Feature

When the Chips are Down

Scientists seek alternatives to a computer technology nearing its limits

By
5:17pm, August 10, 2004
Sponsor Message

Competition to make computer chips smaller, faster, and cheaper has fueled U.S. economic growth, driven a technological revolution, and made your once-flashy personal computer a relic in 2 years' time.

Experts, however, predict this march toward miniaturization will hit a wall by about 2010. That's when transistors as we know them will have shrunk so close to the atomic scale that quantum physics will take over and the old rules of chip design won't hold.

"Every major computer company that wants to stay in business beyond 2012 is doing something in this area," says R. Stanley Williams, a senior scientist at Hewlett-Packard Laboratories in Palo Alto, Calif. "It's kind of a desperate race."

Biology could help speed the electronics industry to the finish line, many researchers say. Nature has been building incredibly tiny, precise, geometric structures ever since life began. Brain cells, for instance, release molecules that fit into their corresponding receptors the way a key fits into a lock. Getting nature to cheaply build a transistor as precisely shaped as a cell's receptor might be one way that 21st-century electronics could vault over the miniaturization wall.

Researchers across disciplines are banding together to attempt the feat. Some are using DNA, as if it were a set of Lego pieces, to make a platform for what could become the smallest computer chips ever. Others are using the building blocks of proteins to assemble the world's tiniest wires. These and other imaginative marriages of biological molecules with electronics could produce a generation of minute devices that look nothing like their 20th-century parents.

Packed with transistors

Today's computer circuits are packed with transistors—the newest Intel Pentium chip has 28 million of them. You can imagine a transistor as a drawbridge and an electron as a car attempting to cross to the other side. Opening or closing the transistor drawbridge controls the flow of electrons through a circuit.

The shorter the distance across each bridge, the more transistors you can squeeze onto a chip, the faster a chip processes data, and the less energy it needs to run. That's why smaller has meant better for decades of electronics manufacturers and consumers.

Today's smallest transistors span a mere 180 nanometers—about 1/500 the width of a human hair. That's getting close to the 25-nm limit at which the laws of quantum physics will allow electrons to leap across transistor drawbridges even when they're open. In other words, the basis of modern computer circuits will soon break down.

It's time to envision something radically different, says Joseph M. Jasinski, the senior manager of computational biology at IBM's research division in Yorktown Heights, N.Y. The question is, What?

The new devices must be on the scale of nanometers and suitable for inexpensive mass production—Mother Nature's forte. The biological world regularly takes nanoscale building blocks of exquisitely precise shapes and sizes and makes them into complex systems. These include the DNA double helix and the cellular machines that build proteins based on the DNA code.

By mimicking biology, researchers hope to design molecules that could act as self-directed construction workers assembling perfect nanoscale circuits, millions or billions at a time. The molecules wouldn't necessarily conduct current, but the crystalline devices they construct would. Like a bricklayer building a house, this so-called bottom-up approach would contrast with the current technique of sculpting microscopic circuit patterns from slabs of silicon. And, researchers say, the new product would make today's fastest circuit components seem unbearably slow.

The right stuff

In their search for just the right stuff to build smaller circuits with, some scientists have zeroed in on DNA. They take advantage of the genetic code—four nucleotide bases labeled A (adenine), C (cytosine), G (guanine), and T (thymine)—to design short DNA strands that self-assemble into what could be the foundation of a circuit board. Since an A in one DNA strand always binds to a T in another, and a C always binds to a G, the scientists can predict how two strands will come together.

Ned Seeman of New York University and his colleagues are trying to commandeer this complementarity for decidedly nonbiological purposes. They're designing DNA strands whose matching sequences zip together into nearly rectangular tiles just 16 nm long and 4 nm wide. The researchers use this method to make different types of DNA building blocks that are identical in shape and size. About 250 trillion of them could fit inside this "o."

The beauty of these structures, according to Seeman, is that they can be combined into a pattern resembling a striped plaza of ceramic tiles. Such a miniature plaza could become the platform of a nanoscale circuit board, according to Seeman. Key to the circuit would be special tiles designed so that complementary DNA strands will stick on and jut into the air like a series of poles.

"We think we can add nanostuff—nanocrystals, nanotubes, and the like—" to the top of each pole, Seeman says. These would be the electrical components of the circuit, he continues, and as long as they're close enough, they should conduct electricity.

Different approach

Veering away from the DNA approach, several scientists have turned to a different molecular pattern maker: peptides, amino acid chains that are shorter than proteins.

Like DNA, peptides don't naturally bind to semiconductors, the crystalline materials that make up computer chips. But through a technique resembling accelerated evolution, materials chemist Angela Belcher of the University of Texas at Austin and physicist Evelyn Hu of the University of California, Santa Barbara are finding peptides that have this unusual property.

Belcher says she became interested in marrying peptides with nanoelectronics during her doctoral research on abalone. "Nature has amazing control over forming materials like shells," she says. Her research group revealed in the mid-1990s that a specific peptide causes calcium carbonate to crystallize into the structure found only in the tough abalone shell.

From that discovery, Belcher and Hu, then her postdoctoral advisor, realized that if they found peptides able to direct the crystal growth of materials such as silicon, they might have a tool for building nanoscale electronics. The problem was that no known peptide interacted with semiconductor materials the way some did with, say, calcium carbonate.

So, Belcher, Hu, and their colleagues grew a random assortment of 1 billion different peptides, each a string of 12 amino acids. They then tested whether any of the peptides stuck to silicon, gallium arsenide, or indium phosphide crystals—three semiconductor materials widely used in the electronics industry.

"There was no guarantee we would've seen anything stick at all," Hu says. To her delight, though, the group found three peptides that not only bound exclusively to one of the crystals in the experiment but also latched onto a particular face of the crystal.

Their research, described in the June 8 Nature, was the first to link peptide self-assemblers with semiconductors, according to chemist Vijay Pande of Stanford University.

Jasinski says, "Angie's work is at the forefront of using biological self-assembly to make nonbiological objects." IBM, Jasinski's employer, is one of the organizations funding her research.

Belcher and her research group are still learning what makes these peptides bind to certain semiconductors and crystal faces. Meanwhile, they continue to discover tricks that certain peptides can do.

For instance, some peptides will prevent a semiconductor crystal from growing in one or two of its three dimensions. This finding is critical because form determines function in both crystals and circuits. Engineers need precise control over the sizes and shapes of circuit components to prevent digital information from getting garbled. And there's even less room for structural imperfection in nanoscale devices than in today's smallest microchips.

"If you really want to assemble a new set of tools—biological molecules that will glom on to all the component materials that go into making a circuit—we need a lot more biological-inorganic pairs," Hu says. Belcher and her group are making progress on that quest. As they've expanded their targets to 20 more semiconductor materials, their stash of crystal-manipulating peptides has ballooned into the hundreds.

They are also designing new peptides that bind to two different crystals at once, acting as a daub of glue. It will take that kind of finesse at the nanoscale to produce self-assembling circuits.

Slow and steady?

As the miniaturization wall looms little more than a decade ahead, computer companies predict that slow and steady won't win the race toward a nanoscale chip. So, some scientists at Hewlett Packard Labs in Palo Alto, Calif., say they're developing a more near-term technique for making molecular electronic devices.

Williams' research team is developing a circuit where molecules acting as switches take on the role that transistors play today (SN: 8/7/99, p. 95). Their circuit will look like a sandwich with molecular switches as the filling.

Two gratings of silicon wires eight atoms wide and 9 nm apart will be the bread of the circuit sandwich, holding specially designed switch molecules in-between. An electric jolt can snap each molecule into one of two shapes.

The upper and lower wires will run in perpendicular directions, creating a grid. In principle, each spot where the wires trap a molecule will be a bit of memory and also a switch, according to Phil Kuekes, a computer architect at Hewlett Packard.

"By taking that regular structure and turning switches on and off, you can program [the circuit] to be as complex as you want it to be," Williams says.

The company's current prototype has micron-size wires—about 1,000 times wider than they'll ultimately need to be—and can do only the simplest of computations, according to Williams. But he says that he and his coworkers have already made silicon wires just 3 nm wide, and by the end of December, his group aims to complete a 16-bit memory chip that fits on a 100-nm square.

Williams adds, "The laws of physics indicate it may be possible to build computing devices that are 1 billion times more efficient than those we have now."

Many questions

Whether nanoscale computer chips have a foundation of DNA, are built by peptides, or run on molecular switches, scientists still have many questions to answer.

How will they hook these tiny components to one another and the rest of a computer? How will they program the devices in sophisticated ways? And will the new chips really turn out to be any better than the best ones engineers can make with existing techniques?

If anything's certain, scientists will keep churning out promising nanoscale ideas and results, and it will take years for the research to mature into a 21st-century computer chip.

More from Science News