Staff Writer, CNET News.com
Molten silicon, designer molecules, and protein globules from a cow. Someday, one of these materials could be used to store data in cell phones and PCs.
A number of start-ups are tinkering with technology that could enhance or replace , and other storage devices. The new technology will benefit consumers, but, just as important, reduce the onerous capital budgets facing manufacturers.
England's , for instance, has developed a method for inserting a magnetized particle inside a sphere of ferritin, a protein produced by animals. Assembled in arrays, the magnetized protein globs can be flipped to represent 1s and 0s, the basic units of digital information.
"I'll go to conferences and people will ask, 'This stuff is a protein. Can you eat it?'" said Eric Mayes, NanoMagnetics' founder and CEO.
On one level, the mission of these companies is a Pyrrhic quest. Hard-drive manufacturers regularly report financial losses and, until recently, profits in flash memory often proved elusive. Moreover, both industries are notoriously conservative when it comes to adopting new technologies.
Advocates, though, believe that circumstances are beginning to congeal that will start to pry the door open to experimentation. For one thing, the cost-benefit equation for producers is getting extraordinarily steep. A gigabyte of hard-drive space currently sells for around 50 cents at retail outlets. Flash memory is also declining in price. Factories to build these devices, however, can cost billions, and research budgets can be arduous.
At the Semicon West conference in San Francisco in July, , director of technology strategy at Intel, noted that the chip industry has taken the first tentative steps toward adopting new manufacturing and design methods because of the cost and scientific difficulties that are ahead.
At the same time, advocates claim they can achieve far greater densities than is possible with existing technologies.
"I can see us doing 20 to 50 times the capacity per (chip) than they do," said CEO Gordon Knight, referring to flash memory makers. His company is a Fremont, Calif.-based start-up that has received a rare venture investment from .
Cranking up the heat
Heat is a problem for most semiconductor manufacturers, but it is the key to Nanochip's technology. A microscopic probe hovering above a piece of silicon quickly heats a point on the silicon to over 600 degrees Celsius. Almost instantly, the silicon crystal beneath the probe becomes amorphous, and thus gets read as a 0. When it cools, it crystallizes, and the area registers as a 1.
Using material this way to store data is part of the field of electronics called "Ovonics." The concept is similar to IBM's technology, but it relies on different processes to change the underlying media and uses fewer probes, Knight said.
While similar to CDs and DVDs, Ovonic media can store more data, according to advocates, because the tools for writing and reading the data—in this case, microscopic probes—are incredibly small, thereby reducing the memory storage location. A red laser, used to read data on a CD, has a beam that's 500 nanometers long. The probes, potentially, could get down to 20 nanometers. (A nanometer is a billionth of a meter.)
While full prototypes do not yet exist, Nanochip has demonstrated to investors that a 1-square-inch chip of its material could contain a terabit of data.
Commercially released chips could contain 50GB of storage space, but, as a Chiclet-size chip, they would be far smaller than the 3.5-inch hard drives inside desktops today. Early versions will likely compete directly with today's minidrives, which contain 1GB to 4GB. Samples should go to customers next year, while volume production could conceivably begin in 2006.
"Speed is a problem because it involves mechanical tips. We're limited by the actuator speed," he said. (The "actuator" is the crane arm that moves the probes into place.) The technology thus will likely compete against NAND flash, the kind found in digital camera cards, or minidrives.
While density would appeal to consumers, the comparatively low cost will appeal to manufacturers. The device's actuators together measure around 10 microns to 20 microns in length, relatively large compared with the 90-nanometer-size features found on today's chips. Ideally, chipmakers could manufacture these chips with tools last used in the early '90s.
"They are tiny from a mechanical standpoint, but they are big from a lithography standpoint," Knight said.
History shows that this could be an uphill climb. Intel co-founder Gordon Moore once predicted a bright future for Ovonics. That prediction was made in the early '70s and has yet to be fulfilled.
Creating one's own molecule
Make every molecule work: That's the concept behind Denver's .
Computer memory essentially holds data by retaining an electric charge, similar to the way a sponge holds water. And, like a household sponge, not every particle gets used to its full potential.
ZettaCore has crafted a complex molecule that can retain or release up to eight electrons. Depending on the number of retained electrons, a given molecule will exhibit a distinct voltage level, which can be read as data. A single molecule can represent 4 bits of data; by contrast, much larger flash memory chips on the market today can retain 2 bits of data at once.
"Molecules act similar, but the fundamental difference is the amount of charge in a given area is greater," said company CEO Randy Levine, who used to teach astrophysics at Harvard University. "Everything is contributing."
As with Nanochip's technology, the payoff comes in two ways: More data can be stored in a smaller space and the chip-manufacturing process, potentially, is cheaper. As with current chips, memory cells have to be drawn through lithography. In ZettaCore's technology, the molecules repel and attract one another until orientated on a substrate. Chemical forces, rather than physical manipulation, orchestrate a chip's design.
Each of the molecules contains several hundred atoms. "They aren't small by (the) chemist's standpoint, but they are extremely small from a manufacturing standpoint," Levine said. The molecule can hold the charge after a host computer is turned off and therefore can be used to replace flash or standard computer memory, called DRAM.
"We have built workable chips at commercially interesting densities," Levine said. "All of the challenges at this point deal with manufacturability."
Of all the companies in this field, ZettaCore arguably has the most star power. Founded by professors from North Carolina State University and the University of California, Riverside, the company has raised $23 million from, among others, Draper Fisher Jurvetson and Intel Capital. Industry luminaries such as Vinod Khosla and Les Vadasz sit on the board.
"It is a conservative industry," Levine said. "But people are talking about it (the transformation). I don't know a major memory vendor that isn't."Should cows get stock options?
Cambridge, England-based NanoMagnetics is one of a number of new start-ups trying to improve electronics through .
"We've got little rust particles that are carried by a protein ball. We can take out the rust and put in platinum alloys or other metallic materials," which can then be used to represent data, explained Mayes, founder and CEO. "Each particle is uniform in size."
Mayes came up with the idea for the company while working on a doctorate, at the University of Bristol, on the interaction between biological agents and inorganic molecules. Bricks and seashells are structurally quite dissimilar, he pointed out, but made of the same material. By exploiting naturally occurring biological phenomenon, NanoMagnetics hopes to reduce the cost of manufacturing storage devices.
True to its roots, the company still obtains its carrier protein, ferritin, the old-fashioned way.
"We get it from animal sources," he said. Humans produce ferritin. But instead of getting it from employees, NanoMagnetics buys its ferritin from collagen manufacturers who get it from cows. The company is currently looking at ways to produce it in the lab "because my wife is a vegetarian," Mayes joked.
In terms of size, ferritin is relatively small. A single sphere measures 12 nanometers in diameter, while the inner cavity measures 8 nanometers. By contrast, an AIDS virus measures around 50 nanometers across. NanoMagnetics is also looking at DNA protection system (DPS) proteins, which are even smaller, as carriers.
Seagate, IBM and others have performed similar experiments by coating magnetic particles in a substance similar to olive oil. Proteins, however, can withstand higher temperatures and therefore maintain their shape and relative position in an array better during high manufacturing temperatures, he said.
While the company has tinkered with different business models, it is now aiming at working with established manufacturers to incorporate the technology into finished products. Agreements may be announced toward the end of the year that could lead to products by the end of 2006. A large Asian manufacturer is currently testing it.
Conceivably, the material could be used to create multiple-gigabyte chips that would allow cell phones to store movies and sitcoms, Mayes said.
Like other companies in this space, however, the uphill challenge is in persuading manufacturers to adopt it. NanoMagnetics' active particles are applied with an inkjet sprayer. Hard-drive makers now use chemical sputterers to apply coatings. Although ink jet spraying will be cheaper, convincing producers to spend money on new methods isn't easy.
"Everyone is a step away from the dustbin," he joked. Most likely, NanoMagnetics will have its first commercial success in water purification, he said. The same particles can be used in the reverse-osmosis process.Ribbons of nanotubes
Woburn, Mass.-based Nantero has come up with a way to make transistors, the on-off switches inside chips, with , hollow tubes of carbon atoms that exhibit a number of remarkable properties.
The company's technology exploits two of these properties: the bendable nature of nanotubes and the strong attraction carbon atoms exhibit for one another.
In Nantero's memory design, ribbons of carbon nanotubes are suspended over pieces of carbon. In the "off" state, the ribbon of nanotubes does not touch the carbon. In the "on" state, the nanotube bends downward and then adheres to the carbon. Electricity flows and the memory cell registers as a 1, in data terms. An electric charge can separate or connect the ribbons.
Nantero's memory is faster than static random access memory (SRAM), an embedded form of memory used for caches on processors, CEO Greg Schmergel said.
Nantero's success in licensing its technology—LSI Logic and aerospace specialist BEA Systems have taken out licenses—lie in a radical shift in architecture that was implemented during the past year. Initially, Nantero proposed making memory cells in which a single nanotube would connect or detach from a perpendicular nanotube below it.
While this would allow for incredibly dense memory chips, most analysts and scientists doubted that the company could come up with a way to erect millions of uniform, microscopic crossbars on a sliver of silicon less than a few square centimeters across.
In the new architecture of the company's chips, a layer of nanotubes is spread onto a piece of carbon. Engineers then use conventional lithography to "draw" electrical contacts that are connected to each other by the thick ribbons of nanotube material and the substrate. This method, however, eliminates many of the cost and size advantages, at least for now.
"We won't be achieving thousandfold density improvement just yet," Schmergel said.
And, like all other new memory advocates, the company will have to face the skepticism of an industry that has seen it all.
"A lot of people are looking for the next material, but we think they already found it in silicon," said Dan Steere, vice president of marketing and sales at Matrix Semiconductor.