A Machine-Learning Interatomic Potential to Understand the Anisotropic Fracture Behavior of Bazro3 Material

Abstract

The complex operating environment severely tests the mechanical stability of the electrolyte material BaZrO3 (BZO), which affects the performance of solid oxide fuel cells(SOFCs). Molecular dynamics (MD) simulation provides an efficient method to research the mechanical behavior of nanocrystalline materials. In the interest of researching the mechanical behavior of BZO materials utilizing a machine learning (ML) MD approach, this article has established the most effective ML potential with DFT reliability. By contrasting the lattice constants and elastic constants with the DFT data, the precision of the potential was confirmed. By simulating the uniaxial tensile mechanical behavior of the BZO models without crack and with a central crack in the crystal orientation of [100] and [110], we found that the BZO material has a significant anisotropic property. In the crack-free model, the BZO crack propagation mode in the [100] crystal orientation is a shear fracture. The central microcrack tilts downward and expands symmetrically in the [110] crystal-oriented BZO model. When the initial structure with a central crack, the fracture mode changes significantly, and in both crystal orientations, the crack fracture along the Ba-O plane, the crack expands horizontally in the BZO model in the [100] crystal orientation and the crack expands zigzag in the BZO model in the [110] crystal orientation. The calculations of fracture strength, critical energy release rate, and fracture toughness show that the [110] crystal-oriented BZO has a stronger resistance to fracture than the [100] crystal-oriented BZO model.

Keywords: Machine learning, Molecular dynamics, Anisotropic property, Crack propagation, Fracture toughness