Quantifying the thickness of WTe2 using atomic-resolution STEM simulations and supervised machine learning.

Area of Science:

• Materials Science
• Condensed Matter Physics
• Nanotechnology

Background:

• Thickness dictates physical and chemical properties of two-dimensional (2D) materials.
• Tungsten ditelluride (WTe2) exhibits thickness-dependent properties, but its complex structure hinders thickness determination.
• Air sensitivity and electron beam damage necessitate direct, non-destructive thickness characterization methods.

Purpose of the Study:

• To develop a direct method for determining the thickness of Td-WTe2 up to ten van der Waals layers.
• To enable precise atomic structure characterization, including local thickness variations and defects, in few-layer 2D materials.

Main Methods:

• Atomic-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image simulation.
• Analysis of intensity line profiles from overlapping atomic columns.
• Development of a standard neural network model trained on line profile features.

Main Results:

• Distinguished even and odd thicknesses up to seven layers by comparing deconvoluted peak intensity or area ratios without machine learning.
• Achieved thickness distinction up to ten layers using a standard neural network model trained on line profile features.
• Demonstrated up to 94% accuracy in thickness determination, even with Gaussian and Poisson noise.

Conclusions:

• The developed method efficiently quantifies Td-WTe2 thickness using HAADF-STEM image simulation and analysis.
• The approach is extendable to other 2D materials with similar structural challenges.
• Provides a pathway for precise characterization of few-layer 2D materials, including thickness variations and atomic defects.