Quantum-accurate SiC growth simulations with atomic-cluster expansion

• Applicant: Privatdozent Dr. Thomas Hammerschmidt
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 546339066

The climate crisis fuels the need for the widespread generation of renewable energy and the electrification of mobility. This requires optimized materials for power electronics to handle the distribution of electrical power between supply, storage and consumption. One of the most suitable materials for precise control and high efficiency of switching electrical power is the IV-IV semiconductor silicon-carbide (SiC). The fabrication of SiC wafers at an industrial scale is challenging due to the formation of several kinds of defects that limit the device performance. Commonly observed planar defects are switches between the numerous different polytypes of SiC. The crystal structures of the SiC polytypes differ only in the stacking order of the SiC-bilayers. Due to this structural similarity, they can exhibit very similar formation energy and compete during synthesis. The typical line defects are threading dislocations and micropipes that cut the Si-C wafer vertically as well as basal plane dislocations parallel to the growth plane. The goal of this project is to understand the atomistic processes during physical vapor transport (PVT) growth of SiC in order to optimize the fabrication of SiC for power-electronic devices. The focus is the initial stage of defect formation that is determined by processes at the atomistic level. The approach are static and dynamic atomistic simulations of SiC growth in the presence of surface and sub-surface defects. This includes the surface evolution at finite temperatures as well as the adsorption, desorption and diffusion of Si, C and Si-C on the surface. The required representation of the interatomic interaction needs to meet the antipodal requirements of a sufficiently accurate description of the local interatomic bond and a sufficient computational efficiency for simulating extended time and length scales. The Si-C system is particularly challenging in this respect as the energy differences of the SiC polytypes that compete during growth are of the order of meV while the size of the experimentally observed defects is of the order of μm. These demands can currently only be met by machine-learning interatomic potentials (MLIP) that are able to deliver an meV-accurate interpolation of punctual DFT calculations of the potential-energy surface. One of the most efficient and accurate MLIP at present is the atomic cluster expansion (ACE). The goal of this project is to develop an ACE model for the Si-C system and to apply it in atomistic simulations of SiC PVT growth in order to reveal the atomistic processes that lead to the formation of the experimentally observed defects.

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr. Ralf Drautz