A Spin–Orbit Torque Pulley Embedded in an Exchange-Spring-Coupled Antiferromagnet/Ferromagnet Bilayer with Pulse-Dependent Plasticity for Efficient Neuromorphic Computing

Monitoring spin–orbit-torque (SOT) switching of exchange bias (Hex) in FM/AFM bilayers is useful for probing AFMs, whereas Hex mainly reflects interfacial spins and masks depth-dependent order formed under an exchange-spring coupling. In this work, we uncover a nonreported hidden order functioning as a magnetic pulley in sputtered Pt/Co/IrMn trilayers and show how it governs SOT switching. Although Hex is fully reversible, driven by SOT, and features a dynamic interface, reducing the driving stimuli, such as small in-plane fields and short current pulses, it reveals a robust subsurface IrMn order inducing a ratchet-type asymmetry: the analog–digital duality. The polarity of the ratchet can be set and reversed by the initial magnetization–Hex state, consistent with an exchange-spring scenario where progressively stronger bulk IrMn moments act as an embedded pinning backbone, although the interfacial Hex remains mobile. By exploiting pronounced pulse-width-dependent plasticity (10–100 μs), we map the measured SOT switching curves for activation functions with tailorable thresholds and steepness, and deploy them in a convolutional neuronal network architecture. A dynamic-annealing modulation (short→long pulses) attains image-recognition accuracy, approaching that acquired by software ReLU. These results uncover an AFM order beyond Hex and establish a reconfigurable, energy-efficient SOT building block for neuromorphic computing.