Artificial Neural Network Based Process Model for Arsenic Diffusion in Si1-xGex Alloys

Abstract

The thermal diffusion behavior of ion-implanted Arsenic (As) in SiGe alloy has been investigated and modeled. This paper introduces a neural network based model consisting of physics-based and process-based parameters for evaluating the effective diffusivity of Arsenic through SiGe accurately. The parameters that served as the input to the neural network included Ge content, diffusion temperature and anneal time. The model was validated for the germanium content of up to 45% with the reported data and the existing simulation models in Silvaco. The model incorporates all the effects associated with the change in the process parameters which affect the diffusivity of As in relaxed-SiGe. The model was found to be extremely accurate in predicting the exact dependencies of As diffusivity on physics-based and process parameters. The proposed empirical process model may find suitable application in prediction of thermal diffusion behavior of As in SiGe process-flow with emphasis on reduced computational time.