advances, MEMS industry still feels growing pains
PARTNER: The incorporation of non-CMOS materials in IMT's processes
requires flexibility and keen attention to defectivity.
BY THEODORE CHI, COURTESY OF IMT
to help resuscitate the human immune system, a biomedical device has 32
motors that enable it to rapidly isolate 200 million to 250 million stem
cells. If it works as designed, the device and its massively parallel
sorting capability will represent a technological leap in cell sorting
that could greatly benefit both chemotherapy patients and victims of radiation
believe this switching mechanism is the fastest MEMS on the planet,"
boasts Monteith Heaton, vice president of marketing and sales for Innovative
Micro Technology (IMT), the Santa Barbara, CA–based fab that is
partnering on the sorter project. "It goes from 0 to 4 meters per
second and back to zero in less than 20 microseconds. It actually accelerates
at about 12,000 Gs." Current cell sorters are 20 times slower and—at
$400,000 apiece—expensive, emphasizes Michael Shillinger, IMT's
vice president of operations.
men claim other advances for the cell-sorting microelectromechanical system.
Because it's encapsulated, "the only thing that touches the fluid
is the chip itself. The chip is disposable,"” Heaton notes. "You
sort one patient, you throw [the device] away. There have been cancer
trials done with banks of those $400,000 cell-sorting systems. And [with
this sorter] you'd be able to do that same job, with the chip in a semidesktop
environment using a less-expensive machine, for a very reasonable cost—to
basically rescue people from those two [medical] situations.
goal here—and we're discussing partnering with a variety of different
biomedical firms—is to build the chips and have someone else build
the machine itself," he explains. "They would sell and market
has been working on the project for approximately a year and a half, Heaton
says. DARPA has provided approximately $7 million in development funds.
An additional $1.5 million has come through the U.S. Army Medical Command.
The timetable for commercialization is sometime in 2007.
the obvious enthusiasm of both Shillinger and Heaton, their excitement
has to be tempered by one frustrating fact for the industry: rapid commercial
introduction of MEMS in high volumes remains an elusive goal.
a 2003 survey, the MEMS Industry Group (MIG) reported that the most significant
problems holding back growth are a segmented market and high entry barriers.
"There is perception and acceptance that MEMS will enable sensor
pervasiveness, leading to significant new markets," the report states.
"While these markets are envisioned, they have yet to be created.
The investment needed in the design and simulation of multidomain devices,
as well as in the fabrication of these highly diverse structures, is high
and has delayed widespread adoption."
hampering commercial prospects are the two separate issues of feasibility
versus manufacturability and of fab capability. On the former point, the
report notes that MEMS designers tend to concentrate on proof-of-concept
demonstrations rather than ease of manufacturing. "This trend results
in quick feasibility demonstration but slow production ramp-up. Because
of that, the fabs do not see their volumes materialize, delaying the creation
of new markets."
the second issue, the report says that most fabs provide expertise in
"nearly every MEMS device category." Among the companies surveyed,
90% of the devices fall into six categories: inertial sensors such as
accelerometers; optical mirrors; microfluidic devices or microarrays such
as reservoirs; components for RF communications; physical sensors such
as pressure or radiation devices; and ink-jet nozzles.
need to cover all these bases requires the use of nonstandard materials
and processes. This diversity of concepts "is one of the major hindrances
to both faster time to production and manufacturing repeatability,"
the report emphasizes.
an indication of the industry's low production level, the report notes
that, on average, the
surveyed fabs were running at 30% capacity. MIG's survey covers what the
authors call "a significant segment" of the industry and more
than 60 noncaptive MEMS fabs around the world. The participants fall into
two categories: worldwide MEMS fabs and U.S. MEMS companies that outsource
his most recent annual report card, Roger Grace, a San Francisco–based
high-tech consultant, gives the MEMS industry a grade of C for its marketing
efforts. Grace, the former president of the Micro and Nanotechnology Commercialization
Education Foundation (MANCEF), says MEMS companies in the main "have
been created by strong, technically oriented people who tend to believe
in the 'build-it-and-they-shall-come' mantra. Not much formal market research
historically has been conducted to determine customer needs and price
contrast, Grace urges the industry to look to its semiconductor counterpart
as a good example of how to improve its business fortunes. "For MEMS
to be truly successful, a major effort must be undertaken by solution
providers to understand customers' needs, to provide unique solutions,
and to provide adequate resources to promote each company's market position,
brand, and approach to solving these issues…. Faster, better, cheaper
is the mantra of the semiconductor industry in direct response to customers'
needs. MEMS producers need to adopt a similar customer-centric attitude."
emphasizes that the MEMS industry has advanced significantly in its commercialization
efforts since 1998, the year he published the first report card. The consultant
says the industry spent robustly on R&D before that year, helped in
particular by DARPA. The defense agency's staff members showed "visionary
leadership" that led to "judicious investments" and a foundation
for commercial progress. Private industry funding has also been robust,
with a C+ grade in 1998, the MEMS industry infrastructure has improved
to an A in Grace's latest assessment. Instead of using chipmakers' hand-me-downs,
most fabs today use 6-in. wafers in high-volume settings. Instead of having
to modify semiconductor process tools, process engineers have access to
some MEMS-specific tools from manufacturers such as EV Group and Suss
a further development, companies such as Coventor and MEMScap market industry-specific
design automation tools for use by the more than 60 MEMS foundries worldwide,
Grace notes. The consultant emphasizes that value is "one of the
most significant barriers to commercialization of MEMS. Packaging and
testing are most costly for MEMS today, while market applications drive
prices to the bottom."
two trends are in conflict, Grace continues, "and one must give in.
When MEMS provides value, such as Pentium processors, or if packaging
and testing can be brought to the level of a few pennies, the cost/value
hurdle will be removed and the floodgates will open for MEMS."
lesson the industry can take from its semiconductor counterpart is the
development of standards, says Grace, who gives the industry a B in this
regard. In the past three years, the industry has increased its support
for standards. In particular, it has begun to develop the first two MEMS
process benchmarks. He also noted eight meetings in the past two years
of the SEMI MEMS standards committee. Both developments bode well because
standards are the mark of a maturing industry, Grace asserts.
more lessons that the semiconductor industry offers are the development
of technology roadmaps and the development of technology clusters. Grace
gives the MEMS industry an A for the two roadmaps in existence. Published
in 2003, MANCEF's International Commercialization Roadmap has 614 pages
covering standards, infrastructure, and other issues. Introduced in September
2003, the Nexus document "is a product-market roadmap created to
a large degree from inputs of the numerous Nexus User-Supplier Clubs."
awards technology cluster development a B. He added the subject to his
report card in 2003 to acknowledge the increasing value that regional
and federal governments place on micro- and nanotechnology as viable businesses.
At least 20 MEMS clusters have been established since 1989, when the first
one was formed in Dortmund, Germany. Again, Grace points out that these
high-tech geographic concentrations have been standard practice in the
semiconductor industry for decades, with Silicon Valley and New England's
Route 128 being two obvious examples.
says its survey turned up a difference in the expectations of fabs and
MEMS companies that outsource their production. The three issues involve
fab service, process standards, and comparisons between the MEMS and semiconductor
industries. MIG notes that fab customers believe that fab offerings are
too broad. An overabundance of design, packaging, and testing services
is not critical to customers and can cause economic hardship for the fabs
themselves, according to the survey.
standardized processes are still controversial for customers who prefer
semicustom processes because the clients believe their devices are unique.
Third, the MEMS industry and its semiconductor counterpart have distinct
differences and different maturity levels. At the current stage of development,
MEMS manufacturers cannot offer the same economic benefits as chipmakers.
The differences mean that treating the industries as one entity only hurts
the MEMS producers, the survey authors write. As a result, venture capital
stays away because of a fundamental misunderstanding of performance expectations.
examination of the survey results led the authors to recommend a course
of action that the MEMS industry should follow in order to maximize production.
The first of three recommendations concerns design for manufacturability.
The suggestions include working with universities to "create MEMS
programs that will drive the efficient use of manufacturing design guidelines."
Further advice is to identify funding in order to commission a group of
experts to produce a handbook on optimized design. The final suggestion
is to establish working groups in important technical areas such as process
control and reliability.
second recommendation concerns standardization. Under that heading, the
survey authors suggest three courses of action. The first is the creation
of a fab matrix that identifies core fab capabilities to help plants specialize
in specific processes. The second is the compilation of a best-practices
fabrication handbook, and the third is to work with SEMI, NIST, and other
standards bodies "to drive the creation of sufficient statistical
data and generate adequate reliability standards for relevant categories
of MEMS devices."
survey makes four recommendations under the economics heading. They are:
track time to market and identify opportunities for reusing processes
and design tools, facilitate the development of cost models for major
MEMS categories, work with groups such as SEMI "to drive consistent
roadmaps and cost models," and establish a MEMS conference as an
itself notes that the MEMS industry has had mixed success over the last
decade. The foundry-customer relationship carries a portion of the blame,
according to the company. Even though funds have been available for new
applications in areas such as optical telecommunications and biomedical
devices, high-volume manufacturing has yet to materialize.
Heaton insists that the company prefers to be considered a manufacturing
partner rather than a foundry. It's a key distinction, he asserts. "We
call it more of a manufacturing partner. . .because the foundry model
says you give us your GDS files [for reticle production] and we pump stuff
out. It just doesn't work that way in MEMS. We call ourselves a partner
because we offer this unique mix of design, prototyping, process development,
and high-volume manufacturing, and of course we have a lot of metrology
and test capability as well."
taking over the 14-acre site of a former data storage plant in 2000, IMT
has signed up 25 projects with 24 partners. One of the important lessons
learned has been that simple rules of fabrication do not apply and that
flexibility and keen attention to contamination issues are required because
of the need to incorporate non-CMOS materials in the process.
control is taken "very, very seriously here at IMT," says Shillinger,
the company's operations chief. "All of us came from production.
We're good at development. Obviously, we've been doing this for five years,
and 24 customers is evidence of that." —JC
Search | Current Issue | MicroArchives
Buyers Guide | Media Kit
Questions/comments about MICRO Magazine? E-mail us at email@example.com.
© 2007 Tom Cheyney
All rights reserved.