Interpretable Structural Evaluation of Metal-Oxide Nanostructures in Scanning Transmission Electron Microscopy (STEM)...

Area of Science:

• Materials Science
• Computational Topology
• Nanotechnology

Background:

• Persistent homology (PH) is vital for structure quantification but lacks interpretability.
• Scanning transmission electron microscopy (STEM) images self-assembled Pt-CeO2 nanostructures.

Purpose of the Study:

• Extract interpretable geometric features from PH persistent diagrams (PDs) of Pt-CeO2 nanostructures.
• Develop a method for classifying nanostructures based on these features.

Main Methods:

• Applied PH to STEM images of Pt-CeO2 nanostructures.
• Extracted features from zeroth and first PDs.
• Utilized hierarchical clustering and principal component analysis (PCA) for inverse analysis.
• Developed a disorder descriptor based on CeO2 phase arcs.

Main Results:

• Identified five interpretable features from PDs.
• Highlighted the significance of small CeO2 arcs in the first PDs.
• Quantified nanostructure disorder using a novel descriptor.
• Successfully classified 12 Pt-CeO2 nanostructures.

Conclusions:

• Interpretable geometric features can be extracted from PDs.
• A novel descriptor effectively quantifies disorder in nanostructures.
• This approach enables clear classification of complex nanostructure patterns.