It was recently experimentally discovered that the transport properties of liquid battery electrolytes, based on the localized high concentration formulations, could be greatly improved through increasing the number of types of solvents or anions, without significantly affecting the solvation thermodynamics. Exactly why the conductivity can be increased by a factor or 2, without changing the interfacial stability and other properties, has not been understood, however.
In order to understand the mechanisms and generality behind these observations, atomistic molecular dynamics is required to provide insight. Directly simulating these systems with ab initio molecular dynamics is not possible because of the number of components required and the time scales involved, and empirical force fields have known accuracy shortcomings for predicting transport properties. Therefore, the aim is to develop a transferable and accurate machine learning (ML) interatomic potential to discover the underlying mechanisms that give rise to the improved transport properties. This will be achieved through using pre-trained ML models to sample structures and accurate DFT functionals in order to have a state-of-the-art ML interatomic potential that can be used for any composition within the studied chemistries, and over a a wide range of operating temperatures and pressures. The predictions of these simulations will be compared against experiments, such as transference numbers from electrophoretic NMR. High entropy electrolytes have also been reported to have greater operating temperatures, but similar solid electrolyte interphase formation at electrified interfaces, as conventional battery electrolytes. These are also areas, the phase behaviour and reactions at interfaces, where the developed ML force field could applied to provide atomistic insights.
For more information contact Dr Zachary Goodwin or Prof. Jonathan Yates.