Dynamic charges effect on infrared dielectric response of polar materials

Predictive modeling of the infrared dielectric function in polar materials is crucial for thermal management and infrared devices design. While the Green–Kubo molecular dynamics (MD) framework provides a nonperturbative route to compute dielectric responses from dipole fluctuations, yet it commonly relies on the fixed-charge approximation that neglects dynamic charge redistribution during atomic motion. Here, we employ a machine-learning neuroevolution potential with dynamic charges combined with Green–Kubo MD to investigate the dynamic charge effect on the infrared dielectric response of rutile TiO₂, a material with large Born effective charges. Our results show that dynamic charge effects become increasingly important at elevated temperatures and are essential for accurately predicting longitudinal optical phonon features and infrared reflectance. This work establishes that accurate prediction of infrared optical properties in polar materials under thermal excitation requires explicit treatment of dynamic charge evolution.