Interpretable evaluation of STEM images of nanostructures via homology analysis

Eguchi Ryuto; Yu Wen; Ayako Hashimoto

ポスター Interpretable evaluation of STEM images of nanostructures via homology analysis

Eguchi Ryuto (Research Center for Energy and Environmental Materials (GREEN)/Battery and Cell Materials Field/Environment-Controlled Microscopy Group, National Institute for Materials Science) ; Yu Wen (Research Center for Advanced Measurement and Characterization/Atomic Structure and Physics Field/In-situ Characterization Technique Development Group, National Institute for Materials Science) ; Ayako Hashimoto (Research Center for Energy and Environmental Materials (GREEN)/Battery and Cell Materials Field/Environment-Controlled Microscopy Group, National Institute for Materials Science)

Download file (402KB)
Download as zip (389KB)

コレクション

引用

Eguchi Ryuto, Yu Wen, Ayako Hashimoto. Interpretable evaluation of STEM images of nanostructures via homology analysis. https://doi.org/10.48505/nims.5080

(BibTeX)

説明:

(abstract)

Persistent homology (PH) quantitatively expresses the "hole" structures of any data in terms of their "birth" and "death" and their scales. We applied PH analysis to STEM images of phase-separated nanostructures, particularly extracting interpretable features of inhomogenous structures. The target image data consisted of HAADF-STEM images of two-phase Pt/CeO2 nanocomposites. This material exhibits different morphologies such as maze-like and striped structures depending on synthesis conditions (temperature and CO/O2 synthesis gas ratio).STEM images were binarized and analyzed focusing on the CeO2 phase. PH analysis extracted features corresponding to the number of isolated CeO2 domains, width and length of CeO2 stripes,
as well as the number of rings and gulf-like structures. Furthermore, using these features in random forest classification revealed the significance of gulf-like structures derived from 1st PDs as structural descriptors.

権利情報: