G E E K   P A G E    Issue 2.05 - May 1996

Giant Magnetoresistance

By Katherine Derbyshire



Full-motion video. Stereo sound. Leading-edge multimedia applications eat disk space for lunch. In 1985, 10Mb hard drives were objects of lust and desire. Now, even 1Gb drives are too small. To keep up, builders of hard drives are searching for ways to cram more bits into less space. The most likely solution is giant magnetoresistance (GMR). It promises to boost storage densities from today's 100Mbits per cm2 to the 5 to 10Gbits needed by the turn of the century, yet relies on familiar magnetic media. Conventional hard drives depend on the close relationship between electricity and magnetism. The read/write head of a drive is a coil of wire. If electrical current is sent through the coil, it induces a magnetic field. A magnet moving across the coil, conversely, creates a current in the wire - its amount depends on how fast the magnet moves.

Floppy drives are a simple example of magnetic storage. The floppy itself is a sheet of plastic with millions of tiny metal shards stuck to its surface. The drive spins the disk while two very small wire coils, called heads, rest on the top and bottom surfaces. A hard drive works on the same principle but has several "platters" of rigid media instead of a single floppy, with at least one head for each side of each platter. Since hard drives retrieve data faster, the platters must spin faster. To avoid damage, the heads hover on a skin of air just microns thick, not touching the platter.

To write data to the disk, a drive sends current through one of the coils, switching it on and off to represent the 1s and 0s of the data stream. The magnetic field induced by the current pulls the magnets on the surface of the disk into alignment, storing the data bits. The same process in reverse reads the data back: as the disk spins, the movement of the tiny magnets induces a cur-rent in the head. The drive interprets changes in the induced current as a stream of binary 1s and 0s.

This is where the problems begin. As storage densities increase, the magnetic region that represents each bit becomes smaller and harder to detect. Writing these tiny magnetic regions can still be done with a conventional write coil, but reading them is much more demanding. The read coil must be positioned closer to the disk and must be wrapped tighter, with more turns. The smaller the spacing between the disk and the head, the more difficult it is to maintain. That's bad: contact between the head and the disk can destroy both the drive and the stored information. Coils with many turns also dissipate more energy. The lost current is converted into heat, which adds electrical noise and makes the data signal difficult to interpret. After a point, adding more turns doesn't improve sensitivity.

To get around these problems, scientists at IBM have been studying an effect called magnetoresistance (MR) since the mid '70s. Magnetoresistive materials have one very special property: when exposed to a magnetic field, they have less resistance to electrical current. So if you put a voltmeter across a disk head made of MR material, the fluctuations in voltage that the meter shows will re-flect the magnetic values stored on the disk. By directly detecting the magnetic field, rather than measuring the change in field as inductive heads do, MR heads gain two advantages. First, they can detect smaller magnetic signals. Second, they don't face the problems with small spacings and thermal noise that currently plague inductive heads.

The simplest magnetoresistive material to make is anisotropic MR. The resistance of an AMR layer depends not only on the presence of a magnetic field but also on the relative orientations (parallel or antiparallel) of the material and the magnets. As the platter rotates past an AMR head, the orientation of the magnets changes, and the resulting resistance change of about 2% is interpreted as data. For storage densities in the 0.155Gbit-per-cm2 range, AMR heads can achieve sensitivities five times greater than inductive heads. Still, 10Gbit storage requires further advances.

Giant magnetoresistance, the next step, is more complicated. In GMR heads, two magnetic layers are separated by a layer of non-magnetic material that is much thinner than that of an AMRhead. Like two bar magnets, the magnetic layers are forced to more parallel, or more antiparallel, positions, by an applied magnetic field. This change in angle affects the head's sensitivity. Today's prototype GMR read heads have twice the MR effect of an AMR head (resistance changes by about 4%), and scientists believe future versions will offer an order-of-magnitude advantage. GMR structures are difficult to manufacture, though. They're made from layers a mere 10 atoms or so thick. Contamination of the thin layers by oxygen or water vapour, or exposure to stray magnetic fields during processing, can seriously degrade the material's MRproperties.

After GMR, the prospects are murky. The next level, colossal magnetoresistance, is a decade or more away from commercial production. In CMR, alternating atomic planes within a single crystal act as magnetic layers. An MR effect thousands of times greater than that seen in GMR occurs, but CMR materials are even more difficult to make reliably. The CMReffect also requires very large applied magnetic fields and cryogenic temperatures. These hurdles must be overcome before the material will be practical for disk drives.

According to some analysts, AMR heads are three times as expensive as inductive heads, so they aren't likely to gain a large share of the market until - at about 0.5Gbits per cm2 - inductive heads just can't compete. Once hard drives that use AMR heads become popular, integration of GMR heads (by 2000 or so) should be a straightforward matter. And with 10Gbit storage density, these drives should be able to handle whatever programmers throw at us. Even Windows 2001.

Katherine Derbyshire (kewms @kew.com) is executive editor of Solid State Technology magazine.