Hearing on Beyond Silicon Computing
Testimony of
Dr. Ruzena Bajcsy, Assistant Director
Computer and Information Science and Engineering
National Science Foundation
Before the House Basic Research Subcommittee
September 12, 2000
Chairman Smith, Ranking Member Johnson, members of
the subcommittee, thank you for inviting me to testify
at this important hearing. I welcome this opportunity
to discuss exciting new directions in computer science
and engineering.
These new directions have the potential to revolutionize
every facet of our lives. President Clinton alluded
to these trends in his State of the Union address
when he spoke about the prospect of "molecular computers
the size of a tear drop with the power of today's
fastest supercomputers."
What had been dismissed as science fiction is now starting
to materialize. The example used by the President
comes from the work of Jim Heath at UCLA and Stan
Williams at Hewlett Packard. This fundamental work
- supported by both NSF and DARPA - aims to build
electronic circuits from the bottom-up, starting at
the molecular level.
The results of the research we are supporting now will
emerge in about 10 years as the pace of improvements
to silicon based computing slows down. As it slows,
we want the nation to be ready to continue to be the
leader in innovations in computing and communication.
That leadership will depend on advances in the new
technologies beyond silicon that we are discussing
today. Molecular and chemical devices, quantum computers
and optical computing and communications are the technologies
that we are exploring now in anticipation that one
or more will be the leadership technologies in ten
or twenty years.
The science and engineering behind these is some of
the most exciting and challenging work today. New
ideas, theories, processes and devices are being developed
unlike any seen before. As just one example, understanding
how the electrical properties of a single molecule
change when it is flexed by another electrical charge
is central to molecular computing; to understand the
science as well as the engineering of building such
devices we must draw on many scientific and engineering
disciplines.
The NSF is particularly well suited to support research
in these areas. NSF focuses on basic research. Most
of our support is for research at universities and
we have many ways to support university researchers
to work across disciplines, with industry researchers,
and with researchers in other nations. NSF is the
principal source of funds that support graduate training
in science and engineering; the young people we support
and train in these emerging technologies will become
the faculty and industry researchers that the nation
will rely on to advance and develop these new computing
methods. And, of course, NSF also funds facilities
that support the university research community and
enable cooperation with industry. NSF's Initiatives
in Information Technology Research and in Nanoscale
Science and Engineering provide mechanisms to address
the interdisciplinary research that these ideas require.
In my comments today, I would like to address five
general points. First is a brief overview of the science
and engineering challenges and accomplishments to
date. My colleagues at this hearing will be able to
address this in greater detail. Then I will speak
to four aspects of the Federal role in these exciting
new areas: how the NSF is supporting and coordinating
this research, multi-agency activities and coordination,
the relationship of federally supported research with
industry, and lastly, international activities in
these areas.
Silicon Chips and Moore's law;
Beyond Silicon's Limits
Since the invention of the silicon integrated circuit
in 1961 to the present, the number of devices that
can be place on a single silicon chip has roughly
doubled every 12 to 18 months. This means that every
ten years, the number of devices on chips increases
about a hundred-fold. This is done by shrinking device
sizes and is achieved by constant improvements in
chemistry, photolithography, clean rooms, and other
efforts. This doubling rate is known as Moore's law.
For the computing industry, the shrinking devices
and increasing density has enabled the information
technology revolution through staggering increases
in speed and functionality of computers accompanied
by astonishing decreases in costs.
We know that this cannot continue for long - the size
of atoms is a very hard limit and very close in time.
Even before that size is reached, small devices on
silicon chips (they are now as small as 200 nanometers
or about 1/50th the diameter of a human
hair) are facing limits due to electron leakage across
small numbers of atoms, uniformity of mixing dopants
in semiconductors, dissipating heat from increasing
numbers of devices on small chips, and other problems.
If we are to continue to see improvements in the performance
and cost of computing, we must go beyond silicon.
The technologies discussed today - chemical and molecular
devices and quantum computing explore these new frontiers
for computing through different approaches.
Chemical and Biomolecular Computing
Chemical principles, such as the switching gates developed
by Heath and Williams, are potentially smaller and
faster than silicon transistors and operate at lower
energy levels. At present, research is identifying
chemical substances with required electrical, mechanical,
and other properties. We anticipate that these devices
cannot be placed precisely and reliably on chips (as
is done with silicon), so research is also addressing
concepts such as self-assembly in which the shape
of molecules dictates that they will form themselves
in regular assemblages. We also support research on
massive fault-tolerance to find techniques for good
devices to assemble and operate as reliable functional
units even in the presence of many faulty units.
Biological principles are the basis for DNA computing.
Much as DNA sequences store information in the genome,
DNA can also store information that represents complex
scheduling problems, data-bases or other information.
Since 1994, researchers with NSF support have been
exploring methods to store and manipulate information
in DNA. These methods are based on DNA fragments mixing
in water solutions rather than ordered materials on
surfaces. These methods have potential for very high
degrees of parallelism with low material cost and
energy requirements. Research challenges include a
new and unfamiliar programming model, finding methods
to detect low concentration "answers" in solutions,
and characterizing the sorts of problems that could
be efficiently solved.
Quantum Computing
Quantum phenomena are the latest in the quest for new
principles to be used for computing purposes. The
mysteries of the quantum world have challenged many
in university classrooms, so this hearing room may
not be the place for an advanced physics lecture.
The quantum world has two mysterious phenomena that
we are exploring for computing and communication.
The first is the notion of quantum "state" as exhibited
by the spins of atomic particles. In silicon devices,
every bit is a "0" or a "1" - an "on" or "off." Atomic
particles have state also spinning clockwise or counterclockwise
- but until that spin is observed, the direction is
a probability of one direction versus the other. Thus
a particle can be in two states at once; these particles
are called qubits, short for quantum bits. Two qubits
can be in four states and 20 particles in a million.
A new field of quantum algorithms has demonstrated
that such devices can solve arithmetic problems (factoring
numbers) and search problems much faster than conventional
computers by exploiting these properties of devices
being in many states at once. In the steps that a
silicon computer uses to seek a single solution for
a complex problem, a quantum computer can potentially
explore all the solutions at once - if our research
shows us the methods to harness the power of these
quantum devices.
The second mystery of the quantum world is entangled
states; two particles can have linked spins even though
they are at a distance. Manipulating one particle
and then reading the spin of the other, linked, particle
is the basis of quantum information teleportation.
This has been demonstrated in laboratory conditions
and appears to be feasible way to securely distribute
cryptographic keys over tens of miles.
The research challenges of quantum computing are enormous.
To mention just a few general areas of research: new
types of algorithms are needed that utilize being
in multiple states, new devices that have coherent
spin states immune to environment hazards are being
invented, and many device forms - such as liquids
for nuclear magnetic resonance manipulation, ion traps,
quantum dots, etc. - are being considered.
My colleagues can expand on their work in these areas,
but I want to emphasize just a few points from this
overview:
- the potential of these chemical and quantum computing
devices is enormous;
- assuring that America is poised for leadership
when these new technologies are ready for development
is essential to our future,
- these areas have attracted some of best and brightest
scientists, engineers and graduate students in
the nation to work on these very exciting ideas,
and
- our support for this work must be stable, long
term, and flexible to assure that this research
will draw as needed on the traditional disciplines
even while it creates new ones.
NSF Support for Computing Beyond
Silicon
The NSF has many modes for support of this research.
I want to mention a few in particular. The Information
Technology Research Initiative, which will announce
its first awards tomorrow, has a "revolutionary computing"
component that will make several awards in these areas.
The initiative allows NSF to identify areas for research,
to encourage our supported scientists to build teams
that cross disciplinary and institutional boundaries,
and to make larger awards with extended durations
that are necessary for longer range research. A second
initiative, Nanoscale Science and Engineering will
also support activities for these areas, especially
the necessary inventions to build and assemble the
small devices that these new computing technologies
will require.
NSF's core funding in disciplinary research is another
important component of our support for these activities.
Program such as Grant Opportunities for Academic Liaison
with Industry (GOALI) provide opportunities for university
researchers to work with industry researchers and
vice versa. One example of GOALI support is the collaboration
between Jim Heath of UCLA and Stan Williams of Hewlett
Packard that led to the chemical gates already mentioned.
Awards for research in computer science, materials
science, physics, electrical engineering and other
disciplines are supporting researchers and graduate
students in these areas.
NSF Centers programs are another avenue for supporting
these efforts. A new center for NanoBioTechnology
at Cornell in New York and the Center for Synthesis,
Growth, and Analysis of Electronic Materials at the
University of Texas are two of the Science and Technology
Centers that are especially relevant to these efforts.
The Center for Discrete Mathematics and Theoretical
Computer Science at Rutgers in New Jersey has conducted
basic research on quantum and DNA computing algorithms.
Several of the Engineering Research Centers perform
supportive research for these new computing technologies,
including the Data Storage Systems Center at Carnegie
Mellon University in Pennsylvania, the Center for
Optoelectronic Computing Systems, at the University
of Colorado, and the Center for Compound Semiconductor
Microelectronics, at the University of Illinois.
The NSF has numerous, flexible mechanisms to capitalize
on these emerging research areas and to support projects
targeting individual researchers, groups of researchers,
graduate training, collaborations with industry.
US Federal Agencies Activities
I will briefly summarize the activities and coordination
of these activities across multiple agencies.
For Quantum Information Science (QIS) there is Coordinating
Oversight group that promotes a coherent national
program. This group consists of representatives of
all the federal sponsors and it meets twice a year.
Participating agencies are Army, Navy, Air Force,
Defense Advanced Research Projects Agency (DARPA),
National Security Agency, National Reconnaissance
Agency, National Science Foundation, NASA, National
Institute for Standards and Technology, and the Department
of Energy. The overall support level for QIS has risen
from about $1M in FY 1995 to over $30M in FY2000.
Approximately 66% of this investment is from mission
agencies and offices, with the remainder from the
NSF. Research thrusts include foundations of QIS,
quantum algorithms and software, quantum communication,
quantum computation, qubit implementation concepts,
applications to clock synchronization, imaging, and
other areas.
The Information Technology Research and Development
activity across multiple agencies is also coordinating
research in new technologies for high-performance
computing. I chair the ITR&D; coordinating committee
and NSF staff are closely involved in these coordination
efforts.
In DNA computing NSF is currently spending approximately
$2.3M for FY2000. We estimate NSF's overall bio-molecular
computing expenditure, to be about $5M. The optical
computing and communication expenditure is hard to
estimate since it is coupled with building the infrastructure,
which is very expensive (approximately $35M).
At NSF we have several programs across the Agency with
emphasis in the Computer and Information Sciences
and Engineering, Mathematics and Physical Sciences,
and Engineering directorates. Furthermore in our NSF
wide initiative programs such as the Information Technology
Research (ITR), Nanoscale Science and Engineering
(NSE) and Science and Technology centers (STC) Programs.
The role of the NSF in this national effort is central
and clear. As articulated in the proceedings of an
NSF-sponsored workshop entitled "Quantum Information
Science", the NSF has a mandate to provide stable
support to fundamental studies of the foundations
of QIS on university campuses, thereby ensuring the
constant flow of new ideas, people, and tools needed
to assume a leading role internationally in this and
other similar promising technologies.
The International Perspective
We have mentioned our concern with leadership for coming
decades. The European Community and Japan are most
prominent in their support for these areas. Their
investments are briefly summarized here:
Quantum Information Processing and Communication In
the European Union
The Quantum Information Processing and Communication
Initiative (QIPC) was launched by the EU in January
2000 with total funding of $15.6 million. The areas
of coverage in QIPC are:
- Cryptography and Communication
- Theory and Algorithms
- Implementation in Atomic Physics and Solid State
physics.
There is support for substantial theoretical work,
and the EU plans to have 5 working groups and a network
of excellence which will link researchers in multiple
universities. Overall the EU program is a long-term
(5-10 year) high-risk, forward looking basic research
program.
Quantum Computer R&D in
Japan:
In Japan, the investment in Quantum Computer R&D
will reach approximately $15M in 2000. The support
is distributed via the Japanese Science and Technology
Corporation (JST), Ministry of Posts and telecommunication
(MPT), the Ministry of Education, Science, Sports
and Culture (MONBUSHO), the Ministry of International
Trade and Industry (MITI) and companies such as NTT,NEC,
Hitachi, Toshiba and Mitsubishi.
Major projects supported by JST are in Functional Evolution
of Materials and Devices based on Electron/Photon
Related Phenomena; Quantum effects, and quantum entanglement.
The Photonic Information technologies are being supported
by MPT basic research laboratories. MONBUSHO supports
a number of university professors who are engaged
in theory of NMR quantum computation, quantum Turing
machines, and related topics.
MITI on the other hand is trying to realize qubits,
that is, establish a reliable system to identify quantum
bits for computing purposes. Quantum computing is
very fashionable now in Japan, and is considered ambitious
and futuristic. With state-of-art electronics technology
and expertise, plus building on their nations strengths
in nanostructure capabilities, Japan hopes to catch
the West and perhaps to make advances in Quantum Computer
developments in the near future.
Conclusion
A new and exciting science has emerged in this intersection
of Computer Science, Physics, Chemistry, Biology,
Mathematics, and Engineering. The motivation has largely
come from the realization that Moore's Law will reach
a dead-end in the next 20 years. The enabling concepts
have been created by a small number of visionaries
such as Richard Feynman, Charles Bennett, Jim Heath,
Leonard Adleman and Peter Shor.
This is a high-risk, high-payoff field, and many years
of basic research into new hardware and software technologies
will be needed to unlock the potential of this science
and technology. Quantum, chemical and DNA computing
are all radically different approaches to information
science and technology. They offer the possibility
of new paradigms in computation and data processing,
data storage and transmission, cryptography and information
security, as well as new quantum-based technologies.
We do not know today with any certainty when and how
these new computing methods will come to fruition.
This is an exquisite example of the importance of
Federal support of fundamental research at the intellectual
frontier. While we earnestly seek the benefits of
the computing capabilities that these advances would
unleash, it is important for the U.S. to maintain
a leadership position in each of these areas.
Federal Agencies across government coordinate their
activities to provide coherent management of this
emerging field, so that each agency's needs are met
without duplication. It is rather difficult to compare
who is spending how much on this research because
comparisons of investments by Japan and Europe for
research on quantum computing and/or photonic computing
are not necessarily equivalent. Neither the EU nor
Japan are accounting for DNA or biomolecular computing.
The NSF support modes have been very important to initiating
this research and its subsequent growth. I have mentioned
the NSF initiatives and the Information Technology
Research program that is announcing its first year's
awards tomorrow with the $90 million that kicked off
the program. I would like to mention a few of the
awards in Revolutionary Computing that will be announced
this week.
One example in the ITR initiative is to the California
Institute of Technology. This five year award with
anticipated funding of $5.0 million will support the
establishment of the Institute for Quantum Information.
Another award to the University of Kansas will support
research on fast superconducting qubit and qugate
devices for quantum computing. A third award to Duke
University will support research on self-assembly
of DNA nanoscale structures for computation. These
are just a few of 11 awards that NSF will announce
in the revolutionary computing area under the ITR
competition.
Mr. Chairman, let me conclude by thanking you for providing
an opportunity to highlight these emerging and exciting
fields of research, and I would be pleased to respond
to any questions that you might have.
|
