Enhanced synaptic characteristics of electrolyte-gated oxide transistors enabled by optimizing interface states at the oxide semiconductor/electrolyte interface

Abstract

Advanced bionic neuromorphic systems hold great promise for e-skin applications due to their efficient parallel processing and sensory capabilities. Metal oxide semiconductor-based electrolyte-gated transistors (MOS-EGTs) offer a low-power platform for artificial neural networks. However, the impact of channel/electrolyte interface traps on synaptic plasticity remains unclear due to the complexity and lack of direct observation methods. In this study, we developed a simplified model to uncover the intrinsic correlation between surface states of oxide semiconductors and the synaptic plasticity of electrolyte-gated transistors (EGTs) by selectively modulating the surface defect state density of the oxide semiconductors. The ITZO EGT with low-density interface traps exhibited enhanced synaptic characteristics, including excitatory post-synaptic current (EPSC), paired-pulse facilitation (PPF), and long-term memory (LTM). Furthermore, neural network simulations using this device achieved a high recognition accuracy of ∼90%. This study deepens the understanding of interface trap effects and provides a feasible approach for constructing high-efficiency, low-power MOS-based artificial neural networks.