by Thomas Ulrich
For more than a half century, the computer revolution has turned on this
simple truth: The shorter the distance between two circuits, the faster the
computation speed. Engineers built computers with vacuum tubes, discrete transistors
and then integrated circuits — those powerful assemblies of silicon transistors
at the heart of modern, high-speed computer design.
But as the performance of the silicon-based integrated circuits approaches
its fundamental physical and financial limits, researchers in HP Labs have
turned to molecular-scale electronics in a quest to create faster, smaller
and less expensive integrated circuits. (Read more about
how the team is extending the life of Moore's Law.)
In a quantum leap for integrated circuit design, Senior HP Fellow Stan
Williams
and his team in the Quantum Science Research
group have harnessed a chemical
reaction that shuttles electrons back and forth between atoms — permitting
molecular-size wire junctions to behave like electronic switches, storing data
or performing logical functions. As a result, researchers can fabricate these
molecular-scale switches more easily and pack them much more densely than conventional
silicon ICs.
 |
|
Packing more devices onto the same-size chip is not the
only advantage to building an integrated circuit with molecular
junctions. Because individual molecules measure less than 10 nanometers
(one nanometer is about the size of five atoms in a row), their
electrical properties are governed by quantum physics, instead
of the classical mechanics that dominate our world. In their nanoscale
world, the physical properties of matter, such as melting point,
color and electrical resistance abide by the laws of quantum mechanics.
Since 1995, Williams and his team have searched for ways to
improve the performance of electronic devices through quantum
physics. The team is a unique one, comprised of scientists and
engineers from all over the world and in such varied disciplines
as chemistry, physics, computer science, materials science and
electrical engineering.
"Our most interesting discoveries come from groups of researchers who combine
a deep understanding of several fields to solve a problem that no scientist alone
could solve," says Williams.
Instead of a single scientific breakthrough, the Quantum Science
Research group has experienced a series of eureka moments.
Phil Kuekes
stood at the entrance to Stan Williams' cubicle for
the first time in 1996. He had led the team that built Teramac,
a defect-tolerant computer that performed a trillion operations
per second. A computer architect who has built several generations
of world-class machines, he is driven to find new ways to boost
computer capacity.
"Everything I knew about computer chips was that they were absolutely perfect,
and if they weren't perfect, they were thrown away," Williams says. "Then
all of a sudden this computer architect shows up and starts telling me about
this crazy machine that he built, and by the way, it has 220,000 broken pieces
and it still works perfectly."
"Every chip contains some defects," Kuekes admitted to Williams that
day. "But, we can correct manufacturing defects by testing for errors, then
downloading software to redirect the flow of data."
Kuekes' discovery convinced Williams that he could create electronic
circuits from a world dominated by quantum phenomena. "Phil
Kuekes gave us permission to think seriously about building circuits
with nanoscale devices, even though the second law of thermodynamics
tells us that it is futile to build a perfect machine with molecular-scale
components."
While scientists could visualize the atomic structure of surfaces
of molecules some 20 years ago, it was not until Stan Williams and
Warren Robinett from the University of North Carolina connected a virtual-reality
system
with a force-feedback haptic interface
to a scanning tunneling microscope that scientists could "feel" individual
atoms.
Scientists from HP Labs now routinely use a scanning tunneling microscope
and an atomic-force microscope to evaluate their nanoscale materials
and devices. They can examine the atomic-scale structure of the surfaces
and molecular films that are used in the fabrication of their devices
to better understand and optimize their function.
But before engineers could build
a complete nano-circuit, Williams and his colleagues had to develop
entirely new fabrication techniques.
They developed and adopted an advanced system of manufacturing called nano-imprint
lithography -- essentially a printing method that allows inexpensive fabrication of nanoscale features across an entire wafer.
Using this technique, the team was able to fabricate circuits based on HP's
patented cross-bar architecture, a checkerboard pattern of horizontal and vertical
wires. In 2002, researchers successfully trapped molecules between platinum
and titanium wires arranged in an eight-by-eight array within a one square
micron area. Each junction performed like an electronic switch — a first
step toward creating a working logic or memory device.
More recently, scientists "grew" nanoscale wires between two electrodes
by exposing a catalytically active particle to gas containing silicon atoms.
(Read
related paper.)
"It is difficult to make contact between the wire and the electrodes
at such a scale," says Ted
Kamins, a Distinguished Scientist in the Quantum Science
Research group. "So we took the opposite approach: Building the electrodes
first and then growing the nanowires between them."

Lateral epitaxial nano-bridges grown across trenches connecting
to opposing sidewalls of silicon.
This self-assembling silicon nano-bridge may allow researchers to integrate
a variety of sensors into conventional circuitry. Integrating the sensors
with the computer makes sense because of the computing power needed to
interpret all the information that sophisticated sensors can deliver.
Sensors
could, for example, be used to detect toxic gases or to determine the mixture
of fuel and air in a carburetor. Another team within the Quantum Science
Research group has used a different technique to build a nanowire sensor that
can detect complementary fragments of DNA.
The next step is to combine both
techniques to create sensors that can detect minute concentrations of biological
and chemical materials.
"We have achieved an important goal by looking
over the horizon," Williams says. "With
these prototypes, designers can begin to integrate molecular scale
concepts into products."
Although HP will not deliver a nanoscale product immediately, product
groups can begin to use the fabrication techniques Williams’ team
has developed to manufacture micron-scale components less expensively.
 |
Changing
the rules |
 |
 |
 |
Stan Williams and his colleagues
at HP Labs are working to extend the life
of Moore's Law, a driving force in integrated circuit design first described
by Intel co-founder Gordon Moore in 1965.
Moore's observation states that
the number of transistors fabricated on an integrated circuit doubles every
18 to 24 months. Solid-state microelectronics has followed that performance
curve for more than 40 years, but as conventional transistors approach
30 nanometers, they will become extremely expensive to produce.
Scientists from the Quantum
Science Research group have developed molecular junctions and a process for
fabricating them that are as simple and inexpensive as contact printing.
Using
nano-imprint lithography and their technique for depositing molecular monolayers,
scientists from HP Labs hope they will be able to fabricate
and demonstrate crossbar circuits with wire-to-wire separation of less than
30 nanometers. That would yield a switch packing density approximately 120
times greater than present-day integrated circuits.
"I think we've picked a winner — something
that will allow Moore's Law to continue for another half century," Williams
says. "I used to think that was impossible. Now I think it is inevitable."
|
 |
 |
 |
|