ÐÓ°ÉÔ­´´

Two researchers at the Natural History Museum in London need samples
of an obscure mineral to help them to explain why some ores of silver are
more sensitive to light than others. Early last year, Alan Criddle and Chris
Stanley, discovered the mineral in a sample of silver ore in Mexico. The
mineral is a compound of silver, tellurium and sulphur (Ag₄TeS), which they
called cervelleite. It exists as a narrow boundary between a core of acanthite,
or silver sulphide (Ag₂S), and a ring of hessite, or silver telluride (Ag₂Te).
When exposed to light, there was a striking, progressive change in the sample’s
surface appearance. Like many acanthites, the core blackened as its silver
atoms precipitated to the surface; in this case, however, the migration
was quicker and more marked than usual. The narrow band of cervelleite turned
from blue to green, and the ring of hessite, which does not normally respond
to light at all, kept changing colour. Criddle and Stanley concluded that
the bizarre tarnishing probably had something to do with the presence of
cervelleite and involved a migration of silver, tellurium and sulphur atoms
between the different mineral grains. The implications of these phenomena
for the development of photosensitive devices persuaded them to pursue their
research. The snag is that only tiny amounts of cervelleite exist naturally.

A few miles away, two researchers at University College, London, Frank
Beech and Ian Boyd, are investigating high-temperature superconductors.
They want to discover if replacing oxygen atoms by fluorine atoms will raise
the temperature at which a material becomes a superconductor, the so-called
transition temperature. The trouble is that the researchers must be able
to implant the fluorine atoms precisely, at constant concentrations in particular
places. This is not possible with conventional techniques because they involve
heat treatment, and fluorine is volatile.

Meanwhile, industry wants electronic circuits to be smaller and faster
to meet the demands of modern technology. Engineers must reduce the size
of the components that make up the circuits of microchip; already in the
laboratory, researchers are testing components that are less than 0.25 micrometres
across, which is a quarter of the size of a typical component in use today.
(The average human hair is about 80 micrometres across). They must also
cram in more circuits and increase the conductivity of the link, or interconnect,
between individual components and circuits. These developments, however,
threaten to turn the minor irritants that come with todays chips – for example,
the need to dissipate the heat and to prevent the interference generated
between components by electronic circuits working closely together – into
serious problems that could limit technological advance.

One way of getting round the difficulties of making circuits smaller
and faster and, also, of producing rare minerals artificially and of manipulating
superconducting materials, is to use high-energy positive ions. The technique,
known as ion beam synthesis, was little more than a research tool just three
years ago (‘Oxygen breathes new life into silicon wafers,’ New ÐÓ°ÉÔ­´´s,
19 November 1987). Now, it could form the basis of a production line.

Among the first things we learn in chemistry is that it is rarely possible
to form a compound simply by combining its constituent elements – an energy
barrier must be overcome before the reaction can proceed. This is what happens
in ion beam synthesis. A particle accelerator fires positive ions at a target,
typically at speeds of around 1500 kilometres per second, although the latest
equipment can accelerate them to speeds up to 10 times as fast. When the
ions enter the target they lose their charge and become atoms; they also
lose their kinetic energy and stop at a depth that depends on their initial
speed. The energy lost by the implanted atoms is sufficient to break bonds
and, because they and the target atoms are intimately mixed, chemical reaction
occur in this region at temperatures well below those at which compounds
usually form. Also, because the technique uses a mass spectrometer to select
the implanted ion and buries the particle below the surface, the synthesised
compound contains only the elements it is supposed to, which makes the process
inherently clean. Contaminants often degrade the properties of layers grown
as films on the surfaces of targets using other techniques, such as molecular
beam expitaxy (MBE) and ultra high vacuum (UHV) deposition. Another advantage
of the technique is that changing the energy of the implanted ions alters
the final depth of the synthesised compound. As a result, engineers can
design in three dimensions.

Faster trips with cobalt

One of the most exciting breakthroughs in the development of the technique
came in December 1987. A team of researchers at AT&T Bell Laboratories
in the US implanted large doses of cobalt ions into single crystal silicon,
which they kept at about 350 °C and, after the implantation, annealed
at 1000 °C for 30 minutes. To their delight, single crystal layers of
cobalt disilicide formed at a depth of 100 nanometres (100 millionths of
a millimetre) beneath the wafer’s surface. These layers were about 150 nanometres
thick, and with the same orientation, or epitaxy, as the silicon wafer.
(Epitaxy is important because a homogeneous crystal structure enables engineers
to control the design of devices accurately.) This was something that researchers
all over the world had been trying to do for the previous 10 years, with
little success.

The microelectronics industry shared the researchers’ excitement. For
a start, industry needed a new material for connecting the elements of an
electronic circuit together. In the past, it used materials such ash polycrystalline
silicon, but as circuits have become smaller this has meant that the electrons
can spend up to half of their time travelling through the material rather
than doing useful work. The resistance of the interconnect had to be reduced,
or the advantage gained by shrinking the device would be lost.

At first sight, silicides such as cobalt disilicide seemed to be the
obvious choice for the next generation of interconnects. They are as conductive
as metals and structurally similar to silicon, which makes them compatible
with the technology used to process silicon and turn it into useful devices.
But industry found that, using conventional techniques such as MBE and UHV,
it could not grow epitaxial films of single crystal silicide on the silicon
that it uses for the majority of its commercial chips. Instead, the films
were polycrystalline, which reduced their conductivity and gave them a ragged
boundary, or interface, with the substrate. This meant that circuit designers
could not be confident of the way that the material would behave in use;
they need homogeneous materials on which to base their designs. Also, silicide
films, because they are on the surface, fill silicon’s processing technologists
with horror – the technologists worry that the metals will contaminate their
production lines. During production, metal atoms could evaporate from the
surface and deposit themselves in equipment where they would be picked up
later by other circuits passing along the production line.

What the team at AT&T Bell did was to show that ion beam synthesis
enabled engineers to grow the silicides they needed, which were those with
the same orientation as the substrate, in the places they wanted, which
was below the surface rather than on it, thus reducing the risk of contamination.

Although growing silicide layers as interconnects is an important application
of the technique if ion beam synthesis, it is by no means the end of the
story. For several years, researchers have wanted to be able to use layers
of silicides to build other features of electronic circuits such as transistors,
which, working as switches, would make circuits operate faster and more
efficiently. Since the transistor’s invention, manufacturers have tried
to increase the rate at which the device can switch a current on and off.
One way of doing this is to reduce the time that electrons spend travelling
from one end of the transistor to the other, from the ’emitter’ to the ‘collector’
through the ‘base,’ either by making the base thinner to shorten the distance
they must travel, or by lowering the base’s resistance to the flow of electrons.
Because silicides are so conductive, researchers are convinced that they
will be able to build transistors with faster switching speeds. Instead
of doping regions of silicon with different impurities to form the junction
in the switch, as manufacturers do for conventional transistors, the junction
in the new device would be the boundary between the silicide and the silicon.
Researchrs could also use ion beam synthesis to build other features of
a transistor, such as the device’s collector, which would then be capable
of gathering the electrons more efficiently. Peter Hunt, a researcher at
Plessey, which is now owned by GEC, designed a hypothetical transistor with
a collector built from silicides. He estimated that the device would operate
at least five times as fast as the fastest silicon transistor now available.

Researchers have also successfully used ion beam synthesis to fire oxygen
ions into silicon, a technique known as separation by implanted oxygen,
or SIMOX. Implanting oxygen under the surface of a silicon wafer creates
an insulating layer that can be used to isolate one circuit element from
another and so allow them to be placed much closer together.

Over the past 18 months, industrial laboratories in Europe, Japan and
the US have used the technique to build computer memory circuits that operate
twice as fast as those made in ordinary silicon wafers. This year, SIMOX
chips were on sale for the first time, and now space scienists reckon that
the next generation of satellite will use them. The buried insulating layers
can be used to protect the electronic circuits from radiation in space;
they will make the chips ‘Radiation-hard.’

SIMOX substrates consist of a thin layer of silicon dioxide, between
400 and 500 nanometres deep, sandwiched between a thin overlayer of single-crystal
silicon, between 100 and 500 nanometres deep, and a thick layer of single-crystal
silicon, about 145 micrometres deep. In most silicon microships, only the
top 100 to 500 nanometres of silicon contain integrated circuits; the rest
of the wafer, which is about 500 micrometres thick, is for mechanical support.
The trouble is that the thicker the silicon the higher the electrical resistance.
In SIMOX structures, the buried silicon dioxide insulates the overlayer
from the support layer below. This means that the currents in the electonics
circuits confront less electrical resistance and so operate faster and consume
less power than similar circuits made in ordinary silicon.

Researchers are now trying to improve the technique. They want to make
the silicon overlayer even thinner – less than 100 nanometres deep – and
use every single bit of it. In the past, the silicon in the overlayer closest
to the oxide was of poor quality and contained particles, or inclusions,
of the oxide, which tended to trap charges. But recent advances in implantation
and annealing have enabled researchers to improve the crystalline structure
of the silicon overlayer and to form abrupt interfaces between the layers
of silicon and silicon dioxide. The results, from using devices made this
way have surpassed expectation; the devices operate up to twice as fast
as conventional SIMOX ones.

SIMOX is also useful for forming small islands of silicon surrounded
by silicon dioxide, which simplifies chip fabrication and produces a structure
that circuit designers can more easily tailor to what they want. Manufacturers
lay a thin film of silicon dioxide onto the surface of the wafer and then
etch holes into this oxide layer where they want the islands to be; these
islands provide the sites for the circuit’s components. Once this is completed,
a particle accelerator fires oxygen ions at the silicon wafer. In the unetched
regions, the ions only just penetrate the oxide layer to oxidise the silicon
beneath; in the etched regions, they travel deeper and form an oxide layer
agout 300 nanometers below the surface. Researchers control the etchant
so that the edges of the etched holes are angled; this allows the oxygen
ions to penetrate to intermediate depths at the periphery and so produce
continuous insulation around the island. John Alderman, a researcher at
Plessey, patented the idea for this type of structure in 1985. Since the,
researchers at Plessey, the University of Surrey, British Telecom and the
US Departament of Defense have been working together to design fast, radiation-hard
circuits based on the technique.

But ion beam synthesis is not a technique for use only by electronics
engineers. It seems ideally suitable for producing rare minerals, which
is how we are trying to use it in collaboration with the researchers at
the Natural History Museum, and for modifying the composition of superconductiong
materials, which is what we have begun to do at University College. We are
awaiting results of the implantations of fluorine into the superconducting
material, neodymium copper oxide. In the case of silver ores at the Natural
History Museum, we have managed to implant sulphur into hessite, and tellurium
into acanthite. We are trying to produce a compound similar to the natural
mineral cervelleite, so that Criddle and Stanley can try to work out what
causes the increased photosensitivity of the sample they found in Mexico.
Early research on the treated hessite had not produced any meaningful results
yet, but studies using the treated acanthite suggest that migrating tellurium
atoms may be chiefly responsible for the unusual tarnishing of the silver
ore. When the illuminate the treated acanthite, the two researchers find
that the implanted tellurium migrates several millimetres over the surface
of the mineral. This migration is accompanied by a change in the mineral’s
surface properties and an increase in its sensitivity to light – for the
experiment, the acanthite used was insensitive to light.

Industry is interested in ion beams synthesis because it is an inherently
clean technique – it buries the ions deep in the silicon substrate. So,
manufacturers can use it without worrying that they will contaminate the
devices they are producing. Another advantage is being able to fabricate
structures that are difficult to produce by more coventional means. Furthermore,
simply by changing the energy of the ions fired into the substrate, manufacturers
can build multi-layered structures for three-dimensional circuits. It may
soon be possible to use the silicon between two buried oxide layers, lying
at different depths below the surface, as a channel for beams of light.
Manufacturers could then build circuits in the top silicon layer and use
light to switch them on and off rapidly.

Putting a stop to high-speed ions

In ion beam synthesis, a particle accelerator fires ions at a crystalline
substrate at speeds of around 1500 kilometres per second, which would be
fast enough to take them around the world in less than half a minute. Once
they are within the substrate, they lose their kinetic energy to the material’s
crystal lattice in two ways, known as electronic and nuclear stopping.

Electronic stopping is like the slowing down of a small ball bearing
in a jar of treacle as the viscosity of the treacle dissipates the ball
bearing’s energy through friction. Similarly, the implanted atoms lose energy
by exciting electrons within the silicon.

Nuclear stopping is similar to what happens when the ball bearing strikes
an ordered array of billiard balls in the jar of treacle and transfers its
energy to one or more of them. Likewise, the implanted atom can lose energy
through collisions with the atoms of the silicon lattice. Displaced atoms
of the lattice can, in turn, collide with others still in the lattice. This
leads to a trail of destruction along and around the path of the implanted
ion, known as the collision cascade.

By incorporating these ideas into a computer model, researchers can
simulate the path of an implanted atom. As more and more ions strike the
target, each generating its own collision cascade, there is an increasing
likelihood that individual cascades will overlap. In semiconductors, this
process can completely destroy the material’s crystal structure in the regions
where implantations have occurred, rendering the material amorphous. In
ion beam synthesis, manufacturers of semi-conductor devices do not want
this to happen because any amorphous material, when heated, becomes polycrystalline
and non-homogeneous, which makes it less conductive and less easy to control.

To overcome this problem, engineers implant ions at temperatures of
between 300 and 700 °C so that there is sufficient energy available
during the process, in the form of heat, to allow many of the displaced
atoms to jump back into their original positions in the lattice. This helps
to reverse any damage done to the crystal structure. After implantation,
to reduce the damage further, engineers heat the material to between 1000
and 1400 °C. This later heat treatment also allows the new buried layer
to scavenge any remaining implanted atoms, from the silicon overlayer and
substrate, to form compounds of the correct chemical composition, or stoichiometry.

Karen Reeson and Russell Gwilliam are members of the Department of Electronic
and Electrical Engineering at the University of Surrey.