News Feature | Published:

2020 computing

Champing at the bits

Naturevolume 440pages398399 (2006) | Download Citation


Despite some remaining hurdles, the mind-bending and frankly weird world of quantum computers is surprisingly close. Philip Ball finds out how these unusual machines will earn their keep.

Five years ago, if you'd have asked anyone working in quantum computing how long it would take to make a genuinely useful machine, they'd probably have said it was too far off even to guess. But not any longer.

“A useful computer by 2020 is realistic,” says Andrew Steane of the quantum-computing group at the University of Oxford, UK. David Deutsch, the Oxford physicist who more or less came up with the idea of quantum computation, agrees. Given recent theoretical advances, he is optimistic that a practical quantum computer “may well be achieved within the next decade”.

This excitement is, however, tempered by the hurdles that have yet to be overcome. Building a quantum computer is still very, very hard to do. This is partly because it involves making quantum systems do things that don't come naturally to them. “There is progress, but it's still very slow,” says physicist Chris Monroe of the University of Michigan in Ann Arbor.

And even if we did have a working quantum computer today, there are hardly any programs that could run on it. In fact, it is likely that even once the machines are available, quantum computers are destined to remain niche products — excellent for certain tasks but not versatile devices like conventional personal computers. “Quantum computers will almost certainly never become general-purpose desktop machines,” concedes Isaac Chuang, a quantum physicist at the Massachusetts Institute of Technology (MIT) in Cambridge.

Credit: J. MAGEE

Nevertheless, as a scientific research tool the quantum computer could be revolutionary because of its ability to simulate other quantum systems. In conventional, or classical, computers, information is stored as strings of bits: binary digits each of which can take the value of 0 or 1. The same is true for quantum computers, except that this time the binary digits — ‘qubits’ — are stored in the quantum states of microscopic systems, such as the electronic state of an atom or ion. So by its very nature, a quantum machine should be much better suited to simulating quantum systems than a classical computer.

A quantum simulator would describe and predict the structure and reactivity of molecules and materials by accurately capturing their fundamental quantum nature. This is the sort of employment the early machines are likely to find: doing calculations of interest to chemists, materials scientists and possibly molecular biologists, says Steane.

“Just a few dozen qubits may shed light on other physics problems that are intractable with conventional computers,” notes Monroe. “There are models of high-temperature superconductivity and other condensed-matter systems that might be approached in such a quantum simulator.”

In a spin

“Computers for specific applications are likely to come before general-purpose devices. But that doesn't rule out the possibility that we'll all be playing quantum Grand Theft Auto in the near future. Seth Lloyd, MIT”

In fact, quantum simulations can already be done using atoms and molecules that store qubits in their nuclear spin and can be probed and manipulated using nuclear magnetic resonance (NMR) techniques. In their own terms, these ‘computers’ “run rings around any classical supercomputer”, says Seth Lloyd, a theorist at MIT. He and his MIT colleague David Cory have been using this technique to simulate a variety of quantum systems in crystals of calcium fluoride and other materials. “As the crystal contains a billion billion spins, these simulations remain out of the reach of the most powerful classical computers,” says Lloyd. The approach remains limited in terms of the different systems it can simulate, although Lloyd anticipates that fully functioning simulators will be readily available by 2020.

The key to the potential success of quantum computers is also the cause of the problems within the field: the quantum nature of data storage and manipulation. In classical computers, bits have clearly defined values of 1 or 0, but the laws of quantum mechanics allow qubits to exist in a ‘superposition’ of states — a mixture of both 1 and 0 that would be impossible in an everyday computer. This means that a quantum computer has much greater capacity for storing information.

Holding pen: this circuit traps ions above its electrodes. Credit: M. BARRETT & J. JOST

A quantum processor can also compute with more than one qubit at once by exploiting another quantum property called entanglement, which makes qubits interdependent. The weird nature of the entangled state means that a measurement on one qubit instantly affects another, even though their previous individual states were undefined until that moment. Entangled states don't readily exist in nature: quantum engineers have to make them by allowing qubits to interact with one another.

By exploiting superpositions, a single quantum computer in effect mimics a whole suite of classical computers running at once, and by using entanglement these ‘parallel computers’ can be linked together. Unfortunately, this powerful parallel processor has an Achilles' heel. A quantum superposition has to remain stable for at least as long as it takes to do the computation. But as soon as qubits interact with their environment, the delicate superposition becomes unstable, a process known as decoherence, which causes information to leak from the quantum computer. Decoherence is especially problematic for entangled states, because then the decoherence of one qubit can affect the others too. Preventing decoherence means reducing uncontrolled interactions with the environment. Cooling the quantum system to very low temperatures helps — but it may also be necessary to shield the qubits from stray electromagnetic fields. In practice, researchers have found it difficult to avoid decoherence of specific qubits for longer than a few seconds. But in principle it should be possible. “For qubits encoded in trapped ions, nobody really believes that we will ever be limited by coherence time,” says Monroe.

Despite the fact that qubits need to be isolated from their environment to avoid decoherence, they must interact strongly with one another, to perform computations. And it must be possible for qubits in superposition to interact strongly with the environment when needed, so that the information can be read out. It is an extraordinarily delicate balancing act, which involves rules that defy intuition and aren't even completely understood.

An easy mistake to make

Decoherence also means that, as they process qubits using logic gates, quantum computers will inevitably incur errors at a much higher rate than classical computers. “The modern transistor has an error rate of less than 1 in 1014 or more switching events. In comparison, the best quantum gates we currently imagine will optimistically have an error rate of something like 1 in 107,”says Chuang. Some researchers thought at first that this would make quantum computers too error-prone to be useful. But thanks to quantum error-correcting codes devised in the 1990s1,2, it is now possible to correct error rates as high as 1 in 105.

By 2002 the key principles behind a quantum computer had been sketched out by theorists (see ‘How to build a quantum computer’), but how best to implement them in a real device remains a wide-open question. Much of the current effort is focused on making quantum computers using atoms or ions that are held in a trap. In an ion-trap computer, the qubits are encoded in the electronic states of ions that are confined by an electromagnetic field. The ions interact with each other through electrostatic repulsion, and can be entangled by using laser beams to make them jiggle in unison. The quantum states of the ions can be read out by using other lasers to excite fluorescence, the wavelength of which depends on the ion's electronic state.


  1. 1

    Shor P. W. . Phys. Rev., A 52 . R2493 - R2496 (1995).

  2. 2

    Steane A. M. . Phys. Rev. Lett., 77 . 793 - 797 (1996).

  3. 3

    Stick D., et al. Nature Phys., 2 . 36 - 39 (2006).

  4. 4

    Seidelin S., et al. preprint at (2006).

  5. 5

    Brickman K.-A., et al. Phys. Rev., A 72, . 050306(R) (2005).

  6. 6

    Hallgren S., et al. in Proc. 34th Annu. ACM Symp. Theor. Comput.. 653 - 658 (ACM, New York). (2002).

  7. 7

    Nielsen M. A., Dowling M. R., Gu M.& Doherty A. C. . Science, 311 . 1133 - 1135 (2006).

  8. 8

    DiVincenzo D. P. . Fortschr. Phys., 48 . 771 - 783 (2000).

Download references

Author information


  1. consultant editor for Nature

    • Philip Ball


  1. Search for Philip Ball in:

Related links

About this article

Publication history


Issue Date


Further reading

  • Computing: report leaps geographical barriers but stumbles over gender

    • Martha E. Pollack
    • , Susanne E. Hambrusch
    • , Carla Schlatter Ellis
    • , Barbara J. Grosz
    • , Jessica Hodgins
    • , Ruzena Bajcsy
    • , Carla E. Brodley
    • , Luigia Carlucci Aiello
    • , Maria Paola Bonacina
    • , Lori A. Clarke
    • , Julia Hirschberg
    • , Manuela M. Veloso
    • , Nancy Amato
    • , Liz Sonenberg
    • , Elaine Weyuker
    • , Lori Pollock
    • , Mary Jane Irwin
    • , Lin Padgham
    • , Barbara G. Ryder
    • , Tiziana Catarci
    • , Kathleen F. McCoy
    • , Maria Klawe
    • , Sandra Carberry
    • , Laura Dillon
    • , Kathleen McKeown
    •  & Mary Lou Soffa

    Nature (2006)


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing