THE lab has the feel of the Cold War about it. At the entrance, security
guards check guests for unauthorised cameras and recording devices. Foreign
nationals must surrender their passports for ID checks. Even the cafeteria has a
government-issue air about it.
This is the Applied Physics Laboratory, a department of Johns Hopkins
University, set in 350 acres of rolling hills not far from Washington DC. Much
of the research is funded by the US Department of Defense, and the work of James
Franson is no exception.
The military are interested in Franson’s work because he may have found a way
to build superfast quantum computers that can carry out calculations impossible
by any other means. The big problem with quantum computing is that any
calculation is ruined if the particle carrying the information is disturbed in
any way. So if you use electrons, atoms or ions to carry information, they have
to be completely isolated, which requires all sorts of expensive equipment.
Advertisement
Light, on the other hand, is much less prone to unwanted disturbances. And it
can be handled with simple equipment such as fibre optic cables, so it’s ideal
for carrying information. The trouble is that to actually make calculations, the
information carriers have to interact—and under ordinary circumstances
photons, the smallest possible bundles of light, simply ignore each other. This
is why light rays pass through each other unaffected. Making photons interact
turns out to be extraordinarily difficult.
But this is just what Franson aims to do. He has worked out that in certain
circumstances, one photon can twist the polarisation of another. And he thinks
he has a simple way to make it happen.
In a small lab across the hall from his tiny office, Franson is testing a
simple device that could play the role of a quantum logic gate that uses photons
as information carriers rather than the electrons used in conventional
computers. Although researchers have been thinking about quantum computers for
decades, nobody has come up with a way to mass-produce quantum logic gates. But
if Franson can make his device work, quantum logic gates could become simple and
cheap to make. And by combining lots of logic gates into a large-scale device,
Franson hopes to become the first person to build a useful quantum computer.
The quantum world at the heart of Franson’s work was discovered in the early
decades of this century. Physicists found that the residents of this world are a
strange bunch. Tiny bits of matter such as electrons and photons can behave as
either waves or particles—depending on the way they are measured. They
also jump instantaneously between clearly defined energy states without passing
through any states in between. And they are harder to pin down than your boss at
pay rise time. Heisenberg’s famous uncertainty principle dictates that you can
know properties such as a particle’s energy or its position, but not both at
once.
In the macroscopic world, however, quantum weirdness plays little or no part,
and useful devices that depend on the antics of quantum particles are few and
far between. But in 1985, a physicist at Oxford University called David Deutsch
worked out how quantum particles could carry information and how a quantum
computer could use this information to carry out calculations. Since then other
researchers have fleshed out how such a machine could solve real problems. It
turns out that quantum computers can solve problems that today’s computers could
never crack. Building one would be a major advance.
A quantum computer performs its prodigious feats by taking advantage of
another strange characteristic of the Universe on the smallest scale. Quantum
particles can exist in two or more states at the same time. This is as strange
as a football being in two places at the same time. When this happens,
physicists say the states are superposed. Electrons, for example, can exist in
more than one energy level at the same time. And photons can exist in two
orientations of polarisation at the same time. Only when pressed by some kind of
outside interaction or measurement does the particle choose one state or the
other.
Some physicists even talk about a particle in such a superposition of states
being in two different universes. In each universe, the particle is in a
different state. When a measurement is made on the particle, this house of cards
collapses. One universe is destroyed while the other is spared.
In a quantum computer, these states represent the 0s and 1s of a digital
code. The revolutionary idea is that when a particle is in a superposition of
states, it can represent both a 0 and a 1 at the same time. This strange bit of
quantum information is known as a qubit.
Calculations are made by passing the qubit through a series of logic gates.
In one universe the calculation occurs as if the bit is a 0 and in another
universe as if it is a 1. With several qubits, a computer could carry out large
numbers of calculations in many different parallel universes. In fact, the
number of calculations that are possible rises exponentially with the number of
qubits.
Parallel universe
“People in regular computing talk about parallel processing, in which a
problem is broken up into pieces that are fed simultaneously to multiple
processors,” says John Dowling, a researcher in quantum optics at the US Army
Aviation and Missile Command at the Redstone Arsenal in Huntsville, Alabama. “In
quantum computing you do the same thing, but the other computers happen to be in
different universes. It sounds strange, but that’s what quantum mechanics is
±ôľ±°ě±đ.”
Physicists have even demonstrated the principles of quantum computing using
single ions and the spin of individual atomic nuclei as qubits. But making a
machine with enough qubits to do anything interesting is a different matter.
“People are confident they can make a two-qubit quantum computer, or maybe one
with 10 qubits,” says David DiVincenzo of IBM’s Watson Research Center near New
York. “But even the most optimistic researchers tend to get a little pale if you
ask for 50.”
So making a quantum computer that can do much more than count its fingers and
toes is extremely difficult. The problem is that particles in a superposition of
states are extraordinarily sensitive to outside influences. The slightest nudge
causes them to collapse back into a single state, ruining any quantum
calculation before it is complete.
Somehow qubits have to be protected from unwanted nudges. But this is
particularly difficult when the qubits have to pass through logic gates. An
electron passing through a material remains in a superposition of states for
less than a nanosecond, nowhere near enough time to carry out useful
calculations.
Photons have far more potential. For a start, it’s quite easy to use their
polarisation to represent the zeros and ones of binary code—a horizontally
polarised photon might represent a zero, for example, while a vertically
polarised photon represents a one. And photons are easy to control using optical
fibres. Most important of all, they are relatively undisturbed by the world
around them. In a vacuum, almost nothing bothers photons. Put them in an
transparent optical fibre and photons can remain coherent for a millisecond or
longer—more than a million times as long as electrons remain coherent.
But photons present difficulties as well. Because they never sit still,
photons are hard to store, although this could be overcome by passing them
through long fibres. But the biggest problem is that they do not easily interact
with each other. Without a strong interaction, logic gates simply do not work.
This is where Franson’s ideas come in.
One of the features of a logic gate is that its output depends on the bits
coming in. In a device known as a conditional-not gate a bit will be flipped
from a 0 to a 1, or vice versa, if the incoming bit is 1. Since all other types
of logic gates—and thus a quantum computer—can be built out of
combinations of conditional-not gates, they are the only kind that physicists
need worry about.
So if photon-based logic gates are ever to work, physicists must find a way
of twisting the polarisation of one photon by 90° if and only if another
photon that enters at the same time is, say, vertically polarised. What’s more,
a quantum computer that packs a decent punch will only be practical if these
gizmos are relatively simple, reliable and easy to link together.
The starting point for Franson’s theory is a well-known phenomenon called the
Kerr effect. Normally, two beams of light ignore one another when they cross.
But in certain crystalline substances called Kerr materials they can interact.
If one beam is bright enough, it changes the refractivity of the
material—how it bends light entering it—and this, in turn, changes
the polarisation of the second beam.
On the atomic scale, a photon from the first beam excites an atom and the
excited atom then twists the polarisation of a photon from the second beam. Of
course, the first beam of light must be intense enough to excite most of the
atoms most of the time in case any one of them is struck by a photon from the
second beam. This hit-and-miss affair makes the traditional Kerr effect
unsuitable for a quantum logic gate involving only two photons.
But Franson realised that a Kerr-like effect might be possible with only a
single pair of photons, and that under certain circumstances this effect could
become magnified until it was large enough to be measured. Once again, the idea
relies on quantum mechanics. First, by defining the wavelength of each photon
precisely, it becomes impossible to pin down its location. This is Heisenberg’s
uncertainty principle, and the result is that photons become “smeared”
throughout the substance they are travelling through. So when a particular
photon interacts with an atom in the substance, it is impossible to know which
atom the photon interacted with.
Now imagine two photons with slightly different wavelengths, A and B, passing
through a medium. Of the many different ways that these photons can interact
with matter, Franson is interested in only one. This is the case when one atom
in the medium absorbs the photon with wavelength A and emits one with wavelength
B while another atom some distance away absorbs B and emits A. The rules of
quantum mechanics dictate that this process of photon-swapping changes the
polarisation of the photons by a tiny amount. However, this change is normally
too small to measure.
But Franson has spotted a trick that can magnify the effect. It depends on
the fact that it is impossible to say just which pair of atoms have exchanged
the incoming photons. “Instead, you are left with a bunch of mutually
indistinguishable possibilities,” he says. “The laws of quantum mechanics
dictate that in this circumstance, each possible exchange contributes to the
overall effect.” In fact, the effect is magnified in proportion to the square of
the number atoms in the sample. And because even a small amount of matter
contains huge numbers of atoms, the overall twist put on the second photon
becomes big enough to be useful. This, at least, is the theory that Franson
published in Physical Review Letters last year (vol 78, p 3852).
Photon trap
For the moment, other physicists are reserving judgment. A year and a half
earlier, physicist Jeff Kimble and his colleagues at the California Institute of
Technology in Pasadena had used a different method to achieve something similar,
that is, to make one photon alter the state of another.
The heart of Kimble’s method is a trap for photons—a tiny cavity
between surfaces so reflective that the photons bounce back and forth about 100
000 times before escaping, greatly magnifying their interaction with the single
caesium atom that Kimble and his colleagues drop into the trap along with the
photons.
Kimble’s approach is extremely demanding. Working with individual photons and
atoms and with the world’s most highly polished mirrors isn’t easy. “Our
experiment is a technically daunting enterprise,” says Kimble, “and there is
still a frightful gulf between lab demonstration and any useful implementation.
Jim [Franson] has come up with a really clever way, at least in principle, to
avoid the complexity of needing to use just one atom and an optical trap. It
remains to be seen if he is right.”
With his theory published, Franson, has shifted his focus to the lab and
begun the hunt for experimental confirmation aided by Todd Pittman, a
postdoctoral fellow at the Applied Physics Laboratory. Another modest place,
their lab looks better suited to teaching an undergraduate optics course than
carrying out cutting edge quantum physics research. It is dominated by a light
table, a kind of oversized billiard table housing a forest of lenses and
mirrors. The entire setup is run by a 15-year-old Apple IIe computer tucked
against the wall. “It still works fine and it’s easy to program,” says Franson,
clearly a practical man.
Franson has chosen to pass his photons through sodium vapour housed in a
glass cell a couple of centimetres long. He and Pittman start with two photon
beams, each with a slightly different wavelength, but close to an excitation
frequency for sodium. First, they carefully filter out all but the right
wavelengths, circularly polarise the first beam and plane polarise the second,
then attenuate the beams until only a small number of each kind of photon is
passing through the cell. On the other side of the cell, Franson has detectors
for the two kinds of photons. Finally, he puts a polarising filter in front of
the detector for the second wavelength of photon.
Every now and then one of each kind of photon happens to pass through the
cell at the same moment. According to his theory, the polarisation of the second
kind of photon should become twisted when this happens. Franson looks for this
rotation by using the polarising filter in front of the second detector to block
any photons that have not been rotated.
The data he collects are statistical, counts of how many times two photons
reach the detectors at the same moment. Since all unaffected photons should be
blocked before they hit the second detector, there would be no events to count
if there were no rotation. But Franson has spotted a few photons coming through
and has worked out that this corresponds to a rotation of about 3°.
The results are encouraging although not yet conclusive. Franson still has to
show that the observed effect varies in the way his theory predicts. For
example, the effect should vary in a specific way with the density of sodium
atoms inside the cell. Franson is looking for this effect and says that he
should have more definitive results within six months.
Earlier this year, he presented his initial data at a NASA-sponsored
conference on quantum computing in Palm Springs, California. His colleagues were
sceptical, but less so than they were when he published his theory. “When Jim
first started talking about this a couple years ago, almost everybody said he
was nuts,” recalls Dowling. “People asked things like: `How come nobody has seen
this before? How come it’s not in the nonlinear optics textbooks?'”
All that has changed. Now physicists are waiting to see if he is right. “Jim
has a pretty good track record on other things. And if he’s right, it will help
all of us make the quantum computer we’re trying to make,” says Dowling.
Code cracker
Even if Franson is right, a useful quantum computer is still years away. But
Franson, in characteristic style, is already thinking about how to move beyond
his somewhat bulky glass sodium vapour cells to a more practical solid-state
device. “It might be possible to connect 100 000 of these things up in a
warehouse somewhere, if that’s what it took,” he says. “But our goal is to
switch to solid state crystals, which could be very small.” His idea is to use
optical fibres, or optical waveguides etched onto a silicon crystal, to connect
the logic gates together. The result would resemble a standard computer
chip.
One organisation that might be willing to back the construction of a
warehouse-sized computer is the National Security Agency, a secretive US
government organisation which already supports Franson’s work.
The NSA is interested in cracking codes, a problem that quantum computers
should be particularly good at. Many codes are based in the fact that it is easy
to multiply two large prime numbers together to get a much larger number, but
very difficult to start with the large number and find the two primes that
produced it. Quantum computers ought to be able to do this in a tiny fraction of
the time an ordinary computer might take.
And if Franson is right, quantum computers might not be that far off. For the
moment, physicists are curious to see if Franson comes up with the experimental
evidence he needs. “Will Jim’s device be a panacea enabling large-scale quantum
computing? Nobody can tell,” says Kimble. Dowling is more confident: “If he’s
right, I think he’ll win the Nobel prize, for sure.”
