“BRUCE’s idea was quite simply brilliant,” says Robert Clark. And he should
know. As the head of the Centre for Quantum Computer Technology at the
University of New South Wales, he understands better than most the challenges of
building the ultimate computer. Physicist Bruce Kane of the University of
Maryland has come up with a chip design that turns common-or-garden silicon into
processing gold. Forget IBM’s Deep Blue supercomputer—Kane’s computer
would defeat Gary Kasparov as easily as the chess grand master would beat a
six-year-old child. And now, behind a veil of secrecy, Clark is attempting to
build it.
The idea is straightforward: place a single atom at the core of a silicon
chip, then use the atom’s electrons for processing and its nucleus as a memory.
In one stroke this unleashes the formidable computing power of the quantum
world; the weird science of quantum mechanics allows these particles to perform
seemingly magical feats of computing. And the masterstroke—putting these
quantum tricks inside standard silicon technology—means that you can turn
the blueprint into reality simply by tapping into the vast experience available
in the conventional computing industry.
When Kane published his proposal for a silicon-based quantum computer in
1998, it wowed researchers in the quantum computing world. They realised this
simple device could be the key to the locked door of quantum computing. Writing
in Nature (vol 393, p113), David DiVincenzo, an expert in quantum
information processing at IBM’s T. J. Watson Research Center in Yorktown
Heights, New York, hailed Kane’s idea as an “audacious vision”. Clark, who knew
of Kane’s idea in advance, had already set to work assembling an international
team of experts to build the device.
Advertisement
Although the design is simple, the techniques involved in making it are
extremely challenging. But Clark’s experts believe they know how to overcome all
the obstacles. While their exact methods are cloaked in secrecy, their goal is
not: they plan to turn Kane’s device into the first fully functioning quantum
computer.
Quantum computing researchers want to create a processor that stores and
manipulates numbers using quantum particles like single atoms and electrons,
rather than the crude technology of capacitors and magnetic disks. Scientists
across the globe are trying all kinds of exotic tricks to achieve this quantum
leap in computing power: trapping single atoms with lasers, playing with the
nuclei of billions of identical atoms and putting materials in strange
superconducting states, all in a bid to harness the strange and powerful nature
of the quantum world.
Kane’s proposal seems ridiculously conventional by comparison. But while it
might look like its familiar and straightforward classical cousin, the weirdness
of the quantum world means that Kane’s chip will act very differently. While
ordinary computers rely on the properties of matter on the large scale—the
bits that your home PC chomps on consist of billions of electrons that are easy
to detect—the quantum equivalent of a bit is stored in a single electron,
atomic nucleus or other quantum particle. Electrons and nuclei have a spin,
which can be either clockwise or anticlockwise relative to the device measuring
it. Clockwise can represent a 1, for example, while anticlockwise represents a
0. The extraordinary thing is that before the measurement takes place, the
particle is in both states at the same time: it is both a 0 and 1
simultaneously, a phenomenon called “superposition”. If the particle takes part
in a computation while it is in this superposition, the calculation occurs for
both a 0 and a 1. This is one of quantum computing’s great powers: the ability
to perform two calculations simultaneously.
Deep connection
But to do really big computations, Kane also needs to harness another quantum
effect known as entanglement. This is a strange state of being in which two
particles are so deeply connected that they share the same existence. Entangle
the individual weirdness of three superpositions together, for example, and
those three 0 bits and three 1 bits become one piece of binary data
simultaneously representing all the numbers from 0 to 7. With entanglement, a
little goes a long way. Entangle 250 quantum bits (know as “qubits”) and you can
simultaneously hold more numbers than there are atoms in the Universe.
This astonishing resource means that a quantum computer can do massive
parallel computations at a single stroke. The experimenter simply carries out a
measurement at the end of the calculation, which destroys the superposition and
entanglement but reveals an answer.
But there is a problem. Superposition and entanglement are extremely fragile
states. Any interaction with the environment has a destructive effect, a problem
known as decoherence. The biggest problem faced by would-be builders of quantum
computers is preventing decoherence taking hold before a calculation is
complete.
All current ideas for making quantum computers fall at this fence. Kane,
however, believes he has found a way over it. First, he says, find an atom whose
electronic and nuclear spins are particularly resistant to disturbance. Then use
the expertise of the electronics industry to surround this atom with
silicon-based electronics that will read its various states. And, hey presto,
you have the perfect building block for a quantum computer.
With this idea in mind, Kane hunted through the data books for his perfect
atom. He found it in a 1959 treatise on donor atoms in silicon: phosphorus
encased in extremely cold silicon (chilled to 1.5 degrees above absolute zero)
has an electron spin lifetime of thousands of seconds, and a nuclear spin
lifetime of more than ten hours. That’s plenty of time to carry out a
calculation.
Kane’s idea is to bury an array of these phosphorus atoms in silicon. Above
this array will be an insulating layer and on top of this, above each atom, an
electrode that can apply a voltage to the atom. “The ideal combination is to use
the electrons for logic operations and the nuclei for memory,” Kane says.
Phosphorus has one unpaired electron in its outer shell which can be used as a
messenger, carrying information to and from the nucleus and exchanging that
information with other atoms in the “if . . . then . . . ” type operations of
logic gates that are the building blocks of computers. The nuclei hold the
qubits in their clockwise or anticlockwise spin. To control the spin of the
nucleus, Kane hits it with radio waves of a specific energy which make it flip
from clockwise to anticlockwise or vice versa.
But you can’t just blast wildly: the radio waves have to be applied with
care, otherwise they flip every nucleus in the array. To selectively control
each phosphorus nucleus, Kane plans to make use of its messenger electron. This
electron interacts with the nuclear spin in a complex way, determining the
relative energies of the nuclear and electronic spins. Applying a voltage to the
atom through the electrode directly above it shifts the plane of the electron’s
spin and changes the energy of both the nuclear and the electronic spin.
Crucially, this also changes the frequency of the radio waves needed to flip the
nucleus. So applying a voltage to a specific electrode changes the resonant
frequency of the nucleus below it; by zapping the whole array with this new
frequency it is possible to address just the target nucleus.
This is a so-called single qubit operation in which a 0 is flipped to a 1 or
vice versa. But to perform the logic operations vital to a quantum computer, two
qubits have to become entangled.
That’s where a second electrode called the J electrode comes in. This sits
above the space between adjacent phosphorus atoms. A carefully controlled
voltage on the J electrode forces the outer electrons of the two phosphorus
atoms to interact. They then form what is essentially a covalent bond—like
the bond between carbon atoms in a diamond. When this happens, the electrons
become entangled and two-qubit operations become possible. Another voltage on
the J electrode shifts the spin of the electrons in a way that reads the values
of the qubits in the nuclei.
Although the blueprint is clear enough, actually building the device is a
significant challenge. But Clark and his collaborators are already working on
it, and they think they know how to overcome the difficulties.
First up is the problem of building the kind of phosphorus atom array Kane
requires. Researchers have long been able to pick up and drop individual atoms
using the tips of scanning tunnelling microscopes. “But you can’t do that with
phosphorus atoms because they bond too strongly with silicon,” says Clark.
What’s more, once the phosphorus atoms have been buried they tend to migrate
through the silicon. Somehow they have to be anchored.
These are difficult problems to solve. “Nobody has the technology for placing
single atoms in specific places over wide areas,” says Kane. But Clark thinks
he’s found a way to do it and has secured funding from the Australian government
to try. For the moment, he is keeping the details of his technique under wraps
to protect it from competitors.
The task of designing a readout for the qubits in the nucleus is being
handled by Kane’s Maryland laboratory. By carefully measuring the spin of the
outer electron, he says, it will be possible to infer the spin of the nucleus.
Measuring the spin of a single electron has never been done before but Kane
believes that placing a sensitive transistor above the atom and measuring the
current passing through the transistor will make it possible. This current is
tiny, however: it consists of just one electron. Kane and his collaborators are
undaunted. “We have the kind of single electron transistors capable of doing the
job,” Clark says.
In principle, there’s no reason why you can’t simply link together as many
logic gates as you need to form a quantum computer that will crunch enough
numbers to satisfy anyone. Doing so may take some time, however: when asked
about the timescales involved, Clark chooses his words carefully. “On the 5 to
10 year timescale, we have a significant chance of making progress on a device
capable of handling a few qubits,” he says. But as Kane points out, the global
computing industry has developed huge expertise in mass producing silicon chips,
so once the basic techniques are finalised, there will be little difficulty in
batch processing the devices.
Even so, other experts are cautious. Jonathan Jones, an expert on nuclear
magnetic resonance technology at the University of Oxford, says: “Whether this
technique will work is up in the air. It involves technologies we don’t have.”
But IBM’s DiVincenzo is more optimistic. “There is a lot of technical art in
using silicon and this is one of the things that make it promising,” he
says.
Kane’s device is not the only one being considered as a model for
silicon-based quantum computing: his innovative idea has triggered a cascade of
proposals. “I’ve been incredibly encouraged by people coming up with new ideas,”
he says. One such proposal, in development at the University of Cambridge, is
also based on the notion of using covalent bonds between electrons in silicon to
carry out quantum logic operations. But its originator, Crispin Barnes, has a
rather more dynamic take on silicon quantum computing. He makes electrons surf
on acoustic waves that move across the surface of a semiconductor. He creates
the waves on the surface of the silicon by passing a voltage across it. By
creating a channel on the surface he can make electrons race along this track at
up to 3000 metres a second, and flip their spins with an electric field from an
electrode placed over the track. This allows him to control the information in
one electron—a one-qubit operation. Racing two electrons down independent
tracks that start off parallel but then get closer together—a kind of
electronic chicane—allows the spins of the two electrons to interact and
forms the same kind of two-qubit covalent bond that Kane wants to achieve. With
both one and two-qubit operations, Barnes can build quantum logic gates out of
specially shaped pairs of electronic racetracks. These logic gates are easy to
link together because they are just extra convolutions in the tracks. An
assembly of these logic gates might look something like a microscopic Scalextric
set.
The real beauty of Barnes’s design is that he can race billions of electrons
along the tracks, one after the other, all performing the same calculation. It
doesn’t matter if a few of them suffer from decoherence: as long as the quantum
properties are preserved in the rest of the electrons throughout the race, he
can measure the flow of spin-up or spin-down electrons at the end. This should
be much easier than measuring the spins of individual electrons. “It looks very
promising,” says DiVincenzo.
Barnes is now looking for funding: although his idea—like
Kane’s—has great potential, making it work is just as difficult. Neither
Kane nor Barnes expect their ideas to lead to a working computer in the near
future. But when they do, they can be sure of a warm welcome in Silicon Valley.
The computer magnates can breathe a sigh of relief: quantum computers aren’t
going to put paid to the silicon chip industry after all. They’re just going to
sprinkle it with a little bit of magic.