Finite temperature dielectric properties of KTaO$_3$ from first principles and machine learning: Phonon spectra, Barrett law, strain engineering and electrostriction

TL;DR

This study combines machine learning and ab initio methods to accurately predict the temperature-dependent dielectric properties, strain effects, and electrostriction in KTaO$_3$, a quantum paraelectric perovskite.

Contribution

It introduces a novel approach integrating machine-learning potentials with quantum lattice dynamics to efficiently model temperature-dependent properties of KTaO$_3$ with high accuracy.

Findings

Softening of polar phonon mode with temperature

Dielectric constant divergence and saturation behavior

Giant electrostriction under strain at room temperature

Abstract

Despite important breakthroughs in the last decade, the calculation of temperature dependent properties of solids still remains a challenging task, especially in the vicinity of structural phase transitions. We show that the combination of machine-learning interatomic potentials with quantum self-consistent ab initio lattice dynamics allows to calculate efficiently the temperature dependence of dielectric properties of the quantum paraelectric perovskite KTaO$_3$, with a precision beyond what could be reasonably achieved using plain density functional theory. We first follow the strong anharmonic softening of the polar mode in this incipient ferroelectric material, and the resulting divergence of the dielectric constant that eventually saturates due to the interplay between temperature and quantum fluctuations. Further, we predict the stability range of the quantum paraelectric state under the application of epitaxial strain at 0 K and 300 K. Finally, we calculate the temperature dependence of electrostrictive tensors for this material and show that giant electrostriction in KTaO$_3$ is to be expected also at room temperature under the condition of strain engineering.