Defects introduced during the growth process greatly affect the device performance of two-dimensional (2D) materials. Here we demonstrate the applicability of employing machine-learning-based analysis to distinguish the monolayer continuous film and defect areas of molybdenum disulfide (MoS2) using position-dependent information extracted from its Raman spectra. The random forest method can analyze multiple Raman features to identify samples, making up for the problem of not being able to effectively identify by using just one certain variable with high recognition accuracy. Even some dispersed nucleation site defects can be predicted, which would commonly be ignored under an optical microscope because of the lower optical contrast. The successful application for classification and analysis highlights the potential for implementing machine learning to tap the depth of classical methods in 2D materials research.
Keywords: 2D materials; Raman spectrum; machine learning; random forest algorithm.