Integrated Nonvolatile Memory and Energy‐Efficient Multifunctional Neuromorphic Simulation Based on ReS ₂ /h‐BN/WTe ₂ Heterostructure

ABSTRACT

Emerging neuromorphic computing, which emulates the parallel operation of neurons and synapses in the human brain, overcomes the high‐power consumption and latency limitations of traditional Von Neumann architecture. In this context, three‐terminal floating‐gate transistors (FGT) based on two‐dimensional materials have emerged as ideal candidates for achieving integrated sensing, memory, and computation due to their high bandwidth, low crosstalk, and multi‐level storage capabilities. Herein, we report a multifunctional FGT based on a ReS ₂ /h‐BN/WTe ₂ van der Waals heterostructure, which integrates non‐volatile memory, synaptic properties, and reconfigurable logic functions altogether in a single device. Leveraging the excellent optoelectronic properties of ReS ₂ and WTe ₂ , the floating‐gate architecture provides a large memory window of 108.42 V (±60 V sweep range), a long retention time of over 10 ⁴ s, and excellent stable endurance above 1000 cycles. Besides accurate emulation of various synaptic behaviors, the device also realizes visual perception, handwritten digit recognition (accuracy: 91.49%), and reconfigurable Boolean logic gate functionalities. Moreover, its synaptic weight modulation enables over 300 distinct states (9 bits) with a low electrical energy consumption of 1.5 pJ per spike. This work demonstrates a highly integrated optoelectronic platform capable of simultaneously performing sensing, processing, and memory, offering strong potential for building efficient, low‐energy artificial neuromorphic systems.