HUMANKIND waited a long time for bronze. But when it arrived, it brought a
revolution that changed civilisation for ever. Then came iron with its hard edge
for swords and ploughshares and, later still, steel helped to usher in the
Industrial Revolution. During the 20th century the pace quickened. In less than
a hundred years, several entirely new classes of material stamped their mark on
the world. Plastics and semiconductors, for example, have spawned revolutions as
far-reaching as those started by bronze, iron and steel. Other materials wait in
the wings鈥攕uperconductors, Bose-Einstein condensates and pancake atoms
will probably make an impression on the 21st century. But more is to come.
Until recently, most new materials were discovered by complete accident, or
by trial and error. The latter strategy involves taking a few metals, mixing
them together in certain ratios, baking them for a while and seeing what comes
out. Researchers gauge the result by prodding, heating and squeezing their new
sample. Obtaining an accurate assessment of a material鈥檚 behaviour under
pressure, at high and low temperatures, in and out of magnetic and electric
fields and in countless other conditions is a process that can take years or
decades. And small changes in the mix can make huge differences in the
properties, as any researcher working with high-temperature superconductors can
testify. Consequently, there are still huge gaps in our understanding of many
materials.
But all this could soon change. Recent advances in mathematics and computing
are making it possible to simulate the properties of materials without ever
making them. The approach turns the whole idea of materials testing on its head.
Instead of asking what a material can do, scientists are deciding what
properties they would like and designing the substance that will do the job.
What鈥檚 more, development work that used to take years can now be done in
months.
Advertisement
The simulations begin with the strange rules of quantum mechanics that govern
matter on the atomic and subatomic scale. 鈥淨uantum mechanics is the most
accurate way of describing materials theoretically,鈥� says Jerry Bernholc of
North Carolina State University at Raleigh. At this level, particles such as
electrons, protons and neutrons behave in ways unheard of at the macroscopic
level. But all their strangeness鈥攖heir wave-particle duality, the
uncertainty in their position and momentum, and their ability to tunnel through
barriers鈥攊s governed by a formula known as the Schr枚dinger equation.
The idea is to plug into this equation details of the desired system鈥攖he
size of an atomic nucleus and the number of electrons surrounding it, for
example. The solution is known as a wave function and it tells a fascinating
tale.
鈥淚n principle, if you have that wave function, you know everything that can
possibly be known about the system,鈥� says Tomas Arias of the Massachusetts
Institute of Technology in Boston. 鈥淚t gives a different probability for any
arrangement of all of those electrons and all of the regions of space where they
could possibly be.鈥�
This is good news for materials scientists because the physical properties of
a material are largely governed by the way its electrons interact. Armed with a
description of the arrangement of nuclei and electrons, scientists can then
calculate the forces between them.
In principle, this sounds easy enough, but in practice the Schr枚dinger
equation is hideously difficult to solve for all but the simplest configurations
of nuclei and electrons. The equation must be written in a form that includes
each particle, and the solution reveals the interactions between all of the
particles in the system. So simulating even a single large atom with hundreds of
electrons is extremely complex. Add a few more atoms and it becomes
unmanageable. 鈥淎 few years ago, a typical large problem could involve 1000
functions and would have taken a billion calculations to solve,鈥� says Mike Teter
or Cornell University in Ithaca, New York. But recent developments are making
these calculations more manageable.
For a start, huge increases in computing power have made larger simulations
possible. And that鈥檚 not all. 鈥淭he algorithms鈥攖he mathematical methods
used to solve these equations鈥攈ave improved at about the same rate as
computing power,鈥� says Teter. For example, one approach distinguishes between an
atom鈥檚 outer or valence electrons, and its inner, 鈥渃ore鈥� electrons. The valence
electrons tend to be the main players when chemical bonds are formed, so
researchers include only these in their calculations. Ignoring the core
electrons vastly reduces the number of equations that have to be dealt with.
Another approach cuts down the number of calculations by using the results of
earlier calculations as starting points for others. With this strategy,
researchers solve the wave function for a material once and then use this result
as the starting point for calculations involving a slightly different
arrangement of atoms. Thanks to such techniques, the number of calculations
needed to solve large numbers of functions has fallen dramatically. Nowadays,
says Teter, a simulation involving 1000 functions would need as few as 3000
calculations.
Taken together, these advances have made a dramatic impact. 鈥淎ll of a sudden,
you can see why there鈥檚 this big revolution in really simulating materials,鈥�
says Teter. Yet despite all the promise, these techniques apply so far only to
specialised situations. 鈥淭hey do not work yet for the general cases,鈥� emphasises
Bernholc.
Where the simulations do work, they make a profound impact on materials
development. One important question that materials scientists often ask is
whether a particular arrangement of atoms is stable. First, they solve the
Schr枚dinger equation and determine the forces between the electrons in the
system. If the forces balance, the structure is stable. If not, the atoms in the
simulation are allowed to 鈥渟ettle鈥� and the new arrangement is the stable
structure. This structure reveals many fundamental properties, such as the
material鈥檚 density and the size of the unit cell if it is a crystal.
The next step is to place the material in a mathematical 鈥渓aboratory鈥� that
measures other properties. The material鈥檚 elasticity is a good example. 鈥淚n the
computer, we force the atoms to have a larger average separation than they would
like, and we see how quickly their energy rises,鈥� says Arias. 鈥淭hat gives you
the elasticity.鈥� Another option is to apply a simulated electric field and
determine the effect it has on the material鈥檚 structure and shape and then
recalculate this as the field increases. If the shape changes, the material is
said to have piezoelectric properties. The results of simulations like these
have amazed researchers often coming within a few percent of the real material鈥檚
properties.
The light touch
Techniques like these are transforming the work of materials scientists.
Using quantum mechanics, John Joannopoulos and his colleagues at MIT have
recently designed what he calls a deliberately modelled material. 鈥淚 took some
specifications,鈥� he says, 鈥渁nd tried to make a material that would meet them,
rather than just putting different types of atoms together to see what
丑补辫辫别苍蝉.鈥�
His goal was to design a material with the optical properties of gallium
arsenide but with a physical structure similar to silicon. Gallium arsenide is
an optical semiconductor鈥攊t can turn optical signals into electronic ones
and vice versa. It is a major component of light-emitting diodes, for example.
What excites researchers about gallium arsenide is its potential to act as a
gateway between optical computers and electronic ones鈥攂ased on
silicon鈥攖hat would convert the signals back and forth.
Growing gallium arsenide on top of silicon chips is very difficult. For one
thing, gallium arsenide and materials like it are polar. In other words, the
positive and negative charges within them are not evenly distributed. This can
generate unwanted electric fields in neighbouring silicon. In addition, the
crystal lattice of gallium arsenide does not line up well with that of
silicon.
So Joannopoulos attempted to invent a nonpolar material with the optical
properties of gallium arsenide and a similar crystal lattice size to silicon. He
started searching by hand for atomic structures that were nonpolar and had
roughly the lattice dimensions he was looking for. This in itself is a large
task since he had to select from more than 20 types of atom, from beryllium to
tungsten, which could be used in almost any combination. From this search he
selected about 40 combinations that gave promising matches with silicon, and
then began to simulate them on a computer to calculate their exact lattice size
and optical properties.
After a decade of work, Joannopoulos says, the breakthrough has finally come.
鈥淚t鈥檚 only in the last two or two and a half years that we鈥檝e really made
progress, and it鈥檚 only been in the last few months or so that we actually
designed something that should have all the right properties.鈥� He and his
student Tairan Wang have made a material composed of repeated layers of
phosphorous-zinc-arsenic-silicon-arsenic-zinc-arsenic-silicon. This is
nonpolar, has a lattice constant that is within 0.08 per cent of silicon鈥檚 and
optical properties similar to gallium arsenide. It also matches the thermal
expansion of silicon to within 0.01 per cent.
Despite the improvements in computing power and algorithms, there is a
practical limit to the number of atoms that can be included in a quantum
simulation. 鈥淲e cannot do quantum mechanics on an automobile,鈥� says William
Goddard of the California Institute of Technology in Pasadena. But it is
possible to build on the quantum mechanical approach.
Once the Schr枚dinger equation has determined the important forces at
work between atoms, these forces can be treated like springs. 鈥淚nstead of the
Schr枚dinger equation, you have Newton鈥檚 equation,鈥� Goddard says. The
Newtonian equation for motion鈥攖hat force equals mass times
acceleration鈥攃an be calculated about three orders of magnitude faster than
the Schr枚dinger equation. Consequently, researchers can simulate systems
containing as many as half a million atoms. This approach is called molecular
dynamics.
Goddard and his colleagues at the Chevron Chemical Company, based in La
Habra, California, have used molecular dynamics to design long-lasting camshafts
for cars. Their challenge was to find a lubricant that could be added to oil in
place of zinc and dithiophosphates, which Chevron would like to eliminate for
environmental reasons. In the past, the effect of additives could only be
measured by stripping down an engine to measure the camshaft wear鈥攁n
expensive, time-consuming process.
So Goddard and his colleagues built a simulation of 13 new wear-inhibiting
compounds that Chevron was considering. They simulated the properties of each
compound and predicted which one would work best. 鈥淲e鈥檙e calculating things like
the coefficient of friction and the equilibrium between the molecules in the
surface and the liquid,鈥� he says. 鈥淲hat I didn鈥檛 know,鈥� he adds, 鈥渨as that they
had actually been testing some of these things. They had eliminated about nine
of them based on our advice. Of the remaining four, they had done some
experiments and they selected the same one that we did.鈥�
For nearly a decade, Uzi Landman and his colleagues at the Georgia Institute
of Technology in Atlanta have been using molecular dynamics to study the
lubrication of devices on the microscopic scale. In a hard disc drive, for
example, the reading stylus and the disc come within a few nanometres of each
other and are lubricated by a liquid layer a few atoms thick. But this lubricant
tends to organise itself into layers. 鈥淵ou get a layered-cake effect, with each
layer being one layer of molecules,鈥� says Landman. The problem is that layers
interact with each other, creating friction and resisting motion.
What might fix this? 鈥淚f you want to disturb something, shake it a little,鈥�
says Landman. He and his colleagues created a simulation of two gold surfaces
separated by 2 nanometres of a wax-like lubricant. As the plates moved over each
other, one of them was forced to vibrate, shaking the molecules in the
lubricant. That tiny oscillation reduced the friction by at least an order of
magnitude. This result was later confirmed experimentally by Jacob Israelachvili
of the University of California, Santa Barbara. 鈥淚n the field of lubrication,鈥�
Landman says, 鈥減eople get patents and become rich sometimes if they produce
something that reduces friction by a factor of two. So a factor of 10 is kind of
unheard of.鈥�
Bernholc and his colleagues are using molecular dynamics to simulate the
strength of nanotubes, which are essentially sheets of graphite wrapped into
cylinders. Nanotubes have not been made in sufficient quantity to test their
bulk properties. But by building a model, Bernholc has already determined how
they deform or break when bent. Recently, a group at the University of North
Carolina at Chapel Hill confirmed their predictions.
快猫短视频s will soon be using quantum mechanics and molecular dynamics to
simulate a huge variety of new materials. 鈥淚t鈥檚 not that you鈥檒l never have to do
another experiment,鈥� says Goddard. 鈥淚t鈥檚 that instead of doing a thousand
experiments, maybe you鈥檒l only have to do ten. And maybe, already of those ten
you know which one is most likely to work, and if it works well enough, you鈥檙e
finished.鈥� Nobody knows what weird and wonderful materials these approaches will
bring or how they might affect society. But if the past is anything to go by,
the 21st century should be a roller coaster ride.