Machine-learning-empowered identification of initial growth modes for 2D transition metal dichalcogenide thin films

Abstract

In situ reflection high-energy electron diffraction (RHEED) is a powerful technique for monitoring surface states and offers invaluable insights into thin film growth. However, extracting hidden features and subtle changes from its vast data remains as a challenge. This work bridges the gap by employing machine learning (ML)-empowered RHEED analysis to elucidate the growth dynamics of two-dimensional (2D) transition metal dichalcogenide (TMDC) thin films grown under two distinct growth modes. Principal component analysis (PCA) and its modified processes were used to separate contributions of the graphene substrate and the MoSe2 film in the RHEED video. The ML-empowered RHEED analysis allowed us to effectively filter out the strong substrate signal and reconstructing RHEED videos solely for the MoSe2 films. This approach enabled detailed monitoring of film growth with unique features, and clearly distinguishing between the layer-by-layer growth mode and the island one. This work demonstrates the potentials of ML-empowered RHEED analysis for revealing complex growth dynamics of 2D TMDC materials, paving the way for advanced thin film monitoring and autonomous control in wider scope of thin film technologies.