Visual information enhancement encoder based on exciton regulation

The strong Coulomb response in two-dimensional transition metal dichalcogenides (TMDCs) leads to the emergence of firmly bound excitons and charged excitons, and their transformation is the key to studying and utilizing the optical and photoelectric properties of two-dimensional materials. Especially in the study of optoelectronic applications, they are beneficial in many valley electron and energy transfer applications. However, their potential for new information processing, such as the combination of visual information and neural networks, remains largely untapped. Here, we study the photoluminescence (PL) properties and exciton behavior of monolayer WS₂. We achieve a wide range of exciton-trion modulation by changing the substrate and pump intensity, extending the dynamic range of visual information processing. At room temperature, the I(X^-)/I(X⁰) ratio of CaF₂/WS₂ samples pumped at high intensity is improved by an order of magnitude, and the three-particle conversion rate can reach 90%. Then, we combined the non-linear response of the WS₂/CaF₂ device and the linear response of the WS₂/SiO₂ device to re-encode the data source through the multilayer neural network algorithm, and the recognition accuracy reached 98.6%, demonstrating the potential of the device in advanced vision technology. Our work provides a new path for future 2D exciton device design.