Physics and modeling of ion implantation induced transient deactivation and diffusion processes in boron doped silicon

The economics of silicon processing requires predictive modeling capabilities for the continued rapid advancement of semiconductor technology. This is because it has become prohibitively expensive to develop a new process by running a large series of test lots through multi-billion dollar fabrication facilities. Effective process modeling requires an accurate physical understanding of the various interacting processes. The complexity of this problem is compounded by highly non-equilibrium phenomena associated with IC fabrication processes such as implantation annealing. Point defect supersaturations of many orders of magnitude are introduced following ion implantation, which is used to introduce the dopants into silicon. Such supersaturations dramatically alter the diffusion of dopants and reduce the electrical activation during the initial phase of the anneal.; Boron is the primary p-type dopant used in silicon and thus understanding and modeling its deactivation/activation and diffusion is critical to predictive process simulation. Since boron is smaller than silicon, boron agglomerates with interstitials becoming electrically inactive. Modeling of boron clusters is complicated, as there is a huge array of potential boron-interstitial cluster compositions. A physical model for boron clustering is derived by identifying dominant clusters and rate limiting steps via atomistic calculations performed at Lawrence Livermore National Labs. The model is then used successfully to match a wide variety of chemical and electrical data. We further apply this model to understand and successfully predict ultra shallow junction formation. We find it is possible to explain some intriguing phenomenon observed during the formation of ultra shallow junctions, like saturation in junction depth despite increasing ramp-up rates.; Researchers are exploring novel experimental processing steps like high energy Si pre-implants to produce highly active and shallow B junctions. To understand and model this phenomenon it is essential to be able to model the evolution of the excess vacancy region formed after the implant. Hence we have developed a detailed model for vacancy clustering built based on atomistic calculations and verified the model by comparison to data from gold labeling experiments. However this model is computationally intractable for routine use in TCAD (Technology Computer Aided Design). Hence, a computationally efficient physical model has been derived from this full model that can be readily incorporated into process simulators.; This research leads to an understanding of a broad range of data applicable to both current and future generation of devices and present a self-consistent model which can be incorporated into TCAD diffusion equation solvers.