Discoverers of giant magnetoresistance win this year’s physics nobel

Secretary General Gunnar Öquist made the announcement: “The Royal Swedish Academy of Sciences has decided to award the Nobel Prize in Physics for the year 2007 jointly to Professor Albert Fert of Université Paris–Sud in France and Professor Peter Grünberg of Forschungszentrum in Jülich, Germany.”

Fert and Grünberg discovered giant magnetoresistance independently. In 1988, they and their collaborators made thin, multilayer structures of ferromagnetic iron and chromium. When the researchers applied a magnetic field, the structures’ electrical resistance fell. In Fert’s 30-layer structure, the drop at 4.2 K was 50%. ¹ He dubbed the new magnetoresistive effect “giant.”

Grünberg saw a more modest drop of 1.5% in his three-layer structure, but at room temperature. ² He grasped the effect’s potential for sensing magnetic fields and promptly filed a patent. Within a decade, thanks mostly to R&D at IBM’s Almaden Research Center, tiny GMR-based sensors were reading data in magnetic disk drives. Disk capacity rose 10-fold as a result.

Before GMR’s discovery, interest in the transport of electrons through magnetic materials was so low that the field failed to meet the threshold—12 papers—for its own session at the annual Magnetism and Magnetic Materials Conference. After GMR, the conference had to run multiple sessions on the topic.

That surge of interest was reflected in the citation rate of Fert’s and Grünberg’s discovery papers. By 1994 the pair had notched up a combined 200 citations per year. That rate has stayed more or less constant ever since.

Magnetoresistance

The origin of GMR lies in the band structure of ferromagnets. In the mid-1930s, Nevill Mott asked himself why the transition metals nickel, palladium, and platinum are worse electrical conductors than their respective neighbors in the periodic table copper, silver, and gold.

In both nickel and copper—to pick one pair—the conduction band is populated by electrons from the atoms’ hybridized 4s and 3p orbitals. Electrons in the 3d band, being more tightly bound, contribute far less to conduction.

Copper’s 3d orbitals are completely full. Nickel’s are not, and the corresponding 3d band has empty states. Those empty states, Mott realized, serve as traps: If a conduction electron scatters off an impurity, defect, or phonon, it can drop into the 3d band. The upshot: increased resistance.

Mott also realized that spin would influence conduction. In nickel and other ferromagnets, neighboring electrons can lower their energy by aligning their spins. The effect is to split and shift the band structure. As the figure on page 13 shows, majority spins end up with fewer unoccupied states in the 3d band than do minority spins. With fewer states to scatter into, majority-spin electrons face less resistance. In effect, two currents flow through the ferromagnet: a larger one of majority spins and a smaller one of minority spins.

In Mott’s picture, conduction through a ferromagnet is like driving current through two parallel sets of resistors: a top set, say, for spin up and a bottom set for spin down. Majority spins would face small resistors; minority spins would face large resistors. If the top resistors are all small—that is, if spin-up electrons are always in the majority—the total current would be optimal. But in a real ferromagnet, the domains point in different directions. The top and bottom resistors would be a mix of large and small. The current would be suboptimal.

In principle, applying a magnetic field to align all the domains could reduce the resistance. In 1964 William Reed and Eric Fawcett of Bell Labs did just that in whiskers of crystalline iron. They switched the alignment of the domains from their lowest-energy configuration—alternating up and down—to near-complete alignment. Resistance plummeted by a factor of 1/7.

Unfortunately for applications, Reed and Fawcett’s effect appeared at the low temperature of 4.2 K and under the high field of 8 MA/m. Moreover, crystalline whiskers make awkward building blocks.

Roughly speaking, Fert’s and Grünberg’s sandwiches work in the same way. Because GMR is manifested through scattering, it doesn’t matter whether the current flows parallel to the layers or perpendicular to them. Regardless of how the layers are configured, the greater the number of layers, the stronger the magnetoresistive effect.

Using nanometer-thick layers rather than whiskers lowers the magnetic field required to change the alignment. It also ensures that most of the scattering that occurs depends on alignment.

To make sure that the highest resistance state, antialignment, persists at room temperature and in zero applied field, a phenomenon called exchange coupling is exploited. Indeed, it was Grünberg’s observation of exchange coupling ³ in 1986 that led him and Fert to look for a magnetoresistive effect in multilayers.

Semiconductors and alloys

Exchange coupling occurs when two thin ferromagnetic layers are separated by a thin nonmagnetic metal layer. The coupling causes the ferromagnets’ magnetizations to antialign. Louis Néel found the effect in 1964 in layers 100 nm thick.

Grünberg’s rediscovery of exchange coupling arose from his work in the 1970s and 1980s on spin waves in europium oxide and other ferromagnetic semiconductors. At first he studied spin waves in bulk samples. Then he became interested in how spin waves in two closely separated layers interact through dipolar coupling. Europium oxide’s ferromagnetism appears below 70 K. Hoping to find stronger interlayer coupling at room temperature, Grünberg began working with ferromagnetic metals.

Fert, by contrast, has always worked with metals, both as a theorist and as an experimenter. In a 1968 paper, he and his thesis adviser Ian Campbell examined what happens to the resistance of nickel when two different dilute impurities are added: “A,” which scatters spin-down electrons more than spin-up, and “B,” which does the opposite. ⁴

By adjusting the proportions of A and B and measuring the change in resistance, Fert and Campbell verified Mott’s two-current model. The adjustment was accomplished by making a fresh batch of alloy—hardly a practical step in a magnetoresistive device—but it gave Fert the idea of using spin to control overall resistance. Exchange coupling gave him the means.

But to exploit exchange coupling, Fert needed to make thin metallic layers. Gary Prinz of the Naval Research Laboratory in Washington, DC, had already provided a recipe: Fit a high-temperature oven to a molecular-beam-epitaxy machine and deposit the metallic layers on a gallium arsenide substrate.

Grünberg followed Prinz’s recipe at his lab in Jülich. Fert, however, lacked his own MBE machine. Fortunately, one of his former students, Alain Friederich, had become a research manager at Thomson, a French electronics company. Friederich and Fert met at a conference in San Diego. After a poolside conversation, they resolved to use one of Thomson’s MBE machines.

Fert and Grünberg made their devices and wrote up their results. Both papers arrived at the Physical Review offices in the summer of 1988. The discoverers found out about each other’s work when they both spoke at that year’s International Colloquium on Magnetic Films and Surfaces in Le Creusot, France.

From GMR to read heads

Although GMR is a robust, reproducible effect, it remains incompletely understood. Sorting out the physics of room-temperature transport in thin layers of less-than-perfect materials is a formidable problem. Even so, GMR soon made its mark in the magnetic storage industry.

Magnetic disk drives work by writing and reading tiny patches of magnetization on a spinning disk. Smaller patches mean more information can be crammed on the disk. They also mean weaker, harder-to-detect fields.

The first generation of disk heads used inductive coils to read and write data. Then, in 1991, IBM introduced read heads based on anisotropic magnetoresistance. Discovered in 1856 by William Thomson (who later became Lord Kelvin), AMR is a few-percent change in resistance when the direction of an applied magnetic field is changed from parallel to perpendicular.

Despite its weakness, AMR has the advantage of being a transport effect. Sensors can be made from thin layers that fly just above the disk. A change in the current flowing through the sensor corresponds to a change (0 → 1 or 1 → 0) in the magnetic bits beneath. AMR-based heads helped to boost storage density 10-fold.

Using GMR in place of AMR was a clear next step. However, the fields that were needed to switch the magnetization in Fert’s and Grünberg’s structures were too strong to bring the hoped-for increase in sensitivity. In a series of experiments, a team of IBM researchers including Virgil Speriosu, Stuart Parkin, Bruce Gurney, and Bernard Dieny figured out how to make GMR-based read heads. And in the process, they also elucidated some of the underlying physics.

Unlike Fert and Grünberg, they used sputtering machines to make their multilayers. Even though sputtered layers have more defects than MBE-grown layers, they found GMR in their structures. And because sputtering is faster and cheaper than MBE, they could readily and quickly explore GMR in a wide range of materials and conditions.

Among the key IBM findings is that the all-important spin-dependent scattering occurs mostly at the interfaces between layers, rather than in the layers themselves.

GMR requires ferromagnetic layers to change their alignment with respect to each other. In a GMR-based read head, that is achieved by depositing a layer with low magnetic anisotropy to respond to an external field and to pin the other layer using exchange anisotropy. An important refinement making the pinning strong was the use of two ferromagnetic layers coupled by a ruthenium layer in between. Because the sandwich has low net magnetization, the coupling is enhanced.

Read heads based on this idea (US patent no. 5,159,513) went into production in 1997, just nine years after GMR was discovered. IBM made billions of dollars in sales and licensing.

GMR-based read heads replaced the older AMR-based heads because the two magnetoresistive effects can be realized with similar technology. By 1999 the GMR read head was the industry standard. More than 5 billion of the heads have been produced. Although still in use today, by 2006 GMR read heads were being supplanted by heads based on another MR phenomenon: tunnel magnetoresistance.

TMR was discovered in the early 1970s. Outwardly, the physics is similar. Both GMR and TMR rely on ferromagnetic layers whose band structure features a spin-dependent shift. But in TMR, it’s tunneling through an insulator, not scattering in a metal, that provides the magnentoresistance.

TMR had languished because it seemed possible only at impractically low temperature. Then, in 1995, MIT’s Jagadeesh Moodera and his collaborators and, independently, Teranobu Miyazaki and Nobuki Tezuka of Tohoku University in Japan saw a GMR-sized effect at room temperature using an aluminum oxide spacer. Nine years later, two groups—one led by Parkin, the other led by Shinji Yuasa of the Nanoelectronics Research Institute in Tsukuba, Japan—replaced aluminum oxide with magnesium oxide and boosted TMR by an order of magnitude. In 2005, Seagate Technology introduced the first drives equipped with TMR read heads.

Although GMR’s read-head reign was short-lived, its influence on physics endures, especially in the broader field of spin-based transport or spintronics. Says Prinz, “Albert and Peter really broke open the field.”