Improved dynamic characteristics of oxide electrolyte-gated transistor for time-delayed reservoir computing

Time-delayed reservoir computing (RC) equipped with prominent superiorities such as easy training and friendly hardware implementation is identified as a high-efficient answer to complex temporal tasks, and thereby draws increasing attention. Oxygen ion-based oxide electrolyte-gated transistor (Ox-E...

Full description

Saved in:
Bibliographic Details
Published in:Applied physics letters 2024-01, Vol.124 (5)
Main Authors: Fang, Renrui, Li, Xufan, Ren, Kuan, Zhang, Woyu, Xu, Han, Wang, Lingfei, Shang, Dashan
Format: Article
Language:English
Subjects:
Computing time
Data processing
Dynamic characteristics
Electrolytes
Handwriting recognition
Hardware
Oxygen ions
Task complexity
Transistors
Waveforms
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Time-delayed reservoir computing (RC) equipped with prominent superiorities such as easy training and friendly hardware implementation is identified as a high-efficient answer to complex temporal tasks, and thereby draws increasing attention. Oxygen ion-based oxide electrolyte-gated transistor (Ox-EGT) with rich ion dynamic characteristics is deemed as a promising candidate for RC. However, it is still a challenge to produce the required dynamic characteristics for RC implementation. Herein, we develop an Ox-EGT with an oxygen vacancy-electron-coupled electric-double-layer at the electrolyte/channel interface to implement time-delayed RC. Effects of oxygen vacancy concentration on the short-term plasticity are investigated, revealing the optimal concentration range of oxygen vacancies for the dynamic characteristics improvement. The underlying physical mechanism is demonstrated by TCAD simulations. Simulations using the waveform classification and handwritten-digit recognition tasks validate the good information processing ability of the Ox-EGT RC system. These results provide a promising approach to exploit Ox-EGT dynamics for large-scale and energy-efficient neuromorphic computing hardware.
ISSN:0003-6951
1077-3118
DOI:10.1063/5.0185402