Work

Flexible Organic Memristor for Robust Physical Reservoir Computing in Diversified Dynamic Tasks

Year: 2025

Type: article

Abstract: ABSTRACT Flexible memristor‐based physical reservoir computing (PRC) holds substantial promise for efficient spatiotemporal information processing. However, the narrow environmental reliability of current PRC systems limits their adaptability and tunability, so the diversity of accessible reservoir states often remains constrained and requires task‐specific and finely tuned control. Here, we present a flexible memristor based on the chloro‐substituted organic small molecule dichloro copper phthalocyanine (Cl 2 CuPc) as an efficient thermally assisted physical reservoir. The device exhibits oxygen‐ion‐driven volatile short‐term synaptic behaviors, in which both electrical pulses and ambient temperature modulate the internal conductance states. By leveraging its strong thermal resilience up to 100°C on flexible substrates, the PRC system achieves 32 distinct reservoir states (5‐bit encoding) under combined control of temperature and pulse parameters. Two complementary encoding schemes, specifically time‐multiplexing and spike‐based preprocessing via pulse amplitude and frequency modulation, enable reliable performance across diversified dynamic tasks. High classification accuracies of 98.4% and 88.6% are obtained for human gait recognition and moving‐object monitoring, respectively. This work demonstrates a lightweight, robust, and multifunctional organic memristor‐based PRC platform, offering a scalable and adaptive solution for analog edge computing in dynamic and complex environments.

Source: Advanced Functional Materials

DOI: 10.1002/adfm.202531718