Single-gate reconfigurable multifunctional devices based on anti-ambipolar van der Waals heterojunctions
The escalating demands of artificial intelligence, machine learning, and neural computing necessitate multifunctional optoelectronic devices capable of integrating sensing, memory, and processing. Two-dimensional van der Waals heterostructures (vdWHs) offer unique advantages, yet their practical deployment is hindered by complex architectures and inefficient mode-switching. Here, we demonstrate a MoTe2/SnS2 anti-ambipolar heterojunction device enabling single-gate reconfiguration among frequency doubling, broadband photodetection, and neuromorphic computing. The device exhibits a peak-to-valley ratio (PVR) of 465, ensuring efficient frequency doubling. As a photodetector, it operates across an exceptionally broad spectral range of 520–2200 nm, with outstanding responsivity and detectivity. Furthermore, the device emulates complete synaptic behaviors, including short-term plasticity (STP), long-term plasticity (LTP), and paired-pulse facilitation (PPF). Integrated into a reservoir computing (RC) system trained on a vehicle motion dataset, it achieves a directional recognition accuracy of 98.7%. This work establishes a paradigm for multifunctional integration and low-power neuromorphic computing, advancing next-generation intelligent optoelectronic systems.