Elastic properties of amorphous LiTaCl${}_6$ solid-state electrolyte

Abstract

Amorphous solid-state electrolytes are attractive candidates for safe, high-energy-density all-solid-state batteries, yet their mechanical properties remain poorly understood from a computational perspective. Here, we investigate the elastic behavior of the recently discovered amorphous superionic Li-ion conductor LiTaCl${}_6$ using density-functional-theory (DFT)-based methods, including unrelaxed static, relaxed static, and strain fluctuations from molecular dynamics (MD) simulations in the isobaric--isothermal (NPT) ensemble with DFT-trained machine-learning force fields (MLFFs). While the unrelaxed static method predicts a Young's modulus an order of magnitude higher than the experiment, the relaxed static method---commonly applied to crystalline electrolytes---still overestimates the modulus by more than 170%. In contrast, the MD approach using MLFFs yields a Young's modulus of $2.84\pm 0.26$ GPa, which quantitatively agrees with the experimental value of $2.91\pm 0.32$ GPa. Using the MLFF-MD approach, we further predict bulk modulus (4.44 GPa), shear modulus (1.02 GPa), and Poisson's ratio (0.39) for amorphous LiTaCl${}_6$ and conclude that elastically it behaves like a soft polymer or gel. These results demonstrate that amorphous superionic materials possess some unique elastic properties and that, among the methods examined, only the MLFF-MD approach yields quantitative agreement with experiment, highlighting the necessity of a dynamical treatment to simulate their elastic response, consistent with recent findings for crystalline superionic conductors.

7 pages, 5 figures, Supplemental Material appended

References in corpus (4)

• Intrinsic Correlation between Hardness and Elasticity in Polycrystalline Materials and Bulk Metallic Glasses
• Accurate determination of crystal structures based on averaged local bond order parameters
• On-the-fly machine learning force field generation: Application to melting points
• Phase transitions of hybrid perovskites simulated by machine-learning force fields trained on-the-fly with Bayesian inference