Applied Material set to deliver 300-mm tools
By Will Wade
EE Times
(10/15/99, 4:22 p.m. EDT)
SANTA CLARA, Calif. ? Anybody walking the floor at Semicon West the past few years could be excused for thinking the long-heralded shift to 300-mm wafers was close at hand. Nearly every booth has featured shiny, automated chip-making systems that smoothly slide 12-inch dummy wafers in and out of various chambers. But those demonstrated tools have already begun to slide into obsolescence with barely any 300-mm fab lines in production, and Applied Materials Inc. is preparing to roll out its next generation of 300-mm chip-making tools, though it's still unclear whether anybody is ready to buy them.
"Most of the tools that have been developed so far are for quarter-micron production, which was fine a few years ago when the chip companies first started talking about using 300-mm wafers," said Dan Hutcheson, president of VLSI Research Inc. (San Jose, Calif.). "What people want now are 0.18-micron or even 0.13-micron tools."
That's exactly what Applied will soon deliver. "We think that the year 2000 will be the year of 300 mm," said Dan Maydan, president of the world's largest equipment maker. Speaking at the annual International Symposium on Semiconductor Manufacturing, he said the company will deliver 0.13-micron 300-mm equipment for evaluation by the middle of next year, and that Applied's entire equipment line will be ready to support 300-mm wafers at that technology node by mid-2001. "There will be a lot of 300-mm wafers in production by 2002," he said.
Many potential customers, however, seemed skeptical about the prospects for larger wafers. "I don't think the 300-mm tool set is completely ready yet," said Steve Appleton, chairman, president and chief executive officer of Micron Technology Inc., delivering a keynote address at the same conference. "The chip companies won't start to adopt larger wafers until they can see a clear economic advantage, and right now there isn't one. We don't believe that 300-mm wafers will be the mainstream technology until about 2003 or 2004."
Applied doesn't provide every piece of equipment in the fab, and other links in the tool chain have not caught up with them. One critical bottleneck is lithography, where step-and-scan throughput is becoming a major limitation. "Lithography is the only part of the chip-making process where the cost to process a square-inch of silicon has actually been increasing," Maydan said.
Yoon Woo Lee, president of Samsung Electronics, said that current step-and-scan technology has not increased its processing speed, and the time required for today's tools to process a single 300-mm (12-inch) wafer is long. "The throughput for 300-mm wafers is just 40 percent of what we are seeing in 200-mm wafers," he said. "So even though you can get much more bits on the bigger wafer, there is no real output benefit."
Complicating the equation even further, Lee noted that the 300-mm tools are more expensive than the equipment for 200-mm (8-inch) wafers, and materials costs are higher. The larger-diameter silicon disks are at least three times the price and demand more gases and other chemicals to process. "There are still many challenges before 300 mm becomes a reality in mass production," he said.
First 300-mm products
But that doesn't change the very immediate need for more chip-making capacity. Lee and Appleton both said their own fabs are running at nearly full utilization, and the major foundries have also reported they are almost 100 percent full.
In the midst of the otherwise somber mood for the 300-mm shift, Semiconductor300, a joint venture of Infineon Technologies and Motorola, announced earlier this week that it has begun shipping its first products, 64-Mbit DRAMs, from its 300-mm fab in Dresden, Germany, at geometries below 0.25 micron.
Sue Billat, equipment industry analyst for BancBoston Robertson Stephens, predicted that six new fab projects could be launched next year, three facilities will begin major upgrades and one new DRAM fab will debut. Nine of these 10 projects will be in the foundry sector, she said, as more semiconductor producers, even large ones with their own fabs, shift at least some production to foundries.
But none of these projects are likely to implement 300-mm wafer lines. Although Taiwan Semiconductor Manufacturing Co. (TSMC), UMC Group and Intel Corp. have all publicly stated plans to move forward with 300-mm fabs, none is expected to use the larger form factor soon. "We will move to the larger wafers in our 0.13-micron generation," said Michael Splinter, senior vice president and general manager at Intel. "But we are depending on the equipment manufacturers to develop tool sets for the 300-mm platforms."
Appleton agreed. "If the tools are not available at least for the 0.15-micron technology node, then they just won't be competitive." He said that Micron expects to use so-called bridge tools, which run 200-mm wafers now, and can be upgraded to process larger wafers later.
While VLSI's Hutcheson noted that these compromise systems could be less efficient than dedicated tools, Micron's approach is probably the least risky and most cost-effective in the long run. Not only does this limit productivity loss during the transition, it gives fab managers freedom to move forward now, when capacity is needed, instead of waiting to see which way the industry moves.
"Nobody wants to build the first 300-mm fab or the last 200-mm fab," said BancBoston's Billat, noting that the higher-priced tools, along with an expected slower ramp because of the new technology, will make 300-mm fabs even more expensive than current ones. "Only the largest foundries and chip companies can afford the cost of a 300-mm fab," she said.
But that is not deterring Applied. "Our customers say they are very interested in 300-mm tools," Maydan said.
Meanwhile, lithography vendors are working to solve their own throughput issues. Klaus Rinnen, lithography analyst for Dataquest Inc. (San Jose), said the next generation of steppers using 193-nm light waves is expected to be pressed into service for the 0.13-micron process generation. Due to the different way that type of light will eat away the resist material, he said 193-nm technology would process wafers even more slowly than today's 248-nm tools.
In overall bit production, however, the tools will be adequate to support the demands of larger wafer production. Not only do the machines save time in the wafer-handling and alignment phase by moving fewer wafers, but the overall bits produced will rise because of the shrinking line widths. "On a per-area basis, those tools will be able to pattern more die in the same time," Rinnen said.
"Intel and TSMC have sent out a message that they are interested in this technology," said Billat, "but now it's time for them to put their money where their mouth is."
eetimes.com |
