Electrooptic Dual-Mode Synergetic Control of Two-Dimensional Nonvolatile Floating-Gate Transistors for Artificial Synapses.

Floating-gate memories based on two-dimensional (2D) materials show great potential for mimicking synapses in artificial neural networks (ANNs), offering a promising pathway toward hardware neuromorphic computing. However, current implementations predominantly rely on electrical modulation, limiting their applications in future artificial visual systems. Here, we demonstrate an electrooptic dual-mode artificial synapse based on a van der Waals heterostructured MoS2 floating-gate transistor. This device uniquely combines electrically controlled nonvolatile programmability with light-tunable multilevel storage states, enabling versatile synaptic functionalities. Specifically, electrical stimulation induces key synaptic behaviors, including excitatory/inhibitory responses and long-term potentiation/depression, while optical excitation facilitates long-term neuroplasticity through persistent multilevel conductance states. It can also work as an optoelectronic heterosynaptic transistor and perform dual-mode modulation through both electrical and optical inputs. With this characteristic, we demonstrate its application in an ANN visual sensor, which achieves a remarkable recognition accuracy of 91.97% in handwritten digit classification.