Engineered surface defects enabling polarization-sensitive optoelectronic synapse in a microwire for glare-resistant neuromorphic vision

Conventional vision sensors suffer from inherent limitations, which have driven the development of neuromorphic systems that integrate sensing and processing functionalities to achieve superior energy efficiency and enhanced operational speed. However, achieving robust performance under complex illumination conditions remains a critical challenge that compromises reliability in safety-critical applications, particularly when subjected to glare induced saturation and contrast degradation. Herein, inspired by biological polarization vision, a monolithic polarization-sensitive optoelectronic synapse based on Ga₂O₃:In microwire is developed to overcome glare interference in neuromorphic devices. The intrinsic anisotropy enables polarization-dependent photoresponse, while post-annealing-induced oxygen vacancies facilitate synaptic plasticity. Light polarization directly controls synaptic weight variation within a single device, enabling bio-inspired encoding where polarization and light stimulation parameters jointly modulate plasticity without external polarizers. This co-integration enables polarization-photocurrent-defined convolutional kernels to outperform software-based counterparts in edge extraction. Under 50% noise interference, the polarization-photocurrent kernels maintain effective edge detection capabilities, whereas the software-defined kernels completely fail. Most significantly, this integrated sensing-processing design achieves high-precision computation, delivering 100% accuracy in facial feature extraction under glare interference. The training time is reduced by 75% compared to a software-based approach. This work establishes a pathway for glare-resistant neuromorphic vision by exploiting the innate properties of anisotropic semiconductors.