Computer users tend to take for granted the existence of magnetic and optical memory devices that offer breathtaking data-storage capacity with extremely high reliability at very low cost. The technological wonders of these rugged devices routinely escape users’ notice, because the internal functioning of disk drives is hidden from view. Nevertheless, magnetic and optical drives are arguably among the world’s most sophisticated electromechanical devices. Equally astonishing is the rapid advance of this technology, far outpacing the highly visible technology sectors of semiconductor electronics and computer software.
Magneto-optical (MO) data storage represents a combination of optical data-storage techniques and magnetic storage media. This article aims to show that this hybridization is an ongoing process, with the most recent developments being among the most exciting in the 40-year history of the technology. Two of the most striking physical characteristics of magnetic and optical drive technology are the exponential growth of the data-storage density in commercial products and the mechanical performance of recording heads following a track of recorded information on the medium.
The storage density metric is called “areal” density, or bit count per unit area on the medium. It has recently been doubling every 8 to 12 months in magnetic disk storage and about every 24 months in MO storage, according to DISK/TREND, Inc., a market research firm in Mountain View, Calif. This advance translates into an exponential drop in the cost of a stored bit of information—value to the customer unheard of in any other industry.
The mechanical operation of disk drives is stunning when one realizes that a “flying” magnet head in the drive is analogous (through scaling) to a 747 aircraft flying over rolling terrain a few inches off the ground. The interface between the aerodynamically supported “slider” and the disk surface containing the user’s precious data must be maintained reliably (without crashes) over the entire average five-year life of the drive. Optical drives are perhaps equally amazing in tracking data “remotely,” with about 50,000 times greater separation between the head and the disk. A mechanical servo-control system provides for a 1µm-diameter focused light beam to follow a comparable-size feature on the disk surface to within 0.1µm accuracy while the surface undergoes vibratory motions in two dimensions with amplitudes in the range of 100µm.
In the 25 years since its earliest commercial development, optical data storage has become an industry that generates about 15% of the total annual data storage market revenues of roughly ~$60 billion worldwide, and it probably accounts for a similar fraction of the total installed worldwide digital data storage capacity (~1018B, according to DISK/TREND and the National Storage Industry Consortium of San Diego, Calif.). The most common optical storage devices today are the ubiquitous compact disc (CD) and the emerging digital versatile disc (DVD) used in audio, video, and computer applications for distributing content and information.
As optical disc storage has developed into a robust technology in the removable media sector over the past 15 years, rewritable optical discs have also progressed. The leading rewritable optical media types today are MO and phase-change, which share somewhat complementary attributes. MO technology has led the way in opening markets for rewritable, removable optical storage, with several generations of International Standards Organization (ISO) drives (commonly found in optical jukeboxes) and the Sony MiniDisc (commonly found in consumer applications primarily in the Far East and Europe). This article offers an overview of MO data storage technology  and looks ahead to emerging directions that will make aspects of MO technology more important in magnetic disk drives in the future.
The usual advantage of optical disc storage is media removability combined with large-capacity random-access data storage. MO is the rewritable optical-recording method best suited for high-performance, extended-lifetime, high-density applications. MO storage material is a thin magnetic film (similar to magnetic disks). The recording process is thermally assisted magnetic recording (atomic magnet reorientation), known to data storage engineers as a rapid and dependable process with practically infinite cyclability. Phase-change media, while providing a popular and viable low-cost alternative write-once or rewritable option, are inferior to MO in both raw performance (data recording rate and storage density) and cyclability. In phase-change recording, the recording/erasure processes involve crystalline-amorphous atomic structural changes (atoms move around); therefore these processes are slower and more prone to wear-out phenomena.
The similarity of MO and magnetic recording is the basis for an expectation of an exciting synergy in the future. The merger and hybridization of these heretofore independently successful technologies is a possibility, as reflected in a string of research and development efforts publicized over the past five years [14, 6, 7].
Role of Optical Data Storage
Optical recording emerged as a viable technology in the 1960s and 1970s when it was made technically feasible by the development of low-cost, compact light sources of adequate power, namely solid-state diode lasers. In classical optical recording, the system designer arranges for an intense, focused light beam to interact with a storage medium (see Figure 1a). Focusing light implies that the location of the light focal point in the medium is relatively remote (a few millimeters) from the optical components that generate and guide the light beam. These so-called “far-field” optics represent a key distinction from magnetic recording, whereby the head is placed very close to the recording medium (today less than 0.1µm).
The optical readback process is performed with the same light beam, though with optical power reduced from the writing level by a factor of from 3 to 6. In reading, the light is used to sense the induced physical change from the writing process, usually through some combination of optical reflection, transmission, and change in the state of the light beam’s polarization, or orientation of the internal electric field in the electromagnetic wave.
Some variants of this standard form of optical recording are known to optical storage engineers. One approach is “near-field recording,” whereby an optical advantage can be gained by utilizing the light in very close proximity to its zone of emergence from the light generation and guidance device. Another approach is to use “volumetric” optical storage. In one example, the storage medium has multiple recording layers. A focused, far-field light beam is easily refocused at variable depths in the medium, making feasible multilayer storage media, provided the individual layers are sufficiently transparent. Volumetric storage can multiply the storage capacity of a disk or tape significantly, since the third dimension of the storage medium is used, instead of just spreading the data over a plane.
Optical Storage Context
To place MO recording in a useful context, the characteristics of today’s common optical storage methodologies need to be defined. Figure 1(a) outlines the process of a focused light beam writing a track of information on a moving optical storage medium. This writing is carried out by modulating the optical power sent to the focusing lens, resulting in local physical change in the recording material. The substrate material carrying the recording medium may be a disk, tape, or card.
The recording configuration in Figure 1(a) is a serial one, because the recorded marks are created sequentially on a track as the medium moves under the focused beam. In a serial digital recording application, this scheme might be called “bit-by-bit” recording. This approach is contrasted with a possibly more parallel method in which multiple beams record multiple data streams concurrently or in which blocks or frames of data might be recorded optically at one instant at various 3D positions in the medium by an extended beam, as in holographic recording (see Orlov’s “Volume Holographic Data Storage” in this section). This article is limited to serial recording systems.
Types of optical recording can be differentiated further according to the functionality provided to the user and by the reversibility of the writing process. Audio and computer CDs are the most common types of read-only optical media (often called ROM for “read-only memory”). In this case, the media are pre-written at the production factory, with specified information (such as recorded music or a set of computer program files) replicated thousands of times. The original information content was recorded onto a master disc by a process very much like that in Figure 1(a), then accurately copied by a mass replication process onto low-cost media the end user can read but cannot write on. The replication process for ROM discs is highly parallel; a single disc is made from the master in a few seconds (~1091010B replicated simultaneously). The physical embodiment is a sequence of small, light-scattering pits along the disc track.
ROM optical media are contrasted with two forms of writable media: write-once, read-many (WORM) and rewritable, erasable, or write/read (W/R). WORM media is exemplified by CD-recordable media (CD-R) allowing the user to write information on the disc one time, then read it back an unlimited number of times. WORM media is preferred when a vendor wants to create a certified body of information that cannot be altered without extraordinary means. Consequently, professional-quality WORM optical media has become a legal standard as an information repository when an audit trail must be established and preserved, since ROM and WORM optical media involve essentially irreversible writing processes.
In order to allow multiple recordings, rewritable optical media must be readable at every stage. Therefore, the recording process must be highly physically reversible. However, few physical processes are perfectly reversible, so a rewritable medium also needs a measure of cyclability, that is, an expected count of the number of reliable rewrites the medium can sustain. Because different physical processes are available to implement optical rewritability, engineers find there are different regimes of cyclability for different embodiments of W/R media. MO media excels by meeting an ISO specification to sustain at least 106 writing cycles and at least 107 reading cycles without unacceptable recording degradation. By contrast, CD-RW exemplifies a consumer-rewritable optical-disc product based on a phase-change medium that may have a factor of 102104 times lower cyclability than typical MO media.
Note that W/R media (magnetic or optical) are used in ROM or WORM mode when adequate software or hardware protections are in place to prevent inadvertent overwriting or erasure of data.
MO recording is a form of magnetic recording in which light is used as a source of medium heating in the writing and erasure processes (thermomagnetic recording) and as a probe of the magnetic state of the medium in reading (see Figure 2). In each of the figure’s three panels, a thin film of magnetic material deposited on a smooth substrate (such as a disc and tape carrier) is shown in cross-section. The magnetic polarization of the MO material is oriented perpendicularly to the film plane, a necessity for MO readout to work. This magnetic orientation requires careful material selection and processing, since it is usually energetically favorable for the magnetic polarization to lie in the plane of the film (longitudinal orientation), as in magnetic disks.
Heating MO media is necessary for writing and erasing, as in Figure 2(a) and (c). In general, magnetic recording is achieved when an applied magnetic field overcomes the medium’s resistance to switching, called its “coercivity.” All magnetic materials steadily lose their magnetic properties, however, including coercivity, as their temperature is elevated. To record information at high density on a surface implies that the region of controlled switching is very small. In MO recording, this condition is achieved by combining a relatively uniform magnetic field from a coil device with strong localized heating from a focused light beam (see Figure 3). (The applied magnetic field is about 600 times stronger than the Earth’s compass-influencing field.) When the medium cools to room temperature, the freshly reversed magnetic polarization is said to be “frozen in.”
Two distinct means of thermomagnetic recording of a magnetization pattern along a track on the moving medium are found in MO drives, one using laser intensity (power) modulation (LIM), the other using magnetic field modulation (MFM) (see Figure 4). In LIM writing, the magnetic field is held constant; in MFM recording, the laser power can be kept on continuously or pulsed at exactly the data clock rate.
A binary data bit sequence can be encoded in magnetic domains on the medium in two ways: In pulse position modulation (PPM), the drive records a binary 1 or 0 corresponding to the existence or absence, respectively, of a small circular magnetic domain. Alternatively, the drive may use pulse width modulation (PWM), whereby a binary 1 or 0 corresponds to the existence or absence, respectively, of a magnetization transition from + to or from to +. This method is used in magnetic recording (see Figure 4 for a comparison of these methods). PWM encoding has advantages for achieving greater linear-bit density, since even the smallest circular domain encodes binary bits at both its leading and trailing edges. Consequently, information can be packed more densely with an equivalent number of written features.
Magneto-Optical and Magnetic Recording Compared
Because the commercial application of MO recording is mostly in disk drive devices, this article compares MO only with magnetic disk drives. For many years, personal computers have included both floppy disk drives and hard disk drives (HDDs). A floppy disk drive is a relatively low-cost, low-performance device that supports removable magnetic media of relatively low capacity (typically 1.44MB per cartridge). Removable media are convenient for nonelectronic transport of user data files between computers.
An HDD, by contrast, is a device within which the magnetic disk(s) is normally sealed; it provides much higher performance (faster data access and throughput) and higher capacity than a floppy drive. Magnetic recording involves close-range interaction between the head and medium. A clean environment is critical for maintaining this mechanical interface. Media removability is an obvious convenience to users but introduces a significant reliability risk for the storage device itself. There have been some widely used products with removable hard disks, and as expected, they offer significantly improved performance and capacity compared to floppy drives but are less dependable than conventional HDDs.
Compare this situation to MO drives. All optical drives excel in media removability with superb reliability—a direct consequence of the far-field head being well removed from the disk. When light is focused through the substrate onto the disk’s second surface, as in Figure 3, the outer surface need not be pristine. Some amount of surface dust and dirt is tolerated by the optical system, because the light beam is roughly a million times more diffuse when passing through the exposed entry surface of the disk than it is at the focal point. The use of second-surface focusing means it is tolerable to not include a protective cartridge around a CD (though gross contamination causes light blockage, absorption, or scattering that is eventually detrimental to performance).
Besides media removability and device reliability, three other characteristics are of interest to users of disk storage devices: capacity, speed, and cost. When comparing the storage capacity of magnetic HDDs and MO drives, areal density has to be considered. Figure 5 compares average areal density over time for HDDs and MO drives. Until recently, MO recording had an advantage, because optical drives had a much higher track density (number of tracks per unit radial distance on the disk). Since 1990, however, magnetic recording has shown a significant increase in the annual growth rate of areal density, initially doubling from the historical average rate of about 30%60% with the widespread introduction of magnetoresistive head technology. More recently, that rate has increased to 100%200% annually, depending on the product. The MO recording industry has been managed more conservatively, reflecting an annual areal density growth rate of about 40% from 1992 to 2000; the same rate is now projected out to 2008.
When comparing MO drives and HDDs, it is important to remember that a huge storage capacity advantage for removable media systems results from the fact that capacity per drive is limited in part by the number of disk cartridges associated with a particular drive—a number without theoretical limit. Drive cost comparisons are somewhat complicated, since an HDD is bundled with dedicated media capacity, while an MO drive is not. MO drives are generally considered to be the most expensive of the optical disk drives, because an optical head for MO readout involves specialized optics for sensing the polarization state of reflected light. Moreover, MO drive manufacturing has probably not yet capitalized on economies of scale, generally due to limited market penetration. (Cost is both the cause and the result of relatively modest product volumes.)
Optimal Readout Process
Readout in MO recording is a matter of detecting the pattern of magnetization in the storage medium utilizing rotation of the plane of polarization upon reflection of polarized light from a magnetic mirror—called the magneto-optic Kerr effect, as in Figures 2(b) and 3. The drive’s reading power is increased enough to raise the signal-to-noise ratio (SNR) as high as possible without heating the track to a level that degrades the written magnetic information. A system’s designer tries to maximize the MO signal while keeping the reflectivity below 30%. The remaining fraction of light power is absorbed during writing and must be high enough to heat the disk efficiently.
A viable data channel in a recording device requires adequate SNR; therefore, the designer must consider noise minimization. The principal noise components (called “media noise”) in an MO recording system are associated with the laser, the read-channel electronics, the light detectors, and the disk. Differential detection, as in Figure 3, cancels some but not all of these noise components. SNR is perhaps the single most important parameter governing data-channel performance, and thus has the greatest influence over the storage system performance metrics of interest to the user—capacity and data-throughput rate. In general, great care must go into a disk’s optical, thermal, and magnetic designs in order to achieve sufficient and balanced system performance.
Since the mid-1990s, several new research thrusts have begun to promise enhanced MO technology applicability and value from this form of rewritable optical storage. There is an important distinction between the schemes to extend the diffraction limit in readout discussed earlier and the more revolutionary approaches to MO recording in terms of system architecture. The following are some of the most notable developments.
High-density MO drive. In late 1995, TeraStor of San Jose, Calif., began developing a novel MO drive using focusing optics with a solid immersion lens—a moderately high refractive index lens placed in close proximity (<<l) to the disk recording surface (not a near-field technique) . This design effectively increases the numerical aperture of the light-incident medium, reducing the spot size, as in Figures 1(c) and (d), and increasing the recording areal-density potential. However, TeraStor disbanded in 2000 without shipping a product.
Flying MO head technology. In 1996, Quinta, a data storage startup company in San Jose, Calif., began developing flying MO head technology. When Seagate acquired Quinta in 1997, this technology was named Optically Assisted Winchester. In it, micro-optics and a microcoil are attached to a flying slider (the carrier for a HDD head), along with an optical-fiber light-delivery system to achieve a low-mass, low-cost means of realizing the first-surface configuration for MO recording. This design introduced MO recording technology into an HDD architecture while preserving HDD performance. Seagate has exhibited prototype drives but has not yet marketed a product based on the original concept.
Hybrid schemes. Hybrid schemes combining magnetic and optical recording elements show promise in the laboratory of outperforming traditional magnetic-only component combinations. Hitachi, Philips, and Sharp have publicized their work in this area [3, 6, 7], while other companies have expressed interest in such approaches. A common element in these methods is the incorporation of laser light into the recording and readout processes. Momentary heating of the recording medium is an effective way to mitigate a looming problem of writability in ordinary magnetic recording by using the thermomagnetic recording concept from MO technology. (Media coercivity in HDDs is being increased steadily to preserve the magnetic stability of ever-shrinking magnetic bit cells; writing heads have a definite physical limitation in their output magnetic switching fields.)
Near-field optical schemes. Research in the early 1990s demonstrated MO recording in domains smaller than 0.1µm by using a near-field optical source. Such an approach overcomes the diffraction limit of light, allowing electromagnetic field dimensions to be determined by the physical extent of the source, such as an aperture in a waveguide and the width of a laser cavity. While this technique allows much smaller optical “spots,” such spots exist only in the proximity of the source, with the allowable separation typically on the same order as the source dimension. Moreover, such near-field schemes normally have the drawback that the transmission efficiency for optical power is extremely low. Therefore, much more efficient near-field optical sources need to be developed to produce energy flux sufficient for thermomagnetic recording.
Some of these recent thrusts could extend optical recording beyond its historical regimes. MO storage has already proven itself in applications requiring very rugged, highly reliable, removable, rewritable optical media. The devices commercialized so far have won acceptance in the professional and consumer markets that demand high storage capacity and moderate random data-access performance with media removability. These applications are fundamentally different from those addressed by HDDs.
Implementation of new approaches in conventional MO recording shows promise in boosting areal density 10-fold. For example, a number of techniques—magnetic super-resolution (MSR), magnetic amplifying MO system (MAMMOS), and domain well dynamic displacement (DWDD)—improve MO readout resolution without needing a smaller focused light spot, which would be physically limited by the available light wavelength and objective lens. Meanwhile, the convergence of optical storage approaches with magnetic storage is an exciting new development. MO recording offers the key technologies of thermomagnetic recording and patterned media, which may be instrumental in alleviating a slowdown in HDD advances [2, 6]. The HDD industry is beginning to appreciate the potential of these approaches to help sustain its historical progress in cost and performance. Such hybridization would be a remarkable development for the HDD industry, which for the past 40 years has relentlessly pursued incremental progress based on scaling, punctuated by the timely introduction of stepwise improvements in component technology. MO technology is well-positioned to continue the flow of engineering solutions to users with an insatiable appetite for storing information.
Figure 4. Two means—LIM and MFM—of thermomagnetic recording of a magnetization pattern along a track on the moving medium. In PPM coding, binary Is are represented by the center position of the small nearby circular marks. In PWM coding, binary Is are represented by edge marks, which can be of N clock lengths, whereby the range of N depends on the particular code being used.