Special Reports

More than 30 years ago, Richard Feynman amazed physicists with his vision of the future. 'Consider the final question as to whether, ultimately - in the great future - we can arrange atoms the way we want; the very atoms, all the way down! What would happen if we could arrange atoms one by one the way we want them?' Feynman was speaking at a meeting of the American Physical Society on 29 December 1959.

What has happened is that scientists have started indulging in microscopic graffiti. The instrument that makes this possible is the scanning tunnelling microscope - invented 11 years ago to produce images of surfaces showing the arrangement of individual atoms. In the past few years scientists have been using the extremely fine tip of the microscope to modify surfaces as well. The temptation to leave their mark in messages only a few atoms high is irresistible.

The graffiti craze was begun in April 1990 by Don Eigler, an IBM researcher working at its Almaden Research Center in San Jose, California. He made the world's smallest advertisement by writing IBM in letters 5 nanometres high - 500 000 times as small as the letters on this page. With the tip of an STM, Eigler arranged 35 xenon atoms into the three letters on the surface of a piece of nickel.

Since then, creative microscopists have turned their tips to everything from antiwar slogans to Valentine messages. Shojiro Asai from the Hitachi Central Research Laboratory in Tokyo produced 'Peace 91' just before the outbreak of the Gulf War, while Munir Nayfeh of the University of Illinois dedicated his heart simply to 'P' (as in Physics). The microscope artists then turned to more representational forms: the Stanford logo, a map of the world and a molecular man.

Though their techniques differ, all the artists rely on the tip of an STM or one of the family of related devices. The tip is extremely fine; sometimes only one atom wide at the point. It also has very precise controls so that researchers can move it over a surface accurately enough to identify individual atoms.

While scrawling miniature messages is fun, the artists have day jobs too. They are developing a new kind of electronics, relying on quantum mechanics and the movements of single particles, which will one day produce devices many times faster and smaller than anything around now. Researchers talk about storing the entire contents of the US Library of Congress on a silicon disc 30 centimetres across, or of producing a computer the size of a pocket calculator but with the power of the biggest supercomputer.

Reseachers in this field, known as nanotechnology, hope that one day they will be able to make tiny sensors that could be implanted inside the body and monitor the constituents of blood, for example. They also foresee minia-ture robots that will travel in the bloodstream and clear blockages in arteries or repair damaged areas of the brain. None of these things will appear until well into the next century, but the precision of the STM and its relatives is allowing scientists to take the first tentative steps towards them.

The easiest way to modify a surface, referred to as a 'tip crash', is to dig the tip into the surface and make an indentation. This usually occurs by accident but Thomas Jung refined the process to write 'Heureka' between the tracks of a compact disc - a technique that Heinrich Rohrer, one of the STM's inventors, dubbed the 'nanometre plough'. Jung, who works at the University of Basel, used a variation of the STM called an atomic force microscope. He rested the tip of the microscope on the disc and applied a small downward force so that it dug into the surface. He then dragged the tip along, vibrating it up and down and side to side thousands of times per second to produce grooves.

Jung says that his plough could be used to store data in a much more compact form than any existing method, as he demonstrated by inscribing 'Heureka' on the CD. Each of the elongated pits either side of the inscription represent a one in the digital code normally used to store data on the disc. A single letter would require eight such pits, so 'Heureka' would need 56. CDs are one of the most dense data storage methods now in use. The main problem with Jung's technique is speed; the tip moves far too slowly for it to be practical for data storage.

Data storage is an obvious use for STMs and their relatives. All that is needed is for the tip to make some kind of mark in a regular pattern - blobs on the surface or the absence of them - to represent a one or zero in digital code. The microscope can then be used in its normal 'imaging' mode to detect the row of blobs and retrieve the data.

Over the past few years, Dan Rugar and Jonathan Mamin have made a concerted effort to develop a memory device from an STM. Like Eigler, they work for IBM at its Almaden Research Center. In their technique, the blobs that represent digital bits of data are blobs of gold - each made up of a few thousand atoms. They showed off its cartographic potential by drawing a map of the world on a scale of 1:10 000 000 000 000.

To deposit their blobs they hold a gold STM tip close to the surface and apply a voltage pulse to it. The pulse causes gold atoms of the tip to become ionised and the electric field between the tip and surface makes them fly off onto the surface. The tip can deposit blobs as it scans over the surface. The pulses can be very quick. It works reasonably reliably, says Rugar, with pulses as short as 0.9 nanoseconds, less than a thousand millionth of a second.

To read back the data, Mamin and Rugar simply use the tip like a microscope and scan over the blobs to detect what is there. They can also erase the data by picking up the blobs: a voltage pulse in the opposite direction makes the gold atoms fly back onto the tip. Importantly, the whole process occurs at room temperature and in air; many STM techniques, such as Eigler's, must be done in a vacuum chamber at a very low temperature.

The IBM researchers' technique can squeeze a huge amount of data into a very small space. They can store 1 terabit of information (1 million million dots) on a square inch of surface (6.5 square centimetres). The complete works of Shakespeare could be stored on a square less than 0.2 millimetres across, giving a data density about 10 000 times better than the best magnetic computer discs on the market.

Mamin and Rugar's technique records bits as physical blobs that can be seen, just as the pits on a compact disc are visible. But digital bits recorded on magnetic discs or tape are simply zones of a particular magnetic orientation, invisible except to a magnetic sensor. It is possible to use similar approaches on a smaller scale using an STM tip as a sensor.

Just such a device, developed last year, uses the tip of an atomic force microscope to create and detect zones of trapped charge beneath the surface of a semiconductor. Its developer, Robert Barrett, formerly at Stanford University in California, used his technique to produce a tiny tribute to his former employers: he inscribed the Stanford logo in trapped charge. Dieter Pohl, head of STM research at IBM's Zurich research laboratory, says: 'It is the best method I know for getting closer to the specifications necessary to design a commercial memory device.'

To make his task easier, Barrett used a semiconductor material that had been used for memory devices in the past, so its properties are well known. Called a nitride-oxide semiconductor, it consists of a silicon substrate covered with a thin layer of silicon dioxide and topped off with a layer of silicon nitride. Silicon dioxide is an electrical insulator so it blocks the movement of charge between the other two layers. But if a strong vertical electric field is applied to the material, some charge can penetrate the silicon dioxide layer.

Barrett placed the microscope tip on the silicon nitride layer and applied a pulsed field through it. This makes the field beneath the tip draw charge from the silicon substrate, up through the insulating layer and into the silicon nitride layer, where it becomes trapped. This small zone of trapped charge then acts as a bit of data. To detect the zones and read back the data, Barrett moved the tip over the surface while keeping them in contact. He applied a small voltage to the tip so that a small current flows between the tip and the silicon nitride layer. This current changes as the tip passes over a zone of trapped charge because the charge reduces the capacity of the material to store charge at that point.

To erase a bit of data, Barrett simply positions the tip over the trapped charge and applies a pulsed electric field in the opposite direction, pushing the charge back through the silicon dioxide layer into the substrate. Barrett's technique can store data at a density of about 2 gigabits (2 thousand million zones) per square centimetre. This is 150 times denser than commercial magnetic discs.

Both these memory techniques are limited by the speed of the tip. Barrett's device reads data 100 times slower than a magnetic computer disc. Rugar and Mamin are working on a fast moving STM that has an extra servomechanism to lift the sample up and down very quickly to help the up and down motion of the tip. The tip can then scan at a speed of1 millimetre per second, a thousand times faster than a conventional STM tip. With it they have been able to make films (at two frames a second) of events on the surface in which layers one atom thick can be seen moving spontaneously.

But electronics needs more than just memory devices, and STMs can be used to make, on a miniature scale, the other essential components: conducting wires with which to build circuits, and transistors, the switches at the heart of digital electronic devices.

Several groups of researchers are using STM tips as a pen to write conducting lines on a surface. In one approach, researchers apply a voltage to the tip and move it near the surface, where it creates a strong electric field over a very small area. The field guides metal ions, the 'ink', into this small area of the surface. Moving the tip produces a line of metal on the surface. Nayfeh used such a pen to dedicate his heart to physics.

To provide the metal ion ink, Nayfeh put his STM inside a vacuum chamber and filled it with trimethyl aluminium gas. He then shone a laser beam in the gap between the tip and the graphite surface onto which he wanted to inscribe his Valentine message. The electric field in the gap and the laser light combine to ionise the gas. The field guided the ionised aluminium down to the surface.

An alternative approach, adopted by Alex de Lozanne from the University of Texas at Austin, is to adsorb a gas containing metal atoms (nickel carbonyl in his case) onto a surface. He applied a sufficiently high voltage to his STM tip so that it emitted a beam of electrons towards the surface. The flow of electrons broke up the gas on the surface leaving a deposit of nickel. By scanning the tip he produced lines that were 50 nanometres across and 95 per cent metallic nickel.

Switches, however, are far more difficult to make than bits of data or conducting wires. They can have many parts that need to be positioned very precisely with respect to each other. Last August, Eigler announced the simplest and probably the smallest switch that it is possible to make (Nature, vol 352, p 600). The on and off states of his 'atomic switch' rely on the position of a single xenon atom.

The atom can be switched between two stable positions, representing on and off. One position is in a kink on the surface of a crystal of nickel, the other is on the tip of an STM held still just a few atomic diameters above the surface. By applying a short pulsed voltage to the tip, Eigler could make the xenon atom jump across the gap from the surface to the tip. A pulse in the other direction made the atom jump back down onto the surface

Poles apart

As the STM tip is used as one of the poles of the switch, it cannot scan over the surface to see whether the xenon atom is in the kink or not. To find out which state the switch was in, Eigler applied a steady and much lower voltage to the tip so that a small current flowed between the tip and the surface. He found that the current increased when the xenon atom was attached to the tip rather than to the surface.

Eigler is quick to emphasise that the atom switch is much more of scientific curiosity than a practical switch and his main concern is to find the mechanism behind it. He believes that electrons, carried by the voltage between the surface and tip, knock against the xenon atom as they pass by. These collisions carry the atom along with the electrons - just as water can roll a stone along a river bed.

While the atom switch is a good demonstration of what is possible with an STM, it is unlikely ever to form a useful device. The xenon atom is held onto the surface only by very weak chemical bonds. This means that the whole process has to take place at temperatures as low as 4 Kelvin so that thermal vibrations do not bounce the xenon around, and in a vacuum so that gas molecules do not disturb it. This makes the whole apparatus very bulky. Modern semiconductor chips cram millions of switches into an area less than a centimetre square and are tough enough to survive in a digital watch or portable computer.

Eigler says that his group is now looking for molecules with the structure of a ready-made switch that would not need the STM tip as one of its electrodes. He wants to investigate cage-like molecules, such as zeolites, that can be manufactured chemically with a metal atom trapped inside. If the structure of the cage had two stable positions for the metal atom to reside in, voltage pulses could switch the atom from one position to the other. Eigler believes that he could then pin down such a molecule on a surface and wire it up with microscopic wires into a tiny switching circuit. The STM would be used 'as a tool to wire up the molecules', he says.

An alternative to a single atom switch is one that relies on a single electron. American researchers made such a transistor several years ago, employing techniques used to make advanced semiconductor chips. Robert Jaklevic, however, has built a single-electron device incorporating an STM tip similar to the way Eigler used a tip as one of the electrodes of his atomic switch. Jaklevic, of the Ford Motor Company's Scientific Laboratory in Dearborn, Michigan, was working with colleagues from the University of Michigan at Ann Arbor.

Jaklevic's team took a semiconductor substrate coated with a thin layer of an insulating material and sprayed minute blobs of indium onto it. It then used the STM tip to locate one of the tiny blobs and positioned the tip over it. The gap between the tip and the blob forms a 'tunnel junction'. This means that in theory the gap forms a barrier and electrons cannot cross it, but quantum mechanics allows some electrons to cheat and tunnel through the barrier. The ease with which electrons can tunnel depends on the size of the gap. The researchers set the gap to make it most likely that electrons would tunnel one at a time from the tip to the blob.

But because the blob is so small, even a single electron tunnelling across to it changes the blob's voltage significantly. This voltage prevents any other electrons from tunnelling from the tip to the blob. The phenomenon is known as a Coulomb blockade. The thin insulating layer between the blob and the substrate acts as another tunnel junction. By manipulating the voltages applied to the tip and the substrate, the researchers can control the movement of single electrons in and out of the blob.

Just as in the atomic switch, it is cumbersome to have the STM forming part of the device. It would be better to make a device flat on the surface of the substrate. Jaklevic and his colleagues are patenting a device that would consist of an array of tiny regions of conducting film fabricated on a substrate and insulated from each other by tunnel junctions. By varying the voltage applied to each region of film it would be possible to move single electrons around from one region to the next. Such a device could be used as a memory, with empty or occupied regions representing digital zeros and ones, as a logic circuit, or as an image detector where light falling on a particular region affects the Coulomb blockade.

The Michigan researchers have not yet made this device because there is no process available to create such tiny and intricate structures. Conventional fabrication techniques are reaching the limits of miniaturisation. 'Right now things look very tough,' Jaklevic says. Using an STM as a tool, rather than as part of the device itself, would be a good way to make experimental devices, he says. 'They have a unique property to make very very small structures.' But the slow speed of the STM tip makes it unsuitable for creating large arrays of devices on a single substrate or for mass-producing chips.

The ability to draw very fine lines with an STM tip may make possible another entirely different type of device. Under certain conditions, according to quantum mechanics, electrons can behave like waves rather than particles. If electrons show wave-like behaviour, such as diffraction, refraction and interference, then a device exploiting this behaviour could control the flow of electrons.

Electrons do not normally behave like waves in electronic materials because to do so they need to move ballistically, as if flying through free space. Normally, vibrations of the atoms in the material scatter the moving electrons so that they do not have a chance to move ballistically for long enough. But at very low temperatures, the vibrations are reduced so some wave-like behaviour is possible.

Researchers at the University of Illinois at Urbana-Champaign, including Nayfeh, are using wave-like electrons to develop a device called a quantum interference transistor. They start with a silicon substrate with two contacts on its surface. The contacts are linked by a channel under the surface, which is made of silicon containing impurities to improve its conductivity. Electrons can flow from one contact to the other through the channel. When liquid nitrogen cools the device to a temperature of -196 °C, the electrons behave as waves with a wavelength of about 10 nanometres.

Between the contacts, in the area over the channel, is a a coating of silicon dioxide, which is an insulator. The researchers plan to use a variation of Nayfeh's technique, which he used to draw lines of aluminium, to draw lines of thicker insulator in a pattern of parallel lines very close together.

The researchers will then cover the striped insulator with another metal contact. Applying a voltage to this contact projects an electric field down through the insulator into the silicon. The insulator weakens the field that reaches the silicon, so the alternating pattern of thick and thin insulator produces an alternating strong and weak field along the channel in the silicon.

When the electron waves meet this alternating field, some will be reflected. When these reflected waves run into more oncoming waves they destructively interfere - the troughs of one wave meeting the peaks of the other will cause the waves to cancel each other out. By adjusting the voltage to the central contact, and hence the strength of the alternating field, they can control the flow of electron waves along the channel or stop it altogether.

At the moment, the closest the researchers can draw the lines is 50 nanometres apart but they need to be as close as the wavelength of the electron waves, 10 nanometres. Joe Lyding, one of the researchers, says that they hope to be able to do this by carrying out the writing in a vacuum rather than in air. Lyding says that they expect to have a working device towards the end of this year. Such quantum interference transistors would be faster than conventional transistors because they are so small - the electrons would have less distance to travel. They would also use less energy because the electrons are scattered less by the atoms of the material. A large part of the energy expended in conventional devices is caused by scattering.

Despite the speed of developments, Lyding believes it will be between 20 and 30 years before devices such as this are commonplace. The problem with using an STM tip as a tool is, as always, speed. If conventional manufacturing techniques can be improved to produce tiny single-electron or quantum interference transistors, then the STM will be invaluable for making experimental devices and prototypes that can then be copied in large quantities.

But the joy of pushing atoms around still remains. 'I just think there's something conceptually boggling about taking an atom under control and saying 'I want it over here or I want it over there',' says Eigler. 'You don't have to look to some glory in the future, the excitement is now.'