Spin-Orbit-Torque-Driven Two-Terminal Giant Magnetoresistance Memristive Devices for In-Memory Computing.

The rising complexity of artificial intelligence and data-intensive applications drives the demand for memristive devices to support high-performance and scalable processing-in-memory (PIM) architectures. While three- and four-terminal spintronic PIM solutions show promise, they face scalability challenges due to large footprints. This study demonstrates a two-terminal spin-orbit torque (SOT)-driven giant magnetoresistance (GMR) memristive device utilizing a Pt/Co/Cu/CoTb stack, integrating data storage and logic functions within a single unit. Ten nonvolatile resistance states are achieved by modulating the SOT current amplitude, and additional states with opposite polarities are created by tuning the CoTb alloy composition. Synaptic plasticity, including long-term potentiation and depression, is demonstrated through current pulse modulation. Simulations of a deep neural network constructed with this GMR device achieve 92% accuracy in handwritten digit recognition and image visualization tasks. Moreover, basic Boolean logic functions further showcase the processing capabilities. This two-terminal GMR device offers a scalable solution for advanced memristive memory and in-memory computing.