0
Select Articles

A Little Knowledge PUBLIC ACCESS

Uninvestigated Factors Govern the Reliability of Micro Devices. What Needs to be Done to Address Them?.

[+] Author Notes

Stuart Brown is director of the Bostoll office of Exponent Inc., an engineering firm.

Mechanical Engineering 125(10), 48-51 (Oct 01, 2003) (4 pages) doi:10.1115/1.2003-OCT-2

This article explains microelectromechanical systems (MEMS) design concepts that are under investigation for their application in different domains. MEMS provide numerous performance advantages. Miniaturization improves packaging and simplifies installation and maintenance. Power consumption can drop dramatically. Analysis shows that appropriate hermeticity and cleanliness remain a challenge, because sufficient contamination to introduce frequency drift may result from the migration of trace contaminants within a package itself. Many reliability issues apply to different MEMS in diverse ways. Work to date indicates that low-stress, hermetically packaged devices pose little concern about crack growth. The increasing maturity of MEMS and the emphasis on reliability represents good news for the industry, as it reflects the natural evolution from initial product design to fabrication technologies to long-term reliability. Despite current challenges, there is no fundamental constraint to reliability improvements as our knowledge of failure mechanisms and countermeasures increases.

Microelectromechanical systems are continuing to replace other technologies for a wide variety of applications, including inertial sensing, signal conditioning, switching, sensing, and biomedicine. MEMS provide numerous performance advantages. Miniaturization improves packaging and simplifies installation and maintenance. Power consumption can drop dramatically. Product lifetime can be extended by combining sensors and signal conditioning within a single package.

The reliability of these MEMS applications can match the technologies they replace. For example, airbag accelerometer lifetimes exceed those of both the distributed sensors and wiring harnesses used previously and will also easily exceed vehicle life. In many cases, the ability to reduce hand assembly and to integrate sensing and electronics translates to immediate improvements in product mean times before failure.

Other applications, however, are challenging limits both of device performance and of our knowledge of the physical phenomena governing reliability. In some cases, reliability is the primary constraint to commercialization.

In the macroscopic world of mechanical switches and relays, product lifetimes are frequently specified in millions of actuation cycles. In the microscopic world of MEMS, radio frequency switches are being fabricated and prototyped for projected actuation lifetimes that can exceed 100 billion cycles-easily pushing four orders of magnitude in expected reliability.

This transition in demand and expectation of reliability is representative of the challenges that MEMS are encountering as they are expected to perform beyond the standard limits of macro products. In such cases, MEMS switch commercialization is revealing new reliability issues that cannot be resolved from previous knowledge of failure mechanisms and macro testing.

It is remarkable that MEMS provides an opportunity for mechanical devices to replace a switching application that was previously met only with solid state devices. Reduced insertion losses and excellent isolation make microswitches obvious replacements for standard solid state switching devices that are very reliable but include certain compromises in electronic performance. Opportunities for replacement of solid state devices include switches. phase shifters, and tunable filters for both commercial and defense applications. The low insertion losses and better isolation translate to lower power consumption, longer battery life, and reduced distortion. The advantages are clear, and many organizations are pursuing different designs.

The challenge for these MEMS switches is reliability. These switches for RF applications can be either direct contact, also called ohmic switches, or capacitive, where high-frequency signals can be directed along different signal paths using conductors that are separated by an insulating dielectric layer. Capacitive switching is possible at high frequencies as signals can jump small distances, removing the need for direct contact.

Ohmic microswitches make direct electrical contact between metal conductors, and wear ultimately limits lifetime. Some performance specifications translate to wear rates of less than one atomic layer per 100,000 make-and-break events. Wear rates this low have not been investigated macroscopically.

Capacitive switches can also experience a limiting phenomenon when the dielectric layer accumulates a static charge. This induced voltage produces a drift in switch behavior and can ultimately cause permanent sticking of the switch.

The remarkable aspect of these failure modes is that reliability data was not available prior to the development of MEMS switches. In the case of ohmic switches, no test data existed on electrodes with 2-square-micrometer contact areas that would make and break for billions of cycles. Questions of material selection, contact pressure, and ideal surface topology have not been investigated in depth. Similarly, although substantial knowledge existed for the charging of dielectrics, relatively little is known about the particular dielectric materials used for MEMS and their voltage histories.

Several initiatives are under way to address MEMS switch reliability. Exponent Inc. has consulted with firms that are currently commercializing MEMS ohmic switches. Capacitive switches, although not as mature commercially, are offering extremely promising reliability.

The Defense Advanced Research Projects Agency has begun initiatives to better understand the physics controlling contact reliability. MEMS switches are in beta testing now as micro relays, and successful resolution of their reliability issues could make them ubiquitous in RF applications.

But there is a challenge, because size matters. Effects that are unrecognized macroscopically become immediately important when devices move on the scale of micrometers.

Static charges and wear rates too small to matter in larger devices can shorten lives of microscale switches (facing page). Natural frequency of l-micrometer resonator (above) can be altered by molecular layer of water.

Grahic Jump LocationStatic charges and wear rates too small to matter in larger devices can shorten lives of microscale switches (facing page). Natural frequency of l-micrometer resonator (above) can be altered by molecular layer of water.

ME MS fatigue device is used to measure very small crack growth rates. A stress concentration on the base promotes crack initiation.

Grahic Jump LocationME MS fatigue device is used to measure very small crack growth rates. A stress concentration on the base promotes crack initiation.

The small size of MEMS immediately magnifies the importance of cleanliness and contamination. This effect is seen throughout MEMS operations, but particularly in resonating applications. Numerous resonating MEMS and nanoscale filter structures have been fabricated to exploit superior bandpass and isolation characteristics. Just as with switches, these micro resonators represent a return to mechanical filters that are competitive in performance with solid state devices. However, these mechanical structures are sensitive to contamination that affects their long-term stability.

Consider a cantilever resonator that is 1 micrometer long, 50 nanometers wide, and 20 nanometers high. Ignoring damping effects, a simple calculation of this cantilever's first order natural frequency yields approximately 27 megahertz. If a single continuous layer of water molecules collects on the outer surface of the resonator, the natural frequency drops by more than 0.6 percent.

This shift is typically excessive for resonator applications. If we shrink the cantilever by a factor of 10 in all dimensions, then a single layer of water would change the natural frequency by 5.5 percent.

Macroscopically, monolayer contaminants do not have a measurable effect. Microscopically, the effect is significant. Some sensing schemes exploit this sensitivity to detect chemicals that will attach to a micro or nanostructure previously coated with a particular chemical's receptor. Mass loading due to the presence of the chemical is easily detected by a change in natural frequency.

Many applications, however, need to avoid any undesirable contamination effects. One method of reducing this sensitivity is to place resonators in a hermetic package. Developers also should include environmental isolation as a part of their design and product planning process.

Appropriate hermeticity and cleanliness remain a challenge, because sufficient contamination to introduce frequency drift may result from the migration of trace contaminants within a package itself.

Another striking example of small changes having large effects is crack growth. Consider the case of a micro pressure sensor used in a flexing bioMEMS cardiovascular application that is intended to last 10 years. Given a typical human heartbeat of 70 beats per minute, a device synchronized to that frequency would flex more than 350 million cycles in that 10-year period. Now suppose that the device has a flexure two micrometers wide with a small surface defect that may grow into a crack. In the macro world, we would certainly consider a component worrisome if a crack grew halfway through its thickness.

Taking one micrometer as the final crack length in the flexure, this one micrometer, over a 10-year period, translates to an average crack growth rate of one atomic distance per day. To our knowledge, this crack growth rate has never been measured.

Exponent and other investigators are using resonating MEMS to measure crack growth rates with an objective of reaching that limit. Using resonating test structures, we can detect small changes in cracks-average growth rates as slow at 300 atoms per day.

Once again, engineers working on a macroscale design would never be concerned about these slow growth rates. However, on a MEMS scale, these crack growth rates are quite important and need to be tested.

Creep rates that could have significant implications on the MEMS scale are so low that engineering has little history of examining them.

Grahic Jump LocationCreep rates that could have significant implications on the MEMS scale are so low that engineering has little history of examining them.

Consider, too, the effect of creep, which is the small-distance rearrangements of atoms resulting from long-term imposed loads and stress. The problem is important because of the continued stress experienced by MEMS devices. For example, many MEMS pressure sensors and suspended structures deliberately include tensile residual stresses. These tensile stresses are introduced during fabrication, both to maintain alignment and to "pull tight" flat structures that would fail if they were allowed to buckle or deflect out of plane.

Creep that we would otherwise ignore becomes important on a MEMS scale because small creep strains can relax those residual stresses and cause failure. Consider a 20-micrometer-long silicon flexure with deliberately introduced 200 MPa tensile stresses that should stay within 20 percent of that stress over 10 years. It would take only 20 atomic distances of additional flexure length to relax those stresses 20 percent, which translates to a very small creep rate of 10-11 per second. Similar to small-scale crack growth, we do not know which materials will be immune from concern about creep, as engineers and researchers have not measured such slow creep rates.

Indeed, we are not sure of the lower temperature limits where we can eliminate creep as a concern. A rule of thumb in macroscopic applications states that significant creep occurs in materials at 0.3 of the homologous temperature, which is the ratio of operating temperature to melting temperature on an absolute temperature scale.

Aluminum, for example, which melts at 933.5 K, would have a homologous temperature of 0.31 at 20°C, or 293.16 K. We should (and do) see creep in aluminum at room temperature. This rule of thumb, however, is based on macroscopically measurable creep rates in materials with relatively large microstructures.

The small creep rates that may be issues for MEMS and nanodevices are unmeasured in many cases. The macroscopic rule of thumb is not dependable, and we do not know what the micro or nano limits should be, as very small grain sizes and diffusion distances may admit creep mechanisms that we would otherwise ignore.

Many reliability issues apply to different MEMS in different ways. Work to date indicates that low-stress, hermetically packaged devices pose little concern about crack growth. Similarly, single crystal silicon devices have not demonstrated issues with creep reliability. In fact, the classic materials of MEMS, namely silicon and silicon nitride, are very resistant to creep and crack growth. Persistent unknowns, however, include long-term performance and reliability limits as operating temperatures change or new materials are incorporated within MEMS.

The increasing maturity of MEMS and the emphasis on reliability represents good news for the industry, as it reflects the natural evolution from initial product design to fabrication technologies to long-term reliability. Despite current challenges, there is no fundamental constraint to reliability improvements as our knowledge of failure mechanisms and countermeasures increases.

We should expect to see similar evolution in nanotechnologies as well, where design concepts are being investigated. Nano applications should be anticipating similar if not greater challenges than MEMS, for many of the phenomena mentioned here will be even more important on a nanoscale. The need for reliability research and development consequently will only grow as MEMS and nanotechnologies become increasingly incorporated into our industrial base.

Creep rates that are normally ignored can result in relaxed stresses in pretension membranes and in buckled structures that fail to perform.

Grahic Jump LocationCreep rates that are normally ignored can result in relaxed stresses in pretension membranes and in buckled structures that fail to perform.

Copyright © 2003 by ASME
View article in PDF format.

References

Figures

Tables

Errata

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In