Fab-wide process control needed for reduced disk drive magnetic spacing
09/01/2007
Even as areal data densities grow toward 350 Gb/in2 by 2009, the magnetic hard-disk drive remains predominantly a mechanical device. The slider (or thin-film magnetic head) assembly rides on an air bearing within a few nanometers of a disk of magnetic media spinning at rates between 7200 and 15,000rpm. The magnetic spacing between the slider and the media is one of the key contributors to drive performance. Success demands a balance between the mechanical requirement for reliable flying heights and the electrical performance of close proximity writing of the magnetic data bit. To achieve 350 Gb/in2 data density, a total magnetic spacing of just 12nm is required.
Magnetic spacing-the distance between the magnetic pole in the head and magnetic storage layers in the media-is always larger than the physical spacing between the lowest surface of the head and the highest surface of the media. Intervening overcoat and lubricant layers, deviations from parallelism of the slider body, recession of the pole tip from the air bearing surface, and surface roughness are the primary factors that contribute to the magnetic separation. Thus, even though shrinking the air gap between the head and the media may not be possible, manufacturers can reduce the total magnetic spacing by addressing other factors that compose it.
Mechanical processes define slider shape
Slider fabrication includes both mechanical steps (e.g. slicing, lapping, and dicing) and vacuum processes like deposition and etch (Fig. 1). All of these contribute to the total budget for magnetic spacing, and all require tight process control. Moreover, variations in one step can easily lead to additional variation in the next.
Slicing, one of the first steps in the process sequence, cuts thin rowbars from the thin-film head wafer. Increasing the available rowbars requires reducing the kerf widths using a thinner blade, which tends to deflect as it slices through the wafer. The angle created by the deflection will affect the shape of all sliders cut from that row, and potentially the adjoining rows as well. To circumvent this problem, Veeco’s ADS160 saw performs a backside edge grinding step in conjunction with the sawing (slicing) operation to achieve a cut edge perpendicularity error of <0.2° (1σ). This results in a perpendicular reference plane for all further processing.
Having a stable reference surface is especially important for the next step: lapping of individual row bars. The lapping steps largely define the performance of the head because they determine such parameters as pole tip recession (PTR), stripe height, throat height, and the crown, camber, and twist of the air bearing surface (ABS). The final shape is typically achieved through a sequence of lapping steps involving dual-sided lapping to flatten the rowbar and remove residual stresses, rough lapping to bring the stripe and throat height within range, fine lapping to converge to the desired throat and stripe heights, and kiss lapping to achieve the final ABS shape, surface roughness, and PTR.
 Figure 1. Slider fabrication includes both mechanical steps (e.g. slicing, lapping, and dicing) and vacuum processes like deposition and etch. |
Of these, fine and kiss lapping are the most critical since they define the final metrics of interest. Achieving the desired precision is especially difficult because the slider is a metal-ceramic sandwich, combining alumina with titanium carbide. These materials polish at different rates, adding surface roughness and additional variability. The latest generation lapping tools include the ability to tilt the polishing angle to square off any tilt remaining from the slicing step, and to fine tune the lapping rate at each location along the rowbar through multipoint bending of the rowbar during lapping. Low-pressure lapping capability ensures a low PTR. With these features, a lapping system can achieve stripe and throat height control of ~5nm with sub-nm control of PTR and reader-writer offset. After lapping, the air bearing surface should be perpendicular to the magnetic pole tip in order to minimize the reader-writer offset (Fig. 2).
Slider coating
After lapping, the slider receives a protective diamond-like carbon (DLC) coating. The DLC deposition brings several additional sources of variability, beginning with the ion beam pre-cleaning etch. This step removes any smearing or particles left behind during lapping that would otherwise degrade the electrical performance of the sensor or form corrosion failure points if trapped under the DLC coat. However, conventional ion beam etching is too aggressive; this step requires an ultra-low energy and uniform ion beam. A low-energy source can reduce PTR from 1.5nm to <1nm with minimal surface damage.
 Figure 2. Lapping determines the pole tip and sensor dimensions. |
The DLC deposition is challenging because it requires a very uniform and pinhole-free layer that is also extremely thin. Uniformity is easier to achieve in thicker films, where statistical fluctuations are small compared to film thickness. Yet the DLC layer lies directly between the head’s magnetic components and the media. DLC thickness directly increases magnetic spacing. A typical DLC film contains 10Å of silicon and 10Å of carbon. Manufacturers would like to achieve a total thickness of 13Å, while cutting variability. Use of a dual-source pulsed filtered cathodic arc (FCA) deposition system achieved both goals.
Compared to a more conventional DC FCA deposition, pulsed deposition (PFCA DLC) increases film density, while focusing coils help improve plume and resultant deposition uniformity. These improvements reduced film thickness while cutting within-wafer nonuniformity from about 2.5Å to 1.5Å, and run-to-run variation from about 4.5Å to 2.5Å. In addition, the total film thickness can be scaled due to the superior film quality of PFCA DLC. In combination, these improvements translate to an effective reduction of 15Å in magnetic spacing.
Low energy ion cleaning and pulsed FCA deposition are combined into a two-chamber cluster tool. The tool maintains a clean environment between processes to avoid cross-contamination.
Fly-height control
Since physical spacing is one of the biggest contributors to magnetic spacing, fly-height control is equally important. While all leading-edge drive technology today utilizes thermal fly-height control, the importance of managing the slider attributes that set the fly-height cannot be overlooked. In particular, fly-height control demands precise control of the depth and profile of the ABS cavity. Roughly speaking, this surface confines the cushion of air on which the slider rests. A good air bearing maintains a small and stable spacing, with low sensitivity to variations in disk velocity or slider attitude. A rough ABS can create turbulence and cause the head to “bounce.” Modern slider designs incorporate multiple etch depths, ranging from 20nm to 2µm or more (see Fig. 3). Shallow areas require precise control of etch depth, while deeper areas require high throughput etching to keep costs low. For all areas, control of surface quality and topography ensures a smooth “ride.” Ion beam etching to shape the ABS is the next step after DLC. Large diameter ion sources offer excellent etch uniformity and repeatability over the typical 9-10 in. pallet.
 Figure 3. Air bearing etch depths range from 20nm for pad areas, to 1-2µm or more in the negative pressure cavity. |
Up until this point, the process works with an entire row of sliders as one unit. Precise metrology and careful process control are required, but can easily be destroyed at the dicing step. The hard dicing blade can create pressure ridges and sidewall debris as it cuts into the ABS. One solution, dual-pass dicing, follows the rough cut with a second, softer blade to minimize chipping and smooth out pressure ridges.
Tight specs demand superior metrology
While metrology systems are traditionally evaluated in terms of repeatability, verifying that such tight specifications have actually been achieved depends on extremely accurate metrology. Tighter specifications also tend to be more difficult to maintain, requiring more frequent process monitoring. Metrology suppliers must deliver better measurements while reducing total cost of ownership driven by capital costs, equipment lifetime, and throughput.
Neither atomic force microscopes (AFM) nor optical profiler measurements individually match all needs of slider metrology. AFM measurements, which are topographical mappings, are accurate, but time consuming, requiring minutes for a single slider profile. AFM tips, which provide the physical profiling, are consumable items, with their variable life depending upon the surface and scan conditions. Tip replacement and qualification adds unproductive time to AFM measurements. One step toward more cost-effective AFM measurements might be intelligent scanning. Rather than doing a time-consuming full surface scan and extracting key process parameters from the data set, the system can perform fast survey scans to identify the structure alignment in order to conduct detailed scans of points of interest. By using intelligent scanning, improving tip lifetime, and automating tip replacement, AFM throughput has been increased by 55% in the past year (Fig. 4).
 Figure 4. Intelligent scanning boosts AFM throughput. (Source: UCSD-CMRR) |
One of the AFM’s primary uses is, and is likely to remain, as a calibration standard for optical metrology. While suppliers seek to reduce AFM measurement times to sub-minute times per slider, current optical measurements typically take ≤6 sec for PTR measurements.
Optical technology now must contend with structures and materials that historically introduced only small variation into the process monitor. Today, magnetic sliders with multiple optically dissimilar materials, DLC coating, and reduced structural sizes are extremely challenging for optical profiles. Overcoming these challenges requires systems that are more insensitive to environmental noise, offer improved optical signal/noise ratios, and enhance the data analysis and extraction.
To achieve more accurate measurements, the optical system can now compensate for n, k, and DLC thickness variations over the pole and shield structures and adjust for the optically dissimilar materials now visible in the AlTiC substrate. Improved accuracy, however, is worth the effort. Since the measured distribution is a combination of process variation and metrology errors, reducing the metrology error improves the process control capabilities and sensitivity, leading to improved yields and profitability.
Conclusion
This article reviewed several ways that improved process and metrology tools can reduce variation and improve the absolute value of the magnetic spacing in thin-film heads. Yet tighter tolerances also require careful attention to parameters that are outside the control of any equipment supplier. When nanometers matter, any source of noise can degrade the results. Inadequate device fixturing and poor sample handling can introduce alignment and orientation errors. Mechanical vibration adds variability to processing and noise to measurements. When nanometers matter, the best processing and metrology tools are only as good as the facilities housing them and the users sitting at the controls.
Ajit Paranjpe is VP of technology, Process Equipment Group, at Veeco Instruments Inc., 3100 Laurelview Ct., Fremont, CA 94538 United States; ph 510/770-9200 ext. 2005, fax 510/770-9320, e-mail [email protected].
Michael Peters is product manager of data storage metrology at Veeco Instruments Inc., 2650 E. Elvira Road, Tucson, AZ 85706 United States; ph 520/741-1044, fax 520/294-1799, e-mail [email protected].