Carbon Nanotubes Take the Heat Off Chips

Purdue scientists find flexible filaments best

Photo: David Umberger/Purdue News Service

WATCHING NANOGRASS GROW

Baratunde A. Cola [left] and Placidus B. Amama grow nanotubes on chips.

Computer chips use only 1 percent of their electrical power to process information. They convert the rest to heat. As chips get smaller and faster, they also get ­hotter, which has engineers looking to carbon nanotubes and other new technologies to keep them cool.

The promise of carbon nanotubes lies in their high thermal conductivity, the ease with which heat flows through them from one end to the other. Researchers at Purdue University, in West Lafayette, Ind., managed to grow forests of nanotubes directly on a chip, and they found that the key to making nanotubes work as heat conductors is to make them flexible.

Most computer cooling systems work by blowing air over a heat sink—a metal plate usually ribbed like a radiator to ­dissipate heat into the air—”but the ­bottleneck is occurring between the heat sink and the chip,” says Baratunde A. Cola, the Purdue doctoral student who coauthored a paper about the work in the 26 September issue of Nanotechnology. You can’t just stick a heat sink directly on a chip, because the sink’s microscopic roughness creates air pockets that resist heat flow [see ”Beat the Heat,” IEEE Spectrum, May 2004]. Current systems rely on thermal interfaces such as grease or solder to fill the gaps, but they are far from ideal.

Figuring nanotubes might do a ­better job, the Purdue team grew between 100 million and 1 billion tubes per square millimeter on test chips. The researchers wanted to see how they could maximize the thermal conductivity of the carbon nanotubes by varying their diameter and defect density. They controlled the tube properties by using a dendrimer template—­essentially a chemical structure with uniformly sized cavities, according to Placidus B. Amama, one of the Purdue researchers. They used the dendrimers to place metal seed nanoparticles atop the chip from which the nanotubes grew. The size of the seed nanoparticles, in turn, determined the diameter of the tubes.

To the team’s surprise, however, controlling the diameter of the individual tubes was less important than controlling how the tubes made contact with the heat sink. ”You need a certain level of ­conductivity,” Cola says, ”but once you get past that threshold, it’s all about contact.”

The interface between the nanotube and the heat sink is like that between the bristles on a toothbrush and your teeth, Cola says. The more the nanotubes can bend, the more they find their way into the nooks and crannies of the heat sink ­surface. To increase the tubes’ ­flexibility, the researchers found that they had to make less conductive, ”lower quality” nanotubes with more defects.

Illustration: Bryan Christie Design

FLEXIBLE FIT

Carbon nanotubes can channel heat from a chip into a heat sink but do it best if they can bend enough to fit into the rough spots on the heat sink.

Such a carbon nanotube interface is several times as conductive as the thermal greases commonly used now, according to IEEE Fellow Avram Bar-Cohen, chairman of mechanical engineering at the University of Maryland, College Park. Bar-Cohen says carbon nanotubes show promise, ­especially for passively cooled devices such as cellphones and personal digital assistants, which lack space for a fan.

”Obviously, people want more and more capability in these personal systems,” Barâ''Cohen says. ”You’d like to run the chips at higher power and yet cool them passively.”

According to Victor Chiriac, a principal scientist at Freescale Semiconductor in Tempe, Ariz., and an expert on thermal management, the Purdue team is among those leading the efforts to make carbon nanotube interfaces practical. Among the other researchers exploring the issue is a team at Stanford University that is experimenting with the concept of growing tubes from both sides of the interface and joining them in the middle.

The nanotube research is still far from seeing use in real products, though, Chiriac says. ”It’s one thing to build in a lab, and another thing altogether to commercially fabricate a device, as current costs could be prohibitive,” he says. He calls the Purdue group’s ability to control the ­diameter, length, and flexibilities of the tubes an important step but just one of many that need to be taken.

Related Stories

Advertisement
Advertisement