Abstract:Artificial neural networks (ANNs), inspired by the structure and function of the human brain, have achieved remarkable success in various fields. However, ANNs implemented using conventional complementary metal oxide semiconductor technology face significant limitations. This has prompted exploration of nonvolatile memory technologies as potential solutions to overcome these limitations by integrating storage and computation within a single device. These emerging technologies can retain resistance values without power, allowing them to serve as analog weights in ANNs, mimicking the behavior of biological synapses. While promising, these nonvolatile devices often exhibit inherent nonlinear relationships between resistance and applied voltage, complicating training processes and potentially impacting learning accuracy. This article proposes a magnetic synapse device based on the spin–orbit torque effect with geometrically controlled linear and nonlinear response characteristics. The device consists of a magnetic multilayer stack patterned into a designed shape, where the width variation along the current flow direction allows for controllable magnetic domain wall propagation. Through finite element method simulations and experimental studies, we demonstrate that by engineering the device geometry, a linear relationship between the applied current and the resulting Hall resistance can be achieved, mimicking the desired linear weight-input behavior in artificial neural networks. Additionally, this study explores the influence of current pulse width on the response curves, revealing a deviation from linearity at longer pulse durations. The geometric tunability of the magnetic synapse device offers a promising approach for realizing reliable and energy-efficient neuromorphic computing architectures.

Radiation-Immune Spintronic Binary Synapse and Neuron for Process-in-Memory Architecture

A Multiplexer-Based High-Capacity Spintronic Synapse

Reconfigurable spintronic logic gate utilizing precessional magnetization switching

Optoelectronic synapses based on hot-electron-induced chemical processes.

Spintronic leaky-integrate-fire spiking neurons with self-reset and winner-takes-all for neuromorphic computing

HXNOR-PBNN: A Scalable and Parallel Spintronics Synaptic Architecture for Probabilistic Binary Neural Networks

The single event upset hardened design of spin transfer torque magnetic random access memory read circuits at 40 nm technology

Enabling architectural innovations using non-volatile memory.

A radiation-hardened hybrid RRAM-based non-volatile latch

Design of SEU Tolerant Read Circuit for Non-Volatile Memories with high reliability

Resistance-Sum Architecture for Voltage-Controlled SOT-MRAM based Computing-in-Memory With Hybrid References

Tunable magnetic synapse for reliable neuromorphic computing

A Novel Computing-in-Memory Platform Based on Hybrid Spintronic/CMOS Memory

Spintronic Neuron Using a Magnetic Tunnel Junction for Low-Power Neuromorphic Computing

High On/Off Ratio Spintronic Multi‐Level Memory Unit for Deep Neural Network

Radiation-Hardened 0.3–0.9-V Voltage-Scalable 14T SRAM and Peripheral Circuit in 28-Nm Technology for Space Applications

Nonvolatile Spintronic Memory Cells for Neural Networks

Spin-Orbit Torque Neuromorphic Fabrics for Low-Leakage Reconfigurable In-Memory Computation

Spin-NeuroMem: A Low-Power Neuromorphic Associative Memory Design Based on Spintronic Devices

Enabling Normally-Off In Situ Computing With a Magneto-Electric FET-Based SRAM Design

Enabling Normally-off In-Situ Computing with a Magneto-Electric FET-based SRAM Design