Unveiling Twist Domains in Monolayer MoS₂ through 4D-STEM and Unsupervised Machine Learning

Area of Science:

• Materials Science
• Condensed Matter Physics
• Nanotechnology

Background:

• Two-dimensional (2D) materials like molybdenum disulfide (MoS2) exhibit unique properties due to their layered structure.
• Understanding crystal structures is vital for optimizing the performance of 2D materials in various applications.
• Conventional methods for analyzing atomic structures have limitations in field of view, hindering comprehensive domain analysis.

Purpose of the Study:

• To analyze the crystal domains of monolayer MoS2 synthesized via metal-organic chemical vapor deposition (MOCVD).
• To employ advanced imaging and machine learning techniques for a wide-field, detailed structural analysis.
• To identify and characterize domain structures, including twist angles and boundaries.

Main Methods:

• Utilized 4D scanning transmission electron microscopy (4D STEM) for high-resolution diffraction pattern acquisition.
• Applied unsupervised machine learning algorithms, including nonnegative matrix factorization (NMF) and hierarchical clustering, to analyze a large dataset (>22k) of diffraction patterns.
• Implemented preprocessing techniques to detect noncentrosymmetry and visualize domain polarities by analyzing violations of Friedel's law.

Main Results:

• Successfully identified twist domains with angles of ±11° in monolayer MoS2.
• Visualized the polarities of distinct MoS2 domains, indicating their crystallographic orientation.
• Characterized the MoS2 specimen, revealing domains of approximately 100 nm in size with numerous mirror-twin boundaries.
• Determined the optimal MOCVD synthesis conditions (e.g., substrate: Al2O3 (0001), temperature: 850 °C).

Conclusions:

• The study demonstrates the efficacy of combining 4D STEM with unsupervised machine learning for comprehensive analysis of 2D material domains.
• The findings provide critical insights into the crystal growth mechanisms and domain structures of MoS2 synthesized by MOCVD.
• This approach offers a pathway for optimizing MOCVD processes to control and enhance the properties of 2D materials for advanced applications.