| Abstract Scope |
Hydrogen-isotope transport in Li2O is often described as diffusion through a static oxide lattice. Our calculations show that this picture is incomplete: the local transport environment evolves with Li-sublattice disorder, native defects, and transient O–H bonding. We develop a DFT-trained atomic cluster expansion potential for Li–O–H and use it to connect first-principles defect mechanisms with transport-scale molecular dynamics. The model is trained on bulk, elastic, thermal, defect, surface, short-range, hydrogen-incorporation, and migration-pathway configurations, and is selected by property-level validation rather than fitting errors alone. It reproduces the Li2O lattice constant, elastic response, phonon stability, native-defect ordering, Li-vacancy migration, Li diffusion, and the onset of superionic Li transport. For hydrogen, the stable isolated-H environment is octahedral-like, while O–H-mediated configurations provide lower-barrier rearrangement pathways. The potential is then used to compare H, D, and T diffusion, solubility, surface stability and LiOH formation. |