Continued scaling of semiconductor devices requires the formation of ever shallower and more abrupt junctions to avoid short-channel effects. Ultra-low-energy ion beams are currently most widely used to introduce dopants into the silicon substrate. Generally this must be followed by high-temperature thermal treatment to eliminate substrate damage generated by energetic ion bombardment and to electrically activate the injected dopants. During the implantation and post-implantation annealing, the injected dopants exhibit transient enhanced diffusion (TED). The consequent doping profile spreading imposes a great difficulty in forming ultrashallow pn junctions in deep submicron device structures. It is now well understood that native defects such as interstitials and vacancies generated during implantation are mainly responsible for the TED. Due to their large mobility even at room temperature, most of interstitials and vacancies are likely to remain in the form of clusters at the onset of annealing. Therefore, it is necessary to better understand the formation, structure and stability of small interstitial and vacancy clusters, in order to develop a predictive kinetic model for evolution of dopant concentration profiles. In this talk, we present an efficient Monte Carlo-based method for determination of the local-minimum configurations of self-interstitials and vacancy clusters in crystalline silicon. Based on first principles density functional calculations of the defect clusters, this presentation will also focus on discussing their thermal stability and bonding mechanisms. This mechanistic understanding greatly contributes to developing improved kinetic models for changes in defect concentration profiles, and in turn dopant diffusion profiling during post-implantation annealing.