Eric E. Fullerton, David T. Margulies, Andreas Moser, Kentaro Takano IBM Research Division, Almaden Research Center, San Jose, California
In current magnetic recording media the signal-to-noise ratio needed for high-density recording is achieved by statistically averaging a large number of weakly interacting magnetic grains per recorded bit. Traditional engineering of magnetic media to achieve higher recording densities involves reducing the grain diameter and film thickness, but this approach is increasingly limited by instabilities due to thermal fluctuations. This phenomenon is thought to limit potential recording densities and is commonly referred to as the "super-paramagnetic limit." This article discusses this problem in the context of traditional recording media and new thin-film media and structures that push back the super-paramagnetic limit.
Areal densities of magnetic recording hard drives the number of recorded bits/unit area on the disk surface are currently growing at a rate >100%/year in both laboratory demonstrations and hard disk drive products , with current ones having recording densities of ~25 Gbits/in2. Advances in the growth and understanding of magnetic thin films are proving to be a key ingredient in this explosive growth. The ability to artificially structure and layer materials on the atomic scale has allowed new phenomena to be engineered into thin films devices. Such quantum engineering of magnetic films has been exploited in advanced recording heads  and holds promise for continued advances in the storage medium.
Figure 1. A longitudinal recording system with a read and write head and the recording medium. t = medium thickness, W = width of the recorded track, B = bit size, and d = height the head flies above the medium. Inset is a magnification of a transition region showing the interplay of the medium magnetic properties with the microstructure where D is media grain size and the transition parameter approximates width of the magnetization reversal region.
Longitudinal recording system
Longitudinal recording systems consist of a recording head composed of separate write and read elements, and the recording medium (Fig. 1). The write head records the data into the recording medium by applying a series of field pulses of opposite polarity resulting in an in-plane magnetization pattern. This geometry is referred to as longitudinal recording. Traditionally the recording media consists of a single magnetic layer and the information is read back via the read head sensor by measuring the dipolar fields arising from the pattern .
In current longitudinal magnetic recording thin-film media the signal-to-noise ratio (SNR) needed for high-density recording is achieved by statistically averaging over a large number of weakly interacting magnetic grains per bit. This is shown in the inset in Fig. 1, where the transitions between two magnetic bits follow the contours of the grains. The weakly interacting granular system allows information to be written on a finer scale than possible in a homogeneous magnetic film.
Increasing areal densities have mainly been achieved by scaling the head and media parameters in Fig. 1 to smaller dimensions. Scaling the media parameters involves both a reduction in grain diameter, D, as well as a reduction in the media's effective magnetic thickness, Mrt, where Mr is the remanent magnetization of the media and t is the physical thickness. Historically, the reduction in magnetic thickness has primarily been achieved by reducing t. Scaling of D is needed in order to keep the number of grains per bit roughly constant and maintain the media SNR. Scaling of Mrt is also needed in order to scale the transition parameter (a, in Fig. 1). The transition parameter is the effective width of the boundary between bits and limits how closely the bits can be recorded.
Shown in Fig. 2 are transmission electron microscope (TEM) images for two different disk media illustrating how the grain structure has changed over time. The TEM image on the left is a magnetic media that supports a data density of about 10 Gbits/in.2 with an average grain diameter of about 13nm. The more advanced magnetic media on the right was used in a 35 Gbits/in2 recording demonstration by IBM . The average grain diameter is about 8.5nm.
Such state-of-the-art longitudinal recording media are sputtered magnetic thin films, typically CoPtCrB magnetic alloys . The choice of these materials is based on the growth-induced chemical segregation of the alloy films into magnetic and nonmagnetic phases. The nonmagnetic segregated material (light areas in the TEM image) is key to the performance of the recording media because it separates the magnetic grains and is the basis for obtaining a weakly interacting granular system. The magnetic interactions between grains need to be weak enough so that the grains remain statistically independent and magnetic transitions can more readily follow the grain contours.
Figure 2. Transmission electron microscope (TEM) images for two different disk media illustrating how the grain structure has changed over time. a) This TEM is a magnetic media that supports a data density of about 10 Gbits/in2. b) This magnetic media was used for IBM's 35 Gbits/in2 demonstration .
The compositional distribution of magnetic grains depends on the microstructure and growth conditions of the films. For granular CoPtCr-alloy films, the Cr diffuses towards the grain boundaries, resulting in a magnetic Co-rich core of the grain with nonmagnetic or weakly magnetic Cr-rich grain boundaries. The addition of boron to the recording alloy has proven to be a particularly important addition to state-of-the-art recording media because it reduces the grain size and helps to further magnetically isolate the grains by forming an amorphous nonmagnetic phase separating the grains. The grains and the amorphous grain boundaries are clearly visible in Fig. 2b.
Although significant progress has been made toward advancing thin-film recording media, thermal instabilities provide significant impediments to the continued growth of areal densities using traditional scaling. This occurs when the volume of an individual grain, V = pD2t/4, decreases to the point where the magnetic energy per grain, KUV, becomes comparable with thermal energies kBT. In this equation, KU =the magnetic anisotropy constant of the grain, kB = Boltzman's constant, and T = temperature. The minimum energy believed to maintain magnetic stability for >5 years in drive conditions is KUV > 55 kBT . This condition is thought to limit the scaling of longitudinal media and is commonly referred to as the "super-paramagnetic limit." Predictions of this limit range from 36-100 Gbits/in2 [5, 6].
The most significant problem encountered when using recording media that are susceptible to thermal instabilities is the decay of the media magnetization over time and the concomitant loss of recorded information. Examples of this effect on traditional longitudinal recording media are shown in Fig. 3 for a series of films with varying magnetic thickness, Mrt, values. In this example, the Mrt variation was achieved by thinning the magnetic layer while keeping the other growth parameters constant. The resulting films have the same grain diameters such that the grain volumes are proportional to the film thickness. The estimated amplitude loss over 10 years as a function of the magnetic film thickness is shown in Fig. 3. This is estimated from magnetization decay measurements over shorter time scales that are extrapolated to 10 years. Acceptable levels of signal decay vary depending on system design, but typically range between 10-20%. The conventional media (Fig. 3) with the largest Mrt value corresponds to the media shown in Fig. 2b and has excellent thermal stability. However, thinning the media by less than 50% results in an increase in amplitude loss above accepted levels. It is this dramatic increase in amplitude loss with reduced grain volume that is at the heart of the super-paramagnetic limit.
Historically, disk drive designers have had only two ways to maintain thermal stability as the media's grain volume decreases with increasing areal density:
1) improve the signal processing and error-correction codes (ECC) so fewer grains are needed per data bit, and
2) develop new magnetic materials with higher anisotropy constants, KU, that strongly resist any change to their magnetization.
However, higher KU alloys also are more difficult to write. While improvements in coding and ECC are ongoing, new approaches are needed to design media that are easy to write at very high areal densities but are much more stable than conventional media. We discuss two such approaches.
Antiferromagnetically coupled media
Antiferromagnetically-coupled (AFC) magnetic recording media is an approach to extend areal densities longitudinally. Developed by IBM Research, this new magnetic media uses multilayer interactions and is expected to permit longitudinal recording for future data density of 100 Gbits/in2 four times the data density of today's products without suffering from the projected data loss due to thermal instabilities. A schematic of the AFC recording media structure is shown in Fig. 4. The recording medium is made up of two ferromagnetic CoPtCrB layers (Mrttop and Mrtbottom) separated by an extraordinarily thin (three atoms) layer of nonmagnetic metal, ruthenium (Ru) [3, 7, 8]. The thickness of the Ru layer is tuned to couple the layers antiferromagnetically such that the magnetization in layers 1 and 2 point in opposite directions. The effective Mrt of the composite structure is given by the difference between the two layers Mrttop - Mrtbottom. This allows the Mrt to be scaled independently of the physical thickness of the media, thus maintaining thermal stability while reducing Mrt .
The magnetic hysteresis loop from an AFC media structure is shown in Fig. 4. The structure consists of two CoPtCrB alloy layers that are AF-coupled with a 6Å Ru layer, the thickness of which was chosen to maximize AF coupling. The magnetic alloy and the underlayer structure are the same as those used for the single-layer film in Figs. 2 and 3. The arrows in Fig. 4 indicate the orientations of the ferromagnetic layers at different points in the hysteresis loop. For large applied fields (during the writing process) the two ferromagnetic layers are parallel to the applied field. As the applied field is reduced to zero, the thinner bottom layer reverses and becomes antiparallel to the top layer and to the applied field resulting in an antiparallel remanent state. In a reverse applied field, the remanent magnetic state is stable until the top layer reverses and both layers become parallel and aligned in the negative saturation state.
Shown in Fig. 3 are the stability measurements for the AFC media compared to the single layer media. In this example, lower Mrt values were achieved by thinning the top layer for a fixed bottom layer thickness. The AFC media shows a weaker dependence of thermal decay on Mrt as compared to the single layer media. For the same Mrt values, the AFC media shows improved thermal stability and allows scaling of the effective Mrt value without compromising the thermal stability.
Figure 4. a) Schematic representation of the AFC media showing the two magnetic layers coupled antiparallel with a single transition shown. b) Magnetic hysteresis loop shown for an AFC media sample.
A major benefit of AFC media is that its writing and readback characteristics are similar to conventional longitudinal media. The output pulse sensed by the recording head is a superposition of the fields from transitions in both the top and bottom magnetic layers. This output is detected as a single pulse as with conventional media, so no modifications to the disk drive's recording head or electronic data channel components are required. The reduction of Mrt reduces the transition parameter, allowing higher density recording. Although the relation of the effective Mrt to the transition parameter is more complicated than for traditional single magnetic layer media, it has been shown that the lower Mrt results in higher bit resolution , demonstrating that AFC media allows continued scaling of longitudinal recording media.
Perpendicular multilayer media
As magnetic recording pushes beyond 100 Gbits/in.2, media with its magnetization perpendicular to the film plane may have a thermal stability advantage over media with its magnetization in-plane or longitudinal media. Unlike longitudinal recording, the perpendicular medium has its remanent magnetization normal to the film plane, has a thick soft underlayer beneath the recording medium, and the write head has a narrow trailing pole (Fig. 5). In this geometry, the magnetic flux of the head travels vertically through the medium into the soft underlayer that provides a return path for the flux. In this dual layer geometry, the highly permeable underlayer creates a mirror image of the pole head and allows the writing of media with higher KU values than is achievable in longitudinal systems. In addition, comparable densities are thought to be achievable with greater medium thickness in perpendicular recording as compared to longitudinal recording. This combination of larger media thickness and higher KU is the basis for improved stability predictions .
These advantages, however, have not been exploited in perpendicular recording systems with higher areal densities [11, 12]. One limiting factor has been creating a high performance recording media with the magnetization perpendicular to the film surface. One promising pathway is the use of magnetic multilayers composed of ultrathin magnetic and nonmagnetic layers (typically with thicknesses less than 5 atomic layers)  (Fig. 6). For sufficiently thin magnetic layers, the magnetization will point perpendicular to the film. Shown in Fig. 6 is the magnetic force microscope image of recorded tracks written into a multilayer media created from alternating layers of 4Å magnetic CoB layer and 7Å nonmagnetic Pd layers repeated 12 times. The structure was grown on a 4000Å Ni45Fe55 soft underlayer. This multilayer had an anisotropy value nearly twice that which can be written in longitudinal recording, clearly demonstrating the improved writability of perpendicular media on soft underlayers. These multilayers show no thermal decay under conditions comparable to those used to test longitudinal media, and the written transitions were unchanged after exposure to external magnetic fields for >100 min. Such results highlight the potential for multilayer structures as thermally stable perpendicular recording media, allowing continued growth of recording densities.
The transition from traditional longitudinal recording media to AFC and potentially toward perpendicular recording poses serious challenges with regard to fabrication of multilayered magnetic architectures. For longitudinal magnetic media it is desirable to have the magnetic anisotropy axis of the magnetic grains aligned parallel to the film plane. This is achieved by controlling the crystallographic orientation of the media layer through proper choices of seed layers grown onto the glass substrates. In the AFC media structure, care must be taken to maintain the microcrystalline growth characteristics in both magnetic layers. Such control can be achieved with Ru interlayers because its crystallographic symmetry matches that of the magnetic layers. Ru has the additional advantage of not diffusing into the media layers during growth, thus maintaining the integrity of the three-atom thick layer.
The key fabrication issue of depositing a three-atom-thick Ru layer with suitable uniformity and repeatability over the entire disk surface has proven to be a significant challenge. If regions of the disk are not coated by the Ru layer, the two magnetic layers will couple ferromagnetically, significantly altering the recording performance in that region of the disk. However, by tight control of the process time and powers, such ultra-thin materials can be reproducibly deposited.
For multilayer structures (Fig. 6) the added complexity presses the capabilities of conventional commercial sputtering tools. The growth of AFC media requires two additional layers whereas Co/Pd perpendicular multilayer media alone consists of more than 20 layers, comprising repeating pairs of two alternating materials. The complexity of depositing multilayered magnetic structures is further increased by additional constraints such as single station deposition, short process time, and uniformity control over an entire substrate.
Conventional sputtering stations are designed for depositing single layers. The ability to deposit two different materials within a single station is currently available only in R&D tools. To address the manufacture of multilayer structures, novel sputtering designs and geometries are being developed. Included are multiple sources placed within concentric rings or mounted on a spindle and rotated to achieve uniformity. Tenth of a second switching and sputtering times of individual layers are required, which means fast power ramp rates, rapid switching of power supplies, and tight software control. This, in part, provides the needed controls to achieve film thicknesses of just a few atoms, 3-10Å, but also addresses throughput issues faced in processing the large, complex disks. Other manufacturing concerns include the control of plasma misfires that result in gaps or thinner layers.
The onset of thermal instabilities with scaling of longitudinal recording media is well established. This suggests that continued increases in recording densities will require new thin-film materials and architectures that deviate from traditional scaling. The development of AFC longitudinal media and perpendicular multilayer media are two such examples. AFC media, in particular, allows scaling of the magnetic thickness independently of physical thickness, maintains thermal stability, and allows continued scaling of longitudinal media. AFC media implementation requires few changes in hard-disk-drive design, and will likely delay consideration of more complex techniques for very high-density magnetic recording like perpendicular recording, patterned media or thermally-assisted recording.
- B.A. Gurney et al., Datatech, Fourth Edition, pp. 41-46, IGC Publishing Ltd., London, 2000.
- H.N. Bertram, 'Theory of magnetic recording", (Cambridge University Press, England) 1994.
- M. Madison, M. Doerner, K. Tang et al., IEEE Trans. Mag., 37, 1052, 2001.
- D. Weller, A. Moser, IEEE Trans., 35, 4423, 1999.
- S.H. Charap, P.L. Lu, Y. He, IEEE Trans. Mag. 33, 978, 1997.
- H.N. Bertram, H. Zhou, R. Gustafson, IEEE Trans. Mag. 34, 1845, 1998.
- E.E. Fullerton, D.T. Margulies, M.E. Schabes, et al., Appl. Phys. Lett., 77, 3806, 2000.
- E. Abarra, et al., Appl. Phys. Lett., 77, 2581, 2000.
- J. Lohau, et al., Appl. Phys. Lett., 78, 2748, 2001.
- N. Bertram, M. Williams, "SNR and Density Limit Estimates: A Comparison of Longitudinal and Perpendicular Recording," Digests of The Mag. Rec. Conf., 9-11, Aug. 1999, San Diego.
- D.A. Thompson, J. Magn. Soc. Japan, 21, 2, 1997.
- K. Ouchi, IEEE Trans. Mag., 36, 16, 2000.
- K. Takano, et al., J. Appl. Phys. 87, 6364, 2000.
Eric E. Fullerton received his BS and PhD in physics from Harvey Mudd College in 1984 and from the University of California, San Diego, in 1991, respectively. He is a research staff member in the Storage Systems and Technology Department at IBM Almaden Research Center, 650 Harry Rd., San Jose, CA 95120; ph 408/927-2430, email eef@ almaden.ibm.com.
David T. Margulies received his PhD from the University of California, San Diego, in the Center for Magnetic Recording Research. He works in the Advanced Recording Media Group at IBM Almaden Research Center.
Andreas Moser received his PhD from the University of Basel, Switzerland, and in 1996, he became a postdoctoral fellow at IBM Almaden Research Center. He subsequently joined IBM as a research staff member.
Kentaro Takano received his BA and PhD in physics from the Univ. of California, Berkeley, in 1990, and from U. Cal., San Diego, in 1998, respectively. He is an advisory engineer at IBM Almaden Research Center.