End of Moore's Law prompts technology search
With the end of Moore's Law looming as a possibility, the search for something to replace today's workhorse CMOS-based silicon is intensifying, researchers told the AVS International Symposium & Exhibition in San Francisco last November. But thus far, no consensus or cost-effective device technology has emerged for the dreaded post-CMOS era less than a decade away, the researchers warned.
That era—when the bulk silicon in today's chip-making fabs no longer scales—could hit by 2015 or sooner, they hinted. By that time, fabs would need a replacement technology, such as carbon nanotubes, nanowires, molecular electronics, quantum computing, three-dimensional transistor designs and spintronics. And the semiconductor industry could see other sweeping and costly changes, including the frightening prospect of 675mm-wafer fabs appearing around 2021.
But the 675mm fab is not seen as the most critical element of the post-CMOS era. More pressing for the industry is narrowing the choices and finding the most feasible next-generation device technology, said C. Michael Garner, manager of the External Materials Research Group at Intel Corp.
"Moore's Law will continue," Garner said, "but the industry needs to define a device technology to replace CMOS-based bulk silicon before 2010 or 2011."
That's easier said than done. Chipmakers, universities and government-backed consortia have already spent millions to research next-generation devices, all of which pose trade-offs and involve manufacturing complexities.
There is still no consensus on where the bulk of the funding should be spent. "It's wide open now," said Paul Nealey, Smith-Bascom professor of chemical and biological engineering for the Nanoscale Science and Engineering Center at the University of Wisconsin. The university is one of several entities looking into pattern-assisted self-assembly, which could be the eventual replacement for optical—and perhaps extreme-ultraviolet—lithography in the next decade.
Leading-edge chipmakers are investigating a breathtaking range of novel and exotic technologies for 2013 and beyond. Intel, for example, is researching trigate transistors, carbon nanotubes, silicon nanowires, III-V-based chips, spintronics, phase change logic devices, interference devices and optical switches.
Among the new technologies, carbon nanotubes—graphene sheets, rolled into cylinders that determine the electronic properties of the device—are generating the most buzz.
Nantero Inc., a startup developing a novel non-volatile memory, said it has become the first company to make carbon nanotubes in production fabs. Its NRAM, a non-volatile RAM, is based on a carbon nanotube matrix structure laid across an etched trench. The company announced it had fabricated and tested a 22nm memory switch in the spring.
At AVS, Intel hinted that carbon nanotubes hold the most promise for logic applications in the post-CMOS era. The chip giant said it has manufactured a 20GHz to 50GHz carbon nanotube device at line width geometries from 2nm to 5nm.
But Intel doesn't expect carbon nanotubes to be used for logic for another decade. The problem is controlling the material properties of the device, said Mike Mayberry, director of components research and VP of Intel's technology manufacturing group. "It can be done on a lab scale," Mayberry said, "but we don't know how to put millions of them on a wafer. It really looks like a daunting challenge."
|Nano-information's four layers split into bio-inspired and quantum categories.|
Others said the industry should move in another direction. "Spin looks to be the most promising," said Jeffrey Welser, director of the Nanoelectronics Research Initiative, which seeks to demonstrate novel devices with critical dimensions below 10nm. By "spin," Welser meant an experimental technology called spintronics, or spin electronics, which refers to the role of electron spin in data storage.
Still to be seen is how to manufacture these next-generation devices. If or when optical lithography runs out of gas, Wisconsin University has proposed an exotic technology: self-assembly.
In January 2006, the university discovered that materials known as block copolymers will spontaneously assemble into intricate 3D shapes when deposited onto particular two-dimensional surface patterns created with photolithography. Lab demonstrations show a strategy for building complex, 3D nanostructures by using standard semiconductor lithography tools.
Another uncertainty is next-generation fabs. The chip industry is in the era of 300mm fabs. The 450mm fab era is projected to emerge in 2012 or so, according to the International Technology Roadmap for Semiconductors.
Some believe 450mm fabs will never get built, in part because of the astronomical costs and lack of equipment for the technology. Intel, however, is expected to move forward with 450mm fabs, which could cost $5 billion or more.
What's next? If one extrapolates beyond 450mm, the next fabs could produce giant 675mm wafers, starting in the 2021 time frame, Intel's Garner said.
For now, the 675mm wafer size is hypothetical, Garner said. "It represents what we would do next, but it's not on the road map."
- Mark LaPedus