Development of a machine-learning interatomic potential for uranium under the moment tensor potential framework

Uranium is attracting growing interest across fundamental science and nuclear research, where atomistic sim-ulations remain challenging. In this study, we developed a novel uranium interatomic potential based on a machine-learning approach embedded in the moment tensor potential package. An active learning framework was utilized in the potential training, using density-functional theory plus Hubbard U (DFT+U) modified data. The potential accurately reproduced the lattice parameters, cohesive energy, elastic, vibrational, and thermo-dynamic properties of & alpha;-U compared to DFT+U calculations and experimental results. The basic properties of other phases, including fl, & gamma;, HCP, and FCC phases, were also validated. In addition, molecular dynamics sim-ulations were used to reproduce the temperature-induced allotropic transformation from & alpha;-U to & gamma;-U, which agrees with experimental observations. Our potential provides a computationally efficient means to study the physical behavior of uranium with nearly DFT accuracy.

Automated summary

Researchers have developed a new interatomic potential for uranium using a machine-learning approach. The potential was trained using density-functional theory plus Hubbard U modified data, and accurately reproduced the properties of α-U and other phases, including fl, γ, HCP, and FCC phases. Molecular dynamics simulations also successfully reproduced the temperature-induced allotropic transformation from α-U to γ-U, in agreement with experimental observations. This potential provides a computationally efficient means to study the physical behavior of uranium with nearly DFT accuracy.