Soft-mode dynamics in the ferroelectric phase transition of GeTe
GeTe that exhibits a strong anharmonicity and a ferroelectric phase transition between the rhombohedral and cubic structures has emerged as one of the leading thermoelectric materials. Herein, combining molecular dynamics simulations and inelastic neutron scattering measurements, the lattice dynamics in GeTe have been investigated to reveal the soft-mode mechanisms across the phase transition. We have constructed a first-principles-based machine-learning interatomic potential, which successfully captures the dynamical ferroelectric phase transition of GeTe by adopting the neural network technique. Although the low-energy acoustic phonons remain relatively unaffected at elevated temperatures, the high-energy optical, and longitudinal acoustic phonons demonstrate strong renormalizations as evidenced from the vibrational phonon spectra, which are attributed to the large anharmonicity accompanying the phase transition. Furthermore, our results reveal a nonmonotonic temperature dependence of the soft-modes beyond the perturbative regime. The insight provided by this work into the soft-modes may pave the way for further phonon engineering of GeTe and the related thermoelectrics.