Hash-Environment-Stable Organic Heterojunction Memristor Enabling Reliable Neuromorphic Computing and Hardware-Aware Neural Network Implementation

Abstract

As Moore’s law approaches its physical and technological limits, neuromorphic computing has emerged as a promising paradigm for energy-efficient and adaptive information processing. However, memristive devices capable of stable analog weight modulation under extreme environmental conditions remain highly desirable yet challenging. Here, we report an extreme-environment-stable organic heterojunction memristor based on n-type hexadecafluoro copper phthalocyanine (F16CuPc) and p-type conjugated polymer poly[2-methoxy-5-(3′,7′-dimethyloctyloxy)-1,4-phenylenevinylene] (MDMO-PPV). The heterojunction architecture enables continuous and reversible conductance modulation governed by a synergistic interplay between interfacial tunneling and trap-modulated space-charge-limited conduction. Remarkably, the device maintains stable memristive characteristics after prolonged exposure to high humidity (70% RH for over three weeks) and across an exceptionally wide temperature range from 50 to 300°C without encapsulation, demonstrating outstanding environmental robustness. The device successfully emulates essential synaptic plasticity functions, including excitatory postsynaptic current, paired-pulse plasticity, and long-term potentiation, enabling reliable analog weight modulation. Furthermore, hardware-aware neural network simulations incorporating experimentally measured conductance states demonstrate stable system-level computing performance. A custom 10-layer convolutional neural network achieves a high classification accuracy of 93.22% on the Fashion-MNIST dataset even when constrained by conductance states measured at 300°C. These results establish a direct link between material-level robustness, device-level synaptic functionality, and system-level neuromorphic computing capability, highlighting the strong potential of organic heterojunction memristors for reliable neuromorphic hardware operating in harsh environments.