 |
July 8, 2003 — Call it a dielectric sandwich.
 |
 |
That’s one approach behind Raytheon Co.’s attempt to commercialize radio frequency MEMS components: Etch a circuit onto a silicon wafer, slap on a layer of dielectric material to act as insulation, and then put a MEMS switch on top of that. Voila — an integrated RF MEMS device.
 |
That morsel of an idea was served up in May, when the U.S. Patent and Trademark Office granted Raytheon a patent on the concept. It’s a potentially huge and lucrative sector, and MEMS businesses everywhere would love to bite into the market.
If they could just make the pesky things work.
The problem? The materials best suited for RF devices are those with high resistivity to electrical current, so they can receive and process radio signals properly. But integrated circuits require materials with low resistivity, since their whole purpose is to conduct an electrical impulse from one place to another. Combining those two contradictory needs into one device means the RF component must be hermetically sealed and separate from the circuit — yet somehow still communicate with the circuit.
“There are some incompatible functions you’re trying to achieve here,” said Al Sidman, an engineer and vice president of MEMS products at Advanced MicroSensors Inc. in Massachusetts. “It’s a daunting task.”
In truth, MEMS companies can create sealed packages for RF devices; it just costs a relative fortune to make them — $20 or more per switch, by some estimates. That limits potential customers to the military, aerospace giants, satellite communications businesses and other vast enterprises. Far richer opportunities in cellular phones, telematics or vehicle tracking systems are out of reach because the packaging is simply too expensive.
“We think about RF MEMS,” said Bob Souloff, business manager for MEMS at Analog Devices Inc., but “we’re still trying to get our arms around the economics of it.”
Analog is not alone. A host of business and academic research labs are exploring ways to manufacture integrated RF devices efficiently, with various degrees of success.
Advanced Microsensors has teamed with another Massachusetts company, Radant MEMS Inc., to develop MEMS switches with a Defense Advanced Research Projects Agency grant. Eventually the switch will go into an 8-foot-long antenna the military will use for radar systems, where each antenna will have about 100,000 switches. The components have a life span of about 10 billion cycles; that’s plenty for commercial customers, but the government wants a life span of 100 billion cycles — about 10 years of use in a low-Earth-orbit satellite.
Rick Morrison, senior engineer with Radant, said his company uses a packaged-wafer with the RF switch encapsulated in a separate container and then attached to the silicon substrate with other circuitry. The switch’s information is sent to the circuit through tiny interconnects commonly called “vias.”
“You have to keep the RF as far away as possible from everything else, and that’s easy,” Morrison said. “What’s tricky is to bring the vias up through the cap or down through the substrate.”
Morrison said it took Radant about six months of research to solve the problem. Radant’s switch costs $10 to $20, a price Morrison admits is expensive. He expects that figure to fall to $1 to $2 each as production increases.
Dielectric layers and reliable packaging are vital for RF components because they prevent stiction, the dreaded phenomenon where a MEMS part simply stops moving. Stiction occurs when electrical impulses “leak” from the nearby circuit layer and charge builds up. Once the amount of charge grows large enough, it jolts the RF switch into immobility.
MagFusion Inc. (formerly called MicroLab) in Arizona is taking another tack. Rather than worry about electrostatic charges that cause stiction, MagFusion has an RF switch that operates using magnetic force. Charles Wheeler, vice president and co-founder, said that allows MagFusion’s switch to operate at lower voltage and cut the chance of charge buildup.
MagFusion’s RF switch was introduced last month.
Magnetic switches have challenges of their own, such as how to control large numbers of switches at once and how to coexist peacefully with other components made of iron, nickel and other metals subject to magnetic force. Still, few doubt that the basic principles work.
Y.C. Lee, a professor at Colorado University, is trying to perfect a way to package RF switches without hermetic seals. If successful, he said, it could cut the cost of packaging down to less than $1 per device.
His technique is called atomic-layer deposition, similar to the chemical vapor deposition used to grow carbon nanotubes, and it allows Lee to drape a compound onto a surface one atomic layer at a time.
That lets him put a conductive layer on a substrate (say, alumina over silicon) and then put a dielectric layer on top of that (zinc oxide, for example). The result is a dielectric layer that separates the RF switch from electrostatic interference, but still has enough conductivity to transmit RF signals to the rest of the circuit.