Polarization‐Driven Optoelectronic Synapse With Programmable Plasticity for Neuromorphic Visual System

ABSTRACT

Optoelectronic synapses that integrate sensing and memory are vital for energy‐efficient artificial vision. However, current devices generally depend on defect trapping or heterojunction interface barriers, both of which can induce uncontrollable carrier dynamics that constrain paired‐pulse facilitation (PPF), energy efficiency, and the transition between short‐term and long‐term plasticity (STP‐LTP). In this study, we introduce a polar optoelectronic synaptic transistor whose intrinsic out‐of‐plane polarization produces a built‐in electric field that effectively separates photogenerated carriers. This results in a high PPF index of 171% and an exceptionally low energy consumption of 25 fJ per event, rivaling the best trade‐off reported for two‐dimensional (2D) material‐based optoelectronic synapses. Notably, this polarization makes recombination kinetics highly responsive to variations in gate voltage and temperature, offering a means to program synaptic memory duration that is distinct from conventional optical or heterojunction engineering. Leveraging this tunability, we demonstrate visual adaptation analogous to an artificial pupil and robust neural network computing, achieving 91.51% accuracy in recognizing handwritten digits and preserving 83.08% accuracy under 40% noise interference. Our work establishes out‐of‐plane polarization as a unified physical parameter that simultaneously optimizes performance, programmability, and functionality, thus offering a simplified design pathway for adaptive neuromorphic vision hardware.