nanotube-based memory wafer using standard semiconductor fabrication processes.

Carbon nanotube memory coming.

Start-up Nantero has built a carbon nanotube-based memory wafer using standard semiconductor fabrication processes. This removes a significant hurdle in commercialising the seemingly exotic NRAM (non-volatile RAM) that could replace DRAM, SRAM and flash memory with a universal memory design.

A carbon nanotube has a wall just one atom wide and a diameter on one billionth of a metre. It is the building block for NRAM chips.
Production of NRAM has faced obstacles in the past, including an inability to position the nanotubes reliably across entire silicon wafers, and contamination of supplied nanotubes that made the material incompatible with hyper-clean CMOS semiconductor fabs.
Nantero has solved these problems and has demonstrated the precise alignment of 10 billion nanotubes on a single wafer manufactured in a standard semiconductor fabrication plant. The process treated the nanotube material as a fabric and used standard semiconductor fab techniques such as spincoating, lithography and etching.
It was started with OEM-sourced carbon nanotubes. These were cleansed of impurities, stray particles for example, and then layered onto a wafer with a foundation of flat tubes parallel to one another in a silicon glue, and a second upright tube layer above it, also with parallel tubes and also in silicon. The tubes can be individually bent by putting an electrical charge through them. A bent tube touching another completes a circuit by lowering electrical resistance and signals a binary one. Straighten the bent tube and you have a zero.
The aim is for Natero to licence its intellectual property to semiconductor fabrication suppliers which would then sell NRAM chips to server, PC, notebook and mobile intelligent device manufacturers.
Nantero has been doing NRAM manufacturing research work with LSI Logic, and this was switched to On Semiconductor in May this year when that firm bought the Gresham, Oregon-based fab from LSI.
In July Nantero and HP anounced they were working together on a printable memory application.
Greg Schmergel, CEO and co-founder of Nantero, said at the time: "A printable NRAM memory could be deposited on flexible substrates to enable very low cost RFID tags."
We might see the introduction of devices using NRAM chips in 2010, possibly earlier.

Oxygen makes nanotube memory
Carbon nanotubes have been used to make experimental transistors, chemical sensors and memory devices that are far smaller than anything available today. But moving from experimental prototypes to practical devices requires overcoming a large hurdle: controlling the way nanotubes grow.

Nanotubes tend to form as mixes of two types -- semiconducting and metallic -- with semiconducting the more technologically desirable. In April, 2001, IBM researchers announced they could weed out metallic nanotubes by sending enough current through a batch of nanotubes to burn up the metallic tubes but not enough to damage the semiconducting ones.

Researchers at the Max Planck Institute in Germany have come up with an alternative method of producing all-semiconducting bundles that, in addition, prepares the microscopic tubes for use in memory devices. The technique allows researchers to oxidize bundles of a few nanotubes or individual nanotubes that measure as small as 2 nanometers in diameter. A nanometer is one millionth of a millimeter, and a line of 20 hydrogen atoms spans two nanometers.

Oxidizing a bundle of nanotubes converts metallic ones to semiconducting, said Marko Burghard, a scientist at the Max Planck Institute for Solid State Research in Germany. Assemble the oxidized bundles into larger arrays and they could be "key building blocks for low-cost memories with ultra-high storage densities," he said.

The process could theoretically produce memory devices that hold one trillion bits per square centimeter, said Burghard. A trillion bits is about 31 DVDs worth of data.

The researchers hit on the process after finding that about half of nanotube bundles left in open air for several months had changed from metallic to semiconducting. This happened because oxygen atoms in the air combined with the carbon atoms in the metallic nanotubes to form a nonmetallic oxide.

The researchers were able to induce the effect by heating the nanotubes in air or treating them with oxygen plasma. A plasma is a gas whose atoms are ionized, meaning they have more or fewer electrons than normal and so can conduct electricity.

The researchers took advantage of a consequence of the oxidation process to make prototype memory devices from the oxidized tubes. The devices use a single oxidized nanotube or bundle of oxidized nanotubes as the semiconducting channel of a transistor. Data is represented by tiny electric charges of one or a few electrons stored in a defect on the surface of a nanotube produced by the oxidation. The defects are tiny clumps of amorphous, or jumbled, carbon attached to the otherwise orderly, crystalline nanotubes.

By sending three volts of electricity through the nanotubes, the researchers stored a charge in a surface defect. In electronic memory, the presence of a charge generally represents a 1 and the absence of a charge a 0. To read the 1s and 0s, the researchers sent a small current through the nanotubes to measure their conductivity. Nanotubes that harbor a stored charge are about 1,000 times more conductive than those without a charge.

Charge storage memory devices based on nanotubes were first developed several years ago; research teams at the University of Maryland and the University of Pennsylvania have recently developed experimental devices.

The Max Planck Institute memory device, however, is able to store charges longer than the other devices, said Burghard. The Maryland researchers reported a charge storage time of 1.4 hours, and the Pennsylvania researchers 16 hours. The Max Planck device is able to store charges for more than 12 days, Burghard said. Stored charge memory devices can be used as nonvolatile computer memory, which retains its data when the power is off.

The research is important work, said Vincent Crespi, an assistant professor of physics at Pennsylvania State University. "It enables a memory device to be implemented within a single nanotube plus three contacts," he said. "The charge trap seems to come along for free."

The researchers will need precise, reproducible control over the character of the charge trap before the device can be used in practical applications, Crespi added.

The researchers plan to study further the oxidation process and the nature of the charge storage defects, said Burghard. Another goal is to search for better, more controllable chemical modifications of the nanotubes, "for example, by electrochemically attaching appropriate chemical residues or small metal clusters, which could then be used for charge storage," he said.

The researchers' nanotube memory element could be used in practical applications in five to ten years, said Burghard.

Burghard's research colleagues were Jingbiao Cui, Roman Sordan and Klaus Kern. They published the research in the October 21, 2002 issue of the journal Applied Physics Letters. The research was funded by the Max Planck Society.