Ultra-low-power lead-free perovskite optoelectronic synapse for artificial visual systems

Bio-inspired artificial visual neural systems have attracted significant interest, yet they often face challenges related to complex fabrication and high energy consumption. This work presents a low-power photonic synaptic device based on the lead-free all-inorganic perovskite Ag₂BiI₅, which is prepared by a simple spin-coating method with an interdigital electrode structure. The resulting device operates at an ultra-low voltage of 0.1 V with an energy consumption of just 1.43 pJ/pulse. It successfully emulates fundamental synaptic functions, including short-term plasticity (STP), long-term plasticity (LTP), paired-pulse facilitation (PPF), and paired-pulse depression (PPD). By modulating light pulses, transitions from short-term to long-term memory are achieved, and behaviors such as Pavlovian conditioning and image memory enhancement are demonstrated. Furthermore, in a 5 × 5 synaptic array, the device mimics human visual memory processes, where memory strength is tuned by the number of training pulses. For application validation, a convolutional neural network based on this device achieves 90.6% accuracy on the Fashion-MNIST dataset, outperforming the electrical approach (87.2%). This study provides an eco-friendly material platform for low-power neuromorphic vision systems and contributes to advances in brain-inspired computing and image processing.