Authors: J.S. Nelson, A.F. Wright and P.A. Schultz
Affilation: Sandia National Laboratory, United States
Pages: 367 - 368
Keywords: at-inition, TCAD, dopand diffusion, silicon
The rapid pace of the silicon microelectronics industry, and its need for physics-based TCAD models of dopant diffusion, is coinciding with the tremendous algorithmic and computational advances occurring within modern ab initio electronic structure methods. These methods have evolved to a level where complex diffusion processes occurring during implantation, and thermal and chemical processing can be accurately calculated. Furthermore, they provide detailed mechanistic insight often difficult to extract from experimental measurements. To illustrate these important advances, we will present results for defectmediated dopant diffusion including quantitative predictions of activation energies, qualitative trends with dopant species, and insight into new dopant diffusion mechanisms. The calculations are based on ab initio pseudopotentials using the local-density approximation (LDA) and the generalized-gradient approximation (GGA) to density functional theory. For vacancy-mediated dopant diffusion, both the first-neighbor binding energies (Fig. 1) and exchange barriers (Fig. 2) are in good agreement with available experimental data and exhibit a systematic trend with dopant. The magnitude of the dopant-vacancy exchange barrier and shape of the interaction potentials (Figs. 3 and 4) are considerably different from what is used in current TCAD models for vacancy-mediated dopant diffusion. For interstitial-mediated phosphorus diffusion we find good agreement with measured activation energies (3.4 eV - 3.6 eV). The formation energy (neutral) for a phosphorus atom in the tetrahedral (T)-, hexagonal (H)-, and  split (X)interstitial sites are 4.10 eV, 3.1 eV, and 3.0 eV, respectively. The lowest energy migration path (neutral charge state) is along the X-H-X path with a migration energy of 0.2 eV and an activation energy of 3.3 eV. In summary, this presentation will show how ab initio level calculations can be used to further the development of physics-based Si process simulators.