Imagine computers orders of magnitude more powerful and far cheaper than today's machines. That's one promise of a field that uses individual molecules as microscopic switches.
By David Rotman
For Mark Reed, the future of molecular electronics has just arrived. A self-described "device guy," Reed, who heads Yale University's electrical engineering department, prides himself on having a distinctly practical bent. Ask him about the possibility of one day using molecules to replace silicon in computers that are billions of times faster than today's PCs or that fit on the head of a pin, and he grimaces. "I don't know how to do that. I don't think anyone does," he says dismissively.
But that doesn't dim the excitement that Reed, a leading researcher in molecular electronics, is feeling. Using molecules synthesized by Rice University chemist James Tour, Reed has fabricated electronic memories and a simple logic element made up of molecules that function as tiny, individual switches. The devices, which rely on small organic molecules tailored by the Rice chemists to have just the right electronic properties, are crude laboratory experiments. But they work-the molecules acting as a component in ultrasmall electronic devices able to turn current "on" and "off." What's more, these early prototypes have already shown hints of performing memory and logic tricks not possible with silicon semiconductors.
Most impressive, says Reed, is that the molecular devices are astonishingly easy-and potentially cheap-to make. You simply dip a silicon wafer lined with metal electrodes into a beaker filled with the right chemicals and give the molecules a few minutes to form on the electrodes. If you're clever enough with the chemistry, it's possible to coax the molecules to spontaneously orient themselves on the electrodes. "It works beautifully-and it works every time," says Reed.
It may work every time, but there's considerable controversy about what these chemical reactions will ever amount to. While true believers envision a world in which microscopic molecular computers made at low cost perform remarkable calculations, skeptics think the field has lost sight of the real world of engineering limits. Meanwhile, "device guys" like Reed think the future-in the form of workable prototypes that can be integrated with conventional silicon technology-is now.
The core advantage of molecular computing is the potential to pack vastly more circuitry onto a microchip than silicon will ever be capable of-and to do it cheaply. Semiconductor-makers can now cram about 28 million transistors on a chip by shrinking the smallest features of the transistors down to about 180 nanometers (billionths of a meter). Using conventional chip-making methods, however, the smaller you make a feature, the more expensive and difficult the process becomes. Many semiconductor experts doubt commercial fabrication methods can economically make silicon transistors much smaller than 100 nanometers. And even if chip-makers could figure out a reasonable way to etch them onto a chip, ultrasmall silicon components probably wouldn't work: At transistor dimensions of around 50 nanometers, the electrons begin to obey odd quantum laws, wandering where they're not supposed to be.
Molecules, on the other hand, are only a few nanometers in size, making possible chips containing billions-even trillions-of switches and components. In initial experiments, scientists have sandwiched a large number of molecules between metal electrodes. The devices work, however, because each molecule operates as a switch. If it were possible to wire a small number of molecules together as individual electronic components to form circuits, the result would change everything in computer design. Molecular memories could have a million times the storage density of today's best semiconductor chips, making it possible to store the experiences of a lifetime in a gadget the size of a wristwatch. Supercomputers could be small enough and cheap enough to incorporate into clothing. Worries that computing technology will soon hit a wall would disappear.
Those applications are decades off-if they ever materialize. Still, Reed argues, some uses for molecular electronics could soon be feasible. Ultrasmall, cheap molecular devices could sit side-by-side with silicon, reducing the number of transistors and the power required by the circuit. "This is something you could use today, something you could sell in Radio Shack," says Reed. "This has a chance to totally change the economics of silicon."
To make that a reality, Reed, Tour and chemists from Pennsylvania State University have co-founded a startup called Molecular Electronics. The group declines to say what the initial products will be, but Tour says having "a working system in a couple years doesn't seem unrealistic."
Until very recently, that prediction would have seemed far-fetched. But in the last year, the field has taken a leap from theory to the realm of the practical. Like their competitors at Yale and Rice, a West Coast collaboration of chemists and computer scientists from Hewlett-Packard and the University of California, Los Angeles, have recently characterized molecules capable of acting as electronic switches and memory (see past issue: "Computing After Silicon," TR September/October 1999). R. Stanley Williams, who heads the effort at HP, says his team expects to build a prototype of a logic circuit that integrates a small number of nanoscale molecular devices within 18 months. "We have the switches and wires-the components to actually make true nanocircuitry," says Williams.
In theory, at least, assembling a molecular electronic device is straightforward. In the version of the recipe favored by the HP/UCLA collaboration, the scientists first make a single monolayer of the right organic molecules in a chemical apparatus called a Langmuir trough; they then dip a silicon substrate covered by a pattern of metal electrodes into the trough. If the chemistry is just right, the molecules will bind to the metal electrodes, neatly orienting themselves. A second set of electrodes is then deposited on the molecules; the result is a monolayer of the organic molecules sandwiched between metal electrodes.
The challenge is that most organic molecules are not electrical conductors at all-never mind having the electronic properties that let them work as an effective switch. What is needed to make the system function electronically are specially tailored molecules that turn on and off repeatedly in a reliable and detectable way (the properties that have made silicon so successful). Coming up with molecules able to do the trick is the domain of chemistry wizards like Rice's Tour and UCLA's James Heath and Fraser Stoddart.
Their wizardry began paying off in a big way last fall. First, the HP/UCLA group published a paper describing what is in effect a molecular fuse-a one-time switch based on a complex, dumbbell-shaped organic molecule called rotaxane; the scientists have subsequently made reversible switches. They also showed how the device could perform simple logic and memory functions. Within months the Yale/Rice collaboration rivaled that feat by describing the synthesis of other organic molecules that act as electronic devices.
Despite the differences in molecular particulars, the two research groups are taking advantage of the same quantum effects that could eventually set fundamental limits on silicon semiconductors. The molecules separating the two electrodes would normally block the flow of current. In the nanoworld of individual molecules, however, electrons can "tunnel" through a barrier that, according to classical physics, should block their path. By manipulating a voltage placed across the electrodes, the scientists can adjust the tunneling rate and thus turn the current on or off.
Reed has already started thinking of ways to use molecular devices in combination with conventional silicon. One type of quantum logic gate that Reed has recently built would, for example, do the same specialized function as seven much larger silicon transistors, significantly reducing the size and power consumption of an integrated circuit. And while fabricating conventional transistors requires complex and expensive processing, the molecular device can be "glued on" to the circuit, says Reed.
Molecules could also provide ultra-cheap electronic memory with some attractive properties. The most common type of semiconductor memory is called DRAM, for dynamic random access memory. (This is the short-term memory your computer relies on when it's running a program.) The problem with DRAM is that the stored information evaporates when the power is shut off-it's "volatile." That's the reason you have to boot up Windows every time you turn your computer on, moving the program from your hard drive to the DRAM chips. But an experimental molecular device Reed made last fall holds data for more than 10 minutes after the power is shut off. "Suppose we can get that up to several years," says Reed. "It would essentially be nonvolatile memory. Imagine how many times you wouldn't have to boot up Windows."
Although these early applications are worlds away from the billion-transistor molecular computers that enthusiasts envision, they could show the value of organic molecules as an electronic material. "They are a camel's nose under the tent," says Reed, adding that "these hybrid devices are already very realistic. They're the first step down the road to more complex [molecular] circuits."
It's likely, however, to be a long road. Even a simple computer made of molecular components is at least a decade away-and then "only if we get really clever," acknowledges Williams. But the HP chemist says his group is already on its way. In their initial prototypes, the California researchers have fabricated the top and bottom metal wires as perpendicular grids, creating a "crossbar" structure with the molecules sitting at the junctions of the wires. So far, the group has made devices with metal contacts that are thousands of nanometers in diameter; there are millions of molecules at each junction. But Williams says that by later this year the group expects to have wires measuring a few nanometers across. "It didn't make sense to do everything hard right away. So we used much larger wires. Now we're doing the experiments to switch to smaller wires and make the measurements."
The nearly perfect candidates for such tiny wires are structures known as carbon nanotubes. These regularly shaped pipes, only a few nanometers in diameter, could be excellent conduits for electrons speeding through a molecular circuit. The problem is that nanotubes tend to form as a tangled mess-far from the neatly ordered arrays needed to fabricate complex circuits. Building any structures with nanotubes "is now an art form," says physicist Paul McEuen of the University of California, Berkeley. "We basically throw them down on the ground and look for [the structure] we want."
The HP/UCLA group is confident they'll solve the wiring problem. "Eventually nanotubes will be used. Their electronic and physical properties are so desirable," says Williams. For now, he says, the group is also working on silicon nanowires. And, promises Williams, with or without carbon nanotubes, by late summer the scientists will scale down the junctions of devices to smaller than 10 nanometers. The near-term targets are a 16-bit memory that is 100 nanometers on a side, and soon after that a similarly sized logic device. These rudimentary circuits may not threaten the reign of silicon, but they could be a milestone in helping prove that molecular electronics is feasible.
But then comes the truly daunting part: turning these simple devices into complex logic circuits, and integrating them into an actual computer. One of the penalties you pay for making microelectronics based on chemistry is that, unlike silicon chips made in high-tech fabrication plants, molecular devices synthesized in vats of chemicals will inherently be full of defects. At the scale of individual molecules, chemistry is given to statistical fluctuations-sometimes it works and sometimes it doesn't. But it's here that the HP/UCLA scientists contend they have made their most important breakthrough.
Their answer: software that overcomes the defects. Several years ago, computer scientists at HP built a supercomputer called Teramac, using defective silicon chips so flawed they were considered worthless. The HP scientists cobbled these rejected chips into a computer by developing a "crossbar" architecture that makes it possible to connect any input with any output. Once the hardware was built, the computer was programmed to identify and route around any defects. The system worked-and its massive parallelism provided an archetype that the California scientists plan to use for their molecular computer.
"A chemist working on a computer is a bizarre thing. You can't go to a chemist and ask them to build a computer," says Heath, one of the UCLA scientists who is helping to synthesize the necessary components. But, he says, the Teramac architecture has provided the HP/UCLA group a clearly defined target. "The software will turn it into a machine," says Heath. That molecular computer "may be a long way off," he acknowledges. "But there's no reason why it won't work."
The World Between
While folks like Heath are sanguine, the technology has its share of doubters. The field of molecular electronics "is in love with itself," says Rick Lytel, a computer scientist at Sun Microsystems. Despite his skepticism, however, Lytel is keeping a sharp eye on the field for Sun and is developing specifications to test and evaluate prototype molecular devices. He believes molecular electronics could eventually find uses as memory devices. But Lytel says many of his colleagues in the field have deluded themselves into thinking that they are "only a step away from the marketplace."
Even believers in the prospects of molecular electronics disagree with one another over the role the technology will play in computation and electronics. Take Mark Ratner, a chemist at Northwestern University who is generally regarded as one of the grandfathers of the field. Ratner doubts molecules will ever compete directly with silicon in complex computational tasks. "You want to use molecules for what they do best" and to compensate for where silicon falls short, says Ratner. In particular, he points to their ability to recognize and respond to other molecules. By combining those functions with the newly developed electronic properties, you might make tiny sensors and actuators that detect and react intelligently to biological and chemical clues. It might, says Ratner, make possible implantable biochips incorporating sensors and actuators made out of molecular electronics that sense the needs of the body and respond by discharging an appropriate dose of medication.
For this pioneer of molecular electronics, the true potential of the field could be realized in bringing the world of microelectronics together with the world of biology and molecules. Molecular electronics, suggests Ratner, could be the piece of the puzzle that finally helps to bridge the material gap between biology and computing.
David Rotman is a Senior Editor at Technology Review