Although appealing, MEMS sensors suffer from uncertainties in the geometrical dimensions, surface topologies, as well as material properties of MEMS devices after production. This affects their behaviour and reduces the production yield, leading to an increase in the manufacturing cost. At the same time, devices such as microphones and accelerometers, used in mobile phones, are under a tremendous price pressure. Since 2004, prices have dropped by 50%. While testing is viewed to be a practical solution to reliability assurance of MEMS, cost for developing effective and reliable testing remains high and is estimated at over 40% of total production costs, mainly because the equipments for tests are really expensive and because the tests are very device specific Already the value added for yield optimization and more efficient test methods is vast.
Numerical models can be used from the design stage to limit these expensive tests. However, due to the multi-physics, multi-scale nature of MEMS and due to the uncertainties in material structures, geometrical dimensions etc, a traditional deterministic approach cannot be used. Thus a stochastic finite-element method (FEM) that uses a sensitivity analysis of the response with respect to the random parameters should be developed. The uncertainty on the properties has to be determined from the uncertainties on the micro-structure of the material (grain size, grain orientation, inclusions…) to characterize the spatial variation of properties at the meso-scale. The stochastic FEM will use these data to compute the macro-scale variation of the structure response, resulting in a 3-scale (micro-meso-macro) methodology.
This project aims at improving the efficiency of the manufacturing process while decreasing the production cost by considering at the design stage the uncertainties in such a way that a range of the MEMS properties can be predicted for the manufactured products. Such a non-deterministic approach can be achieved after introducing the following new achievements as targeted innovation
- Development of original micro-meso-macro stochastic finite element methods
- Database collection of uncertainties for given MEMS devices
- Validation by extensive measurements on produced vibrating micro-sensors
- Evaluation of development/manufacturing/exploitation cost reduction
In particular, validations will be pursued on vibrating micro-sensors, by studying:
-The risk of release stiction of the produced MEMS. Due to their large surface area-to-volume ratio, relatively smooth surfaces properties, and micro/nano scales involved, MEMS are vulnerable to the micro-meter ranged surface forces (capillary, van der Waals, etc), which can lead to permanent adhesion, even at the manufacturing stage (release stiction). Prediction of release stiction risk will account for the uncertainties by using the enhanced contact-adhesion micro-scale model. Released stiction probability predicted will be compared to measurement from beam array productions.
-The statistical range of quality factor actually reached by the manufacturing process. Indeed, quality of the sensors strongly depends on this value, which should reach a minimum threshold. Multi-physics stochastic finite elements (SFE) will be used to determine the quality factor expected range of the micro-sensor, and validation will be achieved by experimental measurements using a laser vibro-meter.