Pressure-Dependent Thermodynamics of Cubic Lu-H-N Solid Solutions by First-Principles Calculations and Graph Neural Networks

Metal hydrides are an important class of materials with high scientific significance, finding applications in various fields such as hydrogen storage, batteries, gas sensors, nuclear reactions and high-temperature superconductivity. Pressure, among others, is an important variable to control their properties. Previous computational studies of metal hydrides under high pressures usually treat them as stoichiometric compounds, without lattice disorder considered. However, when pressure becomes more moderate, lattice disorder gets more significant, as shown in the recent claimed room-temperature superconductor at near-ambient pressures, N-doped Lu hydrides, where three constituents (hydrogen, nitrogen and vacancy) have disordered occupancies in the tetrahedral and octahedral interstitial sites in the fcc Lu lattice. To consider lattice disorder dependent on pressure, in addition to other variables including temperature and composition, first-principles calculations are computationally demanding. In this work, a model of the cubic Lu-H-N solid solutions is developed by combining first-principles calculations and lattice graph neural networks to learn pressure-dependent thermodynamic quantities in the configurational space, and composition-pressure-temperature phase diagrams are derived to describe the relationship between the synthesis conditions and the resulted phase equilibria. This work can improve the thermodynamic understanding of the Lu-H-N system and help rational synthesis of N-doped Lu hydrides, as well as demonstrate an efficient approach to model pressure-dependent thermodynamics of multi-component solid solutions.