Machine Learning Determination of the Twist Angle of Bilayer Graphene by Raman Spectroscopy: Implications for van der Waals Heterostructures

With the increasing interest in twisted bilayer graphene (tBLG) of the past years, fast, reliable, and nondestructive methods to precisely determine the twist angle are required. Raman spectroscopy potentially provides such a method, given the large amount of information about the state of the graphene that is encoded in its Raman spectrum. However, changes in the Raman spectra induced by the stacking order can be very subtle, thus making the angle identification tedious. In this work, we propose the use of machine learning (ML) analysis techniques for the automated classification of the Raman spectrum of tBLG into a selected range of twist angles. The ML classification proposed here is low computationally demanding, providing fast and accurate results with a ∼99% agreement with the manual labeling of the spectra. The flexibility and noninvasive nature of the Raman measurements, paired with the predictive accuracy of the ML, is expected to facilitate the exploration of the emerging research field of twisted van der Waals heterostructures. Moreover, the present work showcases how the currently available open-source tools facilitate the study and integration of ML-based techniques.