Machine learning interatomic potential for predicting the thermal properties of uranium nitride

TL;DR

This work develops a Moment Tensor Potential (MTP) trained on density functional theory data for uranium nitride (UN), achieving accurate predictions of structural, vibrational, and defect energetics, and then leverages MD and relaxation-time approximations to predict thermal properties including expansion, heat capacity, melting behavior, and thermal conductivity. The MTP is validated against DFT benchmarks and experimental data, and is shown to capture phonon dispersion and defect formation energies with high fidelity. The study introduces new low-temperature thermal-conductivity measurements on UN single crystals, demonstrating the importance of four-phonon and electron–phonon scattering in obtaining agreement with experiment. Overall, the developed MTP provides a computationally efficient, predictive tool for UN thermal transport and high-temperature behavior, with potential applications in radiation damage modeling and fuel-performance simulations.

Abstract

We present a combined computational and experimental investigation of the thermal properties of uranium nitride (UN), focusing on the development of a machine learning interatomic potential (MLIP) using the moment tensor potential (MTP) framework. The MLIP was trained on density functional theory (DFT) data and validated against various quantities including energies, forces, elastic constants, phonon dispersion, and defect formation energies, achieving excellent agreement with DFT calculations, prior experimental results and our thermal conductivity measurement. The potential was then employed in molecular dynamics (MD) simulations to predict key thermal properties such as melting point, thermal expansion, specific heat, and thermal conductivity. To further assess model accuracy, we fabricated a UN sample and performed new thermal conductivity measurements representative of single-crystal properties, which showed strong agreement with the MLIP predictions. This work confirms the reliability and predictive capability of the developed potential for determining the thermal properties of UN.

Figures (9)

• Figure 1: (a) Model of FM ordering of the spins of uranium atom. (b) Flowchart of the MLIP training procedure.
• Figure 2: (a) Energies and (b) forces predicted by the developed MTP compared to the DFT reference data in validation set for the corresponding conventional 2$\times$2$\times$2 UN supercells.
• Figure 3: Equation of state predicted by the developed MTP, compared with DFT results calculated using a conventional 2$\times$2$\times$2 UN supercell with the same settings as those employed for database generation.
• Figure 4: Phonon dispersion and density of state (DOS) of UN predicted by the trained MTP and DFT calculations using a conventional 2$\times$2$\times$2 UN supercell with 10$\times$10$\times$10 mesh grid, compared with the experimental data taken from Jackman et al. Jackman1986.
• Figure 5: (a) Lattice parameter and (b) relative linear thermal expansion (rLTE in %) as functions of $T$ predicted by MTP using MD simulation with a conventional 8$\times$8$\times$8 UN supercell, compared with experimental results from Liu et al. (data points taken from Ref LIU2023154215) and Hayes et al. (data points taken from Ref Hayes1990) as well as MD simulation results obtained using the HIP-NN potential (data points taken from Ref. alzate) and the Kocevski potential (data points taken from Ref. Kocevski2022).