Mechanisms of alkali ionic transport in amorphous oxyhalides solid state conductors

TL;DR

This study uses machine learning-enhanced molecular dynamics to elucidate the mechanisms behind high ionic conductivity in amorphous oxyhalides, revealing the roles of structural features and particle correlations in alkali ion transport.

Contribution

It introduces a fine-tuned machine learning interatomic potential to analyze large-scale molecular dynamics, uncovering the structural and dynamical factors influencing ionic transport in amorphous oxyhalides.

Findings

Oxygen anions on metal-anion tetrahedra limit alkali diffusion.

Ionic transport is largely uncorrelated, with Haven ratio close to one.

Diffusion properties are insensitive to variations in atomic chemistry.

Abstract

Amorphous oxyhalides have attracted significant attention due to their relatively high ionic conductivity ($>$1 mS cm$^{-1}$), excellent chemical stability, mechanical softness, and facile synthesis routes via standard solid-state reactions. These materials exhibit an ionic conductivity that is almost independent of the underlying chemistry, in stark contrast to what occurs in crystalline conductors. In this work, we employ an accurately fine-tuned machine learning interatomic potential to construct large-scale molecular dynamics trajectories encompassing hundreds of nanoseconds to obtain statistically converged transport properties. We find that the amorphous state consists of chain fragments of metal-anion tetrahedra of various lenght. By analyzing the residence time of alkali cations migrating around tetrahedrally-coordinated trivalent metal ions, we find that oxygen anions on the metal-anion tetrahedra limit alkali diffusion. By computing the full Einstein expression of the ionic conductivity, we demonstrate that the alkali transference number of these materials is strongly influenced by distinct-particles correlations, while at the same time they are characterized by an alkali Haven ratio close to one, implying that ionic transport is largely dictated by uncorrelated self-diffusion. Finally, by extending this analysis to chemical compositions $AMX_{2.5}\textsf{O}_{0.75}$, spanning different alkaline ($A$ = Li, Na, K), metallic ($M$ = Al, Ga, In), and halogen ($X$ = Cl, Br, I) species, we clarify why the diffusion properties of these materials remain largely insensitive to variations in atomic chemistry.