Ferroelectric Tunnel Junction Enables Superior Neuro-Inspired Computing
Conventional von Neumann computing systems have become obsolete with the rapid development of neuro-inspired computing. Memristor-based synaptic devices, which emulate biological synapses, are considered promising for realizing efficient neurology-inspired computing. However, previously developed memristors either suffered from high power consumption or instability. Ferroelectric Tunnel Junction (FTJ) is a new candidate for constructing memristors due to its stable data storage function, but it does not meet desirable requirements in terms of endurance, power consumption, linearity, etc. .
In a recent book published in Nature Communicationa team led by Prof. Li Xiaoguang and Prof. Yin Yuewei from the University of Science and Technology of China (USTC) of the Chinese Academy of Sciences, has developed a novel FTJ synapse based on Ag/PbZr0.52You0.48O3 (PZT, oriented (111))/Nb:SrTiO3.
The team extensively investigated the properties of the newly developed FTJ synapse. Under low voltage and near dynamic random-access memory (DRAM) operating speed, the FTJ sample displays 256 conductance states with satisfactory linearity and stability. ON/OFF ratio was as high as 200, and stamina up to 109 was also achieved. Even when a voltage pulse close to CPU speed was applied, the sample still exhibited 150 conductance states and little cycle-to-cycle variation.
In order to study the performance of the FTJ synapse under real-world circumstances, the team performed convolutional neural network simulations based on the test result of the FTJ sample. The goal of the simulation was to recognize fashion product images in the F-MNIST dataset, and a high recognition of 94.7% based on 256 states was obtained. Performance was comparable to that achieved with floating-point software.
Noisy images, which are common nowadays, have brought great difficulties in image recognition. Thus, the team then performed a simulation on noisy images with lamella-and-pepper noise or Gaussian noise and the recognition accuracy remained high, demonstrating the reliability of the newly developed memristor based on the FTJ synapse.
These results proved that (111)-oriented FTJs hold promise for neuro-inspired computing.
Neuromorphic computing with memristors
Zhen Luo et al, Subnanosecond Pulse Linear and High Accuracy Weight Updates in Ferroelectric Tunnel Junction for Neuro-Inspired Computing, Nature Communication (2022). DOI: 10.1038/s41467-022-28303-x
Provided by University of Science and Technology of China
Quote: Ferroelectric Tunnel Junction Enables Superior Neuro-Inspired Computing (2022, Feb 25) Retrieved Feb 26, 2022 from https://techxplore.com/news/2022-02-ferroelectric-tunnel-junction-enables-superior.html
This document is subject to copyright. Except for fair use for purposes of private study or research, no part may be reproduced without written permission. The content is provided for information only.