An extremely low-power-consumption reconfigurable two-dimensional tellurene artificial synapse for bio-inspired wearable edge computing

Abstract

Neuromorphic electronics are gaining significant interest as components of next-generation computing systems. However, it is difficult to develop flexible neuromorphic electronics for implementation in various edge applications such as bio-implantable electronics and neuroprosthetics. In this study, we present a reconfigurable 2D tellurene (Te) artificial synaptic transistor on a flexible substrate for neuromorphic edge computing. Single-crystalline 2D Te flexible synaptic transistors exhibit potentiation and depression modulated by gate pulses with an extremely low power consumption of 9 fJ, 93 effective multilevel states, excellent linearity and symmetry, and an accuracy of 93% in recognizing the Modified National Institute of Standards and Technology (MNIST) patterns. Furthermore, it was observed to be a flexible synaptic transistor with outstanding gate tunability and endurance characteristics, even under a 2% curvature in both the concave and convex states. We believe a robust 2D Te flexible artificial synapse will effectively function as a building block for wearable neuromorphic edge computing applications. We fabricated a reconfigurable two-dimensional tellurene artificial synaptic transistor on a flexible substrate for bio-inspired wearable neuromorphic edge computing, showing an extremely low power consumption of 9 fJ and an impressive accuracy of 93% in recognizing MNIST patterns.