Acceleration of dormant storage effects to address the reliability of silicon surface micromachined Micro-Electro-Mechanical Systems (MEMS).

Cox, James V. ; Candelaria, Sam A. ; Dugger, Michael Thomas ; Duesterhaus, Michelle Ann ; Tanner, Danelle Mary ; Timpe, Shannon J. ; Ohlhausen, James Anthony ; Skousen, Troy J. ; Jenkins, Mark W. ; Jokiel, Bernhard, Jr. ; Walraven, Jeremy Allen ; Parson, Ted Blair

Sandia National Laboratories

关键词: Adhesion;    Aging;    Nuclear Weapon Stockpile.-Reliability;    Failure Mode Analysis Materials-Aging.;    36 Materials Science;   

DOI  :  10.2172/923082
RP-ID  :  SAND2006-3362
RP-ID  :  AC04-94AL85000
RP-ID  :  923082

美国|英语

来源: UNT Digital Library

【 摘 要 】

Qualification of microsystems for weapon applications is critically dependent on our ability to build confidence in their performance, by predicting the evolution of their behavior over time in the stockpile. The objective of this work was to accelerate aging mechanisms operative in surface micromachined silicon microelectromechanical systems (MEMS) with contacting surfaces that are stored for many years prior to use, to determine the effects of aging on reliability, and relate those effects to changes in the behavior of interfaces. Hence the main focus was on 'dormant' storage effects on the reliability of devices having mechanical contacts, the first time they must move. A large number ({approx}1000) of modules containing prototype devices and diagnostic structures were packaged using the best available processes for simple electromechanical devices. The packaging processes evolved during the project to better protect surfaces from exposure to contaminants and water vapor. Packages were subjected to accelerated aging and stress tests to explore dormancy and operational environment effects on reliability and performance. Functional tests and quantitative measurements of adhesion and friction demonstrated that the main failure mechanism during dormant storage is change in adhesion and friction, precipitated by loss of the fluorinated monolayer applied after fabrication. The data indicate that damage to the monolayer can occur at water vapor concentrations as low as 500 ppm inside the package. The most common type of failure was attributed to surfaces that were in direct contact during aging. The application of quantitative methods for monolayer lubricant analysis showed that even though the coverage of vapor-deposited monolayers is generally very uniform, even on hidden surfaces, locations of intimate contact can be significantly depleted in initial concentration of lubricating molecules. These areas represent defects in the film prone to adsorption of water or contaminants that can cause movable structures to adhere. These analysis methods also indicated significant variability in the coverage of lubricating molecules from one coating process to another, even for identical processing conditions. The variability was due to residual molecules left in the deposition chamber after incomplete cleaning. The coating process was modified to result in improved uniformity and total coverage. Still, a direct correlation was found between the resulting static friction behavior of MEMS interfaces, and the absolute monolayer coverage. While experimental results indicated that many devices would fail to start after aging, the modeling approach used here predicted that all the devices should start. Adhesion modeling based upon values of adhesion energy from cantilever beams is therefore inadequate. Material deposition that bridged gaps was observed in some devices, and potentially inhibits start-up more than the adhesion model indicates. Advances were made in our ability to model MEMS devices, but additional combined experimental-modeling studies will be needed to advance the work to a point of providing predictive capability. The methodology developed here should prove useful in future assessments of device aging, however. Namely, it consisted of measuring interface properties, determining how they change with time, developing a model of device behavior incorporating interface behavior, and then using the age-aware interface behavior model to predict device function.