Spin–Charge\nConversion-Based Artificial Synaptic\nDevice for Neuromorphic Computing

A synaptic\nfunction is demonstrated using spin–charge conversion\nin a Rashba system. In an asymmetric quantum well channel, fast-moving\ncharges induce a Rashba effective magnetic field, which separates\nspin-up and spin-down potentials. The ferromagnet detects these spin-dependent\npotentials, corresponding to the spin information on the channel.\nThe multiple ferromagnetic electrodes, each with different switching\nfields, probe their respective spin potentials, and the output terminal\nreads the superposition of the detected potentials, thereby realizing\nmultiple voltage states. These multiple states are systematically\nmodulated and changed to any desired state directly, enabling both\nthe potentiation and depression of synaptic behavior. In this memristive\nfunction device, both charge-to-spin and spin-to-charge conversions\nare demonstrated in a single device, consistent with the reciprocal\nrelation. Additionally, neuromorphic pattern recognition is clearly\ndemonstrated by controlling the <i>V</i><sub>max</sub>/<i>V</i><sub>min</sub> ratio and offset resistance.