Pre and post machining and release residual stresses in microelectromechanical systems (MEMS)

Abstract

A major concern in the development of microelectromechanical systems (MEMS) is the presence of residual stress. This stress, which is produced during the fabrication of multi-layer thin-film structures, can significantly affect the performance of micro-scale devices. Though experimental measurement techniques are accurate, actual stress measurements can vary dramatically from run to run and wafer to wafer. For this reason, the modeling of this stress can be a challenging task. Past work has often focused on experimental, static techniques for determining residual-stress levels in single-layer and bi-layer structures. In addition, in prior studies, the focus has primarily been on residual-stress measurements in thin films as they are being deposited and prior to the release of a particular device. In this effort, residual stresses in MEMS resonators are characterized pre- and post-micro-machining and release of the structures. This is accomplished by applying three residual-stress identification techniques. The first technique, which is based on wafer-bow measurements and Stoney's formula, is suited for determining the residual stresses in thin film layers as they are being deposited and before the occurrence of a micro-machining or release process. In the second technique, a static parametric identification technique, device deflection data is made use of to approximate individual device residual stress immediately after release of a structure. The third technique, a dynamic parametric identification technique, which can be based on linear or nonlinear frequency response data can be used to estimate device residual stress immediately after release and after the device has been polarized. The results obtained by using these techniques are used to develop an understanding of how geometry, fabrication, release and polarization of resonators affect the stress state in a piezoelectric device. The results, which show that the stress levels can be quite different after a device has been released and poled, point to the importance of considering parameter identification schemes such as those described in this effort for identifying residual stresses in multi-layer, micro-structures.