¿ìè¶ÌÊÓƵ

For the past forty years, electronics engineers have never failed to
exceed all expectations as they have packed more and more information onto
silicon chips. From 256 kilobits to 1 megabit to 4 megabits . . . even within
the past decade the capacity of memory chips has quadrupled and quadrupled
again as the process of etching minute circuits onto silicon has been pushed
to further limits. But now, as engineers are finding it increasingly hard
to pack yet more components into less and less space, a shift in philosophy
may be the only way forward.

Our group at the University of Birmingham, and a growing band of researchers
elsewhere, consider that a new approach, inspired by chemistry and biology,
might provide a way to construct the ever more compact devices envisaged
for the future. Instead of attempting to continue to use ‘top-down’ engineering,
in which increasingly minute circuitry is etched onto the surface of a silicon
wafer, we might instead create devices on much smaller scales by working
from the ‘bottom-up’, using specially designed molecules that have the capacity
to assemble themselves into the structures required.

The molecules that will form the building blocks of such systems will
have to have highly specialised structures so they can recognise each other
and combine, like the pieces of a self-assembling jigsaw, to form one unique
superstructure. To design and make three-dimensional structures out of molecules
and ions is a staggering challenge for the future. But already, in our
group and in laboratories around the globe, the first very simple synthetic
self-assembling structures have been created, some of them capable of functioning
as miniature molecular switches.

For inspiration in this futuristic project, there is life itself. As
long ago as 1959, the physicist Richard Feynman recognised the elegance
and efficiency of the chemical processes carried out by natural systems,
and saw in them the model for molecular machines that could store and transfer
information. What singles out biological systems as special, and what has
made possible the enormous diversity of living matter in nature, are the
powers of self-organisation, self-assembly and self-replication.

Nature’s best-known self-assembling superstructure is DNA. Its double-stranded
structure is formed by the coming together of two complementary single strands
of DNA. The process is reversible, and the ‘correct’ DNA superstructure
is formed only if the bases attached at regular intervals along each helical
backbone are matched by bases on the other helix. The matching is brought
about by hydrogen bonds, which are stable only for particular base pairings.

This kind of strict self-assembly also operates on a much larger scale.
Perhaps the best understood self-assembling biological nanostructure is
tobacco mosaic virus (TMV). It consists of 2130 identical protein subunits,
each containing 158 amino acid building blocks, that form a helical protein
sheath 300 nanometres long and 18 nanometres across, around a single strand
of RNA made up of 6400 nucleotides. After TMV has been broken up into its
component RNA and protein units it can be reconstituted in the laboratory
to form the fully infective virus – an elegant demonstration of the efficiency
and power of self-assembly.

The seeds of ideas for self-organising, self-replicating and self-assembling
systems were sown by the 1987 Nobel laureates in chemistry, Donald Cram
of the University of California at Los Angeles and Jean-Marie Lehn of the
University of Strasbourg. They realised that certain simple molecules will
associate with each other in a discrete way in solution, and that to choose
molecules that will do this, chemists must understand the processes that
lead to this coming together.

This crucial process of recognition is not based on the sort of interactions
that have underpinned chemistry until now, in which electrons are shared
to form ordinary covalent bonds. It is the much weaker noncovalent interactions
such as hydrogen bonding, van der Waals forces and hydrophobic interactions
that are the key to recognition (see also Box 1). At Birmingham we have
been concentrating on the self-assembly of structures and superstructures
in which the recognition stage features some kind of molecular entanglement.
After all, nature relies on the convolutions of a double helix to pass on
information. Our first goal was to identify small molecular systems to use
as models for exploring how self-assembly can occur. We concentrated on
systems in which two or more components are mechanically interlocked but
not covalently bonded. This was to allow us to investigate the rules of
self-assembly between components that would stay locked together even when
they are not held together by strong chemical bonds.

Two classes of molecule fill the bill: the catenanes, which consist
of two or more interlocking rings, and the rotaxanes, in which one or more
rings are threaded onto a dumbbell-shaped molecule (see Figure 1). In the
rotaxanes, bulky groups of atoms at either end of the dumbbell act as stoppers
that prevent the rings slipping off. As well as being possible models for
self-assembling systems, the (n)rotaxanes, which contain one dumbbell and
n-1 rings, conjure up the prospect of polyrotaxanes – a possible new class
of macromolecular compound that would have all the hallmarks of a molecular
abacus.

The earliest attempts to make rotaxanes and catenanes, back in the 1960s,
did not feature molecular recognition. Unfortunately, the key step in which
a linear molecule threads itself through a cyclic one is rather unlikely
to happen if it is left to chance. But in the mid-1980s we discovered a
way of threading a linear molecule through a cyclic one – a method that
would lead us to find out a lot about self-assembly. It exploits the relatively
strong noncovalent interactions that exist between electron ‘acceptor’ and
‘donor’ units. Acceptors are present in molecules of the nitrogen-containing
compound best known as paraquat. These molecules have two positive charges,
which means that the benzene-like rings are deficient in electrons, making
them more likely to accept electrons from elsewhere.

EFFICIENT ASSEMBLY

Donors are based on another benzene derivative called hydroquinone.
Here, hydroxyl (OH) groups replace two of the hydrogen atoms on opposite
sides of the six-carbon ring. The oxygen of the OH groups tends to push
negative charge towards the ring, making it slightly more electron-rich.
It seemed likely, then, that we might be able to combine donor with acceptor
components in such a way that they would come together spontaneously, in
a process of self-assembly. These components would not be chemically connected
in the conventional sense of covalent bonds, but by weaker noncovalent interactions.

We discovered that a crown ether – a large ring system containing repeating
-OCH₂CH₂– units and incorporating two hydroquinone
donor units arranged symmetrically – will spontaneously encircle and bind
strongly to the rod-like paraquat acceptors. The result is a superstructure
that looks like a rotaxane, but without stoppers. Next, we found that we
could play this game the other way round. If we built the paraquat-like
acceptors into opposite sides of a rigid, box-like molecule, hydroquinone
and other donors would sit in the middle of the box. Could this be the basis
for the efficient molecular self-assembly of rotaxanes and catenanes?

In 1989, two of our group, Neil Spencer and Cristina Vicent, were delighted
to discover that it could. In our first successful attempt, two components
self-assembled to make a (2)catenane – yielding 70 per cent of the maximum
possible. One component, the symmetrical crown ether with its two donor
units, acts as the template for the formation of the other – the open box,
with its two acceptor units (see Figure 2). The open box attaches itself
to the crown ether like an open link being threaded onto a chain – but there
are no covalent bonds keeping it there. Instead, it is held in place by
noncovalent bonding interactions between the stacked donor and acceptor
rings. Finally, the formation of a covalent bond closes the link to produce
a (2)catenane. Since then, we have made subtle structural modifications
to the components to generate an extensive range of self-assembled catenanes
and rotaxanes.

The recognition and order that goes into the self-assembly of these
molecules ‘lives on’ after they have formed. This means that the components
cannot move freely or randomly with respect to each other. Rather, they
switch between two arrangements in which the noncovalent interactions are
equally favoured in energy terms. We believe that this switching between
states is a promising place to start searching for ways of controlling machines
designed from molecules upwards.

MOLECULAR SWITCHES

Catenanes and rotaxanes can be regarded as forerunners of molecular
switches. It works like this. If we make the two donor units in our ring
different, rather than identical, then there will be two different binding
sites available to the box-like acceptor. If we start with the box on one
of these sites, and then use a physical or chemical impulse to move it to
the other site (see Box 2 for more details), we have created a molecular
system that is capable of expressing binary logic – the basis of every digital
computer. It might one day be possible to build a computer out of molecular
‘wires’ and switchable molecules, though that day is most certainly a long
way off.

Working with Vincenzo Balzani of the University of Bologna, we have
been studying a simple molecular machine – a switch that is driven by light.
Our machine is based on a superstructure that spontaneously self-assembles
when a thread-like molecule containing a donor is added to an aqueous solution
of our box-like molecule, with its two acceptor units. When this solution
is irradiated with light under the right conditions, the box is chemically
reduced, causing an electron to be transferred. This destabilises the supramolecular
assembly, so the thread is expelled from the box. The process is reversed
by oxidation: if oxygen is bubbled into the solution in the dark, the thread
worms its way back into the box again. A machine operating on similar principles
could some day play a role in storing and processing information at the
molecular level.

There is still a chasm between the making of small molecules and molecular
assemblies such as catenanes and rotaxanes, and the supramolecular arrays
many nanometres long, containing many thousands of molecule, that could
function as molecular machines. To build such molecular machines from the
bottom up, from matching molecular components, chemists are having to make
a leap of imagination. It is akin to the change in thinking that led to
advances in bridge building in the last century: by abandoning tried and
tested arches in favour of the new cantilever and suspension designs, engineers
were able to create spans longer than ever before.

In the same way, chemists will have to learn to build on conventional
methods of synthetic chemistry using the concepts of self-organisation,
self-assembly and self-replication that come from the biological world.
To succeed, they will need a broad appreciation of science, a flair for
abstract design, a grasp of fundamental principles, familiarity with the
latest techniques – and, not least, a desire to chase fantasies.

David Amabilino is a postdoctoral fellow and Fraser Stoddart is professor
of organic chemistry, both at the University of Birmingham.

* * *

A matter of organisation

Understanding how molecules self-organise is now one of the major goals
of organic chemistry, and chemists tackling the problem have taken a variety
of routes. Chemists have known for some years that amphiphiles – molecules
with watery heads and greasy tails – can organise themselves into superstructures
of single and multiple layers called Langmuir-Blodgett films. During the
1980s, Helmut Ringsdorf of the University of Mainz went a step further by
chopping the amphiphile molecules in two and adding chemical group to the
broken ends that self-organise to re-form the amphiphiles as a supramolecular
system involving weaker interactions.

Meanwhile, teams in the research laboratories of Jurgen-Heinrich Fuhrhop
at the Free University of Berlin and Janos Fendler at Syracuse University
in the US developed synthetic three-dimensional superstructures which resemble
the sort of fluid-filled sacs or ‘vesicles’ found in biological cells. At
the same time, George Whitesides at Harvard University helped to pioneer
the self-organisation of monolayers at gold surfaces, so forming ordered
two-dimensional networks of molecules – essentially two-dimensional organic
crystals, in the size range 100 to 1000 nanometres.

During the past couple of years, Gunter von Kiedrowski at the University
of Freiburg and Julius Rebek Jr at the Massachusetts Institute of Technology
have demonstrated self-replication involving relatively small organic molecules,
similar to components of DNA and RNA. In both experiments, hyd-rogen-bonding
interactions were responsible for assembling their self-replicating molecules.

* * *

From shuttles to switches in simple steps

In 1992, Douglas Philp at the University of Birmingham made a prototype
for a self-assembling family of ‘intelligent’ molecules when from a mixture
of four components – two stoppers, a rod and a ring – he isolated a (2)rotaxane.
In the first step of Philp’s assembly process, a stopper attaches itself
to the rod (see Figures (a)and (b)). This ‘switches on’ a recognition site
for the ring on the rod, to which the ring binds. This is then trapped on
the rod by the attachment of the second stopper, which also has the effect
of ‘switching on’ a second recognition site for the ring within the molecular
assembly. This chem-ical system works by the interplay between the formation
of chemical bonds and molecular recognition.

The ring in the (2)rotaxane occupies the two recognition sites in the
dumbbell equally, moving rapidly between them about 300 000 times per second.
This molecule, which has a molecular weight of just under 3000 daltons
and measures about 5 nanometres from end to end when stretched out, is app-roaching
the scale on which a mole-cular information storage system could operate.
The challenge now, if it is to be useful as a switch, is to find ways of
controlling this molecular shuttling.

One way to control a molecular switch is by removing an electron from
the electron-donor units within the molecule or adding one to acceptor units.
It may be possible to do this electrochemically. For example, suppose that
two different ‘stations’ of about the same size are combined into the dumbbell
component of a (2)rotaxane. If one station is a stronger donor than the
other, then the acceptor unit would be expected to combine with it in the
self-assembly process. This donor should also be the easiest of the two
units in the dumbbell to oxidise, as it will give up an electron more easily.
So a single oxidation of the (2)rotaxane should induce the ring to move
from one recognition site in the dumbbell component to another, providing
the electrochemical mechanism we require to drive the shuttle
(see Figure, left).

At Birmingham, Richard Bissell has self-assembled some prototypical
(2)rotaxanes, and used them to design a molecular shuttle containing two
hydroquinone residues and one tetrathiafulvalene (TTF) unit, which he knew
to be a better donor than hydroquinone. Bissell has since moved to the University
of Miami, where he is working with Angel Kaifer to study ways of controlling
this molecular shuttle electrochemically.