Sandia LabNews

Sandia's precision MEMS reliability tests advancefuture of micromachine systems

Image of back to Lab Newscontents page

Microelectromechanical systems (MEMS), those microscopic marvels that promise to revolutionize the electronics industry, are useless unless they are reliable.

So says Bill Miller, Manager of Reliability Physics Dept. 1728, whose 18-member group is charged with determining the reliability of Sandia’s MEMS.

Danelle Tanner (1728) tests MEMS reliability using one-of-a-kind SHiMMER instrument. (Photo by Randy Montoya)

"We constantly ask the questions – how reliable are they and can we use them in real applications," Bill says.

Now, thanks to instrumentation designed and used by colleagues Norm Smith and Danelle Tanner (both 1728) and a physics-based predictive reliability model developed by Bill, MEMS’ reliability is being tested with results so precise that they have captured the attention of an international audience.

In fact, Bill was one of six invited speakers in May at the fourth International Reliability Challenge in Dublin, Ireland, where he presented the research being conducted by his group.

MEMS, complex machines with micron-size features, can be found in everything from computer game joy sticks to car airbag sensors and from inkjet printers to projection displays. They are so small that they are imperceptible to the human eye and have moving parts no bigger than a grain of pollen.

MEMS future of electronics

"MEMS are the future of electronics," Bill says, "I liken it to the integrated circuit industry at the end of the 1960s – a lot of potential, but no one really knew how big it could get then."

In 1989 MEMS were laboratory curiosities with low power, short lifetimes, and few practical proposed uses. Today, they have taken major roles in several industries and are becoming sophisticated devices that can "think, sense, act, and communicate," Bill says. Industry experts believe that the market for MEMS will grow to more than $30 billion worldwide by early in the next century.

Almost from the beginning, Sandia has been a recognized leader in this emerging field, developing unique technologies that lead to complex mechanical systems on a chip and the integration of these mechanical systems with on-chip control electronics. While Sandia’s interest in MEMS stemmed from their potential use in weapons, the spillover to the commercial arena was inevitable.

"It soon became apparent that one of the greatest challenges for the successful commercialization of this new technology is in proving its reliability," Bill says.

This is true for four reasons, he adds. First, many of the promising applications of MEMS will be in critical systems where the cost of failure is very high. Second, MEMS are a new technology with potentially new and poorly understood failure mechanisms. Third, MEMS technology continues to evolve at a rapid rate. And fourth, design tradeoffs must account for reliability, lest warranty costs grow.

When Bill’s department started MEMS reliability testing two years ago, the general belief was that polysilicon, the material from which micromachines are made, is very brittle. Most reliability concerns centered around material fractures.

The MEMS reliability team led by Danelle did not automatically make that assumption and proceeded to find out the real causes of failure in the micromachines.

"What we have done, which is totally new, is ask the question – how do they really fail. No one had any real data," Bill recalls.

Previous reliability tests had all been done on small scales, looking at one, two, or three devices at a time. The team studied hundreds simultaneously.

SHiMMeR tests micromachines’ reliability

The first step was to build an instrument in which to test the micromachines. The job was done by Norm, who designed and constructed SHiMMeR (Sandia High-volume Micromachine Measurement of Reliability), a Plexiglas enclosure that contains a base for testing as many as 256 MEMS parts at a time and a high-powered optical microscope and video camera to observe and record the failures. Each of the MEMS devices is attached to cables through which signals are sent to activate the micromachines. Humidity, believed to be a major factor in MEMS failure, can be controlled in SHiMMeR. The researchers use the microscope to observe the micromachines operating.

SHiMMeR is a one-of-a-kind machine. Because no commercial system was available to do this type of work, the Sandians had to hand-build the machine. Bill says scientists from around the world come to Sandia to observe SHiMMeR.

By running the micromachines until they break – the current record is seven billion revolutions – and then cutting a cross-section through the gears with a focused ion beam and looking at them through the microscope, Bill and his colleagues could also get a good idea about what makes MEMS fail and when and where the failure occurs.

What they discovered statistically last September, after nearly 1 1/2 years of testing and compiling data, was quite different from previous expectations that microengines malfunctioned because of polysilicon fractures.

"We found they fail by wear, much like a car engine fails without oil. The individual parts get so worn that they jam. The occasional fractures we encountered were the result of the materials wearing thin," Bill says.

Although there are many types of wear that may contribute to MEMS’ failure, the one on which Bill focused was adhesive wear, which involves parts rubbing and causing small pieces to rip off. These pieces attract and stick to each other, particularly in high humidity environments, resulting in regions where the micromachines begin to catch and fail.

The researchers learned, quite contrary to previous beliefs, that the polysilicon at these dimensions was extremely flexible and tough, not at all brittle.

Simultaneous to the stress and failure study, Bill developed a physics-based model that predicted when parts fail. The results derived from the model – an equation taking into account strength, adhesive wear, critical volume, pin joint radius, applied force, resonant frequency, and quality factor – were "remarkably similar" to the actual results from the physical testing.

Bill says the successful predictive model is significant because it now means that MEMS reliability may now be tested without actually waiting the days, weeks, or months that it takes for parts to fail.

Bill notes that his reliability group is only part of a much broader interdisciplinary team of researchers from many departments across Sandia who are focusing on MEMS performance and reliability issues. Expertise on this coordinated team spans packaging, failure analysis, surface science, materials science, processing, design, and optics.

"We can now predict wear-out – how long the device will last – for these tiny machines through accelerated testing and the model," Bill says. "This makes MEMS more feasible for further development and actual use. The floodgates are about open and MEMS will soon be in many applications, partly because they will be all the more reliable due to our work at the Labs."

More information about MEMS can be found at the Website