ÐÓ°ÉÔ­´´

Margarine, petrol and fertiliser appear to have little in common. But
in the chemical processes that yield these disparate products, molecules
must be persuaded to take part in a sequence of well-drilled moves that
would be complex and action-packed enough to satisfy the choreographer of
a West End musical. This show even has its own stars – not the atoms and
molecules that rearrange themselves to form the end product, but small amounts
of metal, catalysts that keep the molecules in step as chemical bonds break
and re-form at lightning speed. Without these catalysts, the reactions that
the chemicals industry relies on would be impractically slow.

ÐÓ°ÉÔ­´´s are even making ‘movies’ of the molecular dance on the surface
of catalysts. Their work will never reach the big screen; the pictures look
more like a sequence of contour maps than close-ups of atoms. But it goes
a long way towards creating a picture of some of the world’s most valuable
chemical reactions. Even more important, these techniques could be the key
to designing the catalysts of the future.

Anyone who has ever experienced the stench of bad eggs as a car with
a cold engine drives off will know that the catalytic converters used to
clean up car exhausts are far from perfect. The same goes for the other
industrial catalysts, despite the billions of dollars that have been spent
on developing and manufacturing them. But chemists developing new catalysts
still work by trial and error. The problem is that to design the perfect
catalyst it is necessary to know exactly how atoms and molecules behave
on a surface. To find out, they have to place them there precisely, like
pieces on a chessboard, then record what they do.

Snapshot at the quantum corral

This degree of control was brought a step closer to being realised in
1990, when Don Eigler and his colleagues at IBM’s Almaden Research Center
in California used a scanning tunnelling microscope (STM) to pick up individual
atoms from a surface and rearrange them into patterns. By late last year
they had even succeeded in building a minute chemical structure on a metal
surface: a ‘quantum corral’, a circle of atoms just 14 nanometres across.

It has become almost routine to use an STM to look at the positions
of individual atoms on a surface, to see how atoms bind there and how they
arrange themselves next to each other. By revealing the positions of atoms
before and after a reaction, this can tell us something about how molecules
react. It is like taking two snapshots and guessing what happened in between.
The STM can take its snapshots a few times a second – but as some reactions
happen in less than a million-millionth of a second this is not nearly
fast enough to reveal how atoms and molecules approach a surface, and each
other, to form bonds. Clearly, another approach is needed to discover how
molecules move during the surface reactions that underpin not only catalysis,
but also corrosion and the growth of semiconductor films.

When a molecule hits a surface during catalysis, or any other sort of
surface reaction, it can split (dissociate) or stick there (adsorb) or bounce
off (scatter). Finding out which of these possibilities dominates a reaction
has important practical consequences. It could help to explain why some
catalytic reactions are so slow, and so give clues as to how they might
be speeded up, or why one metal may be a much better catalyst than another.
Nickel, for instance, is an excellent catalyst for the key process in margarine
manufacture, which involves splitting the bond that holds hydrogen molecules
together. But copper, which is only an electron away from nickel in its
all-important electronic makeup, will not work.

Unfortunately, real catalytic reactions are too complex, and take
place too fast, for it to be possible to follow them directly, as they happen.
But surface scientists can establish some ground rules from studying simpler
systems. They make direct observations of molecules just after they react
– that is before they have chance to interact with anything else, unlike
STM studies, where molecules collide with the surface as many as 1012 times
before they are recorded again – and compare these observations with mathematical
models that can be worked through on a computer to predict how the molecules
will behave under particular circumstances. In our research at the University
of Liverpool we have been following this combined approach.

Until recently, chemists thought they had a reasonably accurate picture
of at least one simple system: the splitting of a diatomic molecule such
as hydrogen (H₂) on a copper surface. In this picture, a hydrogen
molecule splits only if it acquires enough energy to push the two hydrogen
atoms apart. Copper acts as a catalyst for this reaction by weakening the
bond holding the two hydrogen atoms together when the molecule is close
to the metal’s surface. Once over this energy barrier the atoms separate
and are adsorbed onto the surface. To speed up this reaction seemed straightforward
enough: increase the molecule’s velocity so that it hits the surface of
a catalyst harder. This brute force method should help it to overcome the
energy barrier.

But this picture turns out to be too simplistic. It assumes that the
surface is perfectly uniform: that it does not matter where the molecule
strikes. It also assumes that a molecule’s energy depends only on its distance
from the surface: that is, it assumes that the energy increases as the molecule
is pushed towards the surface, until it gets so close that the two hydrogen
atoms can bind to the metal. In particular it ignores the length of the
bond joining the two hydrogen atoms and the orientation of the molecule
as it arrives. Both these parameters change as the molecule rotates and
vibrates.

Chemists attempting to build a more sophisticated mathematical model
of how bonds are made and break in three dimensions face a double challenge.
First they have to calculate the ‘shape’ of the metal surface in terms
of its energy: that is, they have to describe how the energy of a molecule
varies, depending on its position relative to the surface. Then they have
to calculate how molecules move on the surface. Just the first part of this
fiendishly difficult calculation has occupied them for decades.

But a breakthrough may be in sight, thanks to developments over the
past year based on theoretical work by John Harris in the late 1980s at
the Julich Nuclear Research Centre in Germany. When Harris calculated the
interactions between a hydrogen molecule and a cluster of atoms representing
a metal’s surface, he found that it takes more energy to break the molecular
bond at a copper surface than on the surface of other metals such as nickel.
More surprisingly, the bond had to be stretched before it would break.

Reversing the process

Meanwhile Glen Kubiak, Greg Sitz and Dick Zare at Stanford University
were investigating the reverse process: two hydrogen atoms coming together
on copper and leaving as a molecule. They found that part of the energy
released in this process ended up in the molecule’s vibrations. In their
picture, which fits in with Harris’s energy surface, the hydrogen molecule
is initially formed with a stretched bond, and as it leaves the surface
the atoms spring back together, making it oscillate.

If these vibrations influence whether or not the molecule splits, it
should be possible, in theory at least, to control a key step in catalysis
by making the molecule vibrate more. We took up this question in our research
at Liverpool. Mick Hand and one of us (Stephen Holloway) calculated that
a hydrogen molecule in the vicinity of a potential surface like Harris’s
should split when its velocity is increased above a certain threshold. Below
this limit the molecule needs to gain vibrational energy to split.

These calculations have been tested in a flood of experiments that
investigated the shape of the energy surface for thesplitting of hydrogen
on copper (see box ‘Surfaces with potential’). Their results fit in well
with Harris’s predictions for the shape of the energy surface and our theory
of the effect of vibrational motion. Experiments by Klaus Rendulic’s group
at Graz University in Austria and Brian Hayden and Christine Lamont at the
University of Southampton confirmed that there was indeed a threshold kinetic
energy below which hydrogen would not split. Hayden and Lamont also showed
that increasing the vibrational energy of a hydrogen molecule made it more
likely to split, and more recently Charlie Rettner, Hope Michelsen and Dan
Auerbach of Almaden showed that increasing vibrational energy meant that
hydrogen would split at lower velocities.

But there are more complications. Even at high energies, only about
half the molecules split when they collide with the surface. The rest scatter
back from it, intact. Perhaps molecules cannot split if they come down at
the wrong place on the surface, or at the wrong angle – perhaps end on,
for example, with one hydrogen atom close to the surface but the other too
far away to form a bond. Would this multidimensional picture stand up to
experimental scrutiny?

In 1992, one of us at Liverpool (Andrew Hodgson) and the Almaden group
independently measured how a scattered hydrogen molecule divides its energy
between vibrations and translations. We discovered that when hydrogen has
enough energy to split, it can exchange energy easily between these two
motions (see Figure) in the region where the bond is breaking. Our picture
now looks something like this: the hydrogen molecule comes down to the surface,
the bond stretches but does not quite break, the atoms spring back together,
and then the molecule bounces back off, travelling more slowly than before
but with its atoms vibrating rapidly.

The molecule’s rotations take part in this process too. A rapidly rotating
molecule tends to convert less of its vibrational energy into velocity,
and the molecule’s rotation changes the angle it makes with the surface
during the collision. So it seems that the potential surface is sensitive
to this angle – another challenge to the simple one-dimensional model.

Within the past 12 months the Almaden group has discovered that the
rotation of a molecule also subtly affects how fast it has to be moving
for it to split: a little rotation hinders splitting, but as rotational
energy increases it begins to help the molecule to split. To describe this
movement properly, all possible coordinates have to be taken into account
(see box ‘Surfaces with potential’). This in turn calls for better calculations
of the potential surface – calculations that would for the first time include
the effect of the molecule’s rotations.

Last year, two groups, one led by David Bird at the University of Bath
and one by Jens Norskov at the Technical University of Denmark, Lyngby,
near Copenhagen, published such calculations. These will now enable us
to simulate on a computer how rotational energy and the effect of surface
site – such as whether the molecule is on top or between two surface atoms
– affect the reaction.

So how far can ideas developed in studies of model systems be applied
to a real reaction, where bonds are formed, rather than broken? In January
this year, Rettner and Auerbach investigated a system that could be measured
directly, the reaction of hydrogen atoms with chlorine atoms adsorbed on
gold. They scattered a beam of hydrogen atoms off a chlorine-covered surface
and observed the hydrogen chloride formed, using the same techniques as
in the hydrogen on copper model systems.

They discovered that this reaction can occur in two ways. Either the
incoming hydrogen atom reacts with the chlorine atom immediately, as it
approaches the surface, and the hydrogen chloride formed scatters off the
surface in a direction that depends on the angle of the approaching hydrogen
atom. Or the hydrogen atoms are trapped on the gold surface, where they
move around until they meet a chlorine atom, and react where the energy
barrier is lowest to form hydrogen chloride.

The potential surfaces for this reaction have not yet been calculated,
so it is not yet possible to say what makes the molecule select one route
in preference to the other. But this difference has practical implications
for more complex real reactions: if the molecule takes the direct route,
then controlling the energy and the conditions of the incoming gas can help
to modify the path of the reaction. This could mean being able to control
the formation of a particular product – exactly the aim of the chemists
who design catalysts. If, on the other hand, the molecule takes the indirect
route, then all that can be changed is the surface temperature, and there
is much less chance of controlling the reaction.

The dissociation of hydrogen on a copper surface and the reaction of
hydrogen with chlorine serve as important models for the dissociation and
reaction of other small molecules on metal surfaces. One of the main lessons
to emerge from this research is that every system is uniquely complex, so
we are still a long way from unravelling the whole story of catalysis. But
the tools are now available for investigating other systems. Soon we may
be able to rely on calculations to tackle problems which are just too difficult
to study experimentally – and then sitt back to watch the results unfold
as images on a screen.

Andrew Hodgson is a lecturer in chemistry and Stephen Holloway is a
professor of chemical physics at the University of Liverpool. Both are principal
scientists in the IRC in Surface Sceince at the University of Liverpool

* * *

Surfaces with potential

The model on which theorists base their calculations of the forces that
operate as a molecule approaches a surface is called the potential energy
surface. Just as an ordinary contour map describes your height above sea
level as you walk across country, the potential energy surface is a map
which describes how the energy of a molecule changes as it moves towards
a surface.

This map must include all the directions, or coordinates, that the molecule
can move in. These include z (towards the surface), x and y (across the
surface), and r (the distance between the atoms in the molecule). From their
map, surface scientists can calculate the direction of the steepest slope.
This is the direction in which a molecule will initially move, just as a
ball will start to roll down a real incline. Once on the move, the ball
will keep going in the same direction until it meets a slope that either
steers it in another direction, or slows it down and stops it. In just the
same way, a molecule moving near the surface of a metal feels the potential
energy surface steering it this way and that.

Until a few years ago theorists had few reliable energy surfaces on
which to do their modelling, because the calculations needed to produce
them were just too time-consuming and difficult. They had to resort to qualitative
models of the surface to investigate how features of the energy ‘landscape’
influence the reaction. But now they have much more comprehensive maps of
the potential energy surface, from which they can calculate how molecules
move. There are several ways to do this. One is to take each atom in turn,
work out the forces on it and apply Newton’s classical equations of motion
to work out where it will be a short time later. This process can be repeated
for each atom in a molecule over the course of its encounter with the surface,
until either the atoms separate and split apart or the molecule leaves the
surface.

This method works well for relatively heavy molecules such as oxygen
and nitrogen. But for small ones such as hydrogen, classical Newtonian laws
no longer apply, and quantum mechanics starts to take over. So to get realistic
results, the wave nature of matter has to be considered. This involves describing
a beam of molecules mathematically as a bundle of moving waves, called a
time-dependent quantum wave packet. From the way these wave packets evolve
in time, theorists can derive a set of mathematical probabilities of what
will happen to a small molecule – whether it bounces off a surface, splits
into its constituent atoms, or whatever. From the calculations emerges a
picture of how the molecule moves as it approaches the surface and dissociates
– a movie of the reaction on a timescale of less than a million-millionth
of a second.

These calculations require enormous computing power. The most complicated
so far has been a calculation in four dimensions for hydrogen dissociation
on copper, which included four different coordinates simultaneously. Developments
in parallel computing are set to widen the scope of these techniques and
will allow still more complex systems to be modelled.

* * *

Playing squash with molecules

To study what happens as molecules interact on the surface of real crystals,
experimental scientists make a flat, regular, single crystal, and observe
how a beam of molecules scatters from or sticks to it. The beam is produced
by allowing gas at high pressure to expand through a small hole, just like
water squirting out of a high-pressure hose pipe. The pressure of the beam
is then kept extremely low – less than a ten-million-millionth of an atmosphere
– to prevent collisions with impurity atoms, such as oxygen or hydrocarbons,
which would destroy the beam or make the surface dirty.

Experimentalists are interested in varying the translational energy
– the velocity – of the molecules in their beam and their vibrational energy
in turn. It is possible to work out from the temperature of the nozzle how
many molecules have vibrational energy and how many have not. Changing the
temperature therefore allows the vibrational excitation of the beam to be
changed. The nozzle may be heated to temperatures as high as 2000 K. This
increases the velocity and the vibrational energy of the gas. The velocity
can be reduced, independently of the vibrational energy, by mixing or seeding
the molecules in a heavier gas, such as deuterium, neon or argon. Different
combinations give a stream of molecules whose energy and direction is known
quite precisely.

Information about the fraction of the molecules of hydrogen, for example,
that split up on the surface is obtained from the pressure change as the
beam hits the surface and molecules stick there. The energy and angle at
which the gas strikes affects the probability that the molecules will dissociate.
The angular spread of molecules which bounce off the surface tells researchers
about the way energy is exchanged there.

Another possibility is that the collision energy is converted into internal
motion of the molecule – that is, vibrations or rotations. To study this,
experimentalists use a technique called resonance enhanced multi-photon
ionisation, in which a high-powered laser beam excites molecules near the
surface to higher energies. From the absorption spectra of the scattered
molecules, the researchers can deduce the rotational and vibrational energy
of the molecules. They correlate this with the initial collision energy
and the angles at which the molecules arrive and leave the surface. This
information can then be matched against the theorists’ prediction.