Insight into the temperature-dependent lattice thermal properties of CeO₂ stabilized ZrO₂ by machine learning force...

Area of Science:

• Materials Science
• Computational Physics
• Thermodynamics

Background:

• Lattice thermal conductivity (LTC) in ZrO2-CeO2 is vital for thermal protection applications.
• Understanding atomic-scale processes governing LTC and phase transitions is challenging.
• Existing methods like density functional theory have limitations regarding size and temperature effects.

Purpose of the Study:

• To develop a machine learning (ML) interatomic potential for the ZrO2-CeO2 system.
• To investigate thermodynamic processes, including phase transitions and LTC, in large superlattice systems.
• To overcome limitations of traditional methods for simulating thermal properties.

Main Methods:

• Farthest Point Sampling (FPS) for dataset sampling.
• Machine learning (ML) approach to derive interatomic potential.
• Neuroevolution Potential (NEP) framework within molecular dynamics (MD) simulations.
• Homogeneous non-equilibrium molecular dynamics (HNEMD) for thermal conductivity analysis.
• Spectral decomposition for mechanism analysis.

Main Results:

• Successfully derived an ML interatomic potential, overcoming DFT limitations.
• Observed significant first-order and second-order phase transitions during heating.
• Investigated thermal conductivity variations due to lattice size effects.
• Predicted lattice thermal conductivity at different doping concentrations using NEP-MD.
• System exhibits relatively low thermal conductivity at room temperature.

Conclusions:

• The developed ML potential accurately models thermal properties and phase transitions in ZrO2-CeO2.
• The study provides insights into the mechanisms of thermal transport in these materials.
• Findings are crucial for optimizing ZrO2-CeO2 materials for thermal protection and other industrial applications.