Nature of the Superionic Phase Transition of Lithium Nitride from Machine Learning Force Fields

Abstract

Superionic conductors have great potential as solid-state electrolytes, but the physics of type-II superionic transitions remains elusive. In this study, we employed molecular dynamics simulations, using machine learning force fields, to investigate the type-II superionic phase transition in α-Li 3 N. We characterized Li 3 N above and below the superionic phase transition by calculating the heat capacity, Li + ion self-diffusion coefficient, and Li defect concentrations as functions of temperature. Our findings indicate that both the Li + self-diffusion coefficient and Li vacancy concentration follow distinct Arrhenius relationships in the normal and superionic regimes. The activation energies for self-diffusion and Li vacancy formation decrease by a similar proportion across the superionic phase transition. This result suggests that the superionic transition may be driven by a decrease in defect formation energetics rather than changes in Li transport mechanism. This insight may have implications for other type-II superionic materials.

Going against the Grain: Atomistic Modeling of Grain Boundaries in Solid Electrolytes for Solid-State Batteries

“…As a result, all of the chemical behavior is learned from the reference data. Powerful and accurate MLFFs have been developed for a range of topical solid electrolyte materials. − Several exhaustive reviews of MLFFs and their development have been published in recent years. − …”

Harvest the Polyanion Rotation in Sodium Superionic Conductors?

“…Additionally, the high cost of AIMD simulations means that they are typically performed at elevated temperatures to obtain sufficient diffusion statistics, and the AIMD simulation temperatures for reported SEs with anion rotation are all above 600 K. ,,,,, Whether anion rotation can be sustained at room temperature (300 K), which is more relevant to the application of solid-state batteries, requires further exploration. The recently emerging molecular dynamics simulation based on the machine-learning interatomic potentials ,− (MLMD) make it possible to study the anion rotation behavior of SEs at low temperature and long time scale with near density functional theory (DFT) accuracy in energies and forces.…”

Machine-Learning Accelerated First-Principles Accurate Modeling of the Solid–Liquid Phase Transition in MgO under Mantle Conditions

“…Recent advancements in machine learning and deep learning have enabled rapid simulations of materials for different applications. − In particular, deep neural network potentials (DNPs) allowed for the development of transferable and efficient machine-learned interatomic potentials that have accuracy similar to that of DFT but are orders of magnitude computationally faster. − For example, it has been previously demonstrated that DNPs can reliably replicate DFT values across different systems, including elemental and binary , metals, supported metal nanoclusters, hybrid perovskites, and metal oxides . We posit that DNPs can also be successfully applied at extreme temperature and pressures spanning the different phases for MgO and can successfully describe the changes in electronic structure from insulating to metallic behavior as pressure increases to mantle conditions.…”