Metal-Filamentary Memristor-Based Artificial Neurons for Embodied Neuromorphic Intelligence

Embodied neuromorphic intelligence represents a transformative paradigm for next-generation intelligent systems, with broad implications for robotics, autonomous driving, intelligent manufacturing, and smart infrastructures. Neuromorphic chips enable bioinspired computing with sparse, heterogeneous representations and spatiotemporal dynamics, efficiently integrating perception, computation, and actuation. This requires hardware capable of physically emulating neuronal dynamics with high energy efficiency and real-time responsiveness. Metal-filamentary memristors based on the electrochemical metallization (ECM) mechanism have emerged as a promising materials-level platform for this purpose. The dynamic formation and rupture of conductive filaments (CF) intrinsically produce volatility, threshold switching, and stochastic behavior, providing a physical basis for neuron-like temporal integration and spiking dynamics. This review examines the microscopic physics of representative metal-filamentary (i.e., Ag-filamentary) memristors, their formation kinetics, and modulation mechanisms. Materials and interface engineering strategies for optimizing leakage current, threshold voltage, and switching variability are then discussed. Finally, it highlights recent progress in memristor-based artificial neurons and materials-enabled embodied neuromorphic systems integrating sensing and computation at the device level, including tactile, thermal, visual, olfactory, auditory, wind, and humidity perception. Notably, this work provides materials-oriented insights and design guidelines for developing efficient, adaptive, and multimodal neuromorphic hardware toward embodied intelligence.