HP patents way to wire molecular circuits

In 1999, researchers from HP and the University of California at Los Angeles took a major step in a new lineage of computing called molecular electronics, creating a molecule called rotaxane that could be switched on chemically. In 2000, UCLA researchers led by J. Fraser Stoddart advanced the research, creating a molecule called catenane that could be switched off as well, laying the foundation for molecular-scale computer memory.

The new technology has the potential to extend the current trend in computing that leads to new generations of chips twice as powerful as the preceding generation every 18 months, a march of progress called Moore's Law. However, unless there's a way for molecular electronics to communicate with the outside world, the idea will be merely an academic curiosity.

"If we can't hook these things up, it isn't a very interesting technology," said Phil Keukes, a senior scientist in charge of the work at HP Labs. HP, focused on overcoming the engineering challenges of making molecular electronics feasible for mass production, said it's patented a way to wire molecular electronics to conventional chips. The company hopes the technology will be practical in some niche markets in 2006.

Current computer chips, made by etching circuitry onto silicon crystals, have managed to keep pace with Moore's Law. But chip companies are worried about future progress being hampered by physical limitations and the increasing expense of new factories.

Some researchers, notably those at IBM, believe the future of computing will lie in harnessing the peculiar physical laws of quantum mechanics that dominate the universe at very small scales. HP's technology, though, is not quite so revolutionary.

Molecular electronics aren't just useful for storing information, Keukes said. Small collections of the molecules, properly assembled, can act as the "logic" circuitry of chips that actually process instructions, not just store digital information in the form of ones and zeros. Logic circuits such as "and gates" and "or gates," assembled by the millions, let chips handle everything from multiplying numbers to encrypting messages.

Because molecular electronics behave similarly to current silicon chips, today's programmers won't have to worry about adjusting to some radical new world where the basic rules of computing are rewritten, Keukes said. Instead, the people who'll have to worry are those in charge of building the molecular computing devices themselves.

HP's patent addresses one of the key hurdles: getting something as small as a molecule to communicate with today's computers. The wires that extend from grids of molecular electronics circuitry are only 6 atoms wide--about 2 billionths of a meter. Current silicon wires are vastly larger, a comparatively chubby 130 billionths of a meter on today's newest chips.

But other hurdles remain, Keukes said. Among them are making wires 6 atoms across, getting the molecular electronics to work quickly enough, and understanding the chemistry of the computing devices. One compensation, though: There are many different molecules that can be used to store data and process information, he said.

Communicating with the tiny wires of molecular electronics and the information-processing molecules themselves already is understood. First comes a layer of parallel wires in a north-south orientation. On top of that go the information-processing molecules. Finally, on top of that goes a layer of east-west wires.

A signal sent down one north-south wire and another sent down an east-west wire would intersect at a particular molecule, like giving an address.

But linking the tiny wires to the thicker wires of silicon circuits is another challenge. Keukes said HP solved the problem--just theoretically at this point--through a combination of chemistry and computing.

The key is relying not on carefully assembling a network of wires at such minute scales, but rather on creating random connections between the tiny molecular electronics wires and their larger silicon chip brethren, Keukes said.

Chemists have extensive experience at creating regular patterns such as grids of molecules, but "another thing chemists are good at is making a mess," he said. "We make a purely random set of connections to the big wires from the outside world."

After setting up connections between the two sets of wires--potentially by sprinkling flecks of electrically conducting gold between the two--the device would then have to be tested to figure out which wires connect to which addresses, Keukes said. In other words, instead of carefully planning the arrangement of connections in advance, connections would be made randomly and then mapped out after the fact.

"We haven't built this," Keukes said, but "this patent...shows a basic part of our strategy for using chemistry and computing together," eventually creating a process he hopes will outpace current silicon techniques.

UCLA and HP, with funding from the Defense Advanced Research Projects Agency, hope to have a 16-kilobit molecular electronics memory system working by 2005. A year after that could come the earliest use of the technology in some niche markets, such as those requiring low energy consumption and "non-volatile" memory systems that don't lose their data when switched off.

The molecular electronics technology overall could surpass silicon chips for mainstream use in 2010 or 2011, Keukes said.

"Just about the time that silicon should be getting into deep troubles, we should be just about where silicon could be, but dramatically less expensive than silicon," Keukes said.