What challenges remain to achieve heat-assisted magnetic recording?
09/01/2007
Magnetic recording technology in HDDs has advanced remarkably since the invention of this computer peripheral device in 1956. The most concise performance metric of this data storage unit is AD, which over this period has progressed in products shipped from 2 kbit/in2 to about 140 Gbit/in2 today, an increase by a factor of 70 million. This exponential growth roughly parallels the better known Moore’s Law rate of advance of transistor count per device in integrated semiconductor technology since 1971.
However, many feel the HDD record is more impressive, as it relates to an electromechanical device whose characteristic dimensions and clearance between relatively moving components is currently moving below 100 atomic diameters. An oft quoted analogy to a recording head flying over a moving medium with non-flat topography is that it resembles a 747 airliner flying at 500 mph over hilly ground with a clearance of one inch. The refrigerator-sized HDD units of the 1950s and 1960s have shrunk to well below one cubic inch in portable consumer electronic devices of today, and yet they store 10,000 times more digital bits.
This forward march has been so extreme that engineers are continually challenged to address and, if possible, push back a wide range of “limits.” Fortunately, most of these limits are not fundamental physical barriers, but technological limits imposed by the current state-of-the-art. Of course, with the dimensional shrinkage indicated above, it is clear that fundamental limitations such as the atomic structure of matter loom in the future. Nevertheless, HDD technologists estimate that perhaps another factor of 1000 growth in AD is possible, in principle, by cleverly engineering several candidate novel approaches to their full potential [1]. These final phases of HDD technology evolution could take several decades to unfold as advances become more difficult, but the focus of this article is the near-term development of one particularly appealing approach that has potential to deliver the next factor of ten (at least) AD increase: HAMR.
 Figure 1. The basic geometry of recording a track of information on a medium with a magnetic head operating in a) the longitudinal or b) perpendicular mode. |
As its name implies, HAMR adds one new element to the usual procedures of magnetic recording, namely localized heating of the medium during the writing process, followed by rapid cooling back to ambient temperature. The purpose is to reduce the temperature-dependent medium magnetic anisotropy K(T) (responsible for magnetic alignment of media grains or particles in the recording process) by virtue of its negative slope as one warms the disk toward the Curie temperature. High K at ambient temperature is needed to retain the thermal stability of diminishing media grain volumes against spontaneous magnetic reversal under ceaseless thermal agitation of the atoms of the magnetic medium. Suppressing even a slight tendency toward spontaneous magnetic particle reversal (dubbed the “superparamagnetic limit”) is essential to retain an acceptable signal-to-noise ratio (SNR) in the write
ead process at higher AD. Temporarily lowering K(T) is tantamount to reducing the magnetic field strength necessary for switching the polarity of magnetic grains-that is, it boosts (assists) magnetic writability.
How HAMR fits into HDD evolution
At present, magnetic recording in HDDs is completing a transition from the use of longitudinal to perpendicular magnetic orientation of the medium. This terminology refers to the way the tiny magnet particles or grains are aligned in the recording process. Longitudinal recording means that the recorded magnets align along the circumferential tracks on a disk or along the tracks on a tape, and this has been the recording mode in both technologies since their commercial emergence fifty or more years ago (Fig. 1). These magnets thus lie in the plane of the recording layer. An energetic advantage for the perpendicular configuration when packing the regions of opposite magnetic polarity closer and closer (i.e., higher AD), and a worthwhile advantage in writability and SNR potential, now gives a performance edge to perpendicular recording.
What technology extendibility is provided by this change in the recording mode? It is the threat of degrading thermal stability in longitudinal recording that explains this. Perpendicular recording corrects this situation (at least in the short term) with perhaps the least pervasive modification of the longitudinal recording system. However, a weakness of the perpendicular approach is that its AD improvement potential relative to the best commercial longitudinal HDD systems is generally thought to lie in the 3-4 times range-not a huge gain given the relative complexity of the system transition. It is perhaps only the drastic slowdown in the AD compound annual growth rate (CAGR) since the year 2000 (from a peak of ~120% CAGR to the current 20-40% range) that makes this modest extension tolerable.
 Figure 2. Coercivity Hc(T) and saturation magnetization Ms(T) for thermomagnetic recording media with Tcurie = 600K. |
This situation begs the question: “What technology can replace perpendicular recording, and when will it be needed in products?” Of course, the answer to the ‘when’ part depends on the evolution of AD CAGR over the next few years, for it will determine how quickly perpendicular recording will approach its own limits. CAGR trends will depend on market requirement factors that are difficult to forecast, as well as the pace of development of the incumbent and competing technologies. Regarding the ‘what’ part of the question, HAMR is certainly one of a few candidate technologies, but a transition from perpendicular to HAMR is a rather large step. Many feel that it is more likely that we will see incremental improvements in “write-assist” technology that might be technically and economically more feasible to execute without disrupting the $30B per year HDD business than an abrupt jump to HAMR at some future time.
There is another fairly radical technology transition following the evolution of perpendicular recording that is being discussed. It too can address the same set of needs that fuels the pursuit of HAMR. That approach is called “bit patterned media” [BPM], and it would drastically alter the structure of the magnetic storage medium [2]. Currently, and over the better part of the past century, the magnetic recording medium has been a continuous, layered material that is composed of small, weakly coupled magnetic particles (in pigment-type coatings) or grains (in thin films). The distinguishing feature is that this medium extends continuously in a plane to the edge of the physical substrate (disk, tape, card). BPM uses patterned, isolated “islands” of homogeneous magnetic material to store information, or digital bits. There are non-magnetic buffer zones between the magnetic islands. Such a change in configuration significantly enlarges the magnetic “particle” and its thermal stability to extend AD growth capability, and at the same time it enhances writability. The technical barrier for the transition to BPM is one of holding down media cost in the radical reconfiguration of that key component. Additionally, there are new challenges in maintaining reliable high rate dynamic operation when writing must be done only at specific locations on the moving medium (the islands).
The physics of HAMR
The role of magnetic and thermal field gradients. To envision the physical processes involved in HAMR, begin with the present methodology of magnetic recording. A small but strong electromagnet is placed close to a material that can be locally magnetized for long term retention of information. The size of critical features on the head, and the spacing of the head from the medium, must be comparable to the length scale of the information features being recorded to preserve writing and reading resolution (i.e., SNR). When components are scaled this way, the writing magnetic field is well localized, attaining its maximum strength in the vicinity of the magnetic bit to be recorded, and weakening rapidly over the scale of a few bits, which is the situation in all magnetic recording systems in use.
HAMR adds to this configuration a means of rapidly locally heating the recording medium by several hundred degrees Celsius. The thermal profile imparted to the moving medium must be local, and have a spatial variation comparable to the head’s magnetic field currently employed. In HAMR, the size of the thermal profile becomes the critical determinant of AD potential, not the magnetic field extent. As the size of digital bits continues to shrink with AD increase, reducing the scale of this localized heating creates a significant challenge for HAMR in terms of the design of both the heat source and the receiving medium.
What is the dimensional scale of digital bits now and in the future? HDD systems recently passed the AD = 100 Gb/in2 marker, and the next major AD milestone ahead is 1 Tb/in2. At these densities, the bit dimensions are ~25nm (L) × 200nm (W), and 10nm (L) × 65nm (W), respectively. If in the ultimate AD scenario of ~103× the present density, with square or round bits utilizing BPM, the linear dimension of a bit would be ~2.5-3nm, or of the order of 10-12 atoms on an edge [1].
A HAMR-based system with full extendibility across this AD range must anticipate a capability to heat locally on a length scale ranging from perhaps ~100nm to several nanometers. Initial HAMR concepts have keyed off of pioneering developments in optical information storage, which emerged commercially in the 1980s and have now evolved through CD, DVD, MO, and the introduction of blue laser recording [3]. This makes initial consideration of optical heating of a future magnetic medium a natural development, although other methods have been discussed [4]. Use of optical heating beams in the dimension range indicated implies use of near-field (NF) optics rather than the more familiar diffraction-limited techniques used historically in optical storage systems. Near-field optics is less well understood and is hence an active field of current research in its own right.
In HAMR, one employs thermomagnetic recording, the same technology first commercialized in magneto-optical (MO) disk drives from the mid-1980s. The essence of thermomagnetic recording is utilizing a medium whose magnetic properties have a tailored dependence on temperature (Fig. 2), where we see a general characteristic of most ferromagnetic or ferrimagnetic materials. There exists a distinct transition temperature above which cooperative magnetic behavior vanishes (the Curie temperature Tc), and a monotonic decrease in important magnetic parameters such as coercivity (the applied magnetic field required to switch half of the magnetization) and magnetization below Tc. When the local medium temperature is elevated near or above Tc, the magnetism weakens or vanishes, and it can be readily re-established with a desired polarization upon cooling if the applied field is set correctly.
Quantitative insight on the boost in effectiveness gained with thermomagnetic recording can be obtained by considering a simple relation for the gradient (spatial rate of variation) of the field involved. Earlier, we implied that the local variation of magnetic field around a recording head is critical, and indeed this is so. In HAMR we may forego emphasis on the head’s magnetic field variation, and instead rely on an effective recording field gradient as a product of two other gradients:
dH/dx = dH/dTdT/dx
The factors on the right in this equation are respectively the rate of change of the medium anisotropy field (analogous to switching field or coercivity) with temperature and the spatial gradient of temperature in the medium imposed by the heating source. This effective field gradient is potentially much stronger than that available from conventional magnetic heads.
An important design tool for HAMR is computer simulation, as it is in other areas of data recording. Putting together the optical, thermal, and magnetic pieces of the simulation results in an analysis capability that is reasonably comprehensive. This can address the role of a broad set of system parameters, spanning the various components of the system: head, medium, and their interface. We show in Fig. 3 an example of simulated HAMR writing and read back of a single track on a modeled medium. Figure 3a shows the thermal profile, while Figs. 3b and 3c show the head magnetic field and recorded magnetization pattern (isolated transition), respectively. Finally, in Fig. 3d we have a read back signal scan obtained as a magneto-resistive read head is stepped across track.
Medium magnetization dynamics in HAMR. New unique recording mechanisms arise in HAMR. The most important of these from the recording physics viewpoint can be attributed to the rapid elevation and reduction of medium temperature on a local spatial scale. Because magnetic properties such as anisotropy and magnetization have inherent temperature dependence, we need to inquire about the details of magnetization dynamics in a reversal process that now has the added complexity related to the modulation linked to T(x,y,z). This involves fundamental aspects of magnetism in solids since HAMR uses variation in temperature to effectively weaken or destroy ferromagnetism, and then restore it on a sub-nanosecond time scale. This is an active area of scientific research driven in part by interest in the HAMR technology.
HAMR compatibility with other advanced recording technologies. As we have seen, HAMR is a means of assisting magnetic writing by adding thermal energy to the medium. As such, it could be made compatible with other schemes for write assistance. In principle, HAMR should also be fully compatible with BPM to realize the ultimate AD gain potential of very small ferromagnetic particles. This is a statement about recording and thermal stability of stored information, but it does not address the equally important matter of detection SNR [1].
Integrated head challenges
There are three major challenges for the HAMR head design and process: 1) Apply a large magnetic field confined to a spatially small area; 2) Form an intense and spatially small optical spot on the media; 3) Properly align the optical and magnetic fields. Let us consider them in turn. The first challenge can be addressed with a magnetic field delivery system very similar to current magnetic recording write heads (a current-carrying coil wrapped around a magnetic pole where the pole is patterned down to ~100nm and is in close proximity to the media; see Fig. 1b and Fig. 4).
The second challenge can be addressed using a planar solid immersion mirror (PSIM) [5] to form a diffraction limited focal spot, as in Fig. 4 [5-7] PSIM consists of a planar waveguide (WG), which is a high index of refraction (n) layer sandwiched between two low n layers. The WG is then patterned into the shape of a parabola and gratings are formed in the WG to couple light from a laser into the WG. When light strikes the parabola’s edge, it is reflected and focused at the focal point of the parabola, which is placed at the air bearing surface (ABS). The width of the spot in the plane of the WG determines the data track width and is ≈0.5*λ/NA, where λ is the wavelength of the light in a vacuum and NA is the numerical aperture of the PSIM. Using proper WG materials, a spot width of ~λ/4 can be achieved, which would be ~100nm for λ = 405nm light. Since the NA >1, the high spatial frequency components of the light that give the small spot do not propagate in air. Thus, the PSIM needs to be held in close proximity to the media so that the evanescent waves can couple from the PSIM to the media before they decay, which can be achieved with the standard magnetic recording slider that flies <30nm above the media.
 Figure 5. a) SEM and b) Optical images of the ABS of a HAMR head. The 488nm light was coupled into the waveguide when the b) image was taken. |
The third challenge can be addressed by the design shown in Fig. 4, and thousands of such HAMR heads were fabricated using standard thin-film processes. Fig. 5a shows a scanning electron micrograph of the ABS of one of these heads, and Fig. 5b shows an optical image of a HAMR head when λ = 488nm light is coupled into the WG. It can be seen that the light is focused right under the magnetic pole. The head in this image has apertures fabricated during the process that are used to block the side lobes in the intensity profile, which are otherwise present. Figure 6a shows a scanning near-field optical microscopy (SNOM) image of the focused spot at the ABS for a HAMR head with magnetic poles and apertures. The optical full width at half maximum (FWHM) for this device is 140nm and there are almost no side lobes in the intensity profile. λ = 413nm light was also coupled into a PSIM only head, and the FWHM was ~90nm.
Figure 6b shows a magnetic force microscopy (MFM) image of HAMR specific media that was written on a spin-stand with a fully integrated HAMR head with and without light coupled into the WG. It can clearly be seen that HAMR writing is necessary for this high anisotropy media to be recorded. To achieve a spot size smaller than these diffraction-limited spots, a near-field transducer (NFT) using surface plasmon resonance may be employed, such as circular [8, 9], rectangular [10], or “C” apertures [11], or a bow tie [12] or beaked triangle [13], all of which can achieve spot sizes <50nm. The light power needs to be efficiently coupled from the WG to the NFT and then to the media for the NFT to be effective. The NFT will need to fly closer to the media (~5nm) than the PSIM described above to efficiently couple the power in a small spot to the media.
Outlook
HAMR is a future magnetic recording technology being seriously considered on industry roadmaps. It brings excellent promise to help deliver the full stored information density potential allowed by the recording physics of known magnetic materials. It also appears to be compatible with several other contending technologies that can contribute to the assault on the ultimate physical limits of magnetic recording, including BPM. A low-cost, integrated head is a crucial component of a HAMR HDD system.
Acknowledgments
We gratefully acknowledge the collaboration of our Seagate Research colleagues. This work was performed as part of the INSIC HAMR ATP Program, with support of the US Dept. of Commerce, NIST, Advanced Technology Program, Cooperative Agreement #70NANB1H3056.
References
- T.W. McDaniel, “Ultimate Limits to Thermally Assisted Magnetic Recording,” J. Phys.: Cond. Matter 17, R315-R332, 2005.
- Z.Z. Bandic, et al., “Patterned magnetic media: impact on nanoscale patterning on HDDs,” Solid State Technology, Data Storage supplement, Sept. 2006.
- INSIC Optical Data Storage Roadmap, published by the Information Storage Industrial Consortium, San Diego, California, 2006.
- S.J. Greaves, H. Muraoka, “Heat-assisted Recording using Electron Beams,” J. Magn. Soc. Jpn. 30(6)-2, 567, 2006.
- W. Challener, et al., “Miniature Planar Solid Immersion Mirror with Focused Spot Less Than a Quarter Wavelength,” Opt. Exp., 13(18), 7189-7197, 2005.
- C. Peng, et al., “Near-field Optical Recording with A Planar Solid Immersion Mirror,” Appl. Phys. Lett., Vol. 87, pp. 151105-1-151105-3. 2005.
- T. Rausch, et al., “Near-field heat-assisted Magnetic Recording with a Planar Solid Immersion Lens,” Jpn. J. Appl. Phys., Vol. 45, no. 2B, pp. 1314-1320, 2006.
- L.Yin, et al., “Surface Plasmons at Single Nanoholes in Au films,” Appl. Phys. Lett., Vol. 85, no. 3, pp. 467-469, Jul. 2004.
- E. Popov, et al., “Surface Plasmon Excitation on a Single Subwavelength Hole In A Metallic Sheet,” Appl. Opt., Vol. 44, pp. 2332-2337, April 2005.
- X. Shi, L. Hesselink, “Mechanisms for Enhancing Power Throughput from Planar Nano-Apertures for Near-Field Optical Data Storage,” Jpn. J. Appl. Phys., Vol. 41, 1632-1635, 2001.
- B.C. Stipe, et al., “Ridge Waveguide for Thermally Assisted Recording-Optimization, Scaling, and Wavelength Dependence,” paper FB-12, INTERMAG 2006, San Diego, CA.
- R. Grober, S. Bukofsky, S. Selberg, “Application of Near-field Optics to Critical Dimension Metrology,” Appl. Phys. Lett.. Vol. 70, pp. 2368-2370. May 1997.
- T. Matsumoto, et al., “Writing 40-nm Marks by Using a Beaked Metallic Plate Near-field Optical Probe,” ISOM/ODS’05, Honolulu, Hawaii, July 14, 2005.
Terry McDaniel received his BS in physics from Wittenberg U., and his MS and PhD in physics from Michigan State U. He is a research staff member at Seagate Research, 1251 Waterfront Place, Pittsburgh, PA 15222 United States; ph 209/295-6735, e-mail [email protected].
Michael A. Seigler received his BS in electrical and computer engineering from Pennsylvania State U., and his MS and PhD in the same subjects from Carnegie Mellon U. He is the research engineering manager of the Materials & Device Processing Group at Seagate Research.