Multilevel Optical Programming of Intrinsic Vacancies in Solution-Processed MoS2 Films for Retinomorphic Color Differentiation

Abstract

Solution processing provides a scalable route to assemble 2D building blocks into large-area films with low intersheet resistance, enabling scalable device integration. However, chalcogen vacancies introduced during processing often cause unintentional doping, and passivation strategies relying on strong reagents and controlled atmospheres increase process complexity. Here, rather than eliminating vacancies typically regarded as detrimental, we exploit vacancy-localized states in solution-processed MoS2 as essential defects to realize retinomorphic device arrays in which sensing and memory are co-localized within a two-terminal device. Trapping of photoexcited carriers at these vacancy-localized states enhances photogating and induces persistent photoconductivity, writing retentive conductance states without the need for additional trapping layers, floating gates, or complex heterostructures. The stored state is reversible upon oxygen exposure, which promotes de-trapping and restores the dark baseline. Additionally, wavelength-selective conductance modulation enables multilevel accumulation of color-encoded weights for RGB differentiation. Increasing optical pulse number progressively enlarges inter-color conductance contrast; when integrated with a convolutional neural network framework, the encoded states enable color recognition accuracy up to 94%. By exploiting a defect landscape inherent to solution-processed 2D films, this work establishes a scalable materials platform that simultaneously integrates sensing and memory at the pixel level while reducing processing complexity.