Low-frequency bulk and surface generation-recombination noise simulations of semiconductor devices.

摘要

Analysis of semiconductor device noise in the past has generally been separated into two parts. In the first part one tries to understand which type of noise is the dominant noise source and derive its local noise strength in each differential volume. The second part deals with deriving the transfer function, either the trans-conductance or the trans-admittance (depending on the circuit bias condition at the contact), to couple the local noise strength from each differential volume to the external contact. Integrating the product of the local noise strength and the transfer function over the device volume will produce the overall noise at the specified terminal. This type of analysis can only be done to a certain extent, since deriving the analytical expression of the transfer function is limited to only a few simple cases and device structures. To expand the capability of predicting device noise through this type of analysis, we developed a partial differential equation (PDE)-based, generalized-scheme two-dimensional semiconductor device-simulation tool to perform the transfer function calculation.; The purpose of this research is to use a computer-aided device simulation tool to analyze and simulate low frequency semiconductor device noise. To investigate the low-frequency noise, the bulk generation-recombination noise mechanism was implemented. Since this type of noise can only explain Lorentzian spectra observed in resistors but not the 1/f-like noise in MOS devices, interface generation-recombination and the McWhorter-type oxide trapping noise mechanisms were added. The simulation results verified that oxide trapping noise (caused by the carrier fluctuations in the channel because of carrier tunneling between the trap centers in the oxide and at the interface) is indeed the source of the 1/f noise observed in MOS devices.; One application of this tool is noise defect spectroscopy by modeling the defect density distribution in the oxide through inverse engineering. By adjusting the defect density in the bulk and in the oxide to match the noise simulations with their corresponding measured data, one can obtain the defect density profile of the device. In addition, this tool can be used to predict the maximum amount of defects allowed to guarantee excess noise-free device operation.