Toward machine learning interatomic potentials for modeling uranium mononitride

TL;DR

This paper presents the development of the first machine learning interatomic potentials for uranium mononitride, enabling accurate atomic-scale simulations of its thermophysical properties, defects, and radiation effects.

Contribution

The study introduces novel neural network potentials for UN, trained with active learning and DFT data, to accurately model its properties at finite temperatures.

Findings

Successfully reproduce thermophysical properties like lattice parameter and heat capacity.

Accurately predict defect energies and migration barriers.

Reliable for simulating diffusion, impurities, and radiation damage.

Abstract

Uranium mononitride (UN) is a promising accident-tolerant fuel because of its high fissile density and high thermal conductivity. In this study, we developed the first machine learning interatomic potentials for reliable atomic-scale modeling of UN at finite temperatures. We constructed a training set using density functional theory (DFT) calculations that was enriched through an active learning procedure, and two neural network potentials were generated. Both potentials successfully reproduce key thermophysical properties of interest, such as temperature-dependent lattice parameter, specific heat capacity, and bulk modulus. We also evaluated the energy of stoichiometric defect reactions and defect migration barriers and found close agreement with DFT predictions, demonstrating that our potentials can be used for modeling defects in UN. Additional tests provide evidence that our potentials are reliable for simulating diffusion, noble gas impurities, and radiation damage.