杏吧原创

(Image: D-Wave)

With the first quantum computer already on the market, and more in the offing, should you splash the cash? Here鈥檚 our verdict on the best buys out there

IS THERE anything more to quantum computers than talk, talk, talk? You might have given up waiting 鈥� after all, it鈥檚 been more than 30 years since physicist Richard Feynman came up with the idea. He wanted to harness spooky quantum effects to seriously surpass the processing power of any normal computer. In that time, normal computers have become around a billion times faster. Quantum computers, by contrast, are still struggling with primary school arithmetic.

At long last, though, there鈥檚 good news. In laboratories around the world, researchers have been beavering away, and they can鈥檛 help feeling there鈥檚 magic in the air. 鈥淪ome aspects of this field are getting tantalisingly close,鈥� says Matthias Steffen of IBM鈥檚 quantum computing division based in New York. You can even buy a quantum computer right now 鈥� maybe 鈥� but you鈥檒l need deep pockets. It doesn鈥檛 matter what apps you鈥檙e planning to run, this much computational horsepower doesn鈥檛 come cheap.

With that warning in place, let New 杏吧原创 help you make an informed choice 鈥� whether you鈥檙e an online game freak looking to take multiplayer to a whole new level, an engineering powerhouse looking to stay one step ahead, or a security service worried about keeping the nation鈥檚 secrets under lock and key. Over the next five pages you鈥檒l find out what they can do, what different kinds are on offer, and whether you鈥檒l have to turn the back bedroom into a cryogenic coolant plant.

Getting started

Baffled by how a quantum computer is supposed to work? Don鈥檛 worry, some of the biggest brains in physics can鈥檛 figure it out either. Some say such computers run in a swathe of parallel universes; others claim they transcend all normal notions of space and time. Whatever the truth, here鈥檚 a rundown of the basics.

Qubit:

Ordinary computers use 鈥渂its鈥� to process information. The basic unit of quantum computing is the qubit. These are physical systems that can exist in two different states, so they can represent the 1s and 0s that make up the binary code computers run.

A qubit might be an electron held in a magnetic field, or a photon that is polarised so its spin can be easily manipulated. Preparing qubits, as well as reading and writing to them, involves some hardcore hardware. Depending on your choice of technology, you might need a ruby laser, a non-linear crystal or even a pink diamond.

Superposition:

This is where the magic happens! The advantage a qubit has over a normal bit is that it can be put into a superposition state, being 0 and 1 at the same time. But this is tricky to pull off 鈥� any stray heat, electromagnetic noise or physical bump can knock it out again. So you鈥檒l have to invest in some serious refrigeration, tinfoil shielding and tiptoe around your quantum computer, or invest in a state of the art vibration containment system.

Even then, you can only run the computer for a limited time before the superposition collapses. You鈥檒l want to check out this 鈥渃oherence time鈥� carefully, as well as evaluating how many errors are created.

Entanglement:

Okay, we lied, this is where the magic really happens. Thanks to what Einstein termed spooky action at a distance, two subatomic particles can become inextricably interlinked, or entangled. The link lets you manipulate multiple qubits at once. That鈥檚 what makes quantum computers so kickass: just eight qubits, held in superposition and entangled, can simultaneously represent every number from 0 to 255, letting you carry out many operations at once.

So that鈥檚 the other thing to look for when weighing up a purchase: how many entangled qubits can your chosen machine manage at once? Don鈥檛 set your sights too high. At the moment, 14 is the record, achieved in 2012 by Rainer Blatt鈥檚 group at the University of Innsbruck in Austria.

Error correction:

Even normal computers make mistakes. Sometimes a bit can be buffeted by a voltage spike or a passing cosmic ray, changing it from a 0 to a 1, say. Processors deal with this by keeping copies, but this isn鈥檛 an option for qubits, thanks to a law called the no-cloning theorem.

Fortunately there are error correction algorithms to get around this. The drawback is that these need a lot of qubits, anything between 100 and 10,000 times as many as needed for the actual computation you鈥檙e trying to perform. Happily, our ability to assemble arrays of qubits for error correction has come on leaps and bounds. And error rates have been creeping downwards too. In June, IBM unveiled error correcting code that is well suited to the large arrays of qubits expected to outperform regular machines. Essentially we鈥檙e where we need to be to start building interesting quantum computers.

Hardware

Spin or superconductor? That鈥檚 the 鈥淎pple or Android?鈥� of the quantum computing world. Superconducting qubits have been around longer, but spin鈥檚 the ultracold new thing 鈥� and there are a few wild-card options to boot. Here鈥檚 what you need to know.

鈥淪uperconductor or spin? That鈥檚 like deciding between Apple or Android鈥�

Superconducting qubits:

This is the grandaddy of all quantum computer tech. Back in 1962, Cambridge physicist Brian Josephson showed that putting a small gap into a strip of superconductor 鈥� a material that has zero resistance to the flow of electricity at low temperatures 鈥� has a surprising effect. For example, superconducting loops incorporating such a 鈥淛osephson junction鈥� let current flow clockwise and anticlockwise simultaneously. That鈥檚 a superposition of states 鈥� just what you need for a qubit.

What鈥檚 more, these systems are manufactured on the mainstay material of the tech industry: silicon. 鈥淭hat allows you to use standard lithographic tools,鈥� says Steffen. 鈥淵ou鈥檙e not in thrall to natural systems, and once you can do a handful of qubits on a chip reliably, you should be able to put many more onto the same chip.鈥� That makes superconducting a good choice for the buyer looking for a tried-and-tested solution to their quantum computing needs.

If you鈥檙e sold on this approach, you still have a choice to make: transmon or Xmon? Transmons are loop-shaped and, at the moment, up to five of them can be linked together. A standard transmon can maintain its coherence for around 50 microseconds 鈥� long enough to be used in quantum circuits. What鈥檚 more, coherence times twice that length, and transmon arrays of 10 to 20 loops, are just around the corner, according to of Chalmers University in Gothenburg, Sweden.

The Xmons created by a team at the University of California, Santa Barbara (UCSB), are cross-shaped superconducting qubits made from sapphire sitting on aluminium. The UCSB group can connect up five of these to create an array that corrects its own errors, and are working on a nine qubit array. UCSB鈥檚 John Martinis, who has , thinks they can power ahead now: 鈥淢y challenge to the group is to double the number of qubits every year.鈥� That鈥檚 a big ask: the architecture and the kinds of algorithms the machine will run still need work.

Spin qubits:

If you鈥檙e more of an early adopter, you might check out what is up to at the University of New South Wales (UNSW) in Sydney, Australia. Their single atom of phosphorus sitting inside a silicon chip may not sound as impressive as a quintuplet of superconducting loops, but it certainly has its advantages.

Morello鈥檚 team is able to put the atom鈥檚 spin in a superposition, manipulate this blurred quantum state and then read it out by applying a microwave pulse. Morello says the team has maintained coherence for 鈥渢ens of seconds鈥� 鈥� plenty of time to run the kinds of quantum apps being dreamed up. There鈥檚 scope for going longer, too. Researchers at Simon Fraser University in Burnaby, Canada, have managed to get their phosphorus-in-silicon rig to hold for nearly 40 minutes at room temperature. They also preserved the superposition while cycling the material between room temperature and 4.2 kelvin.

At the moment, the UNSW group is squeezing two qubits out of a single atom, using the phosphorus nucleus as the first and one of its electrons as the second. In June, they announced that they could now couple two atoms together and read out all four spins, although they haven鈥檛 yet managed to manipulate them. Once they can, they aim to have a handful of qubits that will let them start making quantum calculations. But that will take three to four years, and the wait for apps will be even longer.

Some researchers are looking for a more high-end solution 鈥� which is where those diamonds come in. When certain diamonds are formed, a nitrogen atom can sneak into the place of a carbon atom to give the gem a slight pink colouring and, at the same time, leave an empty space nearby in the crystal lattice. This combination of nitrogen plus 鈥渧acancy鈥� (NV) can be used to create a qubit; the vacancy has distinct quantum energy levels that can be put into superposition using a pulse of laser light. Further pulses can manipulate the state and read out the result of the operation.

In May this year, researchers from the Delft University of Technology in the Netherlands managed to that were 3 metres apart. This is a baby step towards quantum computing in the cloud and a quantum internet.

鈥淚nformation has been quantum teleported between two diamonds鈥�

Before we get ahead of ourselves, there鈥檚 a catch. You can鈥檛 manufacture these blingtastic NV qubits to order. Putting hundreds of them together, as would be necessary for a useful quantum computer, would give a noisy output. But that鈥檚 not a deal-breaker. Last year, a team led by Simon Benjamin at the University of Oxford showed that you can put just a few NV qubits together in a 鈥渃ell鈥�, then link those cells together with photons that act as the input and output bits. Even at room temperature, the cell can maintain coherence for around a second, and the noisy photon network that connects cells can tolerate a 10 per cent error rate without crashing.

Also available:

If you鈥檙e really into bleeding-edge technology, there are a few other options to consider. Ion trap quantum computing is actually the market leader when it comes to entanglement. This technique has linked together a whopping 14 qubits made from ions 鈥� typically ytterbium 鈥� held in carefully shaped electromagnetic fields and manipulated with laser or microwave pulses. But that鈥檚 still a long way from a useful computer, and researchers are finding it tricky to scale up.

You would have to be brave to put all your money on photonic quantum computing, too. Photons look like they would make good qubits: they are easily superposed and stay coherent for good lengths of time. But although it鈥檚 possible to work with particles that move at the speed of light, it isn鈥檛 easy.

Topological quantum computing, where qubits are encoded in the way subatomic particles move past one another, has its pluses too 鈥� it鈥檚 particularly resistant to environmental disturbances for one thing. Microsoft is beginning to invest heavily in it, but it won鈥檛 be in the shops any time soon. Maybe one to consider when your first quantum computer starts to look a little vintage.

Verdict: Superconducting qubits might attract those who like to play it safe, but spin could just overtake it during the next decade. Everything else is for die-hard experimenters only.

Apps

Hardware is important, but what really matters is the software. So what will you be able to run on your quantum computer? Here鈥檚 our pick of the best apps in the pipeline.

Factorisation:

This is the killer app for quantum computers: finding the factors of a large number. Why? To find a factor, normal computers have to try every possible combination, which takes an incredibly long time. Because of this, factorisation algorithms are used to protect data in situations ranging from banking to internet databases.

Twenty years ago, mathematician Peter Shor devised an algorithm that could break such security with ease. However, Shor鈥檚 algorithm only works on quantum computers and so far the largest number that has been factorised is 21. Don鈥檛 you dare sneer 鈥� when this app takes off it鈥檒l put the revelations of WikiLeaks and the capabilities of the NSA to shame.

Good for: Anyone who wants to read other people鈥檚 secrets.

Backward search:

It鈥檚 no wonder Google鈥檚 PageRank algorithm has been such a money-spinner; searching through an unsorted database is a colossal task for ordinary computers. If there are N possibilities, a standard computer will, on average, find what you鈥檙e looking for in a time related to N/2. Bell Labs researcher Lov Grover showed in 1996 that a quantum computer with the right algorithm can speed up the process, taking a time that is related to the square root of N. It works by fiddling with the qubits to make the object of the search the most likely outcome of a measurement on their superposition.

Good for: Budding dot-com billionaires, people in need of plumbers.

Quantum simulation:

Where it all started. If you want to a simulate a complex quantum system 鈥� a large molecule, say 鈥� there鈥檚 just too much detail for a normal computer, no matter how powerful. So as Richard Feynman realised in 1982, you need a quantum computer. For example, if you have one photon to represent each particle in the system you want to simulate, such as an atomic nucleus, then you can put them through a series of quantum gates that mimic interactions. Handy if you don鈥檛 have the maths or equipment to try it for real.

Good for: Impoverished chemistry students, entrepreneurs searching for everyday superconductors, nuclear warhead designers.

Optimisation:

If you want to baffle a computer, ask it to work out the best shape for the front end of an aeroplane. There are so many variables to consider that you鈥檒l be dead and gone long before it spits out an answer. This kind of problem is perfect for quantum 鈥渁nnealing鈥�, named after the process of shaping materials by cycles of heating and cooling.

Annealing is a way of representing all the different possibilities and their consequences as a landscape of hills and valleys; the ideal solution is the lowest point in that landscape. A quantum computer can survey the whole landscape at once, while a classical machine must run through it repeatedly. That鈥檚 the idea, anyway 鈥� no one really knows if it鈥檒l actually work that way.

Quantum or quicksilver?

Can鈥檛 wait to get your hands on a shiny new quantum computer? The good news is that you can buy one today, if you have $15 million to spare. The bad news is nobody knows if it actually is one.

鈥淵ou can buy a quantum computer today 鈥� if you have $15 million spare鈥�

D-Wave Systems of Burnaby, Canada, is the upstart breaking in on the quantum action. Its flagship model, known as D-Wave Two, or Vesuvius, contains 512 superconducting loops of niobium metal, each containing a Josephson junction.

But be warned, this is no quantum laptop. The whizzy-looking black box it鈥檚 housed in, along with its supporting cryogenic system and supercomputer interface, fills a room 10 metres squared. Perhaps surprisingly then it runs on just 15 kilowatts, less than a thousandth of the power devoured by Tianhe-2, the world鈥檚 fastest supercomputer.

A full 512-qubit performance would leave rivals in the dust, but D-Wave doesn鈥檛 worry about being able to address each qubit individually, with having all the qubits entangled together or even operating properly as Josephson junctions. So it鈥檚 not clear that they actually are qubits, and tests have proved inconclusive as to whether it really outperforms ordinary computers.

鈥淲e鈥檙e not in the business of trying to prove whether it鈥檚 quantum or not,鈥� says Bo Ewald, CEO of D-Wave USA. 鈥淲e don鈥檛 know how much coherence we鈥檝e got, but we鈥檝e shown that eight qubits were entangled and some external work showed 40 were entangled.鈥� He believes all the loops are entangled, but doesn鈥檛 want to spend megabucks proving it: 鈥淭rying to measure this means turning it into an experimental physics device. We鈥檙e more focused on just using it as a computer.鈥�

It鈥檚 a computer with only one application: an optimisation algorithm that searches for the best solution to a given problem. That鈥檚 enough for D-Wave鈥檚 first two customers. Google is using it in machine learning for its head-mounted display Google Glass; so far it has put the D-Wave machine to work finding quicker ways to recognise certain objects in an image. Those can be transplanted back into traditional computers, making them more efficient at the task.

Lockheed Martin is using the machine to find out where its aircraft software might go wrong. The company gives its aircraft control system a bad result 鈥� such as the aircraft nose going in the wrong direction when the pilot pulls up on the stick 鈥� and asks the D-Wave machine to look for input scenarios that might lead there.

D-Wave thinks it鈥檒l find more customers in medical imaging, financial planning and delivery scheduling, but the company is open to offers. It might also be worth noting that Google has now started investing in other quantum technologies: in September the company announced a partnership with UCSB to build an Xmon-based quantum computer.

The verdict

So what鈥檚 right for you (see diagram)? Right now the field is wide open. The experts say it鈥檚 not clear which platform will win out and make it to market, but the smart money is on a hybrid that uses one technology for the computing and another for the networks to connect it up. Exactly when that will happen 鈥� D-Wave and a few experimental and very specialised machines aside 鈥� is up in the air. Insiders say you might not be taking delivery of your general-purpose quantum computer until 2024 or even later. But it鈥檒l be worth the wait to get hold of a machine that can wipe the floor with anything we have today. Don鈥檛 forget, though: such power comes with a hefty price tag. Better start saving.FIG-mg29910901.jpg

This article appeared in print under the headline 鈥淨uantum computing best buys鈥�