Features

Thermal power

So an AQC doesn't have a decoherence time, but it must be almost completely isolated from the world. A small amount of thermal energy jiggles it gently into its actual lowest energy state. The slow change from one state to another is sometimes referred to as 'annealing' - an analogy with the process by which metals are made stronger, by heating an amorphous lump gently, so that it can gradually crystallise into its stronger, lower-energy, crystalline form.

"Adiabatic quantum computing, properly understood, is real quantum computing," says Professor Andrew Steane of Oxford University's Centre for Quantum Computing. "It is fully computationally equivalent to other forms of quantum computing, as long as it can be implemented in a general-purpose way with the right degree of control."

The problem is achieving that control. An adiabatic quantum computer doesn't require total isolation, as other quantum computers do, but it's still pretty demanding. "I doubt this computing method is substantially easier to achieve than any other," says Steane.

The other drawback is that AQCs solve only the problem they are designed for - they are specialised problem solvers and not general-purpose computers. It's possible to take a different problem and map it to the one the AQC solves, using conventional computers to turn it into an equivalent problem.

It's a bit like the way builders use the properties of string and gravity. With a weight tied to its end, the string is a simple machine that will find a vertical. But mediaeval builders mapped a more

ADVERTISEMENT

complex problem on to the same machine - a problem that today would warrant a computer-aided design tool.

The builders anchored the string at two points and let it hang freely. It formed a catenary curve, which is an upside-down image of the perfect load-bearing arch to go between those points. Just like the AQC, the string's solution is its lowest energy state, when it hangs still.

There's a limit to the number of building problems a catenary curve can solve, however, and there must also be a limit to the number of problems an AQC can solve, even with powerful conventional computers mapping them.

It seems that AQCs can't become universal code-breakers, but their advocates claim they can solve NP-complete problems. These are complex but useful problems where it is possible to guess right answers but very difficult to prove they are the best answer. Examples include the 'travelling salesman problem', in which a salesman with a list of customers has to plot the shortest route to all their addresses. It's easy to guess a route, but how would you prove there isn't a shorter one? Many optimisation problems also fall under the NP-complete category.

There are more lucrative NP-complete problems, such as maximising the return on an investment portfolio, or finding the stable shape of a protein molecule in drug design. D-Wave has chosen to solve one from thermodynamic theory, called the two-dimensional Ising problem. Any other NP-complete problem can be mapped to it, says Rose.

When pressed, Rose admits that D-Wave's AQC doesn't come up with an exact solution, but solves NP-complete problems "in the sense that it provides approximate solutions to things that are good enough in that they meet the requirements of the user".

It may be that no machine can completely solve them. "But that's an overly restrictive definition of what 'solving' means," says Rose. "Generally, if a business has one of these problems embedded in its daily operations, it uses what is called a heuristic to solve it, which is a set of rules of thumb that gives good approximate solutions quickly. Our machine has as its intended purpose competition with those heuristics."