Dresden 2026 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 19: Poster Session

MM 19.11: Poster

Dienstag, 10. März 2026, 18:00–20:00, P5

LLZO grain boundaries with doped amorphous domains by adaptively fine-tuned machine-learning interatomic potentials — •yuandong wang, yute chan, karsten reuter, and christoph scheurer — Fritz-Haber-Institut der MPG, Berlin, Germany

Garnet Li₇La₃Zr₂O₁₂ (LLZO) is a highly promising solid-state electrolyte (SSE) for lithium batteries. However, its application faces challenges, primarily arising from Li dendrite formation, the impact of grain boundaries (GBs) on Li transport and stability. Introducing amorphous intergranular domains can mitigate dendrite propagation while enhancing Li-ion mobility in GBs. Moreover, aliovalent cation doping (e.g., Al³⁺, Ga³⁺, Nb⁵⁺, Ta⁵⁺) in both cubic and amorphous LLZO offers additional levers to further enhance key properties. Rationally engineering the morphology of amorphous GBs offers an intriguing approach for tuning electrolyte performance.

In this study, we investigate strategies for generating structural motifs covering Li diffusion, GBs, and amorphous LLZO, which are diversified to include ion hopping transition states, large defects, amorphous motifs, etc. An adaptively fine-tuned MACE machine-learning interatomic potential (MLIP) is trained to accurately model large-scale and realistic nanoscale structures of LLZO with doped amorphous GBs. With the fine-tuned MLIP, the morphology, dopant effects and crystalline-amorphous interactions governing Li-ion diffusion pathways and activation barriers in LLZO can be studied in detail.

Keywords: solid-state battery; machine learning; interface; dopant; nanoscale simulation