"MEMS Across the Valley of Death," Feature Article, April 2006

FEATURE FOCUS: Microsystems

MEMS across the
valley of death

Microelectromechanical systems are taking the world by storm, but only a handful of aerospace applications are now flying. That may be about to change.

by Alan S. Brown, Associate Editor

Milind Pimprikar calls it the "Valley of Death." It is the place between proof of concept and implementation where MEMS go to die. At least, many of the ones designed for aerospace systems.

MEMS, or microelectromechanical systems, are chip-size devices usually carved from semiconductor wafers. Unlike conventional integrated circuits, however, MEMS include microscopically small moving parts that act as sensors or actuators.

"There are hundreds of proven concepts, but only a few MEMS make it across this valley," said Pimprikar, a long-time MEMS developer. "There's nothing wrong with these concepts. It's just a question of having the resources to apply them."

Pimprikar's remarks expose a contradiction in the microworld. The industry is thriving. Analog Devices Inc. of Cambridge, Mass., churns out over one million 4-millimeter-square air bag accelerometers each week and another million chip-size gyroscopes every month.

The MEMS device above is used for radio frequency applications ranging from low frequency wireless to advanced wave antenna technology.

MEMS have become quietly ubiquitous. Many new applications sound like they were lifted from the pages
of science fiction. Small MEMS play an oversize role in one-chip microphones, explosives detectors, and devices that separate proteins based on their electrical conductivity. Rear projection televisions use chips containing micromirrors to bounce images onto a large screen. A similar approach routes Internet traffic over optical cables.

Drop a laptop and a MEMS accelerometer will park your hard drive to prevent damage before it hits the ground. Tilt a cell phone and a MEMS gyroscope will signal the cursor to scroll through a list of names or messages. Lose your Global Positioning System signal in your car and an inertial guidance unit that combines microscale accelerometers and gyros will keep you on track until it returns.

Many of these new applications are only a few years old. Some use innovative MEMS architectures.

Yet in aerospace, the field that gave birth to the very first microsystem in the 1950s and nurtured them with copious funding through their early development, things are quite different. The industry's toolbox—accelerometers, gyros, and pressure sensors—have been largely unchanged for years, although engineers still find innovative ways to use their functions.

"Aerospace is one of the most conservative industries because you can't afford to make a mistake," Pimprikar said. "You cannot fly an application until it's flight-tested, and you have to fly to test it." It also takes money to cross that valley and commercialize a new MEMS technology.

In 2002, Pimprikar founded a group to help developers of promising aerospace MEMS applications bridge the valley. It is called the Canada-Europe-United States Organization on Micro-Nano Technologies for Aerospace Applications, or Caneus. This year it will meet in Toulouse, France, between August 27 and September 1.

"Caneus grew out of my own frustration," explained Pimprikar, who also heads the Center for Large Space Structures and Systems, an R&D initiative funded by the government of Canada. "We had developed a MEMS system to do nondestructive testing of aerospace structures. We had no trouble getting the early proof-of-concept grants, but we needed $12 million to develop it fully. No venue existed to fund the project. We could only get grants of $100,000 or so at a time."

Thomas George, who heads Caneus's U.S. division, said, "The problem is that there is no mesh between new technology and end applications." George also directs MEMS product development at ViaLogy Corp. of Altadena, Calif.

"Technologies end up dying on the vine because developers cannot answer system-level issues about application, performance, and reliability. System developers are suspicious of overblown claims about how technology can do everything, including making sliced bread. They have learned some painful lessons," George said.

The role of Caneus, he said, is to confirm that the technology works, and bring together a consolidated team of inventors, investors, business people, and systems developers to lift the technology into the skies.

In 2004, the Caneus conference helped to line up sponsors and funding for 14 different MEMS concepts. "We looked at the technology, created a concept paper, looked at the viability of the business plan, and debated the concepts at an idea forum in conference," Pimprikar said. This year, he said, the conference will debate actual projects. These range from miniature satellites and harsh environment sensors to astronaut health monitoring devices and systems to improve MEMS reliability.

The Nano, the Pico, and the Satellite

Caneus actually formed a company, Caneus NPS Inc., to help it showcase the industry's ability to move MEMS technologies swiftly and inexpensively from concept to commerce The NPS of the name stands for "nano-picosatellite," the designation given to very small spacecraft that MEMS can make possible. The effort brings together Sweden's Ångström Aerospace Corp. and EADS Astrium (supported by sister EADS companies in France and Germany), potential end users Alcatel, Mitsui Bussan, and the U.S. Air Force, as well as universities and research laboratories.

The small satellites are intended to burst through price barriers and open an entirely new market for Earth imaging and mapping, weather forecasting, and atmospheric and scientific research. Such satellites could also prequalify electrical and mechanical components destined for orbital use by military and commercial customers.

Several years ago, the U.S. Air Force funded three university-based projects to create 1-kilogram satellites. Called Cubesat satellites, they were low-budget experimental systems, not commercial products. "They couldn't transition them to the next level," Pimprikar said. With $36.4 million in U.S. and Canadian government funding, Caneus NPS wants to do just that.

A spider mite dwarfs the micromechanical gears on this complex MEMS structure built at Sandia. Most production MEMS use simpler devices.

Driving Pimprikar's vision is economics. A conventional satellite, he said, weighs more than 10,000 kilograms and costs $150 million to manufacture, $100 million to launch, and $62 million to insure.

MEMS-enabled nanosatellites would weigh 1 to 10 kilograms and cost $3 million to build, $200,000 to launch, and $800,000 to insure. An even smaller picosatellite would weigh less than a kilogram and cost half as much as its nano cousin.

Nanosatellites and picosatellites cost less to build because they use relatively inexpensive MEMS components for specialized tasks. "They do only one job, but they do it well enough," Pimprikar said.

Moreover, they are small and light enough to piggyback into orbit on a variety of launch vechicles. In fact, said Pimprikar, they may not require a traditional launch at all. "We're looking at using missiles fired from a Russian submarine and rockets attached to jet fighters," he said. "That way, a small country like Sweden could launch into a specific orbit rather than wait for a place on the Space Shuttle or Ariane 5, where they're at the mercy of the orbit they're given."

A jet or sub-launched satellite would not reach a high geosynchronous orbit, nor would it pack enough fuel to keep it aloft indefinitely. Still, Pimprikar argues, their flexibility and very low total launch cost make them attractive even if their orbit begins to decay after several months.

Sensors for harsh environments have also shown up on the radar screens of Caneus and many other organizations as well. Not every developer is waiting for Caneus to make its introductions. In 1998, the State of Ohio and NASA Glenn Research Center formed an organization to conduct research into MEMS sensors that could operate at temperatures up to 600°C.

Three years later, the organization spun off Glennan Microsystems Inc. to commercialize its new products. In 2004, Glennan received an Advanced Technology Program grant from the National Institute of Standards and Technology to fund a four-year, $6.3 million program for sensors to reduce nitrogen oxide emissions from jet engines.

Jet engines running lean fuel mixtures are prone to instability.
High-temperature MEMS sensors could improve performance and
fuel mileage while
reducing emissions.

The grant paid for half the program; the rest came from Glennan's partners. They include Delavan Inc. (the turbine fuel unit of Goodrich Corp.), FLX Inc., Zin Technologies, and Case Western Reserve University. Delavan brought in Rolls-Royce, which provided a turbine for the tests, a model A3000, which is used for unpiloted aerial vehicles and business jets.

Glennan's vice president, Rick Earles, makes a compelling case for the technology. Aircraft operators, he said, would like to burn lean fuel mixtures to improve fuel mileage. The ideal lean mixture uses 40 to 45 percent of the fuel of a typical 14:1 air-to-fuel mixture.

"The problem," Earles said, "is that turbines running on such lean mixtures are prone to instabilities, especially during dynamic conditions when you change the thrust of the engine. Lean mixtures also increase NO_x emissions, especially where hot spots form within the engine. When fuel doesn't combust fully, it may go off to cool zones and turn into carbon dioxide."

Glennan's solution is active fuel and emission control. Its engineers have integrated six sensors onto each of the A3000's 16 fuel injectors. The devices sense temperature, pressure, and air-to-fuel ratio, and provide instantaneous feedback on what they find.

"We've created a smart injector that can actually control and continuously optimize combustion," Earles said. "By using them in a distributed control scheme, we can let them control local combustion while controlling their interactions with one another." He figures the technology could improve turbine stability under lean operating conditions while reducing nitrogen oxides by 75 percent and carbon dioxide by 15 percent below current emissions standards.

Jumping the technology gap from concept to system has not been easy. The sensors consist of silicon carbide, a semiconductor that retains its properties at high temperatures. Silicon carbide is notoriously difficult to grow and pattern, however. Adding electronics and wiring that allow each sensor to communicate with the fuel injector was also a challenge.

Still, elements of these technologies have been under development for two decades. The key hurdle for Glennan was building the right team for the project. Goodrich, said Earles, already had a program in active combustion control. It then built a requirements document that spelled out what was necessary for success in the marketplace. That made it easier to bring in Rolls-Royce, which already had close ties with Goodrich.

It also helped Glennan line up additional funding from the Air Force Research Laboratory, the Defense Advanced Research Projects Agency, and the Department of Defense. After proving the technology on jet turbines, Earles plans to adapt it for use in turbine power generators, industrial furnaces, refineries, and chemical plants.

Another MEMS aerospace technology that receives government funding is inertial measurement. This involves the combination of two existing MEMS devices, accelerometers and gyros. The former measure the change in rate of linear motion; the latter sense changes in movement through three dimensions. A processor takes information from up to three accelerometers and three gyros, all at different angles from each other, and uses it to calculate the unit's position.

Such units are already in use. When a car with a Global Positioning System receiver drives through Manhattan, it may lose its satellite signal amid the surrounding skyscrapers. The MEMS inertial measurement unit keeps track of the car's position until it makes contact with the GPS signal again.

Many civilian aircraft and helicopters now use MEMS technology in their primary inertial guidance systems, said Sean Neylon, CEO of Colibrys SA of Neuch‰tel, Switzerland, which supplies MEMS for the aerospace industry. "These aircraft make shorter flights and just need a low-cost, reliable positioning system," he said.

Commercial passenger jets that make long flights need greater precision than MEMS-based units generally provide. The military is also looking for higher precision for its space-limited unpiloted aerial vehicles, especially tactical drones, as well as for missiles and smart bombs.

One problem that MEMS inertial measurement units face involves drift, the tendency to lose position over time. One of the major sources of drift is stress, according to Bob Sulouff, director of business development for the micromachined products division of Analog Devices. Accelerometers and gyros consist of proof masses whose structures respond to changes in motion.

"If you squeeze their packaging, bend their circuit board, or have thermal expansion stresses between their metal and plastic parts, these forces will create signals similar to an acceleration signal," Sulouff said. "They will register on the sensor as the same electrical result." The resulting signal causes the inertial calculator to drift.

Closed, open, and exploded: Three views of a so-called nanosatellite. Based on microtechnology, it would weigh less than 10 kg, making it cheap to build.

There are a number of ways to circumvent the problem. Some developers opt for larger proof masses. "Normally, larger proof masses are more susceptible to stresses, but they also produce stronger signals that are larger than the noise signals," Sulouff said. Others opt to package MEMS on stiff metal assemblies to minimize deformation.

Sulouff's company makes accelerometers and gyros for only a few dollars per unit. Fresh off the assembly line, they are not accurate enough for tactical-grade guidance units used in smart munitions. Yet, he noted that companies are now buying his devices and then testing and recalibrating them. This improves performance by a factor of two or three, and may make them accurate enough for a howitzer shell that spends only a few seconds in the air, he said.

Eric Lautenschlager, a senior research scientist at Honeywell Aerospace's Advanced Technology Organization, takes the opposite approach. "What sets apart our work here is that we're not just interested in reducing cost or consumer products," he said. "We're interested in pushing the performance level of technology."

That isn't easy. Using semiconductor processes, Honeywell carves out comb-like structures that oscillate about 10,000 times per second. Each sensor must be highly reproducible both laterally and in depth, so that the electronics can measure vibration changes of a fraction of an angstrom.

"The biggest issues we face are how to detect the signal and bring it into the macroscopic environment," Lautenschlager said. In addition to stress, resonating sensors respond to changes in temperature, pressure, and flight environment that can mask or degrade performance, he said. The key is to develop amplification techniques that increase signals without introducing noise.

Neylon agrees. He estimates that more reproducible structures explain only 25 percent of the improvement in MEMS performance. Better electronics accounts for the rest.

New Directions

Meanwhile, MEMS technology continues to evolve at a startling pace. At the University of California, Santa Barbara, former DARPA MEMS program manager Noel MacDonald is asking why anyone should limit MEMS structures to silicon.
"Silicon was there, it was inexpensive, and everyone knew how to process it to build lots of structures," he said. The problem is keeping silicon from cracking afterward.

"That's not a problem with microcircuits because they're stationary," he said. "But silicon is a single crystal and it will cleave along an etched line, which is how they cut wafers into chips. It is one of the reasons MEMS yields fall from very high levels to only 5 to 10 percent during packaging."

The problem gets even worse when making three-dimensional structures, such as inertial guidance units. They involve bonding chips into position with heat and pressure. "I don't mean you can't do it, but the yield is low," MacDonald said. "You then need a package that keeps them free from shock, and aerospace is a field where there is always lots of vibration and shock." Making MEMS out of titanium rather than silicon could help resolve some of those issues.

The industry has also begun to build a more innovation-ready infrastructure. Part of the problem, according to longtime MEMS developer Michael Huff, is that making MEMS is not like making integrated circuits.

MEMS structures vary greatly. The post actuators above position mirrors on MEMS used for telecommunication. The oscillating proofmass structures sense angular rotation around an axis in the gyro below.

The latter, he said, use a small set of materials and a fixed sequence of processes. MEMS, on the other hand, involve a much wider range of materials and vastly broader array of processes used to sculpt three-dimensional structures on silicon. "The difference between making an airbag sensor, a microvalve, and a Hall-effect sensor is enormous. Each involves a fully customized process sequence. There is no way you could build a foundry that had everything you needed," he said.

In the late 1990s, the Department of Defense gave Huff money to found the MEMS and Nanotechnology Exchange, a virtual foundry that brings together the production capabilities of MEMS producers around the country.

"We do about 40 to 50 projects every week, mostly prototyping," Huff said. "The average project has about a dozen process steps that take place at three different foundries. Users can go to our Web site and create a run card from hundreds of different processes. We look it over to ensure it makes sense, and then take care of all the logistics and legal aspects. To our users, it looks seamless."

The MEMS and Nanotechnology Exchange provides another way to ease the tortuous path to commercialization. It promises more prototypes and more technologies will flow from MEMS inventors in the future.

Yet bridging Pimprikar's Valley of Death between concept and implementation remains an issue. R&D money has never really been the problem, according to long-time MEMS market guru Roger Grace of Roger Grace Associates in Naples, Fla.

The real problem is that until more MEMS companies begin making money in aerospace, venture capitalists hesitate to fund their companies. "Venture capitalists want to see large volumes, like MEMS microphones for the 900 million cell phones planned for 2006 and the MEMS gyros that stabilize the picture in camera phones," Grace said.

Caneus is trying to give companies a better chance of matching their MEMS technologies with new aerospace programs and applications. Some of those technologies will blossom, like the inertial guidance units developed for aircraft and now used in cars, or the micromirrors once envisioned for phased array radar and now used for display televisions and Internet connections.

They just have to cross that one valley first.