午夜福利1000集合

IN THE quantum steeplechase, the runners are competing for a prize beyond riches: a computer so powerful it can simulate the most complex and mysterious aspects of the Universe without breaking sweat, leaving today鈥檚 supercomputers looking like ageing nags.

Many of the runners have already attracted widespread interest. But could this be a race in which the fences are so menacing, the distances so vast and the water jump so expansive that nobody makes it to the finish?

It wouldn鈥檛 be the first time. Forty years ago, researchers claimed they could harness nuclear fusion to generate cheap, clean power. Today, such power stations are as far off as ever.

Pushing technology beyond the limits of known science is always risky. No one can tell what unimaginable complexities will arise or what laws of physics will emerge to thwart you. Quantum physicists will have several tricky fences to jump before they can build their devices, and their dreams may rapidly change into nightmares.

What makes a quantum computer such a glittering prize? The key is in the nature of the information it would exploit. Think of ordinary information and you probably imagine the 0s and 1s of binary code, but that changes when you work with quantum particles. Information becomes a strange, ethereal substance, because quantum bits or qubits can be both a 0 and a 1 at the same time. And when put to work inside a quantum computer, they can do extraordinary things.

Take the spin of an electron. Think of it as being like the spin of a basketball with its axis pointing either up or down. Let鈥檚 say spin 鈥渦p鈥� corresponds to a 0 and spin 鈥渄own鈥� to a 1. But the electron can also be placed in a dual existence known as a superposition of states in which its spin is both up and down, a 0 and a 1 at the same time. Perform a calculation with this ghostly electron and you compute an answer using both the 0 and the 1. That鈥檚 two calculations for the price of one.

While a single qubit can be in a superposition of two states, a pair of qubits can be in a superposition of four states. The states can be written 00, 01, 10 and 11 (meaning: both spins down; first one down and second one up; and so on). So with two qubits, the system can be in one, some or all of these four states at once. That way you can do four calculations at once.

And then it gets really interesting. By the same logic, three qubits can be in a superposition of eight states, four in 16, five in 32 and so on. This exponential increase means that with only a few hundred qubits it is possible to represent simultaneously more numbers than there are atoms in the Universe.

With this power, quantum computers would make today鈥檚 supercomputers look like pocket calculators. They could crack the most fiendish codes, answer problems once deemed unsolvable and run simulations so authentic they can鈥檛 be distinguished from reality.

And that鈥檚 just the start, says Anton Zei-linger, a qantum physicist at the University of Vienna. 鈥淛ust look at the way ordinary computers were first used and how they are used today. Nobody could have imagined it.鈥� That鈥檚 why physicists the world over are racing to build a quantum computer, backed by nervous governments, power-hungry military organisations and profit-hungry companies.

But creating a quantum computer is a daunting task. The quantum states used to store information are fragile things-you only have to look at a superposition for it to immediately collapse into a single state. This means that just reading out the result of a computation is tricky, and quantum programs have to be designed to make it as easy as possible (see 鈥淨uantum software鈥�).

Interactions with the environment also disturb any quantum state-a problem known as decoherence. So the qubits have to be kept isolated. On the other hand, they have to be somehow connected, because the computer must be able to link up the qubits to perform logical operations.

There are two operations from which all others can be derived. The first flips a single qubit: if the qubit is in a superposition of a little bit of 0 and a lot of 1, it will be flipped to a lot of 0 and a little bit of 1. For example, that could involve passing an electron through a magnetic field that flips its spin from up to down. The second is a controlled-NOT or CNOT gate, which flips one qubit depending on the state of the other. For this to happen, one qubit must somehow feel the presence of the other and influence it. And a quantum computer requires the manipulation of many qubits in this way, each protected from decoherence and able to interact with all the others. It isn鈥檛 easy.

And yet in the mid-1990s, scientists found one way to do it. Using a technique originally developed for medical imaging called nuclear magnetic resonance or NMR, information can be stored deep inside molecules in the spins of the atomic nuclei. The advantage of nuclear spins is that they are almost entirely cut off from the environment-they barely interact with their surroundings. Yet because spinning nuclei act as tiny magnets, they can be controlled using a magnetic field or the electromagnetic fields in radio waves. In a strong magnetic field, the spins wobble at slightly different frequencies depending on their chemical environment. By zapping the molecule with radio waves tuned to these resonant frequencies, you can manipulate each nucleus individually.

Calculate with chloroform

One molecule that can act as a 2-qubit quantum computer is chloroform (CHCl₃). It has one hydrogen and one carbon nucleus. A quantum calculation involves applying a magnetic field and then zapping the molecule with a series of radio pulses that encode a sequence of 1 and 2-qubit operations. Flipping one qubit is relatively straightforward-you just hit it with its resonant radio frequency. The 2-qubit operation is more complicated. The two nuclear magnets don鈥檛 interact directly, but are coupled together by the electrons around them, which act a bit like a spring. The strength of this coupling depends on the two nuclear magnetic states, and that can be exploited to create a CNOT gate. If the hydrogen spin is up, for example, a certain radio pulse will jangle the connecting spring so as to flip the carbon spin. But if the hydrogen spin is down, the connecting spring becomes sensitive to different frequencies, so the same radio pulse won鈥檛 do anything to the carbon spin.

With a bigger molecule, there are more nuclei in different chemical contexts, so you can have more qubits. Indeed, several groups have built working NMR computers that can manipulate up to seven qubits. A 10-qubit machine is even rumoured to exist. Last year, a team at IBM factored the number 15 using a 7-qubit NMR device. But here so-called liquid NMR runs out of steam. 鈥淭he whole show stops at 15 qubits,鈥� says Leo Kouwenhoven, a quantum physicist at the Delft University of Technology in the Netherlands.

That鈥檚 because to read out the result of a calculation, you have to measure the magnetic field produced by the reoriented nuclear spins, and because it鈥檚 such a weak signal you have to use huge numbers of molecules. Worse, the chloroform molecules start out with random spin states, distributed almost equally, which confuses the output signal. The signal fades rapidly with the number of qubits, which means you need still more molecules. NMR computers built so far already need around 10²³ molecules floating in a tub of solvent. Scientists do not expect to be able to handle more than a dozen or so qubits before the signal fades away.

A really useful quantum computer would need maybe 100 qubits. With only 10 or 12, liquid NMR is forever limited to solving school sums. In the quantum computing steeplechase, NMR is out of the race.

Clearly, some technology is needed that allows you to read out the state of a single quantum system, so you don鈥檛 need a whole bucketful. Ions seem promising, because you can use light to read out the spin state of a single ionised atom. What鈥檚 more, scientists have already developed ways to handle individual ions trapped in magnetic fields. Since an ion鈥檚 nucleus can be used to store a qubit, it wasn鈥檛 long before they realised that by coupling two or more ions together they could create a simple quantum logic gate.

The idea is to trap extremely cold ions with electromagnetic fields that hold them tightly in two dimensions but only weakly in the third (see Graphic). The ions repel each other and arrange themselves in a line with equal spacing, like beads on an elastic string.

One scheme uses beryllium ions, each forming a two-qubit logic gate. One qubit is stored in the energy level of an electron orbiting the ion, and a second in the vibration of the ion within the trap. Both of these states can be in quantum superpositions: the electron can have both high and low energy, and the ion can be both vibrating and still.

A laser would be used to couple these two qubits and perform the CNOT operation. Only if the electron is in a particular energy state will a photon in the laser beam have the right energy to resonate with the ion, and so change its vibration.

With more than one ion, the vibration gets passed along the string, acting like a data bus to transfer information. David Wineland, a physicist at the National Institute of Standards and Technology in Boulder, Colorado, has already demonstrated that a string of four ions can be linked in this way, and has performed the equivalent of a 2-qubit logic operation with them. But Wineland and his competitors around the world are rapidly approaching the hurdle of decoherence.

Some quantum states are more fragile than others. The nuclear spins of NMR are quantum fortresses compared with ion vibrational states. Because ions are charged, any stray electric field can set up unwanted vibrations and turn the quantum information into quantum gobbledegook.

The problem of decoherence was once thought so severe that quantum computers wouldn鈥檛 work. But in 1995, Peter Shor at AT&T鈥檚 Bell Laboratories in New Jersey and Andrew Steane at the University of Oxford came up with a plan for taming it. If you pass the quantum information from particle to particle before decoherence has a chance to dissipate it, it can stay intact indefinitely.

But crucially, for this to work the qubits must stay coherent long enough to be passed around. Wineland thinks that should be possible for nuclear spin states, but for vibrational states it鈥檚 not clear.

And there鈥檚 another problem. Without the ability to join logic gates together, an ion trap machine would be no more powerful than a supermarket cash register. So Wineland and his team have been looking for ways to make the link. Adding ions to the string isn鈥檛 an option because the number of modes of vibration rapidly becomes too great to handle. Instead, his hopes are pinned on physically moving ions between logic gates. But so far it hasn鈥檛 worked. Wineland says the ions become overheated during their short journey and this destroys the information they carry. 鈥淲e don鈥檛 know what鈥檚 causing the heating. Nobody has run into this before,鈥� he says.

One alternative to ions is neutral atoms, which cannot be influenced by stray electric fields. The qubit would be stored in the energy states of electrons orbiting their nuclei. An electron can be boosted to a higher energy level when a photon of the right wavelength strikes it, and a photon is generated when the electron falls back to its ground state. So an atom in an excited state could represent a 1, for example, and an atom in its ground state could be a 0. They could then share information via photons held in a mirrored cavity. But the disadvantage is that the atomic states exploited for transmitting and receiving the photons must be highly unstable, as otherwise they would never release their photons-and again this makes them vulnerable to decoherence. If that weren鈥檛 bad enough, the cavities will never be perfect: some light, and hence some information, will always leak out. Neutral atoms don鈥檛 seem like a front-runner.

Silicon dreams

Many researchers think that the techniques for manipulating individual ions and atoms are so clumsy that they can never be scaled up into a large quantum computer operating with hundreds of qubits. For that, they say, you need something solid-the quantum equivalent of a computer chip. David DiVincenzo, a physicist at IBM鈥檚 T. J. Watson Research Center in Yorktown Heights, New Jersey, developed one of the earliest proposals for a quantum computer of this type in 1998, and could be described as a solid state evangelist. 鈥淲e know what we have to do. It鈥檚 just a matter of doing it,鈥� he says.

DiVincenzo鈥檚 idea, which he developed with Daniel Loss at the University of Basle in Switzerland, is to use the spin of single electrons as qubits, and store them on the surface of silicon chips within structures known as quantum dots. Electrodes under each quantum dot can then flip electrons鈥� spin or force two electrons on nearby dots to overlap. That makes them interact, allowing 2-qubit operations. The answer is measured with a magnetic device that distinguishes spin up from spin down. If it works, many millions of dots could be fitted on a single sliver of silicon.

Kouwenhoven has managed to set up a single qubit this way, and thinks he can produce a 3 or 4-qubit processor within four years. But he admits it is going to be hard. Decoherence is the hurdle again: like ions, electrons are susceptible to stray electrical fields.

It鈥檚 possible that quantum states involving large numbers of particles would be more robust. A number of groups are toying with superconductivity, a quantum phenomenon in which current flows with zero resistance. In 1999, a team at the Delft University of Technology in the Netherlands designed a superconducting circuit in which the current flowed both clockwise and anticlockwise in a superposition of states. These circuits can easily be carved onto chips, which is why physicists say they could be joined together into a large-scale computer. Unfortunately, it鈥檚 too soon for anyone to have attempted the 2-qubit operations needed even for small-scale computation, and the fundamental question of whether such a quantum system is more robust than a single quantum particle has yet to be answered.

The most advanced solid state approach is the NMR chip proposed by Bruce Kane at the University of Maryland (New Scientist, 13 January 2001, p 14). The idea is to bury an array of phosphorus atoms in silicon and place an electrode above each. The qubits are stored in the nuclear spin of each atom, and changing the voltage across the atom changes the frequency of the radio waves the nucleus responds to. And since the voltage over each atom can be varied separately, each nucleus can be addressed individually. The phosphorus atoms are close enough for the electrons in their outer shells to interact in a way that can also be controlled by the electrodes. And two-qubit operations can be performed by transferring the qubit from the nucleus to the outer electrons, then allowing those of neighbouring atoms to interact.

At the Centre for Quantum Computer Technology at the University of New South Wales in Australia, Robert Clark heads a team that has taken on the challenge of building Kane鈥檚 computer. They have had to work out how to place the phosphorus atoms in just the right spots and stop them migrating, and align the electrodes so that they sit precisely on top of the atoms. In March the Australian group announced that it had made a simple chip-an impressive achievement. Unfortunately, the group has so far been unable to read the quantum information stored in the chip and so cannot tell whether it works as expected or not (New Scientist, 30 March 2002, p 13). Another group at the University of Cambridge have suggested using sodium atoms instead of phosphorus, as they can be manoeuvred around inside the silicon with electric fields. But this still hasn鈥檛 been turned into a quantum computer.

So neither electrons nor ions nor atoms look like reaching the finishing line any time soon. But there鈥檚 still one quantum particle that physicists feel more comfortable playing with than most-the photon. And photons are extraordinarily robust: under the right circumstances they do not interact easily with other particles, and especially not with each other. With this built-in protection against decoherence, shouldn鈥檛 they be the ideal quantum particles for quantum computation? Actually, no. The inability of photons to interact with each other means that under ordinary circumstances two-qubit operations are almost impossible. It鈥檚 the old conundrum of needing qubits that are aloof, yet interactive.

Some have suggested ways to force the photons into partnership, for example using crystals that mediate an interaction between photons. But there is little progress to report. 鈥淚t鈥檚 a very difficult engineering challenge. Although nothing in principle prevents us, nobody has successfully demonstrated the technique,鈥� says Manny Knill of Los Alamos National Laboratory in New Mexico.

The race for the quantum computer is littered with disqualified runners. Yet there鈥檚 cause for optimism. 鈥淧hysicists believe they can build a quantum computer because the laws of physics allow it. There is no 鈥榥o-go鈥� theorem that says we cannot do it,鈥� says Artur Ekert, a quantum physicist at the University of Oxford. So in principle, problems such as decoherence can be beaten.

At least, that鈥檚 how the laws of physics stand right now. But Ekert argues that even if the drive to build quantum computers uncovers some previously unknown law that puts a spanner in the works, then that will be important too. By this logic, physicists have an each-way bet. If somebody actually builds a quantum computer, they will pace proudly round the winner鈥檚 enclosure displaying their victorious steed. And if somebody proves that nature will never allow useful quantum computation, the race will be quietly called off while physicists rejoice in their newfound understanding of the Universe.

But what of the third outcome, the fate that has dogged scientists working on nuclear fusion? What if the race proves so challenging that the finishing line never looms into view? If so, the runners will continue charging into a gloomy unknown, falling one by one until eventually there are none left.

The prize is a machine powerful enough to take on life, the Universe and everything. Justin Mullins commentates on the race to build a quantum computer