Spinning electrons will speed up chips

A brand new physical effect may prove invaluable to chip makers seeking to speed up their designs, if a £555,000 project announced on Thursday is successful.

Led by the University of Bath, the international three-year effort will investigate the use of microwaves on chip to replace electrical wiring. Instead of relying on complex transmitters and receivers, though, the radio signals will be generated by a planar spin oscillator — a single transistor-like device that uses the breakthrough discovery.

"It's like the signals generated in an NMR brain scan," Dr Alain Nogaret of the University of Bath told ZDNet UK, "but created by a device that can be under 100 nanometres big and integrated with the rest of a silicon circuit."

Nogaret announced this way of producing the effect, called inverse electron spin resonance, last year in the scientific journal Physical Review Letters. "The purpose of the project is to build the device and show it sending and receiving radio signals. We see it being used to link clusters of components on chip through waveguides", he said. "We'll also be in competition with optical interconnects, but it'll be easier to integrate our device with existing circuits than to use optics."

The planar spin oscillator is very low power — each device will produce a nanowatt — but banks of them will produce power that rises exponentially with the number of components. One of the most unusual and useful attributes is that they can be tuned over an exceptionally wide band, from zero to 500GHz, by varying a voltage, opening them up to extremely efficient modulation techniques and high bandwidth transmissions. This should be much more efficient, faster and more flexible than using electrical conductors, freeing chip designers to make swifter and more complicated circuits.

Within the device, electrons are confined to a flat layer and then moved through a carefully designed magnetic field. This causes them to oscillate, which in turn generates the radio signal. Nogaret said that although research would use gallium arsenide semiconductors, the idea would be applicable to standard silicon production processes.

"The principle is so simple, it's quite beautiful and very attractive," Nogaret said. "Although this is the first stage of demonstrating new physics at work, we will be interacting with commercial organisations after the project." He predicted that devices using this may appear five to ten years after the project, and that they could run between 200 and 500 times faster than existing computers.