アイテム詳細

要旨

As the primary anode material of lithium ion batteries, lithium intercalated graphite is one of the central materials behind the transition towards a less CO₂2-intensive energy economy. In spite of that, the atomistic processes governing (dis-)charging cycles, and limiting the speed, safety and reversibility thereof, are still not sufficiently understood. Specifically, diffusion kinetics, relative energetics of stoichiometrically equivalent intercalant-orderings and non-equilibrium phenomena like charge-density gradients due to fast charging speeds require additional research, so that batteries can be further optimized.
In this work, a new semi-empirical Density-Functional Tight-Binding method was parametrized, making use of modern machine-learning for the generation of the repulsion potential. In doing so, an accuracy comparable to that of state-of-the-art, dispersion-corrected Density Functional Theory calculations can be achieved at a fraction of the computational cost. The method was successfully benchmarked against both structural and energetic system properties. Based on it, accurate diffusion barriers, structural properties and the dielectric response were calculated, all in dependence of the state of charge and the semi-local ordering of the charge carriers. At the same time, the process of combining semi-empirical electronic structure based electrostatics with machine-learned repulsion was rigorously investigated and explained – an approach, which provides great promise for many other systems of interest.