Issue



Can fabs be built twice as fast for half the cost?


09/01/2002







If the industry were starting over from zero again, would some of these newfangled ideas make more sense than the way we build fabs today?

If someone could show how to build fabs twice as fast for much less money, double equipment utilization, and cut cost of ownership by 30%, you would think the semiconductor world would be beating a path to his door. Those are some of the claims made by Bevan P.F. Wu, a Palo Alto consultant who was once a manufacturing executive at IBM, for his Buckminster Fuller-like fab design. Yet the industry has not embraced his innovative concepts, except for a pilot project built in Taiwan by ITRI in the late 1990s. He described his concepts in a presentation at the IEEE/Semi Advanced Semiconductor Manufacturing Conference in Boston this spring.

Of course, there is a major hitch. All the industry process tools and automation would have to be repackaged and configured to slide into place in his totally modular roundhouse fab, looking somewhat like a sliced orange with some 36 segments surrounding a central hub. Process modules would be suspended from the spider-like superstructure, linked to five control centers located strategically within the hub-and-spoke configuration. Multi-pod carriers would shuttle FOUPs around on four concentric monorail rings. The superstructure could be in construction while tool modules were being built at many vendor sites, with rails on the tool housings to enable quick insertion. If a tool failed, a replacement module could quickly be inserted, keeping production humming.

Dr. Wu's contention is that process tools will always be less reliable than most other types of machinery because of the nature of the industry. In addition, it is often necessary to upgrade tools. Modular tools that fit into slots could be assembled during the initial fab construction, and then replaced quickly if failures occurred or an upgrade were needed. As part of the ITRI experiment in Taiwan, three modules were assembled, including a thin-film deposition system. They were shipped by truck, installed in five hours, and achieved Class 1 operation, according to Dr. Wu.

There have been other experimental programs to demonstrate alternate fab structures, such as an IBM project showing that automated sequential processing of wafers could be accomplished without human intervention and an Air Force-sponsored program at Texas Instruments involving sealed pods and built-in metrology. Much was learned from these projects, but even the new 300mm fabs continue to be built in traditional style, with cells for individual processing operations such as litho, etch, and ion implantation, and cassettes of wafers moving into stockers to get into the processing queue. While the new fabs are much more automated and use minienvironments to allow more relaxed cleanroom conditions for the operators, the architecture has not changed radically. The risk appears too high to implement a completely new scheme for structuring a semiconductor fab when the required investment is more than $2 billion.

But what if the industry were starting over from zero again? Would some of these newfangled ideas make more sense than the way we build fabs today? If some nation were willing to take on the challenge, provide backing for the investment in a new way to package and assemble process tools to try out an entirely new fab architecture, its chipmakers might be able to gain a significant edge in costs and operating efficiencies if the ideas turned out to be valid. It is unlikely that this radical approach would happen in nations already far along the learning curve in fab construction and operation, such as the US, Japan, South Korea, Taiwan, and those in Europe. Individual companies are leading the way in these countries as traditional business enterprises.

In China, however, where there are somewhat grandiose plans developing for a semiconductor infrastructure to meet massive future needs for electronics, and where very little has actually been built yet, there may be a possibility for a novel approach with great promise. Chinese officials often speak about gaining greater self-sufficiency in all phases of a high-tech economy. If almost everything is being built from scratch, from the process tools to the fabricating plants, they might consider a radical departure, as Dr. Wu is proposing.

There is no indication at present that any such futuristic project is in the wind. When interesting, novel ideas are floating around, however, and the industry faces transitions to more complex system chips, there could be some surprises ahead. We'll keep you posted.

Robert Haavind
Editor in Chief


NEW THIS MONTH: REQUEST INFORMATION ONLINE!
Go to www.onlinecenter.to/sst and click on Online Reader Service. Select by inquiry number or by vendor. Your request for information will automatically be forwarded to the appropriate manufacturer or vendor.