[Paleopsych] Plasmonic computer chips move closer

[Steve Hovland]

Computer chips capable of speeding data around by rippling the electrons on 
the surface of metal wires just got a step closer, researchers say.
Mark Brongersma, at Stanford University in Palo Alto, California, US has 
found a new way to model the three-dimensional propagation of these ripples 
- called plasmons - in two dimensions. He says the new model is much 
simpler and more intuitive than existing simulations and will be crucial in 
the design of plasmonic components for computer chips.
Plasmons travel at the speed of light and are created when light hits a 
metal at a particular angle, causing waves to propagate through electrons 
near the surface.
"Right now, the simulations are so complex that only a few groups in the 
world can carry them out," says Brongersma. "The new model came out of a 
desire for much simpler models."
Currently the biggest application for plasmons is in gold-coated glass 
biosensors, which detect when particular proteins or DNA are present - the 
bio-matter changes the angle at which light hitting the surface produces 
the most intense plasmons.
But scientists would love to use plasmons to ferry data around computer 
chips because they could operate at frequencies 100,000 times faster than 
today's Pentium chips, without requiring thicker wiring.
Light speed
Ordinary light waves can transmit data at similarly high frequencies, but 
using photons to carry data across a computer chip is currently impossible. 
This is because the size of the optical fibre that carries the light waves 
must be about half the wavelength of the light, which is over twice the 
thickness of the wires in modern chips.
"The big advantage of plasmons is that you can make the devices the same 
size as electrical components but give them the speed of photons," says 
Brongersma. Plasmon-carrying wires could also be made out of copper or 
aluminium, like the interconnects on today's computer chips.
Brongersma points out that the speed of these interconnects has become a 
key limiting factor. While transistors have become faster at switching as 
manufacturers find ways to make them smaller and smaller, the wires that 
carry the data are not getting any faster. "We need to find new ways to 
connect transistors together," he says.
Design issues
To develop the new, simpler model, Brongersma showed that the intensity 
pattern of a plasmon travelling across the surface of a metal strip was the 
same as for a light wave travelling through an optical fibre. He says this 
indicates that traditional "ray-tracer" programs for modelling light waves 
should work for plasmons too.
Such models are necessary if devices that generate and route multiple 
plasmons are ever to be designed, Brongersma says. He will publish the work 
in an upcoming issue of Optics Letters.
"Any advance that aids design is a good thing," says Harry Atwater of the 
California Institute of Technology in Pasadena, US. But he warns that a 
bigger hurdle to plasmon-based computer chips is finding plasmon sources 
that are compatible with silicon. "The most important element is not the 
design tools, it is having the ingenuity to know what to do with them," he 
says.