A Digital-Analog Memristor Based on Ru-Polymer Complex for High-Performance, Noise-Resistant AI Applications

Abstract

This study introduces a novel organic memristor based on a blend of ruthenium trichloride (RuCl3) and polyvinylpyridinepolystyrene copolymer (PVPS), designed for neuromorphic computing applications. The device operates on a redox-based mechanism, circumventing the challenges associated with conductive filament formation, and displays both digital and analog switching behaviors. This dual functionality enables rapid binary switching, suitable for digital applications, as well as fine-tuned conductance modulation, mimicking synaptic plasticity in neuromorphic systems. Key features of the device include nonvolatile, a high yield rate of 76%, and a large ION/IOFF ratio of 10 2 , making it ideal for information storage and neuromorphic applications. Furthermore, the memristor demonstrates high robustness in noisy environments, with a recognition accuracy of up to 92.72% in a facial recognition task under 40% Gaussian noise. These results suggest that the Ru-PVPS-based memristor is a promising candidate for AI hardware, offering energy-efficient, scalable, and biologicallyinspired computing solutions.