Artificial Optoelectronic Synapse with Nanolayered GaN/AlN Periodic Structure for Neuromorphic Computing

The rapid development of artificial intelligence requires faster information processing capabilities. However, the traditional von Neumann architecture is limited by transmission speed. The combination of neuromorphic structure and photon calculation will break this limitation and effectively improve the calculation efficiency. Here, we demonstrate an optical synaptic device based on GaN/AlN periodic structure that exhibits strong persistent photoconductivity in ultraviolet (UV) detection, which is caused by the strong polarization of GaN/AlN heterojunction. In our devices, photogenerated carriers are trapped in potential wells and then recombine slowly. This physical process alters the initial distribution of the electric field, thus changing the resistance state of the device. Based on this characteristic, the device exhibits optically tunable synaptic behaviors, such as short/long-term memory, excitatory postsynaptic current, and spike-timing-dependent plasticity stimulated by pulsed light. At the same time, our devices exhibit high response and contrast at high bias voltage, making it sensitive to quite weak UV light. Then, it is demonstrated experimentally that the devices can be applied to image information preprocessing, greatly reducing the amount of input information and subsequent training costs in the task of handwritten number recognition. Overall, this work shows an unreported structure of an optical synaptic device with memory and learning behaviors, with great potential for applications in weak UV light detection and neural computing.