Evaporated Copper‐Based Perovskite Dynamic Memristors for Reservoir Computing Systems

Ruiheng Wang,He Shao,Jianyu Ming,Wei Yang,Jintao Sun,Benxin Liu,Siqi Wu,Haifeng Ling

DOI: https://doi.org/10.1002/admt.202400838

IF: 6.8

2024-09-21

Advanced Materials Technologies

Abstract:Dynamic memristors based on high‐quality CsCu2I3 films are developed via thermal evaporation. Leveraging the dynamic migration characteristics of halogen ions in CsCu2I3, a reservoir computing system is constructed, achieving a letter recognition accuracy of 98.2% with minimal training, thereby offering a promising pathway for future neural network information processing. Dynamic memristors are considered as the optimal hardware devices for reservoir computing (RC) enabled by their nonlinear conductance variations. This significantly reduces the extensive training workload typically required by traditional neural networks. Lead halide perovskites, with their tunable band structure and active ion migration properties, have emerged as highly promising materials for developing dynamic memristors. However, large‐scale and consistently stable production remains a challenge for perovskite functional films, while lead elements' toxicity and environmental impact also partly restrict their practical device utilization. In this work, lead‐free copper‐based perovskite (i.e., CsCu2I3) films are prepared by thermal evaporation for constructing dynamic memristors. The effective conductivity modulation of CsCu2I3‐based memristor can be utilized in artificial neural networks, achieving a high handwritten digit recognition accuracy of 91.2%. In addition, the RC system is also constructed based on the dynamic behavior of the devices, by which a letter recognition accuracy of 98.2% with simple training is achieved. This technology provides a feasible pathway to construct copper‐based perovskite dynamic memristors for future neural network information processing.

materials science, multidisciplinary