The Chemistry of Computing


Computers Made of Molecule-Size Parts Could Build Themselves
by Tim Appenzeller

Source: U.S.News & World Report
U.S.News & World Report Inc. All rights reserved

May 1, 2000

When James Tour thinks about computers, the Rice University scientist pictures big, big numbers-but minuscule machines. That's because he's not a computer designer in the traditional sense at all. He's a chemist, and his notion of computer parts makes today's technology seem as unwieldy as the radio tubes of days past. Indeed, if Tour has his way, the silicon components of today's machines will ultimately go the way of those electronic dinosaurs.

"Consider this," he writes. "There are [a billion trillion] water molecules in just one drop of water." That's hundreds of times the sum total of all the transistors in every computer ever made, he reckons. If just a fraction of the molecules in a speck of matter could be made to act as electronic switches, able to control the electrical currents that are the basic language of computers, present-day computers with their paltry billions of transistors could quickly become obsolete. So would to-day's giant and expensive chip fabrication plants: A molecular computer might build itself in a set of chemical steps.

If this sounds like rank speculation, it's not. Those molecular switches are real. Over the past year, Tour and other researchers have made and tested complex, artful molecules that work like switches you might buy at RadioShack, flipping from "off" to "on" and back when tripped by pulses of voltage. Other groups have made wires that are only atoms across, equipped with the mo-lecular connectors that might soon hook up the tiny switches into computer circuits.

A droplet-size supercomputer will become a reality only after scientists learn how to connect astronomical numbers of molecules into a working computer architecture and link the minuscule computer to the larger-scale world we live in. That could take decades. But simpler devices that use dense patches of molecular switches to soup up conventional chips are already in the works.

Computer switches need to do two jobs: open or close in a flash to process information, and stay open or closed for long enough to act as short-term memory. Last November, Tour, Mark Reed of Yale University, and their colleagues reported that they had met the first requirement by tailoring an electrically conductive molecule. When they positioned a few hundred of the molecules between tiny metal electrodes and increased the voltage across them, they found that at about 2 volts, the molecular switches turned on, allowing current to flow. Above or below that voltage, the switches were off. Now they have modified their switch molecule to create memory: After being flipped on or off, it stays that way for a few minutes. "You can read, write, erase it," Tour says.

At the University of California-Los Angeles, Fraser Stoddart, James Heath, and their colleagues have built similar switches from molecules that actually change shape when they are zapped with a charge. "Think of a dumbbell with a ring that slides back and forth" along a shaft, turning the switch on or off, says Stoddart.

The wiring problem. Connecting these tiny switches with wires as coarse as those on an ordinary microchip would squander much of the molecules' size advantage. So the UCLA group's collaborators at Hewlett-Packard are learning how to make metal or silicon wires just 10 or so atoms across. As a bonus, these wires don't need to be laid down one by one, like the connectors on current silicon chips; instead, they assemble themselves en masse, like long, thin crystals. At Pennsylvania State University, another group is making tiny wires that are predesigned to snap together into larger structures. The researchers coax platinum and gold atoms into the narrow pores in a membrane-then dissolve away the membrane to free breathtakingly thin wires that are striped with bands of the different metals, like a roll of LifeSavers. Molecular switches should stick readily to some of the bands but not others, as do other molecules, including DNA, that serve as glue for linking the wires to one another.

Stanley Williams of Hewlett-Packard says that putting together the first molecule-based devices should take 15 months. By then, he says, "we hope to be able to deliver an entire 16-bit memory unit that could fit on top of the smallest wire in today's integrated circuits, with room to spare." Other groups may also be unveiling prototype molecular processors and memories in two or three years, and one device, from a company founded by Reed, Tour, and others, may even be taking its first steps to market.

Back To Nanotechnology