Introduction
The size of a single transistor
has been reducing in an exponential manner for several decades,
leading to integrated circuits containing tens of millions of
transistors. But as the size of the transistor decreases a physical
limit is encountered where the transistor becomes too small and
quantum effects become significant. When this limit is reached
the exponential growth in computing power that has been characteristic
of the 1980s and 1990s will come to an end. This event is expected
to occur somewhere between 2010 and 2020. This will be the end
of the road for pure silicon technology. At this point completely
new technologies will be needed. So the question is, what is
coming next?
The field of Nano, Bio and
Quantum is concerned with the future, with technologies that
do not yet exist. The field at present is therefore one of science,
but interestingly, it is technology that is enabling this science
of the very small. The technological tools now exist that provide
a means of working at the level beyond micro. These tools are
allowing scientists to interact with and to manipulate matter
at a level that has not been previously possible.
At the technological level
however, there is an increasing realisation that there is not
much to be gained in a practical sense by manipulating single
atoms or molecules. Therefore novel concepts are needed such
as self fabrication and self assembly, which raise problems both
at the physics level but also at the level of architecture, computer
science and systems engineering.
The Challenges
of Working beyond the Small
What are the challenges of
miniaturisation beyond the level of micro technologies? One issue
is when to stop investing in the current micro technologies,
which can be taken down to the nano scale. Miniaturisation of
microelectronics will probably be able to take the size of a
transistor down to 30 nanometres. The next issue would be how
to take the transistor size down to 5 nanometres, for which entirely
new concepts would be needed.
The success of microelectronics
has been in its scalability. As things became smaller the principles
and underlying physics did not change. New steps in the technology
built on previous knowledge and the financing of development
came from the profits from sales of the previous generation of
the technology. Technology development was non-disruptive.
The problem with nanotechnologies
is that they are not completely scalable. Also there is no miniaturisation
perspective. In microelectronics the goal has been to make things
smaller, faster and cheaper. In the nano regime one is at the
level of single electron transistors, single atoms, or single
molecules. So adding value becomes important and that will mean
making things more intelligent.
Interfacing on the nano scale
is going to be a big problem. At this level as soon as an interface
is made to a nano object, its properties change.
The new concepts in the field
of nanotechnology will come from nature. The nano scale operates
at the transition point between condensed matter and molecular
behaviour and quantum effects. Ideas will therefore come from
the fields of biology, chemistry and quantum physics.
The Contribution
of Quantum Physics
One of the new concepts that
may have a role to play in the post microelectronics world is
quantum computation, which is about using the states of atoms
as a basis for computation. Unlike classical logical devices,
which only exist in two states (0 or 1), atoms can have three
states (0 or 1 or 01 where the latter is a superposition of the
first two states).
There are several different
reasons for the interest in quantum computation. Technologists
are interested because they want to address the problems that
arise when the physical limits of silicon are reached. Physicists
on the other hand, want to understand more about quantum mechanics.
Computer scientists however are interested in changes to complexity
classes, where problems that are today seen as difficult may
become easy to solve. Logicians are also interested in this area
because it opens up the possibility of proving logical propositions
through a physical process.
It is very difficult to build
a quantum computer. Single quantum bit (qubit) logic gates are
possible. Two qubit gates (a controlled NOT logic gate) are currently
being investigated. Building single qubit gates using trapped
ion technology is feasible, but the big problem and major difficulty
is creating a 2 qubit gate. The difficulty lies in controlling
the state of a target atom based upon the state of the control
atom, and keeping the system stable over time. Most of the ideas
that have gone into quantum computing have been focused on entangling
two atoms to create a 2 qubit gate. The successful development
of a 2 qubit gate is now very near and it is to be expected that
within a few years 10 qubit gates will have been developed.
A number of competing technologies
are being considered including trapped ions, superconductors
and quantum dots.
Another issue is how to construct
quantum memory. Quantum memory exists and is workable in laboratory
environments. This involves storing and trapping single atoms
in space.
Part of the problems in building
quantum computers lies in the area of engineering the components.
A major difficulty lies in controlling the quantum effects. There
is knowledge about how to control quantum optics in a laboratory
environment, but in the long term the emerging technology may
be solid state physics. At the moment however this field needs
a lot of development work.
Another element of quantum
computation is quantum communications, which involves transmitting
quantum states in a reliable way. One way to do this is to use
photons, and to write the quantum state of an atom onto photons
and then to transmit these. At the receiver these photons are
then used to write the state back onto atoms. The technologies
for building this kind of quantum communication system, including
error correction, will probably be available in a few years time.
The Contribution
of Biology
Moving on from the area of
quantum physics, another discipline that may have a role to play
in shaping future generations of computer technologies, is biology.
Can chemical processes at the bio cell and molecular level be
used for information processing? Can knowledge of biological
molecules help with the development of new information technologies?
These are very much, open questions.
Life has a big advantage in
terms of technology. There is 3.5 billion years of evolutionary
development. DNA provides an example of long term information
storage. It is very compact and replicable, however it is not
very fast. So its use as a model for information processing seems
to be limited. Short-term information storage is handled in biological
systems by energy consuming processes. Brain activity is, for
example, linked to increased energy consumption, but again the
time scales are very slow when compared with microelectronics.
However, in photosynthesis,
photon inputs result in virtually instantaneous (pico second
range) charge separation which then drives the energy making
process in the system. This is an example where biological systems
are very fast. Many biological systems are very efficient electron
or ion carriers.
Life provides examples of sophisticated
biological information mechanisms that might provide models for
technological concepts. But which components of biological systems
provide useful models? Looking to nature for models is a different
approach from trying to incorporate biological materials into
computers. This latter objective may in fact not work because
biological systems tend to be too slow for most information processing
applications. For this reason the emphasis should be more on
using biology to discover how to improve information processing.
The Contribution
of Chemistry
Another disciplinary area that
potentially provides useful concepts for computation is chemistry.
Here the focus is on molecular electronic components based on
single molecules. A few years ago this idea of molecular electronic
was seen as fantasy, but the field is now established and a significant
amount of industrial funding is being directed at the area.
A lot of work in the field
of molecular electronics is focused on reinvesting the transistor
at the molecular level. This seems to be the wrong thing to do,
as the scale is wrong, since transistors are not molecules. Another
way forward is to try and reinvent the computer, to build computers
from molecules. The basic architecture for such a computer already
exists. An interesting feature of the architecture is that it
is default tolerant. Defects in the computer can be ignored.
This is a key requirement for a molecular computer as manufacturing
processes working at the molecular level are always going to
produce defects. Eliminating the defects in manufacturing would
be too difficult and too expensive.
To make this architecture work
switches and wires need to be manufactured using molecules. And
this is now possible. One current question is how small these
switches and wires can be made.
Main Issues
The semiconductor industry
needs to decide when to stop investing in silicon technologies
and when to switch over to the new concepts. There is always
a temptation to go a little further with existing technologies,
when the most appropriate response would be to move on to new
technologies.
One of the implications of
quantum computing is a change to complexity classes, where previously
difficult problems become easy to solve. This has major implications
for security and privacy as current cryptography techniques rely
on current complexity classes. As soon as difficult computational
problems become easy to solve today's security systems become
useless.
In the field of biology the
most fruitful path to follow is using nature as a model to find
better ways of doing information processing. Biology can provide
models that lead to new concepts which would be implemented with
non-biological materials.
Of great interest are the self-organising
capabilities of biological system. If such processes can be better
understood, then this might be an area where biology can make
a significant input to computer technology.
Molecular electronics is based
on building computers out of molecular switches and wires. A
combination of molecular electronics and conventional CMOS seems
to have the potential to extend the exponential growth in computer
power for another 50 years.
Three dimensions systems have
fundamental problems. The interconnections needed to pass information
through three dimensions are very complex. A three dimensional
system is not defect tolerant. Separation of bits is also difficult
in three dimensions.
Interfacing at the nano scale
is difficult. There is a need for protection from fluctuations
in the environment.
The nano scale is more than
just about developing new computer technology. The nano scale
is also about new materials that may serve other purposes, for
example in medical treatments.
A big issue is whether the
computing power that is foreseen will actually be needed. The
answer to this question is that no one really knows. The same
question could have been asked about electricity or Charles Babbage's
mechanical analytical engine. It is an issue of human curiosity.
If more powerful computers are not built then the potential will
never be explored or discovered.
One potential benefit of more
powerful computers would be to get rid of the keyboard and to
be able to hold intelligent conversations with a computer.
Building powerful computers
that are seen as being intelligent is an issue that needs to
be discussed. Does society want computers that are capable of
thinking? This is an open question.
The issue of knowing when to
stop long term research was raised. This is important as there
are examples of research that have been stopped just as the results
were about to become highly relevant. There is a need for faith
in ones work and results and determination to bring technologies
to market.
Many things are open-ended.
It is difficult to say something is impossible. Problems that
are hard should not be dropped because often the solution to
hard problems is very valuable.
Technology roadmaps for technology
can be constructed. But one problem is how to create roadmaps
for algorithms. It is very difficult to predict algorithms. At
the moment there are not many good algorithms for quantum computation.
Computers have improved significantly
over the years, but programmers have not improved to the same
extent. There is a need to build machines that can decide how
to work in parallel with programmers.
The notion of self-assembly
is important. To be able to manufacture regular but defective
structures and then to be able to work around these defects through
self-assembly is a power concept.
One of the benefits of biology
will be to show how to control complex systems and how to handle
complexity, not how to do computing. Biology will provide the
insights into how to do things differently.
Research that is high risk
in its nature and long term and deals with strange ideas needs
public funding. The question of return on investment should not
be asked or a return expected. Public money should support high
quality high-risk work regardless of return. It is an investment
in society. It is impossible to state what is and is not valuable.
The research should just be supported.
Conclusions
and Future Directions
In the longer-term quantum
computation may replace classical computation. These are challenges
that need to be addressed in this particular field. An important
question is how to design larger qubit gates. Scalability is
a central and difficult issue. What is likely is a convergence
between nanotechnology concepts and quantum optics. However whatever
technology is used, the ability to scale up to larger and larger
numbers of qubit gates will be a fundamental requirement of the
technology. Small-scale qubit gates are very near in time, but
large scale is far away and very large scale is a very far way.
Biology shows us that nature
uses many different systems for storage and processing of information.
Generally life does not care about the system as long as it works.
This is an important point and its implications need to be understood.
The direct incorporation of biological molecules in current microelectronic
components is generally difficult if not impossible. The value
of biological systems and molecules is that they offer interesting
models that might provide new concepts for information technology.
The industrial interest in
the area of molecular electronics is concerned with developing
strategic technologies that might be useful in 10 to 15 year's
time. These technologies are high risk, but are ones where the
investment is counted in terms of millions of dollars and the
potential payoff comes in billions of dollars.
The past several decades have
seen a massive improvement in computer technology in terms of
size shrinkage and reduction in the power cost of information
processing. The question arises how much better can things get
and how much further can developments move without resorting
to quantum computing? The answer is that there is no physical
law that says that efficiency improvements of another factor
of one billion cannot be achieved! This could mean that there
is no reason why in one handheld computer, consuming one watt
of power, it should not be possible to achieve computer power
equivalent to all the computers that are in existence in the
year 2000! The challenge is how to do this - avoiding problems
with physics and economics (in terms of the high costs of building
fabrication facilities).
The challenge for molecular
electronics is not the making switches, or wires or circuits.
The main problem is interfacing between the micro scale world
and the nano scale world. |