Tri‐Phase in‐Material Reconfigurable Neuromorphic Devices for Extreme Environmental Adaptive Perception

ABSTRACT Achieving multimodal reconfiguration within a single neuromorphic device is pivotal for low‐power, highly parallel brain‐inspired computing. Conventional reconfigurable memristors rely on conductive filaments, which are stochastic and thermally unstable, limiting device reliability. Here, we report an intrinsically reconfigurable neuromorphic device based on a thermally driven tri‐phase‐interconverted supramolecular liquid crystal (SLC). The device reversibly switches among three distinct molecular phases‐superlattice (SL), lamellar quadruple (LQ), and isotropy (ISO). The SL phase enables linearly programmable conductance modulation, while the LQ phase exhibits bistability within a narrow switching window, achieving excellent cycle‐to‐cycle (>99%) and device‐to‐device (>95%) uniformity, along with stable endurance over 105 cycles at 435 K. The device implements self‐recovery through thermally‐induced molecular re‐organization (>95% consistency over 104 cycles), manifesting robust intrinsic reconfigurability. Furthermore, an environmentally adaptive neuromorphic system is constructed for complex perception and classification tasks in wide‐temperature‐range extreme environments. This three‐states‐in‐one‐device paradigm offers a transformative platform for highly‐efficient, multifunctional, and temperature‐adaptive neuromorphic hardware integrating memory, computation, and encryption.