Phosphorus-defect interactions during thermal annealing of ion implanted silicon

Ion implantation of dopant atoms into silicon generates nonequilibrium levels of crystal defects that can lead to the detrimental effects of transient enhanced diffusion (TED), incomplete dopant activation, and p-n junction leakage. In order to control these effects, it is vital to have a clear understanding of dopant-defect interactions and develop models that account for these interactions. This research focuses on experimentally investigating and modeling the clustering of phosphorus dopant atoms with silicon interstitials.; Damage recovery of 40keV Si+ implants in phosphorus doped wells is experimentally analyzed. The effects of background phosphorus concentration, self implant dose, and anneal temperature are investigated. Phosphorus concentrations ranging from 2.0 × 1017 to 4.0 × 1019 cm−3 and Si+ doses ranging from 5.0 × 1013 cm−2 to 2.0 × 1014 cm−2 are studied during 650–800°C anneals. A dramatic reduction in the number of interstitials bound in {lcub}311{rcub} defects with increasing phosphorus background concentration is observed. It is suggested that the reduction of interstitials in {lcub}311{rcub} defects at high phosphorus concentrations is due to the formation of phosphorus-interstitial clusters (PICs). The critical concentration for clustering (approximately 1.0 × 1019 cm−3 at 750°C) is strongly temperature dependent and in close agreement with the kink concentration of phosphorus diffusion.; Information gained from these “well experiments” is applied to the study of direct phosphorus implantation. An experimental study is conducted on 40keV phosphorus implanted to a dose of 1.0 × 1014 cm−2 during 650–800°C anneals. Electrically inactive PICs are shown to form at concentrations below the solid solubility limit due to high interstitial supersaturations. Data useful for developing a model to accurately predict phosphorus diffusion under nonequilibrium conditions are extracted from the experimental results.; A cluster-mediated diffusion model is developed using the Florida Object Oriented Process Simulator (FLOOPS). The nucleation of defects is controlled by the diffusion-limited competition for excess interstitials between PICs and {lcub}311{rcub} clusters. The release of interstitials is driven by cluster dissolution. Modeling results show a strong correlation to those experimentally observed over a wide temporal and thermal domain using a single set of parameters. Improvements in process simulator accuracy are demonstrated with respect to dopant activation, TED, and dose loss.