Time efficient test method for dimensional parameter determination based on resonant mode detection
High production yield and cost reduction are important for mass production of competitive microelectromechanical systems (MEMS). They can be achieved through highly optimized processes, continuous process control and repeated measurements of process dependent properties of the fabricated structures. The identification of faulty dies as early as possible in the fabrication process saves following costs. Considering all fabrication steps, packaging is one of the most expensive. Therefore measurements on wafer level before the packaging task are very important. Therefore a measurement method is needed for wafer level testing of large quantities that is fast, nondestructive and suitable for cleanroom environment. A solution was introduced with the MEMS parameter identification method which is verified and further developed in this work. The method uses measured undamped natural frequencies and natural frequencies obtained from a parametric finite element (FE) simulation. During an optimization procedure structural and material parameters are determined from both results. The results of the FE simulation are fitted to polynomials to get an analytical description of the undamped natural frequencies as a function of the varied parameters. This time consuming procedure has to be done only once for every type of structure, so that the remaining steps of the method are very time efficient. The results of the optimization process are the chosen structural and material properties of the measured structure. The method was tested on the example of MEMS scanners. They consist of a rectangular mirror plate with flexures on one of their long sites. The interesting parameters are thickness and width of the flexures and the mechanical stress in the structure. Polynomials are calculated from the first, third, fifth and sixth vibration mode. They are chosen because the influence of the unknown parameters onto the corresponding natural frequencies is most different for these modes. The identification results are verified through SEM measurements. Single mirrors from different arrays and different positions on the arrays are extracted from the scanner chips and measured with SEM. The results are displayed and show a good agreement with the identification results. The entire method is implemented in a standalone software.