A novel method for modeling semiconductor by neuro space mapping

Abstract

The drift-diffusion (DD) model is not accurate for simulation of submicrometer semiconductor devices. Using RBF NN, a neuro space mapping is proposed to DD model parameters. The DD model with mapped parameters can produce accurate simulation results of the hydrodynamic model. Simulations of a n-i-n diode confirm the ability of the proposed method for submicron device simulation.