Machine Learning aided Phonon Anarmonicity: the Soft Mode in the Quantum Paraelectric KTaO3

Quantum paraelectric materials are characterized by atypical fluctuations of the polarization due to the anharmonic lattice dynamics at low temperatures [1]. In the incipient ferroelectric KTaO3 perovskite, the low-temperature transition towards a fully ferroelectric phase is driven by the coupling of translation and soft modes. This transition is not accurately described within the conventional harmonic approximation. Here, we adopt the Stochastic Self-Consistent Harmonic Approximation (SSCHA) [3] accelerated by Machine Learning Force Field (MLFF) [4] to determine the full anharmonic energy contribution and the renormalized phonon frequencies in KTaO3. The inclusion of the anharmonic terms leads to a fairly good agreement with the experimental phonon spectrum and shed light on the temperature evolution of the soft mode frequency and the quantum ferroelectic transition. The efficiency and precision of our MLFF-aided stochastic method may open new paths for the study of quantum paraelectrics and phonon instabilities in large systems.

[1] S. E. Rowley et al. Ferroelectric quantum criticality. Nature Phys 10, 367–372 (2014)

[3] L. Monacelli et al. 2021 J. Phys.: Condens. Matter 33 363001

[4] R. Jinnouchi at Al. Phys. Rev. B 100. 014105 (2019)