Patterned magnetic media: impact of nanoscale patterning on hard disk drives
09/01/2006
As the superparamagnetic limit continues to make scaling of conventional magnetic recording technology to higher densities and larger capacities more difficult, novel concepts such as patterned magnetic media are being considered as likely routes to terabit per square inch densities and beyond. Patterned media overcomes the thermal stability problems of conventional magnetic media by implementing single domain magnetic islands for each bit of recorded information. Fabrication of patterned media requires the application of emerging technologies, such as nanoimprint lithography for patterned media disks, and advanced electron beam lithography for master mold fabrication. Practical realization of these new technologies offers challenges in disk fabrication as well as the design of the overall recording system.
Zvonimir Z. Bandić, Elizabeth A. Dobisz, Tsai-wei Wu, Thomas R. Albrecht, Hitachi San Jose Research Center, San Jose, California
Bit areal density is considered to be a key benchmark for measuring progress in magnetic data storage. The discovery of the giant magnetoresistance (GMR) effect, and its implementation in magnetic read sensors, set off nearly a decade of rapid growth; areal density grew at a rate approaching 100% per annum. However, this trend has slowed to 40% in the last few years (Fig. 1), due to difficulties in recording head fabrication and the rapid approach to the magnetic media superparamagnetic limit, one of the key problems facing the hard-disk drive (HDD) industry.
null
Superparamagnetic limit
The nature of the superparamagnetic limit for continuous magnetic media is illustrated in Fig. 2. Conventional magnetic media consist of tiny sub-10nm magnetic grains, each one free to assume its own magnetization state (Fig. 2b). In the case of well-oriented longitudinal media, the magnetization may be “left” or “right,” and in the case of perpendicular media, this orientation may be “up” or “down.” A sufficiently strong external magnetic field can switch the magnetization of individual grains. The magnetic recording head, consisting of an inductive write element and a magnetoresistive read element, and mounted on a hydrodynamic air bearing slider, flies along the data track at a height of around 10nm. The magnetic write head stores data, in the form of 1’s and 0’s, by moving along the data track, and magnetizing media into two possible magnetization directions (shown in red and blue in Fig. 2b). Data is read back by moving the read sensor along the data track, and detecting the varying magnetic flux emanating from the magnetized patterns created by the write head. The signal to noise ratio of the media is roughly proportional to the number of magnetic grains per bit, or SNR ∝ Wbt/Vg, where Wbt is the bit volume (bit width × bit length × media thickness) and Vg is the grain volume [1].
The importance of the grain size can be observed in Fig. 2b, where the roughness of the transition between regions of opposite magnetization improves with reduced grain size. It becomes clear that increasing bit areal density without sacrificing SNR requires reducing the average grain volume Vg in the media. However, the grain volume cannot be arbitrarily reduced-the superparamagnetic limit is reached at the point when grain becomes so small that thermal energy alone can flip its magnetization direction. The critical grain volume Vg that determines the onset of superparamagnetic limit is determined by the condition that the stored magnetic energy KuVg remains about 40-60× larger than the thermal energy kBT, where Ku and kB are the magnetic anisotropy and Boltzmann constant, and T is the temperature [2]. This implies that the size of the thermally stable grains should be larger than approximately 8nm [3]. The current consensus in the HDD industry is that magnetic recording on continuous perpendicular media can be scaled to bit areal density in the range between 500-1000Gb/in2. Beyond that point, alternative technologies such as thermally assisted recording (TAR) and patterned media (PM) are being considered as likely routes to terabit per square inch density and beyond.
Patterned media
Abandoning the requirement for multiple grains for each bit may solve the thermal stability problem. Instead, one can use a single domain predefined magnetic island per bit, and therefore allow for larger volume magnetic entity than the individual random grains used in continuous perpendicular magnetic media. Figure 3 illustrates the PM concept and compares it with continuous media. Creating a functioning data recording system based on PM presents many challenges, such as cost-effective high-volume fabrication of PM disks and accurate individual addressing of each one of the 1012 prepatterned magnetic islands per surface. Typical bit sizes and periodicities as shown in the table clearly illustrate the degree of difficulty required for the nanofabrication of PM. For example, the size and periodicity of the islands required for 500 Gb/in2 bit density (likely the entry point of PM technology into magnetic recording media) is 25nm and 35nm, respectively, and may be required in the 2010-2011 time frame (Fig. 1). The only conventional lithographic technology capable of high enough throughput for fabrication of tens of millions of PM disks is optical lithography. However, the pattern densities shown in the table are generally beyond the reach of photon optics [4].
null
 |
null
Nanoimprint lithography
One emerging technology showing significant promise in achieving both the high required throughput and resolution at reasonable cost is UV-cure nanoimprint lithography [5, 6], which replicates topographic patterns from a master mold into a polymeric resist coating on a disk substrate using a molding-like process. It is in some ways similar to the digital video disk (DVD) manufacturing process [7], except that a UV-cured liquid resist is used instead of relying on thermal softening of a solid polymer.
The proposed process flow for PM disk fabrication using nanoimprinting is shown in Fig. 4. High-resolution e-beam lithography is required to fabricate a high-resolution master mold (Fig. 4a). Since the patterns required for magnetic media have circular symmetry, conventional Cartesian (X-Y) electron beam tools developed for maskmaking in the IC industry cannot be used effectively. Instead, an e-beam column coupled with a high quality mechanical rotary stage is required to meet all pattern specifications. Following the exposure of the e-beam resist coated master (Fig. 4a), the resist is developed, and reactive ion etching (RIE) is used to transfer the pattern into the master mold substrate (Fig. 4b). This topographic master mold is then pressed against a disk substrate coated with liquid nanoimprint resist, which reflows until the resist topography conforms to the master topography (Fig. 4c), at which point ultraviolet (UV) light is applied to cure the resist in place and create a solid nanoimprinted replica. Following UV curing, the resist pattern is transferred into the disk substrate using RIE (Fig. 4d). The imprinted resist topography on the replica is complementary to the stamper topology (i.e. pillars imprint holes, and vice versa). In the final step (Fig. 4e), magnetic media is blanket sputter-deposited over the patterned disk substrate.
A scanning electron microscopy (SEM) image of a typical master pattern used for PM fabrication, and fabricated using a Leica VB6 100KV electron beam lithography tool, is shown in Fig. 5. The density of the pillars corresponds approximately to 300Gb/in2, and the spacing between pillars is 50nm. This type of master, having the same pillar topology as the final pattern on the PM disk, is preferred when additional generation of daughter molds is introduced between master mold and many disk substrate replicas, similar to DVD fabrication strategy.
Nanoimprint lithography, an emerging technology in its early phase, presents many challenges, including selection of the resist photopolymer, resist flow control during imprinting, fidelity of nanoscale features, adhesion of nanoimprinted resist to the substrate, release of the imprinted substrate from the mold without damage to either, control of the thickness of the residual layer, resist curing, the lifetime of the master mold (and/or daughter molds), the etch resistance of the imprinting resist, scale-up to large areas, and last, but not least, double-sided imprinting (since magnetic media disks typically utilize both sides for data recording).
Figure 6a shows a cross-section of nanoimprinted resist on a disk substrate. The residual layer thickness (Fig. 6a) has to be uniform across the substrate and thin compared to the resist thickness, in order to preserve feature fidelity during etching. Molds have to be coated with resist release layers (Fig. 6a), which allow the imprinting resist to wet and conform to the master during the resist reflow, but must have low enough surface energy to reduce the release force between stamper and replica after resist curing. Final selection of the nanoimprinting resist polymer is typically obtained as a compromise between the requirements listed above, the most important being feature fidelity, etch resistance and residual layer thickness. An example of a successful imprint of 50nm period pillar features is shown in Fig. 6b. Unlike in IC applications, nanoimprinting for PM fabrication does not require overlay or tight alignment-only one “mask” is required, and only coarse mechanical alignment with the disk substrate is needed. Since nanoimprinting PM is relatively easy in these respects, it is possible that fabrication of PM disks may be one of the first very large-volume uses of the nanoimprinting lithography.
Following RIE pattern transfer, the disk substrate is stripped, cleaned, and blanket sputter deposited with magnetic media, typically Co/Pt or Co Pd multilayers [8]. A cross-sectional SEM image (Fig. 7a) of a patterned disk substrate sputter-deposited with magnetic media demonstrates that the amount of material deposited on the sidewalls may be controlled. The topographic AFM image (Fig. 7b) and magnetic force microscopy (MFM) image (Fig. 7c) of a patterned media disk at moderate bit areal density of 100Gb/in2 demonstrate that individual islands are successfully magnetically isolated, with two magnetic states clearly visible.
Metrology of PM disks is an additional challenge in PM fabrication. As can be seen in this article, low-throughput scanning probe techniques such as SEM, AFM, and MFM are well suited for detailed analysis of PM structures. For manufacturing process control, however, higher-throughput methods such as advanced optical scatterometry will need to be adopted. Complete PM disks will also likely be tested bit-by-bit for functionality using magnetic write
ead heads on a spin-stand tester by auditing, for example, 1 out of every 100 disks.
Challenges in PM integration
Integration of PM into a HDD recording system (Fig. 8) creates multiple challenges in the head-disk interface (due to the nonplanar surface of patterned media), the track-following servo mechanism (required to keep a magnetic head on the track during write
ead process), and signal processing (since a magnetic write head has to switch the field exactly above the island), just to name a few. Disks will likely have to be planarized following sputter deposition of magnetic media. A low-cost planarization process, much cheaper than conventional CMP, will be required. The track-following servo mechanism must be able to stay precisely in the center of the 35nm wide tracks. This will require implementation of prepatterned servo features that will be created directly in the nanoimprinting master mold. Additionally, a dual-stage actuator arm with microactuator may also be required [9].
 Figure 8. Challenges in integration of PM into hard disk drives. |
null
Conclusion
As the superparamagnetic limit continues to make scaling of conventional magnetic recording technology to higher densities more difficult, new alternative technologies such as PM are being considered as likely routes to 500 Gb/in2 densities and beyond. PM circumvents the thermal stability problem by moving away from the requirement of multiple magnetic grains per bit, and instead implementing larger and stable single domain predefined magnetic islands. Nanoimprint lithography has emerged as a likely candidatefor high-volume fabrication of PM disks, and electron beam lithography as the method required to produce nanoimprinting master stampers. However, many challenges in implementation of nanoimprinting technology into PM disk fabrication process still abound, such as choice of nanoimprinting resist, pattern transfer fidelity and uniformity, and lifetime of master stampers. System-level integration issues also create additional challenges in track-following, head-disk interface, and signal processing.
Acknowledgments
The authors thank many colleagues from Hitachi San Jose Research Center who contributed to this project, including Maggie Best, Dan Kercher, Henry Yang, Michael Xu, Bruce Terris, and Tom Thomson.
References
- R. Wood, “The Feasibility of Magnetic Recording at 1 Terabit per Square Inch,” IEEE Trans. Magn., 36, 36, 2000.
- D. Weller and A. Moser, “Thermal Effect Limits in Ultrahigh-Density Magnetic Recording,” IEEE Trans. Magn., 35, 4423, 1999.
- D. Weller et al, “High K-u Materials Approach to 100 Gbits/in(2),” IEEE Trans. Magn., 36, 10, 2000.
- M. Rothschild et al, “Nanopatterning with UV Optical Lithography,” MRS Bulletin, 30, 942, 2005.
- Stephen Y. Chou, “Nanoimprint Lithography and Lithographically Induced Self-assembly,” MRS Bulletin, 26, 512, 2001.
- M.D. Stewart, C.G. Willson, “Imprint materials for nanoscale devices,” MRS Bulletin, 30, 947, 2005.
- http://www.optical-disc.com/techmat_techpaper.htm.
- B.D. Terris, T. Thomson, “Nanofabricated and Self-assembled Magnetic Structures as Data Storage Media,” J. Phys., D38, R199, 2005.
- T. Hirano, L.S. Fan, J.Q. Gao, W.Y. Lee, “MEMS Milliactuator For Hard-Disk-Drive Tracking Servo,” Journal of Microelectromechanical Systems, 7, 149, 1998.
Zvonimir Z. Bandić is a research staff member at the Hitachi San Jose Research Center, 650 Harry Rd., San Jose, CA 95120; ph 408/323-7206, [email protected].
Elizabeth A. Dobisz is a research staff member at the Hitachi San Jose Research Center.
Tsai-wei Wu joined the Storage System Technology Division at IBM Almaden Research Center in San Jose as a research staff member.
Thomas R. Albrecht manages patterned media research activities at Hitachi San Jose Research Center.