December 08, 2000, 3:17 PM — HearIT, Molecular computing; Jack Vaughan (2:57;
RA,
WM)
NEWS ANALYSIS -- IBM scientists recently furthered the cause of
quantum computing, announcing they had solved an order-finding mathematical problem in
a single cycle using fluorine atoms -- instead of the usual silicon gates -- as the
computing elements. The achievement may be the best evidence to date that an
architecture based on atoms and nuclei can solve multi-step problems that overwhelm
even the most powerful of traditional computers.
Although the researchers advise that the experiment was modest, these and other
advances suggest to some that smaller and abler architectures may arise in the future
to help the computer industry push on to new levels of processing power.
For years, concerns have grown about the limits of semiconductor electronics as the
limits of Gordon Moore's Law have been approached.
Intel founder Moore said that the number of transistors the industry could place on a
silicon chip would double every year or two. The conventional means of pushing hordes
of electrons around in order to do calculations has worked, of course, and smaller and
smaller chip dice have meant larger memory and faster processing.
The most immediate obstacle to further miniaturization is that current chip lithography
techniques are nearing their ultimate resolution. To go smaller, chip makers will have
to employ new x-ray fabrication techniques, which will be quite expensive to implement.
But even with better chip fab technology, experts see an eventual breakdown in that
trend as the size of silicon logic gates shrinks down to the size of atoms.
Qubit power
While pursuing molecular computing research, IBM and other researchers decided to
explore a somewhat non-Boolean approach that is based on the complex states of quantum
matter.
The team included scientists from IBM's Almaden Research Center, Stanford University,
and the University of Calgary. The group fashioned fluorine atoms as qubits, computing
units that exploit the quantum physical properties of matter to represent a spectrum of
states, not just Boolean 0's and 1's as in conventional digital computing.
Isaac "Ike" Chuang, the IBM research staff member who led the team, said the first
applications for quantum computing will probably be on coprocessors for specific
functions such as database lookup. He also sees the technology addressing mathematical
problems such as the Traveling Salesman problem that tries to compute the best route
between many locations. Those problems can overcome conventional computers.
"With quantum computing, you can do the problem in linear time. That could mean a
difference in [computing] time between millions of years and a few weeks," he
said.
The era of quantum computing will begin in about 2020, when "circuit features are
predicted to be the size of atoms and molecules," Chuang projected.
He noted that accelerating word processing or Web surfing would not be well suited to a
quantum computer's capabilities.
Obstacles seen
The complex lab experiment, not reproducible in the usual corporate environment,
entailed the building of five fluorine atoms within a molecule so that the fluorine
nuclei's spins could interact to effect calculations. The atoms were programmed by
radio frequency pulses, and results were detected by nuclear magnetic resonance
instruments, which according to Chuang, are "similar to those commonly found in
hospitals and chemistry labs."
Chuang said the obstacles to commercialization are "huge." At the present time, quantum
computing requires "a lot of expensive equipment," he said. "The apparatus fills half a
lab." Moreover, only five qubits were operative in the experiment. Many more are
required to really tackle tough tasks.
There are important potential benefits to quantum computing, said Linley Gwennap, a
microprocessor industry analyst with the Linley Group of Mountain View, Calif.
"Silicon-based technology is going to taper off and not be able to continue increasing
in performance," he said. "Theoretically, these quantum computers should be able to
operate much faster than conventional transistor-based computers."
Scaling to mass production scale is likely to be the biggest hurdle for commercial
quantum computing. "Right now they are literally dragging atoms around to create
quantum structures, then using very expensive quantum microscopes to observe the
behavior and extract the information," Gwennap said.
