Promising guanidinium layered perovskite photoelectric synaptic transistors for neuromorphic computing and image recognition

Abstract

With the rapid development of bio-inspired hardware and artificial neural networks, synaptic transistors have emerged as core components in neuromorphic computing systems. Two-dimensional halide perovskites exhibit significant potential for use in photoelectric synaptic devices owing to their superior photoelectric response characteristics and structural stability. In this study, we propose a two-dimensional (GA)(MA)₅Pb₅I₁₆ perovskite-based photoelectric synaptic transistor. The device exhibits basic synaptic functions under different light conditions, including excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF), short-term plasticity (STP), and long-term plasticity (LTP), demonstrating inherent learning ability. In addition, it accurately recognizes letters in a 5 × 5 synaptic array and maintains a memory time of over 150 s under a light intensity of 60 mW cm⁻². Finally, using the unique characteristics of photonic potentiation and electric depression, a multi-layer convolutional neural network is constructed to recognize stained blood cell images with an accuracy of 91.55%. This study provides valuable insights for developing neuromorphic systems and advancing artificial intelligence based on two-dimensional perovskite photoelectric synaptic transistors.