Toward Image Processing and Reservoir Computing: Neuromorphic Visual System Enabled by Nanotube‐Confined Tellurium

Abstract The human visual system integrates perception, memory, and processing in a compact, energy‐efficient architecture, inspiring optoelectronic neuromorphic devices. However, most existing systems rely on external gate voltages and complex circuitry, hindering integration and energy efficiency. Moreover, using multiple heterogeneous materials compromises stability and complicates mechanism analysis. Here, a two‐terminal optoelectronic resistive random‐access memory (ORRAM) system based on a tellurium/ carbon nanotube (Te@CNT) heterostructure is presented that integrates optical sensing, information storage, and image preprocessing. Strong interfacial binding and band bending enable nonvolatile optical storage at 0.1 V bias with enhanced stability. The device exhibits tunable synaptic dynamics, classifies 16 binary inputs, and maps digital to analog signals nonlinearly. Integrated into neural networks, the ORRAM array improves feature extraction and noise suppression, achieving 99.17% recognition accuracy within only 10 training epochs. This work provides ideas for low‐power and compact neuromorphic vision systems.