AUSTIN, Texas F. Scott Fitzgerald said there are no second acts in American lives, but International Sematech doesn't believe it. The chip-making research consortium, which began operations here 15 years ago with a manufacturing mandate, now faces a more-complex challenge: how to redefine itself in an era of increasingly sophisticated materials and manufacturing issues, and an environment where money is hard to come by.
Bob Helms, who was appointed president and chief executive officer last year, envisions Sematech applying its modus operandi namely, identifying critical roadblocks and then marshaling the wherewithal throughout the industry to surmount them to what is loosely called "beyond classical CMOS." The Sematech board will look at the issue of stepping up work in this esoteric realm, which includes such technologies as fully depleted silicon-on-insulator, at its next meeting, in November, Helms said.
"We need to harness our ability to identify the gaps, to make sure all the core competencies, the materials, equipment and processes, are available in time," he said. "You could say that this 'beyond classical CMOS' is a new direction for Sematech, which is true, but it leverages off of what we are doing now."
Helms added that there are "plenty of issues around FD fully depleted SOI, in processes, tools and materials, to cause us to take our current programs and make some fairly significant changes over the next couple of years. It is coming, and coming sooner than some people would believe."
Meanwhile, Helms gained a feather for his Stetson by negotiating an agreement with the state of New York and the State University of New York at Albany to establish International Sematech North, a research center that plans to spend $320 million over the next five years. All but $40 million will be supplied by New York state and IBM Corp., with a goal of developing the infrastructure for the coated mask blanks needed for extreme-ultraviolet (EUV) lithography.
By all accounts, Sematech has accomplished its original "Uncle Sam Needs You" mission: to bring the U.S. equipment industry up to snuff, at a time when the federal government feared that the Japanese semiconductor equipment industry was holding back its best machines for early use by Japan's powerhouse chip makers. U.S. chip makers and fabless semiconductor companies now account for about 54 percent of worldwide semiconductor revenues, according to IC Insights. And the U.S. equipment industry is in an equally strong competitive position.
Mark Melliar-Smith, International Sematech's CEO from 1997-2001, noted that Sematech's original role was to bolster a relatively immature group of U.S. equipment suppliers that, back in the mid-1980s, lacked first-rate technical talent. "There has been tremendous improvement by the equipment industry over the past 15 years, an unsung activity. That means that the need for support from Sematech is changing to more rifle-shot activities, such as in mask making or the very, very major challenges involved with new-materials integration or EUV lithography," said Melliar-Smith, now a partner at venture-capital firm Austin Ventures here.
One shining example of the rise of the U.S. equipment industry is Applied Materials Inc., which owns about 16 percent of worldwide equipment industry revenue. While Applied continues to actively participate in Sematech programs, and its machines populate Sematech's 300-mm back-end-of-the-line (interconnect) facility here, Applied has recently opened its own, brand-new process development fab, the Dan Maydan Center, at its Santa Clara, Calif. headquarters and is doing more of its development work there. Ironically, Tokyo Electron Ltd. the world's second-largest equipment company has become the equipment house that works most closely with International Sematech.
Cash is king
During the industry's worst downturn ever, Sematech has found itself scrambling for funds, increasingly asking for donated gear at a time when equipment costs are skyrocketing. The chief technology officer at one midsize U.S. equipment vendor, when asked if his company participates in Sematech, barely concealed his scorn: "Sematech doesn't have any money and can't pay. The only companies that can afford to put equipment there are those with broken business models."
While that view is harsh, it reflects the reality that Sematech's annual budget was in the $250 million range during the consortium's first four years, when Washington supplied half the funds and 14 U.S. chip makers the remainder. Sematech stopped taking federal funds in 1996 and now includes five non-U.S. chip manufacturers among its dozen members. Its annual budget has been about $140 million, plus or minus $5 million, for the last six years. To be sure, that's larger than the budgets at similar consortia in Japan (Selete) or Europe (IMEC). But Sematech has a wider range of activities.
The industry has been paid back in spades for its investments in Sematech; even ardent critics say that its efforts in narrowing down the next-generation-lithography candidates, defining low-k dielectric candidates or searching for new photoresists that will work with 157-nm lithography are priceless.
Nevertheless, Melliar-Smith said he faults himself for not being tougher during his years as CEO, for failing to quickly shut down programs that had outlasted their usefulness.
One manager who often visits Sematech was more blunt: "The problem is that they still try to do too many things, trying to be all things to all people, instead of just focusing on a few areas where they could really make a big difference. They don't have the resources to match what the leading companies can do, like what IBM can do at Fishkill N.Y. or what Intel can do in Hillsboro Ore..
"The budget 10 years ago was double what it is now, at a time when equipment was about 10 times cheaper. Finances are so tight now. If Sematech proposes new programs, they must be funded, and it is awfully easy for people to say 'No' while business conditions are so bad," he said.
More R, less M
Bob Havemann, who earlier managed Sematech's successful copper interconnect program as an assignee from Texas Instruments Inc., agrees that "Sematech definitely needs to move in a new direction. In my opinion," said Havemann, who is now vice president of process integration at equipment maker Novellus Systems Inc., "they need to get involved with more forward-looking research projects, get more involved with universities, and move away from the old Sematech, where the emphasis was on manufacturing."
The cry to trade manufacturing for research is coming in loud and clear among Sematech's bosses. CEO Helms lists lithography, high-k gate insulators and gate structure, and low-k insulators and interconnects as the three largest research areas within Sematech today. By its charter, Sematech does not get involved with designing transistors or in the final process integration steps needed to make those transistors. It leaves that up to the individual member companies. Instead, Sematech's staff of 600 zeroes in on potential stumbling blocks to "advancing the road map" and keeping Moore's Law on track.
Next on the agenda is the frontier beyond CMOS. "If you look at past discussions with our board, the message in past years was that the end of classical CMOS was still several years away. This year, that changed. It is real clear," Helms said. Members are saying, "we need to make sure the infrastructure, the process tools, the new materials and the processes themselves are going to be there to facilitate bringing this into our transistor portfolio."
The longtime Stanford University engineering professor said he puts partially depleted SOI in the standard-CMOS category and FD-SOI in the "beyond classical CMOS" class.
"Fully depleted structures allow you to take advantage of the high-k gate insulators, and by that time the metal gate electrodes also will be needed. The International Technology Roadmap for Semiconductors says that beyond classical CMOS, including metal gates, will be in products by 2009. I will make a prediction: That's way too late. Some of the member companies will have these technologies in products before then. If someone can come in earlier, with fully depleted technology that is yieldable, they will win."
SiGe export controls
Historically, Sematech has played a key role in defining standards for next-generation wafers, and Helms said that expertise will allow it to evaluate FD-SOI wafers. "There are wafer standards issues surrounding thin-body SOI, the uniformity from supplier to supplier, where the wafers need to be in five years areas where standards play a very critical role in the ability to do processing. We need to identify key gaps in the resource space and say, 'Aha! That's a gap, and that is where we want to focus our resources.' "
Asked if, resources permitting, Sematech could play a similar role in the development of strained-silicon technology, Helms said export control issues may hamper such work, since some aspects of silicon germanium are under tight regulations. (In strained silicon, a graded SiGe layer is used to place a strain on the active silicon lattice, enhancing carrier mobility through the channel.) "Strained silicon is part of the board's assessment of new directions," said Helms.
Meanwhile, the Sematech North initiative, which will send dozens of Sematech lithography technologists from Austin to Albany, has some Texans worried. Rinn Cleavelin, who until recently served as Sematech's chief operating officer and director of the front-end processing program while on assignment from TI, said Texas needs to step up to the plate if it wants to keep Sematech in Austin over the long term.
"It should be a concern to the leaders of Texas when they see one of the key programs, like lithography, start to move out to New York because of the subsidies New York is offering," Cleavelin said.
While at Sematech, Cleavelin said he proposed that the state of Texas pony up about $100 million to build a new fab at the Austin site. Over the next decade equipment suppliers could use that fab to sort out issues surrounding the next generation of wafers, expected to measure 450 mm (18 inches) in diameter. (A Sematech spokeswoman said there are no current plans to start research into 450-mm wafers at Sematech.)