Silicon possesses the capability to replace graphite as the standard active material for negative electrodes in lithium-ion-batteries with high energy density. With its over tenfold higher specific capacity compared to graphite, its abundancy in the earth’s crust and its environmental friendliness it has unequivocal benefits. The use of silicon is however complicated by its large volume change of over 300% during cycling and a significant voltage hysteresis, which lead to decreased energy efficiency and a poor cycle life.
The resulting continuous strain, breakage and re-formation of the Solid Electrolyte Interphase (SEI) on the silicon particles is the largest factor slowing down the widespread adoption of silicon as active material. This is why tremendous effort has focused on nanostructuring of silicon and optimizing SEI properties to improve its cycle life. However, a deeper understanding of how SEI composition and its properties influence electrochemical behavior is essential for the targeted design of silicon-based anodes.
To this end, we combine simulations and a model-based experimental design to investigate the interplay between SEI characteristics, mechanical properties, and electrochemical performance. The experimental approach is based on the simulation model by Köbbing et al. […], which proposes a direct causal correlation between the hysteresis and relaxation behaviour of silicon to the mechanical properties of the silicon shell, namely stiffness, thickness and viscosity. In their simulation they explain the voltage hysteresis with the visco-elastoplastic behaviour of the silicon shell and the mechanical pressure that is applied on the particle as a result.
To experimentally investigate the accuracy of the simulation model and validate the effect of the SEI, the SEI properties were systematically modified by using different electrolytes and formation protocols. Furthermore, the effect of different binders and the evolution of the hysteresis upon cycling were examined.
The experimentally acquired results were compared, based on their respective influence on the SEI, with predictions made by the simulation model.