26 April 2001

Better robots through clean living

by Bob Felton

Wear and microdebris are primary causes of MEMS device failures.

Microengines and microrobots will last longer and perform more reliably if designers minimize surface contacts, especially in the area of pin joints, and develop a satisfactory microlubricant, according to a report released by Sandia National Laboratories.

The report, "MEMS Reliability: Infrastructure, Test Structures, Experiments, and Failure Modes", details Sandia National Laboratories' three-year effort to develop a sound theoretical basis for assessing the reliability of microelectromechanical systems (MEMS) devices. Before the study, the conventional wisdom among engineers was that the devices failed because the minute pieces of polysilicon used in MEMS devices were brittle at such small sizes. The researchers learned that's not the case, though; actually, the minuscule pieces are extremely flexible and tough.

Multiple failure causes

The research began with hand construction of a one-of-a-kind test bed for running large, statistically significant numbers of microengines to failure. The bed let engineers control ambient environmental factors, such as temperature and humidity, and monitor the operation and failure of the microengines using microscopes.

"We found they fail by wear," said lead researcher Bill Miller, "much like a car engine fails without oil. The individual parts get so worn that they jam." The study allowed Miller and his team to develop a model that predicts the life of a machine. "We can now predict wear out—how long the device will last—for these tiny machines through accelerated testing and the model," he said.

The researchers found that "by far, the major failure mechanism that we have observed in operating microengines is wear. The polysilicon rubbing surfaces in the pin joint and hub region of the microengines are most susceptible to wear. These regions are quite constrained, with gaps of 0.5 microns or less. Once the wear debris is produced, it cannot escape the region and then participates in three-body wear."

The team was surprised to learn that the intensity of wear is mitigated with increased humidity. Apparently, atmospheric moisture lubricates the tiny devices. "The optimum region to operate a microengine for low wear," the team reported, "is between 30% and 60% relative humidity at 25°C.

A second form of failure, adhesion, occurred when a machine could no longer complete a full revolution. Adhesion failures are of three types:

Sticking arises from localized surface roughness or varying adhesive properties but usually is a precursor to rocking or seizing.
Rocking occurs when the gear cannot rotate through a complete revolution and compensates by rocking through an angle. It is caused by failure of a signal line, a sticky spot, or microdebris buildup that impedes motion.
Seizure is the inability of the gear to turn, and the cause is usually in the region of the gear hub or pin joint. It may be brought on by microdebris or adhesion.

Though microengines are assembled and tested in clean rooms, they work in the real world. There, particulate contamination causes problems, especially for machines subjected to shocks and vibrations. "These environments," the researchers found, "cause the particles to move and can short out working devices. Of particular concern is the edge of the die, where the polysilicon layers are free to peel and crack, which forms rather large debris."

Researchers observed electrical current arcing across comb fingers, too. The result is often that adjoining fingers are welded together, destroying the onboard circuitry. They also learned that capillary forces associated with surface coatings could increase stiction, or the force required to cause a body in contact with another to begin to move. "Some microengines start right up after the release," they wrote, "while others require a 'slight poke' with the manual prober to function. These microengines may be side by side on the same die. There are many unresolved issues surrounding the stiction problem."

The researchers also discovered there's a lot more to learn about the forces at work. "One of the really disturbing failure modes," they reported, "is that of dormancy, when these microengines fail during storage in a benign environment." Relative humidity may be a factor; generally, machines fared better under dry storage. But they also noted that "the issue of intermittence also arises in these dormancy experiments. There have been occasions where a previously failed part begins to function again. The only difference between the two tests was time in storage."

Reliability is mission critical

Boosting MEMS reliability is vital to the technology's chief sponsor, the U.S. military, and to the eventual users of spin-off products in manufacturing facilities. Presently, plans are afoot for a host of tiny battlefield devices, including lab-on-a-chip sensors that will sniff for chemical warfare attacks and even micro air vehicles (MAVs) no larger than a hummingbird that will be used to deliver payloads and conduct aerial reconnaissance.

In a 1997 paper describing the battlefield of the future, Defense Advanced Research Projects Agency (DARPA) program manager James McMichael and U.S. Air Force Colonel Michael Francis discussed the use of MAVs in future conflicts: "The small speck in the sky approaches in virtual silence, unnoticed by the large gathering of soldiers below. In flight, its tiny size and considerable agility evade all but happenstance recognition. After hovering for a few short seconds, it perches on a fifth-floor windowsill, observing the flow of men and machines on the streets below. Several kilometers away, the platoon leader watches the action on his wrist monitor. He sees his target and sends the signal."

Battlefield MEMS won't be the high-end toy of generals, either: "MAVs are envisioned as an asset at the platoon level or below. Locally owned and operated, they will greatly reduce the latency inherent in current reconnaissance assets. They will give the individual soldier on-demand information about his surroundings, resulting in unprecedented situational awareness, greater effectiveness, and fewer casualties."

DARPA is also funding development of a "smart shirt," a garment made of a lightweight fabric with integrated sensors that monitor heart rate, respiration, EKG, temperature, and other vital signs. The garment can pinpoint for battlefield medics the location of wounds and assist triage. The garment also supports voice communications and global positioning systems.

Design recommendations

The researchers concluded that the most important step toward minimizing wear is using a satisfactory microlubricant. The best of the available lubricants are self-assembled monolayers. Humidity can help, too. In the range of 30% to 60% relative humidity, it acts as a lubricant and minimizes the formation of microdebris. Devices last longer at low humidity but are less effective because of wobble and degraded performance. If the devices are sealed in hermetic packages, ambient relative humidity becomes moot, and satisfactory lubrication assumes primary importance.

The second design recommendation, commonsensical but easy to overlook, is to design to minimize surface contacts and the associated normal forces. Candidate areas for redesign include the pin joint region of the drive gear, the hub of the drive gear, the shuttle guides that constrain motion, and the clamping actuator. An additional step toward diminishing the forces at work is to ramp up to full speed over multiple cycles, thereby reducing the magnitude of the impact or impulsive forces. IT