Silicon carbide is one of the most promising wide bandgap semiconductors that offers numerous advantages in power electronics. Superior properties push the performance of electronic devices that can be operated at higher voltages and higher temperatures at reduced power losses compared to conventional semiconductors. Therefore, SiC based electronics can be considered as a key element for future e-mobility applications. The growth of SiC single crystals, however, is much more challenging compare to conventional semiconductors such as e.g. Si. In particular, due to their energetic similarity, a high number of polytypes can be observed.

We present atomistic simulations based on density functional theory (DFT) and interatomic potentials to address SiC polytype energetics. We compare different methods with respect to their prediction of formation energies, lattice parameters and surface energies. Furthermore, we generate a large dataset of defect structures and energies based on DFT which can be used for training interatomic potentials via machine learning approaches. We discuss suitable features for efficiently encoding the geometry and chemistry of the atomistic structures. Our calculations contribute to a better description of SiC polytpye energetics at 0K and elevated temperatures and can serve for optimizing the growth process of high-quality single crystals.