The Big Debate: 450 mm or Not?

It's fairly rare in the semiconductor industry to have such a sharply defined and hotly debated topic as the need to transition from 300 to 450 mm wafers.

It's exactly the same argument from the proponents of change as with every other wafer size change: They would like to produce more die per wafer while keeping equipment cost about the same. Industry economics dictate that such a transition could be required by 2012, according to the International Technology Roadmap for Semiconductors (ITRS). This view, mostly driven by the largest semiconductor companies — Intel, Samsung, TSMC and Toshiba — is also fueled by the thirst for that killer competitive edge. The companies with deep pockets can afford a $10B 450 mm megafab, while others simply cannot, so guess who wins?

The alternative view to 450 mm is that there is still plenty that can be done to drive 300 mm economics (and 200 mm, for that matter) by improving fab efficiency and productivity before a switch to 450 mm wafers is required. The pace of efforts in this regard has recently picked up. Last month, for example, AMD, Freescale Semiconductor, IBM, Qimonda, Renesas Technologies and Spansion gathered for Next Generation Factory (NGF) meetings in Austin in an effort to place a more intense focus on 300 mm fab productivity improvements.

Outside of the 300 mm productivity vs. 450 mm transition debate over economics, there are still plenty of questions regarding the technical feasibility of 450 mm. The assumption seems to be that all of the manufacturing processes — lithography, plating, deposition, thermal processing, etching, implantation, etc. — can be done with equal or better uniformity and comparable throughput over the larger wafer, but few studies have been done.

Even something as basic as an adequate supply of 450 mm wafers is questionable because the economies of the incumbent Czochralski crystal pulling, wafer slicing and polishing processes are not known.

The real stumbling block is that the pain of moving to 300 mm is still vivid among most tool suppliers; it was years before suppliers saw the ROI that was promised. Writing at the end of 1998, two years after the 300 mm generation was to have been in full swing, George A. Lee (one of the main champions of 300 mm) wrote in SI: "The equipment and materials sector had expended huge amounts of energy, time and money to make certain that they were ready for 300 mm demand as soon as their customers, the device manufacturers, raised the green flag. Instead, the caution flag was raised; everyone slowed down; there were hazards on the track. The industry is about to enter 1999, and the green flag has yet to be displayed; there are still impediments to progress in its path." It does not take much of an oracle to envision exactly the same situation happening with 450 mm.

Tool suppliers aren't exactly on the fence on this. They simply aren't doing the development work that is needed now for a 2012 rollout. End of story.

Or is it? History also tells us that money talks. Ex-IBMer and former Sematech COO Paul Castrucci tells a story of how IBM decided it needed to make the move from 5 in. to 8 in. mm wafers and simply presented Applied Materials with a check for $4M; four years later, the first 300 mm tool shipments were on the way (see my "Semi-Conscious" blog on our website for details).

Tom Abell, manager of the 450 mm program at ISMI, said their recent decision to expand the 300 mm productivity-centric 300mmPrime program to include 450 mm was based on nine months of intensive modeling that produced more than 130 full-factory, dynamic simulations. These efforts, representing over 5000 hours of programming and analysis, investigated whether anticipated improvements from 300mmPrime alone could meet ISMI's goals of reducing manufacturing cost per area by 30% and cycle time by 50%, according to ISMI.

"The data shows that 300mmPrime has theoretical potential to achieve the targeted reduction in cycle time, but it does not sufficiently address our members' cost reduction needs," Abell said. "It appears that the key to significant wafer cost reduction can be found only in larger wafers."

The big question is who will pay for it? A footnote in the ITRS captures it well: "Although 450 mm is colored yellow, indicating manufacturable solutions are know, it could have easily been colored red, because there has been no acceptable economic solution for funding identified by the industry."