Application of ab-initio calculations to modeling of nanoscale diffusion and activation in silicon
As ULSI devices enter the nanoscale, ultra-shallow and highly electrically active junctions become necessary. New materials and 3D device structures as well as new process technologies are under exploration to meet the requirements of future devices. A detailed understanding of the atomistic mechanisms of point-defect/dopant interactions which govern diffusion and activation behavior is required to overcome the challenges in building these devices. This dissertation describes how ab-initio calculations can be used to develop physical models of diffusion and activation in silicon. A hierarchy of approaches (ab-initio, kinetic lattice Monte Carlo, continuum) is used to bridge the gaps in time scale and system size between atomistic calculations and nanoscale devices. This modeling approach is demonstrated by investigating two very different challenges in process technology: F co-implantation and stress effects on dopant diffusion/activation.In the first application, ab-initio calculations are used to understand anomalous F diffusion behavior. A set of strongly bound fluorine vacancy complexes (FnVm ) were found. The decoration of vacancies/dangling silicon bonds by fluorine leads to fluorine accumulating in vacancy rich regions, which explains the fluorine redistribution behavior reported experimentally. The revealed interactions of F with point-defects explain the benefits of F co-implantation for B and P activation and diffusion. Based on the insight gained, a simplified F diffusion model at the continuum level (50--100 nm scale) is extracted that accounts for co-implantation effects on B and P for various implant energies and doses.The second application addresses the effect of stress on point-defect/dopant equilibrium concentration, diffusion, and activation. A methodology is developed to extract detailed stress effects from ab-initio calculations. The approach is used to extract induced strains and elasticity tensors for various defects and impurities in order to predict the impact of arbitrary stress tensors (i.e., not just hydrostatic case). The results from first-principles calculations are used in lattice Monte Carlo simulations to quantify stress-driven anisotropies in I and B diffusion. The result is a prediction of strongly anisotropic diffusion of B (as well as I) under biaxial strain. Following the same methodology, a stress dependent B solubility model was developed which predicts large enhancements of B solubility under compressive stress conditions.
- Physics