Hewlett-Packard Co won another US government grant for research into using chemical reactions to build molecular computer chips, a new method that would overhaul chip manufacturing.
The second-biggest computer maker got a US$12.5 million, four-year grant from the Defense Advanced Research Projects Agency, and the company will add US$13.2 million of its own funds. DARPA already had given Hewlett-Packard US$4 million for two years.
Present manufacturing techniques could become obsolete in 10 years, analysts say, as the chip industry improves production methods and shrinks circuit sizes.
This research may provide an alternative. HP Labs is creating transistors and switches it hopes will make up the chips of the future -- storing more data, using less power and costing less to make.
"This is allowing the jump-start of this technology," said Phil Kuekes, an HP Labs researcher working on the project. "DARPA has decided the whole field of molecular electronics deserves more investment. It's quite promising."
Manufacturers today use a complex series of tools to print designs, etch wires and scrape away extra material from the layers of metals and insulators that chips comprise. A fully equipped factory costs at least US$2 billion to build.
The HP Labs team discovered how to use chemical processes to grow wires on silicon, press the device against a bath of liquid to apply switches and then lay the next set of wires to complete the circuit.
This method, which is years from commercial production use, needn't be as pristine as today's super-clean chip factories.
Present chips can be ruined by a speck of dust during production, yet the new technique doesn't have to be perfect to work. That could cut production costs dramatically.
"Our goal is to make chips so cheaply and easily that any 12-year old with a chemistry set could do it," Stan Williams, who runs the team, has said of the project.
The going was tough early on. The group had to invent many of its tools, such as a vibration-isolated granite table to hold the bath at a precise surface tension. They've got it down now. By January, the circuits were piling up in round white dishes in corners of the Palo Alto, California, lab next to huge microscopes and computer monitors.
They met DARPA's original challenge of crafting a working 16-bit memory device by this month or next. That's comparably complex as chips from the 1970s. Now, they've been given a new mission: to have a 16,000-bit part by June 2005.
To do that, researchers need to improve the wires they're building and the connections between those molecular wires and the standard ones that hook up to real-world devices. With a 16-bit part, the wires could be connected by hand, one by one. With the larger goal come hundreds more wires, and the team is working up new techniques to automate making the connections.
The consortium members include the University of California in Los Angeles and Carnegie Mellon University in Pittsburgh.
Eleven bids for the grant were received, DARPA said.
HP Labs, the birthplace of the pocket scientific calculator and the inkjet technology dominating today's printer market, is one of a few corporate centers that can do this type of basic research. Lucent Technologies Inc's Bell Labs and International Business Machines Corp, the biggest computer maker, are among rivals noted for such projects.
RSS