OEMs, chipmakers must improve cooperation
clouds didn't part and the earth didn't shake, but a key message sprang
from SEMI's recent conference on semiconductor materials. Collaboration
and equitable intellectual-property arrangements will be necessary if
the industry is to have a prayer of developing advanced materials in a
longer able to count on "gifts from God"—as one executive calls traditional
chipmaking materials—suppliers, OEMs, and semiconductor manufacturers
must work together to confront the scaling challenges ahead, said several
speakers at the trade association's Strategic Materials Conference in
OF THE PIE: The process materials market (excluding silicon wafers)
grew to $8.5 billion in 2004, with the mask and gases segments
leading the way.
ILLUSTRATION BY LAUREN NAGODA
his keynote speech on emerging device architecture- and technology-driven
requirements for new materials, Intel's Ken David emphasized that the
industry no longer has the luxury of long development times. "We need
better methods to collaborate clearly in an open sort of environment,"
says the director of the components research group at Intel. "The materials
suppliers themselves don't have the equipment or the experience to understand
the integration effects of these films. The only way to shorten the time
to implementation is through much clearer interactions between IDMs and
stay on track with Moore's Law, David advocates a concept he calls equivalent
scaling. The concept involves geometric scaling combined with innovations
that are furthered by strong industry cooperation.
multi-billion-dollar materials segment alone faces several tough tasks
in the coming years, according to analysts who spoke at the Half Moon
Bay, CA, meeting. IP ownership and licensing by OEMs to materials vendors
is just one of these hurdles, notes Karey Holland, a senior managing partner
of Techcet Group.
whose firm specializes in process materials market research and analysis,
enumerated additional challenges. These include continuing consolidation
and reduction in the supplier base, downward pressure on prices, a shift
of materials manufacturing to Asia, and reduced R&D spending.
silicon wafers, the process materials market took in revenues of $8.5
billion in 2004, Holland says. As the chart on page 18 illustrates, photomasks,
gases, photoresists, and CMP consumables composed the major segments of
the overall materials pie at 25% and 23% for the former two categories
and 10% each for the latter two. In order of size, the remaining segments
are wet chemicals, ancillaries, quartz, targets, crucibles, graphite,
and silicon carbide (SiC). Photomasks, gases, CMP, wet chemicals, and
targets show the greatest growth potential.
singled out intellectual property as a big sticking point. "IP ownership
is a real hot button. OEMs are trying to grab the IP. That means a large
price increase gets passed on to the end-user. That is going to continue
to be a problem." She cites patent moves in the CMP sector by Applied
Materials as one example.
search for replacement materials and integrated processes has been somewhat
rocky, and it's unlikely to get any easier. The industry's technology
roadmap is strewn with red-brick roadblocks involving optical lithography
limits, short-channel effects, and scaling limitations of gate dielectrics
as well as junction and silicide scaling, notes conference speaker Mohammad
Mirabedini, manager of advanced technical development at LSI Logic.
whole industry is based on silicon dioxide," says John Poate, CTO of Axcelis
Technologies. Poate, who participated in a panel discussion on the IP
issue, emphasizes that the industry is unsure whether high-k materials
such as hafnium oxide are viable alternatives to silicon dioxide. "There's
been a lot of research in the past five to 10 years, but it's proving
more difficult than people realized. It's really the first time that radically
new materials are being brought in, aside from these gifts from God."
he notes, is "80% an implant company" that by nature is "integrally tied
in with suppliers, except for source-gas materials. But that game is changing,
and that's one reason I was there." Axcelis's curing and strip divisions
are "intimately involved with the materials suppliers, and a couple of
those were on the panel."
came away from the conference with the sense that Axcelis's approach to
handling strategic materials and IP through carefully crafted joint-development
agreements is "a bit of a model," Poate believes. In these JDAs, Axcelis
owns the hardware, the materials suppliers own the materials, and both
sides divvy up the process part.
Robitaille, the chief patent counsel for Axcelis, elaborates: "We protect
our core business with respect to hardware to make it a win-win situation.
We allow the materials companies to protect their domain with respect
to materials, and the process technology that comes out of that is jointly
admits that both sides owning "equal parts of the whole" is a difficult
concept to grasp, and he urges caution: "You don't want to have technology
development in which your joint partner goes out and takes ownership of
technology that you want to exploit and provides it to a competitor. This
is a way to divvy up that technology and the ownership so that you still
protect your core interests."
JDAs in this way furthers Axcelis's goal, which is "to sell our tools
to all the competitors," Poate emphasizes. "Just to put this into context,
the 800-pound gorilla in the closet is Applied Materials. They probably
are trying to get into the materials business as well in order to start
to own this stuff. The materials suppliers are under a lot of pressure."
think that where our approach is somewhat novel," adds Robitaille, "is
that we're willing to acquiesce to the material makers and give them their
value out of the joint development, whereas an Applied Materials might
use a more heavy hand. That's certainly the approach we use with certain
types of arrangements. We are a little more open-minded about it."
suppliers in particular "are under an awful lot of pressure," Poate asserts.
There's the small matter of how to fund the R&D of new materials and
get a fair financial return for your trouble. In the panel discussion
on IP, that topic was a "real issue that wasn't resolved."
overarching principle governs the Axcelis agreements? Well, just look
to your backyard. "Good fences make good neighbors," explains Poate. "We
really get to know our partners well so when the agreements are crafted
they're crafted to work to the model with these watertight compartments.
company is working with Dow Chemical, ATMI, and other suppliers, Poate
says. Axcelis's approach to IP has garnered it several invitations to
appear at seminars "focused on attracting the customer side."
another analyst and speaker, an industry shakeout may be just what the
fab doctor ordered. "You're looking at a $12 billion—maybe a $14 billion—industry
if you include the wafer segment," points out Mike Corbett, director of
electronic chemicals and materials from Kline & Company. Echoing some
of Holland's points, Corbett says chemical suppliers are finding it hard
to generate adequate financial returns and developing the innovative materials
to keep the industry on the technology roadmap.
says many large chemical companies can choose to invest their capacity
dollars in industries other than chipmaking. Several gas vendors, for
instance, are involved in healthcare or in "setting up large hydrogen
plants and refineries. The point is that they have alternative places
where they can invest. If the returns aren't there, we can see some investment
migrating out of the [semiconductor] industry into other end-use applications."
says the industry has "structural problems," which is a euphemism for
too many players. "We need to see some consolidation on the technical
side, but we're really not seeing it happen." Sure, "you hear about acquisitions,"
and he quickly names a few: EKC bought by DuPont, Ashland Chemical by
Air Products, Arch Chemicals by Fuji Film, and recently, Merck's electronics
chemicals unit by BASF. "So we're seeing that go on, but nobody is shutting
any [of the companies] down.
question comes to this: From the chipmakers' perspective, there are seven
large photoresist suppliers in the world and everybody is developing new
193-nm products. Who's really paying for that? The chipmakers are ultimately
paying for these massively parallel development efforts."
only three or four companies were developing these resists, you would
save money because there would no longer be such a duplication of cost
and effort, the analyst asserts. Nevertheless, Corbett acknowledges, "Each
one wants to be a survivor."
believes the melding of Shipley and Rodel into the multiunit Rohm and
Haas Electronic Materials business is a good synergistic example. The
companies are able to focus "a bigger source pool" on innovation and process
efficiency. "Everybody benefits. The chipmaking companies might not say
they're benefiting because they have limited pricing leverage when there's
some dominant suppliers."
competition good for customers, though? "I don't disagree with that,"
Corbett replies. "This is obviously a technology and service—rather than
a capital-intensive—business on the chemical side. You have a lot of
duplication of technology development. You have to remember that chemical
suppliers could be spending 10 to 15% on research and development, and
ultimately the customers are paying for that."
Holland says the recent emergence of a government-supported Japanese chemical
consortium will put even more pressure on producers in the United States.
"If they do this successfully, they could eliminate the ability of the
U.S. chemicals suppliers to sell chemicals."
and end-users need to collaborate if only to prevent the introduction
of unworkable solutions at the cost of millions, according to Holland.
She cites IBM's decision last year to switch from Dow Chemical's SiLK
spin-on dielectric to a more entrenched CVD approach at the 90-nm node
as an example of how tough the integration challenges are. Dow believes
the material can work at the 65-nm node.
and suppliers have got to work together because we never know what we
don't know," Holland says. "If IBM had any idea it was going to fail in
final chip build, they sure as hell would have tested it. You can't just
put all that risk on the poor suppliers."
a prime example of the trials and tribulations of materials integration,
one need look no further than the struggles with copper. "When people
introduced copper damascene five or six years ago, everyone who went into
this, even with more-standard dielectrics, got yields of around 70%,"
Holland points out. "No one was used to 70% yields. Everyone is used to
95 to 100%. Some of the companies say they've made it over the hump, but
I think they made it 'over the hump' at 80%, which is why DRAM companies
have not gone to copper yet."
you do the detailed studies, you just don't understand these things,"
Intel's David emphasizes. He would like to see better tools for integration
such as the existing pore symmetry tools "that can measure the fraction
pore content of an ILD film. Having those data at our disposal allows
us to predict a lot of process interactions to characterize the films
of the "big analytical tools like SEMs and TEMs and SIMS...are very good
at a bulk microscopic level. What I'm looking for is a suite of tools
that give better insight into what's happening at the molecular and structural
levels." These capabilities exist primarily at university research labs.
David remains optimistic. "The challenges of scaling are apparent in a
lot of different areas. We're looking for material solutions in almost
every one. I believe there's progress being made in most areas."—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.