Dresden 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

FM: Fachverband Funktionsmaterialien

FM 10: Crystal Defects and Real Structure Physics in Diamond and Functional Materials II

FM 10.3: Vortrag

Dienstag, 10. März 2026, 14:30–14:45, BEY/0E40

Machine-Learning Potentials Reveal Defect Segregation at Grain Boundaries in Lithium Solid Electrolytes — •Waldemar Kaiser¹, Hyunwon Chu², Jennifer L.M. Rupp^1,2, and David A. Egger¹ — ¹TUM School of Natural Sciences, Technical University of Munich — ²Department of Materials Science and Engineering, Massachusetts Institute of Technology

Grain boundaries (GB) play a central role in ionic and electronic transport in polycrystalline solid electrolytes [1, 2], yet their local defect chemistry remains difficult to resolve. Their structural complexity and chemical heterogeneity introduce a wide distribution of possible defect configurations that are computationally demanding to characterize with first-principles methods. As a result, the origins of variations of ionic and electronic transport often remain unclear.

Here, machine-learning potentials (MLPs) are applied to representative Σ5 and Σ3 GB structures to investigate ionic defects in lithium lanthanum zirconate (LLZO). The MLPs enable the systematic and efficient mapping of lithium- and oxygen-vacancy formation energies, revealing reductions of up to 1 eV within GB cores relative to bulk sites. Structural analysis further identifies locally Zr-deficient coordination environments as preferred incorporation sites for lithium vacancies, and experiments show that Ta-doping fills these bottlenecks and enhances the material’s stability.

[1] B. Gao et al., Adv. Energy Mater. 12, 2102151, 2021 [2] Y. Zhu et al., Nat. Rev. Mater. 6, 313-331, 2021

Keywords: Grain Boundaries; Solid Electrolytes; Defects; Machine Learning Potentials