Ab initio structural prediction of interface and defect reconstructions in group IV crystals

In this thesis, we develop a constrained structural prediction approach for the structural reconstructions of interfaces and point defects in crystals. First, we investigate the structure-property relations for a large and varied family of symmetric and asymmetric tilt grain boundaries in polycrystalline silicon. We find a rich polymorphism in the interface reconstructions, with recurring bonding patterns that we classify in increasing energetic order. We systematically classify the grain boundaries into different types by the structural reconstructions when atoms are removed or added at the interface. Then, we extend the low-energy structures from silicon to other group-IV elementary crystals (carbon, germanium, and tin) and study their stability. A clear relation between bonding patterns and electrically active grain boundary states is unveiled and discussed. Second, we study the reconstructions of diamond interfaces along different directions and focus on the formation of diamond-graphite hybrid structures. We find the number of the graphite layers is limited by the distances between the bonded atoms on the diamond surface. How the electronic properties of diamond are affected by the graphitization is also discussed. Finally, we extend our approach to the study of point defect geometries in hexagonal silicon. We obtain among the lowest-energy structures the hexagonal counterparts of all known defects of cubic silicon, together with other often more complex geometries. Furthermore, due to the reduced symmetry, formation energies can depend on the orientation of the defect with respect to the axis. The density of states of the defective supercells is calculated to determine which defects lead to electronic states in the band gap, potentially affecting the performance of optoelectronic devices based on hexagonal group-IV crystals.