Our need for storage is explosive; fueled by multimedia requirement for text, images, video and audio, storage requirements are growing at an exponential rate and are expected to exceed 1020 bits (12 exabytes) in the year 2000. With an increasing amount of information being generated or captured electronically, a large fraction, perhaps as much as 40%, will be stored digitally. To meet this need, the hierarchy of on-line, near-line and off-line storage systems will be composed of many diverse technologies: magnetic disk drives, magnetic tape drives and tape libraries, optical disk drives and optical libraries. The mix of these sub-systems will be very application-specific so as to optimize the performance and cost of the overall system.
An optical storage system is a particularly attractive component of this hierarchy because it provides data access times that are an intermediate solution between a hard disk drive and a tape drive. Access time is the time, including latency, required to start retrieving a random block of data and typically ranges from less than 10 msec for a hard disk drive, to 30-50 msec for an optical disk drive, and several seconds for a tape drive. It becomes an important link in the chain as data are staged up and down between cpu, memory, and storage.
Perhaps the most enabling feature of optical storage is the removability of the storage medium. With separations of a few millimeters between the recording surface and the optical 'head', and with active servos for focusing and tracking, the medium can be removed and replaced with relatively loose tolerances. The infamous head crashes often experienced in hard disk drives do not occur in optical drives. Data reliability and removability are further enhanced by using the 1.2 mm transparent disk substrate as a protective cover to keep contamination away from the recording surface. (The recently announced digital versatile disk or DVD will have a substrate thickness of only 0.6 mm.)
Removability has created a whole new industry in CD-audio. CD-ROMs have enhanced the efficiency of distribution and use of software, games, and video. These read-only disks containing 680 MB of information can be mass replicated by injection molding in a few seconds for less than 10¢ each, and they are virtually indestructible. One-day express mail of 125 CDs is an effective data rate of 1.0 MB/s.
At the other end of the spectrum, phase change and magneto-optical disks are used in WORM (write-once-read-many) and read/write/erase systems where a single disk can contain almost 5 GB. Using robotics, storage libraries can be assembled with capacities of a terabyte with access to any disk in under 5 seconds.
This review will discuss the fundamentals of optical storage, the components that comprise a system, and the emerging technologies that will allow increased performance, higher storage capacities, and lower cost.
Several international conferences on optical data storage are held each year, and the proceedings of these conferences are a good source of information concerning the latest developments in this field. References (JSAP 1995; OSA 1966; MORIS 1996) provide some information about these conference proceedings and/or digests. There are also several published books in this area, to which the readers are referred for an in-depth coverage of the various subjects (Bouwhius et al. 1985; Marchant 1990; Mansuripur 1995; McDaniel & Victora 1997).
Optical data storage research in the United States dates back to the late 1950s and early 1960s when Bell Labs, IBM, 3M, and Honeywell, among others, developed several key concepts and technologies for optical recording. Several important patents were filed in those early days, which eventually became the property of Disco Vision Associates (DVA), a California-based company with ties to Pioneer of Japan. In recent years, several large U.S. companies (IBM, Eastman Kodak, 3M) have withdrawn from the field of optical storage, either dismantling or selling off their manufacturing facilities in the United States. At the same time a few small companies (MaxOptix, MOST, Laser Byte, Pinnacle Micro), which had carved a niche market for themselves by bringing innovative designs or assembled products to the market, have either failed or are struggling to survive. About 2.5 years ago two startup companies (Quinta and TeraStor) began in California, with the express goal of marrying optical and magnetic recording technologies. Although the promised products of Quinta and TeraStor have not yet materialized in the market, their approach has created a resurgence of interest in the field, especially among the hard drive manufacturers, who see this as a potential threat, as well as an opportunity to overcome the dreaded "superparamagnetic problem."
An optical disk is a plastic (or glass) substrate with one or more thin-film layers coated upon its surface(s). The information may be prerecorded on the surface of the substrate by the disk manufacturer, or it may be recorded on one or more of the thin film layers by the user. Typical diameters of presently available optical disks are as follows:
In the case of read-only media the information is pressed onto the substrate by injection molding of plastics, or by embossing of a layer of photopolymer coated on a glass substrate (Bouwhuis et al. 1985; Marchant 1990). The substrate is then coated with a thin metal layer (e.g., aluminum) to enhance its reflectivity. In other types of optical disks, some information, such as format marks and grooves, may be stamped onto the substrate itself, but then the substrate is coated with a storage layer that can be modified later by the user during recording of information. Typical storage layers are dye-polymer films for write-once applications, tellurium alloys for ablative recording (also write-once), GeSbTe for phase-change rewritable media, and TbFeCo magnetic films for magneto-optical disks (also rewritable) (Marchant 1990; Mansuripur 1995; McDaniel & Victora 1997).
CD technology was developed jointly by Philips of the Netherlands and Sony of Japan in the late 1970s and early 1980s, then introduced to the market in 1983 as the hugely successful compact audio disk. Today most of the research and advanced development work in DVD (the successor to CD) and future generations of read-only media takes place in Japan. Sony, Matsushita, Pioneer, Toshiba, Sharp, and NEC are some of the leading Japanese companies in this area. The MiniDisc (MD) was pioneered by Sony, and despite a sluggish early start, has become a very successful product in Japan. The major players in the field of magneto-optical data storage are Sony, Fujitsu, Hitachi Maxell, Mitsubishi, Nikon, and Sanyo. In phase-change optical recording, the leaders are Matsushita, Toshiba, and NEC, recently joined by Sony. Many other companies are involved in the development of components (lasers, miniature optics, actuators, detectors, substrates, spindle motors, etc.) as well as mastering machines and test equipment for optical recording products. These include some of the major multinational corporations as well as smaller firms.
In a typical disk the storage layer is sandwiched between two dielectric layers, and the stack is capped with a reflector layer (e.g., aluminum or gold), and protected with a lacquer layer. The dielectric layers and the reflective layer perform several tasks: protecting the storage layer, creating an optically tuned structure that has optimized reflectivity and/or absorptivity, allowing the tailoring of the thermal properties of the disk for rapid cooling and reduction of thermal cross-talk during writing, and so on (Mansuripur 1995; McDonald and Victora 1997).
Fig. 5.1 is a schematic diagram showing the basic elements of an optical disk drive. The laser is usually a semiconductor laser diode, whose beam is collimated by a well-corrected lens and directed towards an objective lens through a beam-splitter. The objective lens focuses the beam onto the disk and collects the reflected light. This reflected light is directed (at the beam-splitter) towards the detectors, which produce a data readout signal as well as servo signals for automatic focusing and tracking. The functions and properties of the various elements of this system will be described in some detail in the following sections.
Fig. 5.1. Basic configuration of an optical disk system.
All optical recording technologies rely on lasers as their source of light. The lasers used in optical disk and tape data storage are semiconductor laser diodes of the shortest possible wavelength that can provide sufficient optical power for read/write/erase operations over a period of several thousand hours. Presently the shortest wavelength available in moderate power lasers (around 50 mW) is in the neighborhood of 680 nm; these lasers are being used in CD-R, MO, and PC products. For CD and DVD applications where writing is not a concern, low power lasers (e.g., 5 mW) which emit at 630 nm and 650 nm are being considered. The emphasis on short wavelength lasers for optical recording applications is due to the fact that shorter wavelengths can be focused to smaller spots at the diffraction limit. All things being equal, the diameter of a focused spot scales with its wavelength: a reduction of the wavelength by a factor of 2, for example, will result in a reduction of the focused spot diameter by the same factor and, consequently, a fourfold increase in the data storage density can be realized.
The wavelengths of optical data storage have continuously shrunk during the past 15 years; starting at 830 nm, they are now down to 630 nm, and there is every indication that they will continue to shrink in the foreseeable future. Nichia Chemicals Corporation of Japan recently announced that 50 mW blue lasers operating in the wavelength range of 370 nm to 420 nm would be available before the end of 1998. Several other Japanese companies (e.g., Sony, Matsushita, Pioneer) have demonstrated the feasibility of small, inexpensive, second harmonic generation (SHG) green and blue lasers for use in optical disk drives.
What is needed for optical data storage is compact and inexpensive laser diodes that can be incorporated into small, low-cost drives. The power requirement from such lasers is several milliwatts for read-only media and several tens of milliwatts for recordable media. The lasers should be capable of direct modulation (e.g., by modulating the electrical current input to the laser), otherwise the cost and the size of external modulators may become prohibitive. Spatial coherence and single transverse-mode operation is a requirement, because the beam must be focused to diffraction limit. (In this context vertical cavity surface emitting laser diodes (VCSELs) as well as arrays of such lasers need to be improved, since, at high powers, these lasers tend to operate in high-order modes.) Low-noise operation is very important, especially in applications such as magneto-optical readout, where signal-to-noise ratio is at a premium.
Although operation of laser diodes in several longitudinal modes is presently acceptable, for future devices it may be important to add operation in a single, stable, longitudinal mode to all their other desirable characteristics. Mode hopping, wavelength shifts of several nanometers with temperature fluctuations and with operating current, manufacturing variability of the wavelength from batch to batch, etc., are so severe that, at the present time, it is not possible to consider the use of diffractive lenses either for the collimator or for the objective lens. These high numerical aperture (NA), diffraction-limited lenses are still produced by molding of glass elements. In the future, when microminiaturization becomes a necessity and lenslet arrays begin to appear in commercial products, operation of the laser in a single, stable, longitudinal mode may be required.
Practically all semiconductor diode lasers used in optical storage products come from Japan. Hitachi, Sharp, and Toshiba are the leading producers of semiconductor lasers, and Nichia Chemicals is in a position to dominate the market in the area of GaN-based short-wavelength (green/blue/ultraviolet) lasers. In the United States there are pockets of excellence and know-how both in the industry and among the universities. Spectra Diode Laboratory (SDL), OptoPower and Ortel are manufacturers of specialty laser diodes in this country. Lucent Technologies and Motorola, among others, have significant programs in laser diode arrays and VCSELs. But, by and large, the major source for inexpensive laser diodes is outside the United States.
Presently disk and tape optical recording systems use molded glass lenses for focusing the laser beam to a diffraction-limited spot (see Fig. 5.2). These lenses consist of two aspheric surfaces on a single piece of glass, have fairly large numerical apertures (in the range of 0.4 to 0.6), and are essentially free from aberrations. The numerical aperture of a lens is defined as NA = sin , where is the half-angle subtended by the focused cone of light at its apex. A 0.5 NA lens, for example, will have a focused cone whose full angle is 60°. The diameter of the focused spot is of the order of 0/NA, where 0 is the vacuum wavelength of the laser beam. It is thus clear that higher numerical apertures are desirable if smaller spots (and therefore higher recording densities) are to be attained. Unfortunately, the depth of focus of an objective lens is proportional 0/NA2, which means that the higher the NA, the smaller will be the depth of focus. It thus becomes difficult to work with high NA lenses and maintain focus with the desired accuracy in an optical disk drive.
Fig. 5.2. Single-element aspheric lenses used in optical disk drives.
But a small depth of focus is not the main reason why the present optical drives operate at moderate numerical apertures. The more important reason has to do with the fact that the laser beam is almost invariably focused onto the storage medium through the disk substrate. The disk substrate, being a slab of plastic or glass, has a thickness of 1.2 mm (DVD substrates are only half as thick, or 0.6 mm). When a beam of light is focused through such a substrate it will develop an aberration, known as coma, as soon as the substrate becomes tilted relative to the optical axis of the objective lens. Even a 1° tilt produces unacceptably large values of coma in practice. The magnitude of coma is proportional to NA3, and therefore, higher NA lenses exhibit more sensitivity to disk tilt. Another aberration, caused by the variability of the substrate's thickness from disk to disk, is spherical aberration. This aberration, which scales with the fourth power of NA, is another limiting factor for the numerical aperture. In the future, manufacturers will move toward higher numerical apertures by adopting one or more of the following strategies:
As was mentioned earlier, the use of smaller lenses is always desirable in optical storage technology, particularly if lens arrays are being considered for parallel accessing of multiple tracks in a system. In this respect, gradient-index (GRIN) lenses, holographic optical elements (HOEs), and binary diffractive optical (BDO) lenses are all being considered for future generations of optical storage devices. In the United States, the Florida-based Geltec, Inc. is a major innovator as well as manufacturer of molded glass optics. Many Japanese companies (e.g., Sony) make their own molded glass or plastic lenses for use in optical disk drives. NEC is particularly strong in design and implementation of holographic optical elements (HOEs). Nippon Sheet Glass (NSG) is a major manufacturer of GRIN lenses and other micro-optical elements.
With the rapid changes taking place in the optical recording industry, as exemplified by the convergence of magnetic and optical technologies in the recently announced Quinta and TeraStor products, there will be strong demand for innovative micro-optical elements. In order to stack multiple platters within one drive, it is imperative to reduce the size of the optical head; this can be achieved only if the sizes of the collimator, objective lens, polarizers, beam-shaping optics, etc., are substantially reduced below their current values. With companies like Geltec in the lead, the United States is currently in a good position to take advantage of this window of opportunity. But the Japanese technology is also advancing rapidly, and the U.S. lead may soon evaporate.
A typical optical disk has a plastic substrate that is not perfectly flat, but is slightly warped. Also, when mounted in a drive, small tilts of the axis could cause vertical motions of the disk surface during operation. It is not unusual to find vertical movements as much as ±100 µm during the operation of an optical disk. A typical objective lens has a numerical aperture of 0.5 or higher, and therefore, the focused beam has a depth of focus of a fraction of 0/NA2, which is only a fraction of a micron. The focused spot must remain within the depth of focus while the disk rotates at speeds of several thousand rpm and wobbles in and out of focus by as much as ±100 µm in each revolution. Needless to say, without an autofocus mechanism to maintain the disk continually in focus, the operation of an optical disk drive is unthinkable.
In practice the objective lens is mounted in a voice coil actuator (bandwidth = several kHz), and a feedback mechanism is used to drive the lens towards and away from the disk in such a way as to maintain focus at all times. The signal needed for this feedback mechanism is derived from the light that is reflected from the disk itself. Fig. 5.3 shows a diagram of the astigmatic focus-error detection system used in many practical devices these days. The light reflected from the disk and collected by the objective lens is either convergent or divergent, depending on whether the disk is further away from best focus or closer to the lens than the plane of best focus. This returned beam goes through an astigmatic lens, which normally focuses the incident beam to a symmetric spot halfway between its focal planes. A quad detector placed at this plane (also called the plane of least confusion) then receives equal amounts of light on its four quadrants. When the disk is out of focus, however, the astigmat creates an elongated spot on the detector. Depending on the sign of defocus, this elongated spot may preferentially illuminate quadrants 1 and 3 or quadrants 2 and 4 of the detector. Therefore, the combination signal (S1+S3) - (S2+S4) provides a bipolar focus-error signal, which is fed back to the voice coil for maintaining focus automatically.
Fig. 5.3. Astigmatic focus-error detection system.
Servo design and manufacture has a stronghold in Japan, but other East Asian nations such as Taiwan, South Korea, and Singapore have recently made significant contributions to this area. Sony, Fujitsu, NEC, Pioneer, Sharp, Toshiba, and Samsung Electronics of Korea have a mature base for miniaturized servo-control technologies. As a result, they are in a strong position to develop faster components for future generation optical drives. It is perhaps safe to declare the art of miniaturization of small mechanical devices (e.g., servo motors and actuators) dead in the United States. However, silicon-based micro-electromechanical devices (MEMS) are finding their way into the field of optical storage, and U.S. industry and universities are quite strong in this field. Once again the U.S. position is open to challenge from the Japanese and other East Asians, since they recognize the significance of this field and are moving quickly to close the gap.
The information on an optical disk is recorded either around a series of concentric circular tracks or on a continuous spiral. Manufacturing errors and disk eccentricities caused by mounting errors, thermal expansion of the substrate, etc., will cause a given track to wobble in and out of position as the disk spins. Typically, a given track might be as much as ±100 µm away from its intended position at any given time. The focused spot, of course, is only about 1 µm across and cannot be at the right place at all times. An automatic tracking scheme is therefore desired. Given the mechanical rotation rates of the disks, the frequency response of the actuator needed for track following does not have to cover more then a few kHz, and a voice coil is usually sufficient for the purpose. The feedback signal, for controlling the position of the objective lens within the tracking coil, is again provided by the return beam itself. Several mechanisms have been proposed and have been put to use in commercial devices. Three of these schemes are mentioned here.
The push-pull tracking mechanism relies on the presence of either grooves or a trackful of data on the media. In the case of CD and CD-ROM, the data are prestamped along a spiral on the substrate, and the sequence of marks along the spiral comprises a sort of discontinuous groove structure. The discontinuity is irrelevant to the operation of the tracking servo, however, because it is at a much higher frequency than the tracking servo is designed to follow. Writable media such as CD-R, MO and PC require a tracking mechanism distinct from the data pattern, because prior to the recording of data, the write head must be able to follow the track before it can record anything. Once the data are recorded, the system will have a choice to follow the original tracking mechanism or to follow the recorded data pattern. Continuous grooves are a popular form of preexisting tracks on optical media (see Fig. 5.4a). A typical groove is a fraction of a micron wide (say, 0.4 µm) and one-eighth of a wavelength (/8) deep. As long as the focused beam is centered on a track, diffraction of light from the adjacent grooves will be symmetric. The symmetry of the reflected beam, as sensed by a split detector in the return path, would produce a zero error signal (see Fig. 5.5). However, when the focused spot moves away from the center of the track, an asymmetry develops in the intensity pattern at the split detector. The bipolar signal thus generated from the difference signal is sufficient to return the focused spot to the center of the track.
In read-only media, the three-beam method of tracking has been extremely popular. The laser beam is divided into three beams, one of which follows the track under consideration, while the other two are focused on adjacent tracks, immediately before and after the desired track. Any movement of the central track away from its desired position will cause an increase in the signal from one of the outriggers and, simultaneously, a decrease in the signal from the other outrigger. A comparison of the two outrigger signals provides sufficient information for the track-following servo.
Fig. 5.4. Recordable optical disks: pregrooved media (a) and sampled-servo media (b).
Fig. 5.5. Track-error signal generated by push-pull method.
In yet another possible tracking scheme, the so-called sampled servo scheme, a set of discrete pairs of marks is placed on the media at regular intervals (see Fig. 5.4b). Since these marks are slightly offset from the track center in opposite directions, the reflected light first indicates the arrival of one and then the other of these wobble marks. Depending on the position of the spot on the track, one of these two pulses of reflected light may be stronger than the other, thus indicating the direction of track error.
Much of what was said about servo actuators in the previous section applies to tracking servos and controllers as well. Advances in track following will probably occur in the area of micro-electromechanical devices (MEMS). Tracking actuators are also needed for future generations of hard disk drives, and a convergence of ideas and techniques used in optical and magnetic recording fields is quite likely here. The United States is a powerhouse in hard disk-related technologies with companies such as IBM, Seagate, Quantum, Read-Rite, etc. working feverishly to develop tracking schemes for disk drives. In Japan, Sony, Hitachi, Fujitsu, Pioneer, NEC, and several other companies are quite strong in this field.
Another related subject is the issue of tracking strategies beyond those that are already known and described above. For example, Polaroid Corporation has introduced a new concept that goes beyond the standard sample-servo format and is flexible enough to allow the drive manufacturer to select any one of a number of possible values for the track-pitch. Such innovations will pave the way for future generations of optical disk drives which use advanced techniques (such as near-field optics) and may not be compatible with the older methods of tracking.
Polycarbonate is a fairly strong, inexpensive, moldable plastic that is currently the material of choice in optical disk fabrication. The pattern of pits and grooves is readily impressed onto the surface of this substrate during injection molding. The transparency of polycarbonate at red and near-infrared wavelengths allows the focused beam of light to reach the storage layer through the substrate, an important factor for removable disks, since focusing through the substrate will keep dust, fingerprints, and scratches on the front facet of the disk well out of focus. Thermal and mechanical properties of polycarbonate are acceptable for present day needs, and its multitude of applications outside the field of optical storage has brought its price down to almost negligible values.
On the negative side, polycarbonate is a birefringent plastic and affects the polarization state of the beam as it travels through the substrate. Substrate birefringence is particularly troublesome in MO data storage, where the readout signal is embedded in the polarization state of the return beam. If one denotes the refractive indices of the substrate along the radial, azimuthal, and vertical directions by nr, n, and nz, then in-plane birefringence implies that nr ≠ n and vertical birefringence implies that nr ≠ nz. Careful balancing techniques applied during substrate manufacture have made it possible to reduce the in-plane birefringence to negligible values, but the remnant vertical birefringence and the fact that in-plane birefringence may return at elevated temperatures have kept alive the search for better materials. For example, amorphous polyolephin, which is essentially free from birefringence, has been shown to be an excellent substrate material. Its high price (because of its low volume of applications in other areas), however, has so far been an impediment to its use in optical data storage.
Transparency of plastics will become an issue in the future when very short wavelength lasers become available. Also, if disk flatness happens to be an issue, and recent trends toward flying optical heads and near-field recording indicate that it might be, then glass and aluminum substrates may be more suitable alternatives for these applications. Already several Japanese manufacturers have started R&D programs to develop glass substrates for optical disk applications. Characteristics that are deemed desirable for glass to become an acceptable substrate material are low cost, flatness, hardness, low stress birefringence, and the ability to be patterned with grooves and preformat marks.
In the United States, General Electric Co. is a leading manufacturer of polycarbonate. Other chemical companies (such as Hoechst-Celanese) are also knowledgeable about plastic substrate materials, although they do not seem to have an active development program for optical recording applications. Until recently Eastman Kodak and 3M manufactured their own polycarbonate disk substrates, but these operations have now come to an end. In Japan the work on polycarbonate substrates in many companies has been ongoing for almost two decades, and the level of expertise in this field is quite staggering. Sony, Fujitsu, Pioneer, Mitsubishi, Hitachi-Maxell, NEC, and Nikon, among others, have refined the art and science of substrate manufacturing to the point that they can now routinely fabricate high quality, rigid, low-birefringence substrates for optical disks.
Presently all commercially available MO disks are based on an amorphous terbium-iron-cobalt [Tbx(FeyCo1-y)1-x] magnetic alloy. Typical compositions have a value of x 0.2 and y 0.9. This material belongs to a class of materials known as the rare earth-transition metal alloys. (Terbium is a rare earth element, while iron and cobalt both are transition metals.) The TbFeCo alloy has several interesting properties as described below (Mansuripur 1990).
The only U.S. firm that could manufacture MO disks until recently was 3M, but that program seems to have come to an end. With the recent excitement generated by Quinta and TeraStor corporations, it is likely that some hard disk manufacturers in the United States might have started their own internal research and development activities, but, to this author's knowledge, no one is making commercial-quality MO disks in the United States. In Japan there are powerhouses such as Sony, Hitachi-Maxell, Maxell-Nikon Optical (MNO), Mitsubishi Chemicals, TDK, and several other firms, which have extremely good products on the market. These companies are also in a good position to address the future needs of drive manufacturers, whether these manufacturers demand conventional disks for through-substrate applications, or front-surface disks for near-field applications. The manufacturing processes, as well as the quality of thin magnetic and dielectric films and the substrates, are continually improving in Japan, making it virtually impossible to compete with the low cost and high quality of their MO products.
Presently the medium of choice for erasable phase-change recording is a Ge2Sb2Te5 alloy, which is affectionately referred to as the GST material (Ohara, Akahira and Ishida 1996). This alloy is sputter-deposited on a plastic substrate, with an undercoat and an overcoat of ZnS-SiO2 dielectric layers. The stack is then capped with an aluminum alloy layer for making an antireflection structure. The quadrilayer stack is also effective as a rapid cooling structure, thanks to the heat-sinking properties of the aluminum layer. The as-deposited GST alloy is amorphous. However, each disk is annealed at the factory to transform the recording layer into its polycrystalline state. The recording process turns small regions of the GST medium into amorphous marks, by raising the local temperature above the melting point and allowing a rapid cool down quenching. The reflectivity of the amorphous mark is different from that of the polycrystalline background and, therefore, a signal is developed during readout. Erasure is achieved by using a laser pulse of an intermediate power level (i.e., between the read and write powers). If sufficient time is allowed for the laser beam to dwell on the amorphous mark, the mark will become crystalline once again (annealing). This process is compatible with direct overwrite and is therefore preferable to MO recording, where direct overwrite is harder to achieve.
In the United States, Energy Conversion Devices (ECD), Inc. of Troy, Michigan is the technology leader in the field of phase-change optical media. ECD also owns some of the basic patents on PC materials. In Japan Matsushita Electric Co. is the leader, with Hitachi, Toshiba, and a few other companies close behind. There is significant activity also in S. Korea, Taiwan, and Singapore. The challenge for PC media manufacturers in the next few years is to improve the media in order to increase the recording speed and cyclability. The PC drives are not very different from MO and CD-R drives; therefore, advances in optical components that benefit other areas of optical recording will benefit the PC drives as well. In the area of drives, however, there is currently no U.S. presence, and all commercial products come out of Japan. There are indications, however, that some of the U.S. firms may be gearing up to jump into the fray and produce their own drives. There are certain methods of recording and readout in the near-field that are particularly suited to PC (as opposed to MO) media. It is advisable, therefore, that we in the United States should devote some of our resources to investigating the combination of media and read/write/erase systems in the near future, as we pursue greater densities and higher data rates.
Comparing PC and MO technologies, one can find several advantages and disadvantages for each. PC drives are simpler than MO drives, because they do not need magnets to create external magnetic fields, and also because there is no need for sensitive polarization-detecting optics in PC readout. The read signal is very strong for PC media, so much so that despite the rather large component of media noise of PC, the SNR is still somewhat larger than that of MO media. On the other hand, repeated melting, crystallization, and amorphization of PC media results in material segregation, stress buildup, micro-crack formation, etc. These factors tend to reduce data reliability and cyclability of the PC media. MO disks are guaranteed to sustain over 106 read/write/erase cycles and can probably do better than that in practice, but the corresponding figure for PC media is typically one to two orders of magnitude lower. The maximum temperature reached in MO media during recording and erasure is typically around 300°C, as opposed to 600°C in PC media. The lower temperatures and the fact that magnetization reversal does not produce material fatigue account for the longer life and better cyclability of the MO media. Writing and erasure in MO media can be very fast, fundamentally because spin flips occur on a sub-nanosecond time scale. In contrast, although amorphization in PC media can be very rapid, crystallization is a rather slow process: the atoms must be kept at elevated temperatures long enough to move around and find their place within the crystal lattice. As a result, in principle high data rates are achievable with MO media, but there may be barriers to achieving them in PC media. On the other hand, PC media are directly overwritable, but MO media can be overwritten either with magnetic field modulation (which is a rather slow process) or by using exchange-coupled magnetic multi-layers (which are more difficult to manufacture than the conventional single-magnetic-layer disks).
A new approach to optical disk data storage involves the use of near-field optics in general and the solid immersion lens (SIL) in particular (Kino 1994). The SIL approach requires that a part of the objective lens fly over the surface of the storage medium, as shown in Fig. 5.7. The hemispherical glass of refractive index n receives the rays of light at normal incidence to its surface. These rays come to focus at the center of the hemisphere and form a diffraction-limited spot that is smaller by a factor of n comparedt to what would have been in the absence of the SIL. (This is a well-known fact in microscopy, where oil immersion objectives have been in use for many years.) A typical glass hemisphere having n = 2 will reduce the diameter of the focused spot by a factor of 2, thus increasing the recording density fourfold. To ensure that the smaller spot size does indeed increase the resolution of the system, the bottom of the hemisphere must either be in contact with the active layer of the disk or fly extremely closely to it. For a disk spinning at several thousand rpm, it is possible to keep the SIL at a distance of less than 100 nm above the disk surface.
Fig. 5.7. Solid immersion lens.
The rays of light that are incident at large angles at the bottom of the hemisphere would have been reflected by total internal reflection, except for the fact that light can tunnel through and jump across gaps that are small compared to one wavelength. This tunneling mechanism is known as frustrated total internal reflection, and its presence qualifies the application of SIL in optical data storage as a near-field technique. The price one will have to pay for the increased storage density and data rate afforded by the SIL is the necessity to permanently enclose the disk within the drive, thereby making it non-removable. (The possibility of maintaining removability in an SIL system has been suggested, but it remains to be demonstrated in a practical setting.) The United States is currently in a leading position in this field, with companies like TeraStor pursuing a commercial product vigorously. There was substantial interest and curiosity on the part of WTEC's Japanese hosts in this subject, as evidenced by the numerous questions that they asked WTEC panelists about the technical and business aspects of the TeraStor approach. Although the Japanese seem to be playing catch-up at this point, it will not be long before they can make flying optical heads using SIL or some such technique. After all, many of the components that TeraStor is using in its experimental drive come from Japan, and it is only a matter of time before the Japanese engineers learn how to put these items together and build a complete system. Sony already has a working system with a variant of the solid immersion lens, one that is not working in the near-field yet, but can provide insight into the subtleties of SIL-based systems. Perhaps what makes the SIL concept so attractive to the Japanese industry is its ability to bridge the gap between optical and magnetic recording technologies. Traditionally, hard disk drives have been the domain of American companies, whereas Japan has been strong in optical recording. The SIL has the potential to marry these two technologies, thus giving the Japanese industry an opening to capture at least a fraction of the hard disk market.
It has been found that by making the land and groove of equal width, and by recording the information on both lands and grooves, it is possible to eliminate (or at least substantially attenuate) the cross-talk arising from the pattern of marks recorded on adjacent tracks (Fukumoto, Masuhara and Aratani 1994). Fig. 5.8a shows a typical pattern of recorded data on both lands and grooves. It turns out that for a particular groove depth (typically around /6) the cross-talk from adjacent tracks reaches a minimum. Fig. 5.8b shows the computed cross-talk signal obtained from a theoretical model based on scalar diffraction theory. These results are in excellent agreement with the experimental data, indicating that cross-talk cancellation is a direct consequence of diffraction from the grooved surface and interference among the various diffracted orders.
Land-groove recording works well in phase-change media, and, in fact, this is where it was originally discovered. For MO systems, the birefringence of the substrate creates certain problems; specifically, the presence of in-plane birefringence will make the optimum groove depth for land reading different from that for groove reading. To overcome this problem either better substrates (with smaller in-plane birefringence) should be developed, or a servo system must be deployed to automatically correct the effects of birefringence. So far, the laboratory results of land-groove recording on MO disks have been very encouraging, and there is little doubt that this technique will play a major role in future generations of both PC and MO devices.
The Japanese engineers are quite comfortable with land-groove recording at this point. Major contributions in this area have come from Matsushita, NEC, and Sony, but just about every company in Japan and South Korea seems to have developed laboratory versions of the land-groove disk and studied its properties (Figs. 5.8a, 5.8b). To this author's knowledge, however, no one in the United States has made any contributions to this field. Land-groove recording will be an important aspect of high-density recording in the near future, and deserves more attention.
Figure 5.9a shows a diagram of an exchange-coupled magnetic trilayer. In this system the top and bottom layers have perpendicular magnetic anisotropy, while the intermediate layer is an in-plane magnetized layer. The function of the intermediate layer is to ease the transition from the top to the bottom layer. The top and bottom layers typically have different compositions, magnetic moments, thicknesses, coercivities, etc. By exchange-coupling these two layers together, it is possible to lock their magnetic behavior to one another, so that switching one layer would make the other layer either more likely or less likely to switch (Mansuripur 1995; McDaniel and Victora 1997). When the two layers are magnetized in opposite directions, a domain wall develops between them (going through the intermediate layer, of course). This domain wall, which has energy associated with it, makes the collapse of a domain formed in either of the layers much more likely; in other words, it makes a domain such as that in Fig.5.9b less stable and more susceptible to external disturbances, compared to a domain that goes through the entire thickness of the magnetic stack. Using exchange-coupled magnetic multi-layers, engineers have created media designs with many useful and novel capabilities.
Fig. 5.8a. Typical pattern of land-groove recording.
Fig. 5.8b. Cross-talk signal from a model based on scalar diffraction theory.
One such structure allows light-intensity modulation direct overwrite (LIMDOW) in magneto-optical disks. In these media the recorded domains, being of the type shown in Fig 5.9b, collapse under a focused laser beam of moderate power. Low-power beams that are used for readout do not disturb the recorded domains, and high-power beams create domains that go through the entire thickness of the stack; only moderate-power beams are capable of erasing a pre-existing mark. Thus by switching the laser power between moderate and high levels, one can create a desired pattern of domains irrespective of what data have been previously recorded in that location.
Fig. 5.9a. Exchange-coupled magnetic trilayer.
Fig. 5.9b. Cross-sectional view of magnetic domains in an exchange-coupled magnetic multi-layer.
Another magnetic multi-layer structure enables high resolution readout based on the concept of magnetic super resolution (MSR) (Kaneko and Nakaoki 1996). In this scheme the recorded data is kept in a storage layer; then, during readout, the recorded marks are presented to the read beam one at a time. This is achieved by selective copying of the marks (i.e., magnetic domains) to the read layer using the temperature induced in that layer by the focused laser spot. The essential features of MSR are shown in Fig 5.10. The two schemes shown in this figure are known as front aperture detection (FAD) and rear aperture detection (RAD). In Fig. 5.10a the laser beam heats up the read layer, temporarily erasing the domains within the heated region. The mark in the front aperture region is then detected. In Fig. 5.10b the rear aperture area, which is heated up, yields to the storage layer and accepts its magnetic state, thus receiving the domains and exposing them to the read beam. The advantage of MSR is that, since it presents the domains one at a time to the read beam, it does not suffer from intersysmbol interference and cross-track cross-talk. More important perhaps is the fact that the minimum mark length that is readable in an MSR system is shorter than any that are readable in conventional systems. There is a lower limit, known as optical cutoff, to the mark lengths in conventional optical recording. MSR can overcome this limit and read below the optical cutoff.
More recent work in magnetic multi-layers has involved layers that are separated by thin, non-magnetic dielectric films. The coupling between the various magnetic layers in these structures is no longer mediated by exchange, but it is by means of magneto-static forces, whereby the stray magnetic field from one layer acts on the magnetization of another layer. For example, one proposed MSR technique, referred to as central aperture detection (CAD), has an in-plane-magnetized read layer which is separated from the perpendicularly magnetized storage layer by a thin dielectric film. During readout, the laser beam raises the temperature of the in-plane layer, making it possible for the stray magnetic field from a recorded domain (which resides within the storage layer below) to align the local magnetization of the readout layer with its own (perpendicular) direction of magnetization. A variation on this theme is the newly proposed concept of magnetic amplifying magneto-optical system (MAMMOS). Here, after copying the recorded mark to the readout layer, an external magnetic field is used to expand the copied domain, thereby amplifying the readout signal. Clearly, coupled magnetic multi-layers open the door to many new possibilities, and we expect to see many of these innovations in future generations of MO products.
Fig. 5.10. Exchange-coupled magnetic multi-layer used in magnetic super resolution (MSR) as seen in front aperture (a) and rear aperture (b).
Exchange-coupled magnetic multi-layers are the exclusive domain of the Japanese industry. The original concepts of DOW and MSR were developed by Nikon and Sony, and many contributions were made later by a host of other Japanese companies. Essentially nothing novel has been contributed to this field by the U.S. industry. Hitachi Maxell and MNO (Maxell Nikon Optical) are proficient in the manufacture of exchange-coupled magnetic disks. The future seems to belong to MSR and MAMMOS and their derivatives. Unfortunately, the United States seems to have been left behind in this area, and unless something is done to change the situation drastically, we are bound to be "followers."
Magnetic drive manufacturers generally use a stack of several magnetic disks to achieve the desired storage capacity within the small volume of a hard drive. The small size and the low cost of magnetic heads allows them to use a separate head for each disk surface without compromising the overall size and price of the drive. In contrast, optical heads are very expensive and rather bulky. To achieve volumetric data storage with optical disks, designers have sought methods that rely on a single head for accessing multiple platters. In 1994 IBM researchers demonstrated a system that could read through a stack of six CD surfaces using a single optical head (Rubin et al. 1994). Their system was very similar to a conventional CD player, except that they had taken special care to correct the spherical aberrations that result when the beam of light is focused through a substrate at different depths. The IBM stack consisted of three thin glass disks (thickness 300 Ám) on both surfaces of which the data pits had been embossed. Unlike standard CDs, these disks were not metallized because the focused laser beam had to pass through several such layers before reaching the desired surface. A bare glass surface reflects about 4% of the incident light, and this was apparently sufficient to enable the detection system to retrieve the data and to play back, with high fidelity, the recorded audio and video signals.
Technically, the method described above is straightforward and requires only that the objective lens be corrected for focusing through different thicknesses of the substrate. The separation between adjacent surfaces must be large enough to reduce cross-talk from the data marks recorded on neighboring surfaces. With a 0.5 NA objective lens, a separation of 40 or 50 microns is typically enough to assure acceptable levels of cross-talk from these other surfaces.
In the case of writable media, recording and readout of information on multiple platters is more difficult, primarily because storage layers absorb a significant amount of the laser light. (In their 1994 demonstration, IBM researchers also described a four-layer WORM disk and a two layer MO disk.) Recently double-layer MO and PC disks have been proposed and demonstrated in several industrial laboratories around the world. Again the separation between recording layers has been kept at about 40 microns to reduce cross-talk, and the laser beam has been strong enough to write on the second layer even after half of its power has been absorbed by the first layer. (The power density at the first layer is reduced by a factor of almost 2000; hence, no writing occurs at that layer.) Since focusing through an additional 40 microns of plastic is well within the range of tolerance of typical objective lenses, correction for spherical aberration has not been necessary in these double-layer systems.
Stacked optical disks have obvious advantages in terms of volumetric capacity, but the technological barriers for rewritable media are substantial. Japan is currently in a leading position in this area, with a clear shot at double-layer DVD. Matsushita, Sony, Toshiba, and Pioneer are the leaders, but there is significant know-how in other companies as well.