(Key Person: N. Marzari, EPFL, in collaboration with the Bosch Research and Technology Center)
Solid inorganic Li-ion conductors are ultimately the only solution that can ensure safety of Li-based cells: no liquid organic electrolytes can withstand a runaway thermal event. Also, Li-metal anodes are key to high-energy batteries (Li-ion, Li-S, Li-air), but solid-electrolyte interphase layers formed in-situ in today’s organic electrolytes do not protect the Li anode against morphology changes during cycling. Hence a protective solid membrane is required that is flexible, strong, and has high conductivity and chemical stability. Currently no material satisfies these requirements.
The focus of this pilot is to discover and optimise new classes of solid inorganic Li-ion electrolytes with high ionic and low electronic conductivity, and good electrochemical stability.
We have established a protocol where structures are automatically pulled from crystallographic databases (ICSD, proprietary, and COD, open-access), translated into the appropriate self-consistent first-principles calculation (including if desired extensive Car-Parrinello molecular dynamics), and where the simulations are automatically monitored for optimal performance – either with fast screening, e.g. based on an estimate of the minimal energy difference that allows for a percolating channel, or detailed analysis of the trajectories. Thus, computational screening on ionic transport, electronic structure, and electrochemical properties is automatically deployed on thousands of crystalline and amorphous structures, in order to find candidates that would be otherwise impossible to identify by intuition and experiments.
To overcome the size and time scale limitations of first principles molecular dynamics, which hinder any effective screening approach, we develop a new hybrid/ab-initio potential able to describe the physics of lithium diffusion in a vast range of materials, which we refer to as the pinball model. Our main assumption is that the electronic charge density, in a frozen phonon picture, remains fixed during lithium dynamics.
Thanks to a careful study of the evolution of forces provided by the QUANTUM ESPRESSO quantum engine, we optimized our potential with a procedure based on regression and force-matching. A typical result of this protocol is visualized in Fig. 1. The regression parameters mimic the screening effect of charge polarization.
Figure 1: Forces on lithium in Li10GeP2S12 calculated with DFT against the forces calculated with the pinball model, observing excellent agreement. The panels on the top left and bottom right show the error in the forces made (against DFT) for the pinball model in its previously reported level of approximation and the current level, respectively.
We focus on the creation of workflows to run molecular dynamics and compute diffusion coefficients in an automated fashion.
Calculations prior to the molecular dynamics, such as the creation of supercells and cell relaxations, as well as the following analysis are fully automated and publicly available on BitBucket under MIT license.
We have recently undertaken a new screening procedure using the improved version of the pinball model. The diffusion coefficient is calculated for each material at a selected temperature for structures from the Inorganic Crystallography Structure Database and the Crystallography Open Database and are shown in the fig. below. Some of the structures show potentials as solid-state electrolytes.
Figure 2: Diffusion coefficients obtained after a first screening of structures from the ICSD and COD. Even in this small sample a few structures show potentials as solid-state electrolytes.