Cullity B.D. Introduction to Magnetic Materials. Second Edition (2008)

distinct layers, not counting the magnetic shield layers that provide magnetic isolation from the write head and from the neighboring bits.

In the final structure, the reference layer is magnetized perpendicular to the recorded surface, and the free layer is normally oriented perpendicular to the reference layer. The magnetic flux from the flux transition in the recorded track rotates the direction of the free layer by about 108, resulting in a resistance change of a few percent. It is not desirable to have the magnetization in the free layer jump by 90 or 1808 in the presence of a flux reversal. This would give a bigger signal, but a noisier one.

The write head is also prepared using thin film technology, and is deposited on top of the read head in a continuous operation. More than 10,000 head assemblies are made simultaneously on a single substrate.

15.4.2Colossal Magnetoresistance

Even larger values of magnetoresistance, amounting to a transition from an insulating to a conducting state, have been observed in perovskites containing Mn. Since the term giant magnetoresistance was already taken, this effect is called colossal magnetoresistance. However, very large fields are required to drive the transition, and no applications have yet been announced.

15.4.3Digital Recording Media

Here it is necessary to distinguish between recording on flexible media and on rigid media. That is, between recording on magnetic tape, which can be removed from the recording device and read by a different machine, and recording on a hard disk which is permanently installed in a sealed unit so that the recording and reading are always performed by the same mechanical system. Recording on flexible or “floppy” disks is similar to recording on tape, with the recorded tracks forming a series of concentric circles on the disk. Floppy disks have been substantially replaced by small semiconductor storage units (flash memory) or by magneto-optic disk recording.

For high-density magnetic tape recording, g-Fe2O3 particles have been replaced by metal particles (Fe or CoFe) and inductive read heads have been replaced by magnetoresistive heads. Smoother tape surfaces let the writing and reading heads pass closer to the magnetic media; in tape systems, the heads are in physical contact with the tape surface. Improved positioning systems allow narrower recorded tracks. For a variety of reasons—the flexibility of the magnetic media, the requirement for the tape to be removable and playable in multiple units, and so to be exposed to varying conditions, dust, etc., —the recording density of magnetic tape is typically a factor of 10–100 lower than that of hard disk drives. However, tape remains useful for long-term storage of large amounts of data, e.g. business and medical records, census and weather information, or scientific data from spacecraft. A severe problem with long-term data storage on magnetic tape is that the tape and the data may survive, but the machines to read the tape may disappear as new and better machines are developed. The only obvious solution is to rerecord the data onto newer media before the old machines expire, which can be a very expensive operation.

The density of recorded information on hard disk drives has increased steadily and rapidly by about a factor of 108 since the first hard disk drive was introduced in 1956, and the cost per unit of storage has dropped similarly. A major contribution to this remarkable increase has been the improvement in the recording medium. The disk in a hard disk

512 MAGNETIC MATERIALS FOR RECORDING AND COMPUTERS

Fig. 15.6 Structure of the magnetic layer of a hard disk drive. The heavy line is the boundary between two recorded data bits.

drive consists of a polished aluminum alloy or glass substrate. The earliest disk drives used g-Fe2O3 particles borrowed directly from the technology of magnetic tape. A variety of other materials has been used, including both crystalline and amorphous metallic layers, mostly based on cobalt because of its relatively high magnetization, high anisotropy, and good corrosion resistance.

The current material of choice is cobalt, with additions of Pt to increase the anisotropy and of Cr to lower the magnetization. Control of the seed layer and of the growth conditions allows the magnetic material to form in columnar grains about 10 mm in diameter, separated by nonmagnetic low-density oxide-filled spaces only a few nm thick. Each magnetic particle acts as a single domain, so that the boundary between recorded bits follows an irregular line between the particles, as in Fig. 15.6. The irregularity of this boundary increases the width of the flux reversal, and sets a limit to the minimum length of a recorded bit.

A careful balance is required to keep the anisotropy high enough so that the recorded bits are stable against thermal demagnetization but low enough so that bits can be written with the available field from the write head, while the level of magnetization is high enough so that the recorded bits can be reliably read with the available head technology.

The heads in a hard disk drive are not in contact with the recorded surface, but “fly” just above the surface on a flowing air layer trapped between the head and the disk surface. The flying height must be comparable to the bit length, which means the disk surface must be highly polished and flat. The positioning of the heads with respect to the recorded track also requires an elaborate and fast-acting control system.

15.5PERPENDICULAR RECORDING

For many years, there have been advocates of perpendicular recording; that is, of redesigning the recording system so that the magnetized bits are perpendicular rather than parallel to the disk surface. The arguments for perpendicular recording were plausible, and were taken seriously, but advances in longitudinal recording were so successful that a basic change in the system was unnecessary. Early in the twenty-first century, the situation changed, and perpendicular recording was finally brought into production. The geometry of the recording head is indicated in Fig. 15.7; the read head is a GMR design essentially the same as that used for longitudinal recording.

The major advantages of perpendicular recording are that higher magnetizing fields can be applied to the recording medium, allowing the use of material with higher anisotropy and greater stability, and that the bit length can be reduced, packing more bits per unit length of

Fig. 15.7 Perpendicular recording geometry.

track. A requirement, in addition to the redesign of the write head, is the addition of a soft magnetic underlayer to the disk, between the actual recording layer and the nonmagnetic support layer. This layer provides a path for the magnetic flux to return to the head, as indicated in the figure. The return flux is spread over a much larger area than the area of the writing gap, so the field strength is much lower and is insufficient to disturb the recorded bits.

15.6POSSIBLE FUTURE DEVELOPMENTS

Two methods have been suggested to permit even higher recording densities: thermally assisted writing and patterned media. If the temperature of the recording medium is increased, its anisotropy will be lowered. Therefore a medium with anisotropy too high to be recorded at room temperature could be recordable at say 2008C, and the recorded data would be highly stable when the temperature returns to normal. Local heating could be produced with a laser beam. Patterned media would replace the macroscopically homogeneous recording surface, magnetized locally by the write head, with magnetic tracks or even individual magnetic bits isolated from the rest of the media surface by physical gaps or by nonmagnetic materials. In principle this would allow the tracks to be narrower or the bits smaller, or both. The two approaches are not mutually exclusive.

15.7MAGNETO-OPTIC RECORDING

Most recording on compact disks and digital versatile disks (CDs and DVDs) is purely optical. Single-write disks use a dye layer whose optical properties are changed by highintensity laser illumination and read by low-intensity laser light. Disks capable of multiple writing and rewriting replace the dye layer with a phase-change layer. In both cases no magnetic phenomena are involved. There are, however, also magneto-optic disk recording systems. Typically the magnetic layer is an amorphous alloy of terbium and iron–cobalt, which has a high anisotropy and can be magnetized perpendicular to the disk surface. Digital bits are recorded by simultaneously applying a perpendicular magnetic field to one side of the disk and a thermal pulse from a laser to the other side of the disk, as

514 MAGNETIC MATERIALS FOR RECORDING AND COMPUTERS

Fig. 15.8 Magneto-optic recording.

indicated in Fig. 15.8. The coercive field of the amorphous layer decreases with increasing temperature, so a bit is easier to write at elevated temperature, and will be stable against demagnetizing fields back at room temperature. This is an example of thermo-magnetic writing. The recorded information could be read by an inductive or a magnetoresistive read head, but in practice the Kerr effect is used. The plane of polarization of a reflected laser light beam is rotated in opposite directions, depending on the direction of magnetization of the reflecting surface. This change in polarization direction can be detected and used to read the recorded information.

15.8MAGNETIC MEMORY

15.8.1Brief History

The first large-scale computers after the vacuum-tube era used magnetic memory. Each bit of information was stored as clockwise or counterclockwise magnetization of a small ferrite ring, threaded with wires to provide the fields for writing and reading. The ferrite rings were called cores, and the memory unit was called core memory, a term sometimes still heard. This system did not lend itself to scaling down in size or in cost, and a variety of replacement technologies based on magnetic thin films were suggested and investigated. However, semiconductor technology intervened, and semiconductor random-access memory (RAM) replaced magnetic memory.

In the 1970s, a great deal of research and development was devoted to magnetic bubble memory technology, which was seen as a replacement for semiconductor memory. A large area of single crystal magnetic garnet grown epitaxially on a nonmagnetic garnet substrate can be magnetized perpendicular to the substrate surface. When a reverse field is applied, cylindrical domains of reverse magnetization appear. When viewed from above, these domains look and act somewhat like bubbles, which explains their name (see Fig. 15.9). In a suitably defect-free crystal, the bubbles can move easily in the plane of the device, and the addition of ferromagnetic surface “tracks,” usually thin-film NiFe layers, can confine the bubbles to predefined paths. Finally, the application of an in-plane rotating magnetic field can move the bubbles in a stepwise fashion, one step per revolution of

Fig. 15.9 Magnetic bubbles.

the field, along a closed path. With the addition of facilities to create and destroy individual bubbles, and an optical or magnetic bubble sensor, a bubble domain chip can store and process information. It has the great virtue of being nonvolatile, meaning that the stored information remains safely in place when the power is turned off, eliminating the annoying lengthy start-up sequence required with volatile semiconductor memory. However, bubble memory was never able to compete with semiconductor memory in storage density or cost, and it quietly disappeared after about 10 years of intense effort.

15.8.2Magnetic Random Access Memory

The discovery and exploitation of the giant magnetoresistance (GMR) effect, beginning in the late 1980s, rekindled interest in the possibility of a magnetic random access memory (MRAM). The first commercial MRAM units, available in 2006, made use of magnetic tunnel junctions: two ferromagnetic layers separated by a very thin dielectric tunneling layer, usually magnesium oxide. The physics of the device differs from that of the GMR read head, in that the current between the two magnetic layers tunnels through the barrier layer, but the result is basically the same: The resistance to current flow depends on the relative orientation of the directions of magnetization in the two outer layers. When the magnetizations are parallel, the resistance is minimum. Also in the MRAM device, the magnetization of the free layer is either parallel or antiparallel to that of the fixed layer except during switching. In the GMR head, the free layer magnetization rotates away from its stable orientation, but does not reverse direction. Various ingenious refinements are added to make the device operate reliably, and almost certainly additional changes will be made.

15.8.3Future Possibilities

Information storage using the presence or absence of domain walls in a magnetic channel has been proposed as a way to increase the density of stored information. Furthermore, it may be possible to move domain walls by a current flow, which could eliminate the need for magnetic fields to read and write information. A variety of other (nonmagnetic) high-density information storage systems have been suggested.

518 MAGNETIC PROPERTIES OF SUPERCONDUCTORS

is not caused by the alignment or creation of magnetic moments on the constituent atoms. It results from a real electric current flowing (without resistance) in a thin shell around the outer surface of the superconductor. But, as noted in Chapter 1, a magnetic moment and a current shell are indistinguishable by any external measurement.

An interesting reciprocal relationship thus exists between the representation of magnetic and superconducting behavior. In a magnetic material, the magnetization arises from the alignment of local magnetic moments, or of electron spins, throughout the volume of the material. However, the magnetization may be considered to result from a current flowing in a shell around the outside of the sample; this representation is used in many textbook treatments of magnetization. In a superconductor, the magnetic flux in the sample is cancelled (B ¼ 0) by an actual supercurrent flowing around the surface of the sample. However, the external effects of this current may be represented by a (negative) bulk magnetization of the entire sample; it is therefore customary to attribute a magnetization to a superconductor in a magnetic field. Note that the condition B ¼ 0 in the superconducting state requires that permeability (m ¼ B/H) also be zero, and the condition H þ 4pM ¼ 0 or m0(H þ M ) ¼ 0 requires that susceptibility (x ¼ M/H) be 21/4p (cgs) or 21 (SI).

The superconducting magnetization can be measured by any of the measurement techniques applicable to magnetic materials. Three special considerations apply:

1.The superconducting magnetization is generally fairly small, less than 1 kG or 0.1 T or 800 kA/m. This is not a major problem except when dealing with very small samples, thin films, etc.

2.The measurements must be carried out below room temperature, since no known superconductor remains superconducting much above 130K. Also it is common to make measurements as a function of temperature over a wide temperature range.

3.Because the magnetization is negative, the demagnetizing field adds to the applied field rather than subtracting from it. This takes some mental adjustment, and leads to some magnetic behavior that would not otherwise occur.

Although some important basic measurements on superconductors have been made using ring samples (thus avoiding demagnetizing effects), ring samples are not regularly used. There are several reasons for this: engineering applications of superconductors rarely use a ring geometry; many high-field, high-temperature superconductors are brittle and difficult to make in the form of rings; applying windings to a ring sample is tedious; and it is not easily possible to apply large fields to a ring sample.

Magnetic measurements have significant advantages over direct resistance measurements on superconductors for materials characterization. Magnetic measurements avoid the attaching of voltage and current leads to the sample, which are frequent sources of trouble. Some superconducting materials are difficult to produce in reasonably long lengths with small, uniform cross-sections, as required for resistance measurements, whereas magnetic measurements can be made on small pellets of various shapes.

It the literature of superconductivity negative magnetization is sometimes plotted as a positive quantity, so the initial magnetization curve has positive slope. However, in a book on magnetic materials, it seems preferable to plot negative magnetization as a negative quantity, to emphasize the difference in the behavior of magnetic and superconducting materials.

520 MAGNETIC PROPERTIES OF SUPERCONDUCTORS

Fig. 16.2 Magnetic behavior of an ideal type I superconductor with nonzero demagnetizing factor N.

Fig. 16.3 Hysteretic behavior of a type I superconductor (N = 0).

When the field is decreased, in an ideal sample the curves are exactly retraced. In real samples, there may be some hysteresis, resulting a small amount of trapped flux at Ha ¼ 0, which results in a small positive magnetization, as indicated in Fig. 16.3.

The topography of the intermediate state is variable, and is different at the sample surface and in the interior. There is a positive surface energy associated with the boundary between the superconducting and the normal state, and the magnitude of this energy naturally influences the structure of the intermediate state. There are some parallels between the development of the domain structure of a ferromagnetic sample and the superconducting/normal structure of the intermediate state.

16.3TYPE II SUPERCONDUCTORS

The essential difference between a type I and a type II superconductor is that, in a type II, the surface energy of the superconducting/normal interface is negative. When a field is applied to a type II superconductor, a negative magnetization is produced by a surface current, just as in a type I, but only up to a lower critical field Hc1. At this field,

Fig. 16.4 Magnetic behavior of an ideal type II superconductor (N ¼ 0).

f0 is 2.07 1027 G cm2 (Maxwell) or 2.07 10215 Wb. A type II superconductor containing flux lines is said to be in the mixed state, which is to be distinguished from the intermediate state of a type I superconductor. If the flux lines can move freely, they will enter the sample at H ¼ Hc1 and pack together until they are spaced closely enough to begin to repel one another. The (negative) magnetization thus drops sharply at first, then more slowly as the flux lines are packed more and more tightly together. Finally, at the upper critical field Hc2, the sample becomes completely normal. The magnetization curve has the general shape shown in Fig. 16.4. The curve is retraced on lowering the field back to zero.

However, the flux lines generally do not move freely in practical type II superconductors such as NbTi and Nb3Sn. For engineering applications, it is desirable that the flux lines are pinned by interactions with defects in the structure. If the flux lines are completely pinned, the behavior of the sample is given by the Bean critical state model [C. P. Bean, Phys. Rev. Lett., 8 (1962) p. 250]. In this case, as an external field is applied to a slab of superconductor, the surface shielding current density rises to a maximum value JC characteristic of the material. As the field is increased, this current density extends more and more deeply into the slab, until it occupies the entire volume. The (negative) magnetization produced by this current distribution increases with applied field and reaches a maximum when full current flows throughout the sample (condition H2 in Fig. 16.5a). When the field is reversed, the current density at the surface reverses, giving a field distribution as shown in Fig. 16.5b. The behavior is, of course, highly hysteretic, with large values of “trapped flux” in the sample when the field is reduced to zero, as in Fig. 16.5c. Equations giving the form of the magnetization vs. field behavior for samples of various geometries according to the Bean model are given by R. B. Goldfarb, M. Lelental, and C. A. Thompson, in

Magnetic Susceptibility of Superconductors and Other Spin Systems [R. A. Hein, T. L. Francavilla, and D. H. Liebenberg, eds., Plenum Press (1992)].

For most applications of type II superconductors, specifically to create high magnetic fields, complete flux pinning is desirable, and the Bean model is a reasonable approximation. However, perfect pinning is an ideal case, and often there is some degree of flux creep due to motion of the flux lines. Flux creep appears as a resistance, and therefore as the production of heat, in a superconducting coil. The phenomenon was first clearly identified in a series of experiments by Kim et al., using a superconducting hollow cylinder as a sample, and comparing the magnetic field measured inside and outside the cylinder (see Fig. 16.6).