Gavin writes:
> > Perhaps some on this forum would know.
>
> Bruce and perhaps Wirt probably know more about it than I do.
Unfortunately, like Will Rogers, I only know what I read in the
newspapers.
In that regard, the following article appeared in yesterday's NY Times:
===================================================
March 27, 2001
Computing, One Atom at a Time
By GEORGE JOHNSON
OS ALAMOS, N.M. — The only hint that anything extraordinary is happening
inside the brown stucco building at Los Alamos National Laboratory is a
small metal sign posted in front: "Warning! Magnetic Field in Use.
Remain on Sidewalk." Come much closer and you risk having the magnetic
stripes on your credit cards erased.
The powerful field is emanating from the supercooled superconducting
magnets inside a tanklike machine called a nuclear magnetic resonance
spectrometer.
The device itself is unremarkable. N.M.R. machines are used in chemistry
labs across the world to map the architecture of molecules by sensing
how their atoms dance to the beat of electromagnetic waves. Hospitals
and clinics use the same technology, called magnetic resonance imaging,
or M.R.I., to scan the tissues of the human body.
The machine at Los Alamos has been enlisted on a recent morning for a
grander purpose: to carry out an experiment in quantum computing. By
using radio waves to manipulate atoms like so many quantum abacus beads,
the Los Alamos scientists will coax a molecule called crotonic acid into
executing a simple computer program.
Last year they set a record, carrying out a calculation involving seven
atoms. This year they are shooting for 10. That may not sound like many.
Each atom can be thought of as a little switch, a register that holds a
1 or a 0, and the latest Pentium chip contains 42 million such devices.
But the paradoxical laws of quantum mechanics confer a powerful
advantage: a single atom can do two calculations at once. Two atoms can
do four, three atoms can do eight.
By the time you reach 10, doubling and doubling and doubling along the
way, you have an invisibly tiny computer that can carry out 1,024 (210)
calculations at the same time.
If scientists can find ways to leverage this achievement to embrace 20
atoms, they will be able to execute a million simultaneous calculations.
Double that again to 40 atoms, and 10 trillion calculations can be done
in tandem.
The goal, still but a distant glimmer, is to harness thousands of atoms,
resulting in a machine so powerful that it would easily break codes now
considered impenetrable and solve other problems that are impossible for
even the fastest supercomputer.
"We are at the border of a new territory," said Dr. Raymond Laflamme,
one of the leaders of the Los Alamos project. "All the experiments today
are a very small step, but they show that there is not a wall."
"The big question," he added, "is whether we can make the transition
from theory to practice."
The program that he and a colleague, Dr. Emanuel Knill, are now running
— a procedure for detecting and correcting the errors that inevitably
crop up during the exceedingly delicate quantum calculations — is being
watched with interest by other theorists.
"Quantum error correction is vitally important for future quantum
technologies," said Dr. John Preskill, a physicist at the California
Institute of Technology. "Until the idea of quantum error correction was
discovered in 1995, there was great skepticism about whether large-
scale quantum computers capable of outperforming conventional digital
computers would ever be practical."
Now Dr. Knill and Dr. Laflamme are demonstrating that what was shown to
be true in theory works in practice as well. Their experiment is also a
landmark in another way.
Researchers have recently used N.M.R. to get molecules to execute
rudimentary programs, like searching a database using fewer steps than
required by an ordinary computer. (As a sign of how primitive the
technology remains, the database consisted of a list of only eight
numbers.) Dr. Knill and Dr. Laflamme's error-correcting algorithm is
still quite simple, compared with, say, Microsoft Word, but it is one of
the most complex pieces of quantum software yet run.
Less than a decade ago, quantum computing was just an intellectual
parlor game, a way for theorists to test their mettle by imagining
absurdly small computers with parts the size of individual atoms. At its
root, computation is just a matter of shuffling bits, the 1's and 0's of
binary arithmetic. So suppose an atom pointing up means 1 and an atom
pointing down means 0. Flip around these bits by zapping the atoms with
laser beams or radio waves and the result is an extremely tiny computer.
But that would be just the beginning of its power. Quantum mechanics,
the rules governing subatomic particles, dictates that these quantum
bits, called qubits (pronounced KYEW-bits), can also be in a
"superposition," indicating 1 and 0 at the same time. Two atoms can
simultaneously be in four states: 00, 01, 10 and 11. Three atoms can say
eight things at once: 000, 001, 010, 011, 100, 101, 110 and 111. For
each atom added to the chain, the number of possibilities increases
exponentially, by a power of 2. Put together a few dozen atoms, it
seemed, and they could perform vast numbers of calculations
simultaneously.
All this was of little more than academic interest until 1994, when Dr.
Peter Shor, a researcher at AT&T Laboratories in Florham Park, N.J.,
proved that a quantum computer could rapidly find the factors of long
numbers, a problem that flummoxes human brains and supercomputers. Since
the codes that are used to protect military and financial secrets depend
on the near impossibility of this task, government money began pouring
into places like Los Alamos, allowing theorists like Dr. Laflamme and
Dr. Knill to begin turning the thought experiments into reality.
"There is no fundamental physical barrier that makes quantum computing
impossible," Dr. Knill said. "The technology, as it exists, is a long
way from meeting the goal. But we see no reason in principle why the
goal cannot be met."
For the past few years, laboratories have been using exotic technologies
to isolate small numbers of atoms, prodding them into performing simple
calculations. Dr. Laflamme and Dr. Knill's group is among those that
have been trying a different method: using the off-the- shelf technology
of N.M.R., in which molecules — strings of atoms — are trapped in
intense magnetic fields and manipulated with radio waves.
This approach is possible because the cores of some atoms — the nuclei —
are endowed with a quality called spin. They act like little tops,
rotating in the presence of a magnetic field. If the nucleus is rotating
counterclockwise, its axis of spin points upward. Flip it over and it
rotates clockwise, a condition called downward spin.
Nudging these nuclei with pulses of high-frequency radio waves causes
them to shift between the two positions, up and down. And since the
molecules emit feeble electromagnetic signals, the progress of the
experiment can be monitored on a computer screen.
This technique is a proven tool for chemists, who use N.M.R. to generate
charts called spectra that give clues to the structure of chemical
compounds. Several years ago, scientists at Stanford, the Massachusetts
Institute of Technology, the I.B.M. Almaden Research Center, the
University of Oxford and elsewhere realized that N.M.R. could be used
for a very different purpose. Call up "1" and down "0" and you have a
tiny molecular switch.
As Dr. Laflamme put it: "People had been doing quantum computing all
along. They just didn't know it."
During the recent experiment, he and Dr. Knill sat in front of a
computer workstation that was wired and programmed to control the N.M.R.
machine. Their goal: to get a string of five nuclei to carry out their
error- correcting algorithm.
Errors occur when a bit is accidentally flipped so that it says 1 when
it really means 0, or vice versa. Ordinary computers can protect against
this by using redundancy. In one scheme, data are sent in triplicate, so
101 becomes 111000111. Simple little programs watch out for corrupted
triplets like 010 or 110, restoring the errant bits so they match the
other two.
For quantum computing, error correction is more convoluted. A qubit is
protected by using an intricate scheme that effectively spreads its
value across a cluster of five qubits that are all "entangled" quantum
mechanically. That means that if one of the qubits becomes scrambled,
its original value can be retrieved by analyzing the other four.
In the experiment, the five qubits will be represented by the nuclei of
five atoms in a molecule of crotonic acid. Schematically this can be
thought of as a string of five beads, though the arrangement is slightly
more complex. Four of the beads are carbon nuclei — or actually isotopes
called carbon 13. (Since ordinary carbon 12 is spinless, the physicists
called on a Los Alamos chemist, Dr. Rudy Martinez, to synthesize
crotonic acid using carbon 13.) The other bead on the string actually
consists of a cluster of three hydrogen nuclei, part of a structure
called a methyl group, that is treated as a single processor.
In the scientists' notebooks, the five-qubit sequence is abbreviated
like this: M C1 C2 C3 C4 — a methyl followed by four carbons. These are
the tokens that will be used to compute. As the calculation unfolds, the
tiny signals emitted by the molecules will be monitored and displayed by
the computer as a horizontal zigzag line.
First a small flask of crotonic acid containing about 1021 (a billion
trillion) of the molecules is placed inside the core of the N.M.R.
spectrometer. A bath of liquid nitrogen and liquid helium cools the
machine's superconducting coils, allowing electricity to course through
them unimpeded, generating the intense magnetic field that the sign
outside warns about.
Tapping on the keyboard, Dr. Knill "shims" the magnets — straightening
out the kinks in the electromagnetic field. The effect is not unlike
what happens when a carpenter uses little wooden wedges to shim a window
frame so it is perfectly horizontal.
After the machine is calibrated the experiment can begin. At first, the
nuclei in the molecules are pointing every which way, creating a
predominantly random soup. But the strong magnetic field causes a
fraction of the molecules, about one in a 100,000,000, to line up so
that all their nuclei are pointing up: 11111.
This subset of uniformly aligned molecules — about 10 trillion of them —
will be used to carry out the computation. This is possible because the
five nuclei within the molecules each resonate at a different frequency.
Using pulses of radio waves, an operator sitting at the controls of an
N.M.R. machine can choose an individual nucleus — carbon number 2, for
example — and strike it like a bell. Throughout the flask, trillions of
C2 nuclei will chime in synchrony.
Apply the pulse for just the right duration and the C2's can be rotated
to point down for 0. Another pulse will cause them to point to 1 again.
And a pulse of half the duration will cause them to hover in quantum
superposition, potentially saying 1 and 0 at the same time.
What has been described so far is just the quantum version of a light
switch. The reason a molecule can be used to calculate is that its
nuclei, like the tiny switches inside a computer chip, interact with one
another: a radio pulse will cause a certain nucleus to change from 1 to
0 — but only if the nucleus to its left is 1. In an ordinary computer
these kinds of arrangements are called logic gates, the building blocks
of computation. String enough of them together and any calculation can
be performed.
Dr. Knill starts the quantum algorithm by directing the machine to emit
a short burst of pulses. This causes the first carbon nucleus, C1, to
point down while the other four qubits remain up. Throughout the soup,
trillions of molecules now say 10111, a pattern that can be displayed on
the computer screen as a horizontal line with a peak — the spectrum. A
single qubit of quantum information has been stored.
Another series of pulses then protects this information by encoding it
among the cluster of qubits, flipping them up or down according to the
rules of the correction algorithm.
Then an error is deliberately introduced. A radio pulse is used to flip
one of the qubits — just as might accidentally happen during a
computation. But the original message has not been lost. A final barrage
of pulses is unleashed, which analyzes the cluster, ferrets out the
error and fixes the mistake. On the computer screen the peak of the
spectrum shifts, showing that the experiment has been a success.
There are no whoops of joy or uncorking of Champagne. Since last fall,
the experiment has been repeated many times. But it is always satisfying
for the theorists to see that it really works.
===================================================
Wirt Atmar
|
