Magnetic Storage Industry Continues to Grow and Grow...
By Robert M. White
Figure 1: Cross sections of read sensors used in hard disk drives today. In the bottom synthetic spin valve, the bottom magnetic layer (CoFe) is part of a (two-layer) synthetic antiferromagnet, which is pinned by a “real” antiferromagnet (IrMn). In the spin filter device, there is a layer of copper on top, which provides additional scattering and enhances the magnetoresistance.
Figure 2 - HDD Technology Demonstrations: This graph shows the areal densities demonstrated since 1990 as well as the target of the National Storage Industry Consortium (NSIC), a consortium of companies such as IBM, Seagate, Quantum, Read-Rite, etc. involved in the hard disk drive (HDD) industry.
Storage, like networking, has several levels. At the upper level, there are applications like "datamining" and "data warehousing." As the amount of data grows, there is increasing need for a middle level of automated data management tools. Standards are also being developed to define the management and control of data objects so different storage devices can easily work together on a network. However, for physicists, the interesting story is at the physical level. The figure of merit in storage is how many bits can be stored in a square area. This metric has increased by nearly a factor of ten million since IBM introduced the disk drive in 1956. It is remarkable that this dramatic improvement in storage has been based on an electromechanical technology - the hard disk drive (HDD) that depends upon maintaining a read/write stylus 5 nm from a data surface that is moving at 40 m/s!
Storage densities today are increasing at their highest rate in history, over 100% per year. This growth is leading to lower storage prices, currently approximately 10¢ per megabyte. Another interesting fact about storage is that a 1% decrease in price leads to a 4% increase in demand. The corresponding ratio for semiconductors is only 1.5.
Before describing the most recent developments in recording, let us review the basics. In digital magnetic recording, "1's" and "0's" are stored in the form of magnetic transitions or the absence of magnetic transitions in a longitudinally, magnetized coating. These data may be recorded circumferentially around circular tracks on a disk or serially along a tape. Originally, these magnetic coatings consisted of magnetic particles embedded in a binder. Today, the disk in a hard disk drive consists of a very thin sputtered film on an aluminum or glass substrate. Tapes may still contain iron particles or evaporated films. Data is written onto the medium by an electro-magnetic transducer consisting of a copper coil with a highly permeable core which is photo lithographically produced.
For many years, data was read by inductively sensing the magnetic fields associated with the magnetic transitions in the medium. In 1990, IBM introduced a magnetoresistive sensor. Since the amplitude of the readback signal is proportional to the width of the recording track, introducing the more sensitive magnetoresistive transducer meant it was possible to reduce the track width thereby increasing the track density. The initial magnetoresistive sensors employed permalloy, an alloy of Ni and Fe, which has a magnetoresistance of 2%. That is, there is a 2% change in the resistance when the magnetization changes from being aligned with the current to being perpendicular to it. This is referred to as the anisotropic magnetoresistive effect.
This attack on track density was greatly aided by the discovery in 1988 of the "giant" magnetoresistive effect (GMR). Albert Fert and his colleagues in France discovered that the change in magnetoresistance in the plane of a sandwich of two magnetic films separated by a thin conductor, such as copper, depended upon the relative orientation of the magnetizations in the two magnetic films. Experimental studies indicate that this effect is associated with interfacial scattering. When the magnetizations are parallel, majority spins can scatter both forward into the other magnetic layer or backward into the layer from which the spin originated. Anti-parallel magnetizations restrict the scattering. This structure is therefore referred to as a spin "valve."
This effect was first discovered at low temperatures and required large fields to align the moments in the two magnetic films. Subsequent research led to room temperature operation with magnetoresistances above 10% in fields of a few oersteds. In a recording application, one of the magnetic layers is pinned by placing it adjacent to an antiferromagnet.
In order to reduce the influence of magnetostatic fields from the pinned layer on the free layer, an oppositely magnetized layer is added making the pinned layers look like a synthetic antiferromagnet. Figure 1 shows several cross sections of GMR heads used in drives today. The writing coil and reading sensors are deposited on the back vertical surface of a small block or slide. The bottom surface of the slider contains channels, which guide the flow of air when the slider rides above the spinning disk. This flow of air creates a stiff suspension that maintains the fixed and incredibly small spacing between the writing and reading elements and the data surface.
A major challenge today lies in the media. The thin magnetic film consists of approximately 100 Å of an alloy of cobalt. It is prepared by sputtering which gives it a granular structure. Typical grain sizes are 20 Å. If there are N grains within a bit cell the signal, being coherent across the cell, is proportional to N2 while the noise varies as N. Therefore, the signal-to-noise is proportional to N. As the bit cell decreases with increasing density, if the SNR is to remain constant, the grains must be made smaller. However, if the grains become too small, they become thermally unstable. This is referred to as the superparamagnetic effect and leads to degradation of data.
The first estimate of the superparamagnetic limit to recording density was 40 Gbits/in2. Slightly more than 50 Gbits/in2 has been demonstrated in the lab. This indicates that superparamagnetism may not be as limiting as originally feared. In fact, we believe we should be able to achieve 1 Terabit/in2 (see Figure 2).
There are two strategies currently being pursued toward this goal. One is to increase the volume of the bit cell by making it deeper. One way of doing this is to record perpendicular to the plane of the medium. This will require new writing and reading designs as well as new signal processing, or "channel" techniques.
The second strategy is to increase the coercivity of the medium making it more "resistant" to thermal switching. The difficulty with this approach is that increasing the coercivity also makes it more difficult to write. Nevertheless, there is a very interesting solution - thermally assisted writing! If the material can be "designed" so that its coercivity decreases with temperature, then the simultaneous application of a magnetic field and heating will enable writing on high coercivity media. This is not unlike what occurs in a traditional magnetooptic recording system today. What is particularly appealing about this idea is that the "footprints" associated with the magnetic field gradient and thermal gradient can be made different so that their intersection results in a very small-recorded spot.
The "Terabit Challenge" will require a deeper physical understanding of nanoscale magnetics. It will also very likely require new discovery. It is interesting to note that the continual march of areal density to higher values has benefited from such relatively recent fundamental discoveries as the magnetic force microscope, the giant magnetoresistive effect, and spin-dependent tunneling.
Last year, IBM announced a 250 MByte drive with a one-inch form-factor, the so-called micro drive. One can hardly imagine the storage applications if the perpendicular or hybrid magnetic-optical technologies described above are employed in this form-factor!
Robert M. White is University Professor, Electrical and Computer Engineering, and Director of the Data Storage System Center at Carnegie Mellon University.
©1995 - 2017, AMERICAN PHYSICAL SOCIETY
APS encourages the redistribution of the materials included in this newspaper provided that attribution to the source is noted and the materials are not truncated or changed.
Associate Editor: Jennifer Ouellette