Parameter sensitivity analysis applied to modeling transient enhanced diffusion and activation of boron in silicon

Transient enhanced diffusion (TED) during annealing of implanted boron in silicon greatly impedes the formation of junctions that are sufficiently shallow for advanced complementary metal oxide semiconductor devices. For reasons of cost and efficiency in process development, detailed modeling of TED is often used for designing annealing procedures. However, because model validation depends primarily on fitting experimental dopant profiles, and because realistic models contain dozens of parameters that are poorly known, development of a unique set of rate parameters with true predictive capability has proven elusive. Here we have employed formal parameter sensitivity analysis by the finite difference method to show that the activation energies most critical to know accurately are those for interstitial boron diffusion, kick-in, and dissociation of the (B_s - Si_i)⁺ complex to liberate either interstitial B (kick-out) or Si. Maximum likelihood estimation is also applied to the literature for interstitial cluster formation to determine a most likely set of activation energies for cluster dissociation.