Exploring structures of nanoclusters combining high-dimensional neural network potentials with unsupervised machine learning algorithms
AI Agent · 12.12.2025 · 1 (1)
DOI: 10.20517/aiagent.2025.09 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI Agent, online erschienen am 12.12.2025. Verfasst von Huabing Cai und 4 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 36 Literaturangaben.
Abstract (Originalsprache)
Understanding the structural evolution of catalysts with temperature is crucial for elucidating the real active sites under thermal catalytic conditions. Cerium oxides have emerged as versatile catalysts in such environments; however, the temperature-dependent stability of pristine nanoclusters remains poorly understood, creating a knowledge gap between their idealized and working structures. Herein, we developed a machine learning workflow combining high-dimensional neural network potentials and unsupervised machine learning algorithms to uncover the intrinsic structural evolution of stoichiometric CenO2n (n < 21) clusters. We first identified the most stable configurations for each cluster size and determined several magic numbers (n = 5, 8, 10, 12, 14, 20) with enhanced stability at 0 K. Nanosecond-scale neural network potential molecular dynamics simulations were then employed to explore temperature-dependent behavior, where the most frequently appearing structures extracted via K-means clustering were adopted as stability indicators. The results reveal that the temperature-sensitive stability varies with cluster size: magic-number clusters such as Ce14O28 maintain relative robust stability at elevated temperatures, while non-magic-number clusters such as Ce4O8 undergo pronounced structural fluctuations. This work not only offers new insights into temperature-induced distortion and reconstruction of oxide clusters, but also establishes an efficient workflow for high-precision structural prediction especially under high temperatures.
Zitieren
APA
Cai, H., Xu, A., Rajarathnam, G., Cao, A., & Yan, J. (2025). Exploring structures of nanoclusters combining high-dimensional neural network potentials with unsupervised machine learning algorithms. AI Agent, 1(1). https://doi.org/10.20517/aiagent.2025.09
Vancouver
Cai H, Xu A, Rajarathnam G, Cao A, Yan J. Exploring structures of nanoclusters combining high-dimensional neural network potentials with unsupervised machine learning algorithms. AI Agent. 2025;1(1). doi:10.20517/aiagent.2025.09
BibTeX
@article{Cai2025,
author = {Cai, Huabing and Xu, Aoni and Rajarathnam, Gobinath and Cao, Ang and Yan, Jianhua},
title = {Exploring structures of nanoclusters combining high-dimensional neural network potentials with unsupervised machine learning algorithms},
journal = {AI Agent},
year = {2025},
volume = {1},
number = {1},
doi = {10.20517/aiagent.2025.09}
}
EUVIMED