All-optical modulation self-powered optoelectronic synaptic devices with monochromatic ultraviolet for inhibitory/excitatory synaptic behaviors

Self-powered optoelectronic synaptic devices play an important role for the construction of neuromorphic computing systems. Nevertheless, these devices can only simulate optically excitatory synaptic behaviors, while mimicking optically inhibitory functions remains a challenge, arising from the photo-generated carriers that usually lead to the enhancement of conductivity in these devices. This limitation hinders the application development of the self-powered optoelectronic synaptic devices in neuromorphic computing. In this work, a self-powered optoelectronic synaptic device based on the tin dioxide and cesium silver bismuth bromide heterojunction is developed. Through utilizing the ion migration and photo-generated carrier transport behaviors in the heterojunction, self-powered optically inhibitory and excitatory synaptic behaviors are successfully mimicked under ultraviolet (365 nm) monochromatic all-optical modulation. On this basis, a convolutional neural network has been constructed with the software-based means to recognize images (the absence of real hardware-level image processing) in fashion MNIST dataset and an accuracy of 84.48% is obtained in this network, indicating broad application prospects of the all-optical self-powered systems in the neuromorphic computing.