Flexible Synaptic Memristors With Controlled Rigidity in Zirconium-Oxo Clusters for High-Precision Neuromorphic Computing

Abstract

Flexible memristors are promising candidates for multifunctional neuromorphic computing applications, overcoming the limitations of conventional computing devices. However, unpredictable switching behavior and poor mechanical stability in conventional memristors present significant challenges to achieving device reliability. Here, a reliable and flexible memristor using zirconium-oxo cluster (Zr₆O₄OH₄(OMc)₁₂) as the resistive switching layer is demonstrated. The optimization of the structural rigidity of the hybrid oxo-cluster network by thermal polymerization allows the precise formation of dispersed conductive cluster networks, enhancing the repeatability of the resistive switching with mechanical flexibility. The optimized memristor exhibits endurance of ∼10⁴ cycles and stable memory retention performance up to 10⁴ s, maintaining a high I_ON/I_OFF ratio of 10⁴ under a bending radius of 2.5 mm. Moreover, the device achieves a pattern recognition accuracy of 97.44%, enabled by highly symmetric analog switching with multilevel conductance states. These results highlight that hybrid metal-oxo clusters can provide novel material design principles for flexible and reliable neuromorphic applications, contributing to the development of artificial neural networks.