IBM Corp.Armonk, N.Y.

Dec. 21, 2004
Complementary Metal-Oxide Semiconductor (CMOS)
Silicon-on-Insulator (SOI) Technology

Since the early 1980s, scientists have searched for a way to better insulate the internal structure of the modern semiconductor. Such a development would unlock the door to a generation of microchips that manage power consumption far more efficiently than today's designs. Yet, only a couple of years ago, many scientists -- including those at Intel Corp. -- publicly stated that a breakthrough in so-called SOI technology was almost entirely out of the question. The physical and technical challenges were simply too great. But in the topsy-turvy, breakneck world of semiconductor design, it's never a good idea to bet heavily against any potential technology. In this case the naysayers' predictions were shattered. In August 1998 IBM announced that it had indeed perfected SOI. The accomplishment, which culminated more than 10 years of research and development by Big Blue, has far-reaching repercussions for computers, consumer electronics, and industry in the months and years ahead. The innovation works with mainstream complementary metal-oxide-semiconductor (CMOS) technology to deliver a 25% to 35% performance boost or a reduction in power of 1.7 to 3 times previous levels, depending on how the chip is tuned. That's significant for high-end PCs -- including mainframes, midrange units, and workstations, but also allows designers of hand-held PCs and portable devices to build products that boast far greater battery life and more robust features. "SOI can be used in any type of computing device requiring a microchip. It complements existing technologies and allows designers to squeeze more from what they currently have to work with," notes Bijan Davari, vice president of Advanced Logic Technology in IBM's Microelectronics Division. According to IBM, SOI represents a performance improvement over bulk silicon CMOS chips equivalent to one to two years of scaling (the current SOI chip matches the speed of a conventional bulk CMOS chip two years from now). Equally important: The performance delivered from SOI cannot be achieved using any bulk CMOS technology today. "This type of improvement allows us to continue the rapid pace of development that has driven the semiconductor industry over the last 20 years," says Davari. The biggest challenge for IBM's team of researchers was finding a way to develop the thin layer of defect-free material that's needed to insulate the chip. As recently as 1997, Intel researchers claimed that SOI material is 50 times more prone to defects than bulk silicon and that it would lead to lower chip yields during the manufacturing process. Once IBM had conquered that challenge, it still faced the daunting task of designing the SOI transistor so that it could operate at maximum speed. All previous research in the field had indicated that SOI would slow performance to unacceptable levels. Undaunted, IBM plowed ahead with the project until it found a way to sidestep that problem through a more refined method of fine-tuning. The end result is a device that's already sending ripples through the computing world. The practical benefit of SOI is that it allows designers to build more powerful processors capable of handling advanced technologies such as speech recognition and computer-aided design. Over the next two years, the technology will work its way down from high-end computers to mainstream PCs. On a smaller scale, it will make hand-held devices more practical and powerful, for both consumer and industry. SOI is scheduled to begin shipping in early 1999. The impact from SOI will be significant. Virtually all semiconductor manufacturers are expected to begin using the technology within the next couple of years. Like other developments, including IBM's copper chip (also a Winning Technology), it's helping to alleviate "the fundamental bottlenecks that prevent designers from building faster, smaller, and more efficient chips," says Davari. In fact, he points out that CMOS could hit a roadblock by 2000 if SOI and other breakthroughs did not enter the picture. "Semiconductor manufacturing would run out of steam and hit its fundamental limit. This allows us to keep Moore's Law alive and well," he explains.

Sponsored Recommendations

Voice your opinion!

To join the conversation, and become an exclusive member of IndustryWeek, create an account today!