Integrated electromagnetic/thermal/mechanical analysis and optimization design of RF-MEMS switches.

Abstract

RF MEMS switches have demonstrated high isolation, low insertion loss, low power consumption, and good linearity over conventional <italic>p-n</italic> diode or FET switches. Despite these technological promises, there are major drawbacks with RF MEMS switches, which include low power handling capability, high pull-in voltage requirements, and long switching times. These drawbacks tremendously slow down MEMS switch commercialization. To design reliable RF MEMS devices, it is important to carefully study the electrical, thermal and mechanical characteristics in an integrated manner. As such, we must resort to well-formulated analytical and numerical tools. This thesis presents an efficient integrated method to analyze electromagnetic, thermal and mechanical properties of RF MEMS switches. In addition to the commonly-known challenges associated with the integration of multi-physics principles for MEMS device analysis, there are also numerical issues due to the extremely small (less than lambda/1000) electrical/physical geometrical details. The corresponding matrix system for EM analysis has very high conditional number. Thus, satisfactory accuracy is difficult to attain using available commercial software tools. Modeling a switch within its surrounding environment using traditional methods leads to a huge and impractical number of unknowns. Also, they use of large element sizes and the mixture of large and very small elements implies additional challenges. Further, while the beam is in motion while being turned on and off, any modeling scheme that meshes the volume within and under the beam requires regridding at each movement stage. Current FDTD implementation is therefore poorly-suited to model the beam deflection of the beam in the down state. To address this issue, we introduce a new method, referred to as the extended finite-element boundary-integral (EFE-BI) method, to simulate the EM characteristic of RF MEMS switches. By employing the conventional FE-BI method on the substrate portion and the moment method for the beam surface, we overcome the issues associated with the small spacing between the beam and the substrate. Also beam's movement is handled efficiently. To solve the resulting EFE-BI system, efficient and accurate semi-analytical integration and a Large-Eigenvalue-Sparse Preconditioner (LESP) are also presented. Another important contribution of this thesis is the seamlessly integration of EM and thermal analysis of RF MEMS switches, to predict the failure mechanisms (such as creep and buckling). Using this integrated modeling, a optimization design of RF MEMS switches is presented to achieve high power handling capability, low pull-down voltage and short release time. A given design study provides some practical rules for RF MEMS switch designers.