Machine-learning-assisted analysis of transition metal dichalcogenide thin-film growth

• Published in: Nano convergence Vol. 10; no. 1; p. 10
• Main Authors: Kim, Hyuk Jin, Chong, Minsu, Rhee, Tae Gyu, Khim, Yeong Gwang, Jung, Min-Hyoung, Kim, Young-Min, Jeong, Hu Young, Choi, Byoung Ki, Chang, Young Jun
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 20.02.2023; Springer Nature B.V; SpringerOpen
• Subjects: Chalcogenides; Chemistry and Materials Science; Cluster analysis; Clustering; Electron diffraction; Epitaxial growth; Film growth; Graphene; High energy electrons; K-means clustering; Machine learning; Materials Science; Molecular beam epitaxy; Nanoscale Science and Technology; Nanotechnology; Nanotechnology and Microengineering; Principal component analysis; Principal components analysis; Pulsed laser deposition; Pulsed lasers; ReSe2; RHEED; Substrates; Thin films; TMDC; Transition metal compounds; Two dimensional materials; Vector quantization; ReSe; RHEED; K-means clustering; Machine learning; Principal component analysis; TMDC; ReSe2
• ISSN: 2196-5404; 2196-5404
• DOI: 10.1186/s40580-023-00359-5

In situ reflective high-energy electron diffraction (RHEED) is widely used to monitor the surface crystalline state during thin-film growth by molecular beam epitaxy (MBE) and pulsed laser deposition. With the recent development of machine learning (ML), ML-assisted analysis of RHEED videos aids in interpreting the complete RHEED data of oxide thin films. The quantitative analysis of RHEED data allows us to characterize and categorize the growth modes step by step, and extract hidden knowledge of the epitaxial film growth process. In this study, we employed the ML-assisted RHEED analysis method to investigate the growth of 2D thin films of transition metal dichalcogenides (ReSe 2 ) on graphene substrates by MBE. Principal component analysis (PCA) and K-means clustering were used to separate statistically important patterns and visualize the trend of pattern evolution without any notable loss of information. Using the modified PCA, we could monitor the diffraction intensity of solely the ReSe 2 layers by filtering out the substrate contribution. These findings demonstrate that ML analysis can be successfully employed to examine and understand the film-growth dynamics of 2D materials. Further, the ML-based method can pave the way for the development of advanced real-time monitoring and autonomous material synthesis techniques. Graphical Abstract