The Uncertain Future Of Hard Drives

The future of hard drives has been an iffy subject for nearly a decade. The most recent pressure has come from solid-state drives made from flash memory, which have transformed cell phones and tablets and other portable appliances. But the Internet has driven a staggering demand for data storage and the hard drive business has thrived.

The question of what’s next for hard drives first appeared around 2003, when they were doubling in density every year. In fact, they were projected to hit a hard thermodynamic limit in 2006 for Perpendicular Magnetic Recording (PMR) at 1 terabit per square inch. Since 2003, the growth in drive density has progressively slowed, so that 11 years later, the hard limit is projected for 2016, which is a 10-year delay.

There are two forces at work here. First, getting to 1 Tbit/in² for PMR and the replacement solutions for higher density has proven very challenging. Second, market demand for increased storage in PC hard drives has slowed. Today, storage density growth has slowed (25% CAGR) so much that it actually is less than the world demand for storage (40% CAGR), so the unit count has been growing, according to Seagate and Coughlin Associates. This is good because if the unit count grows, then so does the industry. The slowing of density growth may well be a function of the collective survival instinct of the leaders.

Delays in getting to 1 Tb/in². Based on trends in areal density, published by WD in 2012

For the last 11 years, there have been two possible solutions to density greater then 1 Tbit/in²—Heat Assist Magnetic Recording (HAMR) and Bit Patterned Media (BPM). The limit of 1 Tbit/in² is caused by the written magnetic domain becoming so small that thermal energy can change its state.

HARM does what the name implies. It increases the local temperature of the bit during writing. After writing, the cooled bit is a stable permanent storage element. HAMR requires a new media material, but most of the increase in device complexity is limited to the read/write head. It is not a simple problem. The bit is heated by a laser spot that leads the write head. The optical design takes advantage of the very low and well-controlled (1 to 2nm) flying height of the read-write head. At these distances, a near field optical transducer can be used to create a sub-wavelength spot similar in size to a single bit. The near field transducer consists of a metal foil with a hole; the optimal shape of the hole depends on the geometry and properties of the foil and the media. Finally, the designers have to worry about both the heating of the bit and heat dissipation across neighboring bits, and the reliability consequences of thermal cycling on media and head.

In 2012, both Seagate and TDK announced they had achieved 1 Tbit/in² or greater densities in HAMR. The projected longevity of HAMR has also increased. In 2006, the consensus view was that HAMR would enable a further doubling of density. A 2013 paper from Seagate proposes 5 Tb/in² as a limit, while Western Digital’s vice president of technology suggests 4 Tbit/in².

Bit-patterned media (BPM) is the competing technology that also does what it is says. Rather than relying on multiple naturally occurring small magnetic domains to form a bit, a larger single crystal bit is created by lithography. The additional complexity is that the media must be patterned with very small features (2 Tbit/in² = 10nm wide bits). Molecular Imprints has developed a dedicated imprint system and process that delivers 15nm pitch patterning for less than $1 a platter, as compared to $10 for a typical single-level semiconductor wafer process.

Making a full-size master pattern is still one of the gating challenges. The feature count and resolution are significantly more challenging than state of the art semiconductors. Furthermore, to maximize the volume of magnetic material the bits need to be slightly curved rectangles. The solution has turned into a tour de force of bleeding-edge lithography tricks. You start by writing a pattern of rings, with a separate pattern of radial spokes that takes weeks on a custom electron beam machine. The electron beam machine has a unique rotating stage so that positioning tolerances can be met that enable the read-write head to follow the tracks. The rings and spokes are then at least quadrupled using Directed Self-Assembly and a custom block copolymer. The rings and spoke patterns are then imprinted together to form curved rectangles. Now a first master has been created. This is then copied multiple times to create working imprint molds. The good news is that only one original master is needed. Today, making a full-sized original master still appears to be the biggest barrier to BPM.

In 2010, Toshiba announced it was working on a read-write system for 2.5 Tb/in² with a 17nm pitch. HGST also continues to publish updates on its BPM progress, including a promise from Tom Albright, R&D Manager for Patterned Media at HGST, that a paper coming up in August will show “very nice progress.”

The technical challenges either involve creating heating laser spots of around 10 nm for HAMR, or disks covered with similar sized etched features for BPM.

IDEMA is the hard drive trade association, and manages the industry road map, and precompetitive research. Today, Mark Geenen, chairman of IDEMA, suspects that “HAMR is probably leading the race, although the leader has changed in the last 10 years and might well change again.” However, he adds, “Lots of resources and money are being spent on both technologies. Whichever technology wins, it will become the industry standard for several years because the customers want multiple sourcing.”

Paul Hofemann, chief marketing officer at Molecular Imprints, agrees that HAMR is the PR leader, partly because HAMR has a lower capex . “If HAMR continues to slip then they may re-prioritize back to BPM,” he notes. “Neither technology has published a complete operating disk at a density greater than the 1 Tb/in² barrier.” He adds that, if anything, the published data he has seen show BPM has demonstrated an operating drive over larger fraction of a disk at greater than 1Tb/in2.

It looks as if either HAMR or BPM could get from 1 to 4 Tb/in². One intriguing possibility, detailed in research papers, is to combine HAMR and BPM. That could increase density all the way to 100 Tb/in2 on a 2nm pitch. Certainly the industry believes it has a long-term role in data storage, but it might take many decades to get to 100 Tb/in². From a supplier perspective, the HDD story over the last 10 years is yet another example of the difficulty of positioning a business for technology change. Hofemann believes that “both HAMR and BPM will ultimately be required and are actually very complementary technologies.”

The industry still is committed to extending the life of hard drives well into the future. Whichever technology gets to the first full-sized target density drive will dominate the industry. It is unclear whether HARM or BPM will win, and even less clear when it will happen.