Thin-film head lithography overtaking silicon
09/01/1999
Michael R. Biche, FSI International, Fremont, California
Historically, thin-film head (TFH) lithography resolution has lagged that of silicon, but there have always been special challenges to patterning TFH structures. Today, however, TFH resolution requirements are changing at nearly twice the rate of mainstream silicon, as shown in Fig. 1. At this rate of change, by about 2005, lithography resolution for TFHs will match those of silicon, but in much different circumstances.
|
Figure 1. Critical dimension trends for silicon IC and thin-film head production. (Source: SIA and Ultratech Stepper)
TFH lithography challenges
TFH lithography faces challenges that are not encountered in silicon lithography, including: highly nonplanar surfaces, high aspect ratios, topology larger than depth of focus, square substrates, and a 6- to 9-month product lifetime.
Pressure to continue the historical 60%/yr increase in areal recording density (i.e., the "Moore's Law" of TFHs) drives the TFH industry. Higher areal recording density requires narrower heads with smaller gaps. Modern head designs separate the read and write functions, which must be accurately aligned. The write portion of these heads continues to be inductive, while the read portion uses magnetoresistive (MR) and giant magnetoresistive (GMR) technologies.
Lithographically patterned thin-film magnetic read-write sensors have evolved from: thin-film inductive heads with NiFe yoke geometries controlled by semiconductor processing; to thin-film inductive-write/MR-read heads with NiFe MR films; to currently, thin-film inductive-write/GMR-read heads with antiferromagnetic films. As the areal recording density has increased, TFH CDs have decreased (see table). Current production heads have 1.2µm CDs and are capable of 5.0Gbits/in2 areal recording density. Within the next 12-18 months, two more generations of heads will reach production. With these, CDs will drop to 0.8µm and 0.5µm with areal recording densities of 6.2Gbits/in2 and 8.2Gbits/in2, respectively. The large and increasing aspect ratios imply much different resist processing requirements than for silicon at the same dimensions.
The topology of MR heads does not change appreciably as the CDs shrink. TFHs contain relatively thick structures surrounded by large areas with very thin structures. The thick structures typically project 8-10µm above the thin structures. Accordingly, photoresist coatings must be relatively thick (5-8µm) over the thinner structures to adequately cover the thicker structures.
Depth of focus is quickly becoming a major issue when small features must be printed in the thick photoresist covering the thinner structures. While aspect ratios for silicon rarely exceed 4:1, those for TFHs are rapidly approaching 10:1 (see table). The industry is exploring such options as design changes, chemical mechanical planarization, and multiple exposure imaging.
Square substrate issues
TFHs are fabricated on aluminum titanium carbide substrates. Many manufacturers produce heads on 4-6 in. square substrates. This simplifies the final mechanical processing steps of dicing, lapping, and creating air-bearing surfaces that are unique to TFH production. However, with lithography requirements getting more stringent, uniformly coating square substrates has become an issue. The corners of square substrates cause severe airflow turbulence in typical resist coater bowls. This in turn reduces coating uniformity, an effect especially pronounced in the corners.
|
Several techniques have been developed for coaters to minimize this effect. One, for example, uses a special coater chuck with a recess for the substrate ("nested chuck"). The recess allows the substrate to lie inside the chuck so that the top surface of the substrate is flush with the top surface of the chuck. The outer profile of the chuck is round with a diameter larger than the diagonal of the substrate. This design minimizes turbulence by presenting a nominally round surface to the airflow. Although considerably reduced, the airflow is still affected by small gaps between the chamfered edges of the substrate and the nest.
Another technique positions a flat barrier plate in close proximity over the substrate. The barrier plate produces two effects: it blocks the main air flow, thereby reducing turbulence, and it traps a solvent-rich layer of air between the substrate and the barrier plate. This solvent-rich layer allows the resist to spread to the corners of the substrate without prematurely drying. This is especially important when coating thick layers of photoresist with high solids content.
Two variations on the barrier plate concept have also been developed. The first puts the substrate and chuck into an enclosure that spins with the substrate. This results in solid-body rotation of the air and substrate, thereby eliminating turbulence. However, photoresist build-up on the inside of the spinning enclosure is a significant drawback to this technique.
Another variation on the barrier plate concept uses a barrier with a lower surface that, instead of being flat, is contoured to provide optimum air flow and solvent retention. This approach improves on the flat barrier plate, while retaining its ease of use.
Even with these advances, square substrates are nearing the end of their useable life. A change to round substrates will be required to maintain acceptable yields as the industry moves to deep-UV processes. Some manufacturers are already processing round substrates, a few are in transition, and others are delaying the transition and associated expense as long as possible.
|
Figure 2. Projected IBM MR and GMR spin valve head evolution. (Source: E. Grochowski, IBM Almaden Research Center)
A typical TFH generation has a production life of only 6-9 months (Fig. 2). As a result, manufacturers must be developing processes for the next two to three product generations. Considering accelerated lithography requirements, TFH manufacturers working with i-line exposure systems must also be developing their deep-UV lithography capabilities. It will not be long before they move to even smaller features and advanced techniques.
Lithography resolution for TFH manufacturing continues to approach that of mainstream silicon. This, along with challenges specific to head production, is putting TFH lithography at the leading edge of technology.
Acknowledgment
The author thanks Edward Grochowski, program manager of storage devices at IBM Almaden Research Center, San Jose, CA, for use of Fig. 2.
Author
Michael R. Biché received his BSME from Rensselaer Polytechnic Institute. Biché is director of engineering for the Fremont operations of FSI International, 47003 Mission Falls Ct., Fremont, CA 94539; ph 510/683-8858, fax 510/656-3226.