An Ab Initio Study and Machine Learning Framework to Capture the Motional Effects in Solid-State NMR of Lithium-Ion Conductors

Abstract

Solid-state NMR spectroscopy, when combined with first-principles density functional theory (DFT) calculations, offers a highly sensitive probe of atomic-scale structure and dynamics in solid-state ion conductors, enabling the characterisation of subtle features that govern ionic conductivity. However, current approaches for interpreting NMR spectra rely on a comparison with static DFT reference calculations, which are inadequate for materials exhibiting fast ion dynamics such as lithium battery solid electrolytes. Here, using room-temperature NMR measurements and first-principles calculations, we show that the standard static-structure approach fails to reproduce the experimental 35 Cl isotropic chemical shift (δ iso ) of the fast Li-ion conductor Li 6 PS 5 Cl and substantially overestimates the quadrupolar coupling constant (C Q ). We show that this discrepancy can be resolved using only ten DFT calculations by sampling relaxed configurations representative of Li-ion diffusion from machine-learning molecular dynamics. Compared with vibrational motion, Li-ion hopping around Cl is shown to dominate the motional averaging through reorientation of the NMR tensors. This study therefore provides an efficient computational method to resolve the complexities of the NMR spectra of Li 6 PS 5 Cl, which can be widely applied to other ion-conducting solids.