Chip industry sets a plan for life after silicon

Nanotechnology is officially on the road map.

A handful of futuristic chipmaking technologies at the atomic scale have been added to an industry planning effort that charts the future of the semiconductor manufacturing industry every two years.

The transition to a post-silicon era is forecast in a report called the International Technology Roadmap for Semiconductors, to be issued Saturday. The report, which is produced cooperatively by semiconductor industry associations from Europe, Japan, Korea, Taiwan and the United States, is used by the semiconductor industry as a planning tool to determine how best to spend research and development money for new technology.

The shift away from conventional silicon transistors has become an important part of the industry's thinking, though the use of nanotechnology is not expected to replace current chipmaking processes for another decade.

The urgency in moving to molecular electronics is propelled in part by a recognition that conventional technologies, despite significant advances, will not be able to sustain indefinitely the chip industry dictum, known as Moore's Law, that projects a doubling of computing power roughly every two years.

In recent years, the semiconductor industry has repeatedly found ways to make conventional transistors ever smaller, making it possible to place more transistors on a single chip for increased computing power and capacity. Currently the smallest of modern transistors are no more than a handful of molecules across; the industry view is that it can continue to shrink conventional transistors for only the next decade. But even those minuscule transistors are bigger than the new class of nanoelectronics, composed of components as small as individual molecules. Researchers are experimenting with a variety of new materials beyond silicon, including organic molecules and carbon nanotubes.

What has changed in the industry's road map is the growing confidence in new technologies that make electronic switches from single molecules or even single electrons.

The development of nanoswitches has reached a point where it will be possible to manufacture them reliably at low cost, according to several researchers who have been involved in the preparation of the report. The New York Times obtained a draft of a report chapter titled "Emerging Research Devices."

The transition to new nanotechnology techniques could occur around 2015, when chipmakers will have exhausted their ability to shrink the wires and switches that make up the modern processors and memory storage devices at the heart of the computer, communications and consumer electronics industries.

The industry planning effort, which was concluded this month in Seoul, South Korea, underscores the work of a small but growing group of chemists, physicists and electrical engineers who are striving to build molecular electronics, a realm once considered science fiction.

"In between 2003 and 2005 there has been a tipping point," said Philip J. Kuekes, a physics researcher in the quantum structures research initiative department at Hewlett-Packard Laboratories in Palo Alto, Calif. "All of the buzz is about nanotechnology."

As conventional transistors become no larger than a handful of molecules, strange behavior in the quantum realm comes into play, making it impossible to determine accurately the on or off states of the transistor.

Nanoscale switches are made to be immune to such quantum effects.

"The physics of silicon can carry us only so far," Kuekes said. To replace conventional transistors, the HP laboratory is concentrating on a new class of molecular scale switches that will continue to represent ones and zeroes reliably.

"Our devices only work because of quantum effects," Kuekes said.

Looking ahead at Intel
Paolo A. Gargini, director of technology strategy for Intel, the world's largest chipmaker, echoed the eventual necessity for a transition beyond silicon.

Intel, based in Santa Clara, Calif., is now preparing to make the shift from chips made using a process where the smallest dimensions are 65 nanometers (one nanometer is a billionth of a meter) to 10 nanometers or less. Today's microprocessors already have more than 1 billion transistors. But it is almost certain that new types of switches and new materials will be needed to build chips that have 1,000 times the capacity of current chips, Gargini said.

The goal over the next decade, he added, is to build chips that can hold more than 1 trillion switches. Intel's new chips will be used first in low-cost laptop computers and in home media devices, further evidence that the semiconductor industry is driven by consumer electronics. Those low-cost products with their vast markets are now pushing technology forward rather than supercomputers and other highly specialized machines.

"The main message of the report is that we are broadening the horizon," Gargini said. "If you considered the incubation time for this research being 10 or 15 years, now is the time to pursue these new technologies."

One promising area he cited is an alternative technology known as a spin transistor, which was first developed during the 1990s. Based on the ability of electrons to exhibit one of two states--orientations described as up or down--spin transistors are switches whose state can be detected and altered without applying an electrical charge.

Spin transistors can be far smaller than conventional silicon transistors and are nonvolatile, meaning that they can store information even if power is switched off.

A second approach, called crossbar latch technology, is discussed in the industry report and is being pursued by the HP quantum researchers.

That technology is based on the use of an organic molecule capable of being turned on and off, which could enable researchers to reach the goal of a trillion switches on a chip. It is projected to have a switching speed of 1 trillion times a second, far faster than the three to 4 billion times a second typical of today's fastest microprocessors.