Hard disk-drive technology revolutionizes processing
09/01/2004
This article reviews trends in magnetic hard disk drive (HDD) technology and discusses underlying magnetoresistive phenomena, magnetic media materials issues, and processes for fabricating leading-edge write and read heads for state-of-the-art HDDs.
Over the last decade, information technology's appetite for data storage capacity and data access speed has created a technology explosion in magnetic storage drives. Hard disk-drive storage capacity, as represented by areal information density, has grown exponentially. Although areal density growth has slowed over the past three years — from the 100%/year rate that characterized the last half of the 1990s, to an annualized rate of 30% — acceleration could resume as researchers begin zeroing in on the terabit/in.2 (1000Gb/in.2) level. What lies beyond that? Nature itself provides both the answer and a target: The information density of DNA is estimated at 250Tb/in.2
A number of technological, manufacturing, and business challenges affect the HDD industry. The main goals are to squeeze more data bits into available storage space, shrink form factors to achieve total volume reductions, and increase the speed of access to that data.
Magnetoresistive technologies
An HDD is a complex, interactive, electromechanical system with a range of nano-, micro-, and mesoscale components. Hard disk storage media consist of rigid substrates and multilayer coatings that include magnetic films, conductors, insulating films, and lubricant layers. In an HDD, the disk spins under an assembly that holds both write heads to place digital information on the magnetic medium by changing the local magnetization of small physical regions. Read heads detect the magnetic state of the written regions and convert the stored data into electrical signals that can be digitally processed. Disks are formatted with thousands of circular tracks that contain a sequential array of magnetized regions (the data bits). Positioning of the write and read heads over the disk is controlled by a high-resolution servo system.
Data bits are written inductively, as shown in Fig. 1a for longitudinal recording (in the plane of the medium) and Fig 1b. for perpendicular recording (orthogonal to the plane of the medium). Perpendicular recording enables smaller bit size and increased storage density.
Magnetoresistive (MR) thin-film read heads were introduced in 1990. In all MR types, the sensor element exhibits a relative change in resistance, ΔR/R, when in a magnetic field — namely that of the magnetic data bit in the storage medium. ΔR/R can be monitored as suggested in Fig. 1c. Small bit sizes in high areal-density media imply small fields and require very close proximity of the read sensor to the stored bit on the disk.
When giant magnetoresistive (GMR) effect sensors became practical in the 1990s, they changed the commercial landscape for HDDs. GMR/spin valve (SV) heads are multilayer structures that employ adjacent magnetic layers separated by a nonmagnetic layer with a thickness less than the mean free path for electron scattering. Magnetic fields in the adjacent layers can be aligned in a parallel or antiparallel fashion that couples or decouples their electron spin orientations (hence the alternative term, spin valve) to produce easy-current or hard-current flow conditions, respectively. In various physical and material configurations, GMR ΔR/R values have gradually climbed to 40% under ideal laboratory conditions. In manufacturing, however, values of 15–20% are more common. Compared with other magnetic sensors, GMR sensors generate useful response from bits of much smaller physical size, thus enabling a significant increase in areal storage density.
One relatively recent GMR structure uses a thin Al2O3 layer to separate the magnetic films. It employs spin-dependent tunneling as the conduction mechanism, has demonstrated ΔR/R values up to 50% under ideal conditions, and is of great interest for advancing storage densities in the current perpendicular-to-plane (CPP) configuration to next-generation levels of 140 to 200Gb/in.2 [1].
An even more recent MR effect is ballistic magnetoresistance (BMR). In the BMR regime, sensors are reduced to the size of a cluster of atoms joined by lead wires. The term "ballistic" implies that the sensor dimension is smaller than the electron scattering length and that carriers therefore move in straight-line trajectories, scattered solely by the magnetic effect that makes the readout process highly sensitive. BMR ΔR/R values of >3000% at room temperature have been reported compared to 100% for the best GMR, 1300% for extraordinary MR in semiconductor/metal composites, and 1300% for room-temperature colossal magnetoresistance in oxides films such as LaMnO3 [2]. BMR developments could speed up the arrival of 1Tb/in.2 magnetic storage devices.
 Figure 2. Multilayer CPP head design with in-stack biasing using magnetostatic interaction with an additional pinned layer. |
As bit size is reduced to the nanotechnology domain (CD ≤100nm), random thermal fluctuations can change a bit's magnetic orientation. Called the superparamagnetic effect, this represents a serious problem for long-term storage. Perpendicular recording affords advantages in delaying the onset of this phenomenon, but it is more likely that the front line of defense against superparamagnetism will be the use of higher-coercivity films requiring greater energy to upset magnetic alignment at the microcrystalline level.
Pursuit of longer-range magnetic storage goals will require complex film stacks with 12 or more layers (Fig. 2), near-monolayer deposition, atomic-level interface control, possibly heat-assisted magnetoresistive recording, and media with high thermal stability. For heads, technology features are likely to include nanoscale dimensions, a one- to two-order increase in sensitivity, high signal-to-noise ratios, and possibly laser-assisted media heating capability.
Thin oxides will be required in tunneling and multilayer, stacked structures. Their formation may demand special oxidation/segregation techniques. Increasing emphasis will also likely be placed on material systems that achieve higher coherent spin-scattering. For CPP structures, some stack "defects" (e.g., pinholes, mixed phases, etc.) may be desirable [3]. This suggests the possible use of various thin-film material deposition and post-processing techniques including advanced physical vapor deposition, ion-beam deposition (IBD), atomic-layer deposition (ALD), pulsed-laser deposition, laser surface treatment, and in situ thermal treatment.
Other materials issues include those related to slider structures, the assemblies that attach to the HDD gimbal arm and carry the write and read heads (Fig. 3). But several challenges must be met. As flying height drops to 7nm (about one-quarter of a millionth of an inch), the magnetic-spacing to flying-height difference must decrease without compromising the wear resistance of the slider overcoat. This situation means that reduction or elimination of the seed-layer thickness for the slider overcoat will be desirable. Two approaches that will be explored to achieve this include filtered cathodic arc (FCA) deposition with low energy and carbon alloy capability, and new chemistries with seed-free ALD to achieve thin, highly conformal overcoats.
 Figure 4. Write-head evolution in advance of manufacturing. |
Magnetic write and read heads are complex and difficult to manufacture microstructures. The evolution of writer technology is illustrated in Fig. 4. Combined write-read head designs are gaining favor.
Critical dimension (CD) lithography plays a central role in the processing of write and read heads of all types. Because of the various deposited structures that require lithographic definition, clean lift-off lithography is particularly important. Figs. 5 and 6 illustrate some of the lift-off lithography challenges. Collimated IBD is preferred for fabricating structures like those in the figures [4]. Nevertheless, with shrinking dimensions, Fig. 5 shows an IBD condition that promotes greater film thickness on the inboard side (facing the beam) of a substrate with incursion of active material under the reentrant portion of the lift-off mask. On the
 Figure 6. Near the sensor edges the permanent magnet field dominates, reducing sensitivity. |
outboard side (furthest from the ion source), thinner and offset deposition occurs due to shielding by the mask. Such deposition asymmetry adversely affects sensor performance and reduces yield. In the situation of Fig. 5, advances in IBD equipment with high degrees of collimation have reduced inboard/.outboard asymmetry across a substrate from levels of 15–20% with earlier technology to nonuniformities of <4% (3σ) across a 6-in. substrate for chrome-.cobalt-platinum layers, and <6% (3σ) for Cr and Ta layers [5].
A developing challenge in HDD lithography is that with structures now entering the nanodomain, the evolution of HDD head CDs is exceeding the pace of the Semi lithography roadmap, with high aspect ratios and difficult topography.
Writer pole process tolerance
By 2006, CDs for writer pole structures in magnetic recording heads will be targeted for 100nm and below. As head structure dimensions further approach nanotechnology levels, writer poles will increasingly require slimming/shaping through methods like ion-beam etching (IBE).
From a manufacturing point of view, the looming challenge is to achieve uniform device shapes in recording head structures. Technical success will depend on procedural and equipment-related parameters of the IBE process. Economic success will demand minimum cost-of-ownership for complex process equipment.
A typical total manufacturing tolerance budget for recording head manufacture is 10% (3σ). The tolerance budget for CD control for 250nm writer poles thus will be only 25nm. But absolute etching tolerance for writer pole shape and width distribution is even smaller. If an etch process must remove a total of 100nm of material while slimming a writer pole to its final dimension, then due to the summation of variances, the 3σ target for etch control must typically be 3nm.
Gains in areal bit density on a hard disk require both an increase in linear bit density along the recording tracks and an increase in the number of radial tracks/inch. At an areal density of 120Gb/in.2, representative values will be about 120kbpi (kilobits per inch) linearly along a track, and about 220ktpi (kilotracks per inch) radially. This reduces track width and separation to about 160nm.
Shrinking track separation generates another challenge: magnetic fringing fields around the writer pole increase the chances of erroneously writing data on adjacent tracks. To mitigate this situation, trapezoidally shaped writer poles (Fig. 7) are increasingly employed and IBE has been adopted as the preferred method of pole shaping. In the face of reduced CDs, the need for uniformity in pole size and shape further extends manufacturing process stringency.
 Figure 7. Process sequence for perpendicular writer pole formation. The trapezoidal shape has a slant of 10–20°. View is normal to the plane of the media. |
IBE systems are now available for writer pole fabrication that reduce nonuniformity to <2% (3σ), a level consistent with commercially acceptable manufacturing yield for current- and proximate-generation perpendicular recording-head structures aimed at areal-density levels of 120Gb/in.2 and higher. These data suggest that processes and equipment will be extendible to next-generation products in the drive toward the 1Tb/in.2 target.
As the recording-head manufacturing industry shifts to perpendicular recording and CPP readers, longer product life cycles will likely be encountered. This is a scenario that promotes equipment-focused manufacturing cost-of-ownership issues in the economic balance with pressures to shorten product time-to-market.
With CD reduction broaching the nanoscale, edge atom displacement in device structures must be considered. In narrow track sensors and writers, the ratio of "dead zone" to device size increases. For the 120Gb/in.2 regime, a film structure can have an edge atom content (sidewall dead zones) of 2% even without considering the effect of additional process-related displacements. Low-energy IBE processes that can produce near-vertical sidewalls will become increasingly important. Simulation studies in this area have established best conditions for low-damage etching at 200eV. Researchers are exploring low-pressure reactive ion-beam processes to meet these challenges.
Summary
Technological gains that in principle could enable greater storage densities will come through innovative advances in materials, and through discovery and use of new physical phenomena to produce state changes representing ones and zeros. While laboratory demonstrations are one thing, reduction to commercial practice is quite another.
To enter the technology base, devices and systems first must be manufacturable and then manufactured. Manufacturability challenges device makers and equipment manufacturers to engineer technically and economically practical solutions to materials and process problems. The nascent plunge into nanotechnology is an example.
Manufacturability is a necessary condition but alone is insufficient for business decision making. In the face of rapid technology transitions and almost transient product life cycles, the difficult game of business timing for engineering development, market entry, and product lifetime becomes critical to achieving return on investment, establishing market share, reaching acceptable profitability, and ultimately ensuring the survival of high-tech companies.
References
- S. Mao, "TMR Heads for the Next Generation," Veeco Symposium, Diskcon, Sept. 9, 2003.
- H.D. Chopra, S.Z. Hua, "Ballistic Magnetoresistance Over 3000% in Ni Noncontacts at Room Temperature," Phys. Rev. B 66, June 26, 2002; and http://www.phys.uni.torun.pl/~jkob/physnews/node144.html.
- H. Hegde, "Transition from CIP to CPP, Longitudinal to Perpendicular: Technology and Manufacturing Challenges," Veeco Symposium, Diskcon, Sept. 9, 2003.
- M. Kautzky, J. VanEk, J. Dolejsi, P. Ryan, "The Application of Collimated Sputtering to Abutted Junction Reader Processing," Semiconductor Fabtech, Edition 3.
- B. Druz, et al., "Nexus IBD D/S Hard Bias and Isolation Layer Applications," Veeco Symposium, Diskcon, Sept. 9, 2003.
Kurt E. Williams is director of process technology at Veeco, 1 Terminal Dr., Plainview, NY 11803; ph 516/349-8300 ext. 1474, fax 516/349-7009, e-mail [email protected].