Site: Fujitsu Ltd.
Fujitsu Laboratories, Ltd.
10-1 Morinosato-wakamiya, Atsugi
Kanagawa 243-01, Japan
Date Visited: 11 March 1998
Fujitsu Limited is an electronics and computer powerhouse, with over $35 billion in annual sales. Two thirds of that is in information processing. Fujitsu's storage device sales total about $2.2 billion annually, including hard disk drives (HDDs), magneto-optical drives and libraries, and tape drives. Fujitsu's HDD market share has been increasing. It could exceed 10% this year, with a production rate of about 1.5 million drives per month.
Fujitsu Laboratories has about 1,500 employees. Within that organization, the Peripheral Systems Lab contains the Magnetic Disks Lab at Atsugi, the Optical Disk Media Lab at Akashi, and the M Project Group (MO drive) at Akashi. In the Storage Products Group, the File Memory Lab has about 60 people. Many of these transferred from Fujitsu Laboratories a few years ago in order to accelerate the conversion to magnetoresistive (MR) and spin valve heads and disks. In addition, the research group is in the process of restaffing its hard disk projects.
In the current struggle between MO and phase change optical technologies, Fujitsu is a champion for magneto-optics.
After an introduction by Dr. Ogawa, and a storage business plan presentation by Dr. Miura, the panel heard Mr. Uematsu present Fujitsu's HDD technology roadmap. In response to the panel's interest in the application of perpendicular recording to actual products, the members saw previously unpublished photographs of a 1.8 inch prototype drive intended for a PCMCIA slot in a laptop computer. This work was performed circa 1992 at 200 Mb/in2, and used a Censtor-style in-contact inductive head. It had a steel case for magnetic shielding, and had full LSI inside the enclosure. The WTEC panel had numerous questions about outgassing, magnetic shielding, popcorn noise in the head and soft underlayer, frictional effects on the tracking servo, and so forth. The project was abandoned for business reasons (the 1.8 inch market never materialized), and because the technical problems were not entirely solved.
Fujitsu's technology roadmap for longitudinal recording projected the following growth curve:
The bit cell aspect ratio during this period evolves from about 20x today to about 10x by 2002. A switch to perpendicular recording is considered unlikely before the end of this period. Along the way, the head will evolve from a conventional spin valve, which the company was scheduled to ship in 1998, to an adjacent soft layer spin valve called BCL-type in the year 2001, and possibly a spin tunneling head by 2003. Fujitsu managers expect CoCrPtTa alloys to suffice for the media through 2001, then expect to use a CoPt-SiO2 granular composite longitudinal medium with a soft keeper.
The Fujitsu hosts expect the media grain size to go from 13 nm at the time of this visit at 2 Gb/in2 to 11 nm at 40 Gb/in2. This deviation from simple scaling is driven by the superparamagnetic effect, and is accompanied by a major decrease in the grain size standard deviation. Even with this improvement, however, the media signal-to-noise ratio at 40 Gb/in2 would be expected to be much below today's levels, and to be a challenge to the channel and ECC people. The panel did not hear about coding or ECC, but Fujitsu's roadmap shows decision feedback equalization by the year 2000, and a new technique being sought for the year 2004.
The trend towards lower bit cell ratio will accelerate the need for precision and bandwidth in the actuators. Fujitsu anticipates an evolution that includes piggyback actuators by the year 2000. The company's present design, a magnetic actuator in the middle of the suspension, achieves a 3.3 kHz bandwidth, 0.27 ms settling time, and a 500 G shock resistance. Fujitsu managers estimate a $2 to $3 additional cost per head for this design, but feel that a $1 cost is necessary for future success. For this reason, they expect piezoelectric or MEMS designs to ultimately prevail. They discussed a conceptual design from UC Berkeley of a rotary electrostatic MEMS actuator mounted at the slider/suspension interface. This design would possess a half micron stroke for 50 volt excitation, but Fujitsu's design goal is one micron stroke and a reduced voltage. And, of course, the device must have enhanced shock resistance and a $1 cost.
Another aspect of high track densities that Fujitsu researchers are studying is the problem of servo pattern accuracy for sector servo drives. Rather than dealing with timing jitter and erase band size, they have chosen to pursue a photolithographic patterned media technique that they call ESPAR. This involves pre-etching the (silicon) substrate before depositing a 75 nm CoCrPtTa magnetic servo layer. This is then lapped flat, and the data layer of 17 nm is deposited.
WTEC panelists were excited by this display of high track density technology, and asked if this might accelerate the trend towards a ratio of linear to track density of less than 10:1. The answer was that confidence in the data rate capability, as well as an extrapolated linear density of 630 kb/in at 40 Gb/in2 at a magnetic spacing of 20 nm (to mid-plane of the storage layer), and an expected shield-to-shield spacing in spin valve sensors of only 90 nm, all lead Fujitsu managers to believe that advances in linear density will keep pace with advances in track density, at least until 40 Gb/in2 is reached.
Fujitsu researchers do not favor load/unload technology, and therefore need an anti-stiction technique for very flat sliders and disks. This is accomplished by a patterned slider having four very small (0.09 mm2 area) pads and by using only 2 nm of lubricant. On a very smooth disk, this results in an order of magnitude decrease in stiction force.
Dr. Kobayashi gave a fascinating presentation on his spin tunneling sensors. These could replace spin valves by as early as 2003. His junctions use permalloy magnetic layers, faced with 3.3 nm of Co, a dielectric gap of about 1.3 nm of Al-Al2O3, and the necessary bias and pinning layers. He achieves up to 24% MR ratio with either natural oxidation or plasma oxidation in oxygen of the aluminum, but the required time drops from 500 hours to 60 seconds. There is also an unusual behavior after annealing of his tunnel junction samples, which contain an NiMn pinning layer. The MR ratio of the junction increases with annealing temperatures up to 300°C, where the 24% value is observed. Other issues being pursued with these junctions are the high impedance for very small devices, and the voltage dependence of the MR ratio. Future work will examine half-metallic ferromagnets to increase spin polarization, and double barrier resonant tunneling.
Dr. Oshiki showed the panel an interesting experimental project on spin polarized scanning tunneling microscopy. The basic physics involves polarized photo emission of electrons from single crystal cleaved GaAs tips, using circularly polarized light to stimulate the emission. Previous work has required a transparent subject, since transmitted light was used to stimulate electrons polarized perpendicularly to the surface of the sample. This new technique uses illumination from the side, and can therefore be used with metallic samples. Preliminary data show about a 10% contrast for domains in an NiFe sample. In addition to microscopy, this technique could be used for data readout--if were not for the fact that it requires UHV conditions and a half-day bake-out to get a clean enough tip for operation.
The final presentation of the HDD portion of the program was a talk by Dr. Oshiki on the merits of grazing angle X-ray reflectivity apparatus for analysis of spin valve layered structures. Compared to ordinary X-ray fluorescence, the technique is more sensitive and more accurate. Fujitsu is working with an equipment manufacturer to make such a tool commercially, for the purpose of calibrating production XRF equipment on a more frequent and convenient basis. Comments at lunch included the general feeling at Fujitsu that holographic memory was not practical with present media (the company abandoned it about five years ago); that the probe storage group favored a multiplexed tip approach (Fujitsu already had four tips working simultaneously); and that multilayer optical media were a better approach to volumetric storage than either holography or ultra-high density surface recording methods.
The WTEC team toured the periphery of several clean rooms containing systems for advanced materials research in both media and head materials. Fujitsu has produced experimental media ranging from conventional alloys to ultra-high energy Sm2Co17 and Sm2Fe12 that no conventional head could write. Experimental head materials for advanced spin valves and tunnel junctions are also produced in those laboratories. The team saw an extensive characterization and analysis area, including glancing angle X-ray equipment and a marvelous transmission electron microscope (TEM) adapted for sensitive magnetic imaging.
Fujitsu is actively pursuing high-density HDD technology, magneto-optical storage, and probe storage. The company is well connected with ASET and SRC, and has leading-edge efforts in all three areas.