Monday, Jun. 19, 2000

What Will Replace Silicon?

By Michio Kaku

The economic destiny and prosperity of entire nations may rest on one question: Can silicon-based computer technology sustain Moore's law beyond 2020? Moore's law (see sidebar) is the engine pulling a trillion-dollar industry. It's the reason kids assume that it's their birthright to get a video-game system each Christmas that's almost twice as powerful as the one they got last Christmas. It's the reason you can receive (and later throw away) a musical birthday card that contains more processing power than the combined computers of the Allied Forces in World War II.

The secret behind Moore's law is that chipmakers double every 18 months or so the number of transistors that can be crammed onto a silicon wafer the size of a fingernail. They do this by etching microscopic grooves onto crystalline silicon with beams of ultraviolet radiation. A typical wire in a Pentium chip is now 1/500 the width of a human hair; the insulating layer is only 25 atoms thick.

But the laws of physics suggest that this doubling cannot be sustained forever. Eventually transistors will become so tiny that their silicon components will approach the size of molecules. At these incredibly tiny distances, the bizarre rules of quantum mechanics take over, permitting electrons to jump from one place to another without passing through the space between. Like water from a leaky fire hose, electrons will spurt across atom-size wires and insulators, causing fatal short circuits.

Of course, cyber Cassandras have been tolling the bell for Moore's law for decades. As physicist Carver Mead puts it, "The Chicken Little sky-is-falling articles are a recurring theme." But even Mead admits that by 2014 the laws of physics may have their final revenge. Transistor components are fast approaching the dreaded point-one limit--when the width of transistor components reaches .1 microns and their insulating layers are only a few atoms thick. Last year Intel engineer Paul Packan publicly sounded the alarm in Science magazine, warning that Moore's law could collapse. He wrote, "There are currently no known solutions to these problems."

The key word is known. The search for a successor to silicon has become a kind of crusade; it is the Holy Grail of computation. Among physicists, the race to create the Silicon Valley for the next century has already begun. Some of the theoretical options being explored:

--THE OPTICAL COMPUTER This computer replaces electricity with laser light beams. Unlike wires, light beams can pass through one another, making possible three-dimensional microprocessors. An optical transistor has already been invented; unfortunately, the components are still rather large and clumsy. The optical counterpart of a desktop computer would be the size of a car.

--THE DNA COMPUTER One of the most ingenious ideas being pursued is to compute using DNA, treating the double-stranded molecule as a kind of biological computer tape (except that instead of encoding 0s and 1s in binary, it uses the four nucleic acids, represented by A, T, C, G). This approach holds much promise for crunching big numbers. Hence large banks and institutions may one day use it. However, a DNA computer is an unwieldy contraption, consisting of a jungle of tubes of organic liquid, and is unlikely to replace a laptop in the near future.

--MOLECULAR AND DOT COMPUTERS Other exotic designs include the molecular computer and the quantum dot computer (which replace the silicon transistor with a single molecule and a single electron, respectively). But these approaches face formidable technical problems, such as mass-producing atomic wires and insulators. No viable prototypes yet exist.

--THE QUANTUM COMPUTER The darkest horse to emerge in this race is the quantum computer, sometimes dubbed the ultimate computer. The idea is to direct a laser or radio beam on a carefully arranged collection of atomic nuclei, each of which is spinning like a top. As the beam bounces off the atoms, it flips the spins of some of them. Complex computations can be performed by analyzing how the spins have been flipped.

U.S. Intelligence Agencies are nervously eyeing these new designs. Quantum computers, in particular, could be so powerful that they might one day break the most intricate secret codes the CIA can concoct. Not that a quantum supercomputer is going to leap out of some laboratory and paralyze the CIA anytime soon. These computers seem to be exquisitely sensitive. The tiniest disturbance--even a passing cosmic ray--can change the orientation of their computational atoms, spoiling the calculation. At present, quantum computers can perform only trivial calculations on perhaps five atoms. To do any useful work, they would need to calculate on millions of atoms.

Clearly, none of these designs are ready for prime time. Most are still on the drawing board, and even those with working prototypes are too crude to rival the convenience and efficiency of silicon.

There may be a silver lining to all this. If Moore's law somehow continues unabated, then by some estimates our computers by 2050 will be calculating well beyond 500 trillion bytes per sec., at which point, as Ray Kurzweil suggests (see "Will My PC Be Smarter Than I Am?"), they will be considerably smarter than we are. Evolution says organisms are replaced by species of superior adaptability. When our robots tire of taking orders, they may, if we're lucky, show more compassion to us than we've shown the species we have pushed into oblivion. Perhaps they will put us into zoos, throw peanuts at us and make us dance inside our cages.

Maybe the collapse of Moore's law isn't such a bad thing after all. If none of these exotic designs pan out, our computers won't automatically increase in power every Christmas. But perhaps that's a small price to pay for our freedom.

Michio Kaku is a physics professor at City College of New York and author of Visions: How Science Will Revolutionize the 21st Century