About this Abstract |
Meeting |
TMS Specialty Congress 2025
|
Symposium
|
The 7th International Congress on 3D Materials Science (3DMS 2025)
|
Presentation Title |
Quantitative Elemental Mapping of Bimetallic Nanoparticles From Atomic Scale STEM-HAADF Images |
Author(s) |
Adrien Moncomble, Damien Alloyeau, Maxime Moreaud, Guillaume Wang, Nathaly Ortiz-Peña, Hakim Amara, Riccardo Gatti, Romain Moreau, Christian Ricolleau, Jaysen Nelayah |
On-Site Speaker (Planned) |
Adrien Moncomble |
Abstract Scope |
We propose a deep-learning method for quantifying atomic column composition in bimetallic nanoparticles (NPs) using high-angle annular dark field scanning TEM (HAADF-STEM). When traditional EDX techniques require high-brightness electron sources and are limited by noise at atomic resolution, HAADF-STEM images are promising alternative to get access to the chemical composition of NPs. In this approach, elemental composition is retrieved from HAADF signal intensities using a U-Net like deep-learning model trained on multislice-simulated images and corresponding elemental maps. Multislice simulations reveal that HAADF-STEM intensity is influenced by atomic column composition, configuration, and thickness, requiring a sophisticated model to capture these interactions. Our approach disentangles these parameters, allowing accurate predictions at atomic resolution and providing robust, high-throughput in situ quantification of elemental composition. This method offers a significant advantage over EDX by achieving non-destructive, contamination-free composition analysis, making it highly suitable for real-time experimental conditions. |
Proceedings Inclusion? |
Undecided |