Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 7: Microstructures and Phase Transformations: Oxides & Perovskites
MM 7.2: Talk
Monday, September 5, 2022, 16:00–16:15, H45
Phase transitions and phonon renormalization in CsPbBr3 via a machine learning interatomic potential — •Erik Fransson1, Fredrik Eriksson1, Petter Rosander1, Terumasa Tadano2, and Paul Erhart1 — 1Chalmers University of Technology, Gothenburg, Sweden — 2National Institute for Materials Science, Tsukuba, Japan
Here, we present a study on phase-transition and phonon renormalization in the metal halide perovskite CsPbBr3 using a machine learning (ML) interatomic potential. The ML potential is a neuroevolution-potential constructed using the GPUMD software and is trained on atomic forces, energies and stresses obtained from DFT calculations. We find that the ML potential captures the correct phase-transition from the orthorhombic phase to tetragonal phase and from tetragonal to cubic phase. The phase transition temperatures obtained are slightly underestimated compared to experimental studies, but in good qualitative agreement. These phase-transitions are connected to the so called tilt-modes at the R and M point in the brillouin zone, which corresponds to tilting of the PbBr6 octahedra. The dynamics of these modes are studied with MD simulations and mode projections, and we find that they have are strongly anaharmonic and over-damped character. Furthermore, we investigate and benchmark how different self-consistent phonon methods work for these strongly anharmonic modes.