Fully electronic method of measuring post-release gap and gradient/residual stress of a mems cantilever

Andrew Stephen Kovacs, Purdue University

Abstract

Smartphones and other wireless devices have become ubiquitous over the past decade, and the RF front-end inside of them has become more complex and disproportionately consumes more power compared to other components. Micro-electromechanical systems (MEMS) have a huge potential to reduce these problems while simultaneously offering superior performance compared to current leading-edge technology. However, MEMS technology has difficulty transitioning from the lab to large-scale manufacturing due to the unpredictability of device lifetime and manufacturability issues. This can be mitigated by investigating how critical material or physical parameters (gap, stress, Young's modulus, material thickness, etc.) vary from manufacturing uncertainties and how they change during a device's repeated use. State-of-the-art methods used to measure these parameters are limited by the fact that they must be made optically, which is slow and hampered by opaque packaging used to protect the device. This work presents a method of extracting the post-release gap and either the residual gradient stress for nickel-electroplated MEMS cantilevers, or the residual mean stress of a gold-electroplated MEMS cantilevers using fully electronic methods. The device structure consists of a cantilever beam anchored onto an insulating substrate with two equally-sized contact pads distributed length-wise underneath the beam. The extraction algorithm relies on Finite Element Analysis (FEA) models to simulate the capacitance between the cantilever and the two contacts for a range of gaps and tip deflections. The tip deflections relate either a range of radii of curvature for residual gradient stress extraction, or a range of beam deflection angles with respect to the substrate for residual mean stress extraction. The simulated capacitances are used to create a look-up table to match the measured capacitances to a unique gap and tip deflection. The optically-measured physical parameters, and the electronic capacitive measurement of a calibration device are used in conjunction with the mesh to calculate parasitic capacitance values. With parasitics known, an arbitrary number of subsequent devices can be measured electronically and their corresponding gap and tip deflection extracted. The extracted data is validated with optical measurements from both a confocal laser microscope, and a Laser Doppler Vibrometer (LDV). In addition, the sensitivity of the algorithm to uncertainties in physical parameters are examined. This include: uncertainties in the dielectric constant of the substrate, the thickness and dielectric constant of the material covering the two contacts, the cantilever thickness, and the thickness of the gold comprising the contacts. The statistical variation of these parameters was quantified and then the sensitivity of the model to these variations is examined. The uncertainties created predictable, but small uncertainties in the final extracted parameters of interest. The extracted gap and tip deflection is then used to compute the range of the cantilever's radius of curvature, or its deflection angle, which in turn is used to find the residual gradient and residual mean stress, respectively. A total of eight nickel-electroplated devices and five gold-electroplated devices were measured and characterized for gap and residual gradient and residual mean stress. Good agreement was found between the optically and electronically measured values even when accounting for all known uncertainties. This methodology has been demonstrated to be suitable for a foundry mass production line due to its potential high speed of measurement and accuracy, and minimal reliance on optical observations.

Degree

Ph.D.

Advisors

Peroulis, Purdue University.

Subject Area

Mechanics|Electrical engineering|Meteorology

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS