Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 5: Topical Session: Physics-driven Artificial Intelligence for Materials I
MM 5.4: Talk
Monday, March 9, 2026, 11:15–11:30, SCH/A251
Fantastic Polaronic Peaks and Where to Find Them: Learning Vibrational Spectra of a Disordered Energy Material — •Christoph Dähn1, Yang Wang2, Risov Das2, Bettina V. Lotsch2, Karsten Reuter1, and Christian Carbogno1 — 1Fritz-Haber-Institut der MPG, Berlin — 2MPI für Festkörperforschung, Stuttgart
Vibrational Raman and infrared spectroscopy offers unique opportunities for characterizing microscopic structural and dynamical properties. For energy materials and in particular for solar batteries [1], a straightforward interpretation of such spectra is however hindered by the intrinsic structural and occupational disorder, which includes defects and polarons. At the same time, this also prevents their accurate ab initio simulation, which would require extensive calculations at a hybrid level of density-functional theory (DFT) in a multitude of disordered supercells. In this work, we discuss how machine-learning interatomic potentials trained on high-level DFT data can be used to capture the otherwise inaccessible vibrational dynamics. We demonstrate this approach for a two-dimensional titanium niobate featuring partially occupied metal sites and polarons. By Monte Carlo sampling its configurational disorder, we are able to disentangle polaronic signatures and disorder induced contributions in the spectra. This reveals how local atomic environments control polaron stability and offers insights on how doping can be used to control charge retention in such compounds.
[1] M. Rinaldi et al., J. Phys.: Mater. 8, 031003 (2025).
Keywords: Polaron; Raman and Infrared Spectroscopy; Machine-Learning Interatomic Potentials; Opto-Ionic Materials; Disordered Layered Oxides