Data-driven machine-learning clustering analysis to automatically classify exfoliated graphene flakes from optical microscope images

Machine-learning techniques enable the recognition of a wide range of images to complement the human intelligence. Since the advent of exfoliated graphene on SiO₂/Si substrates, the identification process of graphene has relied on the optical microscopy imaging. Here, we develop the data-driven clustering analysis method to automatically identify the positions, shapes, and thickness of graphene flakes from the large amount of the optical microscope images of exfoliated graphene on SiO₂/Si. Application of the feature extraction algorithm to the images yielded the optical and morphology feature values of the regions surrounded by the edges. The feature values formed the discrete clusters in the optical feature space, which were derived from 1-, 2-, 3-, and 4-layer graphene. The cluster centers are detected by the unsupervised machine-learning algorithm, enabling highly accurate classification of monolayer, bilayer, and trilayer graphene. The analysis can be applied to substrates with differing SiO₂thickness, which demonstrates the generality of the approach.