New approaches to storage can be classified in terms of the degrees of freedom they utilize as shown in Fig. 7.3. One possible degree of freedom that is being considered for future data storage systems is recording multiple values per spot. A typical method that uses this degree of freedom is pit depth recording. Assuming a material that provides 16 distinct pit depths and enough signal separation for detection, this approach may result in a 4x improvement in areal density over conventional methods. However, the dynamic range available from the storage materials limits the impact of this approach.
A powerful direction for increasing the areal density that has driven progress in data storage traditionally is to decrease the bit size. Storage technologies that record and read from the surface of a medium are constrained to this degree of freedom. These technologies include magnetic, near-field optics, and probe storage. Depending on the technology in hand, bit sizes can be reduced significantly, providing a potential for up to 1000x improvement over conventional approaches. However, reliability, servo systems and mostly cost issues may impose practical engineering limits in the future.
In addition, optics can be used to record and retrieve information from the volume of a transparent storage medium. This is an important degree of freedom for optical storage if it can be harnessed at low cost. Although multi-layer phase change disks have been demonstrated and produced (2 layers), and methods are being investigated for magneto-optics to take advantage of this degree of freedom, this approach is not truly scalable for conventional optical storage media. Indeed, optical power considerations as well as aberrations and interlayer crosstalk limit the scalability in the third dimension to less than 16 layers in conventional optical storage. However, two methods--two-photon recording and holographic storage--promise to fully exploit this degree of freedom by providing potential areal densities in the order of Tb/in2 and possibly at low costs. Currently, suitable materials and peripheral optoelectronic devices are still in development.
Fig. 7.3. New degrees of freedom in data storage.
Finally, the advances made in optoelectronic device arrays and MEMS may enable data to be accessed in parallel. This may lead to a significant increase in data transfer rates enabling applications such as content addressed data mining to become practical.
As noted earlier in the introduction, longitudinal magnetic recording, which has been increased in areal density by approximately 2.5 million times since 1957, is predicted to exhibit superparamagnetic thermal instabilities in the recordings when the density reaches 36 Gb/in2, which is less than one order of magnitude higher than the areal density on current disk drives. Some of the Japanese researchers the WTEC team visited predicted that conventional longitudinal magnetic recording may be limited to even lower densities, such as 20 Gb/in2, because of grain size variations in the media. Although a change in the bit aspect ratio from about 20:1 to the order of 4:1 or 6:1 has been predicted to make it possible to achieve densities of the order of 100 Gbit/in2, it is inevitable that the technology must change if the rapid 60% per year increase in areal density, which has been the trend for the past seven years, is to continue to much higher densities.
Various new magnetic storage technologies are being explored to circumvent the barriers that these superparamagnetic effects impose. Alternative magnetic storage technologies that the WTEC team saw being investigated in Japan included perpendicular magnetic recording, thermally or optically assisted magnetic recording and patterned media recording. These technologies, the advantages they may offer and the work being done on them, are discussed in the following sections.
Perpendicular Magnetic Recording
One possible means of extending magnetic recording density beyond what can be achieved using longitudinal magnetic recording is to use perpendicular magnetic recording. The medium is magnetized perpendicular to the film plane, rather than in the plane.
As a result of the perpendicular orientation of the magnetization in the medium, a recorded transition does not contain large magnetostatic charge as it does in a longitudinal medium. Instead, magnetostatic charge is mostly at the bottom and top surfaces of the medium. Consequently, whereas in longitudinal magnetic recording the demagnetizing fields tend to increase with recording density and reduce the thermal stability of the recordings, in perpendicular magnetic recording the demagnetizing fields tend to decrease with the recording density, possibly enabling higher densities to be recorded, if the code used for recording prevents long strings of zeros (no transitions) from being recorded in the medium.
Moreover, many Japanese researchers argue that thicker media may be used in perpendicular magnetic recording. Since grains typically extend through the medium thickness, by using a thicker medium, the grain volume, and therefore the stability of the medium against superparamagnetism, is enhanced.
Finally, with perpendicular magnetic recording, it is, in principle, possible to fabricate the medium on a high permeability magnetic underlayer. This causes an image of the main pole of the recording head to be formed in the underlayer, so that the medium is effectively in the gap of the head. Whereas with conventional longitudinal magnetic recording, the recording is done with the fringing field from the head, with perpendicular recording using a high permeability magnetic underlayer, the perpendicular medium is effectively placed in the gap of the head. This enables much higher write fields from the recording head, which in turn enable much higher anisotropy (and therefore more thermally stable) magnetic materials to be recorded.
Exactly how large the advantage of perpendicular recording is with regard to extending the recording density beyond what can be achieved with longitudinal recording is not yet clear; however, most researchers do admit that it probably does offer some advantage. Although U.S. companies worked on perpendicular magnetic recording in the early 1980s, there has not been much activity since; whereas in Japan there has been a continuous effort since the late 1970s, and it is clear that the Japanese are considerably more knowledgeable about the technology than U.S. researchers in the field. Thus, the U.S. magnetic disk drive industry is at some risk with regard to finding itself behind its Japanese competitors if a switch to perpendicular magnetic recording comes when the superparamagnetic limits of longitudinal recording have been reached.
Thermally Assisted Magnetic Recording
As noted in the introduction, superparamagnetic thermal instability occurs in a medium when the magnetic anisotropy energy KUV of a grain of the medium becomes sufficiently small that thermal energy KBT may cause the magnetization to switch. Hence, to increase the thermal stability of the medium, it is necessary to increase the magnetic anisotropy energy density KU. However, if the magnetic anisotropy becomes too large, it may become impossible to produce the field necessary to record on the medium with available magnetic recording head materials.
One way to overcome this barrier is to use thermally assisted magnetic recording.
Heating a magnetic material generally causes the magnetic anisotropy energy density KU to decrease. Hence, the idea of thermally assisted magnetic recording is to use thermal energy to lower the KU of a medium sufficiently that it may be written with available magnetic recording head materials and then to cool the medium back to ambient conditions, so that the thermal stability of the large KU material is achieved.
Continuously exchange coupled perpendicularly oriented magnetic media like that used for magneto-optic recording may be written in this manner, and recent work has shown that domains as small as 20 nm are stable for times in excess of 10 years in such media.
Thermally assisted magnetic recording is commonly used in magneto-optic recording, where a focused laser beam is used to heat the medium; however, magneto-optic recording uses the Kerr (or Faraday) magneto-optic effects for readout of the recorded information. Since the magneto-optic effects are relatively small, it is difficult to achieve high signal-to-noise ratio when reading out by this technique, and the bandwidth of magneto-optic disk drives is typically restricted to smaller values than magnetic disk drives.
It was suggested by Dr. Miura of Fujitsu Laboratories (during his presentation to WTEC of the research being carried out by the Storage Research Consortium and ASET) that an approach using optically assisted writing and magnetic readout may have merit to overcome both of these limitations.
Patterned Media Magnetic Recording
Yet another method of overcoming the barrier to increasing the areal density of magnetic recording imposed by superparamagnetic effects is to use patterned or structured media.
In conventional magnetic recording, to achieve an acceptable signal-to-noise ratio, it necessary that there be of the order of 100 grains in the volume of a bit cell. This is because the transitions tend to follow the grain boundaries, and therefore have a significant variance in their position relative to where they were intended to be written. By making the grain size smaller with respect to the bit cell dimensions, the variance is reduced.
The idea of patterned or structured media recording is to pattern or structure the media in such a way that the medium consists of small islands of magnetic material in a regular array on a planar surface, presumably with small variance in the location of the bit cells. Ideally, for a rotating disk medium the array would be circumferential, which would be difficult to produce except lithographically, but other symmetries are possible, and self structured arrays are an alternative. Advocates of this approach argue that, in such a structure, it would be possible to have the entire bit cell be magnetically coupled or even a single grain. Thus, a 100-fold increase in areal density could be achieved before superparamagnetic effects set in. Assuming a density of 100 Gb/in2 is possible with longitudinal or perpendicular magnetic recording, this suggests that densities approaching 10 Tb/in2 could be achieved in patterned or structured media recording. If achievable in practical devices, this would enable several orders of magnitude density gain and a couple more decades of growth in the areal density of magnetic storage.
Although there was some mention of patterned media recording by some of the companies the WTEC team visited in Japan, panelists were not shown any work on it.
One of the novel optical storage technologies that may have a major impact in the next millennium is probe storage (Fig. 7.4). A large majority of the Japanese companies visited were investing significant R&D efforts towards developing practical probe storage systems. This area is also actively investigated in U.S. institutions.
Fig. 7.4. Probe storage: a combination of sensing and modulation techniques that can be used with a variety of media for ultra high areal densities.
Probe storage techniques can generally be broken down into those that use a scanning tunneling microscope (STM), field emission probe (FEP), an atomic force microscope (AFM), and near field scanning optical microscope (NSOM). These techniques can employ a variety of modulation techniques including topographic (mechanical), charge, magnetic, conductivity and optical modulation. They have been used and demonstrated with a large variety of materials including organic, ferro-electric, magneto-optic, magnetic and phase change media.
The scanning tunneling microscope takes advantage of the tunneling current that occurs across the gap between two conductors when they are held very close together (on the order of angstroms) (Eigler and Schweizer 1990). With current flow, atoms may be deposited onto a substrate. With an STM it is actually possible to manipulate a single atom, making the approach the highest density storage device demonstrated to date. However, most engineering efforts today concentrate on larger bit sizes (i.e., 10-20 nm size bits, giving a data density of over 1 Tb/in2).
The main drawback to the STM is its slow data rate. Although individual bits may be recorded and read very quickly (using a voltage pulse of less than a nanosecond), the scanning of the STM probe is very slow. The height of the probe tip must be held very constant or else it will lose the tunneling current or crash into the substrate. Therefore scanning speeds are limited by the servo actuating speeds. Data rates as fast as 100 kb/s have been demonstrated (Mamin et. al 1995).
Materials have been developed for read/write/erase applications using STM which undergo phase change (which changes their electrical conductivity); however, these materials may suffer from fatigue and slow recording times (Sato and Tsukamoto 1993).
The atomic force microscope employs a cantilevered tip that is scanned across a sample. In this case the tip is usually in contact with the sample surface. Features on the surface of the sample cause deflections of the tip that can be detected optically by monitoring the back surface of the tip with a laser and split photodiode detector.
Write once read only memory (WORM) writing with an AFM may be achieved using a thermo-mechanical process. The position of the AFM tip is placed concurrent with the focal point of a laser. When a pulse is sent out from the laser it locally heats the area around the tip causing the tip to sink into the polymer substrate creating a pit. Using this technique, pits as small as 100 nm have been recorded giving a recording density of roughly 30 Gb/in2. The recording speed for that experiment was 200 kb/s and readout was up to 1.25 Mb/s.
It may also be possible to produce inexpensive ROM replicas readable by an AFM. Disc "masters" have been fabricated using e-beam lithography (Terris et al. 1996). The masters have then been used to replicate data in a photo-polymer. Features as small as 50 nm have been successfully replicated.
Read/write/erase AFM storage has been done in a number of different materials including charge trapping materials such as nitride-oxide-semiconductor (NOS) layers (Fujiwara, Kojima and Seto 1996) and also phase change materials the same those being used in optical discs (Kado and Tohda 1995). The smallest bits that have been recorded are on the order of 75 nm giving a recording density of over 100 Gb/in2. Readout rates of 30 kb/s have been achieved, but predicted rates of up to 10 Mb/s may be possible.
The major obstacle for AFM storage is tip wear. The tip is slowly worn down as it drags across the sample, reducing the resolution of the readout.
Near-field optical recording includes any technology that allows one to surpass the areal density limit imposed by the diffraction of light in optical data storage. Near field scanning optical microscopy (NSOM) combats the effects of diffraction on the optical spot size by placing the pick up heads very near (about 50 nm above) the media. A near-field probe can produce spots as small as 40 nm in diameter and conceptually can achieve areal densities in the order of 100 Gb/in2. The problem is that the probe must be near contact with the medium, making it difficult to prevent head crashes and support removable media.
One of the original near-field recording techniques was through the use of a tapered fiber (Betzig et al. 1992). The tip of a fiber, which is smaller than the wavelength of the recording light, is positioned within 10 nm of the sample. Using magneto-optic materials, bits as small as 60 nm have been recorded and read out. The tapered fiber approach suffers from very low optical efficiency.
STM, AFM and NSOM can all provide ultra high areal densities. However, the total area available for storage restricts the capacity per device (chip). Assuming a scan area of 1cm2, 30 nm effective spot size yields a total capacity of 10 GB only. Yet using parallel access and MEMS technology, relatively fast access can be achieved. Considering the low power nature of these approaches, we expect this type of probe storage to have a potential market for applications that require portability.
Perhaps a more scalable approach, in terms of capacity per device, is the use of a solid immersion lens (SIL) (Terris et. al 1994; Terris, Mamin and Rugar 1996; Ichimura, Hayashi and Kino 1997). The SIL reduces the actual spot size by both refracting the rays at the sphere surface and by having an increased index of refraction (up to n = 2) within the lens. However, this small spot size can only exist within the SIL; therefore, the evanescent wave from the bottom surface of the lens must be used for recording. This means that the SIL must be held within 100 nm or less of the recording surface. Maintaining such a distance over a rapidly spinning disc is possible using a technique borrowed from hard disk storage: the flying head. The flying head floats on a cushion of air above the disc and is able to hold the head in very close proximity to the disc without crashing. SILs have been fabricated with equivalent numerical aperture of up to 1.8. Using 780 nm wavelength, recording spot sizes on the order of 320 nm have been achieved, giving recording densities on the order of 3 Gb/in2. Using the flying head, data rates of up to 3 Mb/s were achieved, and data rates of up to 15 Mb/s may be possible.
During the WTEC panel's visit to Japan, panelists saw significant progress in various areas of probe storage. This includes hybrid STM/AFM systems, automated Langmuir-Blodgett film deposition systems, new organic polymers (Fig. 7.5), record breaking small recorded spots (Fig. 7.6 and Fig. 7.7), and probe fabrication techniques (Fig. 7.8) that can lead to low cost manufacturable systems. Dr. Ohta at Matsushita perhaps best underlined the expectations of Japanese researchers in probe storage by stating that, "phase change media welcomes probe storage."
Fig. 7.5. Recording by STM/AFM probe in organic LB films (Yano et al. 1996).
Fig. 7.6. Probe storage at Canon: (a) record breaking spot sizes were achieved using organic LB films over an area as large as 4 micron square (Yano et al. 1996).
Fig. 7.7. Probe storage at Matsushita: record breaking erasable spots of 10 nm diameter have been recorded in phase change media using STM (Kado and Tohda 1995).
Fig. 7.8. Manufacturing process for low-cost fabrication of probe tips and their integration with the actuators (Yagi et al. 1997).
Research work in the United States is equally active with IBM's efforts in atom herding and thermo-mechanical recording depicted in Fig. 7.9 (Terris et al. 1996). Most of the work in the United States emphasizes demonstration of parallel access probe storage techniques using FEP techniques (e.g., work carried out at Hewlett Packard) as well as MEMS techniques that utilize phase change, ferro-electric and magnetic media (e.g., work carried out at Cornell (Miller, Turner and MacDonald 1997) and Carnegie Mellon).
Near field optical storage using the SIL techniques is being investigated both in Japan and in the United States. Terastor, a start-up company in the United States, is working towards commercialization of a magneto-optic disk system that utilizes an SIL lens. Several Japanese drive manufacturers are involved with Terastor's activities and are pursuing separate R&D efforts as well. Institutions involved in this direction include NAIR, Hitachi, Fujitsu, and Toshiba.
Fig. 7.9. Thermo-mechanical recording at IBM using rotating plastic disk using an atomic force microscope tip. At selected locations, electrical current was pulsed through the tip causing the tip to be heated. The heat from the tip softens the substrate and the pressure from the tip causes an indentation to be formed. Marks smaller than 0.1 micron can be written in this manner (Mamin et al. 1995).
Fig. 7.10. Classification and demonstrated performance of probe storage and comparison with present storage techniques. Recent results from Canon and Matsushita have been added to the figure (Modified after IBM) (Mamin et al. 1995).
Several approaches to 3D optical storage are being actively investigated both in the United States and Japan. These include the extension of present disk systems based on PC and MO media to a layered format, and 3D layered disk systems recorded by two photon absorption and read by fluorescence and volume holographic storage, where data are recorded in a distributed fashion in the volume. In addition, more futuristic approaches are being investigated to utilize the wavelength as the third dimension.
Layered 3D optical storage is a natural extension of present optical disk systems with a potential to increase the media capacity without significantly affecting the cost of the drives. In this case, the volumetric density (Mb/in3) rather than the areal density is the relevant figure of merit. However, in order to be able to compare layered disks with conventional disks we will use the effective areal density as the figure of merit. This figure corresponds to the number of bits that can be accessed from one side of a disk divided by the top surface area of the disk. This figure must, however, be used with caution and should be applied to volumetric disks with thickness not significantly larger than conventional disks.
Multi-layer recording extends the effective areal recording density and capacity per disk. It therefore enables the penetration of new markets with technologies in hand and eliminates pick-up head duplication. Since data can be made more local to the pick-up head a larger amount of data can be accessed faster. In addition, when the number of layers can be made large, as in the case of two-photon recording, the approach can relax constraints in other dimensions, providing significant cost saving. For example, for larger effective areal density, a multi-layer disk system may use larger spot size and use low cost, lower numerical aperture optics and servo systems.
However, there are also limitations to the third dimension. These include higher optical power requirements depending on the transparency of the material, dealing with aberrations and inter-layer crosstalk while imaging through the volume, extra costs associated with focusing and tracking in three dimensions, and smaller signal levels requiring more complex signal processing and error correction schemes.
Multi Layer Phase Change
In the United States, IBM researchers have shown that the addressing of four to 16 active layers with a single optical head is possible. The number of layers were clearly limited by interlayer crosstalk considerations and the capabilities of the dynamic focusing lens addressing the data in the third dimension. In IBM experiments, several conventional CD-ROM disks were stacked together, each with one active layer.
During discussions with Matsushita representatives, the WTEC panelists learned that a four-layer DVD RAM is also possible by using more transparent recording layers in PC medium. The panelists also noted Hitachi's roadmap for PC disks shown in Fig. 7.11 (Asthana 1994). Chapter 1 of this report covers multi-layer DVD-RAM approaches. Thus the panel expects to see over the next few years two and four layer DVD-RAM systems. However, most researchers believe that a practical limit to the maximum number of layers in PC media would perhaps be 8 layers considering recording power requirements and crosstalk limitations.
Fig. 7.11. Multi-layer optical disk stack (IBM).
The success of double layer DVD-ROM has attracted interest in whether it is possible to use multiple layers with the MO media. A possible approach is shown in Fig. 7.12 (Nakagawa et al. 1997). The disk medium consists of two active TbFeCo MO layers, one modified by the presence of an additional PtCo layer. Thus the characteristics of the two recording layers are slightly different in terms of required power and wavelength. This difference is exploited by wavelength selective thermo-magnetic recording to selectively record onto each layer using two different lasers of different wavelengths. During readout the effect of the two layers on the polarization state of the readout beam is additive on the polarization state of a read-out beam. However, since the total Kerr rotation angle of two reflected waves of two different wavelengths is dependent on the linear combination of the retardation imposed by each layer, by detecting each wave independently it is possible to calculate the state of the recorded bits in each layer. This requires the solution of a linear system of two equations with two unknowns (the bit states) in real time, a rather straightforward operation with present capabilities of electronics.
It is possible in theory to extend this approach to more than two layers using more than two wavelengths. However, the number of lasers, optical components and receiver circuits as well as the head mass grow with growing number of layers. Nevertheless, this approach can significantly benefit from technologies being developed for wavelength division multiplexing (WDM) communication over fiber optic networks and may fuel their manufacturability further. This type of forward-looking R&D in Japan demonstrates the importance attributed to multi-layer recording approaches.
Fig. 7.12. University of Tokyo and Hitachi Maxell Ltd.'s approach to double layer MO disk recording and readout.
Two-Photon Recording: Extending Multi-Layer Recording to More than a Hundred Layers
The two-photon recording approach used by Call/Recall, Inc. in the United States relies on recording bits in a volume by using two-photon absorption (Hunter et al. 1990). As described in Fig. 7.13, a spot is written in the volume of a molded organic polymer only in locations where two beams with sufficient photon energies, one carrying the information and the other specifying the location, intersect temporally and spatially. The recorded bits are then read by fluorescence when excited by single photons absorbed within the written spot volume.
Fig. 7.13. Principle of two-photon recording in 3D (Call/Recall, Inc.).
Using this method, multiple layer ROM disk recording, and readout with a portable unit, have been demonstrated as shown in Fig. 7.14. The results indicate no crosstalk between layers spaced as close as 30 ?m and excellent stability of the written bits at room temperature. The spot size is limited by the recording wavelengths through diffraction as well as by the sensitivity and integration time of the read-out detector used. The approach promises low cost, ultra high effective areal density (1-100 Tb/in2) removable disk media with thousands of layers for image and video applications. In addition to the extremely large volumetric densities achievable with this monolithic 3D disk approach, the stored data can be accessed in parallel leading to high data transfer rates suitable for content-based search operations.
Fig. 7.14. Multi-layer 3D ROM disk reader on the left; side view of a disk showing many recorded layers; and the oscilloscope trace related to detected bit stream with the disk rotating at 5,000 rpm (Call/Recall, Inc.).
However, several issues remain to be resolved. So far only small (1.5 inch diameter) disks have been demonstrated. Although insensitive to media shrinkage, the approach is affected by surface quality and volumetric homogeneity of the media. The cost of short pulse-high intensity lasers required for two-photon recording is still too high to consider their use in a consumer product. Finally, materials capable of writing and erasing without fatigue are yet to be demonstrated for addressing a larger span of applications.
The potential impact of layered 3D optical disks on the capacity per disk can be much greater than the impact of, for example, blue lasers. This is because the capacity is directly proportional to the number of layers. Assuming relaxed areal densities, 3D disks provide the potential for realizing disks with 100 GB to 1 TB capacities before 2005 as shown in Fig. 7.15. In addition to Call/Recall, Inc., two-photon recording is also being investigated at the Jet Propulsion Laboratory in the United States and at Osaka University in Japan (Toriumi, Hermann and Kawata 1997). At this point in time, two-photon recording appears to be a potential contender for future removable ultra-high capacity optical disk storage.
In bit-oriented memories if any portion of the storage media is damaged or blocked, the data stored in that region is lost. This is not the case for holographic storage, where the information about each stored bit is distributed throughout a large region. If a portion of the holographic storage media is damaged or blocked, instead of causing catastrophic loss of some of the data, all of the data are partially degraded. For common types of damage, such as surface dust or smudges, holograms are remarkably robust. This has generated interest in holographic data storage, and despite the more complex optical systems, high cost media, and sources with good spatial coherence required, there has been continued research in the field since the early 1960s (Solymar and Cooke 1981).
Holograms are created by recording the interference pattern of two optical wavefronts. The storage media can record the fringes as index and/or amplitude modulation. When the recording is illuminated by one of the wavefronts (the reference beam), the other wavefront (the object beam) is reproduced.
Fig. 7.15. Potential impact of 3D multi-layer optical storage and its comparison with conventional data storage and new emerging techniques such as SIL (compiled from data obtained from Terastor Web site, OITDA storage roadmap and Call/Recall, Inc. internal reports).
Since the diffraction efficiency of amplitude holograms is limited to 6-7%, and the efficiency of volume amplitude holograms is limited to 33%, volume phase holograms with their diffraction efficiency approaching 100% are best suited for storage applications. In addition, the Bragg selectivity of volume holograms allows the volume multiplexing (by angle or wavelength or combination) of many pages of data with small crosstalk between the pages.
Photorefractive crystals (PRC) and photo-polymers are typically used in demonstration experiments both for non-moving and volumetric disk type applications. These materials are recorded at room temperature, and the inorganic PRCs offer high photocyclicity. Using PRCs, 10,000 images were successfully stored and retrieved (Mok and Stoll 1992). More recently, hero experiments have demonstrated effective areal densities in excess of 64 Gb/in2.
The main drawback of the approach with PRCs is the high cost of materials, the decrease in diffraction efficiency with increasing number of multiplexed holograms and read-out cycles, and the relatively small and slow changes in the index of refraction. To circumvent these problems, significant progress has been made in terms of using recording schedules to improve the uniformity of the recorded data, fixing procedures to permanently store data in PRCs, and new recording geometries to minimize crosstalk. In addition, more recently, photorefractive polymers that exhibit large index of refraction changes under very large electric fields have been demonstrated.
Holographic storage as a read only archival memory can also use polymers such as those available from DuPont and Polaroid. In this case, the shrinkage of the polymer after recording and over time results in a shift in readout wavelength, as well as in increased crosstalk and loss of resolution. Very recently encouraging results on reduced shrinkage materials have been announced.
Holographic storage allows for massively parallel read-write operations. To take advantage of this capability, researchers are considering the potential benefits of using holographic storage as a means for fast access storage using non-moving media. In this case, holographic storage may compete with DRAM banks, solid state disks and probe storage to fill the "access time gap" that exists between the primary and secondary storage. Also for this type of application, in the United States a new start-up company, Templex Technology Inc., is investigating the potentials of using optical beam to record and read "persistent spectral hole burning" materials using the wavelength as the third dimension (see for example http://www.templex.com). Because of the very low temperatures required for this technology, interest in Japan in PSHB only exists within certain university research groups.
Most of the above mentioned progress in holographic storage has been achieved in the United States where the National Storage Industry Consortium carries out a research program with the participation of several U.S. corporations and universities to develop holographic storage media and systems. Key players in the United States include Holoplex, Inc., Optitek Corp., IBM, Rockwell International, Lucent, and Polaroid.
In Japan there is a lesser degree of interest in holographic storage although several university laboratories, NAIR and NTT are involved with it at the fundamental research level. Several Japanese managers remain skeptical, however, about the practicality of the approach.
With the strong demand for capacity for removable systems comes an equally strong demand for high data rates. This demand originates from the wish to quickly transfer large image and video files to direct memory and from the desire to perform fast content based search and image processing with this type of file. The conventional method to increase data rates in a disk system is to increase the rotation speed and to increase the linear bit density. Present rotation speeds are already high and are limited by media integrity and servo speeds. The linear density increases as the square root of the areal density and therefore falls short of satisfying the emerging needs of content-based database search.
Based on progress made in different areas of optoelectronics, newer technologies exploiting the parallel access capabilities of optical storage are emerging to satisfy these new requirements. In Japan, short term and long term approaches to parallel data access on optical disks are under investigation. For example, Fujitsu is investigating the use of laser and detector arrays to access data in parallel from an MO disk as shown in Fig. 7.16.
Fig. 7.16. Multi-beam optical head investigated at Fujitsu.
Issues including heat extraction from laser diode arrays, multi-beam positioning errors, packaging and signal processing circuit complexity for post processing are being investigated.
A more futuristic approach being explored at the University of Tokai by Prof. Goto and his group is shown in Fig. 7.17. This approach involves the use of VCSEL arrays flying directly above an optical disk medium using a near field geometry. The laser is used to both record and directly detect the presence of a bit based on the optical signal that is fed back into the laser. In order to achieve high data rates, massively parallel access through VCSEL array is contemplated as described in Fig. 7.18.
Fig. 7.17. Use of a VCSEL for recording and readout in a near field optics geometry.
In the United States, Quinta, Inc., recently acquired by Seagate, Inc., a manufacturer of magnetic hard drives, has been developing a method to access several optical disk surfaces in parallel using optical fibers. The fibers are used for distributing the laser power to the head and back to the receivers for readout, and a novel optoelectronic switch is used to selectively access data on various plates. Also in the United States, Call/Recall, Inc., jointly with Hewlett Packard Company and Optical Micro Machines, is investigating the use of VCSEL arrays and MEMS to facilitate parallel access to multi-layer disks. In addition, R&D on massively parallel accessing of holographic data storage systems is also underway as described earlier.
It is believed that parallel access is a viable method to increase the data transfer rate of optical and probe storage systems. The issue is rather when the necessary components that can enable massively parallel data transfer may provide a favorable cost entry point to market such systems.
Fig. 7.18. Schematic of a VCSEL array access with a massively parallel near field optics architecture to optical disks promising TB capacities and Tb/s data transfer.
One of the key areas in optical data storage for ROM type applications is the high fidelity replication process. In this area of R&D significant investments are being made by Japan's key CD-ROM and DVD-ROM manufacturers. The mastering effort ranges development of SIL lenses and large area electron beam lithography to the development of ultra short wavelength lasers and probe mastering techniques. Supporting technologies including high precision injection molding and ultra smooth polishing techniques are being developed as well. A potential roadmap derived from OITDA's roadmap for mastering and replication is shown in Fig. 7.19. At this point in time, we do not know of a development effort of this magnitude in this area in the United States.
Fig. 7.19. Roadmap for mastering and replication (compiled based on data available from OITDA, discussions during Sony site visit and various private discussions).