INDUSTRY NEWS

SEMICON WRAP-UP:
New materials, 300-nm tool changeover test industry mettle

For its next trick, the semiconductor industry will emerge from one of its worst down cycles, shift most production to 300-mm wafers, and introduce a host of untried process materials. Drum roll, please.

It's not quite the equivalent of juggling five dinner plates while balancing on one foot. But the message that emerged from selected seminars, conferences, and speeches at Semicon West this year makes it clear that front-end processes from this point will present strong challenges to chipmakers, suppliers, and researchers. Several speakers at the mid-July event insisted that meeting those challenges would require increased collaboration among the various parts of the semiconductor community.

FOCUSED ON GROWTH: Attendees at this year's Semicon West heard forecasts of double-digit market growth in 2004 and 2005. PHOTO BY ERIC HOFFMAN/JOE ORLANDO PHOTOGRAPHY, COURTESY OF SEMI

At least the industry will confront these hurdles under somewhat sunny skies. One big piece of good news came on the trade show's opening day with the announcement that SEMI expects to see a 63% rise in sales in 2004. Member companies responding to the trade association's midyear capital equipment forecast expect to sell $36.3 billion worth of front- and back-end equipment this year.

As the industry recovery gains steam, the tool market is forecast to grow 24% in 2005, taking in $44.8 billion, according to SEMI. A 5% dip in 2006 will give way to sales of $47.5 billion in 2007, a growth rate of 12.5%. Fifty-two percent of the survey's respondents believe the revenue increase will peak in either the second or third quarter of 2005.

A bright spot in the growth pattern is that "it's extremely broad based," said Stanley Myers, SEMI's president and CEO. Equipment bookings and billings are in sync. Noted Myers: "There's no sign of overheating in the chip equipment market."

The transition to 300-mm tools and processes has begun in earnest this year. "The 300-mm crossover predicted for last year will come to pass in 2004," Myers said. Fifty-five percent of the $23.69 billion in global revenues for process tools will come from sales of 300-mm equipment. In 2005, the portion of 300-mm sales will grow to 62% of sales, which are pegged to reach $29.69 billion.

Myers said the 200-mm market has been "much stronger than expected." This persistence has slowed the acceptance of 300-mm process gear and kept the tools from crossing the 50% threshold. The transition has been slowed because adapting to the new size has proven more complicated than expected. Additionally, Myers noted that the industry has had to "climb out of the yield-learning curve."

This issue also surfaced during a panel discussion on IC materials presented by DuPont. Assembled to tackle the topic of challenges facing the migration to the 65-nm process node, five panelists addressed the transition from 90-nm processes, materials integration, the need for new business models, and examples of precompetitive collaboration.

Panel member Bill Rozich, IBM's director of 300-mm semiconductor operations, commented that the chipmaker is wrestling with the heavy demands of 65-nm processes. "We find that the very tight ACLV [across-the-chip linewidth variations] requirements at the gate level at 65 nm are driving such things as tool matching and tool dedication— things we find from a manufacturing perspective that are creating a significant amount of difficulty."

Rozich asked whether manufacturers have definitively decided that the node under discussion will take place on the larger substrates. "Is this still a matter of debate? Is 65 nm for the 300-mm generation, and will 65 nm be built only on 300-mm wafers? Or is there probably still going to be [processing] on 200 mm?

"I think the jury may still be out on that one. Basically, we think it is a 300-mm generation, but if you look again at the tightness of the specifications that have to be held across the wafer, this creates even more of a challenge as you move down the technology roadmap."

Rozich said the transition from 90 nm to 65 nm is "the end of traditional scaling. We see there's going to be much more of an emphasis on material changes. This is fraught with danger. As you scale, you're dealing with an animal that you know. Whenever you start dealing with material changes, you end up with surprises. We have had quite a few surprises. So we see the transition from 90 to 65 as being a very remarkable adventure, and we are moving into a different phase of the technology."

Moderator Ken Monnig, International Sematech's associate director of interconnect, asked what technology changes the migration to 65 nm may require. He also wondered whether "radically new" materials are needed in the change from 90 nm to 65 nm, and specifically what changes may be needed in low-k dielectric films.Mehdi Moussavi, equipment supplier manager for French consortium LETI, replied that the type of product is an important consideration at this node. "High performance, general purpose, or lower power—depending on which product you are looking at, the question will be different. For low power, probably SOI is required, and that's probably the case for high-performance [products]."

Regarding technology shifts, Moussavi said researchers at the Grenoble-based applied research laboratory have worked with both spin-on and CVD low-k dielectrics with a "next-generation as-deposited k-value of around 2.3 or 2.4 and an integrated k-value in the 2.6 to 2.7 range. Our position today with the current maturity of low-k dielectric [is that] we will not introduce lower-k materials at 65 nm." He added that high-k dielectrics for the gate stack are not mature enough "to make manufacturing people comfortable enough to introduce high-k dielectrics right now."

LETI has had to play "more and more tricks as we put more stress and strains on these different films," Moussavi emphasized. "We're not even dealing just with the properties of the films themselves but also the conditions with which they either have to be deposited or stressed."

Panelist Mark McClear of Dow Chemical agreed with IBM's Rozich about the scaling challenges posed by the move to 65-nm processes. Materials sets will have to change, McClear pointed out. "Fundamentally, these materials sets are different from the ones we've used before. In the past 20 or so years we've basically been taking industrial chemicals used in pharmaceuticals and other secular industries, and purifying them and packaging them and delivering them."

McClear, the supplier's global business director for advanced electronic materials, praised the industry's ability to meet "demanding purity levels" for these chemicals. However, he emphasized that manufacturing at the 65-nm node changes the rules of the ball game. "Now we've moved into a different time frame where we're inventing whole new molecules. When you're inventing all-new molecules, it takes real time and real R&D dollars, and it takes a long runway to get an all-new material developed."

The semiconductor industry has seriously underestimated the amount of time needed "to invent the molecules and then to get a tool set perfected that's compatible with that molecule, and then get the integration done and then get the reliability done," McClear insisted. "It takes most of a decade, and then you have to look back at the first seven, eight, nine years when this adventure started and say, 'Was it economically viable?'"

For processing at the 65-nm node and beyond, McClear believes the industry will need many new materials. The challenge, he emphasized, will come in motivating "materials suppliers like Dow and DuPont and other suppliers ... to really put their R&D dollars at risk for doing that, without some sort of balance or different business model."

LaMar Hill, business director of Albany Nanotech, commented that materials integration presents formidable issues. "The challenges of integration at the molecular level, and the commensurate ability to monitor that interface and get that interface to be repeatable, is going to be a challenge at 65 nm and probably become more of a challenge as we move farther down the progression. Another issue is wafer cleaning. You're trying to deal with issues of surface tension and water getting into these structures that are going to become very challenging."

New business models and industry collaboration are two potential roads the industry can take, the panelists agreed. McClear pointed to the investment by Intel's venture capital group in Tri Chemical Laboratories of Japan. "That's the kind of thing that can prime a pump to get the materials companies over this 10-year horizon and actually make the impact downstream on inventing new materials. Addressing the business model question, the thing that has to change is the way the whole materials companies supply-chain transaction works."

Moussavi of LETI would like to see "more collaboration between material companies, R&D centers, and manufacturing companies—IDMs or foundries—to reduce" the 10-year development time. Industry players can alleviate uncertainty over materials integration by developing joint development agreements or other programs with an R&D institute, he said. "Most of the industry is not really willing to introduce exotic materials because of the contamination risks. At the very early stage, you need to work with the R&D centers."

Moussavi said the proximity of three chipmakers—STMicroelectronics, Freescale, and Philips (the partners in the Crolles2 alliance)—to LETI's Grenoble site simplifies its developmental work with materials companies. "This is the plan we have in place: working with material companies to start developing together and, as soon as possible, to take that to short loops on real test wafers" to determine how well the materials work with one another.

Addressing a suppliers' conference later in the week, Sematech's Monnig said the sheer array of new materials other than the basic silicon, oxygen, nitrogen, and aluminum has created great potential for process snags. With the cost of idled fabs approaching $100,000 per hour, Monnig said the custom types of materials that factories need will require collaboration between both materials and tool suppliers. — JC