All-Solid-State Li-Ion-Based Electrolyte-Gated Synaptic Transistor with Dual Modes of Excitatory and Inhibitory Plasticity

Abstract

Artificial synapse is an essential element for building artificial neural networks for neuromorphic computing. Both the excitatory and inhibitory responses are considered to be the well-known synaptic plasticity that contributes equally to the biological nervous system. To date, electrolyte-gated transistor (EGT) has been considered as one of the most promising synaptic devices and proved to be enabled to simultaneously emulate both excitatory and inhibitory synaptic plasticity in a single device. However, the complex architectures or non-compatible solution-based process hinder the development of EGT synaptic devices for further integration. Besides, the reported excitatory and inhibitory responses are mainly modulated through dual voltage supply, which may require complex circuit structure. Herein, an inorganic all-solid-state Li-ion-based EGT with LiPON and IGZO as the electrolyte and channel layer, respectively, is demonstrated using compatible all-sputtering method. The dual modes of excitatory and inhibitory plasticity induced by the inherent predominated electrical double layer or electron trapping effect, respectively, can be simultaneously realized in a single device with the modulation of a single positive voltage supply. The device exhibits a low operation voltage of 2 V and a low energy consumption of ~80 pJ for the inhibitory mode, and 4V and ~800 pJ for the excitatory mode, respectively. In addition, Pavlovian experiment and the learning-forgetting-relearning process of human memory are also successfully achieved. This work may provide a good hardware foundation for complex brain-like computing tasks such as pattern recognition, decision making, and associative learning, and shed new light on the development of neuromorphic devices compatible with mass production process.