb}}_{0.40}$ SAF structure with noncollinear magnetization configuration, we demonstrate that linear and nonlinear memristive switching, driven by spin-orbit torque (SOT) in different operating regimes, can effectively emulate biological synapses and neurons, respectively. In particular, current-induced SOT switching in chirally coupled SAFs exhibits stable and pronounced multilevel switching behavior that current pulses can precisely tune. Using experimentally derived synapse and neuron parameters, we further simulate a three-layer fully connected neural network for handwritten digit recognition, achieving an accuracy of over 94%. This work establishes chirally coupled SAF as a highly promising building block for high-performance neuromorphic spintronic devices.

Close

PrevNext

View figure in article

• Neuromorphic computing
• Spin-orbit torque
• Spintronics
• Ferrimagnets
• Magnetic multilayers
• Memristors
• Synthetic antiferromagnetic multilayers
• Transition-metal rare-earth alloys