INDUSTRY NEWS

MEMS contamination questions starting to yield some answers

Unlike their counterparts in the semiconductor industry, MEMS manufacturers face their most challenging contamination problems at the back-end of the process. Solving those problems is a key step on the industry's path to reaching the multibillion-dollar potential of the market for the technology, say several participants in a recent microelectromechanical systems commercialization conference.

Although manufacturers "don't need things as clean as the new ultracleanrooms for 1-Gb DRAMs, we do need clean, and at further levels out in the process than ICs do," points out Karen Markus, director of the MEMS Technology Application Center at the Microelectronics Center of North Carolina (MCNC) in Research Triangle Park. MEMS devices lack the "density of features" that chips acquire during lithography, etching, and deposition. Hence, Markus continues, "a particle that will kill a DRAM, unless it is absolutely perfectly placed, would be of less concern to a MEMS device.

TESTING: Danelle Tanner of Sandia National Laboratories tests MEMS reliability using the SHiMMeR, a tool developed at the lab. Images at right show MEMS pin joint holes in unstressed and stressed conditions. Photo by Randy Montoya; Images Courtesy of Sandia National Laboratories

"ICs are less sensitive to the environment once the cleanroom process is done. And then you go to the release, dicing, packaging, and assembly stage for MEMS, and particulate contamination becomes ultrasensitive to us. It's the flip side of the IC industry."

MEMS is "an emerging industry. . . that's still trying to define itself," asserts Kenneth R. Farmer II, director of the Microelectronics Research Center at the New Jersey Institute of Technology (NJIT) in Newark. "In addition to being in its infancy, an additional problem is that it's so diverse. You could have MEMS that have fluidics applications, you could have MEMS that have electronics applications or applications in optical and aerospace or in naval or military types of areas in memory storage devices. It's just such a broad, all-encompassing technology. [It can't] define itself like the semiconductor industry, which just bases itself on the transistor. That base is missing in the MEMS industry."

Farmer and Markus were among the industry experts who addressed these growing pains at Commercialization of Microsystems '98, a conference held in San Diego in mid-September. Participants at the third annual conference discussed topics such as the upcoming MEMS roadmap, infrastructure development, technology transfer, and governmental support.

"The industry is at a stage where researchers and companies need to discuss common technology standards in order to meet the future needs for manufacturing," says Steve Walsh, an NJIT professor and MEMS specialist in the School of Industrial Engineering. These standards are necessary if the industry is to realize its potential for system-on-a-chip advancements and other breakthrough applications. Walsh and Markus were two of the cochairpersons for the conference, which was sponsored by NJIT, SEMI, and the Engineering Foundation.

Perry Cook, a principal with Silicon Sense in Culpepper, VA, delivered a paper on behalf of VLSI Standards focusing on the industry's need for standard reference materials. Referring to the flexible silicon membranes used by manufacturers, Cook asked participants, "Do we need to develop. . . a standard that the industry would be able to use in forecasting or characterizing how well the membranes they've made are going to perform?

"The industry does talk a lot about there being a need to improve on yields and to improve on the overall cost of manufacturing, and better yields is one way of doing that. The other huge issue has to do with the packaging of MEMS devices. Every package is different, because every device is different. It's not like memory chip packaging." The industry is conducting a "broadbased reliability study," says Cook, who is also working with Job Elders, chairman of the roadmap committee, on developing a systematic approach to manufacturing that incorporates packaging concepts at the R&D stage that can be implemented easily during manufacturing.

Reliability is a major concern of the "little subgroups" that have sprung up around the different applications in the industry, Farmer says. Despite different applications, there are similarities in fabrication approaches. "MEMS folks who are trying to build on the planar processing. . . whether they're working on one or a dozen application areas, all have equal potential in the market." Even though accelerometers and inertial sensors "came first, they're not going to dominate over microfluidics or DNA analysis devices down the line."

The conference pointed out that manufacturers in all applications worry about reliability issues, "each in their own area," such as wafer bonding, says Farmer. "Wafer bonding is a hybrid between bulk and sacrificial surface processing. It's adding material that might be a bulk type. If you are going to do that for an SOI [silicon-on-insulator] application, and use bonded wafer technology, yield is everything. You may have to have 2000-Å-thick silicon and a 1-µm-thick insulating layer on a standard-thickness slice of silicon. There can't be any voids, any particles, any problems with any of the interfaces. That extends not only to physical but to electrical defects as well. The integrity of the oxide has to be maintained."

Electronic Visions, a company based in Phoenix, has installed its new wafer bonder at NJIT's Microelectronics Research Center. MRC is investigating ultrathin-wafer bonding of single-crystal silicon wafers measuring 4 in. with thicknessess between 2 and 200 µm. The technology "will open up a new class of MEMS possibilities which require single-crystal silicon, rather than the polycrystalline layers deposited by the nonbonding methods," says Thomas Digges Jr. of Virginia Semiconductor. The company has a joint agreement with MRC to explore these possibilities, one of which "would be the integration of MEMS devices with microelectronics on the mechanically active, ultrathin silicon layer."

In applications where thicker layers of silicon are required, "as long as you have a few voids here and there you could live with that," notes Farmer. "One company said, 'We can live with three voids on a bonded interface as long as they're on the edge of the wafer.' That's what MEMS people are saying right now. There's nothing as specific as what the CMOS industry is requiring [for defect levels]. It's much more vague, because we're mainly at the prototype stage."

The industry is already at volume production of accelerometers, rotational sensors, and inertial sensors, Farmer points out. The development of other applications and their concomitant yield requirements will come "when the demand is there," a date he acknowledges is hard to predict. In the biomedical field "there are some things reaching a pretty high degree of maturity," but he adds that market acceptance may hold the key to further development in this area.

He cites the example of a "point-of-care analysis system" for hospital patients, which monitors gases or other chemicals in the bloodstream using just one drop of blood. Centralized laboratories at hospitals may fear losing control over testing and quality control. Using nurses to conduct analysis on one drop of blood with this new device "may really speed things up, but there's fear maybe that there will not be a full need for a centralized lab," Farmer explains.

The overall downturn in the semiconductor industry has not affected MEMS manufacturing to any great extent, says Farmer. The topic of who would "develop these MEMS things" was addressed at the conference, and it was the conclusion at one of the talks that "it's the smaller companies, these start-ups, that have the opportunity to make significant gains. What's happening with the big boys isn't necessarily going to affect what's happening with the smaller companies.

"Yield and defect issues. . . come up in the manufacturing technology that you choose," Farmer points out. "That would be something like we're trying to do here at MRC, to determine what the appropriate infrastructure is in order to support these small companies." He notes that a large semiconductor manufacturer has developed "whole areas where they can quickly make the switch to manufacturing. When MEMS is ready for them they can adjust and be ready for MEMS."

The production technologies of wafer bonding and deep reactive ion etch (RIE) offer benefits and present drawbacks with distinctive yield impacts, says Farmer. "Deep RIE is fast. It can give you nice vertical holes, and you usually work one wafer at a time. It's a serial, not a parallel, type of processing" which he calls "robust, reproducible, and uniform." With wafer bonding, "a major problem is voids."

Automation will be a critical need, particularly for deep-RIE processing, says the center's director. Electronic Visions' automated bonders "can be used cassette to cassette for 8-in. wafers with many different steps." Automation and "good throughput are critical" for maintaining MEMS yields, he notes.

"We're a little more tolerant [of in-line contamination] in comparison with the submicron IC," says Markus of MCNC. "We are significantly orders of magnitude less tolerant in the packaging, assembly, and test phase. I know companies like TI and Analog Devices and any of the companies who are producing commodity products right now are facing big challenges when it comes to [the question of], 'How are we going to test these things?' "

TI moved their back-end test tools into a clean environment for their digital processing devices, Markus says. The concern was where particles might land if they fell from the head of their "$4-million test machines. There're only so many dead pixels that can be tolerated. . . . I think the biggest challenge from [the contamination] perspective is that there's suddenly an area that we never had to seriously worry about. Particulates and handling are now serious yield-impacting issues."

Suppliers need to pitch in by focusing on developing the cleaner tools for the industry, asserts Markus, who was chairperson for a session on infrastructure development and commercialization at the San Diego conference. The problem, Markus hints, is that suppliers lack the economic incentive. "With the IC industry at $365 billion a year without a strong concern about back-end contamination, who is an ATE vendor going to pay attention to?" she asks rhetorically. In contrast, the MEMS market stands at between $8 billion and $32 billion [by 2000], depending on whose estimate you read and how a MEMS device is defined.

"We need to get a strong international focus on this back-end equipment by equipment vendors to recognize that this is going to be a significant market 10 years down the line," asserts Markus.

Some of these questions should be resolved when the roadmap is written. The first draft is due in June 1999 with the final document due by the end of the year, according to Farmer. In the meantime, proclaims Markus, there are no magic bullets for dispatching yield killers. "I have a saying, 'If wishes were knishes, the world would be delicious.' The problems are cost, priorities, and industry involvement."

Parties interested in MEMS packaging issues may contact Karen Markus at markus@mcnc.org. Kenneth Farmer's E-mail address is farmer@tesla.njit.edu.

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