| This thesis consists of four scientific articles concerning atomistic numerical methods and their use to simulate semi-conducting systems where nanometer-scale structures play a crucial role.;We introduce accelerated methods designed to study systems driven by activated events. Afterwards, our first article presents, in depth, the kinetic Activation-Relaxation Technique (kART), an off-lattice, self-learning kinetic Monte Carlo algorithm based on the Activation-Relaxation Technique nouveau (ARTn). This method permits the exact treatment of elastic effects in materials over time-scales reaching one second.;This algorithmic development, combined to recent empirical data, forms the basis of our second article. We explain the origin of heat release by self-implanted crystalline silicon in nanocalorimetry experiments after 3 keV ion bombardment, with the help of kART simulations. We show that the structural relaxation is described by a two-step "Replenish-and-Relax" model. This model is quite general and can potentially explain relaxation in other disordered materials. In the next chapter, i.e. the third article, we push the analysis further and give a complete atomistic description of the mechanisms responsible for structural relaxation during the anneal. We show that punctual defects and small defects complexes control the relaxation, in net contrast with the literature that identify "amorphous pockets" as the drivers of relaxation.;Finally, we study some aspects related to the growth of Ge quantum dots on Si (001). After short chapters explaining the scientific context of this work and methodological details, our fourth article concerns the wetting layer formed by Ge deposition on Si (001), using a QM/MM implementation of the bigDFT-ART code. We characterize the 2xN surface reconstruction atomistic structure and decrease the minimum temperature at which deep Ge intermixing is predicted by ab initio calculations by more than 100 K. |
