Mixed-Dimensional Floating Gate Phototransistors for Mixed-Modal In-Sensor Reservoir Computing

Novel neuromorphic devices constructed from low-dimensional materials have demonstrated significant potential in visual perception and information processing. Colloidal quantum dots (QDs) exhibit strong light absorption and tunable band gaps, while 2D materials provide smooth interfaces and channels with superior charge carrier mobility. However, the potential of devices utilizing QDs as floating gates and 2D materials as channels remains largely underexplored. Herein, a floating-gate phototransistor based on the mixed-dimensional heterostructure of 0D-CsPbBr₃ QDs and 2D-MoS₂ few layer is introduced. By leveraging the optical advantages of 0D-QDs and the electrical properties of 2D materials, mixed-modal in-sensor reservoir computing (RC) is realized. Upon electrical/optical stimulation, the device demonstrates an on/off ratio of 10⁷, over 7-bit multistates, nonlinear memory decay behavior, and dynamics with tunable time scales. Building upon these characteristics, the device enables mixed-modal RC using mixed-inputs of optical and electrical signals. Furthermore, accurate recognition of endangered species under extreme weather conditions is also demonstrated through audio-visual fusion. This study presents a compelling paradigm for utilizing the properties of different-dimensional materials to achieve mixed-modal information fusion and opens new pathways for mimicking biological multisensory fusion.