July 16, 1999, Friday
Business/Financial Desk Tiniest Circuits Hold Prospect Of Explosive Computer Speeds
By JOHN MARKOFF
Plunging deeply into a Lilliputian world that promises ultrafast, low-power computers, a research team has for the first time fashioned simple computing components no thicker than a single molecule.
The achievement, being reported Friday in Science magazine, opens a new window onto a once speculative but now increasingly probable vista of molecular-scale sensors, computers and machines.
The researchers, from the Hewlett-Packard Company and the University of California at Los Angeles, say their work could be a step toward computers 100 billion times as fast as today's most powerful personal computers. And they envision a world in which supercomputing power is so pervasive and inexpensive that it literally becomes an integral part of every man-made object.
Until now, integrated circuits -- the on-off switches that are the basis for all computing -- have been made by etching silicon wafers with beams of light, a process known as photolithography. The ability to make those circuits smaller is ultimately limited by the wavelength of light.
But the team from Hewlett-Packard and U.C.L.A. has found a way to build circuits using chemical processes rather than light, making the switches as small as a molecule. As a result, the researchers believe that they can make components for future computers several orders of magnitude tinier than today's smallest transistors.
Over the next decade, such technology ''holds the promise of vast data storage capability,'' said Phil Kuekes, a physicist and computer designer at Hewlett-Packard, which is based here. And ultimately, he said, it could create a new class of ''Fantastic Voyage''-style machines, like sensors traveling within a person's bloodstream, issuing alerts if health problems are encountered.
The research is also the strongest indication yet that engineers may find new ways to perpetuate the computer industry touchstone known as Moore's Law, which has charted the steady rise in computing power using ever smaller and cheaper chips over the last three decades.
As individual transistors get smaller, they take less electrical power to switch and, in general, can switch on and off more quickly. And more of them can be produced without increasing the cost of production.
One of the next steps by the team at Hewlett-Packard and U.C.L.A. will be to come up with a chemical process to create ultra-thin wires -- no more than several atoms across -- needed to connect all the molecular switches into a complete computer circuit.
In current circuitry such wires are as thin as one-quarter of a micron -- a human hair is about 400 times as thick -- but even so, they are far too thick to connect individually to the new molecular switches.
The molecules themselves, called rotaxanes, are synthetic compounds created by chemists at U.C.L.A. And the switches built from them deliver ''the equivalent capability to a silicon circuit I could have bought in 1970'' that was the size of a fingernail, Mr. Kuekes said.
Computer researchers said that they believed the advance heralded the rapid acceleration of a new field known as moletronics, or molecular electronics. The Pentagon, which was a major underwriter of the work at Hewlett-Packard and U.C.L.A., is interested in such research because of its broad potential in military applications.
''This field is still in its infancy,'' said Mark Reed, a chemist at Yale University, who is doing experimental research in molecular-scale computing. ''But the results are starting to come in faster. Over the last five years we've come from an incredible idea to the point where we might be able to do something.''
The researchers cautioned that what they have created is only the first step toward building molecular computers and that significant barriers remain.
For example, the current device designed by the researchers can only change from one state to another -- essentially, from 1 to 0 -- not switch repeatedly, as it must do to be a practical replacement for silicon-based transistors, rather than just a storage device.
And the molecular switches have so far not exhibited switching speeds faster than today's fastest silicon chips.
But several researchers in the field said they now believed they saw ways to surmount these obstacles to build true molecular-scale computers.
''This is an important stepping stone, but we still have a long way to go,'' said James Tour, a Rice University chemistry professor who is conducting similar molecular scale research. ''I don't want people to think that in three to five years we'll have molecular electronics, but the interesting thing about this work is that we can now see ways to scale past the limits of silicon.''
Molecular computing is rapidly becoming a Holy Grail for many computer scientists because it promises immensely more powerful computers than today's machines, based on the ability to put millions of times the current number of switches in a single machine.
James Heath, a U.C.L.A. chemistry professor who is leading the research with Hewlett-Packard, described the possibility of computers 100 billion times as fast as a Pentium microprocessor. He also suggested that it might some day be possible to replicate the power of 100 computer work stations in a space the size of a grain of salt.
Speculating about whether a new technology like molecular electronics might displace an existing one like silicon is very risky, the researchers said. If the technology does prove workable, however, its first application might be in a specialized form, complementing rather than replacing silicon chips. Similar examples today include gallium arsenide semiconductors, which are used for very-high-speed applications, even though they cost much more than conventional silicon chips.
Indeed, the new era of moletronics is beckoning just as silicon-era technologists are achieving stunning levels of transistor density. For example, computer-memory makers are now beginning production of chips that can store one billion bits on a single chip, 16 times the capacity now typical in personal computers.
But such feats pale compared to the promise of molecular-scale computing.
To illustrate the density possible at the molecular level Mr. Tour said that in 18 milliliters of water -- about one mouthful -- there are enough molecules so that if they were each represented by a sheet of paper the stack would reach from the earth to the sun 400 million times.
''A single molecular computer could conceivably have more transistors than all of the transistors in all of the computers in the world today,'' he said.
Organizations mentioned in this article:
Hewlett-Packard Co; University of California (Los Angeles)
Computers and the Internet; Research; Miniaturization; Supercomputers
You may print or save this article now.
May not be reproduced or
transmitted without permission.
May not be reproduced or transmitted without permission.