Double MgO-Based Perpendicular Magnetic Tunnel Junction for Artificial Neuron

Dong Won Kim,Kei Ashiba,Han Sol Jun,Woo Seok Yi,Jin Young Choi,Jae Joon Kim,Jea Gun Park,Jong Ung Baek

doi:10.3389/fnins.2020.00309

Abstract

A perpendicular spin transfer torque (p-STT)-based neuron was developed for a spiking neural network (SNN). It demonstrated the integration behavior of a typical neuron in an SNN; in particular, the integration behavior corresponding to magnetic resistance change gradually increased with the input spike number. This behavior occurred when the spin electron directions between double Co2Fe6B2 free and pinned layers in the p-STT-based neuron were switched from parallel to antiparallel states. In addition, a neuron circuit for integrate-and-fire operation was proposed. Finally, pattern-recognition simulation was performed for a single-layer SNN.

Highlights

Artificial neural network (ANN)-based artificial intelligence (AI) has been one of the most successful technologies in recent years
Emerging artificial neuron devices have been reported as an alternative to complementary metal oxide semiconductor (CMOS)-based neuron devices such as partially depleted silicon-on-insulator n-MOSFET (PD-SOI n-MOSFET) (Dutta et al, 2017), phase change random-access memory (PCRAM) (Tuma et al, 2016), and magnetic perpendicular spin transfer torque (p-STT) Based Artificial Neuron random-access memory (MRAM) (Grollier et al, 2016; Sengupta et al, 2016; Shim et al, 2017; Srinivasan et al, 2017; Torrejon et al, 2017; Mizrahi et al, 2018; Kurenkov et al, 2019)
MRAM has been proposed as a promising candidate for artificial neuron device due to its high-area efficiency, fast operating speed, and low power consumption (Zhang et al, 2016; Liyanagedera et al, 2017; Hu G. et al, 2018)

Summary

INTRODUCTION

Artificial neural network (ANN)-based artificial intelligence (AI) has been one of the most successful technologies in recent years. Today, it is applied in numerous fields, such as education, security, finance, science, and entertainment. There is a limitation to conventional ANNs working on the von-Neumann architecture. Neuromorphic computing systems that mimic the human brain has been designed to overcome this limitation using complementary metal oxide semiconductor (CMOS)-based artificial neuron devices. MRAM has been proposed as a promising candidate for artificial neuron device due to its high-area efficiency, fast operating speed, and low power consumption (Zhang et al, 2016; Liyanagedera et al, 2017; Hu G. et al, 2018). We conducted a pattern recognition simulation of a spiking neural network (SNN) using our p-STT-based neuron

MATERIALS AND METHODS

RESULTS

DISCUSSION

DATA AVAILABILITY STATEMENT