Defect-engineering induced sub-bandgap stimulation in BeZnO for optoelectronic synapse

Abstract

Optoelectronic synaptic devices offer a promising way to overcome the traditional von Neumann bottleneck, while most studies focused on above-bandgap photons but overlooked sub-bandgap stimulation. In this work, a wide-bandgap Be-doped ZnO (BeZnO) thin film was fabricated to introduce abundant defect levels into the forbidden gap. The BeZnO device exhibits a robust broadband response spanning from the ultraviolet band to the entire visible range, which is essential for an integrated neuromorphic vision system. The key biological synaptic behaviors including excitatory postsynaptic current, paired-pulse facilitation, and the transition from short-term plasticity to long-term plasticity were successfully emulated under optical stimuli by violet, green, and red lights. Furthermore, a VGG-8 convolutional neural network was trained using the below-bandgap stimulated synaptic weights, achieving a high image recognition accuracy of 90.41% on the CIFAR-10 dataset. These results underscore the potential of BeZnO-based optoelectronic synapses for next-generation energy-efficient neuromorphic computing and the broadband artificial visual systems.