Machine-Learning for Optical Identification of Two-Dimensional Structure

Two-dimensional (2D) materials and heterojunctions, with fascinating properties and abundant applications, have attracted numerous interest and triggered revolutions of corresponding device applications. However, facile methods to realize accurate and intelligent characterizations of these 2D structures are still highly desired. Here, we report the successful application of machine-learning strategy in the optical identification of 2D structures, including graphene, molybdenum disulfide and heterojunctions of these two materials. The machine-learning optical identification method (MOI method) relies on trainable and automatic identifications of the RGB information in the optical photograph of 2D structures. The MOI method enables accurate characterizations of 2D structures, including identifications of the thickness, the existence of impurities, and even the stacking order. Together with the progress in optical techniques, this intelligent identification method with significantly high accuracy and high throughput can certainly promote the fundamental research and device application of 2D structures.