Physics-informed computer vision assisted defects classifier in transition metal dichalcogenides using electrical atomic force microscopy

Layered, atomically thin two-dimensional (2D) semiconductors, such as transition metal dichalcogenides (TMDs), have gained significant interest in logic, memory, and photonic applications. In each of these areas, the optimal device performance can be obtained only with the tuning of TMDs’ electronic, structural, and optical properties. However, while TMD growth methods offer a high degree of tunability for the material quality, limited alternatives exist for a rapid screening of material quality when focusing on the electrical properties. To achieve an efficient quality classification for both material selection and optimization, we introduce a combination of computer vision algorithms applicable to electrical atomic force microscopy (AFM). In this article, we introduce a rapid first-pass screening tool that automates the extraction of crucial material parameters, including coverage, defect density, extended defects, and others, by an automated physics-informed analysis flow based on AFM-generated current maps. Presented results offer a pathway to automate the analysis of 2D TMDs, with emphasis on the various electrically active imperfections for the rapid screening of growth conditions in technologically relevant applications.