Multimodal Reservoir Computing Enabled by a Single Ferroelectric Heterojunction With Bio‐Inspired Visual–Olfactory Fusion

ABSTRACT Brain‐inspired computing based on emerging devices offers a route to bypass the von Neumann bottleneck and enables efficient parallel data processing. Bismuth‐based perovskite oxides, which combine ferroelectric, optoelectronic, and semiconducting properties, represent compelling candidates for information‐processing substrates, yet their multifunctional synergies remain underexplored. Here, we demonstrate a dual‐terminal multifunctional sensory synapse based on a single all‐Aurivillius‐phase Bi 2 WO 6 /SrBi 2 Ta 2 O 9 ferroelectric heterojunction that exhibits tunable spatiotemporal dynamics under optical and gaseous stimuli. The interplay among ferroelectric polarization, oxygen‐vacancy migration, and charge trapping/detrapping enables heterosynaptic plasticity with relaxation times scalable via stimulus intensity. We further emulate bio‐inspired visual–olfactory fusion perception and leverage these dynamics to construct a multimodal reservoir computing framework. The system achieves efficient, high‐accuracy image recognition through in‐sensor volatile‐to‐non‐volatile transformation and programmable cross‐modal integration, maintaining robust performance under stochastic noise (up to 50%). This work presents a co‐designed framework that fuses multimodal perceptual intelligence with brain‐inspired computing, offering a scalable pathway toward future adaptive human–machine interaction.