DIGITAL pictures eat up data storage space like a plague of locusts. That is one reason why today鈥檚 PCs, with their limited hard disc space, are not ideal for handling video. But imagine a disc that can store 20 times as much data as now. What brave new multimedia worlds would that open up?
Keep hold of that vision, because reality may only be a few years behind. Computer makers around the world are experimenting with intricate metal structures that respond to the tiniest magnetic fields. Built into the heads that read hard discs, they will bring huge increases in the amount of data you can store. By the year 2000, IBM promises, these read heads will be at the heart of the latest computer technology.
What is special about the materials responsible for this coming revolution is that their electrical resistance changes when they are placed in a magnetic field. This effect, called magnetoresistance, is not particularly new, and it is already widely used in the heads that read data off modern hard discs. Each data bit stored on a disc is represented by a minute magnetised area on its surface. As the surface of the spinning disc passes beneath the read head, it produces a changing resistance, and hence a changing voltage in the head, revealing the stored magnetic pattern. In a modern hard disc, the fluctuating magnetic field changes the electrical resistance of the read head by some 2.5 per cent. But the new materials show resistance swings of more than 200 per cent 鈥� an effect dubbed 鈥済iant magnetoresistance鈥� or GMR.
Advertisement
This means that read heads will be able to sense much smaller magnetic fields than before 鈥� allowing data to be imprinted on smaller magnetic particles, and making it possible to pack information in at much higher densities. At the moment, the highest storage density on the surface of a disc is around 75 megabits per square centimetre. Using materials that show GMR, researchers at IBM鈥檚 laboratories in Almaden, California, are developing a head that can read information that is 20 times as dense: 1.5 gigabits per square centimetre. IBM says that these new heads will be in use by the end of the century.
GMR was discovered in 1988 by Albert Fert and his colleagues at the University of Paris-Sud. They measured the resistance of a 鈥渕agnetic multilayer鈥� consisting of 40 ultrathin layers of iron, which is magnetic, interlaced with even thinner layers of chromium, which is nonmagnetic. They found that the resistance of the multilayer fell by almost 50 per cent when it was placed in a magnetic field. The effect can be understood in terms of a simple model (see 鈥淭he story of GMR鈥�). Put simply, the resistance is high when the magnetisations of alternate magnetic layers point in opposite directions. A strong enough external magnetic field can ensure that the fields in all the layers point in the same direction, so the resistance falls.
GMR is generally represented as the change in resistance of a material, expressed as a percentage of a material鈥檚 lowest resistance. The present record for a magnetic multilayer made of iron and chromium is held by Yvan Bruynseraede and his colleagues at the Catholic Uniyersity of Leuven in Belgium. Their multilayer contained 50 layers of iron, each 0.45 nanometres or about three atoms thick. They interleaved these with 1.2-nanometre layers of chromium. When this structure was cooled to 1.5 K, applying a magnetic field of 11 tesla increased its resistance by 220 per cent.
This effect begins to look insignificant, however, next to a phenomenon discovered in 1994 by Venky Venkatesan and his colleagues from the Center for Superconductivity Research at the University of Maryland, College Park. These researchers prepared films of neodymium, strontium, manganese and oxygen 200 nanometres thick, and found that they exhibit a GMR of 3340 per cent in a magnetic field of 5 tesla. Heating the material to 900 掳C in oxygen for 30 minutes increased the GMR to more than 1 million per cent. Hyperbole seems to be the order of the day: this phenomenon is called colossal magnetoresistance.
Spectacular as this effect is, most researchers believe colossal magnetoresistance materials are not yet the best choice for applications. The phenomenon is highly temperature dependent and the materials that exhibit it are not well understood. Moreover, size isn鈥檛 everything in GMR: sensitivity is much more valuable. A potentially huge increase in resistance isn鈥檛 much use if you have to apply an enormous magnetic field to get it. The magnetic field of 5 tesla used in the colossal magnetoresistance experiments, for instance, is more than 80 000 times the magnetic field of the Earth at the North Pole. The fields generated by the grains on a hard disc are typically only around 1 millitesla.
In the quest to find practical GMR materials, multilayers are not necessarily the best bet. Researchers have found that structural defects in the layers and stray magnetic fields 鈥� such as the Earth鈥檚 鈥� can neutralise the effect of GMR. Alternatives that avoid this problem include clusters of magnetic metals randomly distributed in a nonmagnetic material. One candidate combination is cobalt, which can form nanometre-sized clusters, deposited in silver. With no external field, the magnetisations of the individual cobalt clusters point in different directions. Applying a magnetic field makes them align in the same direction, reducing the resistance of the structure. These alloys are much easier to make than multilayers, but they have a big drawback: very high magnetic fields are needed to align them.
Another promising structure is made up of regular rows and columns of clusters. In 1993 Todd Hylton and his colleagues at Adstar in San Jose, California, slowly heated multilayers of silver and permalloy, a magnetic alloy made of 80 per cent nickel and 20 per cent iron. When originally produced, this structure exhibits virtually no GMR. But the heating causes the silver atoms to diffuse into the permalloy layers, and when the sample is cooled, the magnetic layers form disc-shaped clusters. In a magnetic field, the resistance of this material rises by an unspectacular 5 per cent. But more importantly, this happens at room temperature in a field of less than 1 millitesla 鈥� a record sensitivity at this temperature.
The latest GMR structures are nanowires. In 1994 Luc Piraux and his colleagues at the Catholic University of Leuven grew wires that are 10 micro-metres long and 40 nanometres thick, made of alternating layers of cobalt and copper. They started with a polymer slab, which they bombarded with energetic argon ions and then etched with chemicals to produce hairlike pores in the surface. They then placed the slabs in an electrolysis cell in a solution containing copper and cobalt ions. By carefully controlling the voltage across the cell, they grew stripy wires of copper and cobalt in sequence within the pores (see Diagram, bottom left). These nanowires show a GMR of 19 per cent at a temperature of 4.2 K. Moreover the GMR drops only slightly 鈥� to 15 percent 鈥� at 290 K (17 掳C). This is less spectacular than the effects seen in multilayers, and the field needed is a rather high 0.5 tesla. But the nanowires are relatively cheap to produce and this makes them promising candidates for development.
Building sensitive read heads for a new generation of hard discs is not the only application on the cards for GMR. There is also the possibility of using GMR materials to build 鈥渘on-volatile鈥� solidstate memory 鈥� a type of random access memory (RAM) which does not lose data when the power is turned off. These magnetoresistive memory devices would be useful in pocket organisers, for instance, which do not have room for bulky disc drives. Devices along these lines are being developed by Nonvolatile Electronics in Minneapolis. They have a strongly magnetic base layer, made of an iron or manganese alloy, surmounted by a reference layer with weak magnetisation, such as permalloy, then a nonmagnetic layer such as copper, and finally another weakly magnetisable layer. This top layer is cut into an array of squares or rectangles that can each store a single bit of data.
Magnetoresistive memory stores a 0 when the magnetic fields of the reference and top layers point in the same direction, and a 1 when they point in opposite directions. If there is no external magnetic field, the reference layer has the same magnetisation as the base. Small fields created by a current running nearby can flip the direction of the field in the top layer but not the reference layer, and so change the state of the memory. The memory is read by applying a voltage across each square to find out if the resistance is high or low.
Materials that exhibit GMR are also being used to develop more efficient transistors, the building blocks of microprocessors. These transistors take advantage of the fact that electrons have a quantum property called spin that generates tiny magnetic dipoles. These line up either with or against the magnetic fields in the multilayers. Electrons with spins parallel to the magnetisation are called 鈥渟pin-up鈥� and can flow fairly easily through that layer. Those with antiparallel spins, called 鈥渟pindown鈥�, meet with high resistance. Mark Johnson of the US Naval Research Laboratory in Washington DC is exploiting this in the 鈥渟pin transistor鈥�. His device is made of three layers: a magnetic layer, equivalent to the emitter in a conventional transistor, a middle nonmagnetic layer that corresponds to the base, and another magnetic layer which acts as the collector.
A voltage applied between the emitter and base causes a build-up of either spin-up or spin-down electrons 鈥� whichever move through the emitter more easily 鈥� in the base. The magnetisation direction of the collector layer can be set by running a current through a nearby wire. If emitter and collector magnetisations point in the same direction, current can flow from the base to the collector. But if the magnetisations point in opposite directions, electrons in the base meet with high resistance from the collector. In this case, the current flows from the collector to the base. The device acts as a magnetic switch that could become the basis of a metallic computer.
Spin transistors could have advantages over semiconductor transistors, says Johnson. They are smaller, and could be much less power hungry. Semiconductor transistors continuously consume power 鈥� typically about 50 microwatts 鈥� but spin transistors draw this amount of power only when they are switching.
Researchers still have a lot to learn about GMR. Its effects in tiny films less than 1 micrometre thick are still extremely unpredictable. And to make GMR materials to order, researchers will need a clearer idea of how manufacturing processes affect their properties.
When GMR materials make it into the marketplace, they are sure to find applications beyond computers. According to Sam Bader of the Argonne National Laboratory in the US, GMR materials could monitor currents in transmission lines by sensing the magnetic fields they produce. This would avoid the need for direct connections to the high-voltage lines. Bader predicts that GMR sensors will turn up everywhere from aircraft to kitchens, in control systems for electric motors. At the start of a brand new millennium, GMR is sure to be taking the technological world by storm.
The story of GMR
The first hint of giant magnetoesistance emerged just 10 years ago. In 1986, Peter Gr眉nberg of the J眉lich Research Centre in Germany was investigating the properties of a magnetic multilayer made up of two layers of iron with a chromium layer sandwiched between them. He noticed that the magnetisations of the iron layers aligned in some samples, but pointed in opposite directions in others. The only obvious difference between the different samples was the thickness of the chromium layer. He also found that the antiparallel fields could be forced to point in the same direction if the multilayers were exposed to a strong enough magnetic field.
Two years later, Albert Fert and his colleagues at the University of Paris-Sud found that in a magnetic field of 2 tesla, the resistance of a more complex magnetic multilayer they were investigating fell by almost 50 per cent. Later work by both Gr眉nberg and Fert鈥檚 group showed that their two discoveries seemed to be related: GMR comes into play when the applied magnetic field forces the multilayer鈥檚 fields to align in the same direction.
The change in resistance can be explained by the way an electric current flows through the magnetic layers (see Diagram). The electrons that form the current have a property called spin. This generates tiny magnetic dipoles that line up with or against the magnetic fields in the layers, depending on the direction of the spin. Electrons with spins parallel to the magnetisation are called 鈥渟pin-up鈥�; those with antiparallel spins are called 鈥渟pin-down鈥�.
An electron does not meet with much resistance when its spin axis is aligned with the magnetisation in the layer it is passing through. If the next magnetic layer is magnetised in the same direction, the resistance stays low, but if it points in the opposite direction, the resistance rises. So if the fields in every magnetic layer of a multilayer point in the same direction, the resistance for the spin-up electrons is low in every layer, while the resistance for the spin-down electrons is high.
The spin-up and spin-down electrons can be thought of as forming two separate currents, flowing in parallel. When a low and a high resistance are connected in parallel, the result is a low resistance. But if the multi-layer鈥檚 fields align in opposite directions, then both resistances are fairly large, and the combined resistance in parallel is also large. This is the origin of giant magnetoresistance.
The two-current model envisages the current as flowing through the multilayers at right angles to the layers 鈥� a configuration known as current-perpendicular-to-plane (CPP). This is the situation in some of the latest materials, such as the chains of alternating materials known as nanowires. But in most experiments with multilayers, the current flows along the layers. This is known as the current-in-plane (CIP) geometry. But curiously, the two-current model is still a good guide. This is because the electrons do not stay on a constant course. They tend to change layers so often that the effect is similar to the CPP geometry.
Although the two-current model agrees well with experiment for both geometries, it is a pretty crude one. Spin is a quantum property, so theorists will not be happy until they have developed a clear and detailed quantum mechanical picture of GMR. Another mystery is what causes the fields of magnetic multilayers to flip from parallel to antiparallel simply because the spacer thickness changes. The spacer thickness required to reverse the fields turns out to be dependent on the spacer material: it is 2 nanometres for chromium, but roughly 10 nanometres for most other metals. Working out why is another challenge for the physicists.