Machine learning driven thickness detection in 2D materials

Abstract

In recent times, two-dimensional (2D) materials have attracted significant attention and revolutionized various electronic device applications due to their heterostructure and unique properties not found in their bulk counterparts. However, the lack of large-scale area characterizing methods of accurate and intelligent detection of 2D nanostructures has hindered the rapid development of 2D materials. Thus, a more time efficient and accurate method is required. In this research, we have implemented the combination of machine learning and optical detection of 2D materials and its nanostructures with accurate predictions of certain MoS2 layers. Machine- learning optical identification (MOI) was established to realize accurate predictions of optical microscope images using RGB information of these 2D materials and their nanostructure. The results have proven that the MOI method is efficient in accurate characterizations of the thickness of large-scale area molybdenum disulfide (MoS2) impurities, as well as the identification of adhesive present during the mechanical exfoliation process in the sample preparation stage. With the successful implementation of AI and nanoscience, machine learning driven thickness detection in 2D materials can certainly drive further fundamental research in the 2D material and its wafer- scale device application space.