Predicting the Failure Probability of Device Features in MEMS

In the MEMS industry today, device reliability is commonly evaluated by simulating the loads applied to a device using finite element (FE) analysis and setting a threshold for the allowable first principal stress (
S
₁
). This is a potentially misleading practice for the analysis of brittle structures, but it is used because better reliability tools have not been validated and adopted for MEMS design. In this paper, we detail and validate pragmatic FE-based MEMS analysis capabilities: local (device feature) failure probability prediction and surface-specific failure probability intensity (FPI, μm
^-2
) plotting, which quantitatively identify the most probable locations of fracture on a MEMS device and provide a more meaningful and higher contrast visualization versus corresponding plots of
S
₁
. To validate prediction effectiveness, fractographic analysis was completed on micromirrors (a representative MEMS device) loaded in two distinct complex stress states (primarily, tension or torsion) until failure, thus determining the locations of fracture initiation and their relative probabilities. Our method improves the simulation-driven design of brittle microstructures by providing not only the failure probability of the total structure but also probable failure locations as a function of load, all of which are essential for intelligent design iteration.