A van der Waals Floating-Gate Platform for Broadband Optoelectronic Neuromorphic Computing

Overview & Synopsis

Artificial synapses that integrate nonvolatile memory with broadband optoelectronic learning are essential for next-generation neuromorphic computing. However, their realization remains challenging because conventional floating-gate memories exhibit limited optical functionality and poor scalability. Here, we report a van der Waals ReS2/hBN/InSe floating-gate synaptic transistor that combines efficient charge storage with broadband optoelectronic learning in a single device architecture. The heterostructure incorporates InSe as a floating-gate layer for stable charge storage, hexagonal boron nitride (hBN) as a tunneling dielectric, and ReS2 as a photoresponsive channel, enabling carrier tunneling under electrical and optical stimulation. Through the combined operation of these layers, the device achieves a large memory window, an on/off current ratio exceeding 10^6, stable multilevel conductance states, and reliable retention over 10^4 s. Broadband visible-to-near-infrared optical excitation enables controllable synaptic weight modulation and the emulation of essential synaptic functions, including short- and long-term plasticity, paired-pulse facilitation, and pulse-amplitude- and pulse-duration-dependent plasticity. Optical potentiation and electrical depression further enable bidirectional modulation of synaptic weights. Beyond these synaptic functions, the device demonstrates higher-order learning through associative and discrimination conditioning inspired by the Pavlovian learning paradigm, as well as reconfigurable Boolean logic driven by electro-optical inputs. These results establish a multifunctional van der Waals floating-gate platform that integrates nonvolatile memory, broadband optoelectronic learning, and adaptive information processing within a single transistor, providing a promising strategy for compact, energy-efficient neuromorphic hardware.