Accelerating Global Optimization of Cerium Oxide Nanocluster Structures with High-Dimensional Neural Network...

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Cerium oxide (CeO2) is crucial in catalysis due to its electronic structure and oxygen storage capacity.
• Characterizing complex CeO2 nanocluster structures is challenging.
• Understanding nanocluster structure-property relationships is vital for optimizing catalytic applications.

Purpose of the Study:

• To develop and apply a machine learning approach for efficient global optimization of cerium oxide nanocluster structures.
• To explore the configurational space of small to medium cerium oxide clusters (Ce_nO_{2x}, n=2-18).
• To investigate the size-dependent structural evolution and electronic properties of these nanoclusters.

Main Methods:

• Utilized a high-dimensional neural network potential (HDNNP) for accurate energy calculations.
• Integrated active learning to iteratively refine the HDNNP and explore the configuration space.
• Performed first-principles calculations for validation and comparison.

Main Results:

• The HDNNP achieved accuracy comparable to first-principles calculations.
• Identified distinct structural transitions: compact to multilayered, then to pyramidal structures at specific sizes (n=9, 14).
• Observed continuous growth from pyramidal structures for clusters with n > 14, alongside analysis of electronic structures.

Conclusions:

• The machine learning approach effectively accelerates the structural determination of cerium oxide nanoclusters.
• Structural configurations and electronic properties exhibit significant size dependency.
• Findings offer valuable insights into cerium oxide nanocluster behavior for catalyst design.