Features

With a quantum computer, however, the time taken to solve the problem doesn't increase exponentially. There's an algorithm created by Peter Shor, professor of mathematics at MIT, that, given a working quantum computer of a certain size, will crack a code that was previously uncrackable to all practical purposes.

But can we build a working quantum computer? If you remember Schrodinger's cat, you'll have guessed the drawback. Quantum systems are coherent only if isolated. Any noise or interaction with the outside world, and the quantum states collapse to one. A quantum computer stops working if you look at it.

To make a quantum computer, it seems, you must overcome three problems: building in enough qubits to build a computer faster than conventional systems; keeping the quantum computer isolated and coherent long enough to do actual work (the so-called decoherence time); and reading the quantum states at the end without destroying the result.

It's been slow going. Scientists have come up with various ways of making qubits, including optical components, trapped ions, quantum tunnelling and superconducting quantum interference devices (SQUIDs).

Quantum computing becomes really powerful only when you have many qubits, but scaling quantum computing experiments up to include more qubits has turned out to be difficult. This year, a couple of university projects have verified Shor's algorithm on systems of four qubits, but they are still far short of today's conventional computers.

The maximum number of

ADVERTISEMENT

qubits in any properly reviewed experiment has been 12, and those were not all coherent. This meant the qubits couldn't perform in the way quantum computing was originally envisaged, with all possible quantum states active.

There have been attempts to get longer decoherence times, but these are still only small fractions of a second - too short for useful work.

Quantum entanglement

With these limitations, researchers have had to devise very clever systems using quantum entanglement to get data in and out of a quantum computer. In quantum entanglement, the states of two particles depend on each other, even when they are separated.

Much attention has been paid to reading data from a quantum computer. Most methods accept that they can only arrive at the result with a certain level of probability, but techniques are evolving to push that level of probability higher.

There have been demonstrations of techniques that could lead to the creation of quantum cables. A team of researchers based in Quebec and Zurich created an 8mm linkage that can connect two SQUIDs. It's a small step, but it could allow qubits to connect to larger devices.

All this sounds disappointingly slow, but this is what happens when a subject is put into practice. Researchers started out in the hope that quantum computers could solve complex problems in one go, but as Umesh Vazirani, professor of computer science at Berkeley, pointed out in a letter to The Economist, "this mistaken view was put to rest in the infancy of quantum computation over a decade ago when it was established that the axioms of quantum physics severely restrict the type of information accessible during a measurement". It's still worth doing, he says, even though you don't get an exponential speed increase over classical computers.

Others are gloomier. Michael Dyakonov of the University of Montpellier in France speculates that "since every qubit creates a greater opportunity for thermal noise, this makes the decoherence time shorter. Quantum computing could be more like perpetual motion - an absolute impossibility."