Giant Magnetoresistance in
Layered Magnetic Materials

By Bill Butler, Xiaoguang Zhang, and Don Nicholson

photo

Bill Butler (front), Xiaoguang Zhang, and Don Nicholson use advanced computing techniques to model the giant magnetoresistance effect in magnetized materials. Photograph by John Smith.

By working at the atomic level, performing first-principle calculations of variations of electrical resistivity in layered magnetic alloys, ORNL researchers hope to improve magnetic storage systems. This research in the Metals and Ceramics and Computational Physics and Engineering divisions uses advanced computing techniques to model the giant magnetoresistance (GMR) effect in materials that have structures “layered” at the atomic scale. The GMR effect is a large decrease in electrical resistivity that occurs when the magnetization of two layered samples is aligned by an external magnetic field. A better understanding of GMR could lead to the development of increased data storage density on magnetic disk drives and nonvolatile thin-film magnetic storage devices for computer operating memory whose data would not be destroyed by ionizing radiation or power outages. The goal of this research is to help ORNL’s industrial partners develop new products that meet the needs of the Department of Energy and are competitive in the international marketplace.


 
Building Better Memory Devices

Computer users may no longer need to fear the loss of valuable data every time they hear thunder or see the lights flicker if a project under a new cooperative research and development agreement between ORNL, Honeywell Solid State Electronics Center, and Nonvolatile Electronics, Inc., is successful. Researchers at ORNL are working with their industrial partners to refine a new type of memory based on the giant magnetoresistance (GMR) effect that can survive power interruptions. The new memory system also can endure ionizing radiation, making it especially attractive for military and space applications.

The key to making the new type of memory work efficiently is to be able to deposit high-quality magnetic films. The entire film may be only 50 atoms thick, and some of the layers may be only the thickness of a few atoms. Honeywell and Nonvolatile have called on some of the unique capabilities at ORNL to determine how well these ultrathin films are being deposited so they can learn how to deposit them better. Four different types of experimental techniques are being used to “see” what these films are like at the atomic level.

Gene Ice and Eliot Specht, both scientists in the X-Ray Research Group in ORNL’s Metals and Ceramics Division, are using X-ray diffraction and reflection to get an overall view of the thickness of the films and the perfection of their crystallinity. So far, their work on the films has been performed in their laboratory at ORNL; later, some of the films will be investigated at the ORNL beamline at the Advanced Photon Source at Argonne National Laboratory.

Mike Miller, a scientist in the Microscopy Microanalytical Sciences Group of ORNL’s Metals and Ceramics Division, is using an atom probe field ion microscope to analyze the films atom by atom. Miller’s atom probe removes atoms from a sample one by one and determines their elemental identity and their atomic position. With this information he hopes to develop an atomic-scale picture of the structure of the GMR films.

Steve Pennycook and Yanfa Yan, both scientists with ORNL’s Solid State Division, are using a scanning transmission electron microscope to shoot a tightly focused beam of electrons through the films. The beam is focused down to a size that is as small as or smaller than an atom. By analyzing the intensity of the transmitted electrons, they can see columns of atoms, and by analyzing the energy of the transmitted electrons, they can perform a chemical analysis of these columns.

Edgar Voelkl and Bernard Frost, both scientists at the High Temperature Materials Laboratory, are looking at a different aspect of these ultrathin films. They are able to image the magnetic fields of the microscopic memory cells. This information is important to understanding how reliably and quickly information can be stored in the cells. The information about the structure of the films obtained by these experimental groups will be used to make more realistic models of the GMR effect.

The work at ORNL is supported by DOE’s Office of Energy Research, Laboratory Technology Research Program. The outcome of the work will be the development of better magnetic memory devices.

drawing

This image, obtained from an electron hologram, shows the magnetic fields outside magnetic memory cells in a prototype magnetic random access memory device that uses the giant magnetoresistance effect. The electron hologram was obtained by Bernhard Frost and Edgar Voelkl.

Most of us are familiar with electrons, the tiny negatively charged particles that carry current in a metal wire when it is connected to a battery. The negative charge of the electron is a familiar property, as is its small mass, which enables it to move easily through a metal. However, in addition to charge and mass, electrons have a third property called spin that is often overlooked. It has become possible to take advantage of the spin of electrons to control the way they move through thin metallic conductors. Exciting new technologies may emerge from this newly acquired ability to control electrons, such as computer memory systems that hold more data and are immune to ionizing radiation and power interruptions.

In a rough sense, you can think of an electron as a tiny negatively charged particle that spins about an axis like a child’s top. A spinning charged object creates a magnetic field. However, in most materials, for every electron that spins in one direction there is an electron that spins in the opposite direction so that their magnetic fields cancel exactly. Therefore, most materials are not magnetic. In a few metals—iron, cobalt, and nickel are the most common—electrons gain energy if most of them spin in the same direction. As a result, a net magnetic field is generated in that direction, making the material magnetic.

In the presence of a magnetic field, the laws of quantum mechanics dictate that an electron can spin in one of two directions: Either its own magnetic field, as a result of its spin, aligns with the external field, which physicists call the “up” direction; or its own field aligns opposite to the external field, which is the “down” direction. In a magnetic material most electrons are aligned in the direction of the net magnetic field—that is, there are more up electrons than down electrons. This distinction between up and down electrons is carried over to nonmagnetic materials, although the choice of direction is then arbitrary.

In a nonmagnetic material, the up- and down-spin electrons are expected to respond in exactly the same way to an applied electric field. However, in magnetic metals it was expected that the up- and down-spin electrons might respond quite differently. Unfortunately, there was no way to measure the electrical current carried independently by two kinds of electrons; only the sum of the two currents could be measured.

The discovery of “giant” magnetoresistance (GMR) in 1988 gave scientists the ability to detect and understand the two different kinds of electrons in a metal, opening the possibility of exciting new technological applications (see Applications of Giant Magnetoresistance). Magnetoresistance is a change in the electrical resistivity of a material that is caused by application of a magnetic field. Because magnetoresistance typically is quite small, it was surprising and quite exciting when a group of scientists led by Professor Albert Fert of the University of Paris reported that if a thin metallic film is made of alternating layers of iron and chromium, and if the chromium layers are the right thickness (about 6 atomic layers), the electrical resistivity of this film can be cut in half by placing it in a magnetic field. This GMR was associated with a change in the relative alignment of the net spins on two iron layers. When the net electron spins (or magnetization) on the adjacent iron layers were in opposite directions, the resistance was high; when they were in the same direction, the resistance was low.

Understanding GMR

To understand what might be causing this effect, we calculated how electrons behave in a layered iron-chromium system similar to the one investigated by Professor Fert’s group. We also looked at other systems that show GMR, such as layers of cobalt and copper. These calculations took advantage of a computational approach developed by Professor James MacLaren of Tulane University, called the “Layer Korringa Kohn Rostoker technique.” Korringa, Kohn, and Rostoker were scientists who, during the 1940s and 1950s, devised an elegant technique for calculating how electrons move through a periodic lattice—that is, a system in which all atoms are lined up in precise rows and columns like soldiers in formation. Professor MacLaren extended this technique so that it works very efficiently for layered systems. (Imagine that soldiers in some of the rows are wearing different uniforms). In the article, “Developing a Grand Challenge Materials Application Code,” Bill Shelton and Malcolm Stocks describe another technique based on the Korringa Kohn Rostoker approach, which is even more general.

drawing
drawing
Fig. 1. The number of down-spin electrons per atom hardly changes between the iron and chromium layers, but the cnumber of up-spin electrons is higher for iron than for chromium. Fig. 2. The number of up-spin electrons per atom hardly changes between the cobalt and copper layers, but the number of down-spin electrons is much lower for cobalt than for copper.

Figures 1 and 2 show the number of up- and down-spin electrons that we calculated for each atomic layer of a system consisting of eight chromium layers embedded in iron and for a system consisting of ten copper layers embedded in cobalt. Both of these systems show a large GMR. These figures give an important clue as to what is happening. For the iron-chromium layers, note that the number of down-spin electrons is about the same in the iron and chromium, but the number of up-spin electrons is very different. Conversely, the number of up-spin electrons is very similar on each layer of the copper-cobalt system, but the number of down-spin electrons changes abruptly at the interfaces.

Figures 1 and 2 provide a qualitative understanding of how the GMR effect might arise. The electrical resistance of a metal arises from irregularities and discontinuities in the atomic lattice potential, called defects, as seen by the electrons. Electrons carrying a current are like balls rolling down a hill. If the slope is smooth, then the balls go very fast. But if it has a lot of bumps, the balls will be slowed down. Defects in the atomic lattice are like those bumps. When electrons hit these defects—a process physicists call scattering—the electrons are slowed down. Therefore, scattering can generate electrical resistance.

drawing
Fig. 3. When excess spins in both cobalt layers are aligned, up-spin electrons pass from layer to layer without scattering.

Consider the case in which the ferromagnetic layers are cobalt and the nonmagnetic layer is copper. The up-spin electrons hardly notice any difference in the number of electrons per atom as they travel from the ferromagnetic layer to the nonmagnetic layer (see Fig. 1). To them the lattice potential is smooth and almost defect free. On the other hand, the down-spin electrons see a large difference in electron numbers between atoms of copper and cobalt. They see many bumps at the interface (because copper and cobalt atoms often mix with each other there), so they are very likely to scatter there. When excess spins in the two ferromagnetic layers have the same direction (illustrated by the left-hand side of Fig. 3 for cobalt and copper), up-spin electrons can travel freely from one ferromagnetic (cobalt) layer across the interface to the other ferromagnetic layer, and down-spin electrons will scatter strongly at both interfaces. When the excess spins in the two ferromagnetic layers have opposite directions (illustrated on right-hand side of Fig. 3), both the up- and down-spin electrons will scatter at one of the interfaces. The electrical resistivity for the aligned case should be lower because up-spin electrons in this case experience very little resistance and act like a short circuit. Think of the limiting case in which up spin electrons for the aligned case have no resistance, and down-spin electrons have resistance 2R, with R being the contribution from each interface. Because the up spin provides a short-circuit channel, the resistance of the whole system is zero. For the case in which excess spins on two cobalt layers are aligned in opposite directions, both up and down spins have resistance R because both scatter off one interface. In this case the total resistance is R/2, which is certainly greater than the value zero for the aligned case.

From this oversimplified argument, it is evident that resistance will be lower when excess spins in the magnetic layers are in the same direction. Thus, if excess spins are pointing in different directions in different magnetic layers when no magnetic field is applied, applying an external magnetic field can turn them so that they point in the same direction, resulting in a drop in resistivity.

Calculating Electrical Conductivity

The understanding that came from our initial calculations helped us formulate qualitative ideas to explain possible causes of the GMR effect. However, we needed a more detailed understanding of GMR, so we decided to calculate the electrical conductivity of these materials. Calculation of electrical conductivity, which is the inverse of electrical resistivity, is difficult. In these materials it was particularly tough because the layers must be extremely thin to show a large GMR effect. No one knew how to calculate the electrical conductivity of materials that are inhomogeneous on such a fine scale. In the past, physicists had only attempted to calculate the resistivity of such systems using the simplest of “toy” models. We knew it was important to include information about how electrons interact with and scatter from atoms in the layers. We were able to perform these calculations and develop a realistic description of the resistivity of these layered materials partly because of the availability of parallel supercomputers at ORNL.

The other necessary ingredient was an interlaboratory team that had the experience and capabilities to attempt such a formidable task (and perhaps were sufficiently naive to believe it could be done). In addition to the authors, the other team members are Thomas Schulthess of Lawrence Livermore National Laboratory, Richard Fye of Sandia National Laboratories, and Virgil Speriosu and Bruce Gurney, both of the IBM Almaden Research Center. IBM is interested in the GMR effect as a means of increasing the density of data stored on magnetic disks (see Applications of Giant Magnetoresistance). Speriosu and Gurney provided the experimental measurements that we compared with our calculations. The work was funded by DOE’s Office of Defense Programs with additional support from DOE’s Office of Basic Energy Sciences. Many of the computations were performed on the parallel supercomputers at ORNL’s Center for Computational Sciences.

drawing
Fig. 4. Nonlocal up-spin conductivity for the case in which the excess spins are aligned on the cobalt layers.
drawing
Fig. 5. Nonlocal layer-dependent conductivity for the down-spin electrons for the case in which excess electrons on the cobalt layers are aligned.
drawing
Fig. 6. Conductivity for the case in which excess spins on the cobalt layers are aligned in opposite directions.
drawing
Fig. 7. Giant magnetoconductance (GMC) is the inverse of giant magnetoresistance.

Figure 4 shows some of the results of our calculations of the conductivity of layered cobalt-copper systems. To understand the conductivity of a layered conductor, it is important to realize that the conductivity is nonlocal. We usually think of the conductivity, s, in terms of the formula J = sE, which states that the current at any point, J, is proportional to the applied electric field at that point, E. But in a material made of very thin layers, an electron may be accelerated by a field at one point and then travel a distance, l, before it scatters off an impurity or other irregularity. Because the properties and the fields may change over distances smaller than l, we must consider nonlocal conductivity, which relates the current in a particular atomic layer to the field applied in all of the other atomic layers, J(I) = SJ s(I,J) E(J). Figures 4 through 7 show s(I,J), the current induced in layer I by a field applied in layer J. Figure 4 shows this calculated nonlocal conductivity for the up-spin electrons.

The results are color-coded. Blue represents electrons accelerated by an electric field applied in a cobalt layer which contribute to the conductivity in a cobalt layer. Red represents electrons accelerated in a copper layer that contribute to the current in a copper layer. Purple represents electrons accelerated by a field in a copper layer which contribute to the current in a cobalt layer, or vice versa. Note that the blue region in the near corner represents conductivity from electrons accelerated in one of the cobalt layers which contributes to the conductivity in the other.

Figure 5 shows a similar plot for the down-spin electrons. The conductivity for the down-spin electrons is localized within the layers. If an electron is accelerated by a field in one layer, it contributes only to the current in that layer.

Figure 6 shows the nonlocal conductivity for the case in which the excess spins on the cobalt layers are aligned in opposite directions. In this case the conductivities on the left-hand side of the figure are similar to those of the up-spin electrons of Fig. 4, and those on the right-hand side are similar to the conductivities of the down-spin electrons of Fig. 5.

Figure 7 shows giant magneto-conductance. (Conductance is the inverse of resistance.) Thus, giant magnetoconductance is the difference between total conductivity for the case in which excess spins are parallel (i.e., the sum of the conductivities in Figs. 4 and 5), and the total conductivity for the case in which excess spins are anti-parallel (i.e., Fig. 6 and its mirror image).

Giant magnetoconductance comes from two effects. The blue region in the near corner of Fig. 7 represents electrons accelerated in one cobalt layer which travel through the copper and contribute to the conductivity in the other cobalt layer. This type of contribution to the GMR is not too surprising. It corresponds to the simple picture of GMR illustrated in Fig. 3.

Waveguide Effect for Electrons

The largest contributions to GMR come from electrons accelerated by a field in the copper layer that help produce a current in that layer. This type of contribution was very surprising to us until we traced its origin to an effect well known to physicists and engineers—the waveguide effect.

This phenomenon is becoming more familiar, thanks to advertisements by phone companies that encourage you to use their services because they promise that your conversations will travel along optical fibers. What they are referring to are optical waveguides. Light waves can travel for miles within a thin glass fiber, even around bends and corners if they are not too sharp. An optical waveguide is a glass fiber that has a glass core surrounded by more glass with a lower index of refraction than the core. Light waves within the core traveling approximately parallel to the length of the fiber will be reflected back into the core. Thus, they can travel for several miles if no impurities are present to scatter or absorb them.

Our calculations predict that a similar phenomenon affects up-spin electrons in a copper layer with cobalt on either side. Like light, electrons propagate as waves. Because the effective index of refraction for electrons in copper is higher than for up-spin electrons in cobalt, some up-spin electrons can be confined to the copper because of their wavelength and, because of copper’s low resistivity, they can travel further without scattering. The down-spin electrons for the aligned case will enter the high-resistivity cobalt at both interfaces, and both types of electrons will enter the cobalt at one or the other of the interfaces for the nonaligned case.

The net result is that the waveguide effect can make a big contribution to the GMR for some systems. The size of the channeling effect depends on how nearly perfect the interfaces are between the cobalt and copper layers. We are currently comparing our theoretical calculations with measurements of GMR for cobalt-copper films performed at the IBM Almaden Research Center in California under a cooperative research and development agreement. It may turn out that interfaces in films currently being deposited are not smooth enough to give a strong channeling effect. If so, there will be a strong impetus to increase the effect by finding ways to deposit films with smoother interfaces. An increase in the GMR of the type we predict would have a profound impact on many technologies.


 
Applications of Giant Magnetoresistance

Giant magnetoresistance already has magnetic appeal: It allows more data to be packed on computer disks. If improvements are made in the interfaces between magnetic layers in thin-film structures, the number of new applications could prove irresistible.

For example, it would be possible to make computer operating memories [random access memory (RAM)] that are immune to power disruptions and ionizing radiation. GMR motion sensors could be developed to increase the efficiency and safety of home appliances, automobiles, and factories. Magnetoelectronic devices may someday complement or even replace semiconductor electronic devices.

GMR recently moved out of the laboratory and into our computers with the development of “read sensors” for magnetic disk drives. Because the capacity of disk drives continues to grow rapidly as they shrink in size, GMR read sensors become increasingly important.

Disk drive manufacturers write more and more data into smaller amounts of space. The data are written as tiny regions of magnetization on a disk covered with a thin film of magnetic material. The information (bits of 1 or 0) is stored as the direction of the magnetization of these regions. The information is read by sensing the magnetic fields just above these magnetized regions on the disk. As the density gets higher, these regions get smaller, so the fields that must be sensed to read the data become weaker. Read sensors that employ the GMR effect provide the best technology currently available for detecting the fields from these tiny regions of magnetization. These tiny sensors can be made in such a way that a very small magnetic field causes a detectable change in their resistivity; such changes in the resistivity produce electrical signals corresponding to the data on the disk which are sent to the computer. It is expected that the GMR effect will allow disk drive manufacturers to continue increasing density at least until disk capacity reaches 10 gigabits per square inch. At this density, 120 billion bits could be stored on a typical 3.5-inch disk drive, or the equivalent of about a thousand 30-volume encyclopedias.

GMR also may spur the replacement of RAM in computers with magnetic RAM (MRAM). RAM holds the data that your computer must get to quickly in order to operate. Today’s technology uses semiconductor RAM because it is fast, dense, and relatively inexpensive. In semiconductor RAM, data are stored as small regions that have an excess or deficit of electrons. This use of electrical charges to represent data has two serious drawbacks. First, because these charges leak away, the data must be refreshed several times a second by an electrical circuit. Thus, if the power goes off before the data can be written back to the hard disk for permanent storage, they will be lost. Second, because ionizing radiation temporarily destroys a semiconductor chip’s semiconducting properties, it can destroy data. Using GMR, it may be possible to make thin-film MRAM that would be just as fast, dense, and inexpensive. It would have the additional advantages of being nonvolatile and radiation-resistant. Data would not be lost if the power failed unexpectedly, and the device would continue to function in the presence of ionizing radiation, making it useful for space and defense applications.

One exciting aspect of GMR devices is their extremely small size. Currently, computer and electronics manufacturers are struggling to shrink their devices and keep them working at feature sizes of about 0.5 mm. Operating GMR-based devices are already 50 times smaller than that, and they tend to work better at smaller sizes. It has already been shown that GMR can be used to make a transistor.

The application of GMR in motion sensors is also likely to be important in our homes, automobiles, and factories. It provides a convenient way of sensing the relative motion and position of objects without physical contact. Just attach a magnet to one object and a GMR sensor to another. Alternatively, if one of the objects contains a magnetic material such as iron or steel, the object in motion will alter any magnetic field that is present. These small changes in the magnetic field could be detected by a GMR sensor.

Applications of this effect could become widespread in the industrial, commercial, and military worlds. Here’s a possible list: sensitive detectors for wheel-shaft speed such as those employed in machine-speed controllers, automotive antilock brakes, and auto-traction systems; motion and position sensors for electrical safety devices; current transformers or sensors for measuring direct and alternating current, power, and phase; metal detectors and other security devices; magnetic switches in appliance controls, intrusion alarms, and proximity detectors; motor-flux monitors; level controllers; magnetic-stripe, ink, and tag readers; magnetic accelerometers and vibration probes; automotive engine control systems; highway traffic monitors; industrial counters; equipment interlocks; and dozens of other applications requiring small, low-power, fast sensors of magnetic fields and flux changes. Furthermore, suitable film-deposition processes could also permit fabrication of GMR devices on electronic-circuit chips to produce highly integrated GMR sensors at low cost and high volumes for mass industrial markets. With its promise for tomorrow’s technologies, GMR is bound to attract lots of attention.—Bill Butler and Steve Smith


BIOGRAPHICAL SKETCHES

WILLIAM H. BUTLER has been leader of the Theory Group in ORNL’s Metals and Ceramics Division since 1985 and was manager of computer planning at ORNL in 1984–85. He holds a Ph.D. degree in solid-state physics from the University of California at San Diego. Before coming to ORNL in 1972, he was assistant professor of physics at Auburn University. He received the DOE–Basic Energy Sciences Division of Materials Sciences Award for Outstanding Sustained Research in Metallurgy and Ceramics in 1983 and the DOE–Basic Energy Sciences Division of Materials Sciences Award for Outstanding Scientific Achievement in Metallurgy and Ceramics for 1995 for outstanding computational research on GMR.

XIAOGUANG ZHANG is a member of the research staff in ORNL’s Computational Physics and Engineering Division. He has a Ph.D. degree in physics from Northwestern University. Before joining the ORNL staff in 1995, he worked as a postdoctoral fellow at DOE’s Lawrence Berkeley National Laboratory, as a postdoctoral scholar at the University of Kentucky, and as a postdoctoral research associate in ORNL’s Metals and Ceramics Division. He received the DOE–Basic Energy Sciences Division of Materials Sciences Award for Outstanding Scientific Achievement in Metallurgy and Ceramics for 1995 for outstanding computational research on GMR.

DONALD M. NICHOLSON is a senior research scientist in ORNL’s Computational Physics and Engineering Division. He has a Ph.D. degree in solid-state physics from Brandeis University. He joined the ORNL staff in 1986 as a member of the Metals and Ceramics Division. He received the DOE–Basic Energy Sciences Division of Materials Sciences Award for Outstanding Scientific Achievement in Metallurgy and Ceramics for 1995 for outstanding computational research on GMR.

 

Where to?

Next article | Contents | Search | Mail | Review Home Page | ORNL Home Page