Hidden Markov model for atom-counting from sequential ADF STEM images: Methodology, possibilities and limitations.

Area of Science:

• Materials Science
• Nanotechnology
• Physics

Background:

• Accurate measurement of dynamic atomic structural changes in nanomaterials is essential for understanding their behavior.
• Current atom-counting methods in time-series electron microscopy have limitations in precision and dynamic analysis.

Purpose of the Study:

• To develop a reliable quantitative method for measuring dynamic atomic structure changes in monatomic crystalline nanomaterials.
• To improve atom-counting accuracy in time-series atomic resolution scanning transmission electron microscopy (AR-STEM).

Main Methods:

• Utilized a hidden Markov model (HMM) approach applied to time-series AR-STEM images.
• Estimated atom counts per atomic column for each frame in the image series.
• Incorporated transition probabilities to model atom loss/gain between frames.

Main Results:

• The HMM-based method significantly improves time-series atom-counting performance compared to existing techniques.
• Transition probabilities were identified as critical for enhanced accuracy in dynamic analysis.
• The method demonstrated potential for estimating probabilities and cross-sections of structural changes.

Conclusions:

• The developed quantitative method offers a promising tool for reliable analysis of dynamic processes in nanomaterials.
• Applicable to various dynamic phenomena including surface diffusion, adatom dynamics, beam effects, and in situ experiments.
• Robustness under variable environmental conditions suggests broad applicability.