Rapid identification of two-dimensional materials via machine learning assisted optic microscopy

A combination of Fresnel law and machine learning method is proposed to identify the layer counts of 2D materials. Three indexes, which are optical contrast, red-green-blue, total color difference, are presented to illustrate and simulate the visibility of 2D materials on Si/SiO2 substrate, and the machine learning algorithms, which are k-mean clustering and k-nearest neighbors, are employed to obtain thickness database of 2D material and test the optical images of 2D materials via red-green-blue index. The results show that this method can provide fast, accurate and large-area property of 2D material. With the combination of artificial intelligence and nanoscience, this machine learning assisted method eases the workload and promotes fundamental research of 2D materials. [Display omitted] •Three classic indexes are employed to illustrate the visibility of 2D materials and to select the optimum Si/SiO2 substrate for different 2D materials (graphene and MoS2).•The k-means clustering and k-nearest neighbors algorithm of machine learning are applied to analysis the optical image of 2D materials.•Attempt to combine the traditional analysis, which is three classic indexes of visibility of 2D materials, with modern technique, which is machine learning, in identification of 2D materials.