No abstract

A polysilicon roughening process is developed to reduce the interface impedance of microelectrodes of neural chips. In developing micromachined neural interface systems, one of the basic requirements is to reduce the interface impedance of microelectrodes, because the neuronal signals generally have a very small amplitude and the increased impedance can cause the charge transfer capability of microelectrodes to decrease. The developed process involves forming metal microelectrodes on top of a low pressure chemical vapor deposition (LPCVD) polysilicon film, which is deposited on top of a heavy-phosphorous-content phosphosilicate glass film. The phosphorous inhibits LPCVD polysilicon nucleation and results in very large grains, and hence, very rough film surfaces. This process significantly increases the effective surface area, and the interface impedance can be significantly reduced without increasing the physical size of microelectrodes. By using this process, the interface impedance is significantly lowered. The impedances of conventional gold microelectrodes and the microelectrodes developed in this paper are measured and compared by using a scanning electron microscope, an atomic force microscopy and an impedance spectroscopy system. Experimental results show approximately 50 times lower interface impedance for the developed method.

In this article, a novel and efficient approach for modeling radio-frequency microelectromechanical system (RF MEMS) resonators by using artificial neural network (ANN) modeling is presented. In the proposed methodology, the relationship between physical-input parameters and corresponding electrical-output parameters is obtained by combined circuit/full-wave/ANN modeling. More specifically, in order to predict the electrical responses from a resonator, an analytical representation of the electrical equivalent-network model (EENM) is developed from the well-known electromechanical analogs. Then, the reduced-order, nonlinear, dynamic macromodels from 3D finite-element method (FEM) simulations are generated to provide training, validating, and testing datasets for the ANN model. The developed ANN model provides an accurate prediction of an electrical response for various sets of driving parameters and it is suitable for integration with an RF/microwave circuit simulator. Although the proposed approach is demonstrated on a clamped-clamped (C-C) beam resonator, it can be readily adapted for the analysis of other micromechanical resonators.

No abstract

No abstract