This work presents a multiscale approach to understanding the defect formation during the evolution of solid-phase epitaxy regrowth in Si. A molecular dynamics (MD) simulation technique has been used to elucidate the defect formation mechanisms, as well as to determine their nature. A hybrid lattice kinetic Monte Carlo (LKMC)-finite element method (FEM) model fed by the outcome of MD was subsequently implemented. It scales up the simulation times and sizes, while reproducing the important features of the defected regrowth predicted previously. FEM calculations provide the strain pattern due to the density variation between the amorphous and crystalline phases, which is then taken into account by the LKMC model by including the effect of the strain in the rates of recrystallization. Overall, this multiscale modeling provides a physical explanation of the generation of defects and its relation with the presence of strain. The model also captures the character of formed defects. It distinguishes two types: twins formed at {1 1 1} planes and dislocations produced by the collapse of the two recrystallization fronts. Simulation results are validated by comparing them with significant experiments reported in the literature.

The shear coupled motion of grain boundaries (GBs) is modelled by using two different atomistic simulation techniques: molecular dynamics (MD) and kinetic Monte Carlo (KMC). MD simulations are conducted to identify the elementary mechanisms that take place during the coupled motion of GBs. This process is described on the one hand, in terms of the geometrical approach of the dislocation content in the boundary; and on the other hand, by the thermodynamics of the dislocation passage, shown as a thermal activated process. Relevant MD output is extended into a KMC model that considers the GB migration as a result of a sequence of discrete rare events. The independent motion of each structural unit forming the boundary conforms a single event, having a rate per unit of time to move to the next stable position computed according to the transition state theory. The limited time scale of classical MD is overcome by KMC, that allows to impose realistic deformation velocities up to 10 lm/s.

Current generations of FinFET devices are incorporating SiGe alloys as stressor material in channel regions in order to enhance hole mobility, drive current and channel conductivity. However, the presence of SiGe gives rise to new issues to be controlled during the device fabrication process, such as the strain retention or defectivity control, as they may seriously impact the quality of the strained SiGe channel and so the final device performance. The present work addresses the study of defect formation during the optimized integration of SiGe FinFETs for 10 nm technology nodes, aimed at determining the Ge threshold content for the nucleation of defects. Due to the relevance of atomistic models in determining the mechanisms and nature of defect formation, molecular dynamics (MD) simulations have been performed to emulate the FinFET fabrication process. The channel region is generated by removing a portion of the total volume of the Si substrate, further refilled with Si 1 − x Ge x alloy to form the co-integrated FinFETs. After deposition, the presence of Ge induces a lattice distortion which is expected to be relieved by defect formation. Samples are annealed varying the Ge fraction, allowing to determine that threshold Ge content for the nucleation of defects is x = 0.27. MD provides also the nature of the formed defects, which have been suggested to be twinning developed at {111} planes and 60• misfit dislocations. Simulation results have been compared to experimental observations, both in good agreement.

The solid phase epitaxial regrowth (SPER) of SiGe alloys has been studied using atomistic simulation techniques. Molecular Dynamics (MD) simulations reproduce the recrystallization process of amorphous structures created in two different ways: introducing atoms at random positions according to the crystalline density and carefully relaxing the structure; and using a bond switching algorithm by means of ab initio. Activation energies are confronted, and the first method is validated as an efficient way to generate amorphous-crystalline structures suitable to study SPER processes. The MD extracted results show that the SPER rate does not vary monotonically with the Ge composition; instead, activation energies reveal a non-linear behaviour with the addition of Ge, due to the two-part behaviour of the SPER rate: SPER rate itself and a hypothesized extra strain due to the bond length difference. Since SPER is a thermally activated process, nudged elastic band calculations are carried out in order to extend the previous assumption. The energy barrier for an atom to attach to the crystalline phase is computed. The extracted values confirm the presence of the mentioned strain contribution required for an atom to recrystallize when it is not as the same type of the bulk.

No abstract