The prediction of infrared dielectric function of strongly anharmonic crystals at elevated temperatures remains a nontrivial task. In this study, we target this challenge via Green-Kubo molecular dynamics based on a machine learning potential trained from first-principles data using LiF as a prototype material. The temperature-dependent infrared dielectric function and reflectance of LiF crystals in the range 295–840 K are predicted and show very good agreement with experimental results. We also demonstrate that the classical Lorentz model with input from perturbation theory deviates appreciably from the experiment at high temperatures, even considering the lattice expansion as well as both the third- and fourth-order anharmonicity. These findings highlight the machine learning molecular dynamics as an efficient approach in accurately capturing the physics of infrared optical response in strongly anharmonic materials.

Close

PrevNext

View figure in article

• Anharmonic lattice dynamics
• Dielectric properties
• Optical phonons
• Deep potential molecular dynamics