Caution marks path to 300-mm wafer production
By Anthony Cataldo
EE Times
(12/01/99, 12:48:39 PM EDT)
TOKYO ( ChipWire/EET) -- After several years of delay, a few semiconductor manufacturers are starting to map out definitive plans to install 300-mm wafer production lines now that first-generation 300-mm equipment has been finalized. But the industry is still far from a consensus on when to proceed with the transition to the larger wafers, according to observers here at Semicon Japan.
Hitachi Ltd., for one, has reactivated its 300-mm (12-inch) pilot program at its advanced chip-making factory in Naka, Japan, said Toshiyuki Uchino, senior engineer for front-end manufacturing technology at Hitachi. Uchino declined to say when Hitachi plans to produce 300-mm wafers in volume.
Hitachi's revised plan comes less than six months after Intel Corp. announced plans to build a 300-mm pilot facility in Hillsboro, Ore. Intel will start installing 300-mm equipment in that facility next month, and expects to start production with a 0.13-micron process technology using 300-mm wafers in 2002, said Devadas Pillai, a director at Intel's technology and manufacturing group in Chandler, Ariz.
But interest in shifting to 300-mm wafers -- which will provide 2.3 times more chips per wafer than today's 200-mm wafer -- is not shared by all major chip manufacturers. Even though 300-mm production is capable of yields equal to or better than 200-mm wafers, the tools are only 38% ready, said David Zeglinski, group manager of lithography and etch for Semiconductor300 GmbH, a joint-venture fab of Motorola Inc. and Infineon Technologies AG that produces 64-Mbit DRAMs, and which is currently the only fab producing chips for sale from 300-mm wafers.
Zeglinski said it doesn't make sense to invest in a large-scale 300-mm fab until the wafer cost/centimeter2 is reduced by at least 30%. "We shouldn't spend the money today," he said. "As for next year, we can say that we are hopeful."
Chip companies first started contemplating the move to 300-mm wafers in 1992. That interest later spawned consortia such as the International 300 Initiative, which was formed by Sematech, and Selete, which was formed in Japan. The step to 300-mm wafers was a change from past wafer transitions, which were generally driven by a single company per transition. Once expected to occur by the end of this decade,the transition to 300-mm wafers has been delayed largely by the Asian financial crisis and the semiconductor downturn of 1998.
Companies are now taking a second look at 300-mm wafers, but are cautious about making concrete plans or are expanding existing factories before making the move, observers said. "People are a little bit gun-shy about making bold statements," said Ashwin Ghatalia, director of the International 300-mm Initiative. There are "serious internal debates" at IBM Corp. -- Ghatalia's employer -- over when to make the move, he said.
"A lot of people are standing on the sidelines, and want to be prepared to make the move when they need to," said Robert Hollands, vice president of Steag Electronic Systems Inc. of Tempe, Ariz., which makes rapid thermal processing equipment.
Novellus Systems Inc., which makes metal and dielectric deposition equipment, will spend a significant amount of money on 300-mm R&D next year despite fence-sitting by customers. There's still disagreement in the industry as to when the transition should begin, but the current consensus is that some device makers will start 300-mm wafer production in 2002, said Mark Fissel, vice president and general manager of the integrated metals division at Novellus, based in San Jose.
That hesitancy has created a communications gap between semiconductor equipment suppliers and IC manufacturers. "Even today, the most sophisticated suppliers don't know how IC makers are going to run their [300-mm] factory," Intel's Pillai said.
A 300-mm line now offers a 50% "investment efficiency" advantage over 200-mm wafers, when factors such as capital costs, equipment throughput, factory real estate and chips per wafer are considered, said Hitachi's Uchino. But an improvement of 100% or 110% should be the target, he said. Also, a 300-mm fab line should run in high volume. Unless a 300-mm factory produces at least 3,000 wafers per week, Zeglinski of Semiconductor300 estimates that manufacturing costs will rise relative to 200-mm lines.
Despite delays in settling on a few standards for 300-mm equipment and getting them published, observers said that effort to standardize equipment has been a success, considering that there are more than 50 specifications that have to be met, or more than 10 times the number for 200-mm tools.
But there are still some issues that have yet to be resolved, such as defining a standard for the vessel that transports wafers between pieces of equipment. And some critical process steps have shown poor results on 300-mm wafers, such as chemical mechanical polishing, observers said.
And 300-mm tools still need to be fully evaluated at existing pilot lines. Selete hopes to finish evaluation of the tools by next March. Intel will evaluate the equipment throughout next year and 2001, Pillai said.
"Having standards is not enough," Pillai said. "This is where the rubber hits the road."
|
