In₂O₃ Nanofiber Neuromorphic Transistors for Reservoir Computing

Abstract

In this letter, we propose neuromorphic transistors employing indium oxide (In2O3) nanofibers as the channel layers. Basic synaptic function, such as short-term memory can be emulated by one nanofiber neuromorphic transistor. Nonlinear synaptic function and short-term memory characteristic of such neuromorphic transistors are favorable for reservoir computing (RC) system with high energy-efficiency. Ultra-low energy consumption (15 pJ per reservoir state) and ultra-high accuracy (100%) of speech digital recognition are realized based on such nanofiber neuromorphic transistors, proving a great potential of the RC system for intelligent processing tasks.