Modulation of nonlinearity and asymmetry in a spin–orbit torque driven artificial synapse

• DOI: 10.1063/5.0248325
• Publication Date: 2025-02-25T11:02:48Z
• Authors: Arun Jacob Mathew; John Rex Mohan; Chisato Yamanaka; Kazuki Shintaku; Mojtaba Mohammadi; Hiroyuki Awano; Hironori Asada; Yasuhiro Fukuma
• Keywords: Modulation (music); Orbit (dynamics)

Unconventional computing schemes inspired by biological neural networks are being explored with ever growing interest to eventually replace traditional von Neumann architecture-based computation. Realization of such necessitates the development device analogs neurons and synapses. Particularly, in spin-based artificial synapses, spin–orbit torque (SOT) can be utilized for changing between multiple resistance states synapse. In this work, we demonstrate synaptic behavior, namely long-term potentiation depression a ferrimagnet (GdFe) via SOT generated using heavy metal (Pt). The dependence synapse-like output on input parameters is extensively investigated. Synaptic arrays based experimental results simulated used perform classification handwritten digit dataset. Correlating accuracy experimentally observed performance synapse found depend critical switching currents. Understanding correlation could accelerate spintronic synapses possessing high operation speed, nonvolatility plasticity, thereby enabling efficient compute in-memory systems near future.