by M. David Levenson, editor-in-chief, Microlithography World
Sept. 24, 2008 – The hard disk drive (HDD) industry has long been increasing the aerial density of data storage faster than Moore’s Law for semiconductors. At this year’s Diskcon USA (Sept. 17-18, Santa Clara, CA), a full day symposium explored the lithography implications of sub-ITRS roadmap feature sizes on disk drives. According to kickoff speaker Paul Hofemann, VP of HDD business development at Molecular Imprints Inc., the crossover is happening now, in 4Q08. The challenge is that lithography for hard disk drives must cost 10× less than NAND flash lithography, the lowest-cost semiconductor process.
Current technology based on continuous magnetic media requires data bits to be 75nm wide and 20nm long, reported Robert Fontana of IBM. The 75nm width corresponds to the track and is defined lithographically on the write head (the read head is smaller). The sub-lithographic 20nm bit length is defined by the data rate and linear velocity as the data pattern self-assembles. For the same lithography resolution, the density of magnetic data is 8× that of flash memory. The problem is that the numerous tiny magnetic domains that comprise the bits in continuous media are becoming unstable. Increasing the density beyond 1Terabit/in2 requires using larger domains with one domain/bit, and that means going to patterned disk media.
Neal Robertson of Hitachi Global Storage Technology outlined the patterned media challenges: Each disc will contain 1013 bits, patterned on both sides. The industry must manufacture 109 disks/year (beginning now) and do it at $5/disk. At 1Tb/in2 each data bit will be 25nm in width and length, with 3σ=3nm, if the bit is square and the track pitch 55nm or so. Wider tracks would mean shorter bits, challenging lithography even further. While the density of the data storage region is difficult, hard disks also contain servo sector patterns (Fig.1) with lower density and other features, all of which must be patterned lithographically and transferred using double-sided processes.
Textured disk with servo sectors. (Figure courtesy of Paul Hofemann, MII)
While conventional dry lithographic techniques (with some tricks) seem adequate for fabricating future read-write heads, according to Peter ten Berge of ASML, the media itself will require quite a lot new. The leading candidate for printing patterned media is imprint lithography, but just making the masters for the imprint tools will be a challenge.
Babak Heidari of Obducat described some manufacturing challenges for textured media. Obducat had developed a rotary stage e-beam master writer, but found that the patterned resist film had to be replicated by electroplating nickel as a “father stamp.” The father was then replicated again 25× as “mother stamps” which then printed the polymer working stamps that finally transferred the pattern to the substrate.
A different paradigm, described by Dwayne LaBrake of Molecular Imprints, involves working templates patterned by MII’s SFIL technology from circular masters supplied by Hoya-Elionix. The patterned media discs were then printed using those working templates using MII’s Imprio HD2200 tool. LaBrake reported a TAC time of 20s when proprietary surface treatments were employed on substrate and template.
(Figure courtesy of Paul Hofemann, MII)
MII’s Hofemann had previously shown SFIL results for bit patterned media with 25nm resolution as well as smaller linear patterns for data tracks and complex servo regions, demonstrating most of the capability needed (Fig.2). However, he lamented the complexity of supplying technology to seven data storage media companies, each with its own format (Fig.3). He encouraged industry standardization and cooperation to facilitate the transition to patterned media. Among the opportunities he mentioned:
– Non-proprietary template pattern for process integration and tool development;
– Standardized substrate handling (i.e. automation protocols & cassettes);
– Yield management & defect detection; and
– Process tool specifications (perfection, he noted, is expensive).
Hofemann predicted a busy year on the pilot lines in 2009, with first-generation process tools being debugged and new skills being learned by the hard disk drive industry. By 2013, he predicted that 70% of the disc drives shipped would be using patterned media. — M.D.L.
(Figure courtesy of Paul Hofemann, MII)