Ultrafast & low-power consumption 2D floating-gate devices for opto-electronic hybrid neural networks

Abstract

Optoelectronic hybrid neural networks combine the advantages of electronical and optical neural networks, enabling next-generation neuromorphic computing systems with nanosecond processing speeds, fJ-level energy efficiency, and wafer-scale integration density. Here, we demonstrate a MoS₂/h-BN/graphene based 2D-material floating gate (FG) transistor exhibiting excellent electrical memory characteristics and dual mode photo-response: both positive (PPC) and negative photoconductance (NPC). Utilizing this device, we experimentally demonstrate a three-layer artificial neural network achieving high image recognition accuracy (97.2 %) with ultrafast operation (30 ns) and ultralow energy consumption (3.2 fJ/event). These results indicate that optoelectronic hybrid neural networks implemented with all-2D FG transistors can achieve energy-efficient and high-speed in-memory sensing and computing, showing promising potential in next-generation neuromorphic computing.