GaN-based 3D-stacked heterogeneously integrated optoelectronic neuromorphic system for artificial visual applications

Traditional artificial vision systems, composed of discrete functional units, are fundamentally limited by the von Neumann bottleneck. Inspired by biological vision, we propose a GaN-based optoelectronic heterogeneously integrated platform that three-dimensionally (3D) stacks an AlGaN/GaN heterojunction optoelectronic synaptic metal-oxide-semiconductor field-effect transistor (MOSFET) with an InGaN/GaN multiple quantum well (MQW) LED. This 3D architecture spatially separates photodetection and emission, effectively suppressing optical crosstalk. Under ultraviolet (UV) light pulse stimulation, the synaptic MOSFET exhibits key synaptic behaviors, including excitatory postsynaptic current, paired-pulse facilitation, and the transition from short-term to long-term plasticity. Notably, by integrating the wavelength-selective photoresponse of the synaptic MOSFET with the threshold-dependent emission characteristics of the MQW LED, UV-induced postsynaptic currents are converted into voltage signals to trigger LED emission, whereas the weak responses induced by green illumination remain below the LED threshold and are effectively filtered. Furthermore, the dynamic relaxation process of trapped charges in the synaptic MOSFET is directly translated into the time-dependent decay of LED emission, enabling the optical visualization of synaptic behaviors. This work establishes a viable route toward an integrated "sensing-memory-filtering-visualization" optoelectronic neuromorphic visual system.