After years of hype and false starts, the nanoimprint lithography market is set to explode--or, perhaps, implode.

In one scenario, nanoimprint tool vendors--whose business thus far has been largely limited to small R&D markets--are poised to fulfill past promises, enabling a low-cost paradigm shift for the production of optoelectronics, semiconductors, storage and other products. But nanoimprint vendors could also be on the brink of a shakeout, wrought by too many vendors chasing too small a market.

Business considerations aside, nanoimprint is back in vogue from a technological standpoint after having been dismissed earlier in the decade by some lithography experts. The delays and soaring costs for extreme-ultraviolet (EUV) and other next-generation lithography technologies have rekindled an interest in nanoimprint in the IC world, particularly among the NAND flash community. The storage and LED camps are likewise looking at nanoimprint for the development of next-generation recording media and photonics-based LEDs.

Nanoimprint does not use expensive optics; rather, it employs a simple stamping or hot embossing process to generate features on a device. The process starts with a template or mold. An electron-beam tool then patterns select features on the template, based on a 1:1 reduction scheme. Using a nanoimprint tool--which resembles a giant stamping machine--the template is pressed against a heated, single-layer substrate, creating the tiny feature sizes.

The second looks at nanoimprint have spurred a new wave of business activity. Hitachi Ltd. and Toshiba Corp. are among the well-known names that have entered the nanoimprint market. In May, Nanolithosolution Inc. joined the fray with a nanoimprint technology that was conceived at Hewlett-Packard Co.

Nanoimprint pioneer Nanonex Inc., meanwhile, has entered the semiconductor equipment arena, positioning itself to compete against Molecular Imprints Inc. (MII) and Obducat AB. EV Group and Suss AG also take part in the market. Lux Research (New York) believes that the nanoimprint tool business could grow to nearly $235 million by 2010.

For now, however, nanoimprint remains a small business that has been a money-losing proposition thus far for some players. And it's still unclear whether the technology will move out of the R&D stage, leaving some observers to predict that consolidation will occur in the supplier base.

Nanoimprint remains promising in "applications other than semiconductors,'' said G. Dan Hutcheson, chief executive of VLSI Research Inc. (Santa Clara, Calif.), a research firm.

"In general, nanoimprint is still in the transition stage," said Patrik Lundstrom, chief executive of market leader Obducat (Malmo, Sweden). "We are moving from the R&D environment to the mainstream. I think we'll see some movement in the next 12 to 18 months."

Most observers predict no real action this year. Estimates are that only 30 to 50 nanoimprint machines shipped in 2006. In 2007, shipments are widely expected to be below 50 units; some sources estimate that vendors in total will ship only 10 to 20 real tools this year.

Nanoimprint origins
The nanoimprint lithography business emerged in the late 1990s, with the arrival of Nanonex and Obducat. Subsequently, EV Group, MII and Suss announced tools.

More recently, Hitachi and Toshiba have moved into the arena. Hitachi is using nanoimprint within its own operations, reportedly for disk drives, sources said. Toshiba says it has shipped a few systems, mostly in Japan.

While the numbers for nanoimprint remain modest, the buzz surrounding the technology is anything but. Nanoimprint can pattern devices at feature sizes below 20 nanometers at a fraction of the cost of conventional optical lithography. Today's leading-edge 193-nm immersion scanners sell for as much as $40 million each. In contrast, a state-of-the-art nanoimprint tool may sell for up to $10 million, and most tools in the category sell for far less than that.

But nanoimprint has its share of drawbacks that have kept the technology from going mainstream. Throughput is low; some tools reportedly can process five or fewer wafers an hour, compared with 130 for some optical lithography tools. Overlay--the ability to align and process the various device layers--is another stumbling block: Because nanoimprint involves direct contact between the template and substrate, misalignment can result in unacceptable errors. And the development of 1:1 templates is expensive.

Next-gen questions
Early this decade, suppliers of nanoimprint technology were saying they would soon turn the industry upside down. One vendor predicted nanoimprint would be used for mainstream chip production at the 65-nm node.

That era, of course has arrived, and optical lithography continues to rule semiconductor fabrication. The Information Network (New Tripoli, Pa.), a market research firm, predicts the overall lithography market will grow 11 percent this year, to $7.4 billion. Nanoimprint's share will be negligible.

But the next-generation-lithography front-runners have experienced their share of problems. Immersion lithography could hit a wall at the 32-nm node; EUV has been delayed several times. That has some in the industry considering nanoimprint for the 22-nm node, arriving around 2011.

Chips for 32, 22 nm
"EUV is not ready, and e-beam is too slow," said John Doering, vice president of marketing and business development at MII (Austin, Texas). "The problems for EUV have been larger than anticipated."

NAND flash vendors, in particular, are taking a closer look at nanoimprint. Earlier this year, MII claimed to have delivered a nanoimprint lithography machine to a major memory maker for use in developing chips at 32 and 22 nm. Sources believe MII shipped its Imprio 250 machine to Toshiba for use in NAND flash process development.

The Imprio 250 is based on MII's "step and flash" nanoimprint technology. The tool enables sub-50-nm half-pitch resolution and sub-10-nm alignment.

While it's unclear whether Toshiba will use the tool in production fabs, other NAND flash vendors--reportedly including Hynix Semiconductor, IM Flash Technologies and Samsung Electronics--are expressing interest in the technology. "We're being pulled by a half-dozen semiconductor makers," Doering said.

Larry Koecher, chief operating officer for rival Nanonex (Monmouth Junction, N.J.), also sees a "window of opportunity" for nanoimprint in the semiconductor sector. "We take EUV with a grain of salt," he said. "There is uncertainty whether the optical guys can rise to the occasion" and develop EUV in time.

Nanonex has focused on nonsemiconductor markets, but its NX-3000--a nanoimprint tool based on a new "step and repeat"--technology, is suitable for use with semiconductors, according to the company. The NX-3000 is an R&D tool that's said to process 8-inch wafers at sub-10-nm feature sizes.

In the future, Nanonex plans to expand its line of step-and-repeat products, which bear a slight resemblance to MII's "step and flash" technology. Like MII, Nanonex plans to offer three tools at different price points.

For its part, Obducat recently introduced Sindre, a nanoimprint system for high-volume manufacturing that reportedly can imprint 200-mm wafers at a rate of 30 per hour.

Despite his own company's breakthrough, Obducat's Lundstrom remains skeptical about nanoimprint's near-term suitability for mainstream chip production. Throughput and overlay will remain issues for the technology and could push out nanoimprint's arrival for mainstream chip production from 2011 to 2013, he said.

But he's not worried. "I strongly believe we can do without the semiconductor market," Lundstrom said. "In all honesty, there are so many other applications."

Nanoimprint's big nonsemiconductor applications are storage-patterned media, photonic crystals and wire grid polarizers. Those apps require less-stringent overlays than chip manufacture and may require nanoimprint sooner rather than later.

Consider the disk drive sector, where for decades vendors have increased storage capacity by shrinking the tiny magnetic grains that constitute the data bits in a system. For traditional, longitudinal recording media, data integrity would be "harmed if we continue to shrink the magnetic grains," drive giant Hitachi Global Storage Technologies warned in a recent presentation.

The inability to scale the magnetic gains is called the superparamagnetic effect. Drive vendors have skirted the issue of late by moving to perpendicular recording technology, but now they are looking beyond that technique.

One alternative is the use of bit-patterned media, wherein "each bit is stored in a single, deliberately formed magnetic switching volume," according to the Hitachi presentation. "Since we no longer need on the order of 100 grains per bit, but just one single, grain-sized switching volume, density can be increased by roughly two orders magnitude compared with conventional recording media."

Today's drives have media capacities of 100 Gbits/square inch, with a center-to-center spacing of 86 nm, according to the presentation. Increasing capacity to 1 Tbit/inch² would require center spacing of 27 nm. At 10 Tbits/inch², spacing would be just 9 nm. Hitachi believes optical lithography will run out of gas before the bit-patterned era and is exploring nanoimprint as an alternative.

Another potential market for nanoimprint is high-brightness LEDs, which are enhanced with photonic crystals. Based on gallium nitride (GaN), the souped-up LEDs are emerging in flat-panel displays, lighting systems and automotive headlights.

"These structures look like arrays of contact holes on the surface of the LED and serve to increase both the brightness and the efficiency of the LED," according to a presentation from MII. "The feature sizes need to be less than the wavelength of the LED emission, and the minimum hole spacing can be significantly less than 100 nm. Given the poor surface flatness and 3-inch dimensions of the GaN substrates used for these devices, optical lithography is difficult to use for these dimensions, and imprint is the preferred solution."

The opportunities for nanoimprint thus seem ripe. The question for tool vendors is whether they can survive long enough to cash in.