Scientists make carbon nanotubes more suitable for CPU cooling

Spread the love

Scientists at the Berkeley Lab have developed a technique for efficiently cooling processors using carbon nanotubes. The technique improves the dissipation of heat from a metal surface to the nanotubes by a factor of six.

Carbon nanotubes have long been known to conduct heat very well. The problem is that they are difficult to interact with other materials, which is an obstacle to practicality such as in chip cooling. The weak interaction with other materials ensures a high heat resistance, which reduces the capacity to dissipate heat.

Intel went to the Berkeley Lab to work together to tackle the problem. The researchers at Berkeley Lab’s Materials Sciences Division used organic molecules to create covalent bonds between the carbon nanotubes and metals such as aluminum, gold or copper. The nanotubes are first produced as vertical rows of nanotubes on a silicon wafer substrate. The connection between the nanotubes and the metals was so strong after the formation of the bonds that the nanotubes could be pulled away from the growth substrate. This degree of adhesion provided six times the efficient dissipation of heat than has been achieved so far with carbon nanotubes on metal.

“You can think of the resistance as an extra distance that the heat has to travel through the material,” says Sumanjeet Kaur of the research team. “With carbon nanotubes, the heat resistance interface creates a distance of about 40 microns on both sides of the layer of tubes. our technique, we have succeeded in reducing the resistance interface so that the extra distance is about seven microns.”

The scientists are now going to try to get more nanotubes to bond to the metal, because with the current technology it is still possible that a majority of the tubes do not connect. The covalent compounds are made in gas vapor or liquid chemistry at a low temperature, making it theoretically suitable for implementation in current chip production methods.

You might also like