NanoStructured Materials, Vol. 11, No. 8, pp. 965–986, 1999 Elsevier Science Ltd Copyright © 2000 Acta Metallurgica Inc. Printed in the USA. All rights reserved. 0965-9773/99/$–see front matter
Pergamon
PII S0965-9773(00)00429-3
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE A.K. Menon and B.K. Gupta Read-Rite Corporation, Fremont, CA 94539, USA (Received December 1, 1999) (Accepted December 5, 1999) Abstract—The exponential increase in areal density of recording of magnetic storage devices has reduced the recorded bit size to nano length scale. This has profound implications on all aspects of the storage system including recording device modeling, materials, fabrication, metrology, characterization, and tribology of the head-disk interface. In this paper, the impact of nanotechnology in extending the data storage device storage systems is explored with an emphasis on the fabrication and characterization of nanolayers and structures. ©2000 Acta Metallurgica Inc.
1. Introduction With the advent of the information superhighway, high definition television, replacement of chemical photography with digital images, digitization of images (both still and video motion), simulation of engineering designs and events by finite element modeling, enormous amount of data need to be viewed, manipulated, and stored. The evolution of new applications which require storage capacity in the range of terabyte/disk are shown in Fig. 1 (3). A video ROM two-hour recording requires a storage capacity in the 6 – 8 Gigabytes range. The data storage capacity for future applications is likely to approach tens or hundreds of terabytes. Obviously, the data rate transfer rates need to keep pace with the data storage consumption requirements. For instance, interactive 3-d video will require an areal density of 10 TB.in⫺2 and data transfer rates of 100 Gb/sec, Fig. 2. The success of the hard disk drives originate from a consistent enhancement in storage capacity and performance combined with significant reductions in price per Gigabyte. The focus of the commercial market’s data storage requirements is changing continuously. Today most of the revenue is generated from the 3–10 Gigabytes desktop hard drives. In the near future, the demand for the hard drives will grow in the mobile laptops and highend servers market. By year 2001, the revenue of the hard disk drive industries is expected to reach 40 million US dollars, Fig. 3 (1998 Disk/Trend Report). Because of rapid advances in the recording areal density, the average price per megabyte is continuously falling. Ever shrinking size of the pole tip structure to nanometer dimensions poses new challenges for the head and media design, characterization of films on nano-scale, deposition of films in the subnanometer range, lithography and patterning sequence, and achieving good tribology performance. In this paper, key issues, challenges, and limitations of the nanotechnology that affect the magnetic performance and the tribology of hard disk drives are discussed. 2. Areal density growth The areal density trends of various data storage devices over two decades are shown in Fig. 4. In small personal computers need for storage is growing from hundreds of megabytes to gigabytes, and in larger 965
966
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 1. Evolution of new applications for massive data storage.
systems from tens of gigabytes to terabytes. Correspondingly, the data rate has increased from hundreds of kilobytes per seconds to megabytes per seconds in PC; and from hundreds of megabytes per seconds to gigabytes per second in larger systems. To achieve high areal densities and data transfer rates at lower cost, commercial market utilizes magnetic hard disc, magnetic flexible tape, and optical disc devices. Although several competing technologies for data storage are available, still magnetic hard disc drives continue to be the primary, high performance storage device of choice. In the past five years, areal
Figure 2. Capacity and data rate evolution for various data storage applications (Esener, 1998).
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
967
Figure 3. Worldwide hard disk drives sale revenue (Disk/Trend Report).
density of storage has grown at 60% annually, Fig. 5. The growth is spurred by the introduction of giant magnetoresistive (GMR) head technology and by proportionally reducing all dimensions. In the hard disc drive storage arena, technologies in all aspects of recording are continuously being stretched to the limit to sustain the 60% compound annual growth rate (CGR) (5). Recently, Read-Rite Corporation demonstrated for the first time the feasibility of 13.5 Gb.in⫺2 areal density with the GMR head. Areal densities of 20 and 40 Gb.in⫺2 are expected in production by year 2002 and 2004, respectively (5). The size of the recording bit necessary for higher linear bit and track densities is achieved by scaling down the dimensions of the pole tip structure to nanometer dimensions. The bit aspect ratio reduces from 20 to 4 with an increase in areal density from 0.1 to 100 Gb.in⫺2. Typical lengths and widths of
Figure 4. Areal density growth of rigid and flexible magnetic media, magneto-optical, and near field optical storage devices.
968
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 5. The areal density growth for hard disk drives. The cumulative areal density growth rate (CAGR) for hard disc drives is mainained at 60% since 1990.
recording bits in 1, 10, and 100 Gb.in⫺2 areal densities are 3.0 ⫻ 0.2 m, 0.8 ⫻ 0.066 m, and 0.3 ⫻ 0.075 m, respectively, Fig. 6. Changes in various head, disc, and interface related parameters to achieve increasing areal densities for hard disc drives are listed in Table 1. The form factor (size) of the recording head shrinks with the areal density. Head design parameters such as read gap, reader track width, and stripe height are also scaled down accordingly to the nanometer dimensions. The magnetic spacing decreases from 75 to 12 nm as areal density increases from 1 to 40 Gb.in⫺2. Thickness of the wear coatings on head and disc are reduced to zero at 20 Gb.in⫺2 areal density. Key technologies influencing high areal density are the new GMR heads, positioning servo-mechanisms for high track densities, low noise magnetic media with high coercivity, and low magnetic spacing head-disc interface
Figure 6. Change in bit dimensions with increasing areal density. The bit aspect ratio decreases by a factor of five with an increase in areal density from 0.1 to 100 Gb.in.⫺2.
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
969
TABLE 1 Changes in Various Head, Disc, and Head-Disk Interface Related Parameters with Increasing Areal Density for Hard Disc Drives Attributes Projected Recording technique Head
Media
Design
Type Form factor, % MR Sensitivity, ⌬R/R Read gap, nm Read track width, m Stripe Height, m Overcoat Thickness, nm Substrate Texturing Magnetic film, Thickness, nm Overcoat thickness, nm Lube
Interface
Mechanics Spinning speed, rpm Data rate, MB/s, 65 mm Magnetic spacing nm Loading mechanism
1 Gb.in⫺2
5 Gb.in⫺2
10 Gb.in⫺2 Projected
20 Gb.in⫺2 Projected
40 Gb.in⫺2 Projected
1996 Longitudinal
1998 Longitudinal
2000 Longitudinal
2002 Longitudinal
2004 Longitudinal
Thin Film, MR Inductive 50 MR 2% 250 2.7 1 10 A1-NiP, Glass
MR
Spin Valve GMR 30 GMR 8% 160 0.45 0.4 3 Glass, Si
Advanced GMR 10–15 GMR 32% 140 0.30 0.3 0 Glass, Si
Advanced GMR 10–15 GMR 32% 110 0.19 0.2 0 Glass, Si
Ultra Smooth 10–15
Ultra Smooth 7–10
Ultra Smooth 5–7
5 PFPE ⫹ Additives Fly or Prox. 10,000 150–300 25 Dynamic loading
0 PFPE ⫹ Additives Contact ⬎10,000 210–425 18 Dynamic loading
0 PFPE ⫹ Additives Contact ⬎10,000 300–600 12 Dynamic loading
Mech/Laser 20–25 10–15 PFPE ⫹ Additives Flying 7200 36–57 75 CSS
30 GMR 8% 200 0.7 0.5 6 A1-Ni-P, Glass Ultra Smooth 14–21 10 PFPE ⫹ Additives Flying 10,000 105–210 40 CSS, Dynamic loading
for high linear bit densities. Fabrication of GMR heads for high areal density would require the characterization, deposition, and patterning of nanolayers with controlled thickness in subnanometer range. 3. Magnetoresistive heads Magnetoresistive (MR) head senses magnetic field directly and writes with a separate element. MR heads are based on the anisotropic magnetoresistive (AMR) effect where the resistance of a NiFe (Permalloy) film changes with the magnetization. In the MR head, the magnetic field produced by a recorded bit on the disk rotates the magnetization angle in the magnetoresistive element (MRE). When the magnetization angle () in the MRE is properly biased with a vertical bias field, small changes in resistance, ␦R, are almost directly proportional to small magnetic fields from the recorded bit. The design of a soft adjacent layer (SAL) vertically biased MR head is shown in Fig. 7. In this design, a film of low-coercivity, high-permeability magnetic material is placed adjacent to the MRE. The magnetoresistive coefficient is defined by ␦R/Ro, which is the maximum fractional change in the resistance of the element. This coefficient serves as a figure of merit of the MR materials and MR structures. Higher MR coefficient enables greater change in the output voltage of the MR head. The magnetoresistive coefficient of Permalloy (81% Ni ⫹ 19% Fe) is about 4% in 100 nm thick films and 2% in 2 to 30 nm thick films.
970
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 7. The structure of the soft adjacent layer (SAL) vertically biased anisotropic MR head.
4. Giant magneto-resistive (GMR) heads The main emerging technology for the recording head is giant magneto-resistive (GMR) technology, which was unknown until 1988. The GMR heads will enable areal densities to 15 Gb.in⫺2 and beyond. The primary advantage of the GMR head is greater sensitivity to the magnetic field generated by the recorded bit on the disk. The increased sensitivity makes it possible to detect smaller recorded bits and read those at higher data rates. In the GMR head, resistance changes occur in the thin magnetic films separated by a thin conductor. One film has a variable magnetic orientation and is influenced by the disk’s magnetic field, the second film has a fixed or pinned orientation. The GMR effect is based on the exchange coupling between a multiplicity of thin films of a ferromagnetic material (Fe) separated by thin films of a nonmagnetic material (Cr). The GMR phenomenon is believed to depend principally on the selective scattering of conduction electrons at the interfaces between the ferromagnetic and nonmagnetic layers. When an electrical field is applied, the spin-oriented conduction electrons accelerate until they hit a scattering center. The average distance the conduction electron accelerates is called the coherence length and it is of the order of ten nanometers in the metals used in the GMR superlattices. If the interlayer thickness is less than the coherence length, the conduction electron arrives at the interface of the adjacent ferromagnetic film still carrying its original spin orientation. When the adjacent magnetic layers are magnetized in a parallel manner, the arriving conduction electron has a high probability of entering the adjacent layer with negligible scattering, because its spin orientation matches with the layer’s majority spins, Fig. 8. On the contrary, when the adjacent magnetic layers are magnetized in a antiparallel manner, the majority of the spin-oriented electrons suffers from strong scattering at the interface because they do not match the majority spin orientation. When the magnetic layers are in the ferro state (magnetized parallel) the resistance is low and vice versa in the antiferro state. In the GMR structure, materials used in the interfacial structure must be immiscible and the interlayer must be thinner than the conduction electron coherence length. The GMR coefficient is strongly influenced by the domian structure, layer thickness, and atomic and mechanical defects in the GMR layers. The structure of a GMR head is schematically shown in Fig. 9. The GMR structure consists of a multilayer sandwich of CoFe and Cu with a pinned layer (FeMn, NiMn, PtMN, PdPtMn, IrMn, or NiO) at the top. The GMR coefficient of a multilayer structure depends on the thickness of conducting Cu layer and number of the bilayers, Fig. 10 (9). The two peaks are observed near 1.3 and 2.4 nm in the 14 bilayer structure where the interlayer exchange coupling is expected to be antiferromagnetic. This
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
971
Figure 8. (a) Scattering of majority electrons shown in the film cross-section and (b) top-down view of sensor showing magnetization rotations.
behavior demonstrates that the GMR coefficient is strongly influenced by the number of bilayers and the thickness of the conducting interlayer. Symmetric (or dual) GMR structures offer the possibility of achieving large GMR coefficients in the layers which exhibits relatively low saturation fields. Figure 11 illustrates a symmetric, top, and bottom spin valve structures (2). The symmetric spin valve has two pinned layers (NiO) on the top and bottom. The top or bottom spin valves structures have only one pinned layer either at the top or the bottom of the structure. In conventional spin valves, the pinning film of FeMn acts by providing an exchange bias
Figure 9. Cross-section of the GMR structure.
972
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 10. Magnetoresistance for varying Cu spacer layer thickness and bilayer number. Two peaks at 1.3 and 2.5 nm indicate the antiferromagnetic inter-layer exchange coupling (Kief et al., 1996).
whereas the NiO acts by inducing a very large coercivity in the adjacent Co film. In the bottom spin valve, the top Co film is unpinned and is free to switch magnetically at relatively low fields. In symmetric spin valve the top and bottom Co films are pinned and the central Co film is free. The symmetric spin valve structure exhibits higher GMR coefficient of 23.4% as compared to that of bottom structure (17%). The GMR coefficient is significantly increased by reducing the thickness of Cu layer in the symmetric spin valve structure, Fig. 12. The impediment of thinner Cu film is that the ferromagnetic coupling of the central Co film to the top and bottom Co film rises very sharply for Cu thickness lass than about 1.8 nm.
Figure 11. An illustration of symmetric, top, and bottom spin valve structures (Egelhoff et al., 1996).
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
973
Figure 12. GMR sensitivity versus Cu film thickness for symmetric spin valve structure (Egelhoff et al., 1996).
5. Fabrication of giant magnetoresistive recording heads GMR structure requires deposition and control of thin films contained in the multilayer nanostructure. Future challenges to GMR fabrication involve control of nanometer dimensions and tolerances of ultrathin films, particularly the conducting spacer film which can be only 10 –15 atomic layers thick. GMR involves new film materials that require precise process control to assure stability and compositional uniformity to achieve high GMR coefficient at high areal density. For the deposition of ultrathin films of GMR materials, dc magnetron sputtering and ion beam sputtering techniques are used. The schematics of dc and ion beam sputtering setups are shown in Fig. 13. Ion beam sputtering provides a tighter control over the film thickness and impurities in the films during the deposition process. The reduced track width (700 and 200 nm for 5 and 40 Gb.in⫺2, respectively) poses new challenges for film deposition, lithography and patterning sequence, Fig. 14. With current technology, track widths of about 1.4 m are obtained by using G-line steppers. To achieve track width in a few hundred nanometers range, deep ultraviolet (UV) radiation exposure need to be used in the photolithography process. In the next few years, electron-beam or X-ray lithography techniques may be used for achieving adequate photolithographic resolution and alignment limits. Track width can be further reduced by trimming the writer pole on both sides. The writer pole can be trimmed by using a focused ion beam (FIB) system. The FIB system is schematically shown in Fig. 15. In the FIB process, a focussed beam of gallium ions sputter etch the localized region and reduce the width of the writer pole. 6. Superparamagnetic limit As the areal density increases astronomically, it is necessary to examine some of the fundamental limits and quasi fundamental limits of magnetic recording. The key fundamental limits of the recording are media noise versus thermal stability of the recorded bit, smallest magnetic spacing achievable, and maximum switching speed of current head/media materials. Some of the quasi fundamental limits include scaling of magnetic behavior to submicron head geometries, signal detection at very low signal to noise ratio, and precision of single stage actuator positioning over data tracks. The ultimate limit for room temperature magnetic storage is thermal demagnetization of the recorded bits. In a grain of the magnetic material, the direction of each magnetic moment oscillates in an energy well about its equilibrium direction. There is some finite probability that, in a time t, the thermal disturbance will be sufficient to induce a reversal of the equilibrium direction causing the decay of the recorded dibits. If a number of these reversals occur during the time of a magnetization measurement, the grain is said to
974
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 13. Schematics of (a) dc sputtering system and (b) ion beam sputtering system.
be “superparamagnetic.” Superparamagnetic behavior can be avoided if the energy barrier to reversal, EB, is much greater that kT, where k is Boltzmann constant and T is the absolute temperature. Magnetic thermal stability limits are determined by the superparamagnetic effects of the media. This occurs when the volume V, of the basic switching unit is not adequate to prevent thermal energy from demagnetizing the media. The limit is being approached because in order to maintain constant media signal to noise ratio as the bit size is reduced, it is necessary to make the media with even smaller and more isolated
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
975
Figure 14. Various lithographic processes required for reduced track width for GMR heads.
grains. Thermal decay and increase in the dynamic coercivity are correlated and inseparable. The instabilities can be improved by optimizing the grain size distribution and the anisotropy constants. Today, 5 Gb.in⫺2 recording heads are in large-scale production. In order to achieve 10 Gb.in⫺2 areal density, it is necessary to have thin film media grain size of roughly 10 nm, magnetic spacing of 30 nm, read gap of 150 nm, and track width of 450 nm. Evolution of grain size and coercivity of magnetic film on the disk with areal density is illustrated in Fig. 16. A system to manufacture 10 Gb.in⫺2 drives is feasible with today’s technology and will likely be in production by year 2000. 40 Gb.in⫺2 recording requires sub 5 nm thin film media grain size. Current media materials will not permit stability of the recorded dibit while maintaining sufficient signal to noise ratio. Micromagnetic and Monte Carlo analysis showed that transition from stable to unstable dibits occurs around 40 Gb.in⫺2 in the current head/media materials for longitudinal recording. 7. Reduced magnetic spacing with areal density The ever increasing demand of high areal density leads to a continuous reduction in the magnetic spacing between the head and media, Fig. 17. The areal density (product of bit and track density) of magnetic storage devices is limited by the head width, gap length, and the magnetic/mechanical spacing. The track density depends on the transducer width and bit density depends on the gap length and magnetic spacing. Data rate is determined by the bit density and the disc spinning speed. Loss in the readback amplitude due to magnetic spacing (d) is described by R.L. Wallace (13). Readback signal magnitude is given by the following expression and shown in Fig. 18.
Figure 15. Schematic of the focused ion beam (FIB) system used for trimming the writer pole for reduced track width.
976
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 16. Magnetic media parameters evolution with increasing areal density.
dB ⫽ ⫺
55d
The loss function is given in decibel form (dB) (20 times the log10 of the loss function). The spacing loss in a hard disc drive is controlled by reducing the magnetic spacing between the flying head and the spinning disc surface. Magnetic spacing is decreased by reducing the roughness of head and disc surfaces, thickness of carbon wear coatings on head and disc, thickness of the lubricant, and the mechanical spacing between the flying head and disc. Lower mechanical spacing results in ultralow flying of the head over the disc surface. The formation of air bearing between the head and disc takes place when the disc is spinning at full speed. The head-disc spacing budget for increasing areal density is illustrated in Fig. 19. The magnetic spacing for 15 Gb.in⫺2 areal density is in the sub 25 nanometer range. Note that for 15 Gb.in⫺2 areal density, the carbon overcoat thickness on the head and media is about 1 and 2 nm, respectively. The flying height is reduced from 40 nm to 15 nm as the areal density is increased from 1 to 10 Gb.in⫺2. To keep the contribution of head and disk surface roughness in the magnetic spacing budget to minimum, it is essential to achieve the rms roughness of the head and disk surfaces in the subnanometer range. For achieving rms roughness in the subnanometer range, the lapping technology needs dramatic improvements. The recession of pole-tips (magnetic transducer) has a detrimental effect on the electrical and tribology performance of the head-disc interface since it increases the magnetic spacing. The pole tip
Figure 17. Change in the magnetic spacing with increasing areal density.
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
977
Figure 18. Loss in readback amplitude with increasing magnetic spacing.
recession is defined as the height difference between the writer/shared pole and the air-bearing surface. Larger pole tip recession causes lower electrical readback signal. In order to get optimum electrical performance, the sliders are lapped/polished in a well-controlled way to obtain low PTR. PTR can be measured by stylus profiler (SP), AFM, and non-contact optical profiler (NOP). In the stylus profiler method, placement of cursors on the regions over the surface profile that correspond to the pole-tips and ABS is not always precise, and the variation in selecting the appropriate area on the profile leads to higher sigma of PTR values. In the AFM method, the PTR value is estimated from the topography image over the pole tips after removing the scanner bow by subtracting the pole tip image with that of the ABS image, Fig. 20. AFM method gives accurate values of PTR, if the AFM image is flattened and plane fitted appropriately. The optical profiler is extremely quick but suffers from an inaccuracy introduced by the difference in the optical properties of magnetic pole tip materials, alumina, and slider substrate material. The optical method estimates higher PTR as compared to that of AFM method. The offset among optical and AFM values become more significant when the slider are coated with DLC.
Figure 19. Magnetic separation budget with increasing areal density.
978
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 20. Estimation of pole tip recession (PTR) by using tapping mode AFM images.
The offset value depends on the variations in the optical constants of the pole tips and ABS regions and thickness of DLC coating. 8. Ultrathin carbon overcoats Ultrathin carbon wear coating are applied on head and media to protect the magnetic films against corrosion due to environmental exposure and against mechanical wear due to head dragging over the disc surface prior to take-off and after landing. Currently, a thin nanolayer (5–10 nm thick) of amorphous carbon (also called diamond-like-carbon (dlc)) is applied on the head and media surfaces. In the past few years, dramatic enhancement in tribology of head-disc interface is achieved by applying a thin layer of dlc on the head and disc. Dlc is chosen because of its chemical inertness to environmental exposure, ease of tailoring mechanical and electrical properties, and low cost (1). Since MR/GMR heads consist of multiple coatings of metals and dielectric materials, high adhesion of dlc on these materials is a primary concern. Any delamination or detachment of coating from any region on the head or disc surface can damage the head-disc interface. In the past two decades, a large number of carbon-based films have been deposited with varying hardness, elasticity, friction coefficient, thermal conductivity and stability, electrical resistivity, optical and electric bandgap by modified physical vapor deposition (PVD) and chemical vapor deposition (CVD) processes (1, 12). The sp3 bonded tetrahedric carbon configuration corresponds to a fully interconnected network with very strong C-C covalent bonds (7 eV), exhibits extreme high hardness (100 GPa), high elastic limits and yield stress of diamond. Sp2 bonding corresponds to tetragonal bonds (0.8 eV) and exhibits low hardness of graphite. Sp1 bonding corresponds to polymer like bonding. The sp3/sp2 ratio depends on the kinetic energy of carbon species during the deposition. Properties of the dlc coatings can be tailored to a specific need by varying the sp3/sp2 carbon-bonding ratio which primarily depends on the precursor starting materials, deposition method and conditions, and chemical composition of the dlc film. For enhanced wear resistance, coatings with higher sp3 bonding are preferred because of high hardness. However, there are some tradeoffs among various properties. For instance, extremely hard coatings carry high residual stresses. The relative concentrations of sp3, sp2, and sp1 bonding in the amorphous carbon films deposited by various techniques are compared in Fig. 21. PVD processes involve ion bombardment during film growth and form amorphous carbon films consist of sp3 and sp2 carbon states. The concentration of sp3-bonding depends on the energy and doses of ion bombardment and on the concentration of activated and dissociated carbon atoms that contribute to the film growth. The schematics of direct ion beam and
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
979
Figure 21. Relative concentrations of sp3, sp2, and sp1 bonding in the amorphous carbon films deposited by various techniques.
cathodic arc deposition setups are shown in Fig. 22. Direct ion beam and cathodic arc deposition techniques involve higher kinetic energy (⬃100 eV) and concentration of activated carbon species as compared to those of sputtering or magnetron sputtering (⬃1–10 eV). Therefore, carbon coatings deposited by sputtering exhibit lower hardness (⬃10 GPa) as compared to the coatings deposited by ion beam (⬃25 GPa) and cathodic arc (⬃30 –70 GPa). Laser ablation can also produce high sp3 carbon if the laser pulse energy is high enough and sufficiently focused. The laser pulse form a dense microplasma rich in dissociated, ionized, and activated carbon species, which can result in carbon coatings with more than 50% sp3 content. Currently, this technique suffers from deposition uniformity over irregular surfaces. Figure 23 compares the approximate hardness of amorphous carbon coatings deposited by various techniques with and without hydrogen and varying amount sp3 bonded carbon (12). Films with tetrahedrally bonded carbon exhibit high hardness of about 90 GPa very close to that
Figure 22. Schematics of direct ion beam deposition and filtered cathodic arc deposition systems.
980
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 23. Comparison of hardness of amorphous carbon films deposited by various techniques.
of single and polycrystalline diamond. In general, higher the amount sp3 bonded carbon higher is the hardness and wear resistance, but this is not always true. 9. Tribology and characterization at the nanometer dimensions The reduced flying height necessitates increasingly smoother surfaces for the head and the disk. The smooth surfaces however, give rise to high stiction during the contact start-stop (CSS) operation of the drive. An overcoat of minimum thickness is needed to achieve high readback signal amplitude without sacrificing the friction and wear properties of the head-disk interface. Characterization of the head and disk surfaces at nanometer dimensions is crucial for optimizing the tribological performance (7). Mechanical properties of the carbon overcoats on the head and disk surfaces influence the friction and wear performance of the head-disk interface. Among the mechanical properties of interest are elastic-plastic deformation behavior, hardness, Young’s modulus of elasticity, scratch resistance, overcoat-substrate adhesion, residual stresses, fracture toughness, and fatigue. In the lightly loaded head-disc interface, the friction and wear behavior of sliding surfaces is primarily controlled by the physical and chemical properties of a few surface atomic layers. For understanding and/or estimating the functional behavior of the head-disc interface involving low loads of a couple of tens of milliNewtons, measurements of the mechanical and tribological properties of the thin carbon overcoats at a nano-scale are of crucial importance. In the last decade, a variety of instruments have been developed to measure the mechanical and tribological properties on macro- to nano-scales [Kaneko et al., 1995]. Some of the common instruments are: depth-sensing mechanical properties microscope (nanoindenter), atomic force microscope (AFM), and friction force microscope (FFM). These instruments operate at very low loads and measure properties of the near surface region. In this section, techniques for the characterization of the head-disc interface and DLC overcoats at nanometer dimensions are discussed. Friction mapping at a nano-scale is used to study the spatial variation in the surface chemistry due to formation of thin smears on the disk. To understand the mechanical performance of the substrates, nanoindentations are made on the two-phase Al2O3-TiC substrate. Nanoscratches are made to characterize 5 nm and 7.5 nm thick DLC coatings deposited on the air-bearing surface of the MR head. Nanowear data are correlated with the CSS performance of the DLC coated sliders.
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
981
Figure 24. Topography and friction images of a disk from a failed drive exhibiting high nonrecoverable error rate at outer diameter.
Friction force mapping on nano-scale using friction force microscopy (FFM) provides an in-depth understanding of the stiction behavior of the head-disc interface, lubricant uniformity, presence of contamination, and any spatial variation in the chemistry of the head and disc surfaces. A few monolayers on the head or disc surface are sufficient to change the friction characteristics between the FFM tip and the sample surface. In FFM, the tip is raster scanned in the contact mode over the sample surface and the signal due to torsion or twisting of the cantilever is collected to generate the friction images. The torsion or twisting of the cantilever supporting the tip will increase or decrease depending on the frictional characteristics of the surface (greater twist results from increased friction). Mate [1993] used FFM to study the effect of 2.5 nm thick PFPE lubricant layer on the friction of the DLC coated disc at increasing load. Shown in Fig. 24 are the topography and friction images of a disc from a failed drive exhibiting high nonrecoverable error rate at the outer diameter (OD). During the drive usage, smears are formed on the disk surface at the OD. The smears exhibit lower friction as compared to that of the disk surface. The topography image shows the presence of a thick patch only at one location on the disk surface as a result of the smear formation, while the friction image shows a couple of more tiny patches in addition to the thick patch, shown by arrows in Fig. 24. These tiny patches are significantly thick and apparently contributing towards the additional nonrecoverable errors at the OD. It should be noted that the tiny patches that are not high enough to change the disk surface topography, can be clearly seen in the friction images. On a microscopic scale, the phenomenon of boundary lubrication has been studied using FFM by Meyer et al. [1995]. In the present study, the uniformity of the lubricant layer on a virgin disc is studied by the friction imaging. The topography and friction images of a virgin lubricated disk are shown in Fig. 25. The topography image on the left does not show any nonuniformity in the lubricant layer on the disk surface, while the friction image shows a discontinuity, indicated by an arrow in the right image. This
Figure 25. Topography and friction images of a virgin lubricated disk.
982
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 26. Inverted AFM image (3 m ⫻ 3 m) and load-displacement plots of nanoindentations made at 1500 N on TiC grains of Al2O3-TiC substrate.
example illustrates that friction imaging is indeed a very powerful tool to study the spatial variation in the chemistry due to lack of lubricant on the disc surface. The nanoindentation technique provides an understanding of the deformation behavior of materials in the near surface region at a nano-scale. The elastic-plastic deformation behavior of contacting materials can be correlated to the wear behavior of the head-disc interface, burnishing of the disc
Figure 27. (a) The AFM image of a nanoscratch made at increasing load up to 20 N on the ABS of the 7.5 nm thick dlc coated slider. (b) Lateral force and tip penetration deoth profiles (four profiles) versus normal load during nanoscratching at increasing load.
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
983
Figure 28. AFM images of nanoscratches made at constant loads of 20 N on 7.5 nm thick DLC coated sliders. The DLC coatings were deposited on the Al2O3-TiC substrate at 50 and 175 V using a direct ion beam deposition.
asperities during initial dragging of the slider on the disk, grain pullout due to heavy stiction forces, and catastrophic wear at the head-disk interface. Shown in Fig. 26 are an inverted AFM image and load-displacement plots of nanoindentations made on Al2O3-TiC head substrate using a cube corner diamond tip. Indentations are made at 1500 N loads on TiC grains, grain boundaries, and alumina grains. Insignificant difference in the load-displacement plots indicates that the depth of indentations over different regions do not vary significantly. The macrohardness of TiC and Al2O3 grains is 24 GPa and 18 GPa, respectively. Nanoindentation data suggest that the hardness of the alumina grains is comparable to that of the TiC grains. This is not as expected because TiC grains are expected to exhibit higher hardness. This result can be explained by the fact that there is not enough room for alumina to deform because the indentation region is surrounded by the hard TiC grains. Hardness measurements become increasingly difficult when the indentation depths are on the order of a few nanometers. The tip area function that is used to estimate the hardness and elastic modulus, becomes very sensitive to indentation depth at shallower depths. Any inaccuracy in the area calibration function results in either an increase or a decrease in the hardness value. Thus, an accurate measurement of hardness of 5 nm thick carbon overcoats on heads or disks is almost impossible. In addition to the tip area function inaccuracy, relative contribution of the substrate to the indentation deformation of thin carbon overcoat adds further inaccuracy in the hardness measurements. In nanoscratch measurements, a sharp diamond tip (10 –50 nm tip diameter) is dragged over the sample surface at constant load or increasing load and the lateral force (friction force) is recorded. Weak coatings are delaminated from the substrates and/or cracked during scratching due to poor adhesion or low fracture toughness. Relatively softer coatings exhibit wide and deep scratches and scooping of the material to one end
Figure 29. AFm images (2 ⫻ 2 m) of nanowear patterns (1 ⫻ 1 m) formed at 20 N on 7.5 nm thick dlc coatings deposited at 50 and 175 V using a direct ion beam deposition technique.
984
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
Figure 30. Average depth of wear patterns formed at 20 N on DLC coated and uncoated sliders. The DLC coatings are deposited at 175 V (Coating A), 100 V (Coating B), and 50 V (Coating C) using a direct ion beam deposition technique. (b) Stiction and dynamic friction of dlc coated and uncoated sliders.
during scratching. Harder coatings exhibit low friction at the initial stages and an abrupt increase in friction when the normal load exceeds a critical limit known as critical load. The AFM image of a nanoscratch made at increasing load on the air-bearing surface of the 7.5 nm thick DLC coated slider
Vol. 11, No. 8
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
985
and lateral force and tip penetration depth profile versus normal load plots are shown in Fig. 27. The normal load corresponding to an abrupt increase in the friction force and tip depth is defined as the critical load. When the normal load exceeds the critical load, the tip penetrates through the coating. The critical load is used to compare the mechanical integrity of the films; higher critical load corresponds to a relatively more mechanically robust coating. The AFM images of nanoscratches made at 20 N on 7.5 nm thick DLC coatings deposited at 50 and 175 V using a direct ion beam deposition technique are compared in Fig. 28. We note that 50 V DLC coating is delaminated from the substrate during scratching. It is obvious that 50 V DLC exhibits significantly weaker adhesion as compared to that of 175 V DLC coating. Higher kinetic energy of carbon ions at 175 V is responsible for enhanced adhesion of the DLC coating. In general, the wear behavior of the contacting surfaces is primarily influenced by the normal and shearing loads at the contact points. In most of the cases, higher the load and speed, higher the wear volume. The severity of stresses at the contact points and the physical and mechanical properties of the contacting surfaces determine the wear mechanisms under a specific set of operating conditions. In order to understand the wear mechanisms of very lightly loaded interfaces, such as head-disc interface, it is critical to experimentally study the wear mechanisms of a single asperity contact at nanometer dimensions. In the past few years, STM/AFM/FFM techniques have been used to understand the wear phenomena at an atomic level. With these sophisticated tools, it is possible to study the evolution of wear under a typical set of operating conditions. Kaneko et al. [1995] published a couple of papers on the wear studies of the DLC films and ion-implanted Silicon. In the present study, the contact start-stop (CSS) wear performance of the DLC coated sliders is correlated to the nanowear behavior. The DLC coatings are deposited at 50, 100, and 175 V using a direct ion beam technique on the ABS of the MR heads. The nanowear patterns are formed by sliding a cube corner diamond tip over the dlc coated sliders over two passes at 20 N. The CSS tests are conducted in a clean environment of class 100 at an ambient temperature of 21 °C and a relative humidity of 44%. The AFM images of the wear patterns formed on 7.5 nm thick DLC coatings at 20 N after sliding over two passes are compared in Fig. 29. We note that the DLC coating deposited at 175 V exhibits significantly lower wear as compared to that of the 50 V dlc coating. The nanowear depths of dlc coatings deposited at 175 V (coating A), 100 V (coating B), and 50 V (coating C) and uncoated Al2O3-TiC substrate are compared in Fig. 30 (a). We note that coating A exhibits lower wear depth when compared to coating B followed by coating C. As expected, the wear depth in the uncoated slider is significantly lower because of the higher hardness of the Al2O3-TiC substrate. The stiction during the first one-second of a CSS cycle and dynamic friction of various DLC coated sliders and uncoated sliders in ambient environment are compared in Fig. 30 (b). We note that the sliders coated with the DLC coating A exhibits the lowest stiction and friction when compared to the sliders with other dlc coatings. In addition, Head-disk interface with coating A on the head exhibits least amount of wear debris formation during 20,000 CSS cycles. The correlation between the highest nanowear resistance and best CSS performance (low stiction and low wear) of DLC coatings indicates that the wear properties at a nanoscale can be used as an indirect measure of the CSS wear performance of the carbon overcoats. 10. Closure Areal density of magnetic storage devices is expected to increase at 60% annual growth rate over next five years. Recording density is expected to increase to 20 GB.in⫺2 in production by year 2002. Advances in nanotechnology with regard to fabrication, characterization, and tribology are essential for sustaining 60% growth rate. Thickness and composition control of a couple of atomic layers thick films in the GMR structure plays a key role in optimizing the GMR coefficient (␦R/Ro). The reduced track
986
NANOTECHNOLOGY: A DATA STORAGE PERSPECTIVE
Vol. 11, No. 8
width (200 nm for 40 GB.in⫺2) poses new challenges for photolithography and patterning sequence. The superparamagnetic limit at 40 GB.in⫺2 requires media magnetic film material for high coercivity with the grain size in the sub five-nanometer range. Reduced magnetic spacing (15 nm at 10 GB.in⫺2) leads to ultra thin carbon overcoats, smooth head and disk surfaces, and low flying height. Characterization at the nanometer dimensions is essential to understand the functional CSS performance of the head-disk interface. Friction mapping on the nano-scale can be used to check the uniformity of the lubricant layer and the formation of thin smears of sub-nanometer thickness on the head and disk surfaces. Nanoindentation technique is used to measure the indentation hardness of the TiC and alumina grains. The mechanical integrity of ultra thin DLC is quantitatively evaluated using nanoscratches in terms of the critical load. In addition, AFM imaging of nanoscratches shed some light on the type of physical damage and adhesion failure of the coating. Nanowear simulates a single asperity contact situation and provides information on the evolution of the wear process. The DLC coating exhibiting the lowest nanowear rate performs best in the CSS test. Based on this data, it appears that nanowear measurements on DLC overcoats can be used as an indirect measure of their CSS performance. References 1. 2. 3 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.
B. Bhushan and B. K. Gupta, Handbook of Tribology: Materials, Coatings, and Surface Treatments, McGraw Hill, New York (1991). W. F. Egelhoff, P. J. Chen, Jr., C. J. Powell, M. D. Stiles, R. D. McMichael, C-L. Lin, J. M. Sivertsen, J. H. Judy, K. Takano, A. E. Berkowitz, T. C. Anthony, and J. A. Brug, J. Appl. Phys. 79.8, 5277 (1996). S. Esener, Long Range Alternatives, Proceedings on Workshop on the Future of Data Storage Technologies, International Technology Research Institute, Baltimore (1998). E. Grochowski and D. A. Thompson, IEEE Trans. Mag. 30.6, 3797 (1994). E. Grochowski and H. G. Belleson, MR Head Technology Evolves into Next Generation GMR/Spin Valve Heads, Notes from IBM web site (1997). B. K. Gupta and B. Bhushan, Wear. 190, 110 (1995). B. K. Gupta and A. K. Menon, IEEE Trans. Mag. 35 (1999) R. Kaneko, T. Miyamoto, and E. Hamada, in Micro/Nano Tribology, ed. B. Bhushan, pp. 183–222, CRC Press, Boca Raton, FL (1995). M. T. Kief, J. Bresowar, Q. Leng, J. Appl. Phys. 79.8, 4766 (1996). M. Mate, Surf. Coat. Technol. 62, 373 (1995). E. Meyer, R. M. Overney, and J. Frommer, in Micro/Nano Tribology, ed. B. Bhushan, pp. 223–242, CRC Press, Boca Raton, FL (1995). S. Neuville and A. Matthews, 22(9), 22 (1997). R. L. Wallace, Bell. System. Tech. J. 30, 1145 (1951).