With transistors set to reach their smallest possible size in the next two decades, CITRIS researchers are among those searching for the next big computing substrate.
by Jenn Shreve
…………………………………………..
With transistors set to reach their smallest possible size in the next two
decades, the silicon chip is likely to change dramatically. CITRIS
researchers are among a worldwide body of engineers, physicists, and
chemists racing to find which new computing substrate will allow the
industry to continue on its current path.
That path has been largely shaped by Gordon Moore’s famous prediction that the
number of transistors on a silicon chip would double approximately
every eighteen months. “Moore’s Law” has served as a brass ring for
the semiconductor industry, moving innovation at a steady clip for
decades. Indeed, it is the basis for the International Technology
Roadmap for Semiconductors, which the industry uses to lay out its
plans for the next fifteen years. But, as Jeffrey Bokor, an EECS
professor at UC Berkeley, explains, major roadblocks are just around
the corner.
UC Berkeley EECS Professor Jeffrey Bokor.
“The current 2005 version of the Roadmap has as the final node on there
transistors that are as small as six nanometers,” Bokor explains. “The
new Roadmap is going to come out soon, and it may push a little
further. Who knows? There may be a four-nanometer transistor on there.
But everybody knows they are not going to get much smaller than that.
That is getting pretty close to the absolute limit for standard
conventional transistors as we know them now.”
Make transistors much smaller than that on a silicon chip, and you run up
against numerous difficulties. For example, shrinking the distance
between the two output terminals in a silicon transistor makes it
harder to switch the transistor off, an operation that is crucial to
the operation of logic circuits. Power is another problem. Just because
you can add more transistors does not mean you can scale the amount of
power each transistor uses. “Putting more and more transistors onto the
chips just causes these chips to use up more and more power,” says
Bokor.
An emerging field called spintronics is addressing the power issue by exploring ways to use electron spin
rather than electric charge to process information. Spintronics is the
main focus of the newly formed Western Institute of Nanoelectronics, of
which Bokor is a leader. At this time, silicon is one of several
materials being considered for manipulating electronic spin. “In the
WIN center we have many investigators looking at all sorts of
materials,” Bokor says.
Spintronics is also relevant to the field of quantum computing, which operates on a molecular level.
Classical computing relies on bits, which can be in a state of 0 or 1.
In quantum computing, bits are replaced by quantum bits, or qubits,
which can be in both states at once. That superposition means more
complex processing can occur at much higher speeds. “You can do many
operations at once by operating on a quantum state rather than a
classical state,” explains Birgitta Whaley, a professor of chemistry at
UC Berkeley and co-director of the Berkeley Quantum Information &
Computation (BQIC). That could eventually lead to major breakthroughs
in cryptography, search engines, and modeling of complex biological and
physical systems.
This diagram depicts a system of qubits, the building blocks of quantum computers. Image Courtesy of KB Whaley.
But quantum states are also highly sensitive to interference and readily
revert to classical systems, a reaction called decoherence. Some
isotopes in silicon have nuclear spin, which interferes with electron
spin, causing decoherence to occur, making silicon a less than ideal
substrate for this kind of computing.
If not silicon, then what? Right now, that is anyone’s guess. Whaley says her group is
basing its quantum computing models on a variety of promising
materials, including gallium arsenide and superconductors. She has seen
proposals that use lasers to manipulate ions in the gas phase, and
others “using atoms trapped in potential fields created by light.” She
says, “I don’t think it is clear right now which material is going to
be best. My view is we need to continue to investigate many different
proposals for these different ideas.”
Of course, one need not reach quantum states to see that better materials are
required. “There is already a clear recognition that the transistors
that are made down there at the end of the Roadmap certainly will not
be pure silicon. There are already materials that allow you to make a
better transistor, such as carbon nanotubes,” says Bokor.
Tens of thousands times thinner than a human hair, yet remarkably strong and
flexible, these thin cylinders of carbon could form the basis for
future nanoelectronics. “The advantage of carbon nanotubes is that they
can be formed as perfect single molecules of pure carbon, so electrons
can move along them in a perfect one-dimensional line without being
jostled side-to-side as they are when they move in silicon. This makes
it possible to build faster transistors,” Bokor explains.
None of this is to say that silicon has been ruled out entirely. Bokor’s
group has already successfully incorporated carbon nanotubes into a
functioning integrated silicon circuit. Both Bokor and Whaley are
collaborating with Thomas Schenkel’s group at Lawrence Berkeley
Laboratory to implement quantum computing in silicon. In addition,
Bokor is working with Schenkel on implementing classical spintronics in
silicon.
Both Bokor and Whaley point out that there are compelling reasons to keep silicon around. “There is a huge
industry built up that understand how to work with silicon. So the more
you can take advantage of that infrastructure, the better off you are,”
says Bokor. Industry giants like IBM, HP, and Intel are not taking any
chances; all have announced programs to explore alternative computing
substrates.
Whatever the integrated circuit look like, a lot of the research going towards building it will take place in
CITRIS’s NanoLab Center, which is currently under construction—from
housing research equipment to enabling researchers to manufacture and
try out new materials and approaches.
For more information:
International Technology Roadmap for Semiconductors
Berkeley Quantum Information & Computation Center
Giant leap in small technology: Milestone in the field of nanoelectronics by David Pescovitz (Forefront, Spring, 2004)
Jeffrey Bokor Home Page
Whaley Group Home Page
Exploration and Control of Condensed Matter Qubits (CITRIS Research Projects)
Focus Center in Materials, Structures, and Devices (CITRIS Research Projects)
Making Quantum Computing Work in Silicon, by Paul Preuss (Science at Berkeley Lab, May 2006)
A Nano-Scale Lab with Societal-Scale Impact by Jenn Shreve (CITRIS Newsletter, August 2006)
Beyond Silicon: Intel is exploring different materials for computer chips. by Kate Green (Technology Review , March 15, 2006
IBM researchers look beyond silicon technology (PhysOrg.com, August 3, 2006)
Beyond Silicon: HP Outlines Comprehensive Strategy for Molecular-scale Electronics (PhysOrg.com, March 14, 2005)