A divide over the future of hard drives

Perpendicular hard drive technology, which started appearing last year, currently lets manufacturers increase drive density, or the amount of data stored, by around 50 percent annually. But that pace of progress will likely sputter in about four to five years.

To keep progress going, the first disks based on new technology will need to enter the market around 2011. Competitors differ, however, on how and when ideas for revamping drives should become reality.

Seagate Technologies, the world's largest drive maker, wants to first adopt a concept called "heat-assisted magnetic recording." This involves heating microscopic cells on the disk platters as part of the recording process.

Meanwhile, Hitachi Global Storage Technologies, No. 2 in the industry, favors going forward first with something called "patterned media." In this technique, the cells that store data--which now sit next to each other in a continuous film--would be isolated from each other like dots.

Time is of the essence: Five years--from concept to the first finished products that can be shipped to customers--isn't long. Additionally, Flash memory makers assert that their chips will start to displace drives in notebooks over the coming years. Drive makers scoff at the notion, but agree that technological changes need to occur for drives to protect their turf.

"We need to maintain that 40 percent areal-density growth rate, at a minimum, to stay ahead of flash, and we are dang well going to do it," said Mark Kryder, chief technology officer at Seagate.

Eventually, manufacturers will combine heat-assisted and patterned media to produce drives that will be capable of storing 50 to 100 terabits of data per square inch. That's 280 to 560 times more dense than the 178.8 gigabit-per-square-inch drive coming from Toshiba later this year. (A square inch of 100-terabit material could hold as much data as 12,500 pickup trucks filled with books.)

Seagate and Hitachi, as well as other drive makers, are experimenting with both technologies in their labs. Still, the next step is yet to be determined.

"Most people have thought heat assistance probably would be first, but who knows?" said Jim Porter, president of Disk/Trend, which analyzes the disk drive industry. (Click for photos)

Superparamagneticexpialidocious
The enemy of hard drives is your thermostat. The devices store data in bits, which are microscopic spots on a hard drive platter. The bits themselves are made up of about 50 to 100 cobalt-platinum grains. When the grains get magnetized in a particular direction, the bit represents either a "1" or "0".

To increase the areal density, which is the amount of data a single platter inside a hard drive can hold, engineers have shrunk the size of bits and grains over the years. This has helped PC makers to boost the capacity of hard drives from a few megabytes to more than 100 gigabytes.

Successive years of shrinkage, however, have led to magnetic grains that measure about 8 nanometers long. (A nanometer is a billionth of a meter.)

Reducing the grains further in size could cause them to flip at room temperature and so corrupt the data--an aspect of the "superparamagnetic effect," first identified in the mid-1990s by Stan Charap of Carnegie Mellon University. And cutting back on the number of grains inside each bit, absent further changes, would increase noise and lower reliability.

Drive manufacturers have bought time with perpendicular drives, which stack the bits vertically. But that solution doesn't eliminate the "no more shrinkage" problem.

One or the other
The heat-assisted camp wants to change the grains. Unlike cobalt-platinum grains, iron-platinum grains will not flip at room temperature, Kryder said. To record or erase data, a laser integrated into the drive would heat a particular bit. The data would get recorded or erased, and the bit would quickly cool.

"We'd have to change the (recording) head to add heat, but it's not that big of a deal," Kryder said. Adding a laser wouldn't increase costs much, he noted. More important, the bits could be applied to the platter surfaces through a film, which is how bits are applied today.

Material changes, however, are rarely easy; for example, the switch from aluminum to copper in semiconductors confounded semiconductor makers. For the heat-applied technology, engineers would have to perfect ways to pinpoint the heat from the laser.

"It requires small optical spots. It requires very sharp thermal gradients. It requires new materials," John Best, chief technologist at Hitachi, said, pointing out hurdles in the process.

"You could argue (about) which one's easier to solve, but it looked to us that the practical problems with patterned media meant that we could probably do it first more easily," Best added.

By contrast, the patterned media group wants to keep the current grains. It proposes, instead, reducing the number of grains in each bit from 100 to one, and then isolating the bits from each other to reduce cross-talk and the risk of data corruption, Best said. Initially, the grains in the first patterned media drives would be larger than the grains in today's drives, but the overall size of the bit would be smaller.

"With this, you can get a factor of 100 in increase in density. Of course, you have to scale everything else, so it will take time. But the problem of the temperature of the room reversing magnetization goes away," Best said.

So how do you create a pattern? A master pattern could be drawn with e-beam lithography. That pattern could then be transferred to a mold, which would then be used to stamp out the pattern on hard drive platters though imprint lithography.

Adopting e-beam and imprint lithography into mass manufacturing won't be easy. In fact, patterned media hard drives could easily become the first widescale application for both, Best said.

E-beam, which creates a pattern by firing electrons, was invented years ago to replace traditional lithography in chipmaking, but it never did. Imprint lithography, which makes an impression like that on a signet ring, was only developed in the last few years.

However, lithography of any kind is expensive, particularly when compared to the film-coating processes used today. "We don't have to personalize each bit by patterning it lithographically," Kryder noted, referring to the heat-assisted technique.

Both camps have published papers and lab results, but no one is close to having manufacturing samples. Hitachi, for instance, has created prototype components, but not complete patterned media drives.

Ultimately, the decision could turn on which technology looks easier to bring to mass manufacturing. This year, around 450 million to 460 million drives will leave factories, according to data from Disk/Trend.

"You've got to figure out how to do this, not just in a lab demonstration, but by producing them in the hundreds of millions," said Porter of Disk/Trend. "The good news is that you have people working in both of these camps, and maybe others. There's nano-this and nano-that."

No matter which goes first, the end is not near. Hard drive makers are even examining new materials that could take the grain size below 8 nanometers, although the current candidates are corrosive.

"We can see 50 to 100 terabits being possible," Kryder said. "We are three orders of magnitude from any truly fundamental limits."