Deep Learning-Assisted Quantification of Atomic Dopants and Defects in 2D Materials

Area of Science:

• Materials Science
• Nanotechnology
• Condensed Matter Physics

Background:

• Atomic dopants and defects are essential for novel functionalities in 2D transition metal dichalcogenides (2D TMDs).
• Precise atomic-scale identification and quantification are critical for engineering 2D TMDs for applications in optoelectronics and magnetic semiconductors.
• Scanning transmission electron microscopy (STEM) with sub-Å probe is a powerful tool for visualizing these atomic features, but manual analysis is laborious and subjective.

Purpose of the Study:

• To develop an efficient and automated method for the reliable quantification of dopants and defects in 2D TMDs with single-atom precision.
• To overcome the limitations of manual data analysis in STEM imaging of 2D materials.

Main Methods:

• Utilized a deep learning algorithm, specifically a fully convolutional neural network (CNN), known for its image segmentation capabilities.
• Applied the CNN to analyze scanning transmission electron microscopy (STEM) images of 2D transition metal dichalcogenides.

Main Results:

• Achieved precise mapping of atomic dopants and defects with a detection limit of approximately 1 × 10¹² cm^-2.
• Demonstrated a measurement accuracy of approximately 98% for most atomic sites.
• Validated the methodology's applicability to large datasets for extracting atomic site-specific information.

Conclusions:

• The proposed deep learning approach enables automated, precise, and reliable quantification of atomic dopants and defects in 2D TMDs.
• This methodology facilitates large-scale data analysis, offering insights into defect formation mechanisms.
• The findings pave the way for advanced engineering of 2D materials for diverse technological applications.