Saturday, January 7, 2012

Physics & Math

Thinnest silicon-chip wires refuse to go quantum

Saving Moore's law? <i>(Image: Bent Weber)</i>
Saving Moore's law? (Image: Bent Weber)

Not everything is weird at the nanoscale. Wires so small you'd expect them to obey the strange laws of quantum mechanics have instead displayed the same electrical properties as ordinary electrical interconnects.
The finding bodes well for conventional computers, because these tiny, conductive wires could make chips smaller. It could be bad news, though, for the super-fast quantum computers that are hoped to come next.
So far, conventional computers have followed Moore's law: the density of transistors that a conventional integrated-circuit chip can hold doubles approximately every two years, yielding ever-better performance out of ever-smaller devices.
However, it's getting harder to build smaller interconnects to wire up the devices on the silicon chip. As the width of metal wires drops to few tens of nanometres, their resistivity soars because electrons start interacting with nearby surfaces, dissipating more heat and lowering efficiency.

Phosphorous infusion

Also, as wires get down to nanometre scales, quantum behaviour usually dominates. For instance, the entire wire can exist in a superposition of states because of a property called quantum coherence. The wave behaviour of electrons in the wire might then cause them to interfere with each other, disrupting the electrical properties you would expect to see at larger scales.
Now, Michelle Simmons of the University of New South Wales in Sydney, Australia, and colleagues have etched channels in a silicon chip just 1.5 nanometres wide that behave just like larger wires.
The trick was to infuse them with phosphorus atoms, which provide electrons that can move freely and conduct electricity, turning each channel into a wire. Because the entire wire, except for its ends, was enclosed in the silicon, it was isolated from other surfaces that could disrupt its conductivity.

Coolly classical

The team found that these wires conducted electricity nearly as well as state-of-the-art copper interconnects used in modern microprocessors – despite being much thinner. Moreover, when they built wires of different lengths, the wires followed Ohm's law, in which the resistance of a wire increases with length – a property of non-quantum, or "classical" conductors.
The lack of quantum behaviour surprises David Ferry of Arizona State University in Tempe – especially because the experiments were carried out at a mere 4.2 kelvin. "Usually when you go to [such] low temperatures, you expect quantum mechanics to dominate the world. Here they have Ohm's law, suggesting that it's just like classical behaviour at room temperature," he says.
He reckons the large number of phosphorus atoms in the wire provided a very high density of electrons (1021 per cubic centimetre) and that their mutual scattering destroyed any quantum coherence, leading to classical behaviour.
That bodes well for doing the experiment at higher temperatures. "If they behave classically at low temperature, then they are also likely to behave classically at room temperature," says Simmons.

Coherent problem?

Indeed, Simmons says that the new wires are great news for those hoping for ever-tinier computing devices. "It shows that you can maintain low resistivity and make very thin conducting wires, which is obviously essential for down-scaling devices towards the atomic scale," she says.
The implications for quantum computing are less clear. Simmons's team had already shown that individual phosphorus atoms can exist in a superposition of spin states, making up the quantum bits, or qubits, needed for quantum computation. She thinks that the nanowires could be used to interconnect qubits and help build quantum circuits.
Ferry thinks otherwise. "This lack of quantum coherence is good for Moore's law, but it's bad for quantum computing, because you need quantum coherence for quantum computing. This may make it less likely to occur."

Journal reference: Science, DOI: 10.1126/science.1214319

No comments:

Post a Comment