INDUSTRY NEWS

MEMS industry gearing up
for multibillion-dollar future

Forget: the simple microsensor. Think: an entire MEMS device built into a microchip. Forget: designing a micromachine that amazes only the engineer in the next cubicle. Think: building a GPS-based guidance device that helps dad keep track of his young daughter. Forget: publishing your work in the Journal of Product Innovation Management. Think: Newsweek.

Forget the preceding advice and the growing global MEMS industry may take longer to reach its glowing multibillion-dollar future, industry leaders contend. No one believes that the microelectromechanical systems business is any danger of grinding to a halt. But industry leaders realize that MEMS manufacturing still faces a number of barriers before it can attain its full commercial potential. Some of the above solutions arose during discussions of those obstacles and in papers presented at recent industry conferences.

ON REFLECTION: This Sandia prototype with hinged mirror is part of a weapons safety system. The small drive gear at right measures 76µm in diameter.

Can the barriers be overcome? "Yes, they can," insists Steve Walsh, an assistant professor in the School of Industrial Engineering at the New Jersey Institute of Technology (NJIT) in University Heights, NJ. "Two years ago, when I spoke on MEMS at SEMI's Industry Strategy Symposium it was maybe a half-billion-dollar annual market; today the market is a little over $2 billion." And climbing, according to Walsh, a MEMS maven who projects a $15-billion annual market by the year 2000. He says the industry "is averaging around 45% growth a year," certainly outpacing that of semiconductors. He cites a new Japanese forecast that the industry will reach annual sales of $25 billion by 2004.

"I can tell you most assuredly it is growing, and growing very rapidly," agrees Perry Cook, vice president of marketing for Virginia Semiconductors, a supplier of 2- to 4-in. single-crystal silicon wafers based in Fredericksburg, VA. "Our business has grown by better than 20% a year for the last 10 years, and it has all come from serving what I call the specialty MEMS marketplace." Cook's company has more than 300 worldwide customers that use his poly to make devices such as pressure transducers that trigger deployment of oxygen masks in airplanes.

Barriers to full commercialization fall into four categories, says industry consultant Roger Grace, president of Roger Grace Associates of San Francisco. The concerns are marketing, infrastructure, packaging, and device integration.

In terms of marketing, Grace and Walsh both say that the industry badly needs good P.R. as a means of expanding its potential customer base. "MEMS people don't understand marketing a concept and fulfilling customer needs and desires," Grace says. "You just can't develop a whizbang MEMS for MEMS sake" and expect the outside world to find it inherently interesting. At the Commercialization of Microsystems '96 conference in Kona, Hawaii, last October, participants in working-group discussions suggested that the industry could garner publicity by developing articles for publications like Time and Newsweek.

The industry's infrastructure, though it is expanding, still lacks maturity, says Grace. Only recently, he notes, have companies such as Karl Suss, which makes a double-side aligner, begun selling to the MEMS market. Other positive signs abound, however. Walsh says that more vendors are dipping a toe into the water. A company named GTI is making a stiction-abatement device, and Shipley is developing thick resists, for example. Surface Technology Systems and Plasma-Therm have licensed the Siemens deep-RIE process for making devices, Walsh notes, adding that Applied Materials is considering doing the same. And, according to Grace, three companies—Microcosm, Intellisense, and Tanner Research—have developed CAD/CAM systems for micromachine manufacturing.

The problem with MEMS packaging, says Grace, is that "it's not very sexy." He notes, for example, that pressure sensors for measuring sulfuric acid cannot withstand the harsh chemical on their own. Decent packaging prevents such chemicals from hindering the performance of devices. The interaction of a device with the medium it is sensing or actuating requires several difficult and expensive packaging steps, Grace and other experts point out. The consultant suggests that the industry take "a more interdisciplinary approach" to packaging. This "concurrent engineering" method would involve collaboration among process engineers, manufacturing engineers, and packaging engineers. Unlike chip manufacturing, the back-end must be brought up to the front of the process.

Finally, the industry should further develop the concept of integrating microelectronics and MEMS devices, the "system-on-a-chip." One prime illustration is an accelerometer with on-chip detection circuitry from Analog Devices. In general, though, "MEMS people are looking to a MEMS device independent of anything else," laments Grace, adding, "There's a total lack of wanting to kick down the side of the sandbox" and explore breakthrough applications.

James Smith, supervisor of intelligent micromachine technology development at Sandia National Laboratories in Albuquerque notes that integration can improve MEMS performance and lower manufacturing and packaging costs by combining the devices with an electronic subsystem in the same process. The wafer with the MEMS devices is processed using conventional CMOS techniques, with steps added at the end of the CMOS process to expose and release the embedded devices, says Smith.

This approach avoids the planarity difficulties that can occur in MEMS four-level polysilicon processing, says Smith. His department is "trying to develop a five-level process, which is incrementally more difficult, [because] you have problems very similar to ones the IC industry has had in going to the next level of metal—as you stack more film you run into planarity problems. We're really trying to leverage the IC manufacturing infrastructure. . . and we're using chemical-mechanical polishing technology in order to resolve some of these issues."

MEMS manufacturing "brings you another type of microcontamination problem," notes Walsh of NJIT. "How can you do micromachining in a clean environment and have it be compatible with CMOS, bipolar, or biMOS processing? [Manufacturers] don't need a Class 1 cleanroom. Most of the new facilities coming on-line are Class 10, 6-in. facilities" using one of three manufacturing processes: high-aspect-ratio micromachining, sacrificial surface micromachining, or "traditional" bulk micromachining. Smith says that a primary problem is wafer warpage caused by film stresses common to oxides.

According to most experts, batch-organized planar processing will be one of the deciding factors in the industry's success, because it offers the quality and low cost required to economically manufacture micromachined devices. Involving both etching and deposition steps, the process carries the potential for defect creation. "In bulk micromachining where we etch into the silicon wafer to define the geometries, if there are defects deep into the silicon wafers where they would not conform to the etch pattern, we have problems with defects," notes Mehran Mehregany, an associate professor in the department of electrical engineering and applied physics at Case Western Reserve University in Cleveland.

Sandia has developed its "micromechanics-first approach" to solve the planarity problems by building the MEMS devices before the CMOS process. If the micromechanical processing is done first, planarity is sacrificed, Smith points out. And if CMOS processing is completed first with its metallization needs, the process must endure the high-temperature annealing required in micromechanical manufacturing to reduce stress in the devices.

In collaboration with engineers from the UC Berkeley Sensor and Actuator Center, Sandia has used sacrificial polysilicon surface micromachining with 1.25-µm CMOS processing to develop three-axis accelerometers used as airbag-deployment sensors and manufactured by Analog Devices and Motorola.

Grace says that Sandia, NASA's Jet Propulsion Laboratory in Pasadena, CA, and the UC Berkeley Sensor and Actuator Laboratory "have incredible in-house MEMS capabilities" that the private sector should try to leverage. And the Western Europe­based Network of Excellence in Multifunctional Microsystems (NEXUS) program, launched in 1992, has expanded its information-sharing activities into central and eastern Europe. NEXUS sponsors four centers of excellence.

We're at the point where [William] Shockley was in 1948 when you made discrete devices that solve a singular problem," Walsh contends. "Currently, MEMS manufacturing has not embraced a microfabrication-like planar solution that embodies all of the capabilities of microsystems in one design paradigm. There just isn't one paradigm or packaging structure that's uniform for all MEMS applications."

MEMS manufacturing "disrupts the technologies currently used by the companies making products that a microelectromechanical system would replace," Walsh points out. "Take an accelerometer. Analog's accelerometer costs about $5 to make. It replaces a part that costs $200 from companies like TRW. However, the way Analog manufactures that accelerometer is totally different from how TRW manufactures these big mechanical devices. MEMS technologies totally disrupt the industries that their technologies do well in. TRW was forced to go away from macromachine technology and embrace MEMS. It was do MEMS or perish."

Walsh is working with SEMI to develop a 10-year MEMS roadmap that is scheduled for publication in September 1998. Ron Horwath, director of industrial relations for the trade association, says that the market for MEMS manufacturing tools "is only 10% of the size of the semiconductor market for SEMI members." He acknowledges, however, its growing size, giving impetus to the trade association's commercialization conference activities scheduled for September 1998 in San Diego.

Walsh and Mehregany are copresenters at a MEMS technology tutorial scheduled for July 15 in San Francisco during Semicon West 97. The tutorial will offer an introduction to MEMS with a focus on the role that chip manufacturing tool suppliers have in advancing the technology.

"The MEMS area has always been application driven but technology limited," points out Mehregany. "So a lot of high-volume applications are contemplated. The difficulty is developing the technology to accomplish that at an acceptable price to the market. As the technology matures, some of these barriers will be removed."

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit