Novel Multi-Scale Modeling Framework for Structure and Transport in Complex Battery Electrolytes
Affordable high energy rechargeable batteries are crucial for further electrification of the transport sector, which is necessary in order to contribute to limit our CO2 emissions to acceptable levels. While today’s lithium-ion batteries (LIBs) have indeed initiated the electrification of the transportation section successfully, electric vehicles are still expensive and typically have ranges limited to ca. 100-500 km depending on price class. There are also safety concerns with LIBs and limited abundance of necessary materials why new chemistries, and especially new electrolytes, need to be explored. Emerging classes of electrolytes, such as highly concentrated electrolytes, have more complex structures than conventional electrolytes, with implications for the ion transport mechanism. This complexity necessitates a multi-scale modeling approach starting at the atomic level to gain further fundamental understanding.
This thesis outlines a framework where ab initio molecular dynamics initially is used to simulate small periodic systems (∼100 - 1000 atoms) over relatively short time spans (∼1 ps) to obtain trajectories that subsequently are used to train the parameters of a classical force field by force matching. This optimization is performed over all parameters simultaneously by a genetic algorithm. The force fields developed are then used to simulate larger systems (∼1000 - 100 000 atoms) over longer time scales classically (∼1 ns - 1μs). The resulting trajectories are used to collect statistics for a hierarchical analysis, which resolves the structure in terms of dynamic clusters, and quantifies the life-time distribution, population dynamics, and transport properties of identified clusters and non-covalent bonds. The method is ultimately to be of general use to both qualitatively and quantitatively elucidate the ion transport mechanism in novel types of electrolytes as a function of composition.
force field development