High-Performance Artificial Synapse Device Based on Cs3Bi2Br9/NiO Heterostructure for Bio-Inspired Neuromorphic Computing.

The development of energy-efficient and biocompatible artificial synapses is essential to advance neuromorphic computing. Bismuth-based perovskites are promising candidates to replace toxic lead-based perovskites in resistive switching devices owing to their exceptional optoelectronic properties, high environmental friendliness, and stability. Here, we present a lead-free Cs3Bi2Br9/NiO heterostructure memristor capable of mimicking biological synaptic functions with exceptional robustness. By engineering a heterostructure with a NiO layer, ion migration in Cs3Bi2Br9 is spatially confined, achieving a resistance switching change rate of less than 7.37% between cycles and enhanced long-term stability in ambient air (60 days). This Cs3Bi2Br9/NiO memristor exhibits excellent stability, impressive memory retention time (>7 × 103 s), durability (>100 cycles), good on/off ratio, and basic synaptic behavior. Furthermore, the training results of 200 data achieved an accuracy rate of 95.46% in the MNIST handwritten digit recognition task, which was superior to traditional analog neural networks. This work not only highlights the significant potential of lead-free perovskites for sustainable neuromorphic hardware but also provides a scalable preparation path for biocompatible electronics.