Deep learning–assisted electron microscopy for apparent Z-height mapping of atomically dispersed catalysts

Atomically dispersed catalysts (ADCs) offer maximal metal utilization and local atomic/electronic structures, yet resolving their actual atomic configurations using aberration-corrected scanning transmission electron microscopy (AC-STEM) remains fundamentally constrained by depth ambiguity and complex experimental noise. Existing STEM–machine learning (ML) approaches often rely on ideal simulations or manual annotations, limiting their transferability to realistic ADC images and causing ambiguity between isolated atoms, projected dimers, and small clusters. Here, we develop a physics-informed, domain-adaptive CycleGAN-UNet framework for depth-aware ADC analysis. CycleGAN translates height-labeled simulations into experimental-like images while preserving atomic coordinates and apparent Z-height labels, and a depth-supervised UNet performs simultaneous atom localization and discrete Z-height classification. Support-specific simulations further incorporate support-dependent background contrast, including uniform amorphous carbon (am-C), am-C with local thickness variation, and crystalline supports, thereby improving the applicability of the framework to different ADC imaging conditions. Across diverse metals (Mn, Fe, Cu, Pd, and Pt) and supports (am-C, black phosphorus, and polymeric carbon nitride), the framework achieves high atom detection precision (∼95%) and reliable apparent Z-height classification at ∼2-nanometer binning. By encoding depth-dependent beam broadening features, the framework converts conventional 2D projections into quantitative, apparent Z-height atomic maps. In geminal Pt/PCN catalysts, it resolves the loading-dependent transition from isolated Pt monomers to Pt 2 sites and links these configurations to dehydrogenative aromatization activity. This framework transforms STEM images from projected atom maps into quantitative, depth-aware atomic descriptors for ADC structure–activity analysis.

Authors

Institutions

Publication Details

Journal
Science Advances
Published
2026-11-13
DOI
https://doi.org/10.1126/sciadv.aef7673
Primary Topic
Advanced Electron Microscopy Techniques and Applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Deep learning–assisted electron microscopy for apparent Z-height mapping of atomically dispersed catalysts

Fanqi Meng, Hai Lian Xiao, Xiaoxu Zhao, Xiaocang Han et al.
1 links
Science Advances
Advanced Electron Microscopy Techniques and Applications
article

Deep learning–assisted electron microscopy for apparent Z-height mapping of atomically dispersed catalysts

Fanqi Meng, Hai Lian Xiao, Xiaoxu Zhao, Xiaocang Han, Yuan Meng, Jiong Lu, Zanlin Qiu, Shu Miao, Jin Zhang, Jun Zhao, Zezhong Zhang, Qilong Cai, Yanhui Hong, Kangshu Li, Weiwei Hu
article en

Abstract

Atomically dispersed catalysts (ADCs) offer maximal metal utilization and local atomic/electronic structures, yet resolving their actual atomic configurations using aberration-corrected scanning transmission electron microscopy (AC-STEM) remains fundamentally constrained by depth ambiguity and complex experimental noise. Existing STEM–machine learning (ML) approaches often rely on ideal simulations or manual annotations, limiting their transferability to realistic ADC images and causing ambiguity between isolated atoms, projected dimers, and small clusters. Here, we develop a physics-informed, domain-adaptive CycleGAN-UNet framework for depth-aware ADC analysis. CycleGAN translates height-labeled simulations into experimental-like images while preserving atomic coordinates and apparent Z-height labels, and a depth-supervised UNet performs simultaneous atom localization and discrete Z-height classification. Support-specific simulations further incorporate support-dependent background contrast, including uniform amorphous carbon (am-C), am-C with local thickness variation, and crystalline supports, thereby improving the applicability of the framework to different ADC imaging conditions. Across diverse metals (Mn, Fe, Cu, Pd, and Pt) and supports (am-C, black phosphorus, and polymeric carbon nitride), the framework achieves high atom detection precision (∼95%) and reliable apparent Z-height classification at ∼2-nanometer binning. By encoding depth-dependent beam broadening features, the framework converts conventional 2D projections into quantitative, apparent Z-height atomic maps. In geminal Pt/PCN catalysts, it resolves the loading-dependent transition from isolated Pt monomers to Pt 2 sites and links these configurations to dehydrogenative aromatization activity. This framework transforms STEM images from projected atom maps into quantitative, depth-aware atomic descriptors for ADC structure–activity analysis.

Science AdvancesVol. 12(46)
National University of Singapore (SG), Peking University (CN), JEOL (China) (CN)
National Natural Science Foundation of China, National Key Research and Development Program of China
Advanced Electron Microscopy Techniques and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon