快猫短视频

WHATEVER happened to quantum computers? A few years ago, it seemed, it was just a case of a tweak here, a fiddle there, and some kind of number-crunching Godzilla would be unleashed upon us. Just as digital processors changed our lives in ways hard to imagine a few decades ago, the monstrous information processing power of individual atoms and electrons would mean that computing 鈥� and the world 鈥� would never be the same again.

We鈥檙e still waiting. In 2007, a unveiled what it claimed was a quantum computer that could solve a sudoku puzzle, but there remains whether it is truly a quantum computer. Meanwhile, we seem stuck with the conventional, 鈥渃lassical鈥� computers that rattle and purr away on our desks, toggling currents of electrons in billions of silicon transistors to produce the numbers, words and images that frame our lives.

Paradise lost? No 鈥� merely postponed. Progress might have been slower than many quantum evangelists were predicting a decade or so ago, but after a quiet few years quantum computing is back with a vengeance. 鈥淭he rate of progress has been dramatic,鈥� says David Wineland, a at the US National Institute of Standards and Technology (NIST) in Boulder, Colorado. The stage is now set, he and others claim, for the quantum computer to change our lives in the 21st century just as radically as its classical, digital counterpart did in the 20th.

The premise behind a is simple 鈥� provided you swallow the unpalatable quantum truths that underlie it. One is that objects such as atoms and electrons are not confined to being either this or that, as the objects of our everyday macroscopic world are; they can be both this and that at the same time. They might, for instance, be spinning clockwise and anticlockwise simultaneously, or adopt two different energy states at once. This is known as superposition.

What鈥檚 more, these ambiguous quantum characters can club together so that what you do to one affects the others. This is the phenomenon of entanglement or, if you鈥檙e Einstein, 鈥渟pooky action at a distance鈥�. Together, the characteristics of superposition and entanglement make for a computer of awesome power.

Take a classical computational bit such as a transistor current. It can adopt one of two states: 0 (off) or 1 (on). Not so its quantum counterpart, the qubit. Superposition means a single qubit can simultaneously be 0 and 1, giving you twice the information storage capacity right from the start. Then entanglement kicks in, allowing further bits to share their superposed states in a common pool. The result is that computing power grows exponentially with the number of qubits. While three classical bits are needed to store the number 7, three qubits can store all eight numbers from 0 to 7 simultaneously (see diagram). Just a few hundred qubits could store more numbers than there are thought to be atoms in the universe.

That kind of processor is still a way away. The current record for qubit manipulation is held by practitioners of the ion-trap technique, which uses oscillating electric fields to hold atomic ions in place, like eggs in a box, chilled to within a few degrees of absolute zero. Information is encoded in each ion鈥檚 energy state. Whereas computing in a conventional processor is done by switching transistor currents on and off, the qubits are manipulated by firing a carefully designed laser pulse at the ions to put them into a particular superposition state.

Ions or dots?

In August this year, Wineland鈥檚 group at NIST reported a milestone achievement. Taking two ions, they used a series of carefully calibrated laser pulses to perform some simple computations with them and read out the results. They could also move the ions around the processor without losing the information encoded on them, and repeat the process. In other words, their system does everything that a basic conventional computer should do ().

The team has also managed some, but not yet all, of these feats with arrays of eight or nine trapped ions. There should now be no problem, in principle, with scaling things up to the hundreds or even thousands of qubits that would be necessary to make a useful computer 鈥� it is just a question of acquiring more practice in the art of qubit manipulation. 鈥淚t鈥檚 hard to imagine that we won鈥檛 someday be able to control these things to create a useful device,鈥� Wineland says.

Over the past couple of years, however, the tried-and-tested ion-trap approach has acquired a competitor. It takes the form of dots of aluminium about one-third of a millimetre across, each of which contains billions of atoms. When chilled to extremely low temperatures, the momentum of these atoms is severely reduced. According to the quantum-mechanical uncertainty principle, the more restricted an atom鈥檚 momentum is, the more smeared out is its location in space, and it becomes impossible to tell where one atom ends and the next one starts. The result is a dot that behaves as if it were one giant superatom within which electrons flow freely, encountering no electrical resistance.

These superconducting dots make ideal qubits. In May, a team led by at the University of California, Santa Barbara, showed how their energy levels could be manipulated with exquisite sensitivity by spoon-feeding them with energy in the form of single photons from a controlling microwave field .

Then, in July, and his colleagues at Yale University went a step further, reporting how they used a similar approach to perform computations with the quantum states of two dots . Among other things, they implemented a quantum computational procedure known as Grover鈥檚 algorithm. This boils down to the reverse searching of a database 鈥� like looking for a particular number in a telephone directory that is ordered by name. With a classical computer, it takes as many as N processing steps to search a database with N entries 鈥� you can鈥檛 do much better than laboriously compare your number with those listed in the directory, one by one.

The Yale team showed how with quantum trickery the same search needs only about 鈭�N processing steps. For a database with a million entries, that is 1000 times faster than a classical computer with the same number of bits. For a database with 10 billion entries it would be 100,000 times faster. That鈥檚 the difference between running a computer for an hour and running it for 11 years.

In fairness, that comparison is misleading. Conventional processors operate with a lot more than two bits, and so still seriously outperform today鈥檚 quantum machines. Whether you use ion traps or superconducting dots, scaling things up is going to be the key to unlocking the power of quantum computers.

Here, the superconducting dots could have significant advantages over the ion trap. The Yale team鈥檚 two-qubit processor is built on sapphire wafers that are similar to the silicon wafers that underlie conventional processors. There is nothing in principle to stop you putting hundreds or thousands of qubits on one wafer, just as hundreds of millions of transistors crowd into a single silicon bed. 鈥淥ur processor is a superconducting integrated circuit,鈥� says Leo DiCarlo, a member of the Yale team. 鈥淎t its core, it鈥檚 really just an electronic chip.鈥�

It makes for a technology that is reassuringly familiar and user-friendly, and that can be powered and programmed in ways similar to those used for conventional computers. 鈥淚t has the feel of something you could interface with while sitting at a regular computer,鈥� says of Williams College in Williamstown, Massachusetts. 鈥淚n terms of the technology, I don鈥檛 think anything new has to be invented.鈥�

So what are we waiting for? The fruits of two competing schemes are ripe for harvesting. We can manufacture qubits, control them and run algorithms on them. Soon, we may even be doing those things on significant scales. Can we expect quantum computers to be all over the shops by Christmas?

鈥淲e can manufacture qubits, control them and run algorithms on them. So what are we waiting for?鈥�

鈥楩raid not. Before quantum computers are shipped anywhere, one major hurdle remains to be overcome: decoherence.

Decoherence is akin to what happens when a juggler loses concentration and drops the balls. When a qubit is disturbed by some outside influence 鈥� mechanical vibrations, electromagnetic fields or simply heat 鈥� its delicate superposed state can be wiped out and it falls out of the quantum calculation. As with a juggler鈥檚 balls, the more qubits the computer uses, the more likely it is that one of them will go awry and ruin the entire routine.

Battle for coherence

Decoherence is a problem for all quantum computing schemes. In an ion-trap quantum computer, the ions can be heated or shaken out of their pre-set energy states, or even wiped out by the laser used to process the information. In the NIST team鈥檚 breakthrough last month, only 94 per cent of the quantum manipulations came out right. For the remaining 6 per cent, the unruly control laser knocked the ions out of their carefully prepared quantum states.

at the University of Innsbruck in Austria and his team hold the record for trapped-ion computing accuracy: 99 per cent in a two-ion system. For reliable quantum computing, something like 99.99 per cent is needed, but that extra 0.99 per cent will require a daunting degree of laser control. 鈥淚 think of it as a marathon,鈥� says Wineland. 鈥淎t the moment, it is as if we can look over our shoulders and still see the starting line.鈥�

If the ion-trap people can see the starting line, the superconducting-dot teams hardly have a foot over it. When the probing laser does not miscue and destroy the encoded information, trapped ions can maintain their delicate states for about a second before the surrounding environment overwhelms them. That鈥檚 not long, but time enough to carry out basic computing operations. Superconducting dots, on the other hand, can manage but a meagre microsecond or so.

Why decoherence kicks in quite so quickly in superconducting dots is unclear. Suspects include defects at the interface between the dot and the underlying substrate, but no one has any clinching proof.

The problem needs fixing. For simple algorithms such as those demonstrated by the Yale researchers, processing takes few qubits and little time, so things are not so acute. But anything truly useful, such as searching through enormous piles of unsorted data, will require many more qubits and much more processing time, both of which make a fatal decoherence event much more likely. 鈥淓ven just 10 microseconds would give us significant headroom,鈥� DiCarlo says.

At the moment, no one quite knows how to create that headroom for superconducting dots. It might not even be possible. 鈥淒ecoherence may win, just as friction beat the idea of mechanical computers in the 19th century,鈥� says , a physicist at University College London.

The key, for now, is to keep things as cold as possible. 鈥淣oise affects systems in different ways, but all types of noise are suppressed at lower temperatures,鈥� says Wineland. So if the prototypes are anything to go by, the first truly useful quantum computer will come with a mighty refrigerator attached. The qubits for NIST鈥檚 ion-trap computer, for example, sit in a tiny vacuum chamber, but its peripherals take up an entire room.

That sounds like a wake-up call for those dreaming of a desktop quantum computer, but a quick glance at history suggests we should dream on. The first digital computers were room-sized monsters cooled by arrays of fans and vents, but once the basics were in place, they rapidly shrank. Stoneham already sees an opening for history to repeat itself in compact 鈥渢hermoelectric鈥� coolers that, though not yet hugely efficient, when wrapped around a quantum processor might be good enough to make a desktop machine. 鈥淧ersonally, I think there鈥檚 no need for a cryogenic monster 鈥� at least after the initial development phase,鈥� he says.

So what would we do with a supercooled desktop megaprocessor? Reverse database searching is all very well, but it鈥檚 not the kind of thing that most of us are doing all day. Nor are we cracking security codes, the quantum-computing killer app that gets the national security spooks so worked up. And anyway, aren鈥檛 we doing perfectly well with conventional computers?

History is instructive here, too. The universal potential of digital computers was missed for a long time, and the laser was a solution without a problem for its first 20 years, only coming into its own with the explosion of digital storage media such as CDs. 鈥淭here are probably hundreds of applications for quantum computers that we just don鈥檛 know about yet,鈥� Strauch says.

鈥淭here are probably hundreds of applications for quantum computers we just don鈥檛 know about yet鈥�

Blatt thinks that, given their size and expense compared with cheap and mature silicon technologies, the first quantum computers will be bolt-on subprocessors specialised to certain particularly intensive tasks that conventional processors struggle with, such as recognising patterns in data. Meanwhile, applications that are not hindered by the processing speeds of classical computation will continue to be served by silicon. 鈥淒on鈥檛 expect that your Word or Powerpoint will be run on a quantum computer too soon鈥� says Blatt.

As to when such a machine will be available on the street, those involved are reluctant to leave any hostages to fortune. Strauch hazards a guess at a couple of decades. DiCarlo won鈥檛 be drawn at all. 鈥淚鈥檝e been asked by a few people when they will be able to buy a quantum computer in Radio Shack,鈥� he says. 鈥淚 try not to answer that question.鈥�

And you thought today鈥檚 computer-makers were cagey with their latest products. Then again, perhaps it鈥檚 only fitting that, in the world of the quantum computer, the uncertainty principle reigns supreme.