Understanding Correlations in BaZrO₃: Structure and Dynamics on the Nanoscale

Abstract

Barium zirconate (BaZrO 3 ) is one of few perovskites that is claimed to retain an average cubic structure down to 0 K at ambient pressure while being energetically very close to a tetragonal phase obtained by condensation of a soft phonon mode at the R-point. Previous studies suggest, however, that the local structure of BaZrO 3 may change at low temperature, forming nanodomains or a glass-like phase. Here, we investigate the global and local structure of BaZrO 3 as a function of temperature and pressure via molecular dynamics simulations using machine-learned potential with near density functional theory (DFT) accuracy. We show that the softening of the octahedral tilt mode at the R-point gives rise to weak diffuse superlattice reflections at low temperatures and ambient pressure, which are also observed experimentally. However, we do not observe any static nanodomains but rather soft dynamic fluctuations of the ZrO 6 octahedra with a correlation length of 2−3 nm over time scales of about 1 ps. This soft dynamic behavior is the precursor of a phase transition and explains the emergence of weak superlattice peaks in measurements. On the other hand, when increasing the pressure at 300 K, we find a phase transition from the cubic to the tetragonal phase at around 16 GPa, also in agreement with experimental studies.

Tensorial Properties via the Neuroevolution Potential Framework: Fast Simulation of Infrared and Raman Spectra

“…BaZrO 3 is a perovskite that is being investigated, e.g., as a proton conductor for applications in fuel cells. It has also been the subject of various fundamental studies, as it is a prototypical antiferroelectric perovskite. − It features soft and strongly temperature-dependent phonon modes, , which have been carefully analyzed with Raman spectroscopy, rendering BaZrO 3 an ideal application for the present approach.…”

“…The reference data set comprised cubic and tetragonal supercells with up to 40 atoms. The training structures were taken from MD simulations at different temperatures and pressures, generated using a NEP PES model constructed in an earlier study . In total, the reference data set contained 940 structures.…”

“…The training structures were taken from MD simulations at different temperatures and pressures, generated using a NEP PES model constructed in an earlier study. 89 In total, the reference data set contained 940 structures. 140 structures were randomly placed in a holdout set for validation, while training sets were compiled by the shuffle-split method (random selection with replacement) with 200 to 800 structures and five data sets per training set size.…”

Impact of Organic Spacers and Dimensionality on Templating of Halide Perovskites

“…To obtain a more-detailed understanding of the transitions, we compute ,− (Section S1.5) the distribution over octahedral tilt angles P (θ) along the cooling simulations (Figure ). The tilt angle distribution for a given temperature is averaged over a few snapshots corresponding to a temperature window of ∼1 K in order to improve the statistics.…”