Maximum A Posteriori Estimation of Energetics in Silicon Self-diffusion

Maximum A Posteriori (MAP) estimation was employed to refine estimates of various energetics in silicon self-diffusion. Self-diffusion profiles were augmented with prior parameter estimates determined from previous experimental systems to obtain improved values. Parameter sensitivity analysis was utilized before MAP estimation to determine the relative importance of model parameters. The sensitivities were a strong function of the surface boundary condition. For conditions with high surface loss flux, the energy for exchange between an interstitial and the lattice is the most critical. For conditions with a low surface loss flux, the dissociation energy of large-atom clusters plays a more important role. The incorporation of information from the new experimental system substantially improved the estimates of parameters that could not be identified accurately in previously studied systems.